diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,15799 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 7.0, + "eval_steps": 500, + "global_step": 7161, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.004887585532746823, + "grad_norm": 12.44376256794233, + "learning_rate": 2.2315202231520225e-07, + "loss": 0.775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7660703063011169, + "step": 5, + "valid_targets_mean": 2759.7, + "valid_targets_min": 789 + }, + { + "epoch": 0.009775171065493646, + "grad_norm": 13.083127751453706, + "learning_rate": 5.020920502092051e-07, + "loss": 0.7585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7838925123214722, + "step": 10, + "valid_targets_mean": 2175.1, + "valid_targets_min": 647 + }, + { + "epoch": 0.01466275659824047, + "grad_norm": 10.557856263973706, + "learning_rate": 7.810320781032078e-07, + "loss": 0.7713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7736507654190063, + "step": 15, + "valid_targets_mean": 2148.8, + "valid_targets_min": 803 + }, + { + "epoch": 0.019550342130987292, + "grad_norm": 9.4146088353884, + "learning_rate": 1.0599721059972108e-06, + "loss": 0.7562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7820156812667847, + "step": 20, + "valid_targets_mean": 2958.0, + "valid_targets_min": 945 + }, + { + "epoch": 0.024437927663734114, + "grad_norm": 8.510426871763977, + "learning_rate": 1.3389121338912134e-06, + "loss": 0.7276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7074830532073975, + "step": 25, + "valid_targets_mean": 2418.9, + "valid_targets_min": 734 + }, + { + "epoch": 0.02932551319648094, + "grad_norm": 5.967061937258563, + "learning_rate": 1.6178521617852163e-06, + "loss": 0.6946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7003393769264221, + "step": 30, + "valid_targets_mean": 2274.1, + "valid_targets_min": 724 + }, + { + "epoch": 0.03421309872922776, + "grad_norm": 5.128418930959132, + "learning_rate": 1.8967921896792191e-06, + "loss": 0.6951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6412136554718018, + "step": 35, + "valid_targets_mean": 2286.8, + "valid_targets_min": 1077 + }, + { + "epoch": 0.039100684261974585, + "grad_norm": 3.685105984706391, + "learning_rate": 2.175732217573222e-06, + "loss": 0.622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6155872344970703, + "step": 40, + "valid_targets_mean": 2433.9, + "valid_targets_min": 635 + }, + { + "epoch": 0.04398826979472141, + "grad_norm": 3.552082634936528, + "learning_rate": 2.4546722454672246e-06, + "loss": 0.6131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.622800350189209, + "step": 45, + "valid_targets_mean": 2232.9, + "valid_targets_min": 914 + }, + { + "epoch": 0.04887585532746823, + "grad_norm": 2.6301835879043747, + "learning_rate": 2.7336122733612273e-06, + "loss": 0.5903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.57831209897995, + "step": 50, + "valid_targets_mean": 2736.5, + "valid_targets_min": 1194 + }, + { + "epoch": 0.053763440860215055, + "grad_norm": 1.939146654771752, + "learning_rate": 3.0125523012552303e-06, + "loss": 0.566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5587443113327026, + "step": 55, + "valid_targets_mean": 2471.7, + "valid_targets_min": 753 + }, + { + "epoch": 0.05865102639296188, + "grad_norm": 1.3923623066674402, + "learning_rate": 3.2914923291492334e-06, + "loss": 0.5091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5293725728988647, + "step": 60, + "valid_targets_mean": 3486.3, + "valid_targets_min": 1205 + }, + { + "epoch": 0.0635386119257087, + "grad_norm": 1.2847016539983112, + "learning_rate": 3.570432357043236e-06, + "loss": 0.5174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.48333871364593506, + "step": 65, + "valid_targets_mean": 2133.9, + "valid_targets_min": 639 + }, + { + "epoch": 0.06842619745845552, + "grad_norm": 1.1703945643061187, + "learning_rate": 3.849372384937239e-06, + "loss": 0.4767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.47202277183532715, + "step": 70, + "valid_targets_mean": 2432.9, + "valid_targets_min": 820 + }, + { + "epoch": 0.07331378299120235, + "grad_norm": 1.1061255401631531, + "learning_rate": 4.128312412831242e-06, + "loss": 0.5054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5339841842651367, + "step": 75, + "valid_targets_mean": 2277.9, + "valid_targets_min": 741 + }, + { + "epoch": 0.07820136852394917, + "grad_norm": 1.1039585723531486, + "learning_rate": 4.407252440725244e-06, + "loss": 0.4741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.49983298778533936, + "step": 80, + "valid_targets_mean": 2021.4, + "valid_targets_min": 855 + }, + { + "epoch": 0.08308895405669599, + "grad_norm": 0.9865369304789178, + "learning_rate": 4.6861924686192475e-06, + "loss": 0.4612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42603811621665955, + "step": 85, + "valid_targets_mean": 2219.6, + "valid_targets_min": 734 + }, + { + "epoch": 0.08797653958944282, + "grad_norm": 1.071853710354841, + "learning_rate": 4.9651324965132506e-06, + "loss": 0.4694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.45195892453193665, + "step": 90, + "valid_targets_mean": 2383.9, + "valid_targets_min": 1015 + }, + { + "epoch": 0.09286412512218964, + "grad_norm": 0.9176728913117824, + "learning_rate": 5.244072524407254e-06, + "loss": 0.4449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42296427488327026, + "step": 95, + "valid_targets_mean": 2121.6, + "valid_targets_min": 967 + }, + { + "epoch": 0.09775171065493646, + "grad_norm": 0.9296613279197797, + "learning_rate": 5.523012552301256e-06, + "loss": 0.4326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4197250008583069, + "step": 100, + "valid_targets_mean": 2173.2, + "valid_targets_min": 895 + }, + { + "epoch": 0.10263929618768329, + "grad_norm": 0.9405388278402095, + "learning_rate": 5.801952580195258e-06, + "loss": 0.4436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4366050362586975, + "step": 105, + "valid_targets_mean": 2097.8, + "valid_targets_min": 816 + }, + { + "epoch": 0.10752688172043011, + "grad_norm": 0.7924545346293838, + "learning_rate": 6.080892608089262e-06, + "loss": 0.4115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4147997796535492, + "step": 110, + "valid_targets_mean": 2730.9, + "valid_targets_min": 955 + }, + { + "epoch": 0.11241446725317693, + "grad_norm": 0.789932617535872, + "learning_rate": 6.359832635983264e-06, + "loss": 0.4256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4059978723526001, + "step": 115, + "valid_targets_mean": 2887.8, + "valid_targets_min": 1001 + }, + { + "epoch": 0.11730205278592376, + "grad_norm": 0.8095583193686153, + "learning_rate": 6.6387726638772664e-06, + "loss": 0.4145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4052962064743042, + "step": 120, + "valid_targets_mean": 2514.8, + "valid_targets_min": 847 + }, + { + "epoch": 0.12218963831867058, + "grad_norm": 0.7949836362164371, + "learning_rate": 6.91771269177127e-06, + "loss": 0.4196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4143103063106537, + "step": 125, + "valid_targets_mean": 2648.2, + "valid_targets_min": 826 + }, + { + "epoch": 0.1270772238514174, + "grad_norm": 0.799726720099262, + "learning_rate": 7.1966527196652726e-06, + "loss": 0.4221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4228057265281677, + "step": 130, + "valid_targets_mean": 2594.8, + "valid_targets_min": 698 + }, + { + "epoch": 0.13196480938416422, + "grad_norm": 0.9617534138420096, + "learning_rate": 7.475592747559275e-06, + "loss": 0.4009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42820626497268677, + "step": 135, + "valid_targets_mean": 2055.6, + "valid_targets_min": 632 + }, + { + "epoch": 0.13685239491691104, + "grad_norm": 0.8348186060833679, + "learning_rate": 7.754532775453279e-06, + "loss": 0.4162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4349338114261627, + "step": 140, + "valid_targets_mean": 2389.0, + "valid_targets_min": 1020 + }, + { + "epoch": 0.14173998044965788, + "grad_norm": 0.8157340380976558, + "learning_rate": 8.033472803347281e-06, + "loss": 0.4207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42874521017074585, + "step": 145, + "valid_targets_mean": 2644.0, + "valid_targets_min": 684 + }, + { + "epoch": 0.1466275659824047, + "grad_norm": 0.8180117684614197, + "learning_rate": 8.312412831241283e-06, + "loss": 0.3971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4317898452281952, + "step": 150, + "valid_targets_mean": 2665.6, + "valid_targets_min": 869 + }, + { + "epoch": 0.15151515151515152, + "grad_norm": 0.8885682855346618, + "learning_rate": 8.591352859135287e-06, + "loss": 0.3854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4081709384918213, + "step": 155, + "valid_targets_mean": 2359.7, + "valid_targets_min": 872 + }, + { + "epoch": 0.15640273704789834, + "grad_norm": 0.8080583509361516, + "learning_rate": 8.87029288702929e-06, + "loss": 0.3982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.410560667514801, + "step": 160, + "valid_targets_mean": 2443.3, + "valid_targets_min": 650 + }, + { + "epoch": 0.16129032258064516, + "grad_norm": 1.0075730249123667, + "learning_rate": 9.149232914923292e-06, + "loss": 0.3827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3756175637245178, + "step": 165, + "valid_targets_mean": 2445.9, + "valid_targets_min": 729 + }, + { + "epoch": 0.16617790811339198, + "grad_norm": 0.7988923971464764, + "learning_rate": 9.428172942817295e-06, + "loss": 0.3902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3763277530670166, + "step": 170, + "valid_targets_mean": 2335.2, + "valid_targets_min": 754 + }, + { + "epoch": 0.1710654936461388, + "grad_norm": 0.8801234974461333, + "learning_rate": 9.707112970711298e-06, + "loss": 0.4039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4362441897392273, + "step": 175, + "valid_targets_mean": 2604.2, + "valid_targets_min": 722 + }, + { + "epoch": 0.17595307917888564, + "grad_norm": 1.0573428481191474, + "learning_rate": 9.986052998605302e-06, + "loss": 0.3815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3826301395893097, + "step": 180, + "valid_targets_mean": 1948.0, + "valid_targets_min": 718 + }, + { + "epoch": 0.18084066471163246, + "grad_norm": 0.8244047729823328, + "learning_rate": 1.0264993026499304e-05, + "loss": 0.3677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3717021346092224, + "step": 185, + "valid_targets_mean": 2318.2, + "valid_targets_min": 868 + }, + { + "epoch": 0.18572825024437928, + "grad_norm": 0.8733754607654697, + "learning_rate": 1.0543933054393308e-05, + "loss": 0.3875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41147416830062866, + "step": 190, + "valid_targets_mean": 2373.2, + "valid_targets_min": 855 + }, + { + "epoch": 0.1906158357771261, + "grad_norm": 0.8971397713788575, + "learning_rate": 1.0822873082287308e-05, + "loss": 0.399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4053829312324524, + "step": 195, + "valid_targets_mean": 2233.9, + "valid_targets_min": 731 + }, + { + "epoch": 0.19550342130987292, + "grad_norm": 1.0257634454527333, + "learning_rate": 1.1101813110181312e-05, + "loss": 0.3825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34646230936050415, + "step": 200, + "valid_targets_mean": 1661.8, + "valid_targets_min": 652 + }, + { + "epoch": 0.20039100684261973, + "grad_norm": 0.8750695251551033, + "learning_rate": 1.1380753138075316e-05, + "loss": 0.3884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39979293942451477, + "step": 205, + "valid_targets_mean": 2502.0, + "valid_targets_min": 749 + }, + { + "epoch": 0.20527859237536658, + "grad_norm": 0.8985663536980157, + "learning_rate": 1.1659693165969317e-05, + "loss": 0.3715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3521466553211212, + "step": 210, + "valid_targets_mean": 2193.1, + "valid_targets_min": 1185 + }, + { + "epoch": 0.2101661779081134, + "grad_norm": 0.8690020156723633, + "learning_rate": 1.193863319386332e-05, + "loss": 0.3906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39606279134750366, + "step": 215, + "valid_targets_mean": 2162.5, + "valid_targets_min": 791 + }, + { + "epoch": 0.21505376344086022, + "grad_norm": 0.7851411691827029, + "learning_rate": 1.2217573221757324e-05, + "loss": 0.3534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3474773168563843, + "step": 220, + "valid_targets_mean": 2564.4, + "valid_targets_min": 754 + }, + { + "epoch": 0.21994134897360704, + "grad_norm": 0.8708152091406828, + "learning_rate": 1.2496513249651325e-05, + "loss": 0.3652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3648437261581421, + "step": 225, + "valid_targets_mean": 2014.6, + "valid_targets_min": 544 + }, + { + "epoch": 0.22482893450635386, + "grad_norm": 0.8997257441861602, + "learning_rate": 1.2775453277545329e-05, + "loss": 0.3817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3548926115036011, + "step": 230, + "valid_targets_mean": 1954.0, + "valid_targets_min": 627 + }, + { + "epoch": 0.22971652003910067, + "grad_norm": 0.7622734239422039, + "learning_rate": 1.3054393305439333e-05, + "loss": 0.3576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3857002854347229, + "step": 235, + "valid_targets_mean": 2903.6, + "valid_targets_min": 964 + }, + { + "epoch": 0.23460410557184752, + "grad_norm": 0.8137819570309708, + "learning_rate": 1.3333333333333333e-05, + "loss": 0.367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3650803864002228, + "step": 240, + "valid_targets_mean": 2334.2, + "valid_targets_min": 746 + }, + { + "epoch": 0.23949169110459434, + "grad_norm": 0.9062648393789453, + "learning_rate": 1.3612273361227337e-05, + "loss": 0.3859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.38647863268852234, + "step": 245, + "valid_targets_mean": 2119.2, + "valid_targets_min": 933 + }, + { + "epoch": 0.24437927663734116, + "grad_norm": 0.8593102670864973, + "learning_rate": 1.3891213389121341e-05, + "loss": 0.3547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32468748092651367, + "step": 250, + "valid_targets_mean": 2219.7, + "valid_targets_min": 807 + }, + { + "epoch": 0.24926686217008798, + "grad_norm": 1.6459054094057368, + "learning_rate": 1.4170153417015342e-05, + "loss": 0.3763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3920065760612488, + "step": 255, + "valid_targets_mean": 3263.6, + "valid_targets_min": 886 + }, + { + "epoch": 0.2541544477028348, + "grad_norm": 0.8244627390914492, + "learning_rate": 1.4449093444909346e-05, + "loss": 0.3478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3641440272331238, + "step": 260, + "valid_targets_mean": 2413.2, + "valid_targets_min": 819 + }, + { + "epoch": 0.2590420332355816, + "grad_norm": 0.947648955580425, + "learning_rate": 1.472803347280335e-05, + "loss": 0.3644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3710935413837433, + "step": 265, + "valid_targets_mean": 1834.6, + "valid_targets_min": 668 + }, + { + "epoch": 0.26392961876832843, + "grad_norm": 0.7479773006623937, + "learning_rate": 1.500697350069735e-05, + "loss": 0.3868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3943009376525879, + "step": 270, + "valid_targets_mean": 2967.0, + "valid_targets_min": 712 + }, + { + "epoch": 0.26881720430107525, + "grad_norm": 0.6850159478077383, + "learning_rate": 1.5285913528591356e-05, + "loss": 0.357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33576327562332153, + "step": 275, + "valid_targets_mean": 3043.9, + "valid_targets_min": 648 + }, + { + "epoch": 0.27370478983382207, + "grad_norm": 0.8054999780008923, + "learning_rate": 1.5564853556485358e-05, + "loss": 0.3544, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36960098147392273, + "step": 280, + "valid_targets_mean": 2678.3, + "valid_targets_min": 1027 + }, + { + "epoch": 0.2785923753665689, + "grad_norm": 0.9675054347484953, + "learning_rate": 1.584379358437936e-05, + "loss": 0.378, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3982788324356079, + "step": 285, + "valid_targets_mean": 1821.8, + "valid_targets_min": 701 + }, + { + "epoch": 0.28347996089931576, + "grad_norm": 0.9860609488043531, + "learning_rate": 1.6122733612273362e-05, + "loss": 0.3773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3843904137611389, + "step": 290, + "valid_targets_mean": 2513.7, + "valid_targets_min": 881 + }, + { + "epoch": 0.2883675464320626, + "grad_norm": 0.991847663374839, + "learning_rate": 1.6401673640167365e-05, + "loss": 0.379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4282916784286499, + "step": 295, + "valid_targets_mean": 2874.1, + "valid_targets_min": 1072 + }, + { + "epoch": 0.2932551319648094, + "grad_norm": 0.8307727140975414, + "learning_rate": 1.6680613668061367e-05, + "loss": 0.3548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34740254282951355, + "step": 300, + "valid_targets_mean": 2016.8, + "valid_targets_min": 681 + }, + { + "epoch": 0.2981427174975562, + "grad_norm": 0.7143528489591171, + "learning_rate": 1.6959553695955372e-05, + "loss": 0.3636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.359591007232666, + "step": 305, + "valid_targets_mean": 3070.4, + "valid_targets_min": 774 + }, + { + "epoch": 0.30303030303030304, + "grad_norm": 0.9089966300203021, + "learning_rate": 1.7238493723849375e-05, + "loss": 0.3498, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33375781774520874, + "step": 310, + "valid_targets_mean": 1869.1, + "valid_targets_min": 780 + }, + { + "epoch": 0.30791788856304986, + "grad_norm": 0.8634809970221224, + "learning_rate": 1.7517433751743377e-05, + "loss": 0.3555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37010622024536133, + "step": 315, + "valid_targets_mean": 2224.5, + "valid_targets_min": 754 + }, + { + "epoch": 0.3128054740957967, + "grad_norm": 0.8540663264632482, + "learning_rate": 1.779637377963738e-05, + "loss": 0.3515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3211752772331238, + "step": 320, + "valid_targets_mean": 2225.6, + "valid_targets_min": 777 + }, + { + "epoch": 0.3176930596285435, + "grad_norm": 0.8804814214418613, + "learning_rate": 1.807531380753138e-05, + "loss": 0.3376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35791587829589844, + "step": 325, + "valid_targets_mean": 2065.7, + "valid_targets_min": 548 + }, + { + "epoch": 0.3225806451612903, + "grad_norm": 0.7640563487552252, + "learning_rate": 1.8354253835425383e-05, + "loss": 0.3332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3520016074180603, + "step": 330, + "valid_targets_mean": 2701.5, + "valid_targets_min": 944 + }, + { + "epoch": 0.32746823069403713, + "grad_norm": 0.7296882324617866, + "learning_rate": 1.863319386331939e-05, + "loss": 0.3399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31554561853408813, + "step": 335, + "valid_targets_mean": 3010.9, + "valid_targets_min": 1211 + }, + { + "epoch": 0.33235581622678395, + "grad_norm": 0.7271443111223936, + "learning_rate": 1.891213389121339e-05, + "loss": 0.3478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3308175802230835, + "step": 340, + "valid_targets_mean": 2877.0, + "valid_targets_min": 1375 + }, + { + "epoch": 0.33724340175953077, + "grad_norm": 0.8360036681979609, + "learning_rate": 1.9191073919107393e-05, + "loss": 0.3671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34241533279418945, + "step": 345, + "valid_targets_mean": 2568.8, + "valid_targets_min": 656 + }, + { + "epoch": 0.3421309872922776, + "grad_norm": 0.8550409656130434, + "learning_rate": 1.9470013947001396e-05, + "loss": 0.352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3601570427417755, + "step": 350, + "valid_targets_mean": 2331.7, + "valid_targets_min": 679 + }, + { + "epoch": 0.34701857282502446, + "grad_norm": 0.9155877630691556, + "learning_rate": 1.9748953974895398e-05, + "loss": 0.3635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.38951539993286133, + "step": 355, + "valid_targets_mean": 2071.4, + "valid_targets_min": 760 + }, + { + "epoch": 0.3519061583577713, + "grad_norm": 0.883435595055635, + "learning_rate": 2.0027894002789404e-05, + "loss": 0.36, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3670802414417267, + "step": 360, + "valid_targets_mean": 1985.2, + "valid_targets_min": 592 + }, + { + "epoch": 0.3567937438905181, + "grad_norm": 0.9206852695774652, + "learning_rate": 2.0306834030683406e-05, + "loss": 0.3377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3370826840400696, + "step": 365, + "valid_targets_mean": 2055.2, + "valid_targets_min": 799 + }, + { + "epoch": 0.3616813294232649, + "grad_norm": 0.7244221713153408, + "learning_rate": 2.0585774058577405e-05, + "loss": 0.3354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36608511209487915, + "step": 370, + "valid_targets_mean": 3176.6, + "valid_targets_min": 1237 + }, + { + "epoch": 0.36656891495601174, + "grad_norm": 0.8515795547939923, + "learning_rate": 2.086471408647141e-05, + "loss": 0.3533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3597974181175232, + "step": 375, + "valid_targets_mean": 2919.8, + "valid_targets_min": 1215 + }, + { + "epoch": 0.37145650048875856, + "grad_norm": 0.8264921304378712, + "learning_rate": 2.1143654114365412e-05, + "loss": 0.3603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.364242821931839, + "step": 380, + "valid_targets_mean": 2541.9, + "valid_targets_min": 763 + }, + { + "epoch": 0.3763440860215054, + "grad_norm": 0.907898110015603, + "learning_rate": 2.1422594142259415e-05, + "loss": 0.341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32725921273231506, + "step": 385, + "valid_targets_mean": 1902.6, + "valid_targets_min": 464 + }, + { + "epoch": 0.3812316715542522, + "grad_norm": 0.938573710378532, + "learning_rate": 2.170153417015342e-05, + "loss": 0.3244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34030425548553467, + "step": 390, + "valid_targets_mean": 2139.0, + "valid_targets_min": 627 + }, + { + "epoch": 0.386119257086999, + "grad_norm": 0.9348906560243034, + "learning_rate": 2.1980474198047422e-05, + "loss": 0.3436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3462354838848114, + "step": 395, + "valid_targets_mean": 2194.8, + "valid_targets_min": 738 + }, + { + "epoch": 0.39100684261974583, + "grad_norm": 0.6752427727321951, + "learning_rate": 2.225941422594142e-05, + "loss": 0.3381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33034390211105347, + "step": 400, + "valid_targets_mean": 3065.8, + "valid_targets_min": 796 + }, + { + "epoch": 0.39589442815249265, + "grad_norm": 0.8105931491656845, + "learning_rate": 2.2538354253835427e-05, + "loss": 0.3508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3310199975967407, + "step": 405, + "valid_targets_mean": 2525.2, + "valid_targets_min": 900 + }, + { + "epoch": 0.40078201368523947, + "grad_norm": 0.8829736004010716, + "learning_rate": 2.281729428172943e-05, + "loss": 0.3529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3105641305446625, + "step": 410, + "valid_targets_mean": 2156.9, + "valid_targets_min": 805 + }, + { + "epoch": 0.4056695992179863, + "grad_norm": 1.053696655919182, + "learning_rate": 2.309623430962343e-05, + "loss": 0.3361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3739250898361206, + "step": 415, + "valid_targets_mean": 2282.6, + "valid_targets_min": 1170 + }, + { + "epoch": 0.41055718475073316, + "grad_norm": 0.809739823282899, + "learning_rate": 2.3375174337517437e-05, + "loss": 0.3311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35403841733932495, + "step": 420, + "valid_targets_mean": 2249.1, + "valid_targets_min": 698 + }, + { + "epoch": 0.41544477028348, + "grad_norm": 0.8114625727115713, + "learning_rate": 2.365411436541144e-05, + "loss": 0.3432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3622049391269684, + "step": 425, + "valid_targets_mean": 2313.3, + "valid_targets_min": 766 + }, + { + "epoch": 0.4203323558162268, + "grad_norm": 0.7569678066302454, + "learning_rate": 2.3933054393305438e-05, + "loss": 0.3358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33777353167533875, + "step": 430, + "valid_targets_mean": 2816.7, + "valid_targets_min": 859 + }, + { + "epoch": 0.4252199413489736, + "grad_norm": 0.7327933630295097, + "learning_rate": 2.4211994421199444e-05, + "loss": 0.3466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3273935317993164, + "step": 435, + "valid_targets_mean": 2645.7, + "valid_targets_min": 1098 + }, + { + "epoch": 0.43010752688172044, + "grad_norm": 0.8665670559962085, + "learning_rate": 2.4490934449093446e-05, + "loss": 0.3425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3423697054386139, + "step": 440, + "valid_targets_mean": 2181.8, + "valid_targets_min": 604 + }, + { + "epoch": 0.43499511241446726, + "grad_norm": 0.8580346981985594, + "learning_rate": 2.4769874476987448e-05, + "loss": 0.3384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34883952140808105, + "step": 445, + "valid_targets_mean": 2146.7, + "valid_targets_min": 638 + }, + { + "epoch": 0.4398826979472141, + "grad_norm": 0.8238090290163292, + "learning_rate": 2.5048814504881454e-05, + "loss": 0.3507, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31056028604507446, + "step": 450, + "valid_targets_mean": 2293.9, + "valid_targets_min": 943 + }, + { + "epoch": 0.4447702834799609, + "grad_norm": 0.6463651486346894, + "learning_rate": 2.5327754532775456e-05, + "loss": 0.3203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2803493142127991, + "step": 455, + "valid_targets_mean": 2967.7, + "valid_targets_min": 865 + }, + { + "epoch": 0.4496578690127077, + "grad_norm": 0.941311131272596, + "learning_rate": 2.5606694560669455e-05, + "loss": 0.3452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3155880570411682, + "step": 460, + "valid_targets_mean": 1743.8, + "valid_targets_min": 764 + }, + { + "epoch": 0.45454545454545453, + "grad_norm": 0.736803043029875, + "learning_rate": 2.588563458856346e-05, + "loss": 0.3442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36394965648651123, + "step": 465, + "valid_targets_mean": 2932.3, + "valid_targets_min": 763 + }, + { + "epoch": 0.45943304007820135, + "grad_norm": 0.7921374525507471, + "learning_rate": 2.6164574616457463e-05, + "loss": 0.328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3570030927658081, + "step": 470, + "valid_targets_mean": 3040.9, + "valid_targets_min": 796 + }, + { + "epoch": 0.46432062561094817, + "grad_norm": 0.7900857884761241, + "learning_rate": 2.6443514644351465e-05, + "loss": 0.342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37542372941970825, + "step": 475, + "valid_targets_mean": 3116.3, + "valid_targets_min": 926 + }, + { + "epoch": 0.46920821114369504, + "grad_norm": 1.159582734344674, + "learning_rate": 2.672245467224547e-05, + "loss": 0.3439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36254942417144775, + "step": 480, + "valid_targets_mean": 2185.8, + "valid_targets_min": 713 + }, + { + "epoch": 0.47409579667644186, + "grad_norm": 0.7897072764788433, + "learning_rate": 2.7001394700139473e-05, + "loss": 0.3372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33607053756713867, + "step": 485, + "valid_targets_mean": 2364.0, + "valid_targets_min": 1086 + }, + { + "epoch": 0.4789833822091887, + "grad_norm": 0.8400859552438328, + "learning_rate": 2.728033472803347e-05, + "loss": 0.3359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3387121856212616, + "step": 490, + "valid_targets_mean": 2131.7, + "valid_targets_min": 640 + }, + { + "epoch": 0.4838709677419355, + "grad_norm": 0.7446625512121486, + "learning_rate": 2.7559274755927477e-05, + "loss": 0.334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3753664493560791, + "step": 495, + "valid_targets_mean": 2741.8, + "valid_targets_min": 548 + }, + { + "epoch": 0.4887585532746823, + "grad_norm": 0.7424189633962999, + "learning_rate": 2.783821478382148e-05, + "loss": 0.3449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35922667384147644, + "step": 500, + "valid_targets_mean": 3106.8, + "valid_targets_min": 668 + }, + { + "epoch": 0.49364613880742914, + "grad_norm": 0.8398560921552762, + "learning_rate": 2.811715481171548e-05, + "loss": 0.3178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27231159806251526, + "step": 505, + "valid_targets_mean": 2408.9, + "valid_targets_min": 742 + }, + { + "epoch": 0.49853372434017595, + "grad_norm": 0.8650127139587513, + "learning_rate": 2.8396094839609487e-05, + "loss": 0.3244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31993523240089417, + "step": 510, + "valid_targets_mean": 2272.2, + "valid_targets_min": 672 + }, + { + "epoch": 0.5034213098729228, + "grad_norm": 0.8145474264636156, + "learning_rate": 2.867503486750349e-05, + "loss": 0.3321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3184545040130615, + "step": 515, + "valid_targets_mean": 2181.7, + "valid_targets_min": 676 + }, + { + "epoch": 0.5083088954056696, + "grad_norm": 0.8995961884293948, + "learning_rate": 2.8953974895397488e-05, + "loss": 0.3233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34413954615592957, + "step": 520, + "valid_targets_mean": 2101.1, + "valid_targets_min": 1055 + }, + { + "epoch": 0.5131964809384164, + "grad_norm": 0.7980166217846976, + "learning_rate": 2.9232914923291494e-05, + "loss": 0.3308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3372722864151001, + "step": 525, + "valid_targets_mean": 2395.6, + "valid_targets_min": 1282 + }, + { + "epoch": 0.5180840664711632, + "grad_norm": 1.1201293160921861, + "learning_rate": 2.9511854951185496e-05, + "loss": 0.323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31446385383605957, + "step": 530, + "valid_targets_mean": 2134.9, + "valid_targets_min": 922 + }, + { + "epoch": 0.52297165200391, + "grad_norm": 0.7342360486498539, + "learning_rate": 2.9790794979079498e-05, + "loss": 0.3321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30906954407691956, + "step": 535, + "valid_targets_mean": 2658.1, + "valid_targets_min": 1076 + }, + { + "epoch": 0.5278592375366569, + "grad_norm": 0.7267057803106786, + "learning_rate": 3.0069735006973504e-05, + "loss": 0.32, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3381599187850952, + "step": 540, + "valid_targets_mean": 2644.8, + "valid_targets_min": 814 + }, + { + "epoch": 0.5327468230694037, + "grad_norm": 0.8500045196292703, + "learning_rate": 3.0348675034867506e-05, + "loss": 0.3205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29569292068481445, + "step": 545, + "valid_targets_mean": 1866.9, + "valid_targets_min": 530 + }, + { + "epoch": 0.5376344086021505, + "grad_norm": 0.7412443952817233, + "learning_rate": 3.062761506276151e-05, + "loss": 0.3355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.345531165599823, + "step": 550, + "valid_targets_mean": 2880.6, + "valid_targets_min": 905 + }, + { + "epoch": 0.5425219941348973, + "grad_norm": 0.6993041939598253, + "learning_rate": 3.0906555090655514e-05, + "loss": 0.3124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3102779984474182, + "step": 555, + "valid_targets_mean": 2846.2, + "valid_targets_min": 790 + }, + { + "epoch": 0.5474095796676441, + "grad_norm": 0.7359314473756243, + "learning_rate": 3.118549511854951e-05, + "loss": 0.3203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32391446828842163, + "step": 560, + "valid_targets_mean": 2719.4, + "valid_targets_min": 816 + }, + { + "epoch": 0.552297165200391, + "grad_norm": 0.8113820976460382, + "learning_rate": 3.146443514644352e-05, + "loss": 0.3409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30669841170310974, + "step": 565, + "valid_targets_mean": 2390.4, + "valid_targets_min": 669 + }, + { + "epoch": 0.5571847507331378, + "grad_norm": 0.8096140445105596, + "learning_rate": 3.1743375174337524e-05, + "loss": 0.2977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3120150566101074, + "step": 570, + "valid_targets_mean": 2190.9, + "valid_targets_min": 739 + }, + { + "epoch": 0.5620723362658846, + "grad_norm": 0.9110568245840788, + "learning_rate": 3.202231520223152e-05, + "loss": 0.314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3165706992149353, + "step": 575, + "valid_targets_mean": 1853.4, + "valid_targets_min": 722 + }, + { + "epoch": 0.5669599217986315, + "grad_norm": 0.7684538091148991, + "learning_rate": 3.230125523012553e-05, + "loss": 0.3162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2956627607345581, + "step": 580, + "valid_targets_mean": 2559.9, + "valid_targets_min": 987 + }, + { + "epoch": 0.5718475073313783, + "grad_norm": 0.7441624896883247, + "learning_rate": 3.258019525801953e-05, + "loss": 0.3292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3297075033187866, + "step": 585, + "valid_targets_mean": 2662.0, + "valid_targets_min": 711 + }, + { + "epoch": 0.5767350928641252, + "grad_norm": 0.8719735596323135, + "learning_rate": 3.285913528591353e-05, + "loss": 0.3205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3282514214515686, + "step": 590, + "valid_targets_mean": 2034.2, + "valid_targets_min": 851 + }, + { + "epoch": 0.581622678396872, + "grad_norm": 0.8090870634352967, + "learning_rate": 3.313807531380754e-05, + "loss": 0.3329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31386953592300415, + "step": 595, + "valid_targets_mean": 2551.9, + "valid_targets_min": 914 + }, + { + "epoch": 0.5865102639296188, + "grad_norm": 0.7207352022007785, + "learning_rate": 3.341701534170154e-05, + "loss": 0.3237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3386681079864502, + "step": 600, + "valid_targets_mean": 2835.5, + "valid_targets_min": 661 + }, + { + "epoch": 0.5913978494623656, + "grad_norm": 0.7584008460860305, + "learning_rate": 3.3695955369595536e-05, + "loss": 0.3284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28931134939193726, + "step": 605, + "valid_targets_mean": 2277.4, + "valid_targets_min": 1014 + }, + { + "epoch": 0.5962854349951124, + "grad_norm": 0.792721139679117, + "learning_rate": 3.397489539748954e-05, + "loss": 0.3312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33046597242355347, + "step": 610, + "valid_targets_mean": 2322.7, + "valid_targets_min": 704 + }, + { + "epoch": 0.6011730205278593, + "grad_norm": 0.7784271949093392, + "learning_rate": 3.425383542538355e-05, + "loss": 0.3356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3591078519821167, + "step": 615, + "valid_targets_mean": 2750.8, + "valid_targets_min": 929 + }, + { + "epoch": 0.6060606060606061, + "grad_norm": 0.8776498743873912, + "learning_rate": 3.4532775453277546e-05, + "loss": 0.3143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33385688066482544, + "step": 620, + "valid_targets_mean": 2140.5, + "valid_targets_min": 642 + }, + { + "epoch": 0.6109481915933529, + "grad_norm": 0.746820447052271, + "learning_rate": 3.481171548117155e-05, + "loss": 0.3296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33402806520462036, + "step": 625, + "valid_targets_mean": 2436.4, + "valid_targets_min": 895 + }, + { + "epoch": 0.6158357771260997, + "grad_norm": 0.7897960735140778, + "learning_rate": 3.509065550906556e-05, + "loss": 0.3214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31994372606277466, + "step": 630, + "valid_targets_mean": 2388.4, + "valid_targets_min": 915 + }, + { + "epoch": 0.6207233626588465, + "grad_norm": 0.7255296236144402, + "learning_rate": 3.5369595536959556e-05, + "loss": 0.32, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3037095069885254, + "step": 635, + "valid_targets_mean": 2603.6, + "valid_targets_min": 635 + }, + { + "epoch": 0.6256109481915934, + "grad_norm": 0.8068527468176921, + "learning_rate": 3.564853556485356e-05, + "loss": 0.3123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31424158811569214, + "step": 640, + "valid_targets_mean": 2459.6, + "valid_targets_min": 636 + }, + { + "epoch": 0.6304985337243402, + "grad_norm": 0.7702625587232006, + "learning_rate": 3.592747559274756e-05, + "loss": 0.3358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3524268567562103, + "step": 645, + "valid_targets_mean": 2649.3, + "valid_targets_min": 1140 + }, + { + "epoch": 0.635386119257087, + "grad_norm": 0.7714466133072925, + "learning_rate": 3.6206415620641566e-05, + "loss": 0.3286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31589627265930176, + "step": 650, + "valid_targets_mean": 2330.5, + "valid_targets_min": 657 + }, + { + "epoch": 0.6402737047898338, + "grad_norm": 0.7977408416406456, + "learning_rate": 3.648535564853557e-05, + "loss": 0.3244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29703885316848755, + "step": 655, + "valid_targets_mean": 2082.4, + "valid_targets_min": 725 + }, + { + "epoch": 0.6451612903225806, + "grad_norm": 0.897859600419085, + "learning_rate": 3.676429567642957e-05, + "loss": 0.3072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32600027322769165, + "step": 660, + "valid_targets_mean": 2019.0, + "valid_targets_min": 740 + }, + { + "epoch": 0.6500488758553274, + "grad_norm": 0.8217894407117475, + "learning_rate": 3.704323570432357e-05, + "loss": 0.3197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2989169955253601, + "step": 665, + "valid_targets_mean": 2451.2, + "valid_targets_min": 1105 + }, + { + "epoch": 0.6549364613880743, + "grad_norm": 0.6967345052464025, + "learning_rate": 3.7322175732217575e-05, + "loss": 0.298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28789329528808594, + "step": 670, + "valid_targets_mean": 2827.7, + "valid_targets_min": 692 + }, + { + "epoch": 0.6598240469208211, + "grad_norm": 0.730001904109528, + "learning_rate": 3.760111576011158e-05, + "loss": 0.3068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31676405668258667, + "step": 675, + "valid_targets_mean": 2389.0, + "valid_targets_min": 660 + }, + { + "epoch": 0.6647116324535679, + "grad_norm": 0.7898161538858068, + "learning_rate": 3.788005578800558e-05, + "loss": 0.3168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29914867877960205, + "step": 680, + "valid_targets_mean": 2101.6, + "valid_targets_min": 894 + }, + { + "epoch": 0.6695992179863147, + "grad_norm": 0.7857150534119379, + "learning_rate": 3.8158995815899585e-05, + "loss": 0.3295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3327745199203491, + "step": 685, + "valid_targets_mean": 2511.1, + "valid_targets_min": 744 + }, + { + "epoch": 0.6744868035190615, + "grad_norm": 0.6691932395126847, + "learning_rate": 3.843793584379359e-05, + "loss": 0.3211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3309377431869507, + "step": 690, + "valid_targets_mean": 2951.6, + "valid_targets_min": 644 + }, + { + "epoch": 0.6793743890518084, + "grad_norm": 1.1193866441330809, + "learning_rate": 3.871687587168759e-05, + "loss": 0.3372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36905205249786377, + "step": 695, + "valid_targets_mean": 2194.1, + "valid_targets_min": 818 + }, + { + "epoch": 0.6842619745845552, + "grad_norm": 0.7857310721423778, + "learning_rate": 3.8995815899581595e-05, + "loss": 0.3273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30003082752227783, + "step": 700, + "valid_targets_mean": 2174.9, + "valid_targets_min": 557 + }, + { + "epoch": 0.6891495601173021, + "grad_norm": 0.7397963727859, + "learning_rate": 3.9274755927475594e-05, + "loss": 0.3207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2808341383934021, + "step": 705, + "valid_targets_mean": 2378.6, + "valid_targets_min": 794 + }, + { + "epoch": 0.6940371456500489, + "grad_norm": 1.068416619253061, + "learning_rate": 3.95536959553696e-05, + "loss": 0.3241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32398268580436707, + "step": 710, + "valid_targets_mean": 1853.6, + "valid_targets_min": 791 + }, + { + "epoch": 0.6989247311827957, + "grad_norm": 0.7883949750339846, + "learning_rate": 3.9832635983263605e-05, + "loss": 0.307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3216050863265991, + "step": 715, + "valid_targets_mean": 2201.4, + "valid_targets_min": 661 + }, + { + "epoch": 0.7038123167155426, + "grad_norm": 0.7210396481396164, + "learning_rate": 3.999999049288738e-05, + "loss": 0.3086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32969582080841064, + "step": 720, + "valid_targets_mean": 2724.2, + "valid_targets_min": 572 + }, + { + "epoch": 0.7086999022482894, + "grad_norm": 0.6780724521617006, + "learning_rate": 3.9999883537974135e-05, + "loss": 0.3123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3392641842365265, + "step": 725, + "valid_targets_mean": 2957.1, + "valid_targets_min": 856 + }, + { + "epoch": 0.7135874877810362, + "grad_norm": 1.4001056248209713, + "learning_rate": 3.999965774489451e-05, + "loss": 0.3241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31240707635879517, + "step": 730, + "valid_targets_mean": 2448.2, + "valid_targets_min": 740 + }, + { + "epoch": 0.718475073313783, + "grad_norm": 0.7066267173981833, + "learning_rate": 3.999931311499014e-05, + "loss": 0.2993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2746642827987671, + "step": 735, + "valid_targets_mean": 2677.8, + "valid_targets_min": 1280 + }, + { + "epoch": 0.7233626588465298, + "grad_norm": 0.9463337394740791, + "learning_rate": 3.999884965030881e-05, + "loss": 0.3138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34243521094322205, + "step": 740, + "valid_targets_mean": 1775.4, + "valid_targets_min": 679 + }, + { + "epoch": 0.7282502443792767, + "grad_norm": 0.7034930040606712, + "learning_rate": 3.9998267353604396e-05, + "loss": 0.3083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3177798390388489, + "step": 745, + "valid_targets_mean": 2800.6, + "valid_targets_min": 920 + }, + { + "epoch": 0.7331378299120235, + "grad_norm": 0.6916689022053226, + "learning_rate": 3.9997566228336865e-05, + "loss": 0.3207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.289054274559021, + "step": 750, + "valid_targets_mean": 2406.8, + "valid_targets_min": 856 + }, + { + "epoch": 0.7380254154447703, + "grad_norm": 0.7430166452541812, + "learning_rate": 3.999674627867228e-05, + "loss": 0.3008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29457688331604004, + "step": 755, + "valid_targets_mean": 2496.6, + "valid_targets_min": 804 + }, + { + "epoch": 0.7429130009775171, + "grad_norm": 0.7604153844606187, + "learning_rate": 3.9995807509482715e-05, + "loss": 0.3298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37004536390304565, + "step": 760, + "valid_targets_mean": 2459.6, + "valid_targets_min": 910 + }, + { + "epoch": 0.7478005865102639, + "grad_norm": 0.7783050791408205, + "learning_rate": 3.99947499263463e-05, + "loss": 0.3071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26048505306243896, + "step": 765, + "valid_targets_mean": 1702.1, + "valid_targets_min": 641 + }, + { + "epoch": 0.7526881720430108, + "grad_norm": 0.6842291009472856, + "learning_rate": 3.999357353554712e-05, + "loss": 0.2899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2890620529651642, + "step": 770, + "valid_targets_mean": 2460.7, + "valid_targets_min": 741 + }, + { + "epoch": 0.7575757575757576, + "grad_norm": 0.7667053941772743, + "learning_rate": 3.999227834407523e-05, + "loss": 0.3122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3236725330352783, + "step": 775, + "valid_targets_mean": 2326.1, + "valid_targets_min": 1024 + }, + { + "epoch": 0.7624633431085044, + "grad_norm": 0.9459473272945332, + "learning_rate": 3.9990864359626586e-05, + "loss": 0.2866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2749725580215454, + "step": 780, + "valid_targets_mean": 2354.9, + "valid_targets_min": 914 + }, + { + "epoch": 0.7673509286412512, + "grad_norm": 0.7184999480991484, + "learning_rate": 3.9989331590603e-05, + "loss": 0.3203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3461282253265381, + "step": 785, + "valid_targets_mean": 2811.2, + "valid_targets_min": 725 + }, + { + "epoch": 0.772238514173998, + "grad_norm": 0.6593556008310089, + "learning_rate": 3.998768004611209e-05, + "loss": 0.3225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35281455516815186, + "step": 790, + "valid_targets_mean": 3194.9, + "valid_targets_min": 1180 + }, + { + "epoch": 0.7771260997067448, + "grad_norm": 0.7230785157141715, + "learning_rate": 3.998590973596725e-05, + "loss": 0.3356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3064287304878235, + "step": 795, + "valid_targets_mean": 2452.0, + "valid_targets_min": 786 + }, + { + "epoch": 0.7820136852394917, + "grad_norm": 0.6934289312826055, + "learning_rate": 3.998402067068755e-05, + "loss": 0.3336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3232996165752411, + "step": 800, + "valid_targets_mean": 2586.8, + "valid_targets_min": 1015 + }, + { + "epoch": 0.7869012707722385, + "grad_norm": 0.6558930070169424, + "learning_rate": 3.998201286149772e-05, + "loss": 0.3206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33236560225486755, + "step": 805, + "valid_targets_mean": 3404.1, + "valid_targets_min": 660 + }, + { + "epoch": 0.7917888563049853, + "grad_norm": 0.7295217695663251, + "learning_rate": 3.997988632032805e-05, + "loss": 0.3012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3009946346282959, + "step": 810, + "valid_targets_mean": 2459.5, + "valid_targets_min": 970 + }, + { + "epoch": 0.7966764418377321, + "grad_norm": 0.8584536403730917, + "learning_rate": 3.99776410598143e-05, + "loss": 0.3159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32543686032295227, + "step": 815, + "valid_targets_mean": 2278.4, + "valid_targets_min": 852 + }, + { + "epoch": 0.8015640273704789, + "grad_norm": 0.7449337414001902, + "learning_rate": 3.997527709329772e-05, + "loss": 0.317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3691062927246094, + "step": 820, + "valid_targets_mean": 2001.3, + "valid_targets_min": 625 + }, + { + "epoch": 0.8064516129032258, + "grad_norm": 0.7287752989474822, + "learning_rate": 3.997279443482484e-05, + "loss": 0.3146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3165518045425415, + "step": 825, + "valid_targets_mean": 2824.4, + "valid_targets_min": 1123 + }, + { + "epoch": 0.8113391984359726, + "grad_norm": 0.7414141117963199, + "learning_rate": 3.997019309914747e-05, + "loss": 0.3146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3582077622413635, + "step": 830, + "valid_targets_mean": 2533.0, + "valid_targets_min": 922 + }, + { + "epoch": 0.8162267839687195, + "grad_norm": 0.6604646667995736, + "learning_rate": 3.996747310172261e-05, + "loss": 0.2997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3009747266769409, + "step": 835, + "valid_targets_mean": 2743.4, + "valid_targets_min": 815 + }, + { + "epoch": 0.8211143695014663, + "grad_norm": 0.7869481152483848, + "learning_rate": 3.996463445871233e-05, + "loss": 0.3012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28774186968803406, + "step": 840, + "valid_targets_mean": 2012.9, + "valid_targets_min": 1019 + }, + { + "epoch": 0.8260019550342131, + "grad_norm": 0.7796967007238236, + "learning_rate": 3.996167718698368e-05, + "loss": 0.2991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28823673725128174, + "step": 845, + "valid_targets_mean": 1971.4, + "valid_targets_min": 858 + }, + { + "epoch": 0.83088954056696, + "grad_norm": 0.7921234290446715, + "learning_rate": 3.9958601304108596e-05, + "loss": 0.3121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29436832666397095, + "step": 850, + "valid_targets_mean": 2040.2, + "valid_targets_min": 780 + }, + { + "epoch": 0.8357771260997068, + "grad_norm": 0.7773209045309674, + "learning_rate": 3.995540682836381e-05, + "loss": 0.3047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.320604145526886, + "step": 855, + "valid_targets_mean": 2100.3, + "valid_targets_min": 635 + }, + { + "epoch": 0.8406647116324536, + "grad_norm": 0.7068147985901192, + "learning_rate": 3.99520937787307e-05, + "loss": 0.3111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2897074818611145, + "step": 860, + "valid_targets_mean": 2451.1, + "valid_targets_min": 781 + }, + { + "epoch": 0.8455522971652004, + "grad_norm": 0.6227835458943253, + "learning_rate": 3.994866217489523e-05, + "loss": 0.3086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2862326502799988, + "step": 865, + "valid_targets_mean": 3045.1, + "valid_targets_min": 892 + }, + { + "epoch": 0.8504398826979472, + "grad_norm": 0.8510089182055716, + "learning_rate": 3.994511203724779e-05, + "loss": 0.2996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31772327423095703, + "step": 870, + "valid_targets_mean": 2259.7, + "valid_targets_min": 900 + }, + { + "epoch": 0.855327468230694, + "grad_norm": 0.7031577446655919, + "learning_rate": 3.9941443386883086e-05, + "loss": 0.3074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3305051922798157, + "step": 875, + "valid_targets_mean": 2753.0, + "valid_targets_min": 976 + }, + { + "epoch": 0.8602150537634409, + "grad_norm": 0.7949638614035321, + "learning_rate": 3.9937656245600044e-05, + "loss": 0.3089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3418285846710205, + "step": 880, + "valid_targets_mean": 2385.2, + "valid_targets_min": 804 + }, + { + "epoch": 0.8651026392961877, + "grad_norm": 0.7305187723607021, + "learning_rate": 3.993375063590164e-05, + "loss": 0.3138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3128136992454529, + "step": 885, + "valid_targets_mean": 2648.4, + "valid_targets_min": 1208 + }, + { + "epoch": 0.8699902248289345, + "grad_norm": 1.1746959130401167, + "learning_rate": 3.9929726580994775e-05, + "loss": 0.306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2941882610321045, + "step": 890, + "valid_targets_mean": 2617.6, + "valid_targets_min": 904 + }, + { + "epoch": 0.8748778103616813, + "grad_norm": 0.7402990200201592, + "learning_rate": 3.992558410479015e-05, + "loss": 0.2926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2941339612007141, + "step": 895, + "valid_targets_mean": 2542.5, + "valid_targets_min": 735 + }, + { + "epoch": 0.8797653958944281, + "grad_norm": 0.7737988843199513, + "learning_rate": 3.992132323190214e-05, + "loss": 0.2966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2719045579433441, + "step": 900, + "valid_targets_mean": 1987.3, + "valid_targets_min": 878 + }, + { + "epoch": 0.884652981427175, + "grad_norm": 0.7582651606562361, + "learning_rate": 3.99169439876486e-05, + "loss": 0.3268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3535369634628296, + "step": 905, + "valid_targets_mean": 2560.1, + "valid_targets_min": 870 + }, + { + "epoch": 0.8895405669599218, + "grad_norm": 0.7699462585158438, + "learning_rate": 3.991244639805074e-05, + "loss": 0.3065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3052283823490143, + "step": 910, + "valid_targets_mean": 2282.6, + "valid_targets_min": 900 + }, + { + "epoch": 0.8944281524926686, + "grad_norm": 0.7785182577824628, + "learning_rate": 3.9907830489832993e-05, + "loss": 0.3019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26899123191833496, + "step": 915, + "valid_targets_mean": 2104.2, + "valid_targets_min": 676 + }, + { + "epoch": 0.8993157380254154, + "grad_norm": 0.8574210895658473, + "learning_rate": 3.990309629042281e-05, + "loss": 0.3052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2943177819252014, + "step": 920, + "valid_targets_mean": 2444.8, + "valid_targets_min": 1100 + }, + { + "epoch": 0.9042033235581622, + "grad_norm": 0.7284561255560225, + "learning_rate": 3.989824382795055e-05, + "loss": 0.3046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29493361711502075, + "step": 925, + "valid_targets_mean": 2599.8, + "valid_targets_min": 801 + }, + { + "epoch": 0.9090909090909091, + "grad_norm": 0.6857751641162575, + "learning_rate": 3.989327313124926e-05, + "loss": 0.3186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3466421961784363, + "step": 930, + "valid_targets_mean": 2798.0, + "valid_targets_min": 911 + }, + { + "epoch": 0.9139784946236559, + "grad_norm": 0.7493847771371742, + "learning_rate": 3.9888184229854525e-05, + "loss": 0.3163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32370316982269287, + "step": 935, + "valid_targets_mean": 2422.5, + "valid_targets_min": 703 + }, + { + "epoch": 0.9188660801564027, + "grad_norm": 0.6989351581606708, + "learning_rate": 3.988297715400432e-05, + "loss": 0.2974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2930717468261719, + "step": 940, + "valid_targets_mean": 2921.4, + "valid_targets_min": 1264 + }, + { + "epoch": 0.9237536656891495, + "grad_norm": 0.7430743004739445, + "learning_rate": 3.987765193463879e-05, + "loss": 0.2861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25492438673973083, + "step": 945, + "valid_targets_mean": 2135.5, + "valid_targets_min": 800 + }, + { + "epoch": 0.9286412512218963, + "grad_norm": 0.6670033217086339, + "learning_rate": 3.987220860340009e-05, + "loss": 0.3058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3239936828613281, + "step": 950, + "valid_targets_mean": 3289.5, + "valid_targets_min": 715 + }, + { + "epoch": 0.9335288367546432, + "grad_norm": 0.6702989912107715, + "learning_rate": 3.9866647192632176e-05, + "loss": 0.2958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33095431327819824, + "step": 955, + "valid_targets_mean": 2921.6, + "valid_targets_min": 1062 + }, + { + "epoch": 0.9384164222873901, + "grad_norm": 0.6929820388448079, + "learning_rate": 3.986096773538063e-05, + "loss": 0.3096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3430343270301819, + "step": 960, + "valid_targets_mean": 2701.9, + "valid_targets_min": 1179 + }, + { + "epoch": 0.9433040078201369, + "grad_norm": 0.7185025063166174, + "learning_rate": 3.985517026539247e-05, + "loss": 0.3038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29739582538604736, + "step": 965, + "valid_targets_mean": 2256.4, + "valid_targets_min": 980 + }, + { + "epoch": 0.9481915933528837, + "grad_norm": 0.7330014703291616, + "learning_rate": 3.9849254817115925e-05, + "loss": 0.3009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30947941541671753, + "step": 970, + "valid_targets_mean": 2393.1, + "valid_targets_min": 814 + }, + { + "epoch": 0.9530791788856305, + "grad_norm": 0.692392838579322, + "learning_rate": 3.984322142570026e-05, + "loss": 0.2948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2626379728317261, + "step": 975, + "valid_targets_mean": 2232.8, + "valid_targets_min": 854 + }, + { + "epoch": 0.9579667644183774, + "grad_norm": 0.7858755039683706, + "learning_rate": 3.983707012699554e-05, + "loss": 0.292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31549689173698425, + "step": 980, + "valid_targets_mean": 2162.8, + "valid_targets_min": 967 + }, + { + "epoch": 0.9628543499511242, + "grad_norm": 1.1798359180734284, + "learning_rate": 3.983080095755242e-05, + "loss": 0.2951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30286744236946106, + "step": 985, + "valid_targets_mean": 2232.1, + "valid_targets_min": 612 + }, + { + "epoch": 0.967741935483871, + "grad_norm": 0.6652279763756077, + "learning_rate": 3.982441395462196e-05, + "loss": 0.3031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2915736436843872, + "step": 990, + "valid_targets_mean": 2793.2, + "valid_targets_min": 903 + }, + { + "epoch": 0.9726295210166178, + "grad_norm": 0.7934359151694675, + "learning_rate": 3.9817909156155364e-05, + "loss": 0.2959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3128136396408081, + "step": 995, + "valid_targets_mean": 2334.7, + "valid_targets_min": 642 + }, + { + "epoch": 0.9775171065493646, + "grad_norm": 0.6891936895341518, + "learning_rate": 3.981128660080377e-05, + "loss": 0.2905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2969144880771637, + "step": 1000, + "valid_targets_mean": 2541.0, + "valid_targets_min": 634 + }, + { + "epoch": 0.9824046920821115, + "grad_norm": 0.6463480054438933, + "learning_rate": 3.9804546327918025e-05, + "loss": 0.3045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29501742124557495, + "step": 1005, + "valid_targets_mean": 2564.6, + "valid_targets_min": 714 + }, + { + "epoch": 0.9872922776148583, + "grad_norm": 0.6654896254454947, + "learning_rate": 3.9797688377548443e-05, + "loss": 0.2974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30969366431236267, + "step": 1010, + "valid_targets_mean": 3066.6, + "valid_targets_min": 682 + }, + { + "epoch": 0.9921798631476051, + "grad_norm": 0.6711120956545672, + "learning_rate": 3.979071279044458e-05, + "loss": 0.3167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3256608247756958, + "step": 1015, + "valid_targets_mean": 2601.2, + "valid_targets_min": 1157 + }, + { + "epoch": 0.9970674486803519, + "grad_norm": 0.7414237101887737, + "learning_rate": 3.9783619608054976e-05, + "loss": 0.3275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34451091289520264, + "step": 1020, + "valid_targets_mean": 2595.6, + "valid_targets_min": 1188 + }, + { + "epoch": 1.0019550342130987, + "grad_norm": 0.6693075231319553, + "learning_rate": 3.977640887252691e-05, + "loss": 0.3112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29463478922843933, + "step": 1025, + "valid_targets_mean": 2843.2, + "valid_targets_min": 985 + }, + { + "epoch": 1.0068426197458455, + "grad_norm": 0.7374854812974025, + "learning_rate": 3.976908062670616e-05, + "loss": 0.2789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2664167284965515, + "step": 1030, + "valid_targets_mean": 2574.8, + "valid_targets_min": 1291 + }, + { + "epoch": 1.0117302052785924, + "grad_norm": 0.7176161611789769, + "learning_rate": 3.976163491413675e-05, + "loss": 0.2899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3102620840072632, + "step": 1035, + "valid_targets_mean": 2583.6, + "valid_targets_min": 964 + }, + { + "epoch": 1.0166177908113392, + "grad_norm": 0.6120097146085149, + "learning_rate": 3.9754071779060674e-05, + "loss": 0.2955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29326507449150085, + "step": 1040, + "valid_targets_mean": 3918.3, + "valid_targets_min": 812 + }, + { + "epoch": 1.021505376344086, + "grad_norm": 0.6381885490796083, + "learning_rate": 3.9746391266417664e-05, + "loss": 0.2767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2916497588157654, + "step": 1045, + "valid_targets_mean": 3250.8, + "valid_targets_min": 672 + }, + { + "epoch": 1.0263929618768328, + "grad_norm": 0.6912845827934561, + "learning_rate": 3.973859342184487e-05, + "loss": 0.3046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3102365732192993, + "step": 1050, + "valid_targets_mean": 3456.9, + "valid_targets_min": 1028 + }, + { + "epoch": 1.0312805474095796, + "grad_norm": 0.6994698165553925, + "learning_rate": 3.973067829167664e-05, + "loss": 0.2847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2805289030075073, + "step": 1055, + "valid_targets_mean": 2511.4, + "valid_targets_min": 1131 + }, + { + "epoch": 1.0361681329423265, + "grad_norm": 0.7049190368603762, + "learning_rate": 3.972264592294424e-05, + "loss": 0.285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27423086762428284, + "step": 1060, + "valid_targets_mean": 2325.2, + "valid_targets_min": 617 + }, + { + "epoch": 1.0410557184750733, + "grad_norm": 0.803585152797481, + "learning_rate": 3.971449636337553e-05, + "loss": 0.2836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3054800033569336, + "step": 1065, + "valid_targets_mean": 2130.8, + "valid_targets_min": 633 + }, + { + "epoch": 1.04594330400782, + "grad_norm": 0.8125386640556346, + "learning_rate": 3.970622966139474e-05, + "loss": 0.2796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3040192723274231, + "step": 1070, + "valid_targets_mean": 2165.8, + "valid_targets_min": 726 + }, + { + "epoch": 1.050830889540567, + "grad_norm": 0.6875765301460454, + "learning_rate": 3.969784586612213e-05, + "loss": 0.2682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2739431858062744, + "step": 1075, + "valid_targets_mean": 2306.1, + "valid_targets_min": 750 + }, + { + "epoch": 1.0557184750733137, + "grad_norm": 0.7590315528229628, + "learning_rate": 3.968934502737375e-05, + "loss": 0.2849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28198593854904175, + "step": 1080, + "valid_targets_mean": 2487.6, + "valid_targets_min": 703 + }, + { + "epoch": 1.0606060606060606, + "grad_norm": 0.767363949833768, + "learning_rate": 3.968072719566108e-05, + "loss": 0.2657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2522793114185333, + "step": 1085, + "valid_targets_mean": 1953.9, + "valid_targets_min": 746 + }, + { + "epoch": 1.0654936461388074, + "grad_norm": 0.763111692809638, + "learning_rate": 3.96719924221908e-05, + "loss": 0.2741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31584176421165466, + "step": 1090, + "valid_targets_mean": 2352.2, + "valid_targets_min": 788 + }, + { + "epoch": 1.0703812316715542, + "grad_norm": 0.7106137161882315, + "learning_rate": 3.966314075886442e-05, + "loss": 0.2855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2609940767288208, + "step": 1095, + "valid_targets_mean": 2245.1, + "valid_targets_min": 867 + }, + { + "epoch": 1.075268817204301, + "grad_norm": 0.8801431366949057, + "learning_rate": 3.9654172258278026e-05, + "loss": 0.2708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26888507604599, + "step": 1100, + "valid_targets_mean": 2219.1, + "valid_targets_min": 737 + }, + { + "epoch": 1.0801564027370478, + "grad_norm": 0.9630113421733234, + "learning_rate": 3.964508697372194e-05, + "loss": 0.2862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2783164083957672, + "step": 1105, + "valid_targets_mean": 2106.6, + "valid_targets_min": 670 + }, + { + "epoch": 1.0850439882697946, + "grad_norm": 0.7056624571864765, + "learning_rate": 3.96358849591804e-05, + "loss": 0.2787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2580503821372986, + "step": 1110, + "valid_targets_mean": 2245.5, + "valid_targets_min": 830 + }, + { + "epoch": 1.0899315738025415, + "grad_norm": 0.7476404563873276, + "learning_rate": 3.9626566269331254e-05, + "loss": 0.2637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25160929560661316, + "step": 1115, + "valid_targets_mean": 2236.5, + "valid_targets_min": 824 + }, + { + "epoch": 1.0948191593352883, + "grad_norm": 0.7462231828953746, + "learning_rate": 3.9617130959545614e-05, + "loss": 0.287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28869083523750305, + "step": 1120, + "valid_targets_mean": 2205.5, + "valid_targets_min": 548 + }, + { + "epoch": 1.099706744868035, + "grad_norm": 0.6350210512856392, + "learning_rate": 3.960757908588757e-05, + "loss": 0.2792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29355528950691223, + "step": 1125, + "valid_targets_mean": 2905.8, + "valid_targets_min": 990 + }, + { + "epoch": 1.104594330400782, + "grad_norm": 0.7768451815168522, + "learning_rate": 3.95979107051138e-05, + "loss": 0.298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29628539085388184, + "step": 1130, + "valid_targets_mean": 1995.6, + "valid_targets_min": 796 + }, + { + "epoch": 1.1094819159335287, + "grad_norm": 0.715701114754943, + "learning_rate": 3.958812587467327e-05, + "loss": 0.2946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2757974863052368, + "step": 1135, + "valid_targets_mean": 2513.9, + "valid_targets_min": 795 + }, + { + "epoch": 1.1143695014662756, + "grad_norm": 0.657177833956688, + "learning_rate": 3.957822465270689e-05, + "loss": 0.2853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28396639227867126, + "step": 1140, + "valid_targets_mean": 2740.9, + "valid_targets_min": 1028 + }, + { + "epoch": 1.1192570869990224, + "grad_norm": 0.7265698253166278, + "learning_rate": 3.9568207098047144e-05, + "loss": 0.2835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3029360771179199, + "step": 1145, + "valid_targets_mean": 2664.1, + "valid_targets_min": 1073 + }, + { + "epoch": 1.1241446725317692, + "grad_norm": 0.6515253109002876, + "learning_rate": 3.9558073270217784e-05, + "loss": 0.2764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2731107771396637, + "step": 1150, + "valid_targets_mean": 2830.1, + "valid_targets_min": 815 + }, + { + "epoch": 1.129032258064516, + "grad_norm": 0.7342300835153387, + "learning_rate": 3.954782322943343e-05, + "loss": 0.2647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26124700903892517, + "step": 1155, + "valid_targets_mean": 2346.8, + "valid_targets_min": 728 + }, + { + "epoch": 1.1339198435972628, + "grad_norm": 0.7264943185964359, + "learning_rate": 3.953745703659924e-05, + "loss": 0.2822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29263490438461304, + "step": 1160, + "valid_targets_mean": 2200.6, + "valid_targets_min": 912 + }, + { + "epoch": 1.1388074291300097, + "grad_norm": 0.7531165993253246, + "learning_rate": 3.952697475331054e-05, + "loss": 0.2727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2604370415210724, + "step": 1165, + "valid_targets_mean": 1966.1, + "valid_targets_min": 810 + }, + { + "epoch": 1.1436950146627567, + "grad_norm": 0.5807064795538424, + "learning_rate": 3.951637644185245e-05, + "loss": 0.2779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2689421474933624, + "step": 1170, + "valid_targets_mean": 3077.1, + "valid_targets_min": 663 + }, + { + "epoch": 1.1485826001955035, + "grad_norm": 0.7320623987196705, + "learning_rate": 3.9505662165199546e-05, + "loss": 0.2666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2726061940193176, + "step": 1175, + "valid_targets_mean": 2399.1, + "valid_targets_min": 953 + }, + { + "epoch": 1.1534701857282503, + "grad_norm": 0.718063434455999, + "learning_rate": 3.949483198701544e-05, + "loss": 0.2817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2901926040649414, + "step": 1180, + "valid_targets_mean": 1989.6, + "valid_targets_min": 467 + }, + { + "epoch": 1.1583577712609971, + "grad_norm": 0.7091718018989998, + "learning_rate": 3.948388597165244e-05, + "loss": 0.2802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2814217805862427, + "step": 1185, + "valid_targets_mean": 2682.6, + "valid_targets_min": 977 + }, + { + "epoch": 1.163245356793744, + "grad_norm": 0.7198188497547705, + "learning_rate": 3.947282418415114e-05, + "loss": 0.2678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26243048906326294, + "step": 1190, + "valid_targets_mean": 2464.7, + "valid_targets_min": 1207 + }, + { + "epoch": 1.1681329423264908, + "grad_norm": 0.8032209187501614, + "learning_rate": 3.946164669024005e-05, + "loss": 0.2762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2906745374202728, + "step": 1195, + "valid_targets_mean": 2152.6, + "valid_targets_min": 1274 + }, + { + "epoch": 1.1730205278592376, + "grad_norm": 0.69719975762219, + "learning_rate": 3.945035355633519e-05, + "loss": 0.2792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29841500520706177, + "step": 1200, + "valid_targets_mean": 2345.9, + "valid_targets_min": 886 + }, + { + "epoch": 1.1779081133919844, + "grad_norm": 0.6974558842300727, + "learning_rate": 3.943894484953974e-05, + "loss": 0.2781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29451218247413635, + "step": 1205, + "valid_targets_mean": 2422.9, + "valid_targets_min": 512 + }, + { + "epoch": 1.1827956989247312, + "grad_norm": 0.7729057032255819, + "learning_rate": 3.9427420637643566e-05, + "loss": 0.292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30238640308380127, + "step": 1210, + "valid_targets_mean": 2235.3, + "valid_targets_min": 1257 + }, + { + "epoch": 1.187683284457478, + "grad_norm": 0.7935199270678439, + "learning_rate": 3.941578098912288e-05, + "loss": 0.2917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25925132632255554, + "step": 1215, + "valid_targets_mean": 1819.3, + "valid_targets_min": 868 + }, + { + "epoch": 1.1925708699902249, + "grad_norm": 0.7190594576509517, + "learning_rate": 3.940402597313982e-05, + "loss": 0.2801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30240678787231445, + "step": 1220, + "valid_targets_mean": 2756.8, + "valid_targets_min": 513 + }, + { + "epoch": 1.1974584555229717, + "grad_norm": 0.6468704089059654, + "learning_rate": 3.939215565954201e-05, + "loss": 0.28, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27174267172813416, + "step": 1225, + "valid_targets_mean": 2537.5, + "valid_targets_min": 842 + }, + { + "epoch": 1.2023460410557185, + "grad_norm": 0.6638979450727887, + "learning_rate": 3.938017011886217e-05, + "loss": 0.2752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25439706444740295, + "step": 1230, + "valid_targets_mean": 2504.9, + "valid_targets_min": 857 + }, + { + "epoch": 1.2072336265884653, + "grad_norm": 0.7494764475924144, + "learning_rate": 3.936806942231772e-05, + "loss": 0.283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2995947003364563, + "step": 1235, + "valid_targets_mean": 2547.0, + "valid_targets_min": 667 + }, + { + "epoch": 1.2121212121212122, + "grad_norm": 0.7304687661971925, + "learning_rate": 3.9355853641810286e-05, + "loss": 0.2838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.266486793756485, + "step": 1240, + "valid_targets_mean": 2214.6, + "valid_targets_min": 309 + }, + { + "epoch": 1.217008797653959, + "grad_norm": 0.7022740270462892, + "learning_rate": 3.9343522849925355e-05, + "loss": 0.2734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25830769538879395, + "step": 1245, + "valid_targets_mean": 2205.6, + "valid_targets_min": 728 + }, + { + "epoch": 1.2218963831867058, + "grad_norm": 0.6749560565367534, + "learning_rate": 3.933107711993177e-05, + "loss": 0.2915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3063586354255676, + "step": 1250, + "valid_targets_mean": 2899.1, + "valid_targets_min": 1467 + }, + { + "epoch": 1.2267839687194526, + "grad_norm": 0.640174555103181, + "learning_rate": 3.931851652578137e-05, + "loss": 0.2974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26264217495918274, + "step": 1255, + "valid_targets_mean": 2715.5, + "valid_targets_min": 765 + }, + { + "epoch": 1.2316715542521994, + "grad_norm": 0.7175756349653029, + "learning_rate": 3.930584114210847e-05, + "loss": 0.2743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2843489348888397, + "step": 1260, + "valid_targets_mean": 2394.0, + "valid_targets_min": 1130 + }, + { + "epoch": 1.2365591397849462, + "grad_norm": 0.6471066400554928, + "learning_rate": 3.929305104422949e-05, + "loss": 0.2619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25573495030403137, + "step": 1265, + "valid_targets_mean": 2673.3, + "valid_targets_min": 859 + }, + { + "epoch": 1.241446725317693, + "grad_norm": 0.7057472774221171, + "learning_rate": 3.928014630814245e-05, + "loss": 0.2963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29683718085289, + "step": 1270, + "valid_targets_mean": 2584.6, + "valid_targets_min": 704 + }, + { + "epoch": 1.2463343108504399, + "grad_norm": 0.7027021456995992, + "learning_rate": 3.926712701052656e-05, + "loss": 0.2778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23105719685554504, + "step": 1275, + "valid_targets_mean": 2057.8, + "valid_targets_min": 548 + }, + { + "epoch": 1.2512218963831867, + "grad_norm": 0.8306679586454391, + "learning_rate": 3.925399322874175e-05, + "loss": 0.2842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2782430350780487, + "step": 1280, + "valid_targets_mean": 1920.4, + "valid_targets_min": 782 + }, + { + "epoch": 1.2561094819159335, + "grad_norm": 0.6271862299342814, + "learning_rate": 3.9240745040828196e-05, + "loss": 0.265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2829798460006714, + "step": 1285, + "valid_targets_mean": 3158.1, + "valid_targets_min": 807 + }, + { + "epoch": 1.2609970674486803, + "grad_norm": 0.7221564579455654, + "learning_rate": 3.922738252550587e-05, + "loss": 0.2973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2566811740398407, + "step": 1290, + "valid_targets_mean": 2186.9, + "valid_targets_min": 659 + }, + { + "epoch": 1.2658846529814272, + "grad_norm": 0.7065825530165445, + "learning_rate": 3.921390576217408e-05, + "loss": 0.2722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27165934443473816, + "step": 1295, + "valid_targets_mean": 2100.2, + "valid_targets_min": 698 + }, + { + "epoch": 1.270772238514174, + "grad_norm": 0.9761318767014419, + "learning_rate": 3.9200314830910986e-05, + "loss": 0.2806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2744409441947937, + "step": 1300, + "valid_targets_mean": 1900.2, + "valid_targets_min": 668 + }, + { + "epoch": 1.2756598240469208, + "grad_norm": 0.8388053078423351, + "learning_rate": 3.918660981247312e-05, + "loss": 0.2735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2818601727485657, + "step": 1305, + "valid_targets_mean": 1785.4, + "valid_targets_min": 769 + }, + { + "epoch": 1.2805474095796676, + "grad_norm": 0.613252199578922, + "learning_rate": 3.9172790788294924e-05, + "loss": 0.2964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25405043363571167, + "step": 1310, + "valid_targets_mean": 2779.4, + "valid_targets_min": 875 + }, + { + "epoch": 1.2854349951124144, + "grad_norm": 0.702566015145273, + "learning_rate": 3.9158857840488253e-05, + "loss": 0.2858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2808951437473297, + "step": 1315, + "valid_targets_mean": 2675.1, + "valid_targets_min": 770 + }, + { + "epoch": 1.2903225806451613, + "grad_norm": 0.7219389062248595, + "learning_rate": 3.914481105184188e-05, + "loss": 0.2969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27677303552627563, + "step": 1320, + "valid_targets_mean": 2169.5, + "valid_targets_min": 789 + }, + { + "epoch": 1.295210166177908, + "grad_norm": 0.6230877367684657, + "learning_rate": 3.913065050582103e-05, + "loss": 0.2952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2999587059020996, + "step": 1325, + "valid_targets_mean": 3419.2, + "valid_targets_min": 722 + }, + { + "epoch": 1.300097751710655, + "grad_norm": 0.6833998876959554, + "learning_rate": 3.911637628656685e-05, + "loss": 0.2757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.252625048160553, + "step": 1330, + "valid_targets_mean": 2527.4, + "valid_targets_min": 849 + }, + { + "epoch": 1.3049853372434017, + "grad_norm": 0.7469165884381328, + "learning_rate": 3.910198847889595e-05, + "loss": 0.2834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2838708162307739, + "step": 1335, + "valid_targets_mean": 2485.5, + "valid_targets_min": 622 + }, + { + "epoch": 1.3098729227761485, + "grad_norm": 0.6617163563012934, + "learning_rate": 3.908748716829984e-05, + "loss": 0.277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29781484603881836, + "step": 1340, + "valid_targets_mean": 3058.6, + "valid_targets_min": 961 + }, + { + "epoch": 1.3147605083088953, + "grad_norm": 0.6917934222962383, + "learning_rate": 3.90728724409445e-05, + "loss": 0.2703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23285894095897675, + "step": 1345, + "valid_targets_mean": 2165.1, + "valid_targets_min": 689 + }, + { + "epoch": 1.3196480938416422, + "grad_norm": 0.8112229286729639, + "learning_rate": 3.90581443836698e-05, + "loss": 0.285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2841215133666992, + "step": 1350, + "valid_targets_mean": 2388.8, + "valid_targets_min": 972 + }, + { + "epoch": 1.324535679374389, + "grad_norm": 0.7050055286781818, + "learning_rate": 3.9043303083989e-05, + "loss": 0.2708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2793377935886383, + "step": 1355, + "valid_targets_mean": 2531.8, + "valid_targets_min": 839 + }, + { + "epoch": 1.3294232649071358, + "grad_norm": 0.7692436746915501, + "learning_rate": 3.9028348630088285e-05, + "loss": 0.2801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29042643308639526, + "step": 1360, + "valid_targets_mean": 2408.2, + "valid_targets_min": 972 + }, + { + "epoch": 1.3343108504398826, + "grad_norm": 0.7453876083199784, + "learning_rate": 3.9013281110826146e-05, + "loss": 0.2899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25248265266418457, + "step": 1365, + "valid_targets_mean": 2040.0, + "valid_targets_min": 996 + }, + { + "epoch": 1.3391984359726294, + "grad_norm": 0.6300162995824621, + "learning_rate": 3.899810061573293e-05, + "loss": 0.2904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27708175778388977, + "step": 1370, + "valid_targets_mean": 2868.4, + "valid_targets_min": 1109 + }, + { + "epoch": 1.3440860215053765, + "grad_norm": 0.6688370034426764, + "learning_rate": 3.898280723501027e-05, + "loss": 0.2776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2765743136405945, + "step": 1375, + "valid_targets_mean": 2953.6, + "valid_targets_min": 960 + }, + { + "epoch": 1.3489736070381233, + "grad_norm": 0.6214024065518173, + "learning_rate": 3.896740105953056e-05, + "loss": 0.2946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2931886911392212, + "step": 1380, + "valid_targets_mean": 3082.7, + "valid_targets_min": 1298 + }, + { + "epoch": 1.3538611925708701, + "grad_norm": 0.6750814309638632, + "learning_rate": 3.895188218083641e-05, + "loss": 0.2783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23927871882915497, + "step": 1385, + "valid_targets_mean": 2085.7, + "valid_targets_min": 605 + }, + { + "epoch": 1.358748778103617, + "grad_norm": 1.0887637501386116, + "learning_rate": 3.893625069114013e-05, + "loss": 0.2768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28028884530067444, + "step": 1390, + "valid_targets_mean": 2176.3, + "valid_targets_min": 999 + }, + { + "epoch": 1.3636363636363638, + "grad_norm": 0.773730842008043, + "learning_rate": 3.892050668332312e-05, + "loss": 0.2833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27753153443336487, + "step": 1395, + "valid_targets_mean": 2301.8, + "valid_targets_min": 816 + }, + { + "epoch": 1.3685239491691106, + "grad_norm": 0.7189179648090815, + "learning_rate": 3.890465025093538e-05, + "loss": 0.2808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30436086654663086, + "step": 1400, + "valid_targets_mean": 2609.4, + "valid_targets_min": 869 + }, + { + "epoch": 1.3734115347018574, + "grad_norm": 0.6426850788526689, + "learning_rate": 3.8888681488194934e-05, + "loss": 0.2796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3102635145187378, + "step": 1405, + "valid_targets_mean": 3093.6, + "valid_targets_min": 1107 + }, + { + "epoch": 1.3782991202346042, + "grad_norm": 0.734389179384258, + "learning_rate": 3.887260048998726e-05, + "loss": 0.28, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2662855386734009, + "step": 1410, + "valid_targets_mean": 2011.6, + "valid_targets_min": 684 + }, + { + "epoch": 1.383186705767351, + "grad_norm": 0.6225396878904635, + "learning_rate": 3.885640735186472e-05, + "loss": 0.2618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2845187783241272, + "step": 1415, + "valid_targets_mean": 3417.9, + "valid_targets_min": 1093 + }, + { + "epoch": 1.3880742913000979, + "grad_norm": 0.789834712593994, + "learning_rate": 3.884010217004601e-05, + "loss": 0.2823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29228654503822327, + "step": 1420, + "valid_targets_mean": 2669.8, + "valid_targets_min": 1027 + }, + { + "epoch": 1.3929618768328447, + "grad_norm": 0.7464915480075731, + "learning_rate": 3.882368504141562e-05, + "loss": 0.2937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28257280588150024, + "step": 1425, + "valid_targets_mean": 2574.6, + "valid_targets_min": 1474 + }, + { + "epoch": 1.3978494623655915, + "grad_norm": 0.7059861573051313, + "learning_rate": 3.8807156063523164e-05, + "loss": 0.2657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2998770475387573, + "step": 1430, + "valid_targets_mean": 2475.3, + "valid_targets_min": 570 + }, + { + "epoch": 1.4027370478983383, + "grad_norm": 0.6645964659054957, + "learning_rate": 3.879051533458289e-05, + "loss": 0.2653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2769964933395386, + "step": 1435, + "valid_targets_mean": 2476.9, + "valid_targets_min": 657 + }, + { + "epoch": 1.4076246334310851, + "grad_norm": 0.7558136295073548, + "learning_rate": 3.8773762953473066e-05, + "loss": 0.2831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2779275178909302, + "step": 1440, + "valid_targets_mean": 1848.9, + "valid_targets_min": 762 + }, + { + "epoch": 1.412512218963832, + "grad_norm": 0.6943029966676353, + "learning_rate": 3.8756899019735385e-05, + "loss": 0.2853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2840909957885742, + "step": 1445, + "valid_targets_mean": 2471.9, + "valid_targets_min": 772 + }, + { + "epoch": 1.4173998044965788, + "grad_norm": 0.6342596835576019, + "learning_rate": 3.873992363357437e-05, + "loss": 0.2618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2653612196445465, + "step": 1450, + "valid_targets_mean": 2900.2, + "valid_targets_min": 1140 + }, + { + "epoch": 1.4222873900293256, + "grad_norm": 0.579209777162739, + "learning_rate": 3.8722836895856814e-05, + "loss": 0.2614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24776488542556763, + "step": 1455, + "valid_targets_mean": 2874.0, + "valid_targets_min": 632 + }, + { + "epoch": 1.4271749755620724, + "grad_norm": 0.6564233693786774, + "learning_rate": 3.870563890811111e-05, + "loss": 0.2847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3028268814086914, + "step": 1460, + "valid_targets_mean": 3045.8, + "valid_targets_min": 891 + }, + { + "epoch": 1.4320625610948192, + "grad_norm": 0.6629636773046359, + "learning_rate": 3.868832977252675e-05, + "loss": 0.2769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27365946769714355, + "step": 1465, + "valid_targets_mean": 2735.8, + "valid_targets_min": 799 + }, + { + "epoch": 1.436950146627566, + "grad_norm": 0.6785573010408985, + "learning_rate": 3.86709095919536e-05, + "loss": 0.275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2917429804801941, + "step": 1470, + "valid_targets_mean": 2449.4, + "valid_targets_min": 756 + }, + { + "epoch": 1.4418377321603129, + "grad_norm": 0.6465288698245589, + "learning_rate": 3.86533784699014e-05, + "loss": 0.2799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25123557448387146, + "step": 1475, + "valid_targets_mean": 2389.1, + "valid_targets_min": 696 + }, + { + "epoch": 1.4467253176930597, + "grad_norm": 0.7015331893695984, + "learning_rate": 3.8635736510539065e-05, + "loss": 0.2744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2667387127876282, + "step": 1480, + "valid_targets_mean": 2461.5, + "valid_targets_min": 717 + }, + { + "epoch": 1.4516129032258065, + "grad_norm": 0.7629144665484076, + "learning_rate": 3.861798381869412e-05, + "loss": 0.2821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29424476623535156, + "step": 1485, + "valid_targets_mean": 2159.1, + "valid_targets_min": 763 + }, + { + "epoch": 1.4565004887585533, + "grad_norm": 0.752596217358227, + "learning_rate": 3.860012049985204e-05, + "loss": 0.2733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25024765729904175, + "step": 1490, + "valid_targets_mean": 1827.2, + "valid_targets_min": 888 + }, + { + "epoch": 1.4613880742913001, + "grad_norm": 0.700994546408206, + "learning_rate": 3.858214666015565e-05, + "loss": 0.2822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29054421186447144, + "step": 1495, + "valid_targets_mean": 2365.5, + "valid_targets_min": 695 + }, + { + "epoch": 1.466275659824047, + "grad_norm": 0.7116021660360209, + "learning_rate": 3.856406240640448e-05, + "loss": 0.2755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2625944912433624, + "step": 1500, + "valid_targets_mean": 2130.0, + "valid_targets_min": 684 + }, + { + "epoch": 1.4711632453567938, + "grad_norm": 0.7684258462239429, + "learning_rate": 3.8545867846054126e-05, + "loss": 0.2626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2951001524925232, + "step": 1505, + "valid_targets_mean": 2052.4, + "valid_targets_min": 1064 + }, + { + "epoch": 1.4760508308895406, + "grad_norm": 0.6507083544953332, + "learning_rate": 3.8527563087215634e-05, + "loss": 0.2742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2922488749027252, + "step": 1510, + "valid_targets_mean": 2769.5, + "valid_targets_min": 843 + }, + { + "epoch": 1.4809384164222874, + "grad_norm": 0.7751727870506968, + "learning_rate": 3.8509148238654824e-05, + "loss": 0.2716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24178072810173035, + "step": 1515, + "valid_targets_mean": 1849.9, + "valid_targets_min": 780 + }, + { + "epoch": 1.4858260019550342, + "grad_norm": 0.5854852127584044, + "learning_rate": 3.84906234097917e-05, + "loss": 0.2619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27021411061286926, + "step": 1520, + "valid_targets_mean": 3099.3, + "valid_targets_min": 1029 + }, + { + "epoch": 1.490713587487781, + "grad_norm": 0.6358169347672091, + "learning_rate": 3.8471988710699715e-05, + "loss": 0.2513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26537126302719116, + "step": 1525, + "valid_targets_mean": 2957.4, + "valid_targets_min": 802 + }, + { + "epoch": 1.4956011730205279, + "grad_norm": 0.6204632468552115, + "learning_rate": 3.8453244252105194e-05, + "loss": 0.2797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2769279181957245, + "step": 1530, + "valid_targets_mean": 3130.9, + "valid_targets_min": 737 + }, + { + "epoch": 1.5004887585532747, + "grad_norm": 0.6411624274507086, + "learning_rate": 3.843439014538664e-05, + "loss": 0.2647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26065292954444885, + "step": 1535, + "valid_targets_mean": 2861.7, + "valid_targets_min": 1023 + }, + { + "epoch": 1.5053763440860215, + "grad_norm": 0.7050285259422795, + "learning_rate": 3.841542650257408e-05, + "loss": 0.2647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24625875055789948, + "step": 1540, + "valid_targets_mean": 2171.6, + "valid_targets_min": 1054 + }, + { + "epoch": 1.5102639296187683, + "grad_norm": 0.7521045411153422, + "learning_rate": 3.83963534363484e-05, + "loss": 0.2824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3168919086456299, + "step": 1545, + "valid_targets_mean": 2370.8, + "valid_targets_min": 842 + }, + { + "epoch": 1.5151515151515151, + "grad_norm": 0.7217702499392185, + "learning_rate": 3.8377171060040666e-05, + "loss": 0.2673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27114635705947876, + "step": 1550, + "valid_targets_mean": 2131.9, + "valid_targets_min": 783 + }, + { + "epoch": 1.520039100684262, + "grad_norm": 0.7137936454079413, + "learning_rate": 3.835787948763147e-05, + "loss": 0.2767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25275540351867676, + "step": 1555, + "valid_targets_mean": 2113.8, + "valid_targets_min": 748 + }, + { + "epoch": 1.5249266862170088, + "grad_norm": 0.9545717948909013, + "learning_rate": 3.833847883375022e-05, + "loss": 0.2934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2564704716205597, + "step": 1560, + "valid_targets_mean": 2187.5, + "valid_targets_min": 739 + }, + { + "epoch": 1.5298142717497556, + "grad_norm": 0.7470839745811981, + "learning_rate": 3.831896921367451e-05, + "loss": 0.2821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2791537940502167, + "step": 1565, + "valid_targets_mean": 1975.1, + "valid_targets_min": 686 + }, + { + "epoch": 1.5347018572825024, + "grad_norm": 0.7647266073732814, + "learning_rate": 3.829935074332938e-05, + "loss": 0.2718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24994492530822754, + "step": 1570, + "valid_targets_mean": 2233.8, + "valid_targets_min": 837 + }, + { + "epoch": 1.5395894428152492, + "grad_norm": 0.7045036581823351, + "learning_rate": 3.827962353928667e-05, + "loss": 0.2649, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26855552196502686, + "step": 1575, + "valid_targets_mean": 2261.9, + "valid_targets_min": 763 + }, + { + "epoch": 1.544477028347996, + "grad_norm": 0.7143743531724757, + "learning_rate": 3.825978771876428e-05, + "loss": 0.2873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3018741011619568, + "step": 1580, + "valid_targets_mean": 2336.8, + "valid_targets_min": 663 + }, + { + "epoch": 1.5493646138807429, + "grad_norm": 0.6475010659041159, + "learning_rate": 3.823984339962554e-05, + "loss": 0.276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29907166957855225, + "step": 1585, + "valid_targets_mean": 2830.2, + "valid_targets_min": 732 + }, + { + "epoch": 1.5542521994134897, + "grad_norm": 0.6898759516125273, + "learning_rate": 3.821979070037846e-05, + "loss": 0.2791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2999107241630554, + "step": 1590, + "valid_targets_mean": 2604.1, + "valid_targets_min": 631 + }, + { + "epoch": 1.5591397849462365, + "grad_norm": 0.6572278395987295, + "learning_rate": 3.8199629740175017e-05, + "loss": 0.2681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2730094790458679, + "step": 1595, + "valid_targets_mean": 2485.9, + "valid_targets_min": 804 + }, + { + "epoch": 1.5640273704789833, + "grad_norm": 0.7008441805523876, + "learning_rate": 3.8179360638810503e-05, + "loss": 0.2747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2944326102733612, + "step": 1600, + "valid_targets_mean": 2188.3, + "valid_targets_min": 669 + }, + { + "epoch": 1.5689149560117301, + "grad_norm": 0.5571584829883734, + "learning_rate": 3.815898351672275e-05, + "loss": 0.2667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22917650640010834, + "step": 1605, + "valid_targets_mean": 3037.3, + "valid_targets_min": 765 + }, + { + "epoch": 1.573802541544477, + "grad_norm": 0.7410724781542547, + "learning_rate": 3.813849849499146e-05, + "loss": 0.2616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24475133419036865, + "step": 1610, + "valid_targets_mean": 1836.8, + "valid_targets_min": 741 + }, + { + "epoch": 1.5786901270772238, + "grad_norm": 0.7170995982719762, + "learning_rate": 3.811790569533746e-05, + "loss": 0.2656, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2909032106399536, + "step": 1615, + "valid_targets_mean": 2271.4, + "valid_targets_min": 689 + }, + { + "epoch": 1.5835777126099706, + "grad_norm": 0.7023440780016095, + "learning_rate": 3.8097205240121994e-05, + "loss": 0.2932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3244866132736206, + "step": 1620, + "valid_targets_mean": 2630.7, + "valid_targets_min": 752 + }, + { + "epoch": 1.5884652981427174, + "grad_norm": 0.7145272097523425, + "learning_rate": 3.807639725234598e-05, + "loss": 0.2802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28274455666542053, + "step": 1625, + "valid_targets_mean": 2337.7, + "valid_targets_min": 1045 + }, + { + "epoch": 1.5933528836754642, + "grad_norm": 0.6483060882499405, + "learning_rate": 3.8055481855649295e-05, + "loss": 0.2719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2898852825164795, + "step": 1630, + "valid_targets_mean": 2868.5, + "valid_targets_min": 1718 + }, + { + "epoch": 1.598240469208211, + "grad_norm": 0.6154123007304795, + "learning_rate": 3.8034459174310034e-05, + "loss": 0.2698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24204128980636597, + "step": 1635, + "valid_targets_mean": 2880.8, + "valid_targets_min": 791 + }, + { + "epoch": 1.6031280547409579, + "grad_norm": 0.7242965707991273, + "learning_rate": 3.8013329333243765e-05, + "loss": 0.2746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3221800923347473, + "step": 1640, + "valid_targets_mean": 2574.4, + "valid_targets_min": 836 + }, + { + "epoch": 1.6080156402737047, + "grad_norm": 0.8256849875812563, + "learning_rate": 3.7992092458002813e-05, + "loss": 0.2669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2468244880437851, + "step": 1645, + "valid_targets_mean": 2232.7, + "valid_targets_min": 659 + }, + { + "epoch": 1.6129032258064515, + "grad_norm": 0.8957983328042407, + "learning_rate": 3.797074867477545e-05, + "loss": 0.2941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2847079038619995, + "step": 1650, + "valid_targets_mean": 1988.9, + "valid_targets_min": 1107 + }, + { + "epoch": 1.6177908113391983, + "grad_norm": 0.7320454198900345, + "learning_rate": 3.794929811038525e-05, + "loss": 0.2759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25967937707901, + "step": 1655, + "valid_targets_mean": 2044.0, + "valid_targets_min": 732 + }, + { + "epoch": 1.6226783968719452, + "grad_norm": 0.7944386390316857, + "learning_rate": 3.7927740892290225e-05, + "loss": 0.3063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29300451278686523, + "step": 1660, + "valid_targets_mean": 1941.6, + "valid_targets_min": 584 + }, + { + "epoch": 1.627565982404692, + "grad_norm": 0.7033843489197059, + "learning_rate": 3.790607714858213e-05, + "loss": 0.2718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27778416872024536, + "step": 1665, + "valid_targets_mean": 2372.6, + "valid_targets_min": 625 + }, + { + "epoch": 1.6324535679374388, + "grad_norm": 0.713583282306689, + "learning_rate": 3.78843070079857e-05, + "loss": 0.2706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2961004972457886, + "step": 1670, + "valid_targets_mean": 2199.5, + "valid_targets_min": 867 + }, + { + "epoch": 1.6373411534701856, + "grad_norm": 0.6811625348127058, + "learning_rate": 3.786243059985787e-05, + "loss": 0.2844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24486398696899414, + "step": 1675, + "valid_targets_mean": 2072.9, + "valid_targets_min": 739 + }, + { + "epoch": 1.6422287390029324, + "grad_norm": 0.6214013174128027, + "learning_rate": 3.7840448054187004e-05, + "loss": 0.2698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2856760323047638, + "step": 1680, + "valid_targets_mean": 2886.9, + "valid_targets_min": 868 + }, + { + "epoch": 1.6471163245356792, + "grad_norm": 0.6763894621818294, + "learning_rate": 3.781835950159214e-05, + "loss": 0.2845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2664475739002228, + "step": 1685, + "valid_targets_mean": 2352.6, + "valid_targets_min": 647 + }, + { + "epoch": 1.652003910068426, + "grad_norm": 0.7476808183243178, + "learning_rate": 3.7796165073322194e-05, + "loss": 0.275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24394232034683228, + "step": 1690, + "valid_targets_mean": 1998.0, + "valid_targets_min": 857 + }, + { + "epoch": 1.6568914956011729, + "grad_norm": 0.6654065406280266, + "learning_rate": 3.777386490125518e-05, + "loss": 0.2763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2673056721687317, + "step": 1695, + "valid_targets_mean": 2774.7, + "valid_targets_min": 941 + }, + { + "epoch": 1.6617790811339197, + "grad_norm": 1.0447139479563154, + "learning_rate": 3.7751459117897466e-05, + "loss": 0.2648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29265016317367554, + "step": 1700, + "valid_targets_mean": 2134.6, + "valid_targets_min": 856 + }, + { + "epoch": 1.6666666666666665, + "grad_norm": 0.6756696155203493, + "learning_rate": 3.772894785638292e-05, + "loss": 0.2836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29080730676651, + "step": 1705, + "valid_targets_mean": 2770.5, + "valid_targets_min": 1168 + }, + { + "epoch": 1.6715542521994133, + "grad_norm": 0.6063804591260533, + "learning_rate": 3.770633125047219e-05, + "loss": 0.2756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29159295558929443, + "step": 1710, + "valid_targets_mean": 3060.3, + "valid_targets_min": 1237 + }, + { + "epoch": 1.6764418377321602, + "grad_norm": 0.7088620832664758, + "learning_rate": 3.768360943455185e-05, + "loss": 0.2524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26422443985939026, + "step": 1715, + "valid_targets_mean": 2225.6, + "valid_targets_min": 727 + }, + { + "epoch": 1.681329423264907, + "grad_norm": 0.8121289862781522, + "learning_rate": 3.7660782543633634e-05, + "loss": 0.2691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.271615207195282, + "step": 1720, + "valid_targets_mean": 2280.4, + "valid_targets_min": 639 + }, + { + "epoch": 1.6862170087976538, + "grad_norm": 0.6721466708752135, + "learning_rate": 3.763785071335362e-05, + "loss": 0.2759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2757419943809509, + "step": 1725, + "valid_targets_mean": 2551.2, + "valid_targets_min": 1072 + }, + { + "epoch": 1.6911045943304008, + "grad_norm": 0.7511340865178172, + "learning_rate": 3.761481407997144e-05, + "loss": 0.2787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3051798939704895, + "step": 1730, + "valid_targets_mean": 2232.9, + "valid_targets_min": 884 + }, + { + "epoch": 1.6959921798631477, + "grad_norm": 0.7709735641548016, + "learning_rate": 3.759167278036945e-05, + "loss": 0.2687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28466472029685974, + "step": 1735, + "valid_targets_mean": 2116.8, + "valid_targets_min": 746 + }, + { + "epoch": 1.7008797653958945, + "grad_norm": 0.6541200445160277, + "learning_rate": 3.7568426952051936e-05, + "loss": 0.2688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24498885869979858, + "step": 1740, + "valid_targets_mean": 2278.2, + "valid_targets_min": 615 + }, + { + "epoch": 1.7057673509286413, + "grad_norm": 0.8540951314910903, + "learning_rate": 3.754507673314429e-05, + "loss": 0.2738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2802680730819702, + "step": 1745, + "valid_targets_mean": 2895.8, + "valid_targets_min": 834 + }, + { + "epoch": 1.710654936461388, + "grad_norm": 0.6993627250812203, + "learning_rate": 3.752162226239216e-05, + "loss": 0.2835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2634274363517761, + "step": 1750, + "valid_targets_mean": 2296.5, + "valid_targets_min": 641 + }, + { + "epoch": 1.715542521994135, + "grad_norm": 0.7122786881252015, + "learning_rate": 3.749806367916069e-05, + "loss": 0.2908, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29387813806533813, + "step": 1755, + "valid_targets_mean": 2420.0, + "valid_targets_min": 719 + }, + { + "epoch": 1.7204301075268817, + "grad_norm": 0.6950186797619158, + "learning_rate": 3.747440112343363e-05, + "loss": 0.272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.284950315952301, + "step": 1760, + "valid_targets_mean": 2666.9, + "valid_targets_min": 694 + }, + { + "epoch": 1.7253176930596286, + "grad_norm": 0.6540823928470759, + "learning_rate": 3.745063473581254e-05, + "loss": 0.2801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26305273175239563, + "step": 1765, + "valid_targets_mean": 2482.1, + "valid_targets_min": 522 + }, + { + "epoch": 1.7302052785923754, + "grad_norm": 0.6533343607539351, + "learning_rate": 3.7426764657515926e-05, + "loss": 0.288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32014939188957214, + "step": 1770, + "valid_targets_mean": 2911.1, + "valid_targets_min": 1390 + }, + { + "epoch": 1.7350928641251222, + "grad_norm": 0.681180973864994, + "learning_rate": 3.7402791030378446e-05, + "loss": 0.2609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26099470257759094, + "step": 1775, + "valid_targets_mean": 2321.4, + "valid_targets_min": 514 + }, + { + "epoch": 1.739980449657869, + "grad_norm": 0.8019831916932382, + "learning_rate": 3.737871399685001e-05, + "loss": 0.2651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27410048246383667, + "step": 1780, + "valid_targets_mean": 1795.9, + "valid_targets_min": 760 + }, + { + "epoch": 1.7448680351906158, + "grad_norm": 0.6599982007782187, + "learning_rate": 3.735453369999499e-05, + "loss": 0.2677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2654041647911072, + "step": 1785, + "valid_targets_mean": 2488.9, + "valid_targets_min": 780 + }, + { + "epoch": 1.7497556207233627, + "grad_norm": 0.6985853935310431, + "learning_rate": 3.7330250283491316e-05, + "loss": 0.2678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2721864581108093, + "step": 1790, + "valid_targets_mean": 2234.3, + "valid_targets_min": 892 + }, + { + "epoch": 1.7546432062561095, + "grad_norm": 0.7815807491097794, + "learning_rate": 3.730586389162968e-05, + "loss": 0.2723, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2551916241645813, + "step": 1795, + "valid_targets_mean": 2119.4, + "valid_targets_min": 741 + }, + { + "epoch": 1.7595307917888563, + "grad_norm": 0.6983319157355613, + "learning_rate": 3.728137466931262e-05, + "loss": 0.2646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26075512170791626, + "step": 1800, + "valid_targets_mean": 1934.5, + "valid_targets_min": 790 + }, + { + "epoch": 1.7644183773216031, + "grad_norm": 0.6167013764000374, + "learning_rate": 3.72567827620537e-05, + "loss": 0.2814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2760656476020813, + "step": 1805, + "valid_targets_mean": 3027.8, + "valid_targets_min": 1151 + }, + { + "epoch": 1.76930596285435, + "grad_norm": 0.6509730292742568, + "learning_rate": 3.723208831597663e-05, + "loss": 0.2662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.275715172290802, + "step": 1810, + "valid_targets_mean": 2570.6, + "valid_targets_min": 881 + }, + { + "epoch": 1.7741935483870968, + "grad_norm": 0.6845280114796127, + "learning_rate": 3.720729147781438e-05, + "loss": 0.2678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28714922070503235, + "step": 1815, + "valid_targets_mean": 2509.2, + "valid_targets_min": 842 + }, + { + "epoch": 1.7790811339198436, + "grad_norm": 0.692569209787489, + "learning_rate": 3.718239239490838e-05, + "loss": 0.2746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29387450218200684, + "step": 1820, + "valid_targets_mean": 2449.2, + "valid_targets_min": 648 + }, + { + "epoch": 1.7839687194525904, + "grad_norm": 0.6702820731559377, + "learning_rate": 3.715739121520754e-05, + "loss": 0.2842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28223657608032227, + "step": 1825, + "valid_targets_mean": 2757.1, + "valid_targets_min": 758 + }, + { + "epoch": 1.7888563049853372, + "grad_norm": 0.7241156064899832, + "learning_rate": 3.7132288087267444e-05, + "loss": 0.2754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2525007128715515, + "step": 1830, + "valid_targets_mean": 1929.8, + "valid_targets_min": 824 + }, + { + "epoch": 1.793743890518084, + "grad_norm": 0.6959493990441359, + "learning_rate": 3.710708316024945e-05, + "loss": 0.2691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29006555676460266, + "step": 1835, + "valid_targets_mean": 2522.8, + "valid_targets_min": 887 + }, + { + "epoch": 1.7986314760508308, + "grad_norm": 1.0690505659751939, + "learning_rate": 3.708177658391979e-05, + "loss": 0.283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26409053802490234, + "step": 1840, + "valid_targets_mean": 2085.5, + "valid_targets_min": 898 + }, + { + "epoch": 1.8035190615835777, + "grad_norm": 0.6416781900383685, + "learning_rate": 3.70563685086487e-05, + "loss": 0.2782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2942282557487488, + "step": 1845, + "valid_targets_mean": 2790.2, + "valid_targets_min": 518 + }, + { + "epoch": 1.8084066471163245, + "grad_norm": 0.6292212211636715, + "learning_rate": 3.70308590854095e-05, + "loss": 0.2636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2690730392932892, + "step": 1850, + "valid_targets_mean": 2605.3, + "valid_targets_min": 627 + }, + { + "epoch": 1.8132942326490715, + "grad_norm": 0.7563881415657474, + "learning_rate": 3.7005248465777753e-05, + "loss": 0.2757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26602309942245483, + "step": 1855, + "valid_targets_mean": 1989.0, + "valid_targets_min": 880 + }, + { + "epoch": 1.8181818181818183, + "grad_norm": 0.6636462614146774, + "learning_rate": 3.697953680193028e-05, + "loss": 0.2535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2749035358428955, + "step": 1860, + "valid_targets_mean": 2244.6, + "valid_targets_min": 1017 + }, + { + "epoch": 1.8230694037145652, + "grad_norm": 1.7607977847503449, + "learning_rate": 3.695372424664433e-05, + "loss": 0.2829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29794639348983765, + "step": 1865, + "valid_targets_mean": 2888.2, + "valid_targets_min": 843 + }, + { + "epoch": 1.827956989247312, + "grad_norm": 0.6664158354412304, + "learning_rate": 3.692781095329662e-05, + "loss": 0.2836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31028884649276733, + "step": 1870, + "valid_targets_mean": 2667.8, + "valid_targets_min": 774 + }, + { + "epoch": 1.8328445747800588, + "grad_norm": 0.6484592517147627, + "learning_rate": 3.690179707586247e-05, + "loss": 0.2571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2612004578113556, + "step": 1875, + "valid_targets_mean": 2415.0, + "valid_targets_min": 817 + }, + { + "epoch": 1.8377321603128056, + "grad_norm": 0.6881093759344453, + "learning_rate": 3.6875682768914847e-05, + "loss": 0.2744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2743026912212372, + "step": 1880, + "valid_targets_mean": 3187.6, + "valid_targets_min": 1503 + }, + { + "epoch": 1.8426197458455524, + "grad_norm": 0.7092531777121036, + "learning_rate": 3.684946818762349e-05, + "loss": 0.2766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24219094216823578, + "step": 1885, + "valid_targets_mean": 1849.3, + "valid_targets_min": 538 + }, + { + "epoch": 1.8475073313782993, + "grad_norm": 0.7102034739129834, + "learning_rate": 3.682315348775392e-05, + "loss": 0.2668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2451900839805603, + "step": 1890, + "valid_targets_mean": 2226.6, + "valid_targets_min": 742 + }, + { + "epoch": 1.852394916911046, + "grad_norm": 0.6976435032370457, + "learning_rate": 3.6796738825666606e-05, + "loss": 0.2853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34835338592529297, + "step": 1895, + "valid_targets_mean": 2488.8, + "valid_targets_min": 821 + }, + { + "epoch": 1.857282502443793, + "grad_norm": 0.8345683696586109, + "learning_rate": 3.6770224358315945e-05, + "loss": 0.2892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3025885820388794, + "step": 1900, + "valid_targets_mean": 2416.9, + "valid_targets_min": 544 + }, + { + "epoch": 1.8621700879765397, + "grad_norm": 0.6734416069076511, + "learning_rate": 3.674361024324939e-05, + "loss": 0.2677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2562083601951599, + "step": 1905, + "valid_targets_mean": 2252.0, + "valid_targets_min": 1040 + }, + { + "epoch": 1.8670576735092865, + "grad_norm": 0.7258976397773823, + "learning_rate": 3.67168966386065e-05, + "loss": 0.2825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29352593421936035, + "step": 1910, + "valid_targets_mean": 2220.9, + "valid_targets_min": 770 + }, + { + "epoch": 1.8719452590420333, + "grad_norm": 0.8267516235866793, + "learning_rate": 3.669008370311798e-05, + "loss": 0.2589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26839086413383484, + "step": 1915, + "valid_targets_mean": 2146.9, + "valid_targets_min": 906 + }, + { + "epoch": 1.8768328445747802, + "grad_norm": 0.9090846152816757, + "learning_rate": 3.666317159610478e-05, + "loss": 0.2632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2196088433265686, + "step": 1920, + "valid_targets_mean": 1943.8, + "valid_targets_min": 627 + }, + { + "epoch": 1.881720430107527, + "grad_norm": 0.6328104571501864, + "learning_rate": 3.6636160477477084e-05, + "loss": 0.2789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26850637793540955, + "step": 1925, + "valid_targets_mean": 2655.1, + "valid_targets_min": 1260 + }, + { + "epoch": 1.8866080156402738, + "grad_norm": 0.734256137030361, + "learning_rate": 3.6609050507733434e-05, + "loss": 0.2497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2635309100151062, + "step": 1930, + "valid_targets_mean": 2196.3, + "valid_targets_min": 865 + }, + { + "epoch": 1.8914956011730206, + "grad_norm": 0.6466263966198653, + "learning_rate": 3.6581841847959704e-05, + "loss": 0.2661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23648956418037415, + "step": 1935, + "valid_targets_mean": 2312.8, + "valid_targets_min": 571 + }, + { + "epoch": 1.8963831867057674, + "grad_norm": 0.6012794126830221, + "learning_rate": 3.6554534659828214e-05, + "loss": 0.2612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2779013514518738, + "step": 1940, + "valid_targets_mean": 2929.6, + "valid_targets_min": 682 + }, + { + "epoch": 1.9012707722385143, + "grad_norm": 0.7694228039973774, + "learning_rate": 3.652712910559671e-05, + "loss": 0.2566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2838958203792572, + "step": 1945, + "valid_targets_mean": 2658.0, + "valid_targets_min": 1149 + }, + { + "epoch": 1.906158357771261, + "grad_norm": 0.6786164917053215, + "learning_rate": 3.649962534810743e-05, + "loss": 0.2663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2658728361129761, + "step": 1950, + "valid_targets_mean": 2249.3, + "valid_targets_min": 714 + }, + { + "epoch": 1.911045943304008, + "grad_norm": 0.6234672172314177, + "learning_rate": 3.647202355078614e-05, + "loss": 0.2912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2740452289581299, + "step": 1955, + "valid_targets_mean": 2578.2, + "valid_targets_min": 720 + }, + { + "epoch": 1.9159335288367547, + "grad_norm": 0.6515984804744183, + "learning_rate": 3.644432387764113e-05, + "loss": 0.2743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28338634967803955, + "step": 1960, + "valid_targets_mean": 2537.2, + "valid_targets_min": 781 + }, + { + "epoch": 1.9208211143695015, + "grad_norm": 0.6925882408989057, + "learning_rate": 3.641652649326228e-05, + "loss": 0.2836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28017061948776245, + "step": 1965, + "valid_targets_mean": 2322.9, + "valid_targets_min": 790 + }, + { + "epoch": 1.9257086999022484, + "grad_norm": 0.721478062577497, + "learning_rate": 3.638863156282007e-05, + "loss": 0.2554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2587317228317261, + "step": 1970, + "valid_targets_mean": 1998.3, + "valid_targets_min": 818 + }, + { + "epoch": 1.9305962854349952, + "grad_norm": 0.670482355423357, + "learning_rate": 3.6360639252064576e-05, + "loss": 0.2743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28506842255592346, + "step": 1975, + "valid_targets_mean": 2706.6, + "valid_targets_min": 804 + }, + { + "epoch": 1.935483870967742, + "grad_norm": 0.7503529232237587, + "learning_rate": 3.633254972732452e-05, + "loss": 0.2646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2652524411678314, + "step": 1980, + "valid_targets_mean": 2164.6, + "valid_targets_min": 1087 + }, + { + "epoch": 1.9403714565004888, + "grad_norm": 0.6787558928888673, + "learning_rate": 3.630436315550624e-05, + "loss": 0.2794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24034081399440765, + "step": 1985, + "valid_targets_mean": 2107.4, + "valid_targets_min": 771 + }, + { + "epoch": 1.9452590420332356, + "grad_norm": 0.7914047009085244, + "learning_rate": 3.6276079704092755e-05, + "loss": 0.2695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2761124074459076, + "step": 1990, + "valid_targets_mean": 2557.2, + "valid_targets_min": 836 + }, + { + "epoch": 1.9501466275659824, + "grad_norm": 0.6934610377236305, + "learning_rate": 3.6247699541142705e-05, + "loss": 0.2829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2904479503631592, + "step": 1995, + "valid_targets_mean": 2472.5, + "valid_targets_min": 620 + }, + { + "epoch": 1.9550342130987293, + "grad_norm": 0.7028198944986546, + "learning_rate": 3.62192228352894e-05, + "loss": 0.2753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30306535959243774, + "step": 2000, + "valid_targets_mean": 2393.2, + "valid_targets_min": 820 + }, + { + "epoch": 1.959921798631476, + "grad_norm": 0.6656505685896379, + "learning_rate": 3.6190649755739803e-05, + "loss": 0.2696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27111366391181946, + "step": 2005, + "valid_targets_mean": 2641.5, + "valid_targets_min": 655 + }, + { + "epoch": 1.964809384164223, + "grad_norm": 0.6548401411079302, + "learning_rate": 3.616198047227353e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26000118255615234, + "step": 2010, + "valid_targets_mean": 2489.2, + "valid_targets_min": 1225 + }, + { + "epoch": 1.9696969696969697, + "grad_norm": 0.7088864087296969, + "learning_rate": 3.613321515524181e-05, + "loss": 0.2591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2879897654056549, + "step": 2015, + "valid_targets_mean": 2197.8, + "valid_targets_min": 916 + }, + { + "epoch": 1.9745845552297165, + "grad_norm": 0.7810945047814258, + "learning_rate": 3.610435397556653e-05, + "loss": 0.2669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25487759709358215, + "step": 2020, + "valid_targets_mean": 1809.1, + "valid_targets_min": 1181 + }, + { + "epoch": 1.9794721407624634, + "grad_norm": 0.6618807198434398, + "learning_rate": 3.607539710473916e-05, + "loss": 0.28, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2809375524520874, + "step": 2025, + "valid_targets_mean": 2645.9, + "valid_targets_min": 714 + }, + { + "epoch": 1.9843597262952102, + "grad_norm": 0.6889464103259075, + "learning_rate": 3.6046344714819786e-05, + "loss": 0.2818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27869024872779846, + "step": 2030, + "valid_targets_mean": 2894.4, + "valid_targets_min": 612 + }, + { + "epoch": 1.989247311827957, + "grad_norm": 0.7597436857623857, + "learning_rate": 3.601719697843604e-05, + "loss": 0.2596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24141916632652283, + "step": 2035, + "valid_targets_mean": 1800.8, + "valid_targets_min": 648 + }, + { + "epoch": 1.9941348973607038, + "grad_norm": 0.6357417452126822, + "learning_rate": 3.5987954068782114e-05, + "loss": 0.2673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28626328706741333, + "step": 2040, + "valid_targets_mean": 3067.6, + "valid_targets_min": 824 + }, + { + "epoch": 1.9990224828934506, + "grad_norm": 0.6268686861206388, + "learning_rate": 3.595861615961771e-05, + "loss": 0.2615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25856465101242065, + "step": 2045, + "valid_targets_mean": 2928.6, + "valid_targets_min": 1156 + }, + { + "epoch": 2.0039100684261975, + "grad_norm": 0.5633104622128313, + "learning_rate": 3.5929183425267e-05, + "loss": 0.2354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2239726036787033, + "step": 2050, + "valid_targets_mean": 2999.4, + "valid_targets_min": 631 + }, + { + "epoch": 2.0087976539589443, + "grad_norm": 0.7030082756993576, + "learning_rate": 3.589965604061762e-05, + "loss": 0.241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26195091009140015, + "step": 2055, + "valid_targets_mean": 2908.1, + "valid_targets_min": 793 + }, + { + "epoch": 2.013685239491691, + "grad_norm": 0.7032018693280796, + "learning_rate": 3.5870034181119605e-05, + "loss": 0.2413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23086042702198029, + "step": 2060, + "valid_targets_mean": 2292.1, + "valid_targets_min": 748 + }, + { + "epoch": 2.018572825024438, + "grad_norm": 0.8062986669540131, + "learning_rate": 3.5840318022784345e-05, + "loss": 0.2279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24330705404281616, + "step": 2065, + "valid_targets_mean": 2035.7, + "valid_targets_min": 804 + }, + { + "epoch": 2.0234604105571847, + "grad_norm": 0.7456760467454644, + "learning_rate": 3.5810507742183555e-05, + "loss": 0.2319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24097514152526855, + "step": 2070, + "valid_targets_mean": 2251.5, + "valid_targets_min": 656 + }, + { + "epoch": 2.0283479960899315, + "grad_norm": 0.7974068850160627, + "learning_rate": 3.5780603516448235e-05, + "loss": 0.2364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25509095191955566, + "step": 2075, + "valid_targets_mean": 2308.2, + "valid_targets_min": 842 + }, + { + "epoch": 2.0332355816226784, + "grad_norm": 0.7027608391523197, + "learning_rate": 3.575060552326758e-05, + "loss": 0.2605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26107388734817505, + "step": 2080, + "valid_targets_mean": 2641.1, + "valid_targets_min": 702 + }, + { + "epoch": 2.038123167155425, + "grad_norm": 0.6771926382902913, + "learning_rate": 3.572051394088795e-05, + "loss": 0.2389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.254364550113678, + "step": 2085, + "valid_targets_mean": 3093.8, + "valid_targets_min": 1095 + }, + { + "epoch": 2.043010752688172, + "grad_norm": 0.7920555868975577, + "learning_rate": 3.569032894811182e-05, + "loss": 0.2325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25785136222839355, + "step": 2090, + "valid_targets_mean": 2053.3, + "valid_targets_min": 803 + }, + { + "epoch": 2.047898338220919, + "grad_norm": 0.6836448197530809, + "learning_rate": 3.566005072429668e-05, + "loss": 0.2324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21475102007389069, + "step": 2095, + "valid_targets_mean": 2353.8, + "valid_targets_min": 976 + }, + { + "epoch": 2.0527859237536656, + "grad_norm": 0.712116095288442, + "learning_rate": 3.562967944935402e-05, + "loss": 0.2388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23862826824188232, + "step": 2100, + "valid_targets_mean": 2316.5, + "valid_targets_min": 768 + }, + { + "epoch": 2.0576735092864125, + "grad_norm": 0.7340036820210002, + "learning_rate": 3.559921530374821e-05, + "loss": 0.236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2310110479593277, + "step": 2105, + "valid_targets_mean": 2313.2, + "valid_targets_min": 763 + }, + { + "epoch": 2.0625610948191593, + "grad_norm": 0.7950851146965314, + "learning_rate": 3.556865846849547e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24690885841846466, + "step": 2110, + "valid_targets_mean": 1923.8, + "valid_targets_min": 684 + }, + { + "epoch": 2.067448680351906, + "grad_norm": 0.9485462964048043, + "learning_rate": 3.5538009125162774e-05, + "loss": 0.2383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2360670119524002, + "step": 2115, + "valid_targets_mean": 2597.2, + "valid_targets_min": 1328 + }, + { + "epoch": 2.072336265884653, + "grad_norm": 0.6042905131795798, + "learning_rate": 3.550726745586677e-05, + "loss": 0.2464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24113643169403076, + "step": 2120, + "valid_targets_mean": 3059.4, + "valid_targets_min": 884 + }, + { + "epoch": 2.0772238514173997, + "grad_norm": 0.6939861915048151, + "learning_rate": 3.5476433643272695e-05, + "loss": 0.2375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23294597864151, + "step": 2125, + "valid_targets_mean": 2400.4, + "valid_targets_min": 863 + }, + { + "epoch": 2.0821114369501466, + "grad_norm": 0.7426227574519302, + "learning_rate": 3.5445507870593315e-05, + "loss": 0.2289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2383473813533783, + "step": 2130, + "valid_targets_mean": 2200.8, + "valid_targets_min": 670 + }, + { + "epoch": 2.0869990224828934, + "grad_norm": 0.6868218935296267, + "learning_rate": 3.54144903215878e-05, + "loss": 0.2284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2236156016588211, + "step": 2135, + "valid_targets_mean": 2512.4, + "valid_targets_min": 522 + }, + { + "epoch": 2.09188660801564, + "grad_norm": 0.734071113886848, + "learning_rate": 3.538338118056065e-05, + "loss": 0.2584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26370787620544434, + "step": 2140, + "valid_targets_mean": 2402.6, + "valid_targets_min": 731 + }, + { + "epoch": 2.096774193548387, + "grad_norm": 0.7089461463159339, + "learning_rate": 3.5352180632360614e-05, + "loss": 0.2381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.212519571185112, + "step": 2145, + "valid_targets_mean": 2266.1, + "valid_targets_min": 504 + }, + { + "epoch": 2.101661779081134, + "grad_norm": 0.791032807011372, + "learning_rate": 3.532088886237956e-05, + "loss": 0.2364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2348712533712387, + "step": 2150, + "valid_targets_mean": 2305.6, + "valid_targets_min": 746 + }, + { + "epoch": 2.1065493646138806, + "grad_norm": 0.6579340652769179, + "learning_rate": 3.5289506056551414e-05, + "loss": 0.2562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27510562539100647, + "step": 2155, + "valid_targets_mean": 2961.9, + "valid_targets_min": 680 + }, + { + "epoch": 2.1114369501466275, + "grad_norm": 0.7467367590945104, + "learning_rate": 3.5258032401351005e-05, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22451433539390564, + "step": 2160, + "valid_targets_mean": 2105.4, + "valid_targets_min": 765 + }, + { + "epoch": 2.1163245356793743, + "grad_norm": 0.9754841406665122, + "learning_rate": 3.522646808379299e-05, + "loss": 0.2308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2368217408657074, + "step": 2165, + "valid_targets_mean": 2108.7, + "valid_targets_min": 618 + }, + { + "epoch": 2.121212121212121, + "grad_norm": 0.679630597898811, + "learning_rate": 3.519481329143076e-05, + "loss": 0.2381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23734697699546814, + "step": 2170, + "valid_targets_mean": 2571.5, + "valid_targets_min": 1089 + }, + { + "epoch": 2.126099706744868, + "grad_norm": 0.658153760353656, + "learning_rate": 3.5163068212355274e-05, + "loss": 0.2397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.256761372089386, + "step": 2175, + "valid_targets_mean": 2925.9, + "valid_targets_min": 803 + }, + { + "epoch": 2.1309872922776147, + "grad_norm": 0.7475086838003469, + "learning_rate": 3.513123303519397e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22203174233436584, + "step": 2180, + "valid_targets_mean": 1844.1, + "valid_targets_min": 692 + }, + { + "epoch": 2.1358748778103616, + "grad_norm": 0.6041239999536051, + "learning_rate": 3.509930794910967e-05, + "loss": 0.2367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22519180178642273, + "step": 2185, + "valid_targets_mean": 3204.3, + "valid_targets_min": 946 + }, + { + "epoch": 2.1407624633431084, + "grad_norm": 0.7509582157797575, + "learning_rate": 3.506729314379941e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25604188442230225, + "step": 2190, + "valid_targets_mean": 2394.9, + "valid_targets_min": 677 + }, + { + "epoch": 2.145650048875855, + "grad_norm": 0.6750872876590103, + "learning_rate": 3.5035188809493335e-05, + "loss": 0.2311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22361913323402405, + "step": 2195, + "valid_targets_mean": 2691.1, + "valid_targets_min": 791 + }, + { + "epoch": 2.150537634408602, + "grad_norm": 0.6447663759799113, + "learning_rate": 3.500299513695356e-05, + "loss": 0.2355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22863362729549408, + "step": 2200, + "valid_targets_mean": 2783.2, + "valid_targets_min": 853 + }, + { + "epoch": 2.155425219941349, + "grad_norm": 0.7166550535292393, + "learning_rate": 3.497071231747306e-05, + "loss": 0.2224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25535690784454346, + "step": 2205, + "valid_targets_mean": 2722.8, + "valid_targets_min": 989 + }, + { + "epoch": 2.1603128054740957, + "grad_norm": 0.8862852553501419, + "learning_rate": 3.493834054287449e-05, + "loss": 0.2387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24125589430332184, + "step": 2210, + "valid_targets_mean": 2089.8, + "valid_targets_min": 787 + }, + { + "epoch": 2.1652003910068425, + "grad_norm": 0.7210539973246137, + "learning_rate": 3.490588000550911e-05, + "loss": 0.2345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26080626249313354, + "step": 2215, + "valid_targets_mean": 2307.4, + "valid_targets_min": 995 + }, + { + "epoch": 2.1700879765395893, + "grad_norm": 0.6876101039598905, + "learning_rate": 3.4873330898255556e-05, + "loss": 0.2342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22508588433265686, + "step": 2220, + "valid_targets_mean": 2578.5, + "valid_targets_min": 551 + }, + { + "epoch": 2.174975562072336, + "grad_norm": 0.679616490089902, + "learning_rate": 3.4840693414518776e-05, + "loss": 0.2306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22549769282341003, + "step": 2225, + "valid_targets_mean": 2478.4, + "valid_targets_min": 1293 + }, + { + "epoch": 2.179863147605083, + "grad_norm": 0.6885669807827226, + "learning_rate": 3.4807967748228844e-05, + "loss": 0.2404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21673941612243652, + "step": 2230, + "valid_targets_mean": 2484.1, + "valid_targets_min": 1140 + }, + { + "epoch": 2.1847507331378297, + "grad_norm": 0.6971954496384682, + "learning_rate": 3.477515409383979e-05, + "loss": 0.2313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22069767117500305, + "step": 2235, + "valid_targets_mean": 2294.4, + "valid_targets_min": 914 + }, + { + "epoch": 2.1896383186705766, + "grad_norm": 0.7601614406567037, + "learning_rate": 3.474225264632847e-05, + "loss": 0.2424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20157967507839203, + "step": 2240, + "valid_targets_mean": 1873.1, + "valid_targets_min": 615 + }, + { + "epoch": 2.1945259042033234, + "grad_norm": 0.6527427792093988, + "learning_rate": 3.470926360119343e-05, + "loss": 0.2372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22357343137264252, + "step": 2245, + "valid_targets_mean": 2597.2, + "valid_targets_min": 720 + }, + { + "epoch": 2.19941348973607, + "grad_norm": 1.2913533556146168, + "learning_rate": 3.467618715445367e-05, + "loss": 0.2438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2719864845275879, + "step": 2250, + "valid_targets_mean": 2356.8, + "valid_targets_min": 1038 + }, + { + "epoch": 2.204301075268817, + "grad_norm": 0.6982533310800236, + "learning_rate": 3.464302350264756e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23422983288764954, + "step": 2255, + "valid_targets_mean": 2767.9, + "valid_targets_min": 753 + }, + { + "epoch": 2.209188660801564, + "grad_norm": 0.7221984393019797, + "learning_rate": 3.460977284283162e-05, + "loss": 0.2455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.266290545463562, + "step": 2260, + "valid_targets_mean": 2269.7, + "valid_targets_min": 822 + }, + { + "epoch": 2.2140762463343107, + "grad_norm": 0.7114567359248741, + "learning_rate": 3.457643537257938e-05, + "loss": 0.2313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19123372435569763, + "step": 2265, + "valid_targets_mean": 1872.2, + "valid_targets_min": 570 + }, + { + "epoch": 2.2189638318670575, + "grad_norm": 0.6540588887195975, + "learning_rate": 3.454301128998018e-05, + "loss": 0.2204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2638104259967804, + "step": 2270, + "valid_targets_mean": 3022.8, + "valid_targets_min": 987 + }, + { + "epoch": 2.2238514173998043, + "grad_norm": 0.6898490839976488, + "learning_rate": 3.4509500793638015e-05, + "loss": 0.239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2437189519405365, + "step": 2275, + "valid_targets_mean": 2453.7, + "valid_targets_min": 656 + }, + { + "epoch": 2.228739002932551, + "grad_norm": 0.7463219837340317, + "learning_rate": 3.447590408267034e-05, + "loss": 0.2394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21820449829101562, + "step": 2280, + "valid_targets_mean": 2064.9, + "valid_targets_min": 748 + }, + { + "epoch": 2.233626588465298, + "grad_norm": 0.7810585743007555, + "learning_rate": 3.444222135670688e-05, + "loss": 0.2394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2506566643714905, + "step": 2285, + "valid_targets_mean": 2264.9, + "valid_targets_min": 820 + }, + { + "epoch": 2.2385141739980448, + "grad_norm": 0.7468906966516604, + "learning_rate": 3.44084528158885e-05, + "loss": 0.2261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22076962888240814, + "step": 2290, + "valid_targets_mean": 2084.4, + "valid_targets_min": 721 + }, + { + "epoch": 2.2434017595307916, + "grad_norm": 0.7194591472281033, + "learning_rate": 3.437459866086591e-05, + "loss": 0.2509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23915450274944305, + "step": 2295, + "valid_targets_mean": 2377.3, + "valid_targets_min": 650 + }, + { + "epoch": 2.2482893450635384, + "grad_norm": 0.7180690575974027, + "learning_rate": 3.4340659092798594e-05, + "loss": 0.2362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2506748139858246, + "step": 2300, + "valid_targets_mean": 2737.3, + "valid_targets_min": 567 + }, + { + "epoch": 2.253176930596285, + "grad_norm": 0.6309852432550281, + "learning_rate": 3.4306634313353504e-05, + "loss": 0.2256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2439078688621521, + "step": 2305, + "valid_targets_mean": 3100.8, + "valid_targets_min": 914 + }, + { + "epoch": 2.258064516129032, + "grad_norm": 0.7572600999737557, + "learning_rate": 3.427252452470395e-05, + "loss": 0.2469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2431081384420395, + "step": 2310, + "valid_targets_mean": 1790.9, + "valid_targets_min": 640 + }, + { + "epoch": 2.262952101661779, + "grad_norm": 0.683772308399432, + "learning_rate": 3.423832992952836e-05, + "loss": 0.2364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24556618928909302, + "step": 2315, + "valid_targets_mean": 2683.7, + "valid_targets_min": 801 + }, + { + "epoch": 2.2678396871945257, + "grad_norm": 0.8397527577042548, + "learning_rate": 3.420405073100905e-05, + "loss": 0.2359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.254273384809494, + "step": 2320, + "valid_targets_mean": 2155.0, + "valid_targets_min": 867 + }, + { + "epoch": 2.2727272727272725, + "grad_norm": 0.851065317435724, + "learning_rate": 3.416968713283105e-05, + "loss": 0.2542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2502952814102173, + "step": 2325, + "valid_targets_mean": 2122.0, + "valid_targets_min": 684 + }, + { + "epoch": 2.2776148582600193, + "grad_norm": 0.6985298739440852, + "learning_rate": 3.413523933918093e-05, + "loss": 0.2351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22682811319828033, + "step": 2330, + "valid_targets_mean": 2159.1, + "valid_targets_min": 795 + }, + { + "epoch": 2.2825024437927666, + "grad_norm": 0.8111962594657017, + "learning_rate": 3.4100707554745495e-05, + "loss": 0.2504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24102097749710083, + "step": 2335, + "valid_targets_mean": 2234.1, + "valid_targets_min": 705 + }, + { + "epoch": 2.2873900293255134, + "grad_norm": 0.5993930502127762, + "learning_rate": 3.406609198471064e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18438704311847687, + "step": 2340, + "valid_targets_mean": 2438.3, + "valid_targets_min": 1216 + }, + { + "epoch": 2.29227761485826, + "grad_norm": 0.7265646225215977, + "learning_rate": 3.403139283476011e-05, + "loss": 0.2342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2278384566307068, + "step": 2345, + "valid_targets_mean": 2198.9, + "valid_targets_min": 1056 + }, + { + "epoch": 2.297165200391007, + "grad_norm": 0.6075073209468058, + "learning_rate": 3.3996610311074266e-05, + "loss": 0.2433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2122841477394104, + "step": 2350, + "valid_targets_mean": 2752.8, + "valid_targets_min": 762 + }, + { + "epoch": 2.302052785923754, + "grad_norm": 0.6706963737298545, + "learning_rate": 3.396174462032888e-05, + "loss": 0.2452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.241616889834404, + "step": 2355, + "valid_targets_mean": 2314.3, + "valid_targets_min": 637 + }, + { + "epoch": 2.3069403714565007, + "grad_norm": 1.1053006675938248, + "learning_rate": 3.392679596969391e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.232365682721138, + "step": 2360, + "valid_targets_mean": 2931.2, + "valid_targets_min": 1234 + }, + { + "epoch": 2.3118279569892475, + "grad_norm": 0.68661451987901, + "learning_rate": 3.389176456683222e-05, + "loss": 0.2406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2415533810853958, + "step": 2365, + "valid_targets_mean": 2462.8, + "valid_targets_min": 748 + }, + { + "epoch": 2.3167155425219943, + "grad_norm": 0.7999782604502889, + "learning_rate": 3.385665061989842e-05, + "loss": 0.2181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21598216891288757, + "step": 2370, + "valid_targets_mean": 1724.7, + "valid_targets_min": 662 + }, + { + "epoch": 2.321603128054741, + "grad_norm": 0.7182473780722957, + "learning_rate": 3.382145433753758e-05, + "loss": 0.2471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27340924739837646, + "step": 2375, + "valid_targets_mean": 2582.8, + "valid_targets_min": 774 + }, + { + "epoch": 2.326490713587488, + "grad_norm": 0.7676223492166587, + "learning_rate": 3.3786175928883994e-05, + "loss": 0.2299, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21359305083751678, + "step": 2380, + "valid_targets_mean": 1934.9, + "valid_targets_min": 646 + }, + { + "epoch": 2.3313782991202348, + "grad_norm": 0.701569446945032, + "learning_rate": 3.375081560355996e-05, + "loss": 0.2327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24448227882385254, + "step": 2385, + "valid_targets_mean": 2355.2, + "valid_targets_min": 700 + }, + { + "epoch": 2.3362658846529816, + "grad_norm": 0.668238001789143, + "learning_rate": 3.3715373571674495e-05, + "loss": 0.2522, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24243560433387756, + "step": 2390, + "valid_targets_mean": 2480.6, + "valid_targets_min": 1062 + }, + { + "epoch": 2.3411534701857284, + "grad_norm": 0.7311150815773109, + "learning_rate": 3.367985004382216e-05, + "loss": 0.2438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22406163811683655, + "step": 2395, + "valid_targets_mean": 1759.3, + "valid_targets_min": 1058 + }, + { + "epoch": 2.346041055718475, + "grad_norm": 0.6489801175202864, + "learning_rate": 3.36442452310817e-05, + "loss": 0.2393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2665456533432007, + "step": 2400, + "valid_targets_mean": 2741.1, + "valid_targets_min": 772 + }, + { + "epoch": 2.350928641251222, + "grad_norm": 0.6383014149144618, + "learning_rate": 3.36085593450149e-05, + "loss": 0.2314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23392713069915771, + "step": 2405, + "valid_targets_mean": 2465.8, + "valid_targets_min": 697 + }, + { + "epoch": 2.355816226783969, + "grad_norm": 0.6749587378319567, + "learning_rate": 3.3572792597665244e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2573961019515991, + "step": 2410, + "valid_targets_mean": 2782.9, + "valid_targets_min": 888 + }, + { + "epoch": 2.3607038123167157, + "grad_norm": 0.9440638611034704, + "learning_rate": 3.353694520155672e-05, + "loss": 0.2449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24345912039279938, + "step": 2415, + "valid_targets_mean": 2128.4, + "valid_targets_min": 1012 + }, + { + "epoch": 2.3655913978494625, + "grad_norm": 0.7148978990158092, + "learning_rate": 3.350101736969249e-05, + "loss": 0.2563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2735242247581482, + "step": 2420, + "valid_targets_mean": 2309.9, + "valid_targets_min": 771 + }, + { + "epoch": 2.3704789833822093, + "grad_norm": 0.8017459949388398, + "learning_rate": 3.346500931555369e-05, + "loss": 0.23, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2165670096874237, + "step": 2425, + "valid_targets_mean": 2131.1, + "valid_targets_min": 708 + }, + { + "epoch": 2.375366568914956, + "grad_norm": 0.6221765808255351, + "learning_rate": 3.3428921253098126e-05, + "loss": 0.2281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26252031326293945, + "step": 2430, + "valid_targets_mean": 3323.1, + "valid_targets_min": 1030 + }, + { + "epoch": 2.380254154447703, + "grad_norm": 0.6740618705615318, + "learning_rate": 3.3392753396759e-05, + "loss": 0.2462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23797520995140076, + "step": 2435, + "valid_targets_mean": 2207.9, + "valid_targets_min": 734 + }, + { + "epoch": 2.3851417399804498, + "grad_norm": 0.7273981487589926, + "learning_rate": 3.3356505961443644e-05, + "loss": 0.2382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23226913809776306, + "step": 2440, + "valid_targets_mean": 2119.9, + "valid_targets_min": 685 + }, + { + "epoch": 2.3900293255131966, + "grad_norm": 0.7495183220790104, + "learning_rate": 3.332017916253227e-05, + "loss": 0.2438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2352924793958664, + "step": 2445, + "valid_targets_mean": 2707.6, + "valid_targets_min": 714 + }, + { + "epoch": 2.3949169110459434, + "grad_norm": 0.7842291061024107, + "learning_rate": 3.328377321587662e-05, + "loss": 0.2427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22039464116096497, + "step": 2450, + "valid_targets_mean": 1844.9, + "valid_targets_min": 544 + }, + { + "epoch": 2.39980449657869, + "grad_norm": 0.6173507633048779, + "learning_rate": 3.324728833779876e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23128989338874817, + "step": 2455, + "valid_targets_mean": 2820.6, + "valid_targets_min": 1219 + }, + { + "epoch": 2.404692082111437, + "grad_norm": 0.7009098945991525, + "learning_rate": 3.3210724745089757e-05, + "loss": 0.218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20386609435081482, + "step": 2460, + "valid_targets_mean": 2477.1, + "valid_targets_min": 622 + }, + { + "epoch": 2.409579667644184, + "grad_norm": 0.7651345366780957, + "learning_rate": 3.31740826550084e-05, + "loss": 0.2345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2441290318965912, + "step": 2465, + "valid_targets_mean": 2235.6, + "valid_targets_min": 783 + }, + { + "epoch": 2.4144672531769307, + "grad_norm": 0.7279723174495759, + "learning_rate": 3.313736228527987e-05, + "loss": 0.226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2114153504371643, + "step": 2470, + "valid_targets_mean": 2223.0, + "valid_targets_min": 761 + }, + { + "epoch": 2.4193548387096775, + "grad_norm": 0.7868641978840135, + "learning_rate": 3.310056385409453e-05, + "loss": 0.2494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26033109426498413, + "step": 2475, + "valid_targets_mean": 2057.3, + "valid_targets_min": 770 + }, + { + "epoch": 2.4242424242424243, + "grad_norm": 0.6636689132066698, + "learning_rate": 3.3063687580106544e-05, + "loss": 0.2226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2515316903591156, + "step": 2480, + "valid_targets_mean": 2834.2, + "valid_targets_min": 750 + }, + { + "epoch": 2.429130009775171, + "grad_norm": 0.7145023267491682, + "learning_rate": 3.302673368243264e-05, + "loss": 0.241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2276870310306549, + "step": 2485, + "valid_targets_mean": 2252.9, + "valid_targets_min": 668 + }, + { + "epoch": 2.434017595307918, + "grad_norm": 0.6488982935705684, + "learning_rate": 3.298970238065077e-05, + "loss": 0.243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.250893771648407, + "step": 2490, + "valid_targets_mean": 2848.4, + "valid_targets_min": 1069 + }, + { + "epoch": 2.4389051808406648, + "grad_norm": 0.6730474280370319, + "learning_rate": 3.2952593894798796e-05, + "loss": 0.236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22307783365249634, + "step": 2495, + "valid_targets_mean": 2428.5, + "valid_targets_min": 934 + }, + { + "epoch": 2.4437927663734116, + "grad_norm": 0.7947851421243562, + "learning_rate": 3.291540844537324e-05, + "loss": 0.237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2531839907169342, + "step": 2500, + "valid_targets_mean": 1988.7, + "valid_targets_min": 690 + }, + { + "epoch": 2.4486803519061584, + "grad_norm": 0.6648566470243608, + "learning_rate": 3.287814625332792e-05, + "loss": 0.2428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2527335286140442, + "step": 2505, + "valid_targets_mean": 2601.8, + "valid_targets_min": 1007 + }, + { + "epoch": 2.4535679374389052, + "grad_norm": 0.7237098939276865, + "learning_rate": 3.2840807540072644e-05, + "loss": 0.2483, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24128659069538116, + "step": 2510, + "valid_targets_mean": 2112.2, + "valid_targets_min": 1071 + }, + { + "epoch": 2.458455522971652, + "grad_norm": 0.671690132960215, + "learning_rate": 3.280339252747192e-05, + "loss": 0.2359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24696272611618042, + "step": 2515, + "valid_targets_mean": 2795.6, + "valid_targets_min": 610 + }, + { + "epoch": 2.463343108504399, + "grad_norm": 0.6700935950794794, + "learning_rate": 3.276590143784362e-05, + "loss": 0.2486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21841683983802795, + "step": 2520, + "valid_targets_mean": 2417.8, + "valid_targets_min": 676 + }, + { + "epoch": 2.4682306940371457, + "grad_norm": 0.6719669379309949, + "learning_rate": 3.2728334493957654e-05, + "loss": 0.2381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22499142587184906, + "step": 2525, + "valid_targets_mean": 2297.1, + "valid_targets_min": 699 + }, + { + "epoch": 2.4731182795698925, + "grad_norm": 0.6892705433904501, + "learning_rate": 3.269069191903466e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22186297178268433, + "step": 2530, + "valid_targets_mean": 2068.8, + "valid_targets_min": 740 + }, + { + "epoch": 2.4780058651026393, + "grad_norm": 0.7393767252579292, + "learning_rate": 3.2652973936744667e-05, + "loss": 0.2474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2544059157371521, + "step": 2535, + "valid_targets_mean": 2338.4, + "valid_targets_min": 722 + }, + { + "epoch": 2.482893450635386, + "grad_norm": 0.7561597467756237, + "learning_rate": 3.261518077120578e-05, + "loss": 0.2435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21395118534564972, + "step": 2540, + "valid_targets_mean": 1770.1, + "valid_targets_min": 589 + }, + { + "epoch": 2.487781036168133, + "grad_norm": 0.753579094986984, + "learning_rate": 3.257731264698283e-05, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21489103138446808, + "step": 2545, + "valid_targets_mean": 1672.8, + "valid_targets_min": 755 + }, + { + "epoch": 2.4926686217008798, + "grad_norm": 0.6589858003710982, + "learning_rate": 3.253936978908604e-05, + "loss": 0.2293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24498148262500763, + "step": 2550, + "valid_targets_mean": 3218.2, + "valid_targets_min": 1118 + }, + { + "epoch": 2.4975562072336266, + "grad_norm": 0.7002768152776646, + "learning_rate": 3.250135242296971e-05, + "loss": 0.2514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2629767060279846, + "step": 2555, + "valid_targets_mean": 2427.2, + "valid_targets_min": 796 + }, + { + "epoch": 2.5024437927663734, + "grad_norm": 0.786673139351379, + "learning_rate": 3.246326077453087e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22290992736816406, + "step": 2560, + "valid_targets_mean": 1889.1, + "valid_targets_min": 717 + }, + { + "epoch": 2.5073313782991202, + "grad_norm": 0.6164870341662336, + "learning_rate": 3.242509507010791e-05, + "loss": 0.2447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23814159631729126, + "step": 2565, + "valid_targets_mean": 3004.0, + "valid_targets_min": 792 + }, + { + "epoch": 2.512218963831867, + "grad_norm": 0.7245474670367238, + "learning_rate": 3.238685553647927e-05, + "loss": 0.2446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21552202105522156, + "step": 2570, + "valid_targets_mean": 1887.2, + "valid_targets_min": 753 + }, + { + "epoch": 2.517106549364614, + "grad_norm": 0.7481756056168739, + "learning_rate": 3.2348542400862075e-05, + "loss": 0.2542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2534748911857605, + "step": 2575, + "valid_targets_mean": 2268.9, + "valid_targets_min": 632 + }, + { + "epoch": 2.5219941348973607, + "grad_norm": 0.7224008501321807, + "learning_rate": 3.231015589091079e-05, + "loss": 0.2384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2531392574310303, + "step": 2580, + "valid_targets_mean": 2338.6, + "valid_targets_min": 882 + }, + { + "epoch": 2.5268817204301075, + "grad_norm": 0.7291916308037362, + "learning_rate": 3.2271696234715866e-05, + "loss": 0.2515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2746545374393463, + "step": 2585, + "valid_targets_mean": 2363.4, + "valid_targets_min": 843 + }, + { + "epoch": 2.5317693059628543, + "grad_norm": 1.0380167495946064, + "learning_rate": 3.2233163660802387e-05, + "loss": 0.2162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21444106101989746, + "step": 2590, + "valid_targets_mean": 1598.9, + "valid_targets_min": 732 + }, + { + "epoch": 2.536656891495601, + "grad_norm": 0.6544024445784333, + "learning_rate": 3.219455839812871e-05, + "loss": 0.2382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21233348548412323, + "step": 2595, + "valid_targets_mean": 2456.2, + "valid_targets_min": 1190 + }, + { + "epoch": 2.541544477028348, + "grad_norm": 0.628273229393255, + "learning_rate": 3.21558806760851e-05, + "loss": 0.2384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2087693214416504, + "step": 2600, + "valid_targets_mean": 2745.5, + "valid_targets_min": 700 + }, + { + "epoch": 2.5464320625610948, + "grad_norm": 0.6030343066322286, + "learning_rate": 3.211713072449236e-05, + "loss": 0.2335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22888022661209106, + "step": 2605, + "valid_targets_mean": 2916.0, + "valid_targets_min": 884 + }, + { + "epoch": 2.5513196480938416, + "grad_norm": 0.7152526807779622, + "learning_rate": 3.207830877360051e-05, + "loss": 0.2289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22312462329864502, + "step": 2610, + "valid_targets_mean": 2152.1, + "valid_targets_min": 924 + }, + { + "epoch": 2.5562072336265884, + "grad_norm": 0.6731642169381366, + "learning_rate": 3.203941505408736e-05, + "loss": 0.2477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21658943593502045, + "step": 2615, + "valid_targets_mean": 2331.7, + "valid_targets_min": 911 + }, + { + "epoch": 2.5610948191593352, + "grad_norm": 0.6639073922224608, + "learning_rate": 3.200044979705717e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22162654995918274, + "step": 2620, + "valid_targets_mean": 2536.9, + "valid_targets_min": 765 + }, + { + "epoch": 2.565982404692082, + "grad_norm": 0.7471285143763177, + "learning_rate": 3.1961413234039267e-05, + "loss": 0.2357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24290457367897034, + "step": 2625, + "valid_targets_mean": 2320.9, + "valid_targets_min": 970 + }, + { + "epoch": 2.570869990224829, + "grad_norm": 0.7014404171599978, + "learning_rate": 3.192230559698669e-05, + "loss": 0.2428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2300807237625122, + "step": 2630, + "valid_targets_mean": 2163.2, + "valid_targets_min": 1099 + }, + { + "epoch": 2.5757575757575757, + "grad_norm": 0.7689908471529473, + "learning_rate": 3.1883127118274795e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22766678035259247, + "step": 2635, + "valid_targets_mean": 2223.1, + "valid_targets_min": 758 + }, + { + "epoch": 2.5806451612903225, + "grad_norm": 0.7524156572486015, + "learning_rate": 3.1843878030699843e-05, + "loss": 0.242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22420425713062286, + "step": 2640, + "valid_targets_mean": 2231.3, + "valid_targets_min": 900 + }, + { + "epoch": 2.5855327468230693, + "grad_norm": 0.6206868507870091, + "learning_rate": 3.180455856747768e-05, + "loss": 0.229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23075121641159058, + "step": 2645, + "valid_targets_mean": 2600.9, + "valid_targets_min": 1010 + }, + { + "epoch": 2.590420332355816, + "grad_norm": 0.90695340921089, + "learning_rate": 3.1765168962242314e-05, + "loss": 0.2427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2378099113702774, + "step": 2650, + "valid_targets_mean": 3146.6, + "valid_targets_min": 656 + }, + { + "epoch": 2.595307917888563, + "grad_norm": 0.7522437194650009, + "learning_rate": 3.172570944904454e-05, + "loss": 0.2519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24361330270767212, + "step": 2655, + "valid_targets_mean": 1933.7, + "valid_targets_min": 607 + }, + { + "epoch": 2.60019550342131, + "grad_norm": 0.7385794100286881, + "learning_rate": 3.168618026235052e-05, + "loss": 0.2433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2449553906917572, + "step": 2660, + "valid_targets_mean": 2196.0, + "valid_targets_min": 734 + }, + { + "epoch": 2.6050830889540566, + "grad_norm": 0.6811321355025456, + "learning_rate": 3.164658163704044e-05, + "loss": 0.2357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2384980171918869, + "step": 2665, + "valid_targets_mean": 2501.2, + "valid_targets_min": 826 + }, + { + "epoch": 2.6099706744868034, + "grad_norm": 0.7081141329414362, + "learning_rate": 3.160691380840705e-05, + "loss": 0.2397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22777023911476135, + "step": 2670, + "valid_targets_mean": 2467.4, + "valid_targets_min": 747 + }, + { + "epoch": 2.6148582600195502, + "grad_norm": 0.6830393989705583, + "learning_rate": 3.156717701215433e-05, + "loss": 0.2262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21992099285125732, + "step": 2675, + "valid_targets_mean": 2317.6, + "valid_targets_min": 950 + }, + { + "epoch": 2.619745845552297, + "grad_norm": 0.8012433854096802, + "learning_rate": 3.152737148439608e-05, + "loss": 0.2469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27927660942077637, + "step": 2680, + "valid_targets_mean": 2117.2, + "valid_targets_min": 789 + }, + { + "epoch": 2.624633431085044, + "grad_norm": 0.6934903380108935, + "learning_rate": 3.1487497461654435e-05, + "loss": 0.2424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25259506702423096, + "step": 2685, + "valid_targets_mean": 2417.8, + "valid_targets_min": 791 + }, + { + "epoch": 2.6295210166177907, + "grad_norm": 0.7359216165404469, + "learning_rate": 3.144755518085859e-05, + "loss": 0.234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24347510933876038, + "step": 2690, + "valid_targets_mean": 2227.3, + "valid_targets_min": 1027 + }, + { + "epoch": 2.6344086021505375, + "grad_norm": 0.7767880726332548, + "learning_rate": 3.1407544879343284e-05, + "loss": 0.2373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24625647068023682, + "step": 2695, + "valid_targets_mean": 1754.9, + "valid_targets_min": 689 + }, + { + "epoch": 2.6392961876832843, + "grad_norm": 0.696674510993026, + "learning_rate": 3.136746679484746e-05, + "loss": 0.2366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2380523532629013, + "step": 2700, + "valid_targets_mean": 2332.2, + "valid_targets_min": 790 + }, + { + "epoch": 2.644183773216031, + "grad_norm": 0.6694545044109874, + "learning_rate": 3.1327321165512784e-05, + "loss": 0.2294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2502917945384979, + "step": 2705, + "valid_targets_mean": 2535.8, + "valid_targets_min": 1282 + }, + { + "epoch": 2.649071358748778, + "grad_norm": 0.6987439481563695, + "learning_rate": 3.128710822988231e-05, + "loss": 0.2343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21862663328647614, + "step": 2710, + "valid_targets_mean": 1921.0, + "valid_targets_min": 823 + }, + { + "epoch": 2.653958944281525, + "grad_norm": 0.6914300177187406, + "learning_rate": 3.1246828226899e-05, + "loss": 0.2345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22969259321689606, + "step": 2715, + "valid_targets_mean": 2253.5, + "valid_targets_min": 824 + }, + { + "epoch": 2.6588465298142716, + "grad_norm": 0.602200287873372, + "learning_rate": 3.120648139590435e-05, + "loss": 0.2455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24584728479385376, + "step": 2720, + "valid_targets_mean": 3155.4, + "valid_targets_min": 1278 + }, + { + "epoch": 2.6637341153470184, + "grad_norm": 0.7075447193702563, + "learning_rate": 3.1166067976636906e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2338378131389618, + "step": 2725, + "valid_targets_mean": 2341.6, + "valid_targets_min": 533 + }, + { + "epoch": 2.6686217008797652, + "grad_norm": 0.6716820653818237, + "learning_rate": 3.1125588209230914e-05, + "loss": 0.2499, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.234563410282135, + "step": 2730, + "valid_targets_mean": 2494.7, + "valid_targets_min": 756 + }, + { + "epoch": 2.673509286412512, + "grad_norm": 0.6350189009309046, + "learning_rate": 3.108504233421482e-05, + "loss": 0.2304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2456614375114441, + "step": 2735, + "valid_targets_mean": 2821.6, + "valid_targets_min": 761 + }, + { + "epoch": 2.678396871945259, + "grad_norm": 0.6980749472004388, + "learning_rate": 3.104443059250992e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24124208092689514, + "step": 2740, + "valid_targets_mean": 2521.0, + "valid_targets_min": 992 + }, + { + "epoch": 2.6832844574780057, + "grad_norm": 0.67259645483737, + "learning_rate": 3.1003753225428855e-05, + "loss": 0.2325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22315667569637299, + "step": 2745, + "valid_targets_mean": 2475.1, + "valid_targets_min": 859 + }, + { + "epoch": 2.688172043010753, + "grad_norm": 0.6681313606790765, + "learning_rate": 3.0963010474674217e-05, + "loss": 0.2389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.247651606798172, + "step": 2750, + "valid_targets_mean": 2558.2, + "valid_targets_min": 634 + }, + { + "epoch": 2.6930596285435, + "grad_norm": 0.6859361683493692, + "learning_rate": 3.0922202582337115e-05, + "loss": 0.2164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19534924626350403, + "step": 2755, + "valid_targets_mean": 2257.4, + "valid_targets_min": 876 + }, + { + "epoch": 2.6979472140762466, + "grad_norm": 0.6985818000049671, + "learning_rate": 3.08813297908957e-05, + "loss": 0.2441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25439029932022095, + "step": 2760, + "valid_targets_mean": 2687.1, + "valid_targets_min": 958 + }, + { + "epoch": 2.7028347996089934, + "grad_norm": 0.6566705673923328, + "learning_rate": 3.084039234321379e-05, + "loss": 0.2464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2557886242866516, + "step": 2765, + "valid_targets_mean": 2698.4, + "valid_targets_min": 549 + }, + { + "epoch": 2.7077223851417402, + "grad_norm": 0.6165188909289334, + "learning_rate": 3.079939048253934e-05, + "loss": 0.2493, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24600160121917725, + "step": 2770, + "valid_targets_mean": 2858.0, + "valid_targets_min": 978 + }, + { + "epoch": 2.712609970674487, + "grad_norm": 0.6790410549127608, + "learning_rate": 3.0758324452503065e-05, + "loss": 0.2311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2445409744977951, + "step": 2775, + "valid_targets_mean": 2667.5, + "valid_targets_min": 740 + }, + { + "epoch": 2.717497556207234, + "grad_norm": 0.6974981732520593, + "learning_rate": 3.071719449711699e-05, + "loss": 0.2309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23706629872322083, + "step": 2780, + "valid_targets_mean": 2513.4, + "valid_targets_min": 1093 + }, + { + "epoch": 2.7223851417399807, + "grad_norm": 0.6737555976917173, + "learning_rate": 3.0676000860772956e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2328021228313446, + "step": 2785, + "valid_targets_mean": 2313.6, + "valid_targets_min": 684 + }, + { + "epoch": 2.7272727272727275, + "grad_norm": 1.2575294841809628, + "learning_rate": 3.063474378824119e-05, + "loss": 0.2296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23096132278442383, + "step": 2790, + "valid_targets_mean": 1920.2, + "valid_targets_min": 891 + }, + { + "epoch": 2.7321603128054743, + "grad_norm": 0.6273895966528873, + "learning_rate": 3.0593423524668864e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23225541412830353, + "step": 2795, + "valid_targets_mean": 2795.9, + "valid_targets_min": 815 + }, + { + "epoch": 2.737047898338221, + "grad_norm": 0.750013291851974, + "learning_rate": 3.055204031557863e-05, + "loss": 0.2184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19314704835414886, + "step": 2800, + "valid_targets_mean": 1944.0, + "valid_targets_min": 839 + }, + { + "epoch": 2.741935483870968, + "grad_norm": 0.665083340706297, + "learning_rate": 3.0510594406867153e-05, + "loss": 0.2359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23232056200504303, + "step": 2805, + "valid_targets_mean": 2747.2, + "valid_targets_min": 893 + }, + { + "epoch": 2.746823069403715, + "grad_norm": 0.6430016293534321, + "learning_rate": 3.0469086044803663e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22754496335983276, + "step": 2810, + "valid_targets_mean": 2441.9, + "valid_targets_min": 697 + }, + { + "epoch": 2.7517106549364616, + "grad_norm": 0.6672184525189786, + "learning_rate": 3.0427515476028468e-05, + "loss": 0.2329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2377731204032898, + "step": 2815, + "valid_targets_mean": 2553.1, + "valid_targets_min": 1144 + }, + { + "epoch": 2.7565982404692084, + "grad_norm": 0.6337998207855043, + "learning_rate": 3.038588294755151e-05, + "loss": 0.223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2279859483242035, + "step": 2820, + "valid_targets_mean": 2683.7, + "valid_targets_min": 597 + }, + { + "epoch": 2.7614858260019552, + "grad_norm": 0.6415879365422061, + "learning_rate": 3.034418870675092e-05, + "loss": 0.2518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25901520252227783, + "step": 2825, + "valid_targets_mean": 3471.3, + "valid_targets_min": 771 + }, + { + "epoch": 2.766373411534702, + "grad_norm": 0.6903646046533939, + "learning_rate": 3.0302433001371475e-05, + "loss": 0.2439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2619539499282837, + "step": 2830, + "valid_targets_mean": 2333.2, + "valid_targets_min": 890 + }, + { + "epoch": 2.771260997067449, + "grad_norm": 0.6243991438233805, + "learning_rate": 3.02606160795232e-05, + "loss": 0.2373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2319139838218689, + "step": 2835, + "valid_targets_mean": 2468.5, + "valid_targets_min": 646 + }, + { + "epoch": 2.7761485826001957, + "grad_norm": 0.640200588738358, + "learning_rate": 3.021873818967986e-05, + "loss": 0.2408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20535393059253693, + "step": 2840, + "valid_targets_mean": 3473.6, + "valid_targets_min": 978 + }, + { + "epoch": 2.7810361681329425, + "grad_norm": 0.6766646508609335, + "learning_rate": 3.0176799580677477e-05, + "loss": 0.2389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23535458743572235, + "step": 2845, + "valid_targets_mean": 2393.3, + "valid_targets_min": 861 + }, + { + "epoch": 2.7859237536656893, + "grad_norm": 0.6733786925421605, + "learning_rate": 3.013480050171289e-05, + "loss": 0.2263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23346024751663208, + "step": 2850, + "valid_targets_mean": 2390.7, + "valid_targets_min": 728 + }, + { + "epoch": 2.790811339198436, + "grad_norm": 0.724460745453377, + "learning_rate": 3.009274120234221e-05, + "loss": 0.2305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2303357571363449, + "step": 2855, + "valid_targets_mean": 2362.2, + "valid_targets_min": 786 + }, + { + "epoch": 2.795698924731183, + "grad_norm": 0.6645986334487373, + "learning_rate": 3.0050621932479402e-05, + "loss": 0.2294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22619573771953583, + "step": 2860, + "valid_targets_mean": 2401.0, + "valid_targets_min": 832 + }, + { + "epoch": 2.80058651026393, + "grad_norm": 0.6600838806245943, + "learning_rate": 3.0008442942394763e-05, + "loss": 0.2426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26058539748191833, + "step": 2865, + "valid_targets_mean": 2661.4, + "valid_targets_min": 1131 + }, + { + "epoch": 2.8054740957966766, + "grad_norm": 0.64943249977253, + "learning_rate": 2.9966204482713447e-05, + "loss": 0.2419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22463732957839966, + "step": 2870, + "valid_targets_mean": 2495.8, + "valid_targets_min": 680 + }, + { + "epoch": 2.8103616813294234, + "grad_norm": 0.6786818865745439, + "learning_rate": 2.9923906804413966e-05, + "loss": 0.2326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2178959995508194, + "step": 2875, + "valid_targets_mean": 2419.3, + "valid_targets_min": 826 + }, + { + "epoch": 2.8152492668621703, + "grad_norm": 0.5944406321947248, + "learning_rate": 2.988155015882671e-05, + "loss": 0.2359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23274701833724976, + "step": 2880, + "valid_targets_mean": 3016.2, + "valid_targets_min": 906 + }, + { + "epoch": 2.820136852394917, + "grad_norm": 0.7333620836779235, + "learning_rate": 2.9839134797632448e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2247675657272339, + "step": 2885, + "valid_targets_mean": 1883.5, + "valid_targets_min": 635 + }, + { + "epoch": 2.825024437927664, + "grad_norm": 0.6805311682642338, + "learning_rate": 2.9796660972860843e-05, + "loss": 0.2284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23136526346206665, + "step": 2890, + "valid_targets_mean": 2278.5, + "valid_targets_min": 807 + }, + { + "epoch": 2.8299120234604107, + "grad_norm": 0.7830726431275282, + "learning_rate": 2.975412893688893e-05, + "loss": 0.2374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24189212918281555, + "step": 2895, + "valid_targets_mean": 1769.8, + "valid_targets_min": 663 + }, + { + "epoch": 2.8347996089931575, + "grad_norm": 0.6769811140537966, + "learning_rate": 2.9711538942439637e-05, + "loss": 0.2322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22272557020187378, + "step": 2900, + "valid_targets_mean": 2156.8, + "valid_targets_min": 876 + }, + { + "epoch": 2.8396871945259043, + "grad_norm": 0.7674529053801956, + "learning_rate": 2.9668891242580287e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2575834095478058, + "step": 2905, + "valid_targets_mean": 2216.5, + "valid_targets_min": 744 + }, + { + "epoch": 2.844574780058651, + "grad_norm": 0.6658603801212766, + "learning_rate": 2.962618609072108e-05, + "loss": 0.2373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25467610359191895, + "step": 2910, + "valid_targets_mean": 2915.1, + "valid_targets_min": 694 + }, + { + "epoch": 2.849462365591398, + "grad_norm": 0.62994164165599, + "learning_rate": 2.9583423740613583e-05, + "loss": 0.2469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25572115182876587, + "step": 2915, + "valid_targets_mean": 2733.2, + "valid_targets_min": 1002 + }, + { + "epoch": 2.854349951124145, + "grad_norm": 0.7802082469784795, + "learning_rate": 2.954060444634924e-05, + "loss": 0.2347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21022462844848633, + "step": 2920, + "valid_targets_mean": 1662.9, + "valid_targets_min": 749 + }, + { + "epoch": 2.8592375366568916, + "grad_norm": 0.7025252669117502, + "learning_rate": 2.9497728462357854e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2211170792579651, + "step": 2925, + "valid_targets_mean": 1993.3, + "valid_targets_min": 695 + }, + { + "epoch": 2.8641251221896384, + "grad_norm": 0.5702454921369775, + "learning_rate": 2.9454796043406082e-05, + "loss": 0.2325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22452156245708466, + "step": 2930, + "valid_targets_mean": 3428.3, + "valid_targets_min": 765 + }, + { + "epoch": 2.8690127077223853, + "grad_norm": 0.737348955153242, + "learning_rate": 2.9411807444595903e-05, + "loss": 0.2252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21879072487354279, + "step": 2935, + "valid_targets_mean": 2037.2, + "valid_targets_min": 636 + }, + { + "epoch": 2.873900293255132, + "grad_norm": 0.7194505785174569, + "learning_rate": 2.936876292136311e-05, + "loss": 0.238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2638698220252991, + "step": 2940, + "valid_targets_mean": 2237.4, + "valid_targets_min": 834 + }, + { + "epoch": 2.878787878787879, + "grad_norm": 0.7215626747806759, + "learning_rate": 2.9325662729475808e-05, + "loss": 0.2369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24015557765960693, + "step": 2945, + "valid_targets_mean": 2293.9, + "valid_targets_min": 752 + }, + { + "epoch": 2.8836754643206257, + "grad_norm": 0.6723838618072419, + "learning_rate": 2.928250712503288e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21491815149784088, + "step": 2950, + "valid_targets_mean": 2184.8, + "valid_targets_min": 979 + }, + { + "epoch": 2.8885630498533725, + "grad_norm": 0.7739998699952924, + "learning_rate": 2.9239296364462467e-05, + "loss": 0.238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25200653076171875, + "step": 2955, + "valid_targets_mean": 2104.8, + "valid_targets_min": 815 + }, + { + "epoch": 2.8934506353861194, + "grad_norm": 0.6661883736388912, + "learning_rate": 2.919603070452043e-05, + "loss": 0.2376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24104368686676025, + "step": 2960, + "valid_targets_mean": 2636.2, + "valid_targets_min": 821 + }, + { + "epoch": 2.898338220918866, + "grad_norm": 0.777296307434507, + "learning_rate": 2.915271040228886e-05, + "loss": 0.227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22833505272865295, + "step": 2965, + "valid_targets_mean": 1935.4, + "valid_targets_min": 892 + }, + { + "epoch": 2.903225806451613, + "grad_norm": 0.6089518142750111, + "learning_rate": 2.9109335715174517e-05, + "loss": 0.233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23753002285957336, + "step": 2970, + "valid_targets_mean": 3033.2, + "valid_targets_min": 1109 + }, + { + "epoch": 2.90811339198436, + "grad_norm": 0.782726714582175, + "learning_rate": 2.9065906900907318e-05, + "loss": 0.2335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23450714349746704, + "step": 2975, + "valid_targets_mean": 1985.1, + "valid_targets_min": 770 + }, + { + "epoch": 2.9130009775171066, + "grad_norm": 0.6413112020747412, + "learning_rate": 2.9022424217538797e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2412833422422409, + "step": 2980, + "valid_targets_mean": 2716.9, + "valid_targets_min": 985 + }, + { + "epoch": 2.9178885630498534, + "grad_norm": 0.6854532986235314, + "learning_rate": 2.8978887923440573e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23968929052352905, + "step": 2985, + "valid_targets_mean": 2625.2, + "valid_targets_min": 940 + }, + { + "epoch": 2.9227761485826003, + "grad_norm": 0.6876034052530781, + "learning_rate": 2.8935298277302827e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24147212505340576, + "step": 2990, + "valid_targets_mean": 2415.8, + "valid_targets_min": 727 + }, + { + "epoch": 2.927663734115347, + "grad_norm": 0.6592399161754122, + "learning_rate": 2.8891655538132747e-05, + "loss": 0.2374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24528644979000092, + "step": 2995, + "valid_targets_mean": 2493.1, + "valid_targets_min": 879 + }, + { + "epoch": 2.932551319648094, + "grad_norm": 0.7330835802360961, + "learning_rate": 2.8847959965252997e-05, + "loss": 0.2362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24110592901706696, + "step": 3000, + "valid_targets_mean": 2464.8, + "valid_targets_min": 714 + }, + { + "epoch": 2.9374389051808407, + "grad_norm": 0.9963806436549506, + "learning_rate": 2.8804211818300175e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24917423725128174, + "step": 3005, + "valid_targets_mean": 2112.3, + "valid_targets_min": 635 + }, + { + "epoch": 2.9423264907135875, + "grad_norm": 0.7093645468792228, + "learning_rate": 2.8760411357223274e-05, + "loss": 0.2496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.256412148475647, + "step": 3010, + "valid_targets_mean": 2230.4, + "valid_targets_min": 514 + }, + { + "epoch": 2.9472140762463344, + "grad_norm": 0.6313278112245672, + "learning_rate": 2.8716558842282133e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21212854981422424, + "step": 3015, + "valid_targets_mean": 2637.0, + "valid_targets_min": 916 + }, + { + "epoch": 2.952101661779081, + "grad_norm": 0.5943727879588198, + "learning_rate": 2.8672654534045893e-05, + "loss": 0.2323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22816044092178345, + "step": 3020, + "valid_targets_mean": 3090.9, + "valid_targets_min": 819 + }, + { + "epoch": 2.956989247311828, + "grad_norm": 0.6972022185831341, + "learning_rate": 2.8628698693391454e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24018850922584534, + "step": 3025, + "valid_targets_mean": 2340.3, + "valid_targets_min": 807 + }, + { + "epoch": 2.961876832844575, + "grad_norm": 0.6777221287793921, + "learning_rate": 2.8584691581501904e-05, + "loss": 0.2387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22792454063892365, + "step": 3030, + "valid_targets_mean": 2147.8, + "valid_targets_min": 929 + }, + { + "epoch": 2.9667644183773216, + "grad_norm": 0.608921349958511, + "learning_rate": 2.8540633459864984e-05, + "loss": 0.2218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22310322523117065, + "step": 3035, + "valid_targets_mean": 2694.1, + "valid_targets_min": 714 + }, + { + "epoch": 2.9716520039100685, + "grad_norm": 0.6960978391554294, + "learning_rate": 2.8496524590271558e-05, + "loss": 0.2428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25981682538986206, + "step": 3040, + "valid_targets_mean": 2517.1, + "valid_targets_min": 940 + }, + { + "epoch": 2.9765395894428153, + "grad_norm": 0.7378583701878564, + "learning_rate": 2.8452365234813992e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1961268186569214, + "step": 3045, + "valid_targets_mean": 2311.8, + "valid_targets_min": 740 + }, + { + "epoch": 2.981427174975562, + "grad_norm": 0.6720248721061973, + "learning_rate": 2.8408155655884666e-05, + "loss": 0.2336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2390064001083374, + "step": 3050, + "valid_targets_mean": 2484.6, + "valid_targets_min": 1076 + }, + { + "epoch": 2.986314760508309, + "grad_norm": 0.6878264817723515, + "learning_rate": 2.836389611617437e-05, + "loss": 0.241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2337898164987564, + "step": 3055, + "valid_targets_mean": 2284.5, + "valid_targets_min": 1266 + }, + { + "epoch": 2.9912023460410557, + "grad_norm": 0.6966111553176655, + "learning_rate": 2.8319586878670767e-05, + "loss": 0.246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22837823629379272, + "step": 3060, + "valid_targets_mean": 2068.6, + "valid_targets_min": 772 + }, + { + "epoch": 2.9960899315738025, + "grad_norm": 0.6945196129457747, + "learning_rate": 2.827522820665681e-05, + "loss": 0.2375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2056921422481537, + "step": 3065, + "valid_targets_mean": 2009.1, + "valid_targets_min": 612 + }, + { + "epoch": 3.0009775171065494, + "grad_norm": 0.6719993247544616, + "learning_rate": 2.82308203637092e-05, + "loss": 0.2409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2267838567495346, + "step": 3070, + "valid_targets_mean": 2419.8, + "valid_targets_min": 732 + }, + { + "epoch": 3.005865102639296, + "grad_norm": 0.6667645399427046, + "learning_rate": 2.8186363613696807e-05, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22658130526542664, + "step": 3075, + "valid_targets_mean": 2837.1, + "valid_targets_min": 648 + }, + { + "epoch": 3.010752688172043, + "grad_norm": 0.6995064995237391, + "learning_rate": 2.81418582207791e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21117915213108063, + "step": 3080, + "valid_targets_mean": 2488.8, + "valid_targets_min": 746 + }, + { + "epoch": 3.01564027370479, + "grad_norm": 0.7816935143677926, + "learning_rate": 2.809730444940459e-05, + "loss": 0.2033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19485579431056976, + "step": 3085, + "valid_targets_mean": 1877.2, + "valid_targets_min": 585 + }, + { + "epoch": 3.0205278592375366, + "grad_norm": 0.6280789848534472, + "learning_rate": 2.8052702564309232e-05, + "loss": 0.1944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18880394101142883, + "step": 3090, + "valid_targets_mean": 2680.8, + "valid_targets_min": 754 + }, + { + "epoch": 3.0254154447702835, + "grad_norm": 0.7246655595987659, + "learning_rate": 2.8008052830514882e-05, + "loss": 0.2121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21295440196990967, + "step": 3095, + "valid_targets_mean": 2274.8, + "valid_targets_min": 788 + }, + { + "epoch": 3.0303030303030303, + "grad_norm": 0.7042756737696498, + "learning_rate": 2.7963355513327717e-05, + "loss": 0.2091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23234760761260986, + "step": 3100, + "valid_targets_mean": 2904.4, + "valid_targets_min": 708 + }, + { + "epoch": 3.035190615835777, + "grad_norm": 0.7467994069578198, + "learning_rate": 2.7918610878336644e-05, + "loss": 0.2076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22012445330619812, + "step": 3105, + "valid_targets_mean": 2543.5, + "valid_targets_min": 893 + }, + { + "epoch": 3.040078201368524, + "grad_norm": 0.6581294580719984, + "learning_rate": 2.7873819191411732e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1838349997997284, + "step": 3110, + "valid_targets_mean": 2560.5, + "valid_targets_min": 1434 + }, + { + "epoch": 3.0449657869012707, + "grad_norm": 0.6275202980378985, + "learning_rate": 2.782898071870261e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1899479627609253, + "step": 3115, + "valid_targets_mean": 2978.4, + "valid_targets_min": 723 + }, + { + "epoch": 3.0498533724340176, + "grad_norm": 0.708666227527697, + "learning_rate": 2.7784095726636945e-05, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2058933973312378, + "step": 3120, + "valid_targets_mean": 2619.5, + "valid_targets_min": 1038 + }, + { + "epoch": 3.0547409579667644, + "grad_norm": 0.7252823344406546, + "learning_rate": 2.773916448191879e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2228691130876541, + "step": 3125, + "valid_targets_mean": 2502.4, + "valid_targets_min": 855 + }, + { + "epoch": 3.059628543499511, + "grad_norm": 0.6775893290948455, + "learning_rate": 2.7694187251527034e-05, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20811831951141357, + "step": 3130, + "valid_targets_mean": 2852.7, + "valid_targets_min": 834 + }, + { + "epoch": 3.064516129032258, + "grad_norm": 0.6750667682794524, + "learning_rate": 2.7649164302713818e-05, + "loss": 0.1971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2103792279958725, + "step": 3135, + "valid_targets_mean": 2695.4, + "valid_targets_min": 1083 + }, + { + "epoch": 3.069403714565005, + "grad_norm": 0.6452750671952329, + "learning_rate": 2.7604095903002925e-05, + "loss": 0.2068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19791629910469055, + "step": 3140, + "valid_targets_mean": 2622.8, + "valid_targets_min": 851 + }, + { + "epoch": 3.0742913000977516, + "grad_norm": 0.6855575675327805, + "learning_rate": 2.7558982320188228e-05, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20340736210346222, + "step": 3145, + "valid_targets_mean": 2924.1, + "valid_targets_min": 858 + }, + { + "epoch": 3.0791788856304985, + "grad_norm": 0.7823747155692843, + "learning_rate": 2.7513823822332044e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19628237187862396, + "step": 3150, + "valid_targets_mean": 1840.7, + "valid_targets_min": 759 + }, + { + "epoch": 3.0840664711632453, + "grad_norm": 0.8146578281379756, + "learning_rate": 2.74686206777636e-05, + "loss": 0.1991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20766061544418335, + "step": 3155, + "valid_targets_mean": 1969.3, + "valid_targets_min": 918 + }, + { + "epoch": 3.088954056695992, + "grad_norm": 0.6883738938370201, + "learning_rate": 2.7423373155077403e-05, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18747478723526, + "step": 3160, + "valid_targets_mean": 2373.4, + "valid_targets_min": 856 + }, + { + "epoch": 3.093841642228739, + "grad_norm": 0.7077726637761635, + "learning_rate": 2.7378081523131644e-05, + "loss": 0.1966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20626375079154968, + "step": 3165, + "valid_targets_mean": 2385.8, + "valid_targets_min": 683 + }, + { + "epoch": 3.0987292277614857, + "grad_norm": 0.8627305595439234, + "learning_rate": 2.7332746051046616e-05, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21347269415855408, + "step": 3170, + "valid_targets_mean": 2014.2, + "valid_targets_min": 734 + }, + { + "epoch": 3.1036168132942326, + "grad_norm": 0.7893777089075921, + "learning_rate": 2.72873670082031e-05, + "loss": 0.2108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2146587371826172, + "step": 3175, + "valid_targets_mean": 2289.4, + "valid_targets_min": 642 + }, + { + "epoch": 3.1085043988269794, + "grad_norm": 0.8273291714086848, + "learning_rate": 2.7241944664240792e-05, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16076461970806122, + "step": 3180, + "valid_targets_mean": 1337.8, + "valid_targets_min": 514 + }, + { + "epoch": 3.113391984359726, + "grad_norm": 0.729684484993837, + "learning_rate": 2.719647928905666e-05, + "loss": 0.1985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20329692959785461, + "step": 3185, + "valid_targets_mean": 2202.4, + "valid_targets_min": 1251 + }, + { + "epoch": 3.118279569892473, + "grad_norm": 0.6526019445711841, + "learning_rate": 2.715097115280337e-05, + "loss": 0.2076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19159898161888123, + "step": 3190, + "valid_targets_mean": 2702.8, + "valid_targets_min": 793 + }, + { + "epoch": 3.12316715542522, + "grad_norm": 0.7498464955317224, + "learning_rate": 2.7105420525887667e-05, + "loss": 0.1991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2095525860786438, + "step": 3195, + "valid_targets_mean": 2154.8, + "valid_targets_min": 916 + }, + { + "epoch": 3.1280547409579667, + "grad_norm": 0.7063138314214019, + "learning_rate": 2.7059827678968775e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19605648517608643, + "step": 3200, + "valid_targets_mean": 2327.6, + "valid_targets_min": 916 + }, + { + "epoch": 3.1329423264907135, + "grad_norm": 0.759819492933154, + "learning_rate": 2.7014192882956795e-05, + "loss": 0.204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2000146508216858, + "step": 3205, + "valid_targets_mean": 2217.5, + "valid_targets_min": 682 + }, + { + "epoch": 3.1378299120234603, + "grad_norm": 0.7249197896837959, + "learning_rate": 2.6968516409011073e-05, + "loss": 0.2096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2277396023273468, + "step": 3210, + "valid_targets_mean": 2899.7, + "valid_targets_min": 905 + }, + { + "epoch": 3.142717497556207, + "grad_norm": 0.6645341527478739, + "learning_rate": 2.6922798528538615e-05, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1898479163646698, + "step": 3215, + "valid_targets_mean": 2493.7, + "valid_targets_min": 697 + }, + { + "epoch": 3.147605083088954, + "grad_norm": 0.7890579447239405, + "learning_rate": 2.6877039513192452e-05, + "loss": 0.1952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1876201033592224, + "step": 3220, + "valid_targets_mean": 2037.3, + "valid_targets_min": 916 + }, + { + "epoch": 3.1524926686217007, + "grad_norm": 0.7907814247477751, + "learning_rate": 2.6831239634870044e-05, + "loss": 0.2053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19796785712242126, + "step": 3225, + "valid_targets_mean": 1733.1, + "valid_targets_min": 742 + }, + { + "epoch": 3.1573802541544476, + "grad_norm": 0.7070573564286917, + "learning_rate": 2.6785399165711662e-05, + "loss": 0.2101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19448038935661316, + "step": 3230, + "valid_targets_mean": 2516.5, + "valid_targets_min": 1012 + }, + { + "epoch": 3.1622678396871944, + "grad_norm": 0.7496403178879768, + "learning_rate": 2.673951837809874e-05, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20808371901512146, + "step": 3235, + "valid_targets_mean": 2226.3, + "valid_targets_min": 702 + }, + { + "epoch": 3.167155425219941, + "grad_norm": 0.7481396583011426, + "learning_rate": 2.669359754465231e-05, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20283767580986023, + "step": 3240, + "valid_targets_mean": 2323.2, + "valid_targets_min": 618 + }, + { + "epoch": 3.172043010752688, + "grad_norm": 0.7195578782630686, + "learning_rate": 2.6647636938231342e-05, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21234679222106934, + "step": 3245, + "valid_targets_mean": 2472.0, + "valid_targets_min": 642 + }, + { + "epoch": 3.176930596285435, + "grad_norm": 0.7513145461378141, + "learning_rate": 2.6601636831931134e-05, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21580851078033447, + "step": 3250, + "valid_targets_mean": 2293.7, + "valid_targets_min": 521 + }, + { + "epoch": 3.1818181818181817, + "grad_norm": 0.7021834829360021, + "learning_rate": 2.6555597499081694e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2086678445339203, + "step": 3255, + "valid_targets_mean": 2655.6, + "valid_targets_min": 921 + }, + { + "epoch": 3.1867057673509285, + "grad_norm": 0.7337259433387169, + "learning_rate": 2.6509519213246107e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19709116220474243, + "step": 3260, + "valid_targets_mean": 2152.6, + "valid_targets_min": 661 + }, + { + "epoch": 3.1915933528836753, + "grad_norm": 0.7732965615576772, + "learning_rate": 2.646340224821892e-05, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17325511574745178, + "step": 3265, + "valid_targets_mean": 1718.6, + "valid_targets_min": 757 + }, + { + "epoch": 3.196480938416422, + "grad_norm": 0.7642020533927616, + "learning_rate": 2.64172468780245e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21912142634391785, + "step": 3270, + "valid_targets_mean": 1914.9, + "valid_targets_min": 762 + }, + { + "epoch": 3.201368523949169, + "grad_norm": 0.7137081449723205, + "learning_rate": 2.6371053376915442e-05, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1930808126926422, + "step": 3275, + "valid_targets_mean": 2530.1, + "valid_targets_min": 1259 + }, + { + "epoch": 3.2062561094819158, + "grad_norm": 0.7965252099837399, + "learning_rate": 2.6324822019370874e-05, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20748776197433472, + "step": 3280, + "valid_targets_mean": 2264.4, + "valid_targets_min": 801 + }, + { + "epoch": 3.2111436950146626, + "grad_norm": 0.7194214666441809, + "learning_rate": 2.6278553080094887e-05, + "loss": 0.2069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22017285227775574, + "step": 3285, + "valid_targets_mean": 2692.5, + "valid_targets_min": 794 + }, + { + "epoch": 3.2160312805474094, + "grad_norm": 0.7245863184684226, + "learning_rate": 2.6232246834014876e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1813446581363678, + "step": 3290, + "valid_targets_mean": 2046.6, + "valid_targets_min": 659 + }, + { + "epoch": 3.220918866080156, + "grad_norm": 0.7908976384173254, + "learning_rate": 2.618590355627992e-05, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2101239562034607, + "step": 3295, + "valid_targets_mean": 2258.9, + "valid_targets_min": 1006 + }, + { + "epoch": 3.225806451612903, + "grad_norm": 0.7977860033215038, + "learning_rate": 2.613952352225912e-05, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18104560673236847, + "step": 3300, + "valid_targets_mean": 1767.8, + "valid_targets_min": 484 + }, + { + "epoch": 3.23069403714565, + "grad_norm": 0.6115674983154584, + "learning_rate": 2.6093107007539998e-05, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17804718017578125, + "step": 3305, + "valid_targets_mean": 3087.1, + "valid_targets_min": 818 + }, + { + "epoch": 3.2355816226783967, + "grad_norm": 0.6099669179614807, + "learning_rate": 2.604665428792683e-05, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20128144323825836, + "step": 3310, + "valid_targets_mean": 3071.8, + "valid_targets_min": 1057 + }, + { + "epoch": 3.2404692082111435, + "grad_norm": 0.7754454106550883, + "learning_rate": 2.600016563943904e-05, + "loss": 0.2188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.216938778758049, + "step": 3315, + "valid_targets_mean": 2073.1, + "valid_targets_min": 740 + }, + { + "epoch": 3.2453567937438903, + "grad_norm": 0.7143546354659018, + "learning_rate": 2.5953641338309496e-05, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19034463167190552, + "step": 3320, + "valid_targets_mean": 2202.1, + "valid_targets_min": 737 + }, + { + "epoch": 3.250244379276637, + "grad_norm": 0.8320544699825392, + "learning_rate": 2.590708166098296e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20066097378730774, + "step": 3325, + "valid_targets_mean": 2138.4, + "valid_targets_min": 985 + }, + { + "epoch": 3.255131964809384, + "grad_norm": 0.7253333750300228, + "learning_rate": 2.586048688411437e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23125603795051575, + "step": 3330, + "valid_targets_mean": 2893.1, + "valid_targets_min": 1018 + }, + { + "epoch": 3.2600195503421308, + "grad_norm": 0.730875427222105, + "learning_rate": 2.5813857284567237e-05, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19738978147506714, + "step": 3335, + "valid_targets_mean": 2522.3, + "valid_targets_min": 859 + }, + { + "epoch": 3.2649071358748776, + "grad_norm": 0.7298430353490984, + "learning_rate": 2.5767193139411964e-05, + "loss": 0.1981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20270711183547974, + "step": 3340, + "valid_targets_mean": 2434.2, + "valid_targets_min": 467 + }, + { + "epoch": 3.2697947214076244, + "grad_norm": 1.083616521939471, + "learning_rate": 2.5720494725924257e-05, + "loss": 0.2122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21499906480312347, + "step": 3345, + "valid_targets_mean": 1869.0, + "valid_targets_min": 724 + }, + { + "epoch": 3.274682306940371, + "grad_norm": 0.6154395147501789, + "learning_rate": 2.5673762321583415e-05, + "loss": 0.2117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21717801690101624, + "step": 3350, + "valid_targets_mean": 3228.8, + "valid_targets_min": 974 + }, + { + "epoch": 3.279569892473118, + "grad_norm": 0.7777446446492675, + "learning_rate": 2.5626996204070714e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.207222580909729, + "step": 3355, + "valid_targets_mean": 2294.8, + "valid_targets_min": 705 + }, + { + "epoch": 3.2844574780058653, + "grad_norm": 0.626315557736335, + "learning_rate": 2.558019665126777e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18110740184783936, + "step": 3360, + "valid_targets_mean": 2922.9, + "valid_targets_min": 607 + }, + { + "epoch": 3.289345063538612, + "grad_norm": 0.6898177043479149, + "learning_rate": 2.5533363941254826e-05, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18781378865242004, + "step": 3365, + "valid_targets_mean": 2249.1, + "valid_targets_min": 740 + }, + { + "epoch": 3.294232649071359, + "grad_norm": 0.7219158135860761, + "learning_rate": 2.5486498352309194e-05, + "loss": 0.1993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19829457998275757, + "step": 3370, + "valid_targets_mean": 2316.6, + "valid_targets_min": 896 + }, + { + "epoch": 3.2991202346041058, + "grad_norm": 0.7339876884452256, + "learning_rate": 2.5439600162903527e-05, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2177194505929947, + "step": 3375, + "valid_targets_mean": 2257.8, + "valid_targets_min": 726 + }, + { + "epoch": 3.3040078201368526, + "grad_norm": 0.6776390716226277, + "learning_rate": 2.5392669651704185e-05, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19257496297359467, + "step": 3380, + "valid_targets_mean": 2925.8, + "valid_targets_min": 1060 + }, + { + "epoch": 3.3088954056695994, + "grad_norm": 0.759226279377635, + "learning_rate": 2.5345707097569586e-05, + "loss": 0.2066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1948852837085724, + "step": 3385, + "valid_targets_mean": 2035.1, + "valid_targets_min": 679 + }, + { + "epoch": 3.313782991202346, + "grad_norm": 0.7681252041823609, + "learning_rate": 2.529871277954854e-05, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.219600647687912, + "step": 3390, + "valid_targets_mean": 2099.7, + "valid_targets_min": 798 + }, + { + "epoch": 3.318670576735093, + "grad_norm": 0.7587652122936347, + "learning_rate": 2.5251686976878618e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18549835681915283, + "step": 3395, + "valid_targets_mean": 2503.0, + "valid_targets_min": 847 + }, + { + "epoch": 3.32355816226784, + "grad_norm": 0.6848113837428216, + "learning_rate": 2.5204629968984443e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21900366246700287, + "step": 3400, + "valid_targets_mean": 2764.0, + "valid_targets_min": 651 + }, + { + "epoch": 3.3284457478005867, + "grad_norm": 0.6897084263388162, + "learning_rate": 2.5157542035476077e-05, + "loss": 0.1946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19517606496810913, + "step": 3405, + "valid_targets_mean": 2661.6, + "valid_targets_min": 960 + }, + { + "epoch": 3.3333333333333335, + "grad_norm": 0.7517965693921952, + "learning_rate": 2.5110423456147322e-05, + "loss": 0.2087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21612215042114258, + "step": 3410, + "valid_targets_mean": 2433.1, + "valid_targets_min": 975 + }, + { + "epoch": 3.3382209188660803, + "grad_norm": 0.7318661225517749, + "learning_rate": 2.50632745109741e-05, + "loss": 0.2108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1917639672756195, + "step": 3415, + "valid_targets_mean": 1934.4, + "valid_targets_min": 771 + }, + { + "epoch": 3.343108504398827, + "grad_norm": 0.7445802916995302, + "learning_rate": 2.501609548011275e-05, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1921059787273407, + "step": 3420, + "valid_targets_mean": 2087.8, + "valid_targets_min": 692 + }, + { + "epoch": 3.347996089931574, + "grad_norm": 0.6673062277647884, + "learning_rate": 2.4968886643898366e-05, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18176676332950592, + "step": 3425, + "valid_targets_mean": 2963.6, + "valid_targets_min": 1007 + }, + { + "epoch": 3.3528836754643208, + "grad_norm": 0.9097050470991745, + "learning_rate": 2.4921648282843176e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18103784322738647, + "step": 3430, + "valid_targets_mean": 2004.6, + "valid_targets_min": 628 + }, + { + "epoch": 3.3577712609970676, + "grad_norm": 0.7122931954083914, + "learning_rate": 2.4874380677634817e-05, + "loss": 0.2066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19356971979141235, + "step": 3435, + "valid_targets_mean": 2382.2, + "valid_targets_min": 914 + }, + { + "epoch": 3.3626588465298144, + "grad_norm": 0.7088171184773836, + "learning_rate": 2.4827084109134697e-05, + "loss": 0.2163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20455177128314972, + "step": 3440, + "valid_targets_mean": 2174.6, + "valid_targets_min": 657 + }, + { + "epoch": 3.367546432062561, + "grad_norm": 0.6612005144692485, + "learning_rate": 2.477975885837634e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21015259623527527, + "step": 3445, + "valid_targets_mean": 2880.4, + "valid_targets_min": 648 + }, + { + "epoch": 3.372434017595308, + "grad_norm": 0.7655924949272598, + "learning_rate": 2.473240520656367e-05, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2177879959344864, + "step": 3450, + "valid_targets_mean": 2229.4, + "valid_targets_min": 836 + }, + { + "epoch": 3.377321603128055, + "grad_norm": 0.7618724000532252, + "learning_rate": 2.4685023435069388e-05, + "loss": 0.2165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23734396696090698, + "step": 3455, + "valid_targets_mean": 2332.8, + "valid_targets_min": 621 + }, + { + "epoch": 3.3822091886608017, + "grad_norm": 0.8143707345252376, + "learning_rate": 2.4637613825433277e-05, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19915053248405457, + "step": 3460, + "valid_targets_mean": 1512.4, + "valid_targets_min": 668 + }, + { + "epoch": 3.3870967741935485, + "grad_norm": 0.8158876268350879, + "learning_rate": 2.4590176659360538e-05, + "loss": 0.2088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22826993465423584, + "step": 3465, + "valid_targets_mean": 2035.4, + "valid_targets_min": 650 + }, + { + "epoch": 3.3919843597262953, + "grad_norm": 0.6759338161481376, + "learning_rate": 2.4542712218720093e-05, + "loss": 0.2043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18781697750091553, + "step": 3470, + "valid_targets_mean": 2423.6, + "valid_targets_min": 738 + }, + { + "epoch": 3.396871945259042, + "grad_norm": 0.6931462841032442, + "learning_rate": 2.449522078554294e-05, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20876827836036682, + "step": 3475, + "valid_targets_mean": 2486.4, + "valid_targets_min": 610 + }, + { + "epoch": 3.401759530791789, + "grad_norm": 0.656880638760444, + "learning_rate": 2.444770264202047e-05, + "loss": 0.2045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21965506672859192, + "step": 3480, + "valid_targets_mean": 2932.4, + "valid_targets_min": 924 + }, + { + "epoch": 3.4066471163245358, + "grad_norm": 0.7793846246855756, + "learning_rate": 2.4400158070502773e-05, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2179027497768402, + "step": 3485, + "valid_targets_mean": 2128.7, + "valid_targets_min": 807 + }, + { + "epoch": 3.4115347018572826, + "grad_norm": 0.6366727057551457, + "learning_rate": 2.435258735349699e-05, + "loss": 0.2043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.207027405500412, + "step": 3490, + "valid_targets_mean": 3005.8, + "valid_targets_min": 901 + }, + { + "epoch": 3.4164222873900294, + "grad_norm": 0.8083254712829566, + "learning_rate": 2.4304990773665587e-05, + "loss": 0.2094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22201141715049744, + "step": 3495, + "valid_targets_mean": 2266.4, + "valid_targets_min": 790 + }, + { + "epoch": 3.421309872922776, + "grad_norm": 0.7883137521426757, + "learning_rate": 2.4257368613824745e-05, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18446215987205505, + "step": 3500, + "valid_targets_mean": 1863.5, + "valid_targets_min": 1075 + }, + { + "epoch": 3.426197458455523, + "grad_norm": 0.7729853988513622, + "learning_rate": 2.4209721156942615e-05, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1607511341571808, + "step": 3505, + "valid_targets_mean": 1680.6, + "valid_targets_min": 792 + }, + { + "epoch": 3.43108504398827, + "grad_norm": 0.7615871115872819, + "learning_rate": 2.416204868613765e-05, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2097199559211731, + "step": 3510, + "valid_targets_mean": 2369.9, + "valid_targets_min": 809 + }, + { + "epoch": 3.4359726295210167, + "grad_norm": 0.7886947243326079, + "learning_rate": 2.4114351484676975e-05, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22828155755996704, + "step": 3515, + "valid_targets_mean": 2259.6, + "valid_targets_min": 702 + }, + { + "epoch": 3.4408602150537635, + "grad_norm": 0.6428893882305308, + "learning_rate": 2.4066629835974614e-05, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22016070783138275, + "step": 3520, + "valid_targets_mean": 3275.2, + "valid_targets_min": 1265 + }, + { + "epoch": 3.4457478005865103, + "grad_norm": 0.822974697294548, + "learning_rate": 2.40188840235899e-05, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20060139894485474, + "step": 3525, + "valid_targets_mean": 1830.9, + "valid_targets_min": 692 + }, + { + "epoch": 3.450635386119257, + "grad_norm": 0.6770812059975508, + "learning_rate": 2.3971114331225715e-05, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20556485652923584, + "step": 3530, + "valid_targets_mean": 2694.8, + "valid_targets_min": 642 + }, + { + "epoch": 3.455522971652004, + "grad_norm": 0.7786397671656108, + "learning_rate": 2.392332104272684e-05, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2165582925081253, + "step": 3535, + "valid_targets_mean": 2489.4, + "valid_targets_min": 786 + }, + { + "epoch": 3.4604105571847508, + "grad_norm": 0.7323225666730152, + "learning_rate": 2.3875504442078282e-05, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21573954820632935, + "step": 3540, + "valid_targets_mean": 2259.0, + "valid_targets_min": 713 + }, + { + "epoch": 3.4652981427174976, + "grad_norm": 0.6799054356460841, + "learning_rate": 2.3827664813403537e-05, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1902822107076645, + "step": 3545, + "valid_targets_mean": 2551.8, + "valid_targets_min": 627 + }, + { + "epoch": 3.4701857282502444, + "grad_norm": 0.7472013601982304, + "learning_rate": 2.3779802440962958e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20308758318424225, + "step": 3550, + "valid_targets_mean": 2375.5, + "valid_targets_min": 982 + }, + { + "epoch": 3.4750733137829912, + "grad_norm": 0.7391959529865234, + "learning_rate": 2.3731917609152023e-05, + "loss": 0.2097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21505077183246613, + "step": 3555, + "valid_targets_mean": 2454.9, + "valid_targets_min": 704 + }, + { + "epoch": 3.479960899315738, + "grad_norm": 0.6408227234028441, + "learning_rate": 2.3684010602499675e-05, + "loss": 0.204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18841466307640076, + "step": 3560, + "valid_targets_mean": 2844.4, + "valid_targets_min": 736 + }, + { + "epoch": 3.484848484848485, + "grad_norm": 0.6995274259824635, + "learning_rate": 2.363608170566661e-05, + "loss": 0.1934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19288122653961182, + "step": 3565, + "valid_targets_mean": 2298.1, + "valid_targets_min": 593 + }, + { + "epoch": 3.4897360703812317, + "grad_norm": 0.712559649636414, + "learning_rate": 2.35881312034436e-05, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18562278151512146, + "step": 3570, + "valid_targets_mean": 2527.5, + "valid_targets_min": 1190 + }, + { + "epoch": 3.4946236559139785, + "grad_norm": 0.7601682130608852, + "learning_rate": 2.3540159380749794e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18290041387081146, + "step": 3575, + "valid_targets_mean": 1959.9, + "valid_targets_min": 244 + }, + { + "epoch": 3.4995112414467253, + "grad_norm": 0.6518733256350043, + "learning_rate": 2.349216652263101e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1879890263080597, + "step": 3580, + "valid_targets_mean": 2892.5, + "valid_targets_min": 788 + }, + { + "epoch": 3.504398826979472, + "grad_norm": 0.6632097525792074, + "learning_rate": 2.3444152914258087e-05, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19737496972084045, + "step": 3585, + "valid_targets_mean": 2592.2, + "valid_targets_min": 694 + }, + { + "epoch": 3.509286412512219, + "grad_norm": 0.701914967587876, + "learning_rate": 2.339611884092514e-05, + "loss": 0.1971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21197064220905304, + "step": 3590, + "valid_targets_mean": 2489.3, + "valid_targets_min": 763 + }, + { + "epoch": 3.5141739980449658, + "grad_norm": 0.7664935207947431, + "learning_rate": 2.3348064588047893e-05, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19324423372745514, + "step": 3595, + "valid_targets_mean": 1855.3, + "valid_targets_min": 635 + }, + { + "epoch": 3.5190615835777126, + "grad_norm": 0.7141259014048177, + "learning_rate": 2.329999044116197e-05, + "loss": 0.2085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21774733066558838, + "step": 3600, + "valid_targets_mean": 2392.5, + "valid_targets_min": 682 + }, + { + "epoch": 3.5239491691104594, + "grad_norm": 0.6730829479768565, + "learning_rate": 2.3251896685921205e-05, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19916030764579773, + "step": 3605, + "valid_targets_mean": 2920.2, + "valid_targets_min": 821 + }, + { + "epoch": 3.5288367546432062, + "grad_norm": 0.7671142376386617, + "learning_rate": 2.3203783608095954e-05, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23348473012447357, + "step": 3610, + "valid_targets_mean": 2295.7, + "valid_targets_min": 959 + }, + { + "epoch": 3.533724340175953, + "grad_norm": 0.7689072632328403, + "learning_rate": 2.3155651493571382e-05, + "loss": 0.2053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20937813818454742, + "step": 3615, + "valid_targets_mean": 2182.8, + "valid_targets_min": 1082 + }, + { + "epoch": 3.5386119257087, + "grad_norm": 0.7292242265860919, + "learning_rate": 2.3107500628345756e-05, + "loss": 0.2075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22746799886226654, + "step": 3620, + "valid_targets_mean": 2486.5, + "valid_targets_min": 730 + }, + { + "epoch": 3.5434995112414467, + "grad_norm": 0.7976787109769994, + "learning_rate": 2.3059331298528776e-05, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2245236039161682, + "step": 3625, + "valid_targets_mean": 1958.8, + "valid_targets_min": 1215 + }, + { + "epoch": 3.5483870967741935, + "grad_norm": 0.7468841302577356, + "learning_rate": 2.301114379033984e-05, + "loss": 0.201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18756568431854248, + "step": 3630, + "valid_targets_mean": 1843.2, + "valid_targets_min": 900 + }, + { + "epoch": 3.5532746823069403, + "grad_norm": 0.7052365263985373, + "learning_rate": 2.2962938390106388e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.190373957157135, + "step": 3635, + "valid_targets_mean": 2208.5, + "valid_targets_min": 672 + }, + { + "epoch": 3.558162267839687, + "grad_norm": 0.7861655797997658, + "learning_rate": 2.291471538426214e-05, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21023714542388916, + "step": 3640, + "valid_targets_mean": 2122.3, + "valid_targets_min": 654 + }, + { + "epoch": 3.563049853372434, + "grad_norm": 0.7654010417442486, + "learning_rate": 2.2866475059345453e-05, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21987277269363403, + "step": 3645, + "valid_targets_mean": 2314.6, + "valid_targets_min": 872 + }, + { + "epoch": 3.567937438905181, + "grad_norm": 0.7329153965811843, + "learning_rate": 2.2818217701997577e-05, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19848138093948364, + "step": 3650, + "valid_targets_mean": 2325.4, + "valid_targets_min": 815 + }, + { + "epoch": 3.5728250244379276, + "grad_norm": 0.6582062601819844, + "learning_rate": 2.276994359896099e-05, + "loss": 0.2144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20447036623954773, + "step": 3655, + "valid_targets_mean": 2623.2, + "valid_targets_min": 985 + }, + { + "epoch": 3.5777126099706744, + "grad_norm": 0.6885840936099205, + "learning_rate": 2.2721653037077645e-05, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2211674004793167, + "step": 3660, + "valid_targets_mean": 2750.7, + "valid_targets_min": 701 + }, + { + "epoch": 3.5826001955034212, + "grad_norm": 0.7022727114163723, + "learning_rate": 2.2673346303287297e-05, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19418203830718994, + "step": 3665, + "valid_targets_mean": 2241.1, + "valid_targets_min": 548 + }, + { + "epoch": 3.587487781036168, + "grad_norm": 0.6929148877932964, + "learning_rate": 2.262502368462582e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22029688954353333, + "step": 3670, + "valid_targets_mean": 2660.6, + "valid_targets_min": 729 + }, + { + "epoch": 3.592375366568915, + "grad_norm": 0.6443170990968766, + "learning_rate": 2.2576685468223444e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18446439504623413, + "step": 3675, + "valid_targets_mean": 2688.0, + "valid_targets_min": 692 + }, + { + "epoch": 3.5972629521016617, + "grad_norm": 0.600440286447225, + "learning_rate": 2.2528331941303093e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19318996369838715, + "step": 3680, + "valid_targets_mean": 3134.5, + "valid_targets_min": 1041 + }, + { + "epoch": 3.6021505376344085, + "grad_norm": 0.7844338847583994, + "learning_rate": 2.2479963391178657e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21118606626987457, + "step": 3685, + "valid_targets_mean": 2066.8, + "valid_targets_min": 814 + }, + { + "epoch": 3.6070381231671553, + "grad_norm": 0.7215085805201092, + "learning_rate": 2.2431580105253313e-05, + "loss": 0.2149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20135241746902466, + "step": 3690, + "valid_targets_mean": 2184.1, + "valid_targets_min": 722 + }, + { + "epoch": 3.611925708699902, + "grad_norm": 0.6656818554741202, + "learning_rate": 2.238318237101778e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20373103022575378, + "step": 3695, + "valid_targets_mean": 2753.5, + "valid_targets_min": 975 + }, + { + "epoch": 3.616813294232649, + "grad_norm": 0.7180482881493666, + "learning_rate": 2.233477047604863e-05, + "loss": 0.2223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1976693719625473, + "step": 3700, + "valid_targets_mean": 2586.9, + "valid_targets_min": 692 + }, + { + "epoch": 3.621700879765396, + "grad_norm": 0.6461550790717755, + "learning_rate": 2.2286344708006578e-05, + "loss": 0.2125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20432640612125397, + "step": 3705, + "valid_targets_mean": 3076.8, + "valid_targets_min": 952 + }, + { + "epoch": 3.6265884652981426, + "grad_norm": 0.7135365897281873, + "learning_rate": 2.2237905354634775e-05, + "loss": 0.213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21222619712352753, + "step": 3710, + "valid_targets_mean": 2544.2, + "valid_targets_min": 1058 + }, + { + "epoch": 3.6314760508308894, + "grad_norm": 0.7106647598992838, + "learning_rate": 2.2189452703757097e-05, + "loss": 0.2045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1919505000114441, + "step": 3715, + "valid_targets_mean": 2239.9, + "valid_targets_min": 700 + }, + { + "epoch": 3.6363636363636362, + "grad_norm": 0.7370346403284481, + "learning_rate": 2.214098704327643e-05, + "loss": 0.2085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2074413299560547, + "step": 3720, + "valid_targets_mean": 2248.9, + "valid_targets_min": 504 + }, + { + "epoch": 3.641251221896383, + "grad_norm": 0.6946560980334792, + "learning_rate": 2.2092508661172957e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18183276057243347, + "step": 3725, + "valid_targets_mean": 2272.8, + "valid_targets_min": 746 + }, + { + "epoch": 3.64613880742913, + "grad_norm": 0.6384613289501591, + "learning_rate": 2.204401784550246e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18616612255573273, + "step": 3730, + "valid_targets_mean": 2584.4, + "valid_targets_min": 684 + }, + { + "epoch": 3.6510263929618767, + "grad_norm": 0.6909767548097899, + "learning_rate": 2.1995514884394596e-05, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1910645067691803, + "step": 3735, + "valid_targets_mean": 2587.4, + "valid_targets_min": 771 + }, + { + "epoch": 3.6559139784946235, + "grad_norm": 0.8148464145542923, + "learning_rate": 2.1947000066051204e-05, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20611697435379028, + "step": 3740, + "valid_targets_mean": 1887.4, + "valid_targets_min": 730 + }, + { + "epoch": 3.6608015640273703, + "grad_norm": 0.7056151602431089, + "learning_rate": 2.189847367874454e-05, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21500477194786072, + "step": 3745, + "valid_targets_mean": 2486.1, + "valid_targets_min": 1001 + }, + { + "epoch": 3.665689149560117, + "grad_norm": 1.0531242937391503, + "learning_rate": 2.1849936010815653e-05, + "loss": 0.212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21067142486572266, + "step": 3750, + "valid_targets_mean": 1737.8, + "valid_targets_min": 714 + }, + { + "epoch": 3.670576735092864, + "grad_norm": 0.7901950772244742, + "learning_rate": 2.1801387350672577e-05, + "loss": 0.2038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20664237439632416, + "step": 3755, + "valid_targets_mean": 2177.6, + "valid_targets_min": 627 + }, + { + "epoch": 3.675464320625611, + "grad_norm": 0.7840380182400347, + "learning_rate": 2.175282798678869e-05, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21524584293365479, + "step": 3760, + "valid_targets_mean": 2213.3, + "valid_targets_min": 722 + }, + { + "epoch": 3.6803519061583576, + "grad_norm": 0.7134090806306669, + "learning_rate": 2.170425820770096e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20987609028816223, + "step": 3765, + "valid_targets_mean": 2505.4, + "valid_targets_min": 1124 + }, + { + "epoch": 3.6852394916911044, + "grad_norm": 0.7246436271430389, + "learning_rate": 2.1655678302008232e-05, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20697399973869324, + "step": 3770, + "valid_targets_mean": 2323.0, + "valid_targets_min": 753 + }, + { + "epoch": 3.6901270772238517, + "grad_norm": 0.6770013885437635, + "learning_rate": 2.160708855836954e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1982710063457489, + "step": 3775, + "valid_targets_mean": 2790.8, + "valid_targets_min": 715 + }, + { + "epoch": 3.6950146627565985, + "grad_norm": 0.799738720478482, + "learning_rate": 2.1558489265502373e-05, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.220750093460083, + "step": 3780, + "valid_targets_mean": 2273.8, + "valid_targets_min": 750 + }, + { + "epoch": 3.6999022482893453, + "grad_norm": 0.6508927329830108, + "learning_rate": 2.1509880712180946e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20443478226661682, + "step": 3785, + "valid_targets_mean": 2857.2, + "valid_targets_min": 874 + }, + { + "epoch": 3.704789833822092, + "grad_norm": 0.6802733863979533, + "learning_rate": 2.1461263187234515e-05, + "loss": 0.213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20990541577339172, + "step": 3790, + "valid_targets_mean": 2561.4, + "valid_targets_min": 705 + }, + { + "epoch": 3.709677419354839, + "grad_norm": 0.7028949057842757, + "learning_rate": 2.141263697954563e-05, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2133583277463913, + "step": 3795, + "valid_targets_mean": 2545.5, + "valid_targets_min": 881 + }, + { + "epoch": 3.714565004887586, + "grad_norm": 0.6847821139141917, + "learning_rate": 2.1364002378048446e-05, + "loss": 0.2063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20359432697296143, + "step": 3800, + "valid_targets_mean": 2529.1, + "valid_targets_min": 914 + }, + { + "epoch": 3.7194525904203326, + "grad_norm": 0.7683431448503167, + "learning_rate": 2.1315359671726994e-05, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20140613615512848, + "step": 3805, + "valid_targets_mean": 2015.9, + "valid_targets_min": 740 + }, + { + "epoch": 3.7243401759530794, + "grad_norm": 0.6765983248131439, + "learning_rate": 2.1266709149613448e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19246436655521393, + "step": 3810, + "valid_targets_mean": 2428.9, + "valid_targets_min": 870 + }, + { + "epoch": 3.7292277614858262, + "grad_norm": 0.9226737395699486, + "learning_rate": 2.1218051100786443e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1814860701560974, + "step": 3815, + "valid_targets_mean": 1909.5, + "valid_targets_min": 779 + }, + { + "epoch": 3.734115347018573, + "grad_norm": 0.8000270533236249, + "learning_rate": 2.1169385814369317e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22426359355449677, + "step": 3820, + "valid_targets_mean": 1909.4, + "valid_targets_min": 711 + }, + { + "epoch": 3.73900293255132, + "grad_norm": 0.6753953556756007, + "learning_rate": 2.112071357952844e-05, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1894885152578354, + "step": 3825, + "valid_targets_mean": 2391.7, + "valid_targets_min": 696 + }, + { + "epoch": 3.7438905180840667, + "grad_norm": 0.7747404427802933, + "learning_rate": 2.1072034685471442e-05, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2056509256362915, + "step": 3830, + "valid_targets_mean": 1816.9, + "valid_targets_min": 548 + }, + { + "epoch": 3.7487781036168135, + "grad_norm": 0.7637718123929285, + "learning_rate": 2.1023349421445532e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20844748616218567, + "step": 3835, + "valid_targets_mean": 2643.7, + "valid_targets_min": 793 + }, + { + "epoch": 3.7536656891495603, + "grad_norm": 0.6354837562399616, + "learning_rate": 2.0974658076735767e-05, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1894562840461731, + "step": 3840, + "valid_targets_mean": 2781.6, + "valid_targets_min": 1309 + }, + { + "epoch": 3.758553274682307, + "grad_norm": 0.6766758246484479, + "learning_rate": 2.0925960940663356e-05, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19174224138259888, + "step": 3845, + "valid_targets_mean": 2529.0, + "valid_targets_min": 987 + }, + { + "epoch": 3.763440860215054, + "grad_norm": 0.6892947666726518, + "learning_rate": 2.087725830258389e-05, + "loss": 0.2063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2084968388080597, + "step": 3850, + "valid_targets_mean": 2646.6, + "valid_targets_min": 655 + }, + { + "epoch": 3.768328445747801, + "grad_norm": 0.70291002815775, + "learning_rate": 2.082855045188565e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20856565237045288, + "step": 3855, + "valid_targets_mean": 2704.9, + "valid_targets_min": 789 + }, + { + "epoch": 3.7732160312805476, + "grad_norm": 0.7692739550745331, + "learning_rate": 2.0779837677987927e-05, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19723054766654968, + "step": 3860, + "valid_targets_mean": 1855.1, + "valid_targets_min": 1118 + }, + { + "epoch": 3.7781036168132944, + "grad_norm": 0.7101378522156258, + "learning_rate": 2.073112027033923e-05, + "loss": 0.2175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21212387084960938, + "step": 3865, + "valid_targets_mean": 2229.6, + "valid_targets_min": 1150 + }, + { + "epoch": 3.7829912023460412, + "grad_norm": 0.795515076642536, + "learning_rate": 2.068239851841563e-05, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20189426839351654, + "step": 3870, + "valid_targets_mean": 2010.2, + "valid_targets_min": 768 + }, + { + "epoch": 3.787878787878788, + "grad_norm": 0.8748951610329812, + "learning_rate": 2.0633672711718975e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20840033888816833, + "step": 3875, + "valid_targets_mean": 2443.8, + "valid_targets_min": 1158 + }, + { + "epoch": 3.792766373411535, + "grad_norm": 0.6476901318697162, + "learning_rate": 2.0584943139775237e-05, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1948857605457306, + "step": 3880, + "valid_targets_mean": 2858.7, + "valid_targets_min": 882 + }, + { + "epoch": 3.7976539589442817, + "grad_norm": 0.7397815710205294, + "learning_rate": 2.053621009213276e-05, + "loss": 0.2054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21412993967533112, + "step": 3885, + "valid_targets_mean": 2184.9, + "valid_targets_min": 997 + }, + { + "epoch": 3.8025415444770285, + "grad_norm": 0.7146555857159458, + "learning_rate": 2.0487473858360515e-05, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18842893838882446, + "step": 3890, + "valid_targets_mean": 2270.2, + "valid_targets_min": 713 + }, + { + "epoch": 3.8074291300097753, + "grad_norm": 0.65940562297892, + "learning_rate": 2.0438734728046432e-05, + "loss": 0.1959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18981792032718658, + "step": 3895, + "valid_targets_mean": 2287.4, + "valid_targets_min": 1016 + }, + { + "epoch": 3.812316715542522, + "grad_norm": 0.8426072348378402, + "learning_rate": 2.0389992990795632e-05, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22454825043678284, + "step": 3900, + "valid_targets_mean": 1866.6, + "valid_targets_min": 537 + }, + { + "epoch": 3.817204301075269, + "grad_norm": 0.7342198247089646, + "learning_rate": 2.0341248936228748e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2228454351425171, + "step": 3905, + "valid_targets_mean": 2335.3, + "valid_targets_min": 1021 + }, + { + "epoch": 3.822091886608016, + "grad_norm": 0.6636543015733303, + "learning_rate": 2.0292502853980164e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20288674533367157, + "step": 3910, + "valid_targets_mean": 2667.2, + "valid_targets_min": 791 + }, + { + "epoch": 3.8269794721407626, + "grad_norm": 0.7610090323169763, + "learning_rate": 2.0243755033696313e-05, + "loss": 0.1981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20183882117271423, + "step": 3915, + "valid_targets_mean": 2155.2, + "valid_targets_min": 632 + }, + { + "epoch": 3.8318670576735094, + "grad_norm": 0.6831492165378048, + "learning_rate": 2.019500576503397e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1926528513431549, + "step": 3920, + "valid_targets_mean": 2656.8, + "valid_targets_min": 1210 + }, + { + "epoch": 3.8367546432062563, + "grad_norm": 0.7402425571751615, + "learning_rate": 2.0146255337658494e-05, + "loss": 0.1996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21684202551841736, + "step": 3925, + "valid_targets_mean": 2248.5, + "valid_targets_min": 420 + }, + { + "epoch": 3.841642228739003, + "grad_norm": 0.6921612269978583, + "learning_rate": 2.009750404124216e-05, + "loss": 0.1975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20540589094161987, + "step": 3930, + "valid_targets_mean": 2690.6, + "valid_targets_min": 697 + }, + { + "epoch": 3.84652981427175, + "grad_norm": 0.7529950655294566, + "learning_rate": 2.0048752165462373e-05, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21762652695178986, + "step": 3935, + "valid_targets_mean": 2733.9, + "valid_targets_min": 844 + }, + { + "epoch": 3.8514173998044967, + "grad_norm": 0.6580479552711862, + "learning_rate": 2e-05, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1937936544418335, + "step": 3940, + "valid_targets_mean": 2622.9, + "valid_targets_min": 632 + }, + { + "epoch": 3.8563049853372435, + "grad_norm": 0.8321134334974649, + "learning_rate": 1.9951247834537634e-05, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21654708683490753, + "step": 3945, + "valid_targets_mean": 2118.2, + "valid_targets_min": 594 + }, + { + "epoch": 3.8611925708699903, + "grad_norm": 0.8110075847407984, + "learning_rate": 1.9902495958757844e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21147048473358154, + "step": 3950, + "valid_targets_mean": 2464.2, + "valid_targets_min": 775 + }, + { + "epoch": 3.866080156402737, + "grad_norm": 0.8481527474330096, + "learning_rate": 1.985374466234151e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1752730906009674, + "step": 3955, + "valid_targets_mean": 1691.5, + "valid_targets_min": 680 + }, + { + "epoch": 3.870967741935484, + "grad_norm": 0.7391679144111151, + "learning_rate": 1.9804994234966036e-05, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2055065929889679, + "step": 3960, + "valid_targets_mean": 2255.8, + "valid_targets_min": 1006 + }, + { + "epoch": 3.875855327468231, + "grad_norm": 0.6879643282262666, + "learning_rate": 1.9756244966303693e-05, + "loss": 0.2096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21640875935554504, + "step": 3965, + "valid_targets_mean": 2545.9, + "valid_targets_min": 750 + }, + { + "epoch": 3.8807429130009776, + "grad_norm": 0.682330103660924, + "learning_rate": 1.9707497146019846e-05, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17964300513267517, + "step": 3970, + "valid_targets_mean": 2946.8, + "valid_targets_min": 750 + }, + { + "epoch": 3.8856304985337244, + "grad_norm": 0.6624662717730248, + "learning_rate": 1.9658751063771255e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20151904225349426, + "step": 3975, + "valid_targets_mean": 3265.1, + "valid_targets_min": 711 + }, + { + "epoch": 3.8905180840664713, + "grad_norm": 0.7423253354159347, + "learning_rate": 1.961000700920437e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19609767198562622, + "step": 3980, + "valid_targets_mean": 2088.9, + "valid_targets_min": 780 + }, + { + "epoch": 3.895405669599218, + "grad_norm": 0.9813652565552392, + "learning_rate": 1.956126527195358e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2142195701599121, + "step": 3985, + "valid_targets_mean": 2652.4, + "valid_targets_min": 759 + }, + { + "epoch": 3.900293255131965, + "grad_norm": 0.5712086066191886, + "learning_rate": 1.9512526141639492e-05, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16019731760025024, + "step": 3990, + "valid_targets_mean": 3155.6, + "valid_targets_min": 770 + }, + { + "epoch": 3.9051808406647117, + "grad_norm": 0.7587001616327813, + "learning_rate": 1.9463789907867248e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25113722681999207, + "step": 3995, + "valid_targets_mean": 2617.2, + "valid_targets_min": 869 + }, + { + "epoch": 3.9100684261974585, + "grad_norm": 0.7046294004069471, + "learning_rate": 1.941505686022476e-05, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18923306465148926, + "step": 4000, + "valid_targets_mean": 2173.3, + "valid_targets_min": 929 + }, + { + "epoch": 3.9149560117302054, + "grad_norm": 0.6934724921753709, + "learning_rate": 1.936632728828103e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20533259212970734, + "step": 4005, + "valid_targets_mean": 2358.2, + "valid_targets_min": 898 + }, + { + "epoch": 3.919843597262952, + "grad_norm": 0.8196595601053237, + "learning_rate": 1.9317601481584383e-05, + "loss": 0.208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2104458212852478, + "step": 4010, + "valid_targets_mean": 1972.3, + "valid_targets_min": 725 + }, + { + "epoch": 3.924731182795699, + "grad_norm": 0.7849322123412164, + "learning_rate": 1.926887972966077e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.269350528717041, + "step": 4015, + "valid_targets_mean": 2177.9, + "valid_targets_min": 819 + }, + { + "epoch": 3.929618768328446, + "grad_norm": 0.7525483327532918, + "learning_rate": 1.922016232201208e-05, + "loss": 0.2074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21186742186546326, + "step": 4020, + "valid_targets_mean": 2130.6, + "valid_targets_min": 548 + }, + { + "epoch": 3.9345063538611926, + "grad_norm": 0.6367950186766983, + "learning_rate": 1.917144954811435e-05, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19727790355682373, + "step": 4025, + "valid_targets_mean": 2766.4, + "valid_targets_min": 817 + }, + { + "epoch": 3.9393939393939394, + "grad_norm": 0.7362334474471227, + "learning_rate": 1.9122741697416122e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.206711083650589, + "step": 4030, + "valid_targets_mean": 2190.1, + "valid_targets_min": 771 + }, + { + "epoch": 3.9442815249266863, + "grad_norm": 0.6272342072403345, + "learning_rate": 1.9074039059336654e-05, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19984391331672668, + "step": 4035, + "valid_targets_mean": 2838.1, + "valid_targets_min": 614 + }, + { + "epoch": 3.949169110459433, + "grad_norm": 0.6758572832963811, + "learning_rate": 1.902534192326423e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21952173113822937, + "step": 4040, + "valid_targets_mean": 2742.6, + "valid_targets_min": 1552 + }, + { + "epoch": 3.95405669599218, + "grad_norm": 0.6455921997950347, + "learning_rate": 1.8976650578554475e-05, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17914989590644836, + "step": 4045, + "valid_targets_mean": 2462.2, + "valid_targets_min": 1012 + }, + { + "epoch": 3.9589442815249267, + "grad_norm": 0.7064643550905159, + "learning_rate": 1.892796531452857e-05, + "loss": 0.2141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22437229752540588, + "step": 4050, + "valid_targets_mean": 2798.8, + "valid_targets_min": 572 + }, + { + "epoch": 3.9638318670576735, + "grad_norm": 0.7269435964183898, + "learning_rate": 1.8879286420471562e-05, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22988593578338623, + "step": 4055, + "valid_targets_mean": 2328.2, + "valid_targets_min": 584 + }, + { + "epoch": 3.9687194525904204, + "grad_norm": 0.6243634215768921, + "learning_rate": 1.8830614185630686e-05, + "loss": 0.2017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19760319590568542, + "step": 4060, + "valid_targets_mean": 3104.0, + "valid_targets_min": 665 + }, + { + "epoch": 3.973607038123167, + "grad_norm": 0.7229402835170321, + "learning_rate": 1.8781948899213563e-05, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19763663411140442, + "step": 4065, + "valid_targets_mean": 2293.1, + "valid_targets_min": 1080 + }, + { + "epoch": 3.978494623655914, + "grad_norm": 0.7692945667783568, + "learning_rate": 1.8733290850386556e-05, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2350245714187622, + "step": 4070, + "valid_targets_mean": 2153.9, + "valid_targets_min": 522 + }, + { + "epoch": 3.983382209188661, + "grad_norm": 0.7485634972600307, + "learning_rate": 1.8684640328273013e-05, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2132551074028015, + "step": 4075, + "valid_targets_mean": 2146.8, + "valid_targets_min": 533 + }, + { + "epoch": 3.9882697947214076, + "grad_norm": 0.7868285556946475, + "learning_rate": 1.8635997621951554e-05, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21789923310279846, + "step": 4080, + "valid_targets_mean": 1982.0, + "valid_targets_min": 1146 + }, + { + "epoch": 3.9931573802541545, + "grad_norm": 0.6520687750578001, + "learning_rate": 1.858736302045438e-05, + "loss": 0.2106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2020658552646637, + "step": 4085, + "valid_targets_mean": 2626.5, + "valid_targets_min": 1214 + }, + { + "epoch": 3.9980449657869013, + "grad_norm": 0.6705009639669899, + "learning_rate": 1.8538736812765495e-05, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2510734796524048, + "step": 4090, + "valid_targets_mean": 2900.7, + "valid_targets_min": 823 + }, + { + "epoch": 4.002932551319648, + "grad_norm": 0.6983984686970028, + "learning_rate": 1.8490119287819058e-05, + "loss": 0.1852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1645195633172989, + "step": 4095, + "valid_targets_mean": 2202.6, + "valid_targets_min": 706 + }, + { + "epoch": 4.007820136852395, + "grad_norm": 0.7590415869647605, + "learning_rate": 1.844151073449763e-05, + "loss": 0.1727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1695270985364914, + "step": 4100, + "valid_targets_mean": 2045.4, + "valid_targets_min": 484 + }, + { + "epoch": 4.012707722385142, + "grad_norm": 0.8500063167328339, + "learning_rate": 1.8392911441630458e-05, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1960211545228958, + "step": 4105, + "valid_targets_mean": 1932.2, + "valid_targets_min": 886 + }, + { + "epoch": 4.0175953079178885, + "grad_norm": 0.6876315958843118, + "learning_rate": 1.8344321697991775e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17042392492294312, + "step": 4110, + "valid_targets_mean": 2488.9, + "valid_targets_min": 984 + }, + { + "epoch": 4.022482893450635, + "grad_norm": 0.6267560608434691, + "learning_rate": 1.8295741792299054e-05, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1570637822151184, + "step": 4115, + "valid_targets_mean": 2911.5, + "valid_targets_min": 998 + }, + { + "epoch": 4.027370478983382, + "grad_norm": 0.6860515522522642, + "learning_rate": 1.8247172013211316e-05, + "loss": 0.17, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17035432159900665, + "step": 4120, + "valid_targets_mean": 2804.3, + "valid_targets_min": 1062 + }, + { + "epoch": 4.032258064516129, + "grad_norm": 0.658990742688263, + "learning_rate": 1.819861264932743e-05, + "loss": 0.1688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1729648858308792, + "step": 4125, + "valid_targets_mean": 3242.4, + "valid_targets_min": 952 + }, + { + "epoch": 4.037145650048876, + "grad_norm": 0.8278627891112406, + "learning_rate": 1.815006398918435e-05, + "loss": 0.1708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17281252145767212, + "step": 4130, + "valid_targets_mean": 1954.6, + "valid_targets_min": 746 + }, + { + "epoch": 4.042033235581623, + "grad_norm": 0.7070945916212567, + "learning_rate": 1.8101526321255464e-05, + "loss": 0.1886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18609067797660828, + "step": 4135, + "valid_targets_mean": 2809.1, + "valid_targets_min": 725 + }, + { + "epoch": 4.0469208211143695, + "grad_norm": 0.732269604211766, + "learning_rate": 1.805299993394881e-05, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1601986289024353, + "step": 4140, + "valid_targets_mean": 2153.6, + "valid_targets_min": 895 + }, + { + "epoch": 4.051808406647116, + "grad_norm": 0.7644694368274649, + "learning_rate": 1.8004485115605407e-05, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17084315419197083, + "step": 4145, + "valid_targets_mean": 2271.1, + "valid_targets_min": 900 + }, + { + "epoch": 4.056695992179863, + "grad_norm": 0.680277205771116, + "learning_rate": 1.7955982154497545e-05, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18231913447380066, + "step": 4150, + "valid_targets_mean": 2801.8, + "valid_targets_min": 1152 + }, + { + "epoch": 4.06158357771261, + "grad_norm": 0.7619020216747033, + "learning_rate": 1.7907491338827053e-05, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17396214604377747, + "step": 4155, + "valid_targets_mean": 2283.8, + "valid_targets_min": 649 + }, + { + "epoch": 4.066471163245357, + "grad_norm": 0.7954884604214785, + "learning_rate": 1.7859012956723577e-05, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17803660035133362, + "step": 4160, + "valid_targets_mean": 2164.9, + "valid_targets_min": 784 + }, + { + "epoch": 4.0713587487781036, + "grad_norm": 0.70348085226897, + "learning_rate": 1.781054729624291e-05, + "loss": 0.1718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16010019183158875, + "step": 4165, + "valid_targets_mean": 2553.6, + "valid_targets_min": 843 + }, + { + "epoch": 4.07624633431085, + "grad_norm": 1.026152425313603, + "learning_rate": 1.7762094645365225e-05, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18153509497642517, + "step": 4170, + "valid_targets_mean": 2432.5, + "valid_targets_min": 615 + }, + { + "epoch": 4.081133919843597, + "grad_norm": 0.7254831762036433, + "learning_rate": 1.7713655291993426e-05, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20568543672561646, + "step": 4175, + "valid_targets_mean": 2840.9, + "valid_targets_min": 881 + }, + { + "epoch": 4.086021505376344, + "grad_norm": 0.697671859067687, + "learning_rate": 1.766522952395138e-05, + "loss": 0.1728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19269169867038727, + "step": 4180, + "valid_targets_mean": 3086.3, + "valid_targets_min": 746 + }, + { + "epoch": 4.090909090909091, + "grad_norm": 0.7479636389963256, + "learning_rate": 1.7616817628982225e-05, + "loss": 0.1604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16542398929595947, + "step": 4185, + "valid_targets_mean": 2449.9, + "valid_targets_min": 1025 + }, + { + "epoch": 4.095796676441838, + "grad_norm": 0.8012001808352684, + "learning_rate": 1.7568419894746694e-05, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15177610516548157, + "step": 4190, + "valid_targets_mean": 1801.7, + "valid_targets_min": 567 + }, + { + "epoch": 4.1006842619745845, + "grad_norm": 0.6599199969261322, + "learning_rate": 1.7520036608821343e-05, + "loss": 0.1764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16316524147987366, + "step": 4195, + "valid_targets_mean": 2802.2, + "valid_targets_min": 607 + }, + { + "epoch": 4.105571847507331, + "grad_norm": 0.660475255360126, + "learning_rate": 1.7471668058696917e-05, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16640618443489075, + "step": 4200, + "valid_targets_mean": 2976.4, + "valid_targets_min": 821 + }, + { + "epoch": 4.110459433040078, + "grad_norm": 0.7767208027973069, + "learning_rate": 1.7423314531776566e-05, + "loss": 0.1769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16882777214050293, + "step": 4205, + "valid_targets_mean": 2222.6, + "valid_targets_min": 627 + }, + { + "epoch": 4.115347018572825, + "grad_norm": 0.6976329404563815, + "learning_rate": 1.737497631537418e-05, + "loss": 0.172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17232577502727509, + "step": 4210, + "valid_targets_mean": 2611.0, + "valid_targets_min": 869 + }, + { + "epoch": 4.120234604105572, + "grad_norm": 0.8087989717723046, + "learning_rate": 1.7326653696712706e-05, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1771789789199829, + "step": 4215, + "valid_targets_mean": 2261.4, + "valid_targets_min": 551 + }, + { + "epoch": 4.125122189638319, + "grad_norm": 0.7223726297736563, + "learning_rate": 1.7278346962922365e-05, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17196297645568848, + "step": 4220, + "valid_targets_mean": 2553.8, + "valid_targets_min": 791 + }, + { + "epoch": 4.130009775171065, + "grad_norm": 0.8038120785795847, + "learning_rate": 1.7230056401039014e-05, + "loss": 0.1663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1827443540096283, + "step": 4225, + "valid_targets_mean": 1950.0, + "valid_targets_min": 744 + }, + { + "epoch": 4.134897360703812, + "grad_norm": 0.8507666081146841, + "learning_rate": 1.7181782298002426e-05, + "loss": 0.1919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1930236965417862, + "step": 4230, + "valid_targets_mean": 1955.8, + "valid_targets_min": 837 + }, + { + "epoch": 4.139784946236559, + "grad_norm": 0.72829350980729, + "learning_rate": 1.713352494065455e-05, + "loss": 0.1802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1770409643650055, + "step": 4235, + "valid_targets_mean": 2629.4, + "valid_targets_min": 728 + }, + { + "epoch": 4.144672531769306, + "grad_norm": 0.8739112255295849, + "learning_rate": 1.708528461573787e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17912176251411438, + "step": 4240, + "valid_targets_mean": 1731.5, + "valid_targets_min": 829 + }, + { + "epoch": 4.149560117302053, + "grad_norm": 0.6736533471138151, + "learning_rate": 1.703706160989362e-05, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17516711354255676, + "step": 4245, + "valid_targets_mean": 3023.8, + "valid_targets_min": 726 + }, + { + "epoch": 4.1544477028347995, + "grad_norm": 0.7146049815510215, + "learning_rate": 1.6988856209660162e-05, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18740811944007874, + "step": 4250, + "valid_targets_mean": 3117.8, + "valid_targets_min": 895 + }, + { + "epoch": 4.159335288367546, + "grad_norm": 0.8287505524072367, + "learning_rate": 1.694066870147123e-05, + "loss": 0.165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17309562861919403, + "step": 4255, + "valid_targets_mean": 2429.8, + "valid_targets_min": 613 + }, + { + "epoch": 4.164222873900293, + "grad_norm": 0.8384682014231488, + "learning_rate": 1.689249937165425e-05, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17695944011211395, + "step": 4260, + "valid_targets_mean": 1893.6, + "valid_targets_min": 622 + }, + { + "epoch": 4.16911045943304, + "grad_norm": 0.8023347194519758, + "learning_rate": 1.6844348506428624e-05, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1862003207206726, + "step": 4265, + "valid_targets_mean": 2379.9, + "valid_targets_min": 996 + }, + { + "epoch": 4.173998044965787, + "grad_norm": 0.7189754762701044, + "learning_rate": 1.679621639190405e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18278726935386658, + "step": 4270, + "valid_targets_mean": 2820.7, + "valid_targets_min": 981 + }, + { + "epoch": 4.178885630498534, + "grad_norm": 0.7689815704124234, + "learning_rate": 1.6748103314078795e-05, + "loss": 0.1737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15576958656311035, + "step": 4275, + "valid_targets_mean": 2550.9, + "valid_targets_min": 830 + }, + { + "epoch": 4.18377321603128, + "grad_norm": 0.7790069715413338, + "learning_rate": 1.6700009558838034e-05, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1859605610370636, + "step": 4280, + "valid_targets_mean": 2325.7, + "valid_targets_min": 765 + }, + { + "epoch": 4.188660801564027, + "grad_norm": 0.7059772286515155, + "learning_rate": 1.6651935411952117e-05, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18116068840026855, + "step": 4285, + "valid_targets_mean": 2637.1, + "valid_targets_min": 981 + }, + { + "epoch": 4.193548387096774, + "grad_norm": 0.8478026456180225, + "learning_rate": 1.6603881159074864e-05, + "loss": 0.1696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1635878086090088, + "step": 4290, + "valid_targets_mean": 2021.2, + "valid_targets_min": 728 + }, + { + "epoch": 4.198435972629521, + "grad_norm": 0.7892671310342947, + "learning_rate": 1.655584708574192e-05, + "loss": 0.1728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16756808757781982, + "step": 4295, + "valid_targets_mean": 2019.6, + "valid_targets_min": 694 + }, + { + "epoch": 4.203323558162268, + "grad_norm": 0.7866619457000387, + "learning_rate": 1.6507833477368994e-05, + "loss": 0.1628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1662808358669281, + "step": 4300, + "valid_targets_mean": 2019.4, + "valid_targets_min": 824 + }, + { + "epoch": 4.2082111436950145, + "grad_norm": 0.8459519999111245, + "learning_rate": 1.6459840619250217e-05, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1930123269557953, + "step": 4305, + "valid_targets_mean": 2161.1, + "valid_targets_min": 817 + }, + { + "epoch": 4.213098729227761, + "grad_norm": 0.8021367267841367, + "learning_rate": 1.6411868796556408e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18006908893585205, + "step": 4310, + "valid_targets_mean": 2118.6, + "valid_targets_min": 572 + }, + { + "epoch": 4.217986314760508, + "grad_norm": 0.8090359446901784, + "learning_rate": 1.6363918294333395e-05, + "loss": 0.162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15488529205322266, + "step": 4315, + "valid_targets_mean": 1977.8, + "valid_targets_min": 727 + }, + { + "epoch": 4.222873900293255, + "grad_norm": 0.6855591904703289, + "learning_rate": 1.6315989397500332e-05, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16280516982078552, + "step": 4320, + "valid_targets_mean": 3256.6, + "valid_targets_min": 958 + }, + { + "epoch": 4.227761485826002, + "grad_norm": 0.809105744787128, + "learning_rate": 1.6268082390847983e-05, + "loss": 0.1685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15723910927772522, + "step": 4325, + "valid_targets_mean": 1953.2, + "valid_targets_min": 746 + }, + { + "epoch": 4.232649071358749, + "grad_norm": 0.7131155432880343, + "learning_rate": 1.6220197559037045e-05, + "loss": 0.1744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18179753422737122, + "step": 4330, + "valid_targets_mean": 2626.1, + "valid_targets_min": 1086 + }, + { + "epoch": 4.237536656891495, + "grad_norm": 0.7455797655769655, + "learning_rate": 1.617233518659647e-05, + "loss": 0.1711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16530537605285645, + "step": 4335, + "valid_targets_mean": 2392.8, + "valid_targets_min": 770 + }, + { + "epoch": 4.242424242424242, + "grad_norm": 0.7439354451578698, + "learning_rate": 1.6124495557921725e-05, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.173561230301857, + "step": 4340, + "valid_targets_mean": 2214.4, + "valid_targets_min": 671 + }, + { + "epoch": 4.247311827956989, + "grad_norm": 0.7373073806467637, + "learning_rate": 1.6076678957273162e-05, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17753678560256958, + "step": 4345, + "valid_targets_mean": 2657.6, + "valid_targets_min": 734 + }, + { + "epoch": 4.252199413489736, + "grad_norm": 0.7000837897806548, + "learning_rate": 1.602888566877429e-05, + "loss": 0.1781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17242294549942017, + "step": 4350, + "valid_targets_mean": 2792.0, + "valid_targets_min": 770 + }, + { + "epoch": 4.257086999022483, + "grad_norm": 0.7271682426991912, + "learning_rate": 1.59811159764101e-05, + "loss": 0.1721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16193613409996033, + "step": 4355, + "valid_targets_mean": 2416.5, + "valid_targets_min": 907 + }, + { + "epoch": 4.2619745845552295, + "grad_norm": 0.7944206346753406, + "learning_rate": 1.593337016402539e-05, + "loss": 0.1729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17991730570793152, + "step": 4360, + "valid_targets_mean": 2064.8, + "valid_targets_min": 655 + }, + { + "epoch": 4.266862170087976, + "grad_norm": 0.7010001037382477, + "learning_rate": 1.5885648515323028e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16409370303153992, + "step": 4365, + "valid_targets_mean": 2635.2, + "valid_targets_min": 930 + }, + { + "epoch": 4.271749755620723, + "grad_norm": 0.6568280133407176, + "learning_rate": 1.5837951313862352e-05, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18202292919158936, + "step": 4370, + "valid_targets_mean": 3231.6, + "valid_targets_min": 671 + }, + { + "epoch": 4.27663734115347, + "grad_norm": 0.8085208993831905, + "learning_rate": 1.5790278843057395e-05, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18638502061367035, + "step": 4375, + "valid_targets_mean": 2588.2, + "valid_targets_min": 665 + }, + { + "epoch": 4.281524926686217, + "grad_norm": 0.8087448136164485, + "learning_rate": 1.5742631386175255e-05, + "loss": 0.1752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17984755337238312, + "step": 4380, + "valid_targets_mean": 2140.7, + "valid_targets_min": 701 + }, + { + "epoch": 4.286412512218964, + "grad_norm": 0.7073051473557163, + "learning_rate": 1.569500922633442e-05, + "loss": 0.169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16911262273788452, + "step": 4385, + "valid_targets_mean": 2626.2, + "valid_targets_min": 694 + }, + { + "epoch": 4.29130009775171, + "grad_norm": 0.7422208850214889, + "learning_rate": 1.5647412646503025e-05, + "loss": 0.1685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17553281784057617, + "step": 4390, + "valid_targets_mean": 2183.2, + "valid_targets_min": 1272 + }, + { + "epoch": 4.296187683284457, + "grad_norm": 0.7592007757916336, + "learning_rate": 1.5599841929497234e-05, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17773813009262085, + "step": 4395, + "valid_targets_mean": 2256.4, + "valid_targets_min": 771 + }, + { + "epoch": 4.301075268817204, + "grad_norm": 0.7553316641615674, + "learning_rate": 1.555229735797954e-05, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18115384876728058, + "step": 4400, + "valid_targets_mean": 2220.1, + "valid_targets_min": 733 + }, + { + "epoch": 4.305962854349951, + "grad_norm": 0.7811092431468514, + "learning_rate": 1.5504779214457064e-05, + "loss": 0.1695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17034941911697388, + "step": 4405, + "valid_targets_mean": 1985.4, + "valid_targets_min": 872 + }, + { + "epoch": 4.310850439882698, + "grad_norm": 0.6751967572463863, + "learning_rate": 1.5457287781279914e-05, + "loss": 0.1739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1524028480052948, + "step": 4410, + "valid_targets_mean": 2459.4, + "valid_targets_min": 768 + }, + { + "epoch": 4.3157380254154445, + "grad_norm": 0.7006365562513612, + "learning_rate": 1.5409823340639472e-05, + "loss": 0.174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1530303657054901, + "step": 4415, + "valid_targets_mean": 2240.3, + "valid_targets_min": 809 + }, + { + "epoch": 4.320625610948191, + "grad_norm": 0.8127996699553531, + "learning_rate": 1.5362386174566727e-05, + "loss": 0.1674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1497165858745575, + "step": 4420, + "valid_targets_mean": 1762.8, + "valid_targets_min": 672 + }, + { + "epoch": 4.325513196480938, + "grad_norm": 0.7509544790056238, + "learning_rate": 1.5314976564930616e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.215011328458786, + "step": 4425, + "valid_targets_mean": 2781.2, + "valid_targets_min": 870 + }, + { + "epoch": 4.330400782013685, + "grad_norm": 0.7344537959086179, + "learning_rate": 1.526759479343634e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1870063692331314, + "step": 4430, + "valid_targets_mean": 2725.8, + "valid_targets_min": 1240 + }, + { + "epoch": 4.335288367546432, + "grad_norm": 0.672200316438301, + "learning_rate": 1.5220241141623666e-05, + "loss": 0.1661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1690032184123993, + "step": 4435, + "valid_targets_mean": 3031.5, + "valid_targets_min": 1266 + }, + { + "epoch": 4.340175953079179, + "grad_norm": 0.6666959910156269, + "learning_rate": 1.517291589086531e-05, + "loss": 0.1752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15352541208267212, + "step": 4440, + "valid_targets_mean": 2541.2, + "valid_targets_min": 749 + }, + { + "epoch": 4.345063538611925, + "grad_norm": 0.7100554914799727, + "learning_rate": 1.512561932236519e-05, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16762804985046387, + "step": 4445, + "valid_targets_mean": 2549.6, + "valid_targets_min": 978 + }, + { + "epoch": 4.349951124144672, + "grad_norm": 0.7579351582413887, + "learning_rate": 1.507835171715683e-05, + "loss": 0.1772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19213947653770447, + "step": 4450, + "valid_targets_mean": 2302.4, + "valid_targets_min": 893 + }, + { + "epoch": 4.354838709677419, + "grad_norm": 0.7076551024584983, + "learning_rate": 1.503111335610164e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16405513882637024, + "step": 4455, + "valid_targets_mean": 2656.4, + "valid_targets_min": 662 + }, + { + "epoch": 4.359726295210166, + "grad_norm": 0.6577102500804493, + "learning_rate": 1.4983904519887257e-05, + "loss": 0.1773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18623942136764526, + "step": 4460, + "valid_targets_mean": 3397.5, + "valid_targets_min": 869 + }, + { + "epoch": 4.364613880742913, + "grad_norm": 0.8364738050383415, + "learning_rate": 1.4936725489025907e-05, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18252325057983398, + "step": 4465, + "valid_targets_mean": 2765.1, + "valid_targets_min": 740 + }, + { + "epoch": 4.3695014662756595, + "grad_norm": 0.7332207981929358, + "learning_rate": 1.4889576543852679e-05, + "loss": 0.1773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18655236065387726, + "step": 4470, + "valid_targets_mean": 2426.7, + "valid_targets_min": 759 + }, + { + "epoch": 4.374389051808406, + "grad_norm": 0.7291042324534501, + "learning_rate": 1.4842457964523932e-05, + "loss": 0.1743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1768685281276703, + "step": 4475, + "valid_targets_mean": 2585.4, + "valid_targets_min": 776 + }, + { + "epoch": 4.379276637341153, + "grad_norm": 0.7241032428052671, + "learning_rate": 1.4795370031015562e-05, + "loss": 0.1686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15364965796470642, + "step": 4480, + "valid_targets_mean": 2301.1, + "valid_targets_min": 968 + }, + { + "epoch": 4.3841642228739, + "grad_norm": 0.8029472975208416, + "learning_rate": 1.4748313023121385e-05, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17943759262561798, + "step": 4485, + "valid_targets_mean": 2096.1, + "valid_targets_min": 698 + }, + { + "epoch": 4.389051808406647, + "grad_norm": 0.7738811474250595, + "learning_rate": 1.4701287220451463e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18864920735359192, + "step": 4490, + "valid_targets_mean": 2364.0, + "valid_targets_min": 761 + }, + { + "epoch": 4.393939393939394, + "grad_norm": 0.8073103577600892, + "learning_rate": 1.4654292902430424e-05, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1911054402589798, + "step": 4495, + "valid_targets_mean": 2996.8, + "valid_targets_min": 912 + }, + { + "epoch": 4.39882697947214, + "grad_norm": 0.7781332848101631, + "learning_rate": 1.4607330348295821e-05, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20065414905548096, + "step": 4500, + "valid_targets_mean": 2334.9, + "valid_targets_min": 620 + }, + { + "epoch": 4.403714565004887, + "grad_norm": 0.8238449326675337, + "learning_rate": 1.4560399837096478e-05, + "loss": 0.1629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1719534993171692, + "step": 4505, + "valid_targets_mean": 1942.7, + "valid_targets_min": 1107 + }, + { + "epoch": 4.408602150537634, + "grad_norm": 0.7909535037694309, + "learning_rate": 1.4513501647690804e-05, + "loss": 0.1696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15581440925598145, + "step": 4510, + "valid_targets_mean": 2048.7, + "valid_targets_min": 680 + }, + { + "epoch": 4.413489736070381, + "grad_norm": 0.667844663877056, + "learning_rate": 1.4466636058745179e-05, + "loss": 0.1826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16047389805316925, + "step": 4515, + "valid_targets_mean": 2790.1, + "valid_targets_min": 697 + }, + { + "epoch": 4.418377321603128, + "grad_norm": 0.7179178500193523, + "learning_rate": 1.4419803348732245e-05, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16572578251361847, + "step": 4520, + "valid_targets_mean": 2572.5, + "valid_targets_min": 923 + }, + { + "epoch": 4.4232649071358745, + "grad_norm": 0.7482497307829565, + "learning_rate": 1.4373003795929287e-05, + "loss": 0.1781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17924658954143524, + "step": 4525, + "valid_targets_mean": 2537.1, + "valid_targets_min": 1124 + }, + { + "epoch": 4.428152492668621, + "grad_norm": 0.7983711495940576, + "learning_rate": 1.4326237678416588e-05, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15910127758979797, + "step": 4530, + "valid_targets_mean": 2120.4, + "valid_targets_min": 865 + }, + { + "epoch": 4.433040078201368, + "grad_norm": 0.7219070798034924, + "learning_rate": 1.4279505274075743e-05, + "loss": 0.165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17049354314804077, + "step": 4535, + "valid_targets_mean": 2468.0, + "valid_targets_min": 1040 + }, + { + "epoch": 4.437927663734115, + "grad_norm": 0.6820006015545326, + "learning_rate": 1.423280686058804e-05, + "loss": 0.1751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17498445510864258, + "step": 4540, + "valid_targets_mean": 2878.1, + "valid_targets_min": 814 + }, + { + "epoch": 4.442815249266862, + "grad_norm": 0.7956975533298152, + "learning_rate": 1.4186142715432775e-05, + "loss": 0.1618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15380510687828064, + "step": 4545, + "valid_targets_mean": 1974.9, + "valid_targets_min": 788 + }, + { + "epoch": 4.447702834799609, + "grad_norm": 0.6688865610533883, + "learning_rate": 1.4139513115885635e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18294396996498108, + "step": 4550, + "valid_targets_mean": 3151.7, + "valid_targets_min": 1199 + }, + { + "epoch": 4.452590420332355, + "grad_norm": 0.7510655980124383, + "learning_rate": 1.4092918339017046e-05, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17913363873958588, + "step": 4555, + "valid_targets_mean": 2332.2, + "valid_targets_min": 875 + }, + { + "epoch": 4.457478005865102, + "grad_norm": 0.7147653866732739, + "learning_rate": 1.4046358661690515e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15671703219413757, + "step": 4560, + "valid_targets_mean": 2355.0, + "valid_targets_min": 648 + }, + { + "epoch": 4.462365591397849, + "grad_norm": 0.8146269315094486, + "learning_rate": 1.3999834360560969e-05, + "loss": 0.1769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18449649214744568, + "step": 4565, + "valid_targets_mean": 2228.4, + "valid_targets_min": 798 + }, + { + "epoch": 4.467253176930596, + "grad_norm": 0.8549012646003764, + "learning_rate": 1.3953345712073172e-05, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20259252190589905, + "step": 4570, + "valid_targets_mean": 1910.8, + "valid_targets_min": 698 + }, + { + "epoch": 4.472140762463344, + "grad_norm": 0.6930221111338045, + "learning_rate": 1.3906892992460004e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19013884663581848, + "step": 4575, + "valid_targets_mean": 2861.4, + "valid_targets_min": 1011 + }, + { + "epoch": 4.4770283479960895, + "grad_norm": 0.8136412327729154, + "learning_rate": 1.3860476477740883e-05, + "loss": 0.1672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1534624695777893, + "step": 4580, + "valid_targets_mean": 1744.6, + "valid_targets_min": 650 + }, + { + "epoch": 4.481915933528837, + "grad_norm": 0.7069363998150793, + "learning_rate": 1.3814096443720091e-05, + "loss": 0.1769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16891230642795563, + "step": 4585, + "valid_targets_mean": 2515.7, + "valid_targets_min": 696 + }, + { + "epoch": 4.486803519061583, + "grad_norm": 0.801344597413106, + "learning_rate": 1.3767753165985126e-05, + "loss": 0.1663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14025087654590607, + "step": 4590, + "valid_targets_mean": 1840.0, + "valid_targets_min": 818 + }, + { + "epoch": 4.491691104594331, + "grad_norm": 0.7759235567696053, + "learning_rate": 1.3721446919905121e-05, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16802477836608887, + "step": 4595, + "valid_targets_mean": 2318.0, + "valid_targets_min": 698 + }, + { + "epoch": 4.496578690127077, + "grad_norm": 0.7746563734745292, + "learning_rate": 1.3675177980629135e-05, + "loss": 0.161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17537719011306763, + "step": 4600, + "valid_targets_mean": 2127.3, + "valid_targets_min": 625 + }, + { + "epoch": 4.5014662756598245, + "grad_norm": 0.692029862169816, + "learning_rate": 1.3628946623084563e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1666170358657837, + "step": 4605, + "valid_targets_mean": 2804.1, + "valid_targets_min": 1352 + }, + { + "epoch": 4.50635386119257, + "grad_norm": 0.7747220358157921, + "learning_rate": 1.3582753121975502e-05, + "loss": 0.1754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16754350066184998, + "step": 4610, + "valid_targets_mean": 2106.8, + "valid_targets_min": 1041 + }, + { + "epoch": 4.511241446725318, + "grad_norm": 0.794400474753371, + "learning_rate": 1.3536597751781082e-05, + "loss": 0.1706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15418118238449097, + "step": 4615, + "valid_targets_mean": 2122.1, + "valid_targets_min": 623 + }, + { + "epoch": 4.516129032258064, + "grad_norm": 0.8630490069683876, + "learning_rate": 1.3490480786753901e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1692330241203308, + "step": 4620, + "valid_targets_mean": 1878.2, + "valid_targets_min": 864 + }, + { + "epoch": 4.521016617790812, + "grad_norm": 0.6949929703480787, + "learning_rate": 1.3444402500918314e-05, + "loss": 0.1747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19203145802021027, + "step": 4625, + "valid_targets_mean": 2929.6, + "valid_targets_min": 1109 + }, + { + "epoch": 4.525904203323558, + "grad_norm": 0.7212230561742061, + "learning_rate": 1.3398363168068871e-05, + "loss": 0.1655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1773679554462433, + "step": 4630, + "valid_targets_mean": 3018.6, + "valid_targets_min": 852 + }, + { + "epoch": 4.530791788856305, + "grad_norm": 0.7491901926461985, + "learning_rate": 1.3352363061768668e-05, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19477087259292603, + "step": 4635, + "valid_targets_mean": 2341.3, + "valid_targets_min": 548 + }, + { + "epoch": 4.535679374389051, + "grad_norm": 0.7218327291631365, + "learning_rate": 1.3306402455347693e-05, + "loss": 0.1749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15920904278755188, + "step": 4640, + "valid_targets_mean": 2138.7, + "valid_targets_min": 749 + }, + { + "epoch": 4.540566959921799, + "grad_norm": 0.7886843070822219, + "learning_rate": 1.3260481621901269e-05, + "loss": 0.1691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20281338691711426, + "step": 4645, + "valid_targets_mean": 2345.5, + "valid_targets_min": 700 + }, + { + "epoch": 4.545454545454545, + "grad_norm": 0.7254460885818595, + "learning_rate": 1.3214600834288351e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18300187587738037, + "step": 4650, + "valid_targets_mean": 2600.0, + "valid_targets_min": 818 + }, + { + "epoch": 4.550342130987293, + "grad_norm": 0.7303419088042947, + "learning_rate": 1.3168760365129958e-05, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1488429307937622, + "step": 4655, + "valid_targets_mean": 2088.1, + "valid_targets_min": 957 + }, + { + "epoch": 4.555229716520039, + "grad_norm": 0.8187576278617511, + "learning_rate": 1.3122960486807551e-05, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1666831523180008, + "step": 4660, + "valid_targets_mean": 2028.4, + "valid_targets_min": 734 + }, + { + "epoch": 4.560117302052786, + "grad_norm": 0.7689970727875466, + "learning_rate": 1.3077201471461396e-05, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19278375804424286, + "step": 4665, + "valid_targets_mean": 2721.1, + "valid_targets_min": 1050 + }, + { + "epoch": 4.565004887585533, + "grad_norm": 0.7136026943772088, + "learning_rate": 1.3031483590988932e-05, + "loss": 0.1832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18920962512493134, + "step": 4670, + "valid_targets_mean": 3069.2, + "valid_targets_min": 703 + }, + { + "epoch": 4.56989247311828, + "grad_norm": 0.6768328311866438, + "learning_rate": 1.2985807117043212e-05, + "loss": 0.1732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.170124351978302, + "step": 4675, + "valid_targets_mean": 3036.4, + "valid_targets_min": 750 + }, + { + "epoch": 4.574780058651027, + "grad_norm": 0.7065627099385119, + "learning_rate": 1.2940172321031228e-05, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17484335601329803, + "step": 4680, + "valid_targets_mean": 2361.8, + "valid_targets_min": 985 + }, + { + "epoch": 4.579667644183774, + "grad_norm": 0.7870866877498359, + "learning_rate": 1.2894579474112336e-05, + "loss": 0.1712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17841269075870514, + "step": 4685, + "valid_targets_mean": 2178.9, + "valid_targets_min": 673 + }, + { + "epoch": 4.58455522971652, + "grad_norm": 0.7992856974999708, + "learning_rate": 1.2849028847196642e-05, + "loss": 0.1832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18455563485622406, + "step": 4690, + "valid_targets_mean": 2120.1, + "valid_targets_min": 522 + }, + { + "epoch": 4.589442815249267, + "grad_norm": 0.7236365680919709, + "learning_rate": 1.2803520710943345e-05, + "loss": 0.1702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1834551990032196, + "step": 4695, + "valid_targets_mean": 2806.9, + "valid_targets_min": 683 + }, + { + "epoch": 4.594330400782014, + "grad_norm": 0.7328506697251148, + "learning_rate": 1.2758055335759214e-05, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16841992735862732, + "step": 4700, + "valid_targets_mean": 2405.8, + "valid_targets_min": 830 + }, + { + "epoch": 4.599217986314761, + "grad_norm": 0.7488482580131285, + "learning_rate": 1.27126329917969e-05, + "loss": 0.1752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20795467495918274, + "step": 4705, + "valid_targets_mean": 2736.6, + "valid_targets_min": 740 + }, + { + "epoch": 4.604105571847508, + "grad_norm": 0.701924815848274, + "learning_rate": 1.2667253948953392e-05, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16952826082706451, + "step": 4710, + "valid_targets_mean": 2894.4, + "valid_targets_min": 805 + }, + { + "epoch": 4.6089931573802545, + "grad_norm": 0.7638018479267547, + "learning_rate": 1.262191847686837e-05, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17446717619895935, + "step": 4715, + "valid_targets_mean": 2210.6, + "valid_targets_min": 1220 + }, + { + "epoch": 4.613880742913001, + "grad_norm": 0.9146808729960212, + "learning_rate": 1.2576626844922602e-05, + "loss": 0.1796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19167935848236084, + "step": 4720, + "valid_targets_mean": 2399.1, + "valid_targets_min": 743 + }, + { + "epoch": 4.618768328445748, + "grad_norm": 0.7745900925140338, + "learning_rate": 1.2531379322236403e-05, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17140185832977295, + "step": 4725, + "valid_targets_mean": 2393.5, + "valid_targets_min": 850 + }, + { + "epoch": 4.623655913978495, + "grad_norm": 0.776866808386388, + "learning_rate": 1.2486176177667963e-05, + "loss": 0.178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1710333526134491, + "step": 4730, + "valid_targets_mean": 2295.1, + "valid_targets_min": 635 + }, + { + "epoch": 4.628543499511242, + "grad_norm": 0.9138199933637996, + "learning_rate": 1.2441017679811777e-05, + "loss": 0.1711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17593233287334442, + "step": 4735, + "valid_targets_mean": 1879.2, + "valid_targets_min": 661 + }, + { + "epoch": 4.633431085043989, + "grad_norm": 0.7080266048721636, + "learning_rate": 1.2395904096997082e-05, + "loss": 0.1688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17759348452091217, + "step": 4740, + "valid_targets_mean": 2714.9, + "valid_targets_min": 715 + }, + { + "epoch": 4.638318670576735, + "grad_norm": 0.7107351831118239, + "learning_rate": 1.2350835697286185e-05, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1701585203409195, + "step": 4745, + "valid_targets_mean": 2465.2, + "valid_targets_min": 836 + }, + { + "epoch": 4.643206256109482, + "grad_norm": 0.7110148275686181, + "learning_rate": 1.2305812748472969e-05, + "loss": 0.1729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16965368390083313, + "step": 4750, + "valid_targets_mean": 2887.4, + "valid_targets_min": 1056 + }, + { + "epoch": 4.648093841642229, + "grad_norm": 0.7808205504531988, + "learning_rate": 1.2260835518081216e-05, + "loss": 0.179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1801660656929016, + "step": 4755, + "valid_targets_mean": 2264.4, + "valid_targets_min": 537 + }, + { + "epoch": 4.652981427174976, + "grad_norm": 0.6736174438322065, + "learning_rate": 1.2215904273363055e-05, + "loss": 0.1741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16034579277038574, + "step": 4760, + "valid_targets_mean": 2747.2, + "valid_targets_min": 878 + }, + { + "epoch": 4.657869012707723, + "grad_norm": 0.8172833092442401, + "learning_rate": 1.2171019281297397e-05, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18934208154678345, + "step": 4765, + "valid_targets_mean": 2011.7, + "valid_targets_min": 689 + }, + { + "epoch": 4.6627565982404695, + "grad_norm": 0.7810540506761577, + "learning_rate": 1.2126180808588281e-05, + "loss": 0.181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19872036576271057, + "step": 4770, + "valid_targets_mean": 2411.7, + "valid_targets_min": 858 + }, + { + "epoch": 4.667644183773216, + "grad_norm": 0.7414123362578164, + "learning_rate": 1.2081389121663361e-05, + "loss": 0.1841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19137006998062134, + "step": 4775, + "valid_targets_mean": 2429.5, + "valid_targets_min": 996 + }, + { + "epoch": 4.672531769305963, + "grad_norm": 0.7992138573951394, + "learning_rate": 1.2036644486672287e-05, + "loss": 0.1756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18999941647052765, + "step": 4780, + "valid_targets_mean": 2264.3, + "valid_targets_min": 800 + }, + { + "epoch": 4.67741935483871, + "grad_norm": 0.7629820940559762, + "learning_rate": 1.199194716948512e-05, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18513910472393036, + "step": 4785, + "valid_targets_mean": 2405.6, + "valid_targets_min": 548 + }, + { + "epoch": 4.682306940371457, + "grad_norm": 0.712753687411961, + "learning_rate": 1.1947297435690772e-05, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18980765342712402, + "step": 4790, + "valid_targets_mean": 2899.6, + "valid_targets_min": 916 + }, + { + "epoch": 4.687194525904204, + "grad_norm": 0.7285016261064938, + "learning_rate": 1.190269555059542e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17685630917549133, + "step": 4795, + "valid_targets_mean": 2543.8, + "valid_targets_min": 779 + }, + { + "epoch": 4.69208211143695, + "grad_norm": 0.7186549073765208, + "learning_rate": 1.1858141779220902e-05, + "loss": 0.1665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17428386211395264, + "step": 4800, + "valid_targets_mean": 2633.2, + "valid_targets_min": 781 + }, + { + "epoch": 4.696969696969697, + "grad_norm": 0.7081619736005064, + "learning_rate": 1.1813636386303196e-05, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1797652244567871, + "step": 4805, + "valid_targets_mean": 2718.8, + "valid_targets_min": 776 + }, + { + "epoch": 4.701857282502444, + "grad_norm": 0.8451059412061002, + "learning_rate": 1.1769179636290801e-05, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18409138917922974, + "step": 4810, + "valid_targets_mean": 2124.2, + "valid_targets_min": 847 + }, + { + "epoch": 4.706744868035191, + "grad_norm": 0.8202631061205068, + "learning_rate": 1.1724771793343192e-05, + "loss": 0.17, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17531917989253998, + "step": 4815, + "valid_targets_mean": 2052.2, + "valid_targets_min": 700 + }, + { + "epoch": 4.711632453567938, + "grad_norm": 0.6836107410471749, + "learning_rate": 1.1680413121329243e-05, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17793884873390198, + "step": 4820, + "valid_targets_mean": 3199.6, + "valid_targets_min": 1120 + }, + { + "epoch": 4.7165200391006845, + "grad_norm": 0.7484449202460228, + "learning_rate": 1.1636103883825636e-05, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1955634206533432, + "step": 4825, + "valid_targets_mean": 2796.8, + "valid_targets_min": 1038 + }, + { + "epoch": 4.721407624633431, + "grad_norm": 0.8765114876948064, + "learning_rate": 1.1591844344115346e-05, + "loss": 0.1737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16911789774894714, + "step": 4830, + "valid_targets_mean": 1774.6, + "valid_targets_min": 637 + }, + { + "epoch": 4.726295210166178, + "grad_norm": 0.730247866586368, + "learning_rate": 1.1547634765186016e-05, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16497325897216797, + "step": 4835, + "valid_targets_mean": 2376.2, + "valid_targets_min": 818 + }, + { + "epoch": 4.731182795698925, + "grad_norm": 0.8018221641880683, + "learning_rate": 1.1503475409728448e-05, + "loss": 0.1814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17386886477470398, + "step": 4840, + "valid_targets_mean": 2089.1, + "valid_targets_min": 1032 + }, + { + "epoch": 4.736070381231672, + "grad_norm": 0.7294276842415255, + "learning_rate": 1.1459366540135019e-05, + "loss": 0.1667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1633351594209671, + "step": 4845, + "valid_targets_mean": 2460.2, + "valid_targets_min": 870 + }, + { + "epoch": 4.740957966764419, + "grad_norm": 0.7012095976697436, + "learning_rate": 1.1415308418498104e-05, + "loss": 0.1773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1733730584383011, + "step": 4850, + "valid_targets_mean": 2616.0, + "valid_targets_min": 644 + }, + { + "epoch": 4.745845552297165, + "grad_norm": 0.6630187702111296, + "learning_rate": 1.137130130660855e-05, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16590818762779236, + "step": 4855, + "valid_targets_mean": 3160.9, + "valid_targets_min": 1147 + }, + { + "epoch": 4.750733137829912, + "grad_norm": 0.9413915512130215, + "learning_rate": 1.1327345465954109e-05, + "loss": 0.1788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1672978401184082, + "step": 4860, + "valid_targets_mean": 1737.2, + "valid_targets_min": 744 + }, + { + "epoch": 4.755620723362659, + "grad_norm": 0.7740192526311673, + "learning_rate": 1.128344115771787e-05, + "loss": 0.1653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14710134267807007, + "step": 4865, + "valid_targets_mean": 2162.9, + "valid_targets_min": 612 + }, + { + "epoch": 4.760508308895406, + "grad_norm": 0.7042101370968502, + "learning_rate": 1.1239588642776736e-05, + "loss": 0.178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1820552498102188, + "step": 4870, + "valid_targets_mean": 2700.5, + "valid_targets_min": 856 + }, + { + "epoch": 4.765395894428153, + "grad_norm": 0.846516777206178, + "learning_rate": 1.1195788181699829e-05, + "loss": 0.1713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16076403856277466, + "step": 4875, + "valid_targets_mean": 1912.6, + "valid_targets_min": 607 + }, + { + "epoch": 4.7702834799608995, + "grad_norm": 0.8117401631836421, + "learning_rate": 1.1152040034747005e-05, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19138172268867493, + "step": 4880, + "valid_targets_mean": 2053.1, + "valid_targets_min": 679 + }, + { + "epoch": 4.775171065493646, + "grad_norm": 0.8090125882585804, + "learning_rate": 1.1108344461867256e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18099914491176605, + "step": 4885, + "valid_targets_mean": 2086.2, + "valid_targets_min": 875 + }, + { + "epoch": 4.780058651026393, + "grad_norm": 0.7869706404872734, + "learning_rate": 1.1064701722697171e-05, + "loss": 0.1669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17875008285045624, + "step": 4890, + "valid_targets_mean": 2636.2, + "valid_targets_min": 789 + }, + { + "epoch": 4.78494623655914, + "grad_norm": 0.8338533349319971, + "learning_rate": 1.102111207655943e-05, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1614495813846588, + "step": 4895, + "valid_targets_mean": 2010.5, + "valid_targets_min": 1178 + }, + { + "epoch": 4.789833822091887, + "grad_norm": 0.8181674038081255, + "learning_rate": 1.0977575782461215e-05, + "loss": 0.169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16135358810424805, + "step": 4900, + "valid_targets_mean": 1836.6, + "valid_targets_min": 704 + }, + { + "epoch": 4.794721407624634, + "grad_norm": 0.7134765702982729, + "learning_rate": 1.0934093099092684e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18349812924861908, + "step": 4905, + "valid_targets_mean": 2678.9, + "valid_targets_min": 1270 + }, + { + "epoch": 4.79960899315738, + "grad_norm": 0.7321067232072485, + "learning_rate": 1.0890664284825488e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1972254067659378, + "step": 4910, + "valid_targets_mean": 2744.4, + "valid_targets_min": 991 + }, + { + "epoch": 4.804496578690127, + "grad_norm": 0.7586910224470423, + "learning_rate": 1.0847289597711144e-05, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1841929405927658, + "step": 4915, + "valid_targets_mean": 2455.1, + "valid_targets_min": 881 + }, + { + "epoch": 4.809384164222874, + "grad_norm": 0.7967660614928856, + "learning_rate": 1.0803969295479577e-05, + "loss": 0.1799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16611838340759277, + "step": 4920, + "valid_targets_mean": 1992.2, + "valid_targets_min": 699 + }, + { + "epoch": 4.814271749755621, + "grad_norm": 0.7051736092585311, + "learning_rate": 1.076070363553754e-05, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14659056067466736, + "step": 4925, + "valid_targets_mean": 2598.4, + "valid_targets_min": 904 + }, + { + "epoch": 4.819159335288368, + "grad_norm": 0.8074487953089629, + "learning_rate": 1.071749287496712e-05, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18871314823627472, + "step": 4930, + "valid_targets_mean": 2446.9, + "valid_targets_min": 917 + }, + { + "epoch": 4.8240469208211145, + "grad_norm": 0.7358191683646678, + "learning_rate": 1.0674337270524195e-05, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15642912685871124, + "step": 4935, + "valid_targets_mean": 2261.7, + "valid_targets_min": 659 + }, + { + "epoch": 4.828934506353861, + "grad_norm": 0.7772322735086833, + "learning_rate": 1.06312370786369e-05, + "loss": 0.1689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16359858214855194, + "step": 4940, + "valid_targets_mean": 2106.1, + "valid_targets_min": 706 + }, + { + "epoch": 4.833822091886608, + "grad_norm": 0.7984734484679462, + "learning_rate": 1.0588192555404106e-05, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1786036491394043, + "step": 4945, + "valid_targets_mean": 1923.9, + "valid_targets_min": 1242 + }, + { + "epoch": 4.838709677419355, + "grad_norm": 0.7509801685683009, + "learning_rate": 1.0545203956593922e-05, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16603411734104156, + "step": 4950, + "valid_targets_mean": 2316.3, + "valid_targets_min": 917 + }, + { + "epoch": 4.843597262952102, + "grad_norm": 0.6961656843543814, + "learning_rate": 1.0502271537642144e-05, + "loss": 0.1735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17675796151161194, + "step": 4955, + "valid_targets_mean": 2698.9, + "valid_targets_min": 825 + }, + { + "epoch": 4.848484848484849, + "grad_norm": 0.8051523733028211, + "learning_rate": 1.0459395553650767e-05, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18461185693740845, + "step": 4960, + "valid_targets_mean": 1971.4, + "valid_targets_min": 727 + }, + { + "epoch": 4.853372434017595, + "grad_norm": 0.8334702591345402, + "learning_rate": 1.041657625938643e-05, + "loss": 0.1942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19902193546295166, + "step": 4965, + "valid_targets_mean": 2607.7, + "valid_targets_min": 1300 + }, + { + "epoch": 4.858260019550342, + "grad_norm": 0.7868464417093073, + "learning_rate": 1.0373813909278926e-05, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20459884405136108, + "step": 4970, + "valid_targets_mean": 2319.3, + "valid_targets_min": 861 + }, + { + "epoch": 4.863147605083089, + "grad_norm": 0.757600709591517, + "learning_rate": 1.0331108757419715e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1869640350341797, + "step": 4975, + "valid_targets_mean": 2491.2, + "valid_targets_min": 567 + }, + { + "epoch": 4.868035190615836, + "grad_norm": 0.7221664526957879, + "learning_rate": 1.0288461057560361e-05, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17515313625335693, + "step": 4980, + "valid_targets_mean": 2669.4, + "valid_targets_min": 1520 + }, + { + "epoch": 4.872922776148583, + "grad_norm": 0.7579422903647014, + "learning_rate": 1.0245871063111077e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1622183918952942, + "step": 4985, + "valid_targets_mean": 2357.6, + "valid_targets_min": 912 + }, + { + "epoch": 4.8778103616813295, + "grad_norm": 0.8067790989051661, + "learning_rate": 1.0203339027139167e-05, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16633889079093933, + "step": 4990, + "valid_targets_mean": 1989.5, + "valid_targets_min": 632 + }, + { + "epoch": 4.882697947214076, + "grad_norm": 0.7758521837897394, + "learning_rate": 1.0160865202367556e-05, + "loss": 0.1769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1709386706352234, + "step": 4995, + "valid_targets_mean": 2203.2, + "valid_targets_min": 829 + }, + { + "epoch": 4.887585532746823, + "grad_norm": 0.8248140895342541, + "learning_rate": 1.0118449841173294e-05, + "loss": 0.1699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18402284383773804, + "step": 5000, + "valid_targets_mean": 2114.4, + "valid_targets_min": 720 + }, + { + "epoch": 4.89247311827957, + "grad_norm": 0.6897458976588339, + "learning_rate": 1.0076093195586043e-05, + "loss": 0.1735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16674119234085083, + "step": 5005, + "valid_targets_mean": 2625.1, + "valid_targets_min": 708 + }, + { + "epoch": 4.897360703812317, + "grad_norm": 0.6687153559969188, + "learning_rate": 1.0033795517286558e-05, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17733493447303772, + "step": 5010, + "valid_targets_mean": 2868.8, + "valid_targets_min": 922 + }, + { + "epoch": 4.902248289345064, + "grad_norm": 0.8576695321729364, + "learning_rate": 9.991557057605244e-06, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18968823552131653, + "step": 5015, + "valid_targets_mean": 1945.0, + "valid_targets_min": 704 + }, + { + "epoch": 4.9071358748778104, + "grad_norm": 0.639640416335708, + "learning_rate": 9.949378067520601e-06, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14256887137889862, + "step": 5020, + "valid_targets_mean": 2694.2, + "valid_targets_min": 656 + }, + { + "epoch": 4.912023460410557, + "grad_norm": 0.6732914788931557, + "learning_rate": 9.90725879765779e-06, + "loss": 0.1771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1720367670059204, + "step": 5025, + "valid_targets_mean": 3006.6, + "valid_targets_min": 1307 + }, + { + "epoch": 4.916911045943304, + "grad_norm": 0.7100316647560853, + "learning_rate": 9.865199498287118e-06, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16838201880455017, + "step": 5030, + "valid_targets_mean": 2566.2, + "valid_targets_min": 826 + }, + { + "epoch": 4.921798631476051, + "grad_norm": 0.7460391948002998, + "learning_rate": 9.823200419322523e-06, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16970403492450714, + "step": 5035, + "valid_targets_mean": 2189.6, + "valid_targets_min": 737 + }, + { + "epoch": 4.926686217008798, + "grad_norm": 0.7470863844930459, + "learning_rate": 9.781261810320151e-06, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17340749502182007, + "step": 5040, + "valid_targets_mean": 2428.0, + "valid_targets_min": 647 + }, + { + "epoch": 4.9315738025415445, + "grad_norm": 0.7125845379627102, + "learning_rate": 9.739383920476806e-06, + "loss": 0.1699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16167551279067993, + "step": 5045, + "valid_targets_mean": 2554.8, + "valid_targets_min": 544 + }, + { + "epoch": 4.936461388074291, + "grad_norm": 0.7572669890789546, + "learning_rate": 9.69756699862853e-06, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22088664770126343, + "step": 5050, + "valid_targets_mean": 2523.8, + "valid_targets_min": 944 + }, + { + "epoch": 4.941348973607038, + "grad_norm": 0.7472631938735368, + "learning_rate": 9.65581129324909e-06, + "loss": 0.176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17298045754432678, + "step": 5055, + "valid_targets_mean": 2148.9, + "valid_targets_min": 591 + }, + { + "epoch": 4.946236559139785, + "grad_norm": 0.7406653369418025, + "learning_rate": 9.614117052448488e-06, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12865734100341797, + "step": 5060, + "valid_targets_mean": 2030.6, + "valid_targets_min": 842 + }, + { + "epoch": 4.951124144672532, + "grad_norm": 0.7621843768822362, + "learning_rate": 9.572484523971544e-06, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17717863619327545, + "step": 5065, + "valid_targets_mean": 2430.3, + "valid_targets_min": 704 + }, + { + "epoch": 4.956011730205279, + "grad_norm": 0.7498657834356395, + "learning_rate": 9.530913955196344e-06, + "loss": 0.1754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18062527477741241, + "step": 5070, + "valid_targets_mean": 2419.0, + "valid_targets_min": 824 + }, + { + "epoch": 4.9608993157380255, + "grad_norm": 0.8141850556767372, + "learning_rate": 9.489405593132846e-06, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15304099023342133, + "step": 5075, + "valid_targets_mean": 2178.5, + "valid_targets_min": 780 + }, + { + "epoch": 4.965786901270772, + "grad_norm": 0.6380568455457596, + "learning_rate": 9.447959684421374e-06, + "loss": 0.1594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16023944318294525, + "step": 5080, + "valid_targets_mean": 3038.6, + "valid_targets_min": 1137 + }, + { + "epoch": 4.970674486803519, + "grad_norm": 0.7983753521526253, + "learning_rate": 9.40657647533114e-06, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18773408234119415, + "step": 5085, + "valid_targets_mean": 2392.6, + "valid_targets_min": 753 + }, + { + "epoch": 4.975562072336266, + "grad_norm": 0.7019559841529516, + "learning_rate": 9.365256211758819e-06, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18119125068187714, + "step": 5090, + "valid_targets_mean": 2740.1, + "valid_targets_min": 797 + }, + { + "epoch": 4.980449657869013, + "grad_norm": 0.8008073853070248, + "learning_rate": 9.323999139227051e-06, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18562883138656616, + "step": 5095, + "valid_targets_mean": 2072.8, + "valid_targets_min": 764 + }, + { + "epoch": 4.9853372434017595, + "grad_norm": 0.7313841730070882, + "learning_rate": 9.282805502883007e-06, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1831430047750473, + "step": 5100, + "valid_targets_mean": 2517.4, + "valid_targets_min": 754 + }, + { + "epoch": 4.990224828934506, + "grad_norm": 0.8078162384967679, + "learning_rate": 9.241675547496938e-06, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17006579041481018, + "step": 5105, + "valid_targets_mean": 2003.2, + "valid_targets_min": 714 + }, + { + "epoch": 4.995112414467253, + "grad_norm": 0.8091777933799764, + "learning_rate": 9.200609517460675e-06, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18514806032180786, + "step": 5110, + "valid_targets_mean": 2168.2, + "valid_targets_min": 680 + }, + { + "epoch": 5.0, + "grad_norm": 0.7870915330741979, + "learning_rate": 9.159607656786222e-06, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16797024011611938, + "step": 5115, + "valid_targets_mean": 2128.8, + "valid_targets_min": 887 + }, + { + "epoch": 5.004887585532747, + "grad_norm": 0.7747894970742039, + "learning_rate": 9.1186702091043e-06, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16766156256198883, + "step": 5120, + "valid_targets_mean": 2120.5, + "valid_targets_min": 675 + }, + { + "epoch": 5.009775171065494, + "grad_norm": 0.6475219132787013, + "learning_rate": 9.077797417662888e-06, + "loss": 0.1449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16814623773097992, + "step": 5125, + "valid_targets_mean": 3438.9, + "valid_targets_min": 621 + }, + { + "epoch": 5.0146627565982405, + "grad_norm": 0.6877209083156904, + "learning_rate": 9.036989525325788e-06, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.151102676987648, + "step": 5130, + "valid_targets_mean": 3217.8, + "valid_targets_min": 768 + }, + { + "epoch": 5.019550342130987, + "grad_norm": 0.7324938421863219, + "learning_rate": 8.996246774571159e-06, + "loss": 0.1452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12554004788398743, + "step": 5135, + "valid_targets_mean": 2168.3, + "valid_targets_min": 1037 + }, + { + "epoch": 5.024437927663734, + "grad_norm": 0.7753610298141373, + "learning_rate": 8.95556940749009e-06, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14286714792251587, + "step": 5140, + "valid_targets_mean": 2078.7, + "valid_targets_min": 668 + }, + { + "epoch": 5.029325513196481, + "grad_norm": 0.7395423291733766, + "learning_rate": 8.914957665785184e-06, + "loss": 0.1575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16619789600372314, + "step": 5145, + "valid_targets_mean": 2746.6, + "valid_targets_min": 986 + }, + { + "epoch": 5.034213098729228, + "grad_norm": 0.6439026975349222, + "learning_rate": 8.874411790769093e-06, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12593859434127808, + "step": 5150, + "valid_targets_mean": 2663.3, + "valid_targets_min": 1334 + }, + { + "epoch": 5.0391006842619745, + "grad_norm": 0.7324203667912003, + "learning_rate": 8.833932023363096e-06, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16145989298820496, + "step": 5155, + "valid_targets_mean": 2840.1, + "valid_targets_min": 1147 + }, + { + "epoch": 5.043988269794721, + "grad_norm": 1.2076159147090062, + "learning_rate": 8.793518604095659e-06, + "loss": 0.1641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16184774041175842, + "step": 5160, + "valid_targets_mean": 2118.9, + "valid_targets_min": 985 + }, + { + "epoch": 5.048875855327468, + "grad_norm": 0.8370647289954567, + "learning_rate": 8.753171773101e-06, + "loss": 0.1527, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15503370761871338, + "step": 5165, + "valid_targets_mean": 2071.8, + "valid_targets_min": 741 + }, + { + "epoch": 5.053763440860215, + "grad_norm": 1.3707229780441046, + "learning_rate": 8.712891770117693e-06, + "loss": 0.1583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1608889102935791, + "step": 5170, + "valid_targets_mean": 2661.7, + "valid_targets_min": 807 + }, + { + "epoch": 5.058651026392962, + "grad_norm": 0.8818514323414521, + "learning_rate": 8.672678834487228e-06, + "loss": 0.1424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1271866112947464, + "step": 5175, + "valid_targets_mean": 1782.8, + "valid_targets_min": 933 + }, + { + "epoch": 5.063538611925709, + "grad_norm": 0.7996369140231598, + "learning_rate": 8.632533205152553e-06, + "loss": 0.1491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14621879160404205, + "step": 5180, + "valid_targets_mean": 2618.7, + "valid_targets_min": 771 + }, + { + "epoch": 5.0684261974584555, + "grad_norm": 0.7632406562623092, + "learning_rate": 8.592455120656726e-06, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15845827758312225, + "step": 5185, + "valid_targets_mean": 2956.1, + "valid_targets_min": 636 + }, + { + "epoch": 5.073313782991202, + "grad_norm": 0.8154709547280357, + "learning_rate": 8.552444819141413e-06, + "loss": 0.1533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15961962938308716, + "step": 5190, + "valid_targets_mean": 2535.6, + "valid_targets_min": 727 + }, + { + "epoch": 5.078201368523949, + "grad_norm": 0.7508882761061626, + "learning_rate": 8.512502538345568e-06, + "loss": 0.1534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13347077369689941, + "step": 5195, + "valid_targets_mean": 2393.2, + "valid_targets_min": 625 + }, + { + "epoch": 5.083088954056696, + "grad_norm": 0.7503877902892744, + "learning_rate": 8.472628515603936e-06, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1839996576309204, + "step": 5200, + "valid_targets_mean": 3139.9, + "valid_targets_min": 1167 + }, + { + "epoch": 5.087976539589443, + "grad_norm": 0.814409890229622, + "learning_rate": 8.43282298784567e-06, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16277194023132324, + "step": 5205, + "valid_targets_mean": 2501.2, + "valid_targets_min": 949 + }, + { + "epoch": 5.0928641251221896, + "grad_norm": 0.7086407404966241, + "learning_rate": 8.393086191592963e-06, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15189936757087708, + "step": 5210, + "valid_targets_mean": 3098.6, + "valid_targets_min": 1159 + }, + { + "epoch": 5.097751710654936, + "grad_norm": 0.8089395622560438, + "learning_rate": 8.353418362959564e-06, + "loss": 0.1541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1489260196685791, + "step": 5215, + "valid_targets_mean": 2491.1, + "valid_targets_min": 644 + }, + { + "epoch": 5.102639296187683, + "grad_norm": 0.8056121720254122, + "learning_rate": 8.313819737649483e-06, + "loss": 0.1679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15654730796813965, + "step": 5220, + "valid_targets_mean": 2591.2, + "valid_targets_min": 801 + }, + { + "epoch": 5.10752688172043, + "grad_norm": 0.8044880648557395, + "learning_rate": 8.274290550955468e-06, + "loss": 0.1521, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13206225633621216, + "step": 5225, + "valid_targets_mean": 2005.5, + "valid_targets_min": 1094 + }, + { + "epoch": 5.112414467253177, + "grad_norm": 0.7613818292118435, + "learning_rate": 8.234831037757687e-06, + "loss": 0.1502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17657440900802612, + "step": 5230, + "valid_targets_mean": 2638.0, + "valid_targets_min": 765 + }, + { + "epoch": 5.117302052785924, + "grad_norm": 0.6877071433658863, + "learning_rate": 8.195441432522324e-06, + "loss": 0.1448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1490596979856491, + "step": 5235, + "valid_targets_mean": 3027.0, + "valid_targets_min": 874 + }, + { + "epoch": 5.1221896383186705, + "grad_norm": 0.7403615217739022, + "learning_rate": 8.156121969300168e-06, + "loss": 0.1435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1425008326768875, + "step": 5240, + "valid_targets_mean": 2339.8, + "valid_targets_min": 659 + }, + { + "epoch": 5.127077223851417, + "grad_norm": 0.7526140740334464, + "learning_rate": 8.116872881725213e-06, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1776733696460724, + "step": 5245, + "valid_targets_mean": 3384.4, + "valid_targets_min": 1031 + }, + { + "epoch": 5.131964809384164, + "grad_norm": 0.7793267157176477, + "learning_rate": 8.077694403013312e-06, + "loss": 0.147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1542745679616928, + "step": 5250, + "valid_targets_mean": 2364.7, + "valid_targets_min": 712 + }, + { + "epoch": 5.136852394916911, + "grad_norm": 0.7971711756751064, + "learning_rate": 8.038586765960735e-06, + "loss": 0.1555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1412791758775711, + "step": 5255, + "valid_targets_mean": 1963.4, + "valid_targets_min": 641 + }, + { + "epoch": 5.141739980449658, + "grad_norm": 0.7033271354224208, + "learning_rate": 7.999550202942833e-06, + "loss": 0.1653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16525672376155853, + "step": 5260, + "valid_targets_mean": 3219.7, + "valid_targets_min": 809 + }, + { + "epoch": 5.146627565982405, + "grad_norm": 0.7247341620856336, + "learning_rate": 7.960584945912649e-06, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16547167301177979, + "step": 5265, + "valid_targets_mean": 2777.9, + "valid_targets_min": 744 + }, + { + "epoch": 5.151515151515151, + "grad_norm": 0.8520459168069531, + "learning_rate": 7.921691226399492e-06, + "loss": 0.1575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15888334810733795, + "step": 5270, + "valid_targets_mean": 2529.7, + "valid_targets_min": 816 + }, + { + "epoch": 5.156402737047898, + "grad_norm": 0.7059577585280569, + "learning_rate": 7.882869275507648e-06, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13361497223377228, + "step": 5275, + "valid_targets_mean": 2845.0, + "valid_targets_min": 704 + }, + { + "epoch": 5.161290322580645, + "grad_norm": 0.6859160374477156, + "learning_rate": 7.844119323914918e-06, + "loss": 0.1521, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14880457520484924, + "step": 5280, + "valid_targets_mean": 3074.0, + "valid_targets_min": 764 + }, + { + "epoch": 5.166177908113392, + "grad_norm": 0.6951765797767238, + "learning_rate": 7.805441601871292e-06, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1490635871887207, + "step": 5285, + "valid_targets_mean": 3103.2, + "valid_targets_min": 1018 + }, + { + "epoch": 5.171065493646139, + "grad_norm": 0.7792674902926341, + "learning_rate": 7.766836339197617e-06, + "loss": 0.155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1745825558900833, + "step": 5290, + "valid_targets_mean": 2414.8, + "valid_targets_min": 972 + }, + { + "epoch": 5.1759530791788855, + "grad_norm": 0.7243392727575753, + "learning_rate": 7.728303765284136e-06, + "loss": 0.1587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1535375863313675, + "step": 5295, + "valid_targets_mean": 2897.2, + "valid_targets_min": 932 + }, + { + "epoch": 5.180840664711632, + "grad_norm": 0.7297677901128055, + "learning_rate": 7.689844109089217e-06, + "loss": 0.1495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14912715554237366, + "step": 5300, + "valid_targets_mean": 2656.3, + "valid_targets_min": 1004 + }, + { + "epoch": 5.185728250244379, + "grad_norm": 0.7467264254993835, + "learning_rate": 7.651457599137929e-06, + "loss": 0.1498, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14847451448440552, + "step": 5305, + "valid_targets_mean": 2904.3, + "valid_targets_min": 638 + }, + { + "epoch": 5.190615835777126, + "grad_norm": 0.8720410023528405, + "learning_rate": 7.613144463520732e-06, + "loss": 0.1555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1656719148159027, + "step": 5310, + "valid_targets_mean": 2167.3, + "valid_targets_min": 734 + }, + { + "epoch": 5.195503421309873, + "grad_norm": 0.7908689871901836, + "learning_rate": 7.574904929892093e-06, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16172878444194794, + "step": 5315, + "valid_targets_mean": 2394.0, + "valid_targets_min": 898 + }, + { + "epoch": 5.20039100684262, + "grad_norm": 0.755139638023909, + "learning_rate": 7.53673922546913e-06, + "loss": 0.1367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14658725261688232, + "step": 5320, + "valid_targets_mean": 2478.6, + "valid_targets_min": 810 + }, + { + "epoch": 5.205278592375366, + "grad_norm": 0.8778718952166356, + "learning_rate": 7.498647577030291e-06, + "loss": 0.161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14963701367378235, + "step": 5325, + "valid_targets_mean": 2121.1, + "valid_targets_min": 635 + }, + { + "epoch": 5.210166177908113, + "grad_norm": 0.8376481931091295, + "learning_rate": 7.460630210913964e-06, + "loss": 0.1529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14989005029201508, + "step": 5330, + "valid_targets_mean": 1952.4, + "valid_targets_min": 648 + }, + { + "epoch": 5.21505376344086, + "grad_norm": 0.8652589359238804, + "learning_rate": 7.422687353017173e-06, + "loss": 0.1486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15695667266845703, + "step": 5335, + "valid_targets_mean": 1941.1, + "valid_targets_min": 795 + }, + { + "epoch": 5.219941348973607, + "grad_norm": 0.8041415349342634, + "learning_rate": 7.384819228794222e-06, + "loss": 0.1609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1654549539089203, + "step": 5340, + "valid_targets_mean": 2202.2, + "valid_targets_min": 1217 + }, + { + "epoch": 5.224828934506354, + "grad_norm": 0.7245646950191174, + "learning_rate": 7.347026063255338e-06, + "loss": 0.147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14550118148326874, + "step": 5345, + "valid_targets_mean": 2840.0, + "valid_targets_min": 1225 + }, + { + "epoch": 5.2297165200391005, + "grad_norm": 0.761083682415506, + "learning_rate": 7.309308080965345e-06, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15238893032073975, + "step": 5350, + "valid_targets_mean": 2448.1, + "valid_targets_min": 953 + }, + { + "epoch": 5.234604105571847, + "grad_norm": 0.8141359878753163, + "learning_rate": 7.271665506042349e-06, + "loss": 0.1536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1655212789773941, + "step": 5355, + "valid_targets_mean": 2414.3, + "valid_targets_min": 750 + }, + { + "epoch": 5.239491691104594, + "grad_norm": 0.7925319949059126, + "learning_rate": 7.234098562156382e-06, + "loss": 0.1478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14529694616794586, + "step": 5360, + "valid_targets_mean": 2259.9, + "valid_targets_min": 796 + }, + { + "epoch": 5.244379276637341, + "grad_norm": 0.8281189537598208, + "learning_rate": 7.196607472528083e-06, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15511846542358398, + "step": 5365, + "valid_targets_mean": 2057.9, + "valid_targets_min": 696 + }, + { + "epoch": 5.249266862170088, + "grad_norm": 0.9202487657550533, + "learning_rate": 7.1591924599273646e-06, + "loss": 0.1623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16870905458927155, + "step": 5370, + "valid_targets_mean": 2043.6, + "valid_targets_min": 806 + }, + { + "epoch": 5.254154447702835, + "grad_norm": 0.7490177593952363, + "learning_rate": 7.121853746672087e-06, + "loss": 0.1565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1420806646347046, + "step": 5375, + "valid_targets_mean": 2500.4, + "valid_targets_min": 1353 + }, + { + "epoch": 5.259042033235581, + "grad_norm": 0.7953194730901839, + "learning_rate": 7.084591554626761e-06, + "loss": 0.1596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15275171399116516, + "step": 5380, + "valid_targets_mean": 2575.5, + "valid_targets_min": 920 + }, + { + "epoch": 5.263929618768328, + "grad_norm": 0.8052676980006547, + "learning_rate": 7.047406105201202e-06, + "loss": 0.1584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14699678122997284, + "step": 5385, + "valid_targets_mean": 2143.2, + "valid_targets_min": 514 + }, + { + "epoch": 5.268817204301075, + "grad_norm": 0.8178962169352656, + "learning_rate": 7.010297619349238e-06, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14148107171058655, + "step": 5390, + "valid_targets_mean": 1963.5, + "valid_targets_min": 642 + }, + { + "epoch": 5.273704789833822, + "grad_norm": 0.7553088050684637, + "learning_rate": 6.973266317567364e-06, + "loss": 0.1657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1751079261302948, + "step": 5395, + "valid_targets_mean": 2179.1, + "valid_targets_min": 829 + }, + { + "epoch": 5.278592375366569, + "grad_norm": 0.7418479058847184, + "learning_rate": 6.936312419893458e-06, + "loss": 0.1552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1365339756011963, + "step": 5400, + "valid_targets_mean": 2537.8, + "valid_targets_min": 923 + }, + { + "epoch": 5.2834799608993155, + "grad_norm": 0.8557491666172627, + "learning_rate": 6.899436145905474e-06, + "loss": 0.16, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15646493434906006, + "step": 5405, + "valid_targets_mean": 2075.5, + "valid_targets_min": 648 + }, + { + "epoch": 5.288367546432062, + "grad_norm": 0.8021756331273079, + "learning_rate": 6.862637714720138e-06, + "loss": 0.1409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1329081803560257, + "step": 5410, + "valid_targets_mean": 1917.8, + "valid_targets_min": 682 + }, + { + "epoch": 5.293255131964809, + "grad_norm": 0.7229260402895247, + "learning_rate": 6.825917344991611e-06, + "loss": 0.1584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15381160378456116, + "step": 5415, + "valid_targets_mean": 2806.7, + "valid_targets_min": 994 + }, + { + "epoch": 5.298142717497556, + "grad_norm": 0.8535602266976685, + "learning_rate": 6.789275254910246e-06, + "loss": 0.1464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13848789036273956, + "step": 5420, + "valid_targets_mean": 1859.3, + "valid_targets_min": 713 + }, + { + "epoch": 5.303030303030303, + "grad_norm": 0.8074188447585233, + "learning_rate": 6.752711662201241e-06, + "loss": 0.1486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15609613060951233, + "step": 5425, + "valid_targets_mean": 2341.6, + "valid_targets_min": 806 + }, + { + "epoch": 5.30791788856305, + "grad_norm": 0.877202839044622, + "learning_rate": 6.716226784123383e-06, + "loss": 0.1538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16460809111595154, + "step": 5430, + "valid_targets_mean": 2276.0, + "valid_targets_min": 662 + }, + { + "epoch": 5.312805474095796, + "grad_norm": 0.7588859531837344, + "learning_rate": 6.6798208374677384e-06, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16514082252979279, + "step": 5435, + "valid_targets_mean": 2550.6, + "valid_targets_min": 761 + }, + { + "epoch": 5.317693059628543, + "grad_norm": 0.7981982898217473, + "learning_rate": 6.643494038556355e-06, + "loss": 0.1547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17266137897968292, + "step": 5440, + "valid_targets_mean": 2877.4, + "valid_targets_min": 1372 + }, + { + "epoch": 5.32258064516129, + "grad_norm": 0.7362910731569113, + "learning_rate": 6.607246603241009e-06, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15483227372169495, + "step": 5445, + "valid_targets_mean": 2524.4, + "valid_targets_min": 884 + }, + { + "epoch": 5.327468230694037, + "grad_norm": 0.7656377794863742, + "learning_rate": 6.571078746901878e-06, + "loss": 0.1497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14594011008739471, + "step": 5450, + "valid_targets_mean": 2367.2, + "valid_targets_min": 768 + }, + { + "epoch": 5.332355816226784, + "grad_norm": 0.6850050225914522, + "learning_rate": 6.53499068444631e-06, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1416652500629425, + "step": 5455, + "valid_targets_mean": 2953.6, + "valid_targets_min": 704 + }, + { + "epoch": 5.3372434017595305, + "grad_norm": 0.7391512591400116, + "learning_rate": 6.498982630307518e-06, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1710202693939209, + "step": 5460, + "valid_targets_mean": 2936.5, + "valid_targets_min": 863 + }, + { + "epoch": 5.342130987292277, + "grad_norm": 0.7660416451265245, + "learning_rate": 6.463054798443287e-06, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16152888536453247, + "step": 5465, + "valid_targets_mean": 2735.8, + "valid_targets_min": 1110 + }, + { + "epoch": 5.347018572825024, + "grad_norm": 0.8252295444761306, + "learning_rate": 6.4272074023347606e-06, + "loss": 0.1535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15773028135299683, + "step": 5470, + "valid_targets_mean": 2095.6, + "valid_targets_min": 893 + }, + { + "epoch": 5.351906158357771, + "grad_norm": 0.9055114549440841, + "learning_rate": 6.391440654985106e-06, + "loss": 0.1536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17055881023406982, + "step": 5475, + "valid_targets_mean": 1950.2, + "valid_targets_min": 684 + }, + { + "epoch": 5.356793743890518, + "grad_norm": 0.6588127218396698, + "learning_rate": 6.355754768918301e-06, + "loss": 0.1511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13662315905094147, + "step": 5480, + "valid_targets_mean": 3307.9, + "valid_targets_min": 747 + }, + { + "epoch": 5.361681329423265, + "grad_norm": 0.7472617276476732, + "learning_rate": 6.32014995617785e-06, + "loss": 0.1561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15279489755630493, + "step": 5485, + "valid_targets_mean": 2531.1, + "valid_targets_min": 645 + }, + { + "epoch": 5.366568914956011, + "grad_norm": 0.7945430057967098, + "learning_rate": 6.284626428325504e-06, + "loss": 0.1472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13775640726089478, + "step": 5490, + "valid_targets_mean": 2237.1, + "valid_targets_min": 793 + }, + { + "epoch": 5.371456500488758, + "grad_norm": 1.1165361040355197, + "learning_rate": 6.249184396440053e-06, + "loss": 0.1582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15144461393356323, + "step": 5495, + "valid_targets_mean": 2468.8, + "valid_targets_min": 615 + }, + { + "epoch": 5.376344086021505, + "grad_norm": 0.7782637843895414, + "learning_rate": 6.2138240711160125e-06, + "loss": 0.1661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1696222573518753, + "step": 5500, + "valid_targets_mean": 2696.6, + "valid_targets_min": 699 + }, + { + "epoch": 5.381231671554252, + "grad_norm": 0.7825515135644441, + "learning_rate": 6.178545662462425e-06, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16258756816387177, + "step": 5505, + "valid_targets_mean": 2587.4, + "valid_targets_min": 754 + }, + { + "epoch": 5.386119257086999, + "grad_norm": 0.8421569052905954, + "learning_rate": 6.1433493801015865e-06, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13311320543289185, + "step": 5510, + "valid_targets_mean": 1930.4, + "valid_targets_min": 617 + }, + { + "epoch": 5.3910068426197455, + "grad_norm": 0.8261937566397005, + "learning_rate": 6.108235433167791e-06, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16040858626365662, + "step": 5515, + "valid_targets_mean": 2021.9, + "valid_targets_min": 684 + }, + { + "epoch": 5.395894428152492, + "grad_norm": 0.7936724009453671, + "learning_rate": 6.073204030306101e-06, + "loss": 0.1465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14740975201129913, + "step": 5520, + "valid_targets_mean": 2411.1, + "valid_targets_min": 636 + }, + { + "epoch": 5.400782013685239, + "grad_norm": 0.8088963586124284, + "learning_rate": 6.038255379671121e-06, + "loss": 0.1492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15532401204109192, + "step": 5525, + "valid_targets_mean": 2203.5, + "valid_targets_min": 678 + }, + { + "epoch": 5.405669599217986, + "grad_norm": 0.8249172911213923, + "learning_rate": 6.003389688925736e-06, + "loss": 0.1515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13733899593353271, + "step": 5530, + "valid_targets_mean": 2478.9, + "valid_targets_min": 738 + }, + { + "epoch": 5.410557184750733, + "grad_norm": 0.6793605880624324, + "learning_rate": 5.968607165239897e-06, + "loss": 0.1466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14237245917320251, + "step": 5535, + "valid_targets_mean": 3130.7, + "valid_targets_min": 1230 + }, + { + "epoch": 5.41544477028348, + "grad_norm": 0.9530584429559311, + "learning_rate": 5.933908015289367e-06, + "loss": 0.1615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17519810795783997, + "step": 5540, + "valid_targets_mean": 1629.4, + "valid_targets_min": 512 + }, + { + "epoch": 5.420332355816226, + "grad_norm": 0.7526914536401272, + "learning_rate": 5.89929244525451e-06, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16908365488052368, + "step": 5545, + "valid_targets_mean": 2734.1, + "valid_targets_min": 1191 + }, + { + "epoch": 5.425219941348973, + "grad_norm": 0.7626748112290224, + "learning_rate": 5.864760660819073e-06, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16685664653778076, + "step": 5550, + "valid_targets_mean": 2915.1, + "valid_targets_min": 1098 + }, + { + "epoch": 5.43010752688172, + "grad_norm": 0.8034530660024547, + "learning_rate": 5.830312867168948e-06, + "loss": 0.1482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14739468693733215, + "step": 5555, + "valid_targets_mean": 2265.8, + "valid_targets_min": 692 + }, + { + "epoch": 5.434995112414467, + "grad_norm": 0.7652321619302118, + "learning_rate": 5.795949268990964e-06, + "loss": 0.1504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14698611199855804, + "step": 5560, + "valid_targets_mean": 2454.2, + "valid_targets_min": 717 + }, + { + "epoch": 5.439882697947214, + "grad_norm": 0.8333089448649366, + "learning_rate": 5.7616700704716545e-06, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15633103251457214, + "step": 5565, + "valid_targets_mean": 2116.5, + "valid_targets_min": 848 + }, + { + "epoch": 5.4447702834799605, + "grad_norm": 0.8009884901164667, + "learning_rate": 5.727475475296047e-06, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15122680366039276, + "step": 5570, + "valid_targets_mean": 2134.5, + "valid_targets_min": 900 + }, + { + "epoch": 5.449657869012707, + "grad_norm": 0.8124584053736166, + "learning_rate": 5.693365686646501e-06, + "loss": 0.157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16159634292125702, + "step": 5575, + "valid_targets_mean": 2173.5, + "valid_targets_min": 809 + }, + { + "epoch": 5.454545454545454, + "grad_norm": 0.7935256921202335, + "learning_rate": 5.659340907201418e-06, + "loss": 0.1534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16111275553703308, + "step": 5580, + "valid_targets_mean": 2267.2, + "valid_targets_min": 701 + }, + { + "epoch": 5.459433040078201, + "grad_norm": 0.6835989696344518, + "learning_rate": 5.625401339134093e-06, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14738526940345764, + "step": 5585, + "valid_targets_mean": 3000.6, + "valid_targets_min": 879 + }, + { + "epoch": 5.464320625610948, + "grad_norm": 0.7930964833671504, + "learning_rate": 5.591547184111514e-06, + "loss": 0.1594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1674892157316208, + "step": 5590, + "valid_targets_mean": 2696.9, + "valid_targets_min": 687 + }, + { + "epoch": 5.469208211143695, + "grad_norm": 0.744501626898747, + "learning_rate": 5.557778643293117e-06, + "loss": 0.1618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16441041231155396, + "step": 5595, + "valid_targets_mean": 2941.3, + "valid_targets_min": 1195 + }, + { + "epoch": 5.474095796676442, + "grad_norm": 0.7910032892994638, + "learning_rate": 5.524095917329668e-06, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16341263055801392, + "step": 5600, + "valid_targets_mean": 2293.9, + "valid_targets_min": 1001 + }, + { + "epoch": 5.478983382209188, + "grad_norm": 0.8189803335488413, + "learning_rate": 5.490499206361992e-06, + "loss": 0.1655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1621970534324646, + "step": 5605, + "valid_targets_mean": 2192.6, + "valid_targets_min": 1044 + }, + { + "epoch": 5.483870967741936, + "grad_norm": 0.8435168373600971, + "learning_rate": 5.456988710019822e-06, + "loss": 0.1571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1423211395740509, + "step": 5610, + "valid_targets_mean": 1718.4, + "valid_targets_min": 756 + }, + { + "epoch": 5.488758553274682, + "grad_norm": 0.8432435824638052, + "learning_rate": 5.4235646274206185e-06, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15067586302757263, + "step": 5615, + "valid_targets_mean": 2142.1, + "valid_targets_min": 632 + }, + { + "epoch": 5.49364613880743, + "grad_norm": 0.8434772077306247, + "learning_rate": 5.3902271571683815e-06, + "loss": 0.1661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15619194507598877, + "step": 5620, + "valid_targets_mean": 2239.2, + "valid_targets_min": 719 + }, + { + "epoch": 5.4985337243401755, + "grad_norm": 0.7932516543133425, + "learning_rate": 5.356976497352442e-06, + "loss": 0.1608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14296786487102509, + "step": 5625, + "valid_targets_mean": 2430.8, + "valid_targets_min": 878 + }, + { + "epoch": 5.503421309872923, + "grad_norm": 0.7340907560380051, + "learning_rate": 5.323812845546334e-06, + "loss": 0.158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15341924130916595, + "step": 5630, + "valid_targets_mean": 2661.9, + "valid_targets_min": 889 + }, + { + "epoch": 5.508308895405669, + "grad_norm": 0.9584349909909398, + "learning_rate": 5.290736398806575e-06, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17308732867240906, + "step": 5635, + "valid_targets_mean": 1577.6, + "valid_targets_min": 848 + }, + { + "epoch": 5.513196480938417, + "grad_norm": 0.8019713328945052, + "learning_rate": 5.2577473536715275e-06, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16516554355621338, + "step": 5640, + "valid_targets_mean": 2460.1, + "valid_targets_min": 1066 + }, + { + "epoch": 5.518084066471163, + "grad_norm": 0.8191884705772617, + "learning_rate": 5.224845906160219e-06, + "loss": 0.1692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15267640352249146, + "step": 5645, + "valid_targets_mean": 2165.6, + "valid_targets_min": 752 + }, + { + "epoch": 5.5229716520039105, + "grad_norm": 0.8330339846872703, + "learning_rate": 5.1920322517711615e-06, + "loss": 0.1552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15274596214294434, + "step": 5650, + "valid_targets_mean": 2050.2, + "valid_targets_min": 591 + }, + { + "epoch": 5.527859237536656, + "grad_norm": 0.7824632706097082, + "learning_rate": 5.1593065854812294e-06, + "loss": 0.1508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1320272833108902, + "step": 5655, + "valid_targets_mean": 2084.6, + "valid_targets_min": 789 + }, + { + "epoch": 5.532746823069404, + "grad_norm": 0.8015604744159438, + "learning_rate": 5.1266691017444505e-06, + "loss": 0.1525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1577574610710144, + "step": 5660, + "valid_targets_mean": 2387.2, + "valid_targets_min": 800 + }, + { + "epoch": 5.53763440860215, + "grad_norm": 0.7592693450999322, + "learning_rate": 5.094119994490898e-06, + "loss": 0.1595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16440163552761078, + "step": 5665, + "valid_targets_mean": 2866.8, + "valid_targets_min": 772 + }, + { + "epoch": 5.542521994134898, + "grad_norm": 0.7921173779072441, + "learning_rate": 5.061659457125514e-06, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15453504025936127, + "step": 5670, + "valid_targets_mean": 2380.8, + "valid_targets_min": 1188 + }, + { + "epoch": 5.547409579667644, + "grad_norm": 0.8120586505606648, + "learning_rate": 5.029287682526949e-06, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16854724287986755, + "step": 5675, + "valid_targets_mean": 2460.8, + "valid_targets_min": 742 + }, + { + "epoch": 5.552297165200391, + "grad_norm": 0.8499742573768506, + "learning_rate": 4.997004863046446e-06, + "loss": 0.1609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1677948236465454, + "step": 5680, + "valid_targets_mean": 2218.1, + "valid_targets_min": 1013 + }, + { + "epoch": 5.557184750733137, + "grad_norm": 0.8783081648919876, + "learning_rate": 4.964811190506671e-06, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18013803660869598, + "step": 5685, + "valid_targets_mean": 2322.4, + "valid_targets_min": 692 + }, + { + "epoch": 5.562072336265885, + "grad_norm": 0.7709090075615438, + "learning_rate": 4.932706856200589e-06, + "loss": 0.1511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14455710351467133, + "step": 5690, + "valid_targets_mean": 2390.8, + "valid_targets_min": 626 + }, + { + "epoch": 5.566959921798632, + "grad_norm": 0.8439698983111165, + "learning_rate": 4.90069205089033e-06, + "loss": 0.1532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16360528767108917, + "step": 5695, + "valid_targets_mean": 2045.9, + "valid_targets_min": 708 + }, + { + "epoch": 5.571847507331379, + "grad_norm": 0.8445073532119768, + "learning_rate": 4.868766964806029e-06, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16692796349525452, + "step": 5700, + "valid_targets_mean": 2184.8, + "valid_targets_min": 970 + }, + { + "epoch": 5.5767350928641255, + "grad_norm": 0.7247679212006615, + "learning_rate": 4.8369317876447365e-06, + "loss": 0.1518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15431565046310425, + "step": 5705, + "valid_targets_mean": 3089.9, + "valid_targets_min": 1388 + }, + { + "epoch": 5.581622678396872, + "grad_norm": 0.7600528639537043, + "learning_rate": 4.805186708569245e-06, + "loss": 0.1563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17368043959140778, + "step": 5710, + "valid_targets_mean": 2844.1, + "valid_targets_min": 958 + }, + { + "epoch": 5.586510263929619, + "grad_norm": 0.7391855729545126, + "learning_rate": 4.773531916207008e-06, + "loss": 0.161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13605889678001404, + "step": 5715, + "valid_targets_mean": 2471.4, + "valid_targets_min": 992 + }, + { + "epoch": 5.591397849462366, + "grad_norm": 0.769129841502073, + "learning_rate": 4.741967598649002e-06, + "loss": 0.1519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15731492638587952, + "step": 5720, + "valid_targets_mean": 2548.4, + "valid_targets_min": 711 + }, + { + "epoch": 5.596285434995113, + "grad_norm": 1.1287691604840524, + "learning_rate": 4.7104939434485884e-06, + "loss": 0.1416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1409187912940979, + "step": 5725, + "valid_targets_mean": 2582.1, + "valid_targets_min": 892 + }, + { + "epoch": 5.60117302052786, + "grad_norm": 0.9156130590977292, + "learning_rate": 4.679111137620442e-06, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1329200565814972, + "step": 5730, + "valid_targets_mean": 1692.3, + "valid_targets_min": 591 + }, + { + "epoch": 5.606060606060606, + "grad_norm": 0.8263554263382872, + "learning_rate": 4.64781936763939e-06, + "loss": 0.1502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14420738816261292, + "step": 5735, + "valid_targets_mean": 2274.1, + "valid_targets_min": 830 + }, + { + "epoch": 5.610948191593353, + "grad_norm": 0.7412321177836615, + "learning_rate": 4.616618819439353e-06, + "loss": 0.1501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1455959975719452, + "step": 5740, + "valid_targets_mean": 2443.1, + "valid_targets_min": 875 + }, + { + "epoch": 5.6158357771261, + "grad_norm": 0.7056612270287848, + "learning_rate": 4.585509678412208e-06, + "loss": 0.1519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15663772821426392, + "step": 5745, + "valid_targets_mean": 3141.4, + "valid_targets_min": 1052 + }, + { + "epoch": 5.620723362658847, + "grad_norm": 0.8069910406282396, + "learning_rate": 4.5544921294066935e-06, + "loss": 0.1541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14563411474227905, + "step": 5750, + "valid_targets_mean": 2318.9, + "valid_targets_min": 1035 + }, + { + "epoch": 5.625610948191594, + "grad_norm": 0.8114201491577269, + "learning_rate": 4.523566356727309e-06, + "loss": 0.1546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1693604439496994, + "step": 5755, + "valid_targets_mean": 2370.1, + "valid_targets_min": 869 + }, + { + "epoch": 5.6304985337243405, + "grad_norm": 0.8443036168376865, + "learning_rate": 4.492732544133236e-06, + "loss": 0.158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15634265542030334, + "step": 5760, + "valid_targets_mean": 1911.4, + "valid_targets_min": 700 + }, + { + "epoch": 5.635386119257087, + "grad_norm": 0.787586816101732, + "learning_rate": 4.461990874837227e-06, + "loss": 0.1502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15353325009346008, + "step": 5765, + "valid_targets_mean": 2510.8, + "valid_targets_min": 762 + }, + { + "epoch": 5.640273704789834, + "grad_norm": 0.8265538830295129, + "learning_rate": 4.431341531504536e-06, + "loss": 0.1529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17224039137363434, + "step": 5770, + "valid_targets_mean": 2606.2, + "valid_targets_min": 897 + }, + { + "epoch": 5.645161290322581, + "grad_norm": 0.9987858614241965, + "learning_rate": 4.400784696251801e-06, + "loss": 0.1596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1915130913257599, + "step": 5775, + "valid_targets_mean": 2068.2, + "valid_targets_min": 1206 + }, + { + "epoch": 5.650048875855328, + "grad_norm": 0.8197168076239274, + "learning_rate": 4.37032055064599e-06, + "loss": 0.1472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1514715850353241, + "step": 5780, + "valid_targets_mean": 2108.2, + "valid_targets_min": 629 + }, + { + "epoch": 5.654936461388075, + "grad_norm": 0.7371445886865936, + "learning_rate": 4.3399492757033235e-06, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20320241153240204, + "step": 5785, + "valid_targets_mean": 2937.3, + "valid_targets_min": 720 + }, + { + "epoch": 5.659824046920821, + "grad_norm": 0.860820707417868, + "learning_rate": 4.309671051888187e-06, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14680787920951843, + "step": 5790, + "valid_targets_mean": 1897.9, + "valid_targets_min": 534 + }, + { + "epoch": 5.664711632453568, + "grad_norm": 0.7425530568167907, + "learning_rate": 4.27948605911205e-06, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18525880575180054, + "step": 5795, + "valid_targets_mean": 2515.6, + "valid_targets_min": 752 + }, + { + "epoch": 5.669599217986315, + "grad_norm": 0.7672403302530817, + "learning_rate": 4.2493944767324245e-06, + "loss": 0.1587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17380796372890472, + "step": 5800, + "valid_targets_mean": 2673.0, + "valid_targets_min": 701 + }, + { + "epoch": 5.674486803519062, + "grad_norm": 0.8190685848445297, + "learning_rate": 4.2193964835517656e-06, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1683279573917389, + "step": 5805, + "valid_targets_mean": 2120.9, + "valid_targets_min": 965 + }, + { + "epoch": 5.679374389051809, + "grad_norm": 1.002164296776056, + "learning_rate": 4.189492257816443e-06, + "loss": 0.1609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14651206135749817, + "step": 5810, + "valid_targets_mean": 2480.2, + "valid_targets_min": 740 + }, + { + "epoch": 5.6842619745845555, + "grad_norm": 0.8097394342502438, + "learning_rate": 4.159681977215663e-06, + "loss": 0.1575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17866715788841248, + "step": 5815, + "valid_targets_mean": 2452.8, + "valid_targets_min": 867 + }, + { + "epoch": 5.689149560117302, + "grad_norm": 0.8423732358599466, + "learning_rate": 4.1299658188804015e-06, + "loss": 0.158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1501942276954651, + "step": 5820, + "valid_targets_mean": 2100.3, + "valid_targets_min": 741 + }, + { + "epoch": 5.694037145650049, + "grad_norm": 0.7237945688434769, + "learning_rate": 4.1003439593823865e-06, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15869738161563873, + "step": 5825, + "valid_targets_mean": 2729.0, + "valid_targets_min": 986 + }, + { + "epoch": 5.698924731182796, + "grad_norm": 0.7467148363318494, + "learning_rate": 4.070816574733003e-06, + "loss": 0.1454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14755885303020477, + "step": 5830, + "valid_targets_mean": 2749.3, + "valid_targets_min": 980 + }, + { + "epoch": 5.703812316715543, + "grad_norm": 0.7867120070266118, + "learning_rate": 4.041383840382294e-06, + "loss": 0.1499, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.161052405834198, + "step": 5835, + "valid_targets_mean": 2601.4, + "valid_targets_min": 1062 + }, + { + "epoch": 5.70869990224829, + "grad_norm": 0.9049628718004302, + "learning_rate": 4.01204593121789e-06, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15240147709846497, + "step": 5840, + "valid_targets_mean": 1931.8, + "valid_targets_min": 722 + }, + { + "epoch": 5.713587487781036, + "grad_norm": 0.8698769453803369, + "learning_rate": 3.982803021563961e-06, + "loss": 0.1532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14392434060573578, + "step": 5845, + "valid_targets_mean": 2359.8, + "valid_targets_min": 790 + }, + { + "epoch": 5.718475073313783, + "grad_norm": 0.7228885339388987, + "learning_rate": 3.953655285180223e-06, + "loss": 0.148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14013831317424774, + "step": 5850, + "valid_targets_mean": 2797.0, + "valid_targets_min": 640 + }, + { + "epoch": 5.72336265884653, + "grad_norm": 0.7537686797769533, + "learning_rate": 3.924602895260843e-06, + "loss": 0.1486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12795598804950714, + "step": 5855, + "valid_targets_mean": 2317.8, + "valid_targets_min": 715 + }, + { + "epoch": 5.728250244379277, + "grad_norm": 0.7471691429860486, + "learning_rate": 3.895646024433475e-06, + "loss": 0.1483, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15418082475662231, + "step": 5860, + "valid_targets_mean": 2598.8, + "valid_targets_min": 734 + }, + { + "epoch": 5.733137829912024, + "grad_norm": 0.7549943500132472, + "learning_rate": 3.866784844758196e-06, + "loss": 0.1504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1698361039161682, + "step": 5865, + "valid_targets_mean": 2571.3, + "valid_targets_min": 1006 + }, + { + "epoch": 5.7380254154447705, + "grad_norm": 0.7902697956389423, + "learning_rate": 3.838019527726478e-06, + "loss": 0.1514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15650413930416107, + "step": 5870, + "valid_targets_mean": 2332.9, + "valid_targets_min": 748 + }, + { + "epoch": 5.742913000977517, + "grad_norm": 0.7626167489885958, + "learning_rate": 3.8093502442602037e-06, + "loss": 0.1578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1520456075668335, + "step": 5875, + "valid_targets_mean": 2732.3, + "valid_targets_min": 851 + }, + { + "epoch": 5.747800586510264, + "grad_norm": 0.890203614110944, + "learning_rate": 3.7807771647106073e-06, + "loss": 0.1597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16833049058914185, + "step": 5880, + "valid_targets_mean": 2514.6, + "valid_targets_min": 1051 + }, + { + "epoch": 5.752688172043011, + "grad_norm": 0.7072501670025816, + "learning_rate": 3.752300458857303e-06, + "loss": 0.1538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15029340982437134, + "step": 5885, + "valid_targets_mean": 3031.2, + "valid_targets_min": 837 + }, + { + "epoch": 5.757575757575758, + "grad_norm": 0.7640980023033417, + "learning_rate": 3.723920295907255e-06, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1618524044752121, + "step": 5890, + "valid_targets_mean": 2569.4, + "valid_targets_min": 728 + }, + { + "epoch": 5.762463343108505, + "grad_norm": 0.7319035255044318, + "learning_rate": 3.695636844493764e-06, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16255182027816772, + "step": 5895, + "valid_targets_mean": 3114.8, + "valid_targets_min": 951 + }, + { + "epoch": 5.767350928641251, + "grad_norm": 0.8063750882434633, + "learning_rate": 3.66745027267549e-06, + "loss": 0.155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1666271686553955, + "step": 5900, + "valid_targets_mean": 2507.8, + "valid_targets_min": 986 + }, + { + "epoch": 5.772238514173998, + "grad_norm": 0.8345912032783375, + "learning_rate": 3.6393607479354275e-06, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15384799242019653, + "step": 5905, + "valid_targets_mean": 1862.2, + "valid_targets_min": 635 + }, + { + "epoch": 5.777126099706745, + "grad_norm": 0.7061819113437781, + "learning_rate": 3.611368437179934e-06, + "loss": 0.1607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18276137113571167, + "step": 5910, + "valid_targets_mean": 3247.3, + "valid_targets_min": 1489 + }, + { + "epoch": 5.782013685239492, + "grad_norm": 0.8167005148421289, + "learning_rate": 3.583473506737727e-06, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1597469002008438, + "step": 5915, + "valid_targets_mean": 2362.9, + "valid_targets_min": 759 + }, + { + "epoch": 5.786901270772239, + "grad_norm": 0.7711204073304082, + "learning_rate": 3.555676122358884e-06, + "loss": 0.1597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1491527557373047, + "step": 5920, + "valid_targets_mean": 2585.6, + "valid_targets_min": 1306 + }, + { + "epoch": 5.7917888563049855, + "grad_norm": 0.7283811601045469, + "learning_rate": 3.5279764492138724e-06, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17027896642684937, + "step": 5925, + "valid_targets_mean": 3067.6, + "valid_targets_min": 522 + }, + { + "epoch": 5.796676441837732, + "grad_norm": 0.7011679340720517, + "learning_rate": 3.500374651892573e-06, + "loss": 0.1603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1561456024646759, + "step": 5930, + "valid_targets_mean": 2874.4, + "valid_targets_min": 851 + }, + { + "epoch": 5.801564027370479, + "grad_norm": 0.9979178900846251, + "learning_rate": 3.472870894403291e-06, + "loss": 0.1679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1694871187210083, + "step": 5935, + "valid_targets_mean": 2362.9, + "valid_targets_min": 513 + }, + { + "epoch": 5.806451612903226, + "grad_norm": 0.8639990166797461, + "learning_rate": 3.4454653401717896e-06, + "loss": 0.1527, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12851397693157196, + "step": 5940, + "valid_targets_mean": 1915.4, + "valid_targets_min": 746 + }, + { + "epoch": 5.811339198435973, + "grad_norm": 0.6986837988916482, + "learning_rate": 3.4181581520403005e-06, + "loss": 0.1518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15324917435646057, + "step": 5945, + "valid_targets_mean": 3119.9, + "valid_targets_min": 1011 + }, + { + "epoch": 5.81622678396872, + "grad_norm": 0.8313260438961377, + "learning_rate": 3.390949492266569e-06, + "loss": 0.1481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1447916030883789, + "step": 5950, + "valid_targets_mean": 2159.3, + "valid_targets_min": 800 + }, + { + "epoch": 5.821114369501466, + "grad_norm": 0.885127583979189, + "learning_rate": 3.363839522522916e-06, + "loss": 0.1591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15982311964035034, + "step": 5955, + "valid_targets_mean": 1842.9, + "valid_targets_min": 607 + }, + { + "epoch": 5.826001955034213, + "grad_norm": 0.7808780924503399, + "learning_rate": 3.336828403895227e-06, + "loss": 0.1536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1592644453048706, + "step": 5960, + "valid_targets_mean": 2727.3, + "valid_targets_min": 1050 + }, + { + "epoch": 5.83088954056696, + "grad_norm": 0.685353500101023, + "learning_rate": 3.309916296882021e-06, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1466306447982788, + "step": 5965, + "valid_targets_mean": 2775.5, + "valid_targets_min": 627 + }, + { + "epoch": 5.835777126099707, + "grad_norm": 1.2671880923497407, + "learning_rate": 3.2831033613935092e-06, + "loss": 0.1496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15228471159934998, + "step": 5970, + "valid_targets_mean": 2204.8, + "valid_targets_min": 1143 + }, + { + "epoch": 5.840664711632454, + "grad_norm": 0.7532578861942046, + "learning_rate": 3.256389756750613e-06, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13797354698181152, + "step": 5975, + "valid_targets_mean": 2369.1, + "valid_targets_min": 788 + }, + { + "epoch": 5.8455522971652005, + "grad_norm": 0.7743267560564064, + "learning_rate": 3.229775641684063e-06, + "loss": 0.1586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16579267382621765, + "step": 5980, + "valid_targets_mean": 2343.3, + "valid_targets_min": 888 + }, + { + "epoch": 5.850439882697947, + "grad_norm": 0.7015726455253567, + "learning_rate": 3.203261174333403e-06, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14441198110580444, + "step": 5985, + "valid_targets_mean": 2980.8, + "valid_targets_min": 807 + }, + { + "epoch": 5.855327468230694, + "grad_norm": 0.7070459951461793, + "learning_rate": 3.1768465122460813e-06, + "loss": 0.1549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1486390233039856, + "step": 5990, + "valid_targets_mean": 2794.2, + "valid_targets_min": 824 + }, + { + "epoch": 5.860215053763441, + "grad_norm": 0.7969197667556421, + "learning_rate": 3.1505318123765137e-06, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16264227032661438, + "step": 5995, + "valid_targets_mean": 2405.8, + "valid_targets_min": 1005 + }, + { + "epoch": 5.865102639296188, + "grad_norm": 0.7190738114648545, + "learning_rate": 3.12431723108515e-06, + "loss": 0.1601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1532568335533142, + "step": 6000, + "valid_targets_mean": 2788.5, + "valid_targets_min": 903 + }, + { + "epoch": 5.869990224828935, + "grad_norm": 0.7936029883530649, + "learning_rate": 3.0982029241375343e-06, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16097313165664673, + "step": 6005, + "valid_targets_mean": 2423.9, + "valid_targets_min": 911 + }, + { + "epoch": 5.874877810361681, + "grad_norm": 0.7946199748419005, + "learning_rate": 3.0721890467033866e-06, + "loss": 0.1469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1583392471075058, + "step": 6010, + "valid_targets_mean": 2257.8, + "valid_targets_min": 654 + }, + { + "epoch": 5.879765395894428, + "grad_norm": 0.8074320913746822, + "learning_rate": 3.0462757533556784e-06, + "loss": 0.1466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13972869515419006, + "step": 6015, + "valid_targets_mean": 2338.6, + "valid_targets_min": 708 + }, + { + "epoch": 5.884652981427175, + "grad_norm": 0.8929139568223418, + "learning_rate": 3.020463198069721e-06, + "loss": 0.1529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16818490624427795, + "step": 6020, + "valid_targets_mean": 2117.4, + "valid_targets_min": 727 + }, + { + "epoch": 5.889540566959922, + "grad_norm": 0.6713747500739194, + "learning_rate": 2.994751534222251e-06, + "loss": 0.1476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1403338611125946, + "step": 6025, + "valid_targets_mean": 2784.2, + "valid_targets_min": 1212 + }, + { + "epoch": 5.894428152492669, + "grad_norm": 0.7757304779081556, + "learning_rate": 2.969140914590498e-06, + "loss": 0.1488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14841482043266296, + "step": 6030, + "valid_targets_mean": 2193.8, + "valid_targets_min": 790 + }, + { + "epoch": 5.8993157380254155, + "grad_norm": 0.8629820125294086, + "learning_rate": 2.943631491351311e-06, + "loss": 0.1544, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15556024014949799, + "step": 6035, + "valid_targets_mean": 2002.3, + "valid_targets_min": 1163 + }, + { + "epoch": 5.904203323558162, + "grad_norm": 0.8188534807515834, + "learning_rate": 2.918223416080219e-06, + "loss": 0.1502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16173824667930603, + "step": 6040, + "valid_targets_mean": 2232.9, + "valid_targets_min": 891 + }, + { + "epoch": 5.909090909090909, + "grad_norm": 0.8132660493437572, + "learning_rate": 2.892916839750557e-06, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1494334191083908, + "step": 6045, + "valid_targets_mean": 2253.3, + "valid_targets_min": 732 + }, + { + "epoch": 5.913978494623656, + "grad_norm": 0.654182287461871, + "learning_rate": 2.8677119127325625e-06, + "loss": 0.1574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15798476338386536, + "step": 6050, + "valid_targets_mean": 3594.9, + "valid_targets_min": 1150 + }, + { + "epoch": 5.918866080156403, + "grad_norm": 0.6693246221511635, + "learning_rate": 2.8426087847924643e-06, + "loss": 0.1523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13518929481506348, + "step": 6055, + "valid_targets_mean": 2902.4, + "valid_targets_min": 1123 + }, + { + "epoch": 5.92375366568915, + "grad_norm": 0.8107053745570154, + "learning_rate": 2.8176076050916255e-06, + "loss": 0.1463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15758317708969116, + "step": 6060, + "valid_targets_mean": 2398.6, + "valid_targets_min": 711 + }, + { + "epoch": 5.9286412512218964, + "grad_norm": 0.9021761140325874, + "learning_rate": 2.7927085221856185e-06, + "loss": 0.1442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15211939811706543, + "step": 6065, + "valid_targets_mean": 1915.1, + "valid_targets_min": 619 + }, + { + "epoch": 5.933528836754643, + "grad_norm": 0.7392044459110118, + "learning_rate": 2.7679116840233788e-06, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1442590355873108, + "step": 6070, + "valid_targets_mean": 2682.1, + "valid_targets_min": 837 + }, + { + "epoch": 5.93841642228739, + "grad_norm": 0.7909706387505339, + "learning_rate": 2.7432172379463073e-06, + "loss": 0.1481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1507616639137268, + "step": 6075, + "valid_targets_mean": 2728.3, + "valid_targets_min": 1133 + }, + { + "epoch": 5.943304007820137, + "grad_norm": 0.6858517709489104, + "learning_rate": 2.718625330687381e-06, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1569923758506775, + "step": 6080, + "valid_targets_mean": 3103.2, + "valid_targets_min": 698 + }, + { + "epoch": 5.948191593352884, + "grad_norm": 0.7282489397560563, + "learning_rate": 2.6941361083703244e-06, + "loss": 0.155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15824103355407715, + "step": 6085, + "valid_targets_mean": 3084.2, + "valid_targets_min": 1098 + }, + { + "epoch": 5.9530791788856305, + "grad_norm": 0.8200856694692574, + "learning_rate": 2.6697497165086826e-06, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14971165359020233, + "step": 6090, + "valid_targets_mean": 1849.7, + "valid_targets_min": 690 + }, + { + "epoch": 5.957966764418377, + "grad_norm": 0.8145406485221162, + "learning_rate": 2.645466300005013e-06, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14848312735557556, + "step": 6095, + "valid_targets_mean": 2086.6, + "valid_targets_min": 765 + }, + { + "epoch": 5.962854349951124, + "grad_norm": 0.7660068981755783, + "learning_rate": 2.6212860031499942e-06, + "loss": 0.1524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14158686995506287, + "step": 6100, + "valid_targets_mean": 2279.2, + "valid_targets_min": 1033 + }, + { + "epoch": 5.967741935483871, + "grad_norm": 0.8064378082219088, + "learning_rate": 2.597208969621561e-06, + "loss": 0.1637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14924579858779907, + "step": 6105, + "valid_targets_mean": 2187.1, + "valid_targets_min": 777 + }, + { + "epoch": 5.972629521016618, + "grad_norm": 0.7501577303291025, + "learning_rate": 2.5732353424840817e-06, + "loss": 0.1564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14925900101661682, + "step": 6110, + "valid_targets_mean": 2335.9, + "valid_targets_min": 634 + }, + { + "epoch": 5.977517106549365, + "grad_norm": 0.7898048269613115, + "learning_rate": 2.549365264187469e-06, + "loss": 0.1528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15726542472839355, + "step": 6115, + "valid_targets_mean": 2276.5, + "valid_targets_min": 1190 + }, + { + "epoch": 5.9824046920821115, + "grad_norm": 0.7940500094930575, + "learning_rate": 2.525598876566371e-06, + "loss": 0.1486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1534472405910492, + "step": 6120, + "valid_targets_mean": 2346.0, + "valid_targets_min": 809 + }, + { + "epoch": 5.987292277614858, + "grad_norm": 0.882710433816707, + "learning_rate": 2.5019363208393133e-06, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18637877702713013, + "step": 6125, + "valid_targets_mean": 1959.3, + "valid_targets_min": 1010 + }, + { + "epoch": 5.992179863147605, + "grad_norm": 0.798333765699043, + "learning_rate": 2.4783777376078443e-06, + "loss": 0.1664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15766102075576782, + "step": 6130, + "valid_targets_mean": 2160.6, + "valid_targets_min": 1137 + }, + { + "epoch": 5.997067448680352, + "grad_norm": 0.8188870991896096, + "learning_rate": 2.4549232668557154e-06, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15294690430164337, + "step": 6135, + "valid_targets_mean": 2066.3, + "valid_targets_min": 733 + }, + { + "epoch": 6.001955034213099, + "grad_norm": 0.772371500976364, + "learning_rate": 2.4315730479480614e-06, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14228251576423645, + "step": 6140, + "valid_targets_mean": 2113.5, + "valid_targets_min": 941 + }, + { + "epoch": 6.0068426197458455, + "grad_norm": 0.6951812269111425, + "learning_rate": 2.4083272196305483e-06, + "loss": 0.146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12359236180782318, + "step": 6145, + "valid_targets_mean": 2279.8, + "valid_targets_min": 901 + }, + { + "epoch": 6.011730205278592, + "grad_norm": 0.7341125048499366, + "learning_rate": 2.385185920028563e-06, + "loss": 0.1344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12021429091691971, + "step": 6150, + "valid_targets_mean": 2476.9, + "valid_targets_min": 834 + }, + { + "epoch": 6.016617790811339, + "grad_norm": 0.8580978920271629, + "learning_rate": 2.3621492866463845e-06, + "loss": 0.1476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14868280291557312, + "step": 6155, + "valid_targets_mean": 2026.8, + "valid_targets_min": 776 + }, + { + "epoch": 6.021505376344086, + "grad_norm": 0.8292027396293019, + "learning_rate": 2.3392174563663716e-06, + "loss": 0.1455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.154433935880661, + "step": 6160, + "valid_targets_mean": 2073.4, + "valid_targets_min": 610 + }, + { + "epoch": 6.026392961876833, + "grad_norm": 0.8083140379382482, + "learning_rate": 2.3163905654481524e-06, + "loss": 0.1393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15254239737987518, + "step": 6165, + "valid_targets_mean": 2362.4, + "valid_targets_min": 988 + }, + { + "epoch": 6.03128054740958, + "grad_norm": 0.711249329248669, + "learning_rate": 2.293668749527809e-06, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13382339477539062, + "step": 6170, + "valid_targets_mean": 2892.4, + "valid_targets_min": 761 + }, + { + "epoch": 6.0361681329423265, + "grad_norm": 0.7271334674837908, + "learning_rate": 2.2710521436170807e-06, + "loss": 0.1465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15734508633613586, + "step": 6175, + "valid_targets_mean": 3120.2, + "valid_targets_min": 1168 + }, + { + "epoch": 6.041055718475073, + "grad_norm": 0.8320574708134784, + "learning_rate": 2.2485408821025435e-06, + "loss": 0.1496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15673065185546875, + "step": 6180, + "valid_targets_mean": 2314.8, + "valid_targets_min": 676 + }, + { + "epoch": 6.04594330400782, + "grad_norm": 0.7618387380481915, + "learning_rate": 2.226135098744824e-06, + "loss": 0.1469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15294227004051208, + "step": 6185, + "valid_targets_mean": 2752.2, + "valid_targets_min": 1263 + }, + { + "epoch": 6.050830889540567, + "grad_norm": 0.7682499575279256, + "learning_rate": 2.2038349266778146e-06, + "loss": 0.1508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16213871538639069, + "step": 6190, + "valid_targets_mean": 2590.8, + "valid_targets_min": 764 + }, + { + "epoch": 6.055718475073314, + "grad_norm": 0.835630554054457, + "learning_rate": 2.1816404984078664e-06, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14952757954597473, + "step": 6195, + "valid_targets_mean": 2388.6, + "valid_targets_min": 678 + }, + { + "epoch": 6.0606060606060606, + "grad_norm": 0.7123162200549785, + "learning_rate": 2.1595519458129967e-06, + "loss": 0.1377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13014379143714905, + "step": 6200, + "valid_targets_mean": 2729.5, + "valid_targets_min": 815 + }, + { + "epoch": 6.065493646138807, + "grad_norm": 0.8818566517853015, + "learning_rate": 2.1375694001421343e-06, + "loss": 0.1493, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15553683042526245, + "step": 6205, + "valid_targets_mean": 1978.7, + "valid_targets_min": 761 + }, + { + "epoch": 6.070381231671554, + "grad_norm": 0.7715315446565182, + "learning_rate": 2.115692992014304e-06, + "loss": 0.1596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15126900374889374, + "step": 6210, + "valid_targets_mean": 2411.3, + "valid_targets_min": 1339 + }, + { + "epoch": 6.075268817204301, + "grad_norm": 0.8184332591539848, + "learning_rate": 2.0939228514178735e-06, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.146785169839859, + "step": 6215, + "valid_targets_mean": 2621.6, + "valid_targets_min": 892 + }, + { + "epoch": 6.080156402737048, + "grad_norm": 0.8703651745159651, + "learning_rate": 2.0722591077097844e-06, + "loss": 0.1442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14825935661792755, + "step": 6220, + "valid_targets_mean": 1934.5, + "valid_targets_min": 621 + }, + { + "epoch": 6.085043988269795, + "grad_norm": 0.8574526478965231, + "learning_rate": 2.0507018896147525e-06, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1470494568347931, + "step": 6225, + "valid_targets_mean": 2009.9, + "valid_targets_min": 591 + }, + { + "epoch": 6.0899315738025415, + "grad_norm": 0.8481048091940785, + "learning_rate": 2.0292513252245507e-06, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15715302526950836, + "step": 6230, + "valid_targets_mean": 2249.1, + "valid_targets_min": 750 + }, + { + "epoch": 6.094819159335288, + "grad_norm": 0.8473826531669545, + "learning_rate": 2.0079075419971937e-06, + "loss": 0.1385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15390710532665253, + "step": 6235, + "valid_targets_mean": 2229.1, + "valid_targets_min": 659 + }, + { + "epoch": 6.099706744868035, + "grad_norm": 0.813390106542457, + "learning_rate": 1.986670666756234e-06, + "loss": 0.1407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1278359740972519, + "step": 6240, + "valid_targets_mean": 2046.5, + "valid_targets_min": 814 + }, + { + "epoch": 6.104594330400782, + "grad_norm": 0.8400309752046771, + "learning_rate": 1.9655408256899713e-06, + "loss": 0.1385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15281391143798828, + "step": 6245, + "valid_targets_mean": 2291.9, + "valid_targets_min": 733 + }, + { + "epoch": 6.109481915933529, + "grad_norm": 0.7250372273390747, + "learning_rate": 1.944518144350709e-06, + "loss": 0.1456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1458112895488739, + "step": 6250, + "valid_targets_mean": 2883.2, + "valid_targets_min": 804 + }, + { + "epoch": 6.114369501466276, + "grad_norm": 0.8632663254658965, + "learning_rate": 1.9236027476540276e-06, + "loss": 0.1397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14336003363132477, + "step": 6255, + "valid_targets_mean": 2101.6, + "valid_targets_min": 817 + }, + { + "epoch": 6.119257086999022, + "grad_norm": 0.7798497119811234, + "learning_rate": 1.9027947598780127e-06, + "loss": 0.127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14085283875465393, + "step": 6260, + "valid_targets_mean": 2175.4, + "valid_targets_min": 714 + }, + { + "epoch": 6.124144672531769, + "grad_norm": 0.8683686286891048, + "learning_rate": 1.8820943046625429e-06, + "loss": 0.144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12134305387735367, + "step": 6265, + "valid_targets_mean": 1859.7, + "valid_targets_min": 776 + }, + { + "epoch": 6.129032258064516, + "grad_norm": 0.8639340556226129, + "learning_rate": 1.861501505008545e-06, + "loss": 0.1397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14260660111904144, + "step": 6270, + "valid_targets_mean": 2105.2, + "valid_targets_min": 905 + }, + { + "epoch": 6.133919843597263, + "grad_norm": 0.7652388086076658, + "learning_rate": 1.8410164832772536e-06, + "loss": 0.1365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1462109088897705, + "step": 6275, + "valid_targets_mean": 2521.6, + "valid_targets_min": 689 + }, + { + "epoch": 6.13880742913001, + "grad_norm": 0.8599010636007501, + "learning_rate": 1.8206393611895045e-06, + "loss": 0.1552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15619584918022156, + "step": 6280, + "valid_targets_mean": 2249.8, + "valid_targets_min": 1113 + }, + { + "epoch": 6.1436950146627565, + "grad_norm": 0.8398925161179197, + "learning_rate": 1.8003702598249861e-06, + "loss": 0.1447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14089694619178772, + "step": 6285, + "valid_targets_mean": 2057.7, + "valid_targets_min": 1225 + }, + { + "epoch": 6.148582600195503, + "grad_norm": 0.9006101111923522, + "learning_rate": 1.780209299621547e-06, + "loss": 0.1417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1570209264755249, + "step": 6290, + "valid_targets_mean": 2732.9, + "valid_targets_min": 922 + }, + { + "epoch": 6.15347018572825, + "grad_norm": 0.7691620511866091, + "learning_rate": 1.7601566003744631e-06, + "loss": 0.1364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1531471312046051, + "step": 6295, + "valid_targets_mean": 2870.6, + "valid_targets_min": 947 + }, + { + "epoch": 6.158357771260997, + "grad_norm": 0.6860071399944331, + "learning_rate": 1.740212281235727e-06, + "loss": 0.1378, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13493847846984863, + "step": 6300, + "valid_targets_mean": 3376.0, + "valid_targets_min": 753 + }, + { + "epoch": 6.163245356793744, + "grad_norm": 0.7541769441863807, + "learning_rate": 1.7203764607133377e-06, + "loss": 0.155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15232732892036438, + "step": 6305, + "valid_targets_mean": 2786.0, + "valid_targets_min": 671 + }, + { + "epoch": 6.168132942326491, + "grad_norm": 0.7651389425645014, + "learning_rate": 1.7006492566706233e-06, + "loss": 0.1424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13763144612312317, + "step": 6310, + "valid_targets_mean": 2695.4, + "valid_targets_min": 1083 + }, + { + "epoch": 6.173020527859237, + "grad_norm": 0.8366808292024842, + "learning_rate": 1.6810307863254927e-06, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14807304739952087, + "step": 6315, + "valid_targets_mean": 2219.9, + "valid_targets_min": 827 + }, + { + "epoch": 6.177908113391984, + "grad_norm": 0.8457539648171782, + "learning_rate": 1.6615211662497822e-06, + "loss": 0.1458, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1576795130968094, + "step": 6320, + "valid_targets_mean": 2348.2, + "valid_targets_min": 757 + }, + { + "epoch": 6.182795698924731, + "grad_norm": 0.80734473246753, + "learning_rate": 1.6421205123685392e-06, + "loss": 0.1507, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13625971972942352, + "step": 6325, + "valid_targets_mean": 2120.9, + "valid_targets_min": 806 + }, + { + "epoch": 6.187683284457478, + "grad_norm": 0.9804288667902588, + "learning_rate": 1.6228289399593334e-06, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14446699619293213, + "step": 6330, + "valid_targets_mean": 1873.2, + "valid_targets_min": 888 + }, + { + "epoch": 6.192570869990225, + "grad_norm": 0.7529261993671538, + "learning_rate": 1.603646563651604e-06, + "loss": 0.1455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14811274409294128, + "step": 6335, + "valid_targets_mean": 2624.8, + "valid_targets_min": 796 + }, + { + "epoch": 6.1974584555229715, + "grad_norm": 0.7758608591461221, + "learning_rate": 1.5845734974259207e-06, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13113583624362946, + "step": 6340, + "valid_targets_mean": 2221.9, + "valid_targets_min": 900 + }, + { + "epoch": 6.202346041055718, + "grad_norm": 0.7977758438125597, + "learning_rate": 1.5656098546133658e-06, + "loss": 0.1496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15472404658794403, + "step": 6345, + "valid_targets_mean": 2333.7, + "valid_targets_min": 777 + }, + { + "epoch": 6.207233626588465, + "grad_norm": 0.7496625060594996, + "learning_rate": 1.5467557478948148e-06, + "loss": 0.1468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15827788412570953, + "step": 6350, + "valid_targets_mean": 2944.5, + "valid_targets_min": 659 + }, + { + "epoch": 6.212121212121212, + "grad_norm": 0.7617727567525714, + "learning_rate": 1.5280112893002908e-06, + "loss": 0.1433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13309329748153687, + "step": 6355, + "valid_targets_mean": 2364.3, + "valid_targets_min": 988 + }, + { + "epoch": 6.217008797653959, + "grad_norm": 0.7687353152394582, + "learning_rate": 1.5093765902083069e-06, + "loss": 0.1404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15288090705871582, + "step": 6360, + "valid_targets_mean": 2643.9, + "valid_targets_min": 697 + }, + { + "epoch": 6.221896383186706, + "grad_norm": 0.8946465205196672, + "learning_rate": 1.4908517613451778e-06, + "loss": 0.1476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14633601903915405, + "step": 6365, + "valid_targets_mean": 2086.7, + "valid_targets_min": 607 + }, + { + "epoch": 6.226783968719452, + "grad_norm": 0.8280125163578745, + "learning_rate": 1.472436912784374e-06, + "loss": 0.1444, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13586950302124023, + "step": 6370, + "valid_targets_mean": 2219.1, + "valid_targets_min": 1104 + }, + { + "epoch": 6.231671554252199, + "grad_norm": 0.9360214005794244, + "learning_rate": 1.4541321539458775e-06, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14993751049041748, + "step": 6375, + "valid_targets_mean": 1659.2, + "valid_targets_min": 686 + }, + { + "epoch": 6.236559139784946, + "grad_norm": 0.762660547954432, + "learning_rate": 1.4359375935955222e-06, + "loss": 0.1484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1589202582836151, + "step": 6380, + "valid_targets_mean": 2716.4, + "valid_targets_min": 668 + }, + { + "epoch": 6.241446725317693, + "grad_norm": 0.8996541456011813, + "learning_rate": 1.4178533398443507e-06, + "loss": 0.1459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14994928240776062, + "step": 6385, + "valid_targets_mean": 1683.6, + "valid_targets_min": 731 + }, + { + "epoch": 6.24633431085044, + "grad_norm": 0.7402483966290739, + "learning_rate": 1.3998795001479605e-06, + "loss": 0.1401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15734237432479858, + "step": 6390, + "valid_targets_mean": 2690.1, + "valid_targets_min": 776 + }, + { + "epoch": 6.2512218963831865, + "grad_norm": 0.8314893917101926, + "learning_rate": 1.3820161813058786e-06, + "loss": 0.1452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1409616619348526, + "step": 6395, + "valid_targets_mean": 1984.4, + "valid_targets_min": 617 + }, + { + "epoch": 6.256109481915933, + "grad_norm": 0.7192255814903107, + "learning_rate": 1.3642634894609308e-06, + "loss": 0.1453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14671871066093445, + "step": 6400, + "valid_targets_mean": 2923.3, + "valid_targets_min": 567 + }, + { + "epoch": 6.26099706744868, + "grad_norm": 0.7999646416770593, + "learning_rate": 1.3466215300985996e-06, + "loss": 0.1406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13374221324920654, + "step": 6405, + "valid_targets_mean": 2253.6, + "valid_targets_min": 701 + }, + { + "epoch": 6.265884652981427, + "grad_norm": 0.7701307733257204, + "learning_rate": 1.3290904080464007e-06, + "loss": 0.1466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15439356863498688, + "step": 6410, + "valid_targets_mean": 2676.4, + "valid_targets_min": 869 + }, + { + "epoch": 6.270772238514174, + "grad_norm": 0.717925366016105, + "learning_rate": 1.3116702274732585e-06, + "loss": 0.1441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1328277885913849, + "step": 6415, + "valid_targets_mean": 2629.2, + "valid_targets_min": 888 + }, + { + "epoch": 6.275659824046921, + "grad_norm": 0.8180028485476835, + "learning_rate": 1.2943610918888893e-06, + "loss": 0.1385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15177467465400696, + "step": 6420, + "valid_targets_mean": 2332.1, + "valid_targets_min": 680 + }, + { + "epoch": 6.280547409579667, + "grad_norm": 0.8165966121783467, + "learning_rate": 1.277163104143191e-06, + "loss": 0.1394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14720627665519714, + "step": 6425, + "valid_targets_mean": 2315.4, + "valid_targets_min": 934 + }, + { + "epoch": 6.285434995112414, + "grad_norm": 0.8235251951027993, + "learning_rate": 1.2600763664256288e-06, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15732279419898987, + "step": 6430, + "valid_targets_mean": 2338.2, + "valid_targets_min": 719 + }, + { + "epoch": 6.290322580645161, + "grad_norm": 0.8537813636704863, + "learning_rate": 1.2431009802646177e-06, + "loss": 0.1495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15671692788600922, + "step": 6435, + "valid_targets_mean": 2427.2, + "valid_targets_min": 614 + }, + { + "epoch": 6.295210166177908, + "grad_norm": 0.7787776965511892, + "learning_rate": 1.2262370465269368e-06, + "loss": 0.1413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14795002341270447, + "step": 6440, + "valid_targets_mean": 2533.7, + "valid_targets_min": 867 + }, + { + "epoch": 6.300097751710655, + "grad_norm": 0.7466171619616006, + "learning_rate": 1.2094846654171133e-06, + "loss": 0.1438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1360429972410202, + "step": 6445, + "valid_targets_mean": 2684.0, + "valid_targets_min": 930 + }, + { + "epoch": 6.3049853372434015, + "grad_norm": 0.8453955068732553, + "learning_rate": 1.1928439364768418e-06, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1500052511692047, + "step": 6450, + "valid_targets_mean": 2271.9, + "valid_targets_min": 656 + }, + { + "epoch": 6.309872922776148, + "grad_norm": 0.7381224772665309, + "learning_rate": 1.176314958584388e-06, + "loss": 0.1472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1459953933954239, + "step": 6455, + "valid_targets_mean": 2728.8, + "valid_targets_min": 945 + }, + { + "epoch": 6.314760508308895, + "grad_norm": 0.7664923578206903, + "learning_rate": 1.1598978299539886e-06, + "loss": 0.1374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11979406327009201, + "step": 6460, + "valid_targets_mean": 2113.8, + "valid_targets_min": 740 + }, + { + "epoch": 6.319648093841642, + "grad_norm": 0.774864233400857, + "learning_rate": 1.143592648135292e-06, + "loss": 0.1475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13750594854354858, + "step": 6465, + "valid_targets_mean": 2761.6, + "valid_targets_min": 847 + }, + { + "epoch": 6.324535679374389, + "grad_norm": 0.887880805842084, + "learning_rate": 1.1273995100127478e-06, + "loss": 0.1547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1429017186164856, + "step": 6470, + "valid_targets_mean": 1817.9, + "valid_targets_min": 739 + }, + { + "epoch": 6.329423264907136, + "grad_norm": 0.7426599652356065, + "learning_rate": 1.1113185118050662e-06, + "loss": 0.1451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1407342255115509, + "step": 6475, + "valid_targets_mean": 2687.5, + "valid_targets_min": 830 + }, + { + "epoch": 6.334310850439882, + "grad_norm": 0.8069295804769829, + "learning_rate": 1.0953497490646204e-06, + "loss": 0.1356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1433069258928299, + "step": 6480, + "valid_targets_mean": 2190.9, + "valid_targets_min": 635 + }, + { + "epoch": 6.339198435972629, + "grad_norm": 0.7177391118014483, + "learning_rate": 1.0794933166768828e-06, + "loss": 0.1408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1430281698703766, + "step": 6485, + "valid_targets_mean": 3095.9, + "valid_targets_min": 770 + }, + { + "epoch": 6.344086021505376, + "grad_norm": 0.7687796424872444, + "learning_rate": 1.063749308859876e-06, + "loss": 0.1395, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13334771990776062, + "step": 6490, + "valid_targets_mean": 2388.4, + "valid_targets_min": 966 + }, + { + "epoch": 6.348973607038123, + "grad_norm": 0.7612117876020217, + "learning_rate": 1.0481178191635876e-06, + "loss": 0.1404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1389297991991043, + "step": 6495, + "valid_targets_mean": 2652.2, + "valid_targets_min": 777 + }, + { + "epoch": 6.35386119257087, + "grad_norm": 0.7514521694555615, + "learning_rate": 1.0325989404694424e-06, + "loss": 0.1563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16383954882621765, + "step": 6500, + "valid_targets_mean": 2935.7, + "valid_targets_min": 1179 + }, + { + "epoch": 6.3587487781036165, + "grad_norm": 0.8203426377026969, + "learning_rate": 1.017192764989734e-06, + "loss": 0.15, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15453344583511353, + "step": 6505, + "valid_targets_mean": 2347.6, + "valid_targets_min": 721 + }, + { + "epoch": 6.363636363636363, + "grad_norm": 0.8292620788591387, + "learning_rate": 1.001899384267071e-06, + "loss": 0.1476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15466228127479553, + "step": 6510, + "valid_targets_mean": 2221.9, + "valid_targets_min": 750 + }, + { + "epoch": 6.36852394916911, + "grad_norm": 0.829931504832341, + "learning_rate": 9.867188891738544e-07, + "loss": 0.1325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13335032761096954, + "step": 6515, + "valid_targets_mean": 2455.7, + "valid_targets_min": 1143 + }, + { + "epoch": 6.373411534701857, + "grad_norm": 0.7012016461939861, + "learning_rate": 9.716513699117148e-07, + "loss": 0.1406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12033946067094803, + "step": 6520, + "valid_targets_mean": 2587.4, + "valid_targets_min": 1035 + }, + { + "epoch": 6.378299120234604, + "grad_norm": 0.7608167779950807, + "learning_rate": 9.56696916010995e-07, + "loss": 0.1481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15372420847415924, + "step": 6525, + "valid_targets_mean": 2700.6, + "valid_targets_min": 1255 + }, + { + "epoch": 6.383186705767351, + "grad_norm": 0.7875236266722555, + "learning_rate": 9.418556163302073e-07, + "loss": 0.1376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1496465802192688, + "step": 6530, + "valid_targets_mean": 2373.9, + "valid_targets_min": 950 + }, + { + "epoch": 6.388074291300097, + "grad_norm": 0.9345295844327403, + "learning_rate": 9.271275590555073e-07, + "loss": 0.1383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13679920136928558, + "step": 6535, + "valid_targets_mean": 2371.4, + "valid_targets_min": 744 + }, + { + "epoch": 6.392961876832844, + "grad_norm": 0.8165642457777276, + "learning_rate": 9.125128317001653e-07, + "loss": 0.1456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14877143502235413, + "step": 6540, + "valid_targets_mean": 2413.6, + "valid_targets_min": 692 + }, + { + "epoch": 6.397849462365591, + "grad_norm": 0.8293551530994266, + "learning_rate": 8.980115211040607e-07, + "loss": 0.1359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12983205914497375, + "step": 6545, + "valid_targets_mean": 1852.1, + "valid_targets_min": 635 + }, + { + "epoch": 6.402737047898338, + "grad_norm": 0.8185584829200172, + "learning_rate": 8.836237134331527e-07, + "loss": 0.1485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1399831473827362, + "step": 6550, + "valid_targets_mean": 2156.9, + "valid_targets_min": 878 + }, + { + "epoch": 6.407624633431085, + "grad_norm": 0.8752733032904388, + "learning_rate": 8.693494941789748e-07, + "loss": 0.1352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14943121373653412, + "step": 6555, + "valid_targets_mean": 2125.4, + "valid_targets_min": 791 + }, + { + "epoch": 6.4125122189638315, + "grad_norm": 0.771428128135249, + "learning_rate": 8.551889481581255e-07, + "loss": 0.1494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13312771916389465, + "step": 6560, + "valid_targets_mean": 2736.9, + "valid_targets_min": 1206 + }, + { + "epoch": 6.417399804496578, + "grad_norm": 0.8363887006600691, + "learning_rate": 8.411421595117519e-07, + "loss": 0.1368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1484169363975525, + "step": 6565, + "valid_targets_mean": 2258.8, + "valid_targets_min": 612 + }, + { + "epoch": 6.422287390029325, + "grad_norm": 0.841896164198332, + "learning_rate": 8.272092117050778e-07, + "loss": 0.1485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14031308889389038, + "step": 6570, + "valid_targets_mean": 2081.2, + "valid_targets_min": 741 + }, + { + "epoch": 6.427174975562072, + "grad_norm": 0.8344208378276725, + "learning_rate": 8.133901875268813e-07, + "loss": 0.142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16051240265369415, + "step": 6575, + "valid_targets_mean": 2340.0, + "valid_targets_min": 793 + }, + { + "epoch": 6.432062561094819, + "grad_norm": 0.739011240056262, + "learning_rate": 7.996851690890195e-07, + "loss": 0.1514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1327826976776123, + "step": 6580, + "valid_targets_mean": 2467.3, + "valid_targets_min": 1040 + }, + { + "epoch": 6.436950146627566, + "grad_norm": 0.8388926850354799, + "learning_rate": 7.860942378259251e-07, + "loss": 0.1429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13930006325244904, + "step": 6585, + "valid_targets_mean": 2314.1, + "valid_targets_min": 687 + }, + { + "epoch": 6.441837732160312, + "grad_norm": 0.736849529441786, + "learning_rate": 7.726174744941351e-07, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1561793088912964, + "step": 6590, + "valid_targets_mean": 3114.3, + "valid_targets_min": 1144 + }, + { + "epoch": 6.446725317693059, + "grad_norm": 0.7374996366190613, + "learning_rate": 7.592549591718091e-07, + "loss": 0.1466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14413076639175415, + "step": 6595, + "valid_targets_mean": 3133.8, + "valid_targets_min": 1182 + }, + { + "epoch": 6.451612903225806, + "grad_norm": 0.8220802356280166, + "learning_rate": 7.460067712582519e-07, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1294611543416977, + "step": 6600, + "valid_targets_mean": 2196.5, + "valid_targets_min": 712 + }, + { + "epoch": 6.456500488758553, + "grad_norm": 0.7893984126724283, + "learning_rate": 7.328729894734388e-07, + "loss": 0.1505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1567493975162506, + "step": 6605, + "valid_targets_mean": 2438.6, + "valid_targets_min": 703 + }, + { + "epoch": 6.4613880742913, + "grad_norm": 0.8117586298458785, + "learning_rate": 7.198536918575527e-07, + "loss": 0.1492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1428253948688507, + "step": 6610, + "valid_targets_mean": 2303.7, + "valid_targets_min": 632 + }, + { + "epoch": 6.4662756598240465, + "grad_norm": 0.7826755426811578, + "learning_rate": 7.069489557705145e-07, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15663695335388184, + "step": 6615, + "valid_targets_mean": 2520.6, + "valid_targets_min": 655 + }, + { + "epoch": 6.471163245356793, + "grad_norm": 0.8284462615884153, + "learning_rate": 6.941588578915315e-07, + "loss": 0.1491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1673087477684021, + "step": 6620, + "valid_targets_mean": 2560.1, + "valid_targets_min": 1093 + }, + { + "epoch": 6.476050830889541, + "grad_norm": 0.6413200188770708, + "learning_rate": 6.814834742186361e-07, + "loss": 0.148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14878563582897186, + "step": 6625, + "valid_targets_mean": 4005.8, + "valid_targets_min": 696 + }, + { + "epoch": 6.480938416422287, + "grad_norm": 0.7615467726204301, + "learning_rate": 6.689228800682301e-07, + "loss": 0.1464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14344452321529388, + "step": 6630, + "valid_targets_mean": 2507.8, + "valid_targets_min": 899 + }, + { + "epoch": 6.485826001955035, + "grad_norm": 0.7837667879512825, + "learning_rate": 6.564771500746525e-07, + "loss": 0.1442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13829047977924347, + "step": 6635, + "valid_targets_mean": 2337.0, + "valid_targets_min": 800 + }, + { + "epoch": 6.490713587487781, + "grad_norm": 0.7681966736659308, + "learning_rate": 6.441463581897167e-07, + "loss": 0.1438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1388791799545288, + "step": 6640, + "valid_targets_mean": 2536.2, + "valid_targets_min": 960 + }, + { + "epoch": 6.495601173020528, + "grad_norm": 0.7657991833749364, + "learning_rate": 6.319305776822848e-07, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13757102191448212, + "step": 6645, + "valid_targets_mean": 2367.8, + "valid_targets_min": 720 + }, + { + "epoch": 6.500488758553274, + "grad_norm": 0.8617641880740021, + "learning_rate": 6.198298811378278e-07, + "loss": 0.1459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1516299545764923, + "step": 6650, + "valid_targets_mean": 2099.6, + "valid_targets_min": 627 + }, + { + "epoch": 6.505376344086022, + "grad_norm": 0.8509825074669289, + "learning_rate": 6.078443404579948e-07, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.160349503159523, + "step": 6655, + "valid_targets_mean": 2118.5, + "valid_targets_min": 764 + }, + { + "epoch": 6.510263929618768, + "grad_norm": 0.7385237771877197, + "learning_rate": 5.959740268601843e-07, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14553092420101166, + "step": 6660, + "valid_targets_mean": 2765.1, + "valid_targets_min": 799 + }, + { + "epoch": 6.515151515151516, + "grad_norm": 0.7414225190431922, + "learning_rate": 5.842190108771206e-07, + "loss": 0.1494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.158996120095253, + "step": 6665, + "valid_targets_mean": 2852.4, + "valid_targets_min": 1218 + }, + { + "epoch": 6.5200391006842615, + "grad_norm": 0.9920100082894623, + "learning_rate": 5.725793623564379e-07, + "loss": 0.1446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.143795907497406, + "step": 6670, + "valid_targets_mean": 2288.8, + "valid_targets_min": 981 + }, + { + "epoch": 6.524926686217009, + "grad_norm": 0.7440670427532977, + "learning_rate": 5.610551504602657e-07, + "loss": 0.1561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1507560908794403, + "step": 6675, + "valid_targets_mean": 2652.2, + "valid_targets_min": 921 + }, + { + "epoch": 6.529814271749755, + "grad_norm": 0.8286332715981629, + "learning_rate": 5.496464436648108e-07, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13288316130638123, + "step": 6680, + "valid_targets_mean": 1858.8, + "valid_targets_min": 777 + }, + { + "epoch": 6.534701857282503, + "grad_norm": 0.8232783710376684, + "learning_rate": 5.383533097599558e-07, + "loss": 0.1412, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13944336771965027, + "step": 6685, + "valid_targets_mean": 2516.9, + "valid_targets_min": 925 + }, + { + "epoch": 6.539589442815249, + "grad_norm": 0.7654399610127802, + "learning_rate": 5.271758158488638e-07, + "loss": 0.1456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13826604187488556, + "step": 6690, + "valid_targets_mean": 2636.9, + "valid_targets_min": 902 + }, + { + "epoch": 6.5444770283479965, + "grad_norm": 0.8184911988154895, + "learning_rate": 5.161140283475608e-07, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1462034285068512, + "step": 6695, + "valid_targets_mean": 2207.8, + "valid_targets_min": 710 + }, + { + "epoch": 6.549364613880742, + "grad_norm": 0.7337833239715681, + "learning_rate": 5.051680129845605e-07, + "loss": 0.1366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12600910663604736, + "step": 6700, + "valid_targets_mean": 2454.8, + "valid_targets_min": 864 + }, + { + "epoch": 6.55425219941349, + "grad_norm": 0.8445779401373558, + "learning_rate": 4.94337834800458e-07, + "loss": 0.144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1528075784444809, + "step": 6705, + "valid_targets_mean": 2297.4, + "valid_targets_min": 716 + }, + { + "epoch": 6.559139784946236, + "grad_norm": 0.7101947557300763, + "learning_rate": 4.836235581475523e-07, + "loss": 0.1381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12695491313934326, + "step": 6710, + "valid_targets_mean": 2736.0, + "valid_targets_min": 801 + }, + { + "epoch": 6.564027370478984, + "grad_norm": 0.8639233831130482, + "learning_rate": 4.7302524668946916e-07, + "loss": 0.1365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14430125057697296, + "step": 6715, + "valid_targets_mean": 2032.4, + "valid_targets_min": 846 + }, + { + "epoch": 6.568914956011731, + "grad_norm": 0.7461099525991259, + "learning_rate": 4.6254296340076497e-07, + "loss": 0.1361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13411906361579895, + "step": 6720, + "valid_targets_mean": 2808.7, + "valid_targets_min": 648 + }, + { + "epoch": 6.573802541544477, + "grad_norm": 0.7989702744987931, + "learning_rate": 4.521767705665747e-07, + "loss": 0.1408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15190516412258148, + "step": 6725, + "valid_targets_mean": 2514.4, + "valid_targets_min": 740 + }, + { + "epoch": 6.578690127077224, + "grad_norm": 0.8220944285398598, + "learning_rate": 4.419267297822205e-07, + "loss": 0.1471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1572040170431137, + "step": 6730, + "valid_targets_mean": 2357.9, + "valid_targets_min": 790 + }, + { + "epoch": 6.583577712609971, + "grad_norm": 0.6585139462243008, + "learning_rate": 4.317929019528566e-07, + "loss": 0.1451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13705238699913025, + "step": 6735, + "valid_targets_mean": 3228.4, + "valid_targets_min": 752 + }, + { + "epoch": 6.588465298142718, + "grad_norm": 0.8169152817627165, + "learning_rate": 4.217753472931141e-07, + "loss": 0.1393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.139057457447052, + "step": 6740, + "valid_targets_mean": 2172.5, + "valid_targets_min": 640 + }, + { + "epoch": 6.593352883675465, + "grad_norm": 0.7809355920094504, + "learning_rate": 4.118741253267322e-07, + "loss": 0.1459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14593061804771423, + "step": 6745, + "valid_targets_mean": 2661.4, + "valid_targets_min": 780 + }, + { + "epoch": 6.5982404692082115, + "grad_norm": 0.7768920952770774, + "learning_rate": 4.020892948862032e-07, + "loss": 0.1566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13564547896385193, + "step": 6750, + "valid_targets_mean": 2473.9, + "valid_targets_min": 537 + }, + { + "epoch": 6.603128054740958, + "grad_norm": 0.7722557323473561, + "learning_rate": 3.9242091411243245e-07, + "loss": 0.1459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1392875760793686, + "step": 6755, + "valid_targets_mean": 2471.2, + "valid_targets_min": 790 + }, + { + "epoch": 6.608015640273705, + "grad_norm": 0.7539682369302184, + "learning_rate": 3.8286904045438553e-07, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1369071751832962, + "step": 6760, + "valid_targets_mean": 2399.9, + "valid_targets_min": 610 + }, + { + "epoch": 6.612903225806452, + "grad_norm": 0.6424299073579681, + "learning_rate": 3.73433730668753e-07, + "loss": 0.1472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13768397271633148, + "step": 6765, + "valid_targets_mean": 3869.8, + "valid_targets_min": 1204 + }, + { + "epoch": 6.617790811339199, + "grad_norm": 0.6880284888599075, + "learning_rate": 3.641150408196037e-07, + "loss": 0.1459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12130261957645416, + "step": 6770, + "valid_targets_mean": 2982.4, + "valid_targets_min": 815 + }, + { + "epoch": 6.622678396871946, + "grad_norm": 0.8219148218793978, + "learning_rate": 3.5491302627806314e-07, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15362417697906494, + "step": 6775, + "valid_targets_mean": 2819.4, + "valid_targets_min": 919 + }, + { + "epoch": 6.627565982404692, + "grad_norm": 0.7921515154766982, + "learning_rate": 3.4582774172197576e-07, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13740390539169312, + "step": 6780, + "valid_targets_mean": 2345.8, + "valid_targets_min": 790 + }, + { + "epoch": 6.632453567937439, + "grad_norm": 0.7739946089387783, + "learning_rate": 3.368592411355831e-07, + "loss": 0.1392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1267765611410141, + "step": 6785, + "valid_targets_mean": 2384.4, + "valid_targets_min": 771 + }, + { + "epoch": 6.637341153470186, + "grad_norm": 0.7562090861343093, + "learning_rate": 3.2800757780920846e-07, + "loss": 0.1443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1409822702407837, + "step": 6790, + "valid_targets_mean": 2635.0, + "valid_targets_min": 594 + }, + { + "epoch": 6.642228739002933, + "grad_norm": 0.8014153886361279, + "learning_rate": 3.192728043389237e-07, + "loss": 0.15, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15787863731384277, + "step": 6795, + "valid_targets_mean": 2474.6, + "valid_targets_min": 1065 + }, + { + "epoch": 6.64711632453568, + "grad_norm": 0.8105759791663855, + "learning_rate": 3.106549726262542e-07, + "loss": 0.1445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15971817076206207, + "step": 6800, + "valid_targets_mean": 2614.4, + "valid_targets_min": 1129 + }, + { + "epoch": 6.6520039100684265, + "grad_norm": 0.7732426818990713, + "learning_rate": 3.0215413387787e-07, + "loss": 0.1406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1353302150964737, + "step": 6805, + "valid_targets_mean": 2363.2, + "valid_targets_min": 725 + }, + { + "epoch": 6.656891495601173, + "grad_norm": 0.6933846730883887, + "learning_rate": 2.937703386052637e-07, + "loss": 0.1379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12826628983020782, + "step": 6810, + "valid_targets_mean": 3215.6, + "valid_targets_min": 821 + }, + { + "epoch": 6.66177908113392, + "grad_norm": 0.8557865157908506, + "learning_rate": 2.8550363662447347e-07, + "loss": 0.1435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1560901701450348, + "step": 6815, + "valid_targets_mean": 2224.4, + "valid_targets_min": 803 + }, + { + "epoch": 6.666666666666667, + "grad_norm": 0.8732910114888374, + "learning_rate": 2.77354077055767e-07, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.145709827542305, + "step": 6820, + "valid_targets_mean": 2143.8, + "valid_targets_min": 729 + }, + { + "epoch": 6.671554252199414, + "grad_norm": 0.780368314719552, + "learning_rate": 2.6932170832336016e-07, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14553119242191315, + "step": 6825, + "valid_targets_mean": 2521.6, + "valid_targets_min": 934 + }, + { + "epoch": 6.676441837732161, + "grad_norm": 0.7148221396324418, + "learning_rate": 2.614065781551345e-07, + "loss": 0.1503, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1638277769088745, + "step": 6830, + "valid_targets_mean": 3283.7, + "valid_targets_min": 1090 + }, + { + "epoch": 6.681329423264907, + "grad_norm": 0.7948993329236099, + "learning_rate": 2.5360873358234004e-07, + "loss": 0.1419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14200204610824585, + "step": 6835, + "valid_targets_mean": 2449.1, + "valid_targets_min": 738 + }, + { + "epoch": 6.686217008797654, + "grad_norm": 0.7070697581069597, + "learning_rate": 2.459282209393243e-07, + "loss": 0.1456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1273440271615982, + "step": 6840, + "valid_targets_mean": 2714.1, + "valid_targets_min": 793 + }, + { + "epoch": 6.691104594330401, + "grad_norm": 0.804428556746242, + "learning_rate": 2.3836508586325246e-07, + "loss": 0.134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13834835588932037, + "step": 6845, + "valid_targets_mean": 2142.0, + "valid_targets_min": 684 + }, + { + "epoch": 6.695992179863148, + "grad_norm": 0.8003383717639464, + "learning_rate": 2.309193732938453e-07, + "loss": 0.135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15829630196094513, + "step": 6850, + "valid_targets_mean": 2366.2, + "valid_targets_min": 744 + }, + { + "epoch": 6.700879765395895, + "grad_norm": 0.8634766410996613, + "learning_rate": 2.235911274730973e-07, + "loss": 0.1436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16072383522987366, + "step": 6855, + "valid_targets_mean": 2273.4, + "valid_targets_min": 680 + }, + { + "epoch": 6.7057673509286415, + "grad_norm": 0.8555538490006295, + "learning_rate": 2.1638039194503246e-07, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15848436951637268, + "step": 6860, + "valid_targets_mean": 2117.4, + "valid_targets_min": 718 + }, + { + "epoch": 6.710654936461388, + "grad_norm": 0.8279297269051833, + "learning_rate": 2.0928720955542436e-07, + "loss": 0.1427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14095677435398102, + "step": 6865, + "valid_targets_mean": 2048.8, + "valid_targets_min": 754 + }, + { + "epoch": 6.715542521994135, + "grad_norm": 0.7949216593255324, + "learning_rate": 2.0231162245156088e-07, + "loss": 0.1456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15973398089408875, + "step": 6870, + "valid_targets_mean": 2616.4, + "valid_targets_min": 898 + }, + { + "epoch": 6.720430107526882, + "grad_norm": 0.7776529385814106, + "learning_rate": 1.9545367208197995e-07, + "loss": 0.1435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13128122687339783, + "step": 6875, + "valid_targets_mean": 2468.4, + "valid_targets_min": 773 + }, + { + "epoch": 6.725317693059629, + "grad_norm": 0.7470593281900991, + "learning_rate": 1.887133991962342e-07, + "loss": 0.1472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13984917104244232, + "step": 6880, + "valid_targets_mean": 2672.4, + "valid_targets_min": 918 + }, + { + "epoch": 6.730205278592376, + "grad_norm": 0.8327481491216713, + "learning_rate": 1.8209084384464005e-07, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14572235941886902, + "step": 6885, + "valid_targets_mean": 2151.9, + "valid_targets_min": 812 + }, + { + "epoch": 6.735092864125122, + "grad_norm": 0.7666404720709251, + "learning_rate": 1.7558604537804224e-07, + "loss": 0.1429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1454159915447235, + "step": 6890, + "valid_targets_mean": 2545.7, + "valid_targets_min": 759 + }, + { + "epoch": 6.739980449657869, + "grad_norm": 0.7742410628801816, + "learning_rate": 1.6919904244758311e-07, + "loss": 0.1486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15060219168663025, + "step": 6895, + "valid_targets_mean": 2724.8, + "valid_targets_min": 826 + }, + { + "epoch": 6.744868035190616, + "grad_norm": 0.8696550747312377, + "learning_rate": 1.6292987300446705e-07, + "loss": 0.1423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14360955357551575, + "step": 6900, + "valid_targets_mean": 1997.2, + "valid_targets_min": 688 + }, + { + "epoch": 6.749755620723363, + "grad_norm": 0.669193884708062, + "learning_rate": 1.5677857429974296e-07, + "loss": 0.1462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12194126844406128, + "step": 6905, + "valid_targets_mean": 2778.8, + "valid_targets_min": 753 + }, + { + "epoch": 6.75464320625611, + "grad_norm": 0.8129272991005836, + "learning_rate": 1.507451828840756e-07, + "loss": 0.1381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12596674263477325, + "step": 6910, + "valid_targets_mean": 2184.9, + "valid_targets_min": 533 + }, + { + "epoch": 6.7595307917888565, + "grad_norm": 0.8272536163498828, + "learning_rate": 1.4482973460753446e-07, + "loss": 0.1525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16095013916492462, + "step": 6915, + "valid_targets_mean": 2635.4, + "valid_targets_min": 1072 + }, + { + "epoch": 6.764418377321603, + "grad_norm": 0.8546760713602463, + "learning_rate": 1.3903226461937424e-07, + "loss": 0.1398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1363896131515503, + "step": 6920, + "valid_targets_mean": 2176.2, + "valid_targets_min": 663 + }, + { + "epoch": 6.76930596285435, + "grad_norm": 0.7403266010437409, + "learning_rate": 1.3335280736783028e-07, + "loss": 0.1459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15206800401210785, + "step": 6925, + "valid_targets_mean": 2854.7, + "valid_targets_min": 1169 + }, + { + "epoch": 6.774193548387097, + "grad_norm": 0.7772690818458694, + "learning_rate": 1.2779139659991224e-07, + "loss": 0.135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1384185254573822, + "step": 6930, + "valid_targets_mean": 2430.2, + "valid_targets_min": 708 + }, + { + "epoch": 6.779081133919844, + "grad_norm": 0.8678197977791959, + "learning_rate": 1.2234806536120857e-07, + "loss": 0.1488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1501295566558838, + "step": 6935, + "valid_targets_mean": 1968.8, + "valid_targets_min": 832 + }, + { + "epoch": 6.783968719452591, + "grad_norm": 1.006053072526263, + "learning_rate": 1.1702284599568014e-07, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15837866067886353, + "step": 6940, + "valid_targets_mean": 1922.8, + "valid_targets_min": 692 + }, + { + "epoch": 6.788856304985337, + "grad_norm": 0.7781597084300925, + "learning_rate": 1.1181577014547807e-07, + "loss": 0.144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14804880321025848, + "step": 6945, + "valid_targets_mean": 2636.6, + "valid_targets_min": 874 + }, + { + "epoch": 6.793743890518084, + "grad_norm": 0.7361777760679318, + "learning_rate": 1.0672686875074834e-07, + "loss": 0.1425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12848952412605286, + "step": 6950, + "valid_targets_mean": 2487.4, + "valid_targets_min": 880 + }, + { + "epoch": 6.798631476050831, + "grad_norm": 0.9048873371832021, + "learning_rate": 1.0175617204945421e-07, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1462620347738266, + "step": 6955, + "valid_targets_mean": 1957.2, + "valid_targets_min": 802 + }, + { + "epoch": 6.803519061583578, + "grad_norm": 0.8115943754772563, + "learning_rate": 9.690370957718965e-08, + "loss": 0.1342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14204561710357666, + "step": 6960, + "valid_targets_mean": 2568.1, + "valid_targets_min": 692 + }, + { + "epoch": 6.808406647116325, + "grad_norm": 0.7847252093602829, + "learning_rate": 9.216951016701281e-08, + "loss": 0.1462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1391274333000183, + "step": 6965, + "valid_targets_mean": 2426.6, + "valid_targets_min": 820 + }, + { + "epoch": 6.8132942326490715, + "grad_norm": 0.7599174927993604, + "learning_rate": 8.755360194926399e-08, + "loss": 0.1344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15206477046012878, + "step": 6970, + "valid_targets_mean": 2830.7, + "valid_targets_min": 847 + }, + { + "epoch": 6.818181818181818, + "grad_norm": 0.7366744359812574, + "learning_rate": 8.30560123514057e-08, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14298483729362488, + "step": 6975, + "valid_targets_mean": 2676.4, + "valid_targets_min": 1298 + }, + { + "epoch": 6.823069403714565, + "grad_norm": 0.7873925943117703, + "learning_rate": 7.867676809786284e-08, + "loss": 0.1449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14373791217803955, + "step": 6980, + "valid_targets_mean": 2348.8, + "valid_targets_min": 737 + }, + { + "epoch": 6.827956989247312, + "grad_norm": 0.7786043991826431, + "learning_rate": 7.441589520984726e-08, + "loss": 0.1604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1594143807888031, + "step": 6985, + "valid_targets_mean": 2935.5, + "valid_targets_min": 921 + }, + { + "epoch": 6.832844574780059, + "grad_norm": 0.7362339661403121, + "learning_rate": 7.027341900523122e-08, + "loss": 0.1398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1562596559524536, + "step": 6990, + "valid_targets_mean": 3039.9, + "valid_targets_min": 1045 + }, + { + "epoch": 6.837732160312806, + "grad_norm": 0.8183556025387007, + "learning_rate": 6.624936409836524e-08, + "loss": 0.1482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15503844618797302, + "step": 6995, + "valid_targets_mean": 2337.7, + "valid_targets_min": 727 + }, + { + "epoch": 6.842619745845552, + "grad_norm": 0.7856265162452774, + "learning_rate": 6.23437543999561e-08, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1516461968421936, + "step": 7000, + "valid_targets_mean": 2516.7, + "valid_targets_min": 1013 + }, + { + "epoch": 6.847507331378299, + "grad_norm": 0.7797885305733583, + "learning_rate": 5.855661311691574e-08, + "loss": 0.1497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16110171377658844, + "step": 7005, + "valid_targets_mean": 2726.2, + "valid_targets_min": 990 + }, + { + "epoch": 6.852394916911046, + "grad_norm": 0.8996911264100806, + "learning_rate": 5.4887962752216975e-08, + "loss": 0.1373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14116908609867096, + "step": 7010, + "valid_targets_mean": 2133.9, + "valid_targets_min": 967 + }, + { + "epoch": 6.857282502443793, + "grad_norm": 0.7855480605188909, + "learning_rate": 5.1337825104775805e-08, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13040246069431305, + "step": 7015, + "valid_targets_mean": 2354.3, + "valid_targets_min": 815 + }, + { + "epoch": 6.86217008797654, + "grad_norm": 0.778868414355739, + "learning_rate": 4.790622126930489e-08, + "loss": 0.1468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16201795637607574, + "step": 7020, + "valid_targets_mean": 2563.2, + "valid_targets_min": 1043 + }, + { + "epoch": 6.8670576735092865, + "grad_norm": 0.8949061085156262, + "learning_rate": 4.459317163619803e-08, + "loss": 0.1387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1284133791923523, + "step": 7025, + "valid_targets_mean": 1841.8, + "valid_targets_min": 958 + }, + { + "epoch": 6.871945259042033, + "grad_norm": 0.9097660526010823, + "learning_rate": 4.13986958914081e-08, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15148180723190308, + "step": 7030, + "valid_targets_mean": 1931.8, + "valid_targets_min": 679 + }, + { + "epoch": 6.87683284457478, + "grad_norm": 0.8079953101764225, + "learning_rate": 3.832281301632712e-08, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1472192108631134, + "step": 7035, + "valid_targets_mean": 2672.9, + "valid_targets_min": 670 + }, + { + "epoch": 6.881720430107527, + "grad_norm": 0.8377820814625377, + "learning_rate": 3.536554128767522e-08, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11620961874723434, + "step": 7040, + "valid_targets_mean": 1944.7, + "valid_targets_min": 917 + }, + { + "epoch": 6.886608015640274, + "grad_norm": 0.9780072356339135, + "learning_rate": 3.252689827739186e-08, + "loss": 0.1492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13305175304412842, + "step": 7045, + "valid_targets_mean": 1505.8, + "valid_targets_min": 921 + }, + { + "epoch": 6.891495601173021, + "grad_norm": 0.7630678184839079, + "learning_rate": 2.980690085253368e-08, + "loss": 0.1486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14864769577980042, + "step": 7050, + "valid_targets_mean": 2593.9, + "valid_targets_min": 625 + }, + { + "epoch": 6.896383186705767, + "grad_norm": 0.7665055215312272, + "learning_rate": 2.7205565175167925e-08, + "loss": 0.1484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14561477303504944, + "step": 7055, + "valid_targets_mean": 3042.3, + "valid_targets_min": 963 + }, + { + "epoch": 6.901270772238514, + "grad_norm": 0.7931423401009772, + "learning_rate": 2.472290670228361e-08, + "loss": 0.143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14829900860786438, + "step": 7060, + "valid_targets_mean": 2566.6, + "valid_targets_min": 672 + }, + { + "epoch": 6.906158357771261, + "grad_norm": 0.7348681150492757, + "learning_rate": 2.2358940185698285e-08, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15821123123168945, + "step": 7065, + "valid_targets_mean": 3052.2, + "valid_targets_min": 1009 + }, + { + "epoch": 6.911045943304008, + "grad_norm": 0.7851239048108627, + "learning_rate": 2.0113679671960317e-08, + "loss": 0.1524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15088459849357605, + "step": 7070, + "valid_targets_mean": 2557.6, + "valid_targets_min": 514 + }, + { + "epoch": 6.915933528836755, + "grad_norm": 0.7257205703096241, + "learning_rate": 1.7987138502284508e-08, + "loss": 0.1451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14671635627746582, + "step": 7075, + "valid_targets_mean": 2951.8, + "valid_targets_min": 1216 + }, + { + "epoch": 6.9208211143695015, + "grad_norm": 0.7158651863598146, + "learning_rate": 1.5979329312456603e-08, + "loss": 0.1498, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.156073197722435, + "step": 7080, + "valid_targets_mean": 3083.7, + "valid_targets_min": 987 + }, + { + "epoch": 6.925708699902248, + "grad_norm": 0.7570891695989537, + "learning_rate": 1.4090264032760037e-08, + "loss": 0.1387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.133244588971138, + "step": 7085, + "valid_targets_mean": 2580.4, + "valid_targets_min": 832 + }, + { + "epoch": 6.930596285434995, + "grad_norm": 0.8832686071227391, + "learning_rate": 1.2319953887918179e-08, + "loss": 0.142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13915324211120605, + "step": 7090, + "valid_targets_mean": 2498.4, + "valid_targets_min": 632 + }, + { + "epoch": 6.935483870967742, + "grad_norm": 0.7469142291910154, + "learning_rate": 1.0668409397009972e-08, + "loss": 0.1439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13527530431747437, + "step": 7095, + "valid_targets_mean": 3218.7, + "valid_targets_min": 794 + }, + { + "epoch": 6.940371456500489, + "grad_norm": 0.7764886081432477, + "learning_rate": 9.135640373418853e-09, + "loss": 0.1391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15149667859077454, + "step": 7100, + "valid_targets_mean": 2654.7, + "valid_targets_min": 1291 + }, + { + "epoch": 6.945259042033236, + "grad_norm": 0.792380332907017, + "learning_rate": 7.721655924770588e-09, + "loss": 0.142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14253517985343933, + "step": 7105, + "valid_targets_mean": 2680.3, + "valid_targets_min": 1109 + }, + { + "epoch": 6.9501466275659824, + "grad_norm": 0.8708382788168486, + "learning_rate": 6.426464452879977e-09, + "loss": 0.1543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15370717644691467, + "step": 7110, + "valid_targets_mean": 1872.2, + "valid_targets_min": 824 + }, + { + "epoch": 6.955034213098729, + "grad_norm": 0.7697531302927278, + "learning_rate": 5.250073653702004e-09, + "loss": 0.1465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14434030652046204, + "step": 7115, + "valid_targets_mean": 2634.2, + "valid_targets_min": 882 + }, + { + "epoch": 6.959921798631476, + "grad_norm": 0.909840589624624, + "learning_rate": 4.192490517285208e-09, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1433998942375183, + "step": 7120, + "valid_targets_mean": 1897.3, + "valid_targets_min": 721 + }, + { + "epoch": 6.964809384164223, + "grad_norm": 0.880665464777133, + "learning_rate": 3.2537213277228364e-09, + "loss": 0.1438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15554025769233704, + "step": 7125, + "valid_targets_mean": 2096.7, + "valid_targets_min": 754 + }, + { + "epoch": 6.96969696969697, + "grad_norm": 0.7411591407393181, + "learning_rate": 2.4337716631328555e-09, + "loss": 0.1394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13086152076721191, + "step": 7130, + "valid_targets_mean": 2635.5, + "valid_targets_min": 826 + }, + { + "epoch": 6.9745845552297165, + "grad_norm": 0.8057506293460369, + "learning_rate": 1.732646395606885e-09, + "loss": 0.1429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14949795603752136, + "step": 7135, + "valid_targets_mean": 2405.2, + "valid_targets_min": 748 + }, + { + "epoch": 6.979472140762463, + "grad_norm": 0.8097660910794996, + "learning_rate": 1.1503496911924316e-09, + "loss": 0.1471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.134006530046463, + "step": 7140, + "valid_targets_mean": 2312.4, + "valid_targets_min": 997 + }, + { + "epoch": 6.98435972629521, + "grad_norm": 0.7666411016932615, + "learning_rate": 6.868850098618041e-10, + "loss": 0.1485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1432676464319229, + "step": 7145, + "valid_targets_mean": 2855.9, + "valid_targets_min": 763 + }, + { + "epoch": 6.989247311827957, + "grad_norm": 0.8085737681168628, + "learning_rate": 3.4225510549656947e-10, + "loss": 0.1442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13285693526268005, + "step": 7150, + "valid_targets_mean": 2373.7, + "valid_targets_min": 684 + }, + { + "epoch": 6.994134897360704, + "grad_norm": 0.6798769668315741, + "learning_rate": 1.1646202586756973e-10, + "loss": 0.1453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15072640776634216, + "step": 7155, + "valid_targets_mean": 3628.1, + "valid_targets_min": 1272 + }, + { + "epoch": 6.999022482893451, + "grad_norm": 0.8430786453318077, + "learning_rate": 9.507112626039316e-12, + "loss": 0.1443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11695080995559692, + "step": 7160, + "valid_targets_mean": 2002.8, + "valid_targets_min": 715 + }, + { + "epoch": 7.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1567835956811905, + "step": 7161, + "total_flos": 1242507420499968.0, + "train_loss": 0.2217263549632857, + "train_runtime": 30173.9773, + "train_samples_per_second": 3.796, + "train_steps_per_second": 0.237, + "valid_targets_mean": 1853.1, + "valid_targets_min": 622 + } + ], + "logging_steps": 5, + "max_steps": 7161, + "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": true + }, + "attributes": {} + } + }, + "total_flos": 1242507420499968.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}