diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,14699 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 7.0, + "eval_steps": 500, + "global_step": 6664, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.005252100840336135, + "grad_norm": 14.420694283658325, + "learning_rate": 2.39880059970015e-07, + "loss": 0.8126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.790475606918335, + "step": 5, + "valid_targets_mean": 1423.4, + "valid_targets_min": 715 + }, + { + "epoch": 0.01050420168067227, + "grad_norm": 15.311496330131531, + "learning_rate": 5.397301349325338e-07, + "loss": 0.8157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.8297374248504639, + "step": 10, + "valid_targets_mean": 1407.8, + "valid_targets_min": 612 + }, + { + "epoch": 0.015756302521008403, + "grad_norm": 13.45958528833613, + "learning_rate": 8.395802098950526e-07, + "loss": 0.7879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7610906362533569, + "step": 15, + "valid_targets_mean": 1413.5, + "valid_targets_min": 714 + }, + { + "epoch": 0.02100840336134454, + "grad_norm": 11.905230272753508, + "learning_rate": 1.1394302848575713e-06, + "loss": 0.7735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.753049373626709, + "step": 20, + "valid_targets_mean": 1450.1, + "valid_targets_min": 754 + }, + { + "epoch": 0.026260504201680673, + "grad_norm": 9.744011308309286, + "learning_rate": 1.43928035982009e-06, + "loss": 0.744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.7382731437683105, + "step": 25, + "valid_targets_mean": 1481.1, + "valid_targets_min": 718 + }, + { + "epoch": 0.031512605042016806, + "grad_norm": 7.094821641510122, + "learning_rate": 1.7391304347826088e-06, + "loss": 0.7128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6914137601852417, + "step": 30, + "valid_targets_mean": 1530.8, + "valid_targets_min": 731 + }, + { + "epoch": 0.03676470588235294, + "grad_norm": 6.368906220458843, + "learning_rate": 2.0389805097451275e-06, + "loss": 0.6643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6466853022575378, + "step": 35, + "valid_targets_mean": 1416.6, + "valid_targets_min": 725 + }, + { + "epoch": 0.04201680672268908, + "grad_norm": 5.399102696758742, + "learning_rate": 2.3388305847076464e-06, + "loss": 0.6382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.6274276971817017, + "step": 40, + "valid_targets_mean": 1371.5, + "valid_targets_min": 610 + }, + { + "epoch": 0.04726890756302521, + "grad_norm": 4.306253658085107, + "learning_rate": 2.6386806596701653e-06, + "loss": 0.5772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5616586208343506, + "step": 45, + "valid_targets_mean": 1484.7, + "valid_targets_min": 720 + }, + { + "epoch": 0.052521008403361345, + "grad_norm": 3.7077951595748515, + "learning_rate": 2.9385307346326843e-06, + "loss": 0.527, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4847095012664795, + "step": 50, + "valid_targets_mean": 1365.5, + "valid_targets_min": 765 + }, + { + "epoch": 0.05777310924369748, + "grad_norm": 2.363263225041768, + "learning_rate": 3.2383808095952024e-06, + "loss": 0.4935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.5059503316879272, + "step": 55, + "valid_targets_mean": 1234.7, + "valid_targets_min": 593 + }, + { + "epoch": 0.06302521008403361, + "grad_norm": 1.6467637243933408, + "learning_rate": 3.5382308845577213e-06, + "loss": 0.4679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.44630974531173706, + "step": 60, + "valid_targets_mean": 1540.4, + "valid_targets_min": 704 + }, + { + "epoch": 0.06827731092436974, + "grad_norm": 1.5411024922038776, + "learning_rate": 3.83808095952024e-06, + "loss": 0.4446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.41912999749183655, + "step": 65, + "valid_targets_mean": 1329.9, + "valid_targets_min": 722 + }, + { + "epoch": 0.07352941176470588, + "grad_norm": 1.244534525186305, + "learning_rate": 4.137931034482759e-06, + "loss": 0.4205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39288026094436646, + "step": 70, + "valid_targets_mean": 1549.6, + "valid_targets_min": 865 + }, + { + "epoch": 0.07878151260504201, + "grad_norm": 1.1996703760120404, + "learning_rate": 4.437781109445278e-06, + "loss": 0.4055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.40949755907058716, + "step": 75, + "valid_targets_mean": 1507.4, + "valid_targets_min": 797 + }, + { + "epoch": 0.08403361344537816, + "grad_norm": 1.2896717545023444, + "learning_rate": 4.737631184407796e-06, + "loss": 0.4011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.43845802545547485, + "step": 80, + "valid_targets_mean": 1542.4, + "valid_targets_min": 728 + }, + { + "epoch": 0.08928571428571429, + "grad_norm": 1.0177436395323942, + "learning_rate": 5.037481259370315e-06, + "loss": 0.3951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.38916561007499695, + "step": 85, + "valid_targets_mean": 1595.2, + "valid_targets_min": 683 + }, + { + "epoch": 0.09453781512605042, + "grad_norm": 1.9249372968791647, + "learning_rate": 5.337331334332834e-06, + "loss": 0.3869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39647072553634644, + "step": 90, + "valid_targets_mean": 1401.6, + "valid_targets_min": 706 + }, + { + "epoch": 0.09978991596638656, + "grad_norm": 1.284016705259006, + "learning_rate": 5.6371814092953526e-06, + "loss": 0.3607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3547477126121521, + "step": 95, + "valid_targets_mean": 1298.4, + "valid_targets_min": 625 + }, + { + "epoch": 0.10504201680672269, + "grad_norm": 1.0443779966069502, + "learning_rate": 5.937031484257871e-06, + "loss": 0.3572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35236796736717224, + "step": 100, + "valid_targets_mean": 1659.7, + "valid_targets_min": 790 + }, + { + "epoch": 0.11029411764705882, + "grad_norm": 1.0855951515574154, + "learning_rate": 6.2368815592203904e-06, + "loss": 0.3587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3708180785179138, + "step": 105, + "valid_targets_mean": 1600.4, + "valid_targets_min": 684 + }, + { + "epoch": 0.11554621848739496, + "grad_norm": 1.1032891548957378, + "learning_rate": 6.536731634182909e-06, + "loss": 0.3475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.322293221950531, + "step": 110, + "valid_targets_mean": 1338.3, + "valid_targets_min": 739 + }, + { + "epoch": 0.1207983193277311, + "grad_norm": 0.9768878378529005, + "learning_rate": 6.8365817091454274e-06, + "loss": 0.3369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3234562277793884, + "step": 115, + "valid_targets_mean": 1456.2, + "valid_targets_min": 657 + }, + { + "epoch": 0.12605042016806722, + "grad_norm": 1.0487060724275885, + "learning_rate": 7.136431784107947e-06, + "loss": 0.329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32765859365463257, + "step": 120, + "valid_targets_mean": 1340.2, + "valid_targets_min": 783 + }, + { + "epoch": 0.13130252100840337, + "grad_norm": 1.1049013021741143, + "learning_rate": 7.436281859070465e-06, + "loss": 0.3235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31769877672195435, + "step": 125, + "valid_targets_mean": 1394.4, + "valid_targets_min": 734 + }, + { + "epoch": 0.13655462184873948, + "grad_norm": 1.0426803049134692, + "learning_rate": 7.736131934032984e-06, + "loss": 0.319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3211321234703064, + "step": 130, + "valid_targets_mean": 1501.8, + "valid_targets_min": 678 + }, + { + "epoch": 0.14180672268907563, + "grad_norm": 1.0513598327816873, + "learning_rate": 8.035982008995503e-06, + "loss": 0.3162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31629741191864014, + "step": 135, + "valid_targets_mean": 1651.7, + "valid_targets_min": 843 + }, + { + "epoch": 0.14705882352941177, + "grad_norm": 1.0551559578309457, + "learning_rate": 8.335832083958023e-06, + "loss": 0.3204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3266269564628601, + "step": 140, + "valid_targets_mean": 1632.1, + "valid_targets_min": 782 + }, + { + "epoch": 0.15231092436974789, + "grad_norm": 0.9612333107113046, + "learning_rate": 8.63568215892054e-06, + "loss": 0.306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33223849534988403, + "step": 145, + "valid_targets_mean": 1615.3, + "valid_targets_min": 957 + }, + { + "epoch": 0.15756302521008403, + "grad_norm": 1.0081404887340868, + "learning_rate": 8.93553223388306e-06, + "loss": 0.315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32516705989837646, + "step": 150, + "valid_targets_mean": 1482.9, + "valid_targets_min": 713 + }, + { + "epoch": 0.16281512605042017, + "grad_norm": 1.0468696423891182, + "learning_rate": 9.235382308845579e-06, + "loss": 0.3071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3112391531467438, + "step": 155, + "valid_targets_mean": 1552.8, + "valid_targets_min": 726 + }, + { + "epoch": 0.16806722689075632, + "grad_norm": 0.9973569578712165, + "learning_rate": 9.535232383808097e-06, + "loss": 0.2979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2998279333114624, + "step": 160, + "valid_targets_mean": 1590.4, + "valid_targets_min": 1147 + }, + { + "epoch": 0.17331932773109243, + "grad_norm": 1.2443677749688786, + "learning_rate": 9.835082458770614e-06, + "loss": 0.2954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30380943417549133, + "step": 165, + "valid_targets_mean": 1417.6, + "valid_targets_min": 700 + }, + { + "epoch": 0.17857142857142858, + "grad_norm": 1.1254552821617871, + "learning_rate": 1.0134932533733135e-05, + "loss": 0.284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2681633234024048, + "step": 170, + "valid_targets_mean": 1255.4, + "valid_targets_min": 640 + }, + { + "epoch": 0.18382352941176472, + "grad_norm": 0.9851494239413889, + "learning_rate": 1.0434782608695653e-05, + "loss": 0.2881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2868928611278534, + "step": 175, + "valid_targets_mean": 1566.6, + "valid_targets_min": 817 + }, + { + "epoch": 0.18907563025210083, + "grad_norm": 1.098473835442716, + "learning_rate": 1.0734632683658172e-05, + "loss": 0.286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27357298135757446, + "step": 180, + "valid_targets_mean": 1440.9, + "valid_targets_min": 724 + }, + { + "epoch": 0.19432773109243698, + "grad_norm": 1.1557321572729486, + "learning_rate": 1.103448275862069e-05, + "loss": 0.2938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29238957166671753, + "step": 185, + "valid_targets_mean": 1399.1, + "valid_targets_min": 657 + }, + { + "epoch": 0.19957983193277312, + "grad_norm": 1.1396871077620272, + "learning_rate": 1.1334332833583211e-05, + "loss": 0.2932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2987218499183655, + "step": 190, + "valid_targets_mean": 1334.3, + "valid_targets_min": 746 + }, + { + "epoch": 0.20483193277310924, + "grad_norm": 1.1178689873643397, + "learning_rate": 1.1634182908545729e-05, + "loss": 0.2941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33798643946647644, + "step": 195, + "valid_targets_mean": 1452.8, + "valid_targets_min": 692 + }, + { + "epoch": 0.21008403361344538, + "grad_norm": 1.0854720395667985, + "learning_rate": 1.1934032983508246e-05, + "loss": 0.2816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25798511505126953, + "step": 200, + "valid_targets_mean": 1338.4, + "valid_targets_min": 682 + }, + { + "epoch": 0.21533613445378152, + "grad_norm": 1.1508679180169188, + "learning_rate": 1.2233883058470766e-05, + "loss": 0.2723, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.276319682598114, + "step": 205, + "valid_targets_mean": 1360.2, + "valid_targets_min": 703 + }, + { + "epoch": 0.22058823529411764, + "grad_norm": 0.990266456581399, + "learning_rate": 1.2533733133433283e-05, + "loss": 0.2973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2741560935974121, + "step": 210, + "valid_targets_mean": 1520.9, + "valid_targets_min": 864 + }, + { + "epoch": 0.22584033613445378, + "grad_norm": 0.9411807898115928, + "learning_rate": 1.2833583208395803e-05, + "loss": 0.2983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3527371883392334, + "step": 215, + "valid_targets_mean": 1814.6, + "valid_targets_min": 734 + }, + { + "epoch": 0.23109243697478993, + "grad_norm": 0.9509244068358998, + "learning_rate": 1.313343328335832e-05, + "loss": 0.2758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27980518341064453, + "step": 220, + "valid_targets_mean": 1478.3, + "valid_targets_min": 607 + }, + { + "epoch": 0.23634453781512604, + "grad_norm": 0.9802646448806872, + "learning_rate": 1.3433283358320841e-05, + "loss": 0.2797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2757881283760071, + "step": 225, + "valid_targets_mean": 1497.9, + "valid_targets_min": 551 + }, + { + "epoch": 0.2415966386554622, + "grad_norm": 1.1416789358107773, + "learning_rate": 1.3733133433283359e-05, + "loss": 0.2842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2863280773162842, + "step": 230, + "valid_targets_mean": 1550.5, + "valid_targets_min": 676 + }, + { + "epoch": 0.24684873949579833, + "grad_norm": 1.2243896073517977, + "learning_rate": 1.4032983508245878e-05, + "loss": 0.28, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2647148370742798, + "step": 235, + "valid_targets_mean": 1372.5, + "valid_targets_min": 904 + }, + { + "epoch": 0.25210084033613445, + "grad_norm": 1.0671177129673626, + "learning_rate": 1.4332833583208396e-05, + "loss": 0.2667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2796405553817749, + "step": 240, + "valid_targets_mean": 1479.0, + "valid_targets_min": 859 + }, + { + "epoch": 0.25735294117647056, + "grad_norm": 1.1155587555825344, + "learning_rate": 1.4632683658170917e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2673705816268921, + "step": 245, + "valid_targets_mean": 1234.8, + "valid_targets_min": 737 + }, + { + "epoch": 0.26260504201680673, + "grad_norm": 1.0111523665697042, + "learning_rate": 1.4932533733133435e-05, + "loss": 0.2816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3024107813835144, + "step": 250, + "valid_targets_mean": 1363.2, + "valid_targets_min": 791 + }, + { + "epoch": 0.26785714285714285, + "grad_norm": 0.9398095246969382, + "learning_rate": 1.5232383808095954e-05, + "loss": 0.2773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28389376401901245, + "step": 255, + "valid_targets_mean": 1532.1, + "valid_targets_min": 786 + }, + { + "epoch": 0.27310924369747897, + "grad_norm": 0.9520954006331536, + "learning_rate": 1.5532233883058472e-05, + "loss": 0.2619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.262344628572464, + "step": 260, + "valid_targets_mean": 1471.0, + "valid_targets_min": 944 + }, + { + "epoch": 0.27836134453781514, + "grad_norm": 0.9231810982153198, + "learning_rate": 1.5832083958020993e-05, + "loss": 0.268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25394904613494873, + "step": 265, + "valid_targets_mean": 1494.1, + "valid_targets_min": 1075 + }, + { + "epoch": 0.28361344537815125, + "grad_norm": 1.036465235131244, + "learning_rate": 1.613193403298351e-05, + "loss": 0.2693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2542717456817627, + "step": 270, + "valid_targets_mean": 1340.9, + "valid_targets_min": 740 + }, + { + "epoch": 0.28886554621848737, + "grad_norm": 0.969535346268365, + "learning_rate": 1.6431784107946028e-05, + "loss": 0.2674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2644956707954407, + "step": 275, + "valid_targets_mean": 1530.9, + "valid_targets_min": 727 + }, + { + "epoch": 0.29411764705882354, + "grad_norm": 1.095458178867704, + "learning_rate": 1.6731634182908546e-05, + "loss": 0.257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2714339792728424, + "step": 280, + "valid_targets_mean": 1233.0, + "valid_targets_min": 636 + }, + { + "epoch": 0.29936974789915966, + "grad_norm": 0.989855094627666, + "learning_rate": 1.7031484257871064e-05, + "loss": 0.2571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2507407069206238, + "step": 285, + "valid_targets_mean": 1246.8, + "valid_targets_min": 684 + }, + { + "epoch": 0.30462184873949577, + "grad_norm": 1.0834380820761222, + "learning_rate": 1.7331334332833585e-05, + "loss": 0.2479, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26474529504776, + "step": 290, + "valid_targets_mean": 1392.5, + "valid_targets_min": 798 + }, + { + "epoch": 0.30987394957983194, + "grad_norm": 1.035539563074315, + "learning_rate": 1.7631184407796102e-05, + "loss": 0.2635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24842077493667603, + "step": 295, + "valid_targets_mean": 1319.9, + "valid_targets_min": 661 + }, + { + "epoch": 0.31512605042016806, + "grad_norm": 1.0392297368060632, + "learning_rate": 1.7931034482758623e-05, + "loss": 0.2567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2532370090484619, + "step": 300, + "valid_targets_mean": 1554.4, + "valid_targets_min": 665 + }, + { + "epoch": 0.32037815126050423, + "grad_norm": 1.1437803903133303, + "learning_rate": 1.823088455772114e-05, + "loss": 0.2547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26745492219924927, + "step": 305, + "valid_targets_mean": 1417.8, + "valid_targets_min": 692 + }, + { + "epoch": 0.32563025210084034, + "grad_norm": 1.0470695391841682, + "learning_rate": 1.853073463268366e-05, + "loss": 0.2622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27369868755340576, + "step": 310, + "valid_targets_mean": 1223.8, + "valid_targets_min": 632 + }, + { + "epoch": 0.33088235294117646, + "grad_norm": 0.99682750228871, + "learning_rate": 1.8830584707646176e-05, + "loss": 0.2577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2700139284133911, + "step": 315, + "valid_targets_mean": 1550.6, + "valid_targets_min": 922 + }, + { + "epoch": 0.33613445378151263, + "grad_norm": 0.9922010090048154, + "learning_rate": 1.9130434782608697e-05, + "loss": 0.2509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23878350853919983, + "step": 320, + "valid_targets_mean": 1542.2, + "valid_targets_min": 710 + }, + { + "epoch": 0.34138655462184875, + "grad_norm": 0.9969922099846812, + "learning_rate": 1.9430284857571215e-05, + "loss": 0.2564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24636775255203247, + "step": 325, + "valid_targets_mean": 1390.6, + "valid_targets_min": 750 + }, + { + "epoch": 0.34663865546218486, + "grad_norm": 0.9249347847273258, + "learning_rate": 1.9730134932533736e-05, + "loss": 0.2576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25145918130874634, + "step": 330, + "valid_targets_mean": 1515.5, + "valid_targets_min": 766 + }, + { + "epoch": 0.35189075630252103, + "grad_norm": 0.9954869233754978, + "learning_rate": 2.0029985007496254e-05, + "loss": 0.2675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23031079769134521, + "step": 335, + "valid_targets_mean": 1333.3, + "valid_targets_min": 748 + }, + { + "epoch": 0.35714285714285715, + "grad_norm": 1.0229626657604847, + "learning_rate": 2.0329835082458775e-05, + "loss": 0.2635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2601286470890045, + "step": 340, + "valid_targets_mean": 1395.2, + "valid_targets_min": 682 + }, + { + "epoch": 0.36239495798319327, + "grad_norm": 0.9742753314709449, + "learning_rate": 2.0629685157421292e-05, + "loss": 0.2644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2516604959964752, + "step": 345, + "valid_targets_mean": 1515.6, + "valid_targets_min": 738 + }, + { + "epoch": 0.36764705882352944, + "grad_norm": 1.0984493157523587, + "learning_rate": 2.092953523238381e-05, + "loss": 0.263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26009607315063477, + "step": 350, + "valid_targets_mean": 1312.8, + "valid_targets_min": 731 + }, + { + "epoch": 0.37289915966386555, + "grad_norm": 1.0124670927846027, + "learning_rate": 2.1229385307346328e-05, + "loss": 0.2634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2756904661655426, + "step": 355, + "valid_targets_mean": 1492.9, + "valid_targets_min": 646 + }, + { + "epoch": 0.37815126050420167, + "grad_norm": 0.9303311281634132, + "learning_rate": 2.152923538230885e-05, + "loss": 0.2475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23849308490753174, + "step": 360, + "valid_targets_mean": 1369.1, + "valid_targets_min": 780 + }, + { + "epoch": 0.38340336134453784, + "grad_norm": 0.938627487391855, + "learning_rate": 2.1829085457271363e-05, + "loss": 0.2506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24323752522468567, + "step": 365, + "valid_targets_mean": 1521.9, + "valid_targets_min": 676 + }, + { + "epoch": 0.38865546218487396, + "grad_norm": 1.0170934463633303, + "learning_rate": 2.2128935532233884e-05, + "loss": 0.2615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23910865187644958, + "step": 370, + "valid_targets_mean": 1277.9, + "valid_targets_min": 687 + }, + { + "epoch": 0.3939075630252101, + "grad_norm": 0.9747309964507797, + "learning_rate": 2.2428785607196405e-05, + "loss": 0.2519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2347027212381363, + "step": 375, + "valid_targets_mean": 1336.5, + "valid_targets_min": 703 + }, + { + "epoch": 0.39915966386554624, + "grad_norm": 1.0544024939800751, + "learning_rate": 2.2728635682158923e-05, + "loss": 0.2461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22451592981815338, + "step": 380, + "valid_targets_mean": 1288.1, + "valid_targets_min": 522 + }, + { + "epoch": 0.40441176470588236, + "grad_norm": 1.0293634003316077, + "learning_rate": 2.302848575712144e-05, + "loss": 0.2399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2373955398797989, + "step": 385, + "valid_targets_mean": 1526.9, + "valid_targets_min": 764 + }, + { + "epoch": 0.4096638655462185, + "grad_norm": 1.0510698011217983, + "learning_rate": 2.3328335832083958e-05, + "loss": 0.2701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.272568941116333, + "step": 390, + "valid_targets_mean": 1372.4, + "valid_targets_min": 548 + }, + { + "epoch": 0.41491596638655465, + "grad_norm": 1.0256066743434804, + "learning_rate": 2.362818590704648e-05, + "loss": 0.237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2508437931537628, + "step": 395, + "valid_targets_mean": 1356.8, + "valid_targets_min": 694 + }, + { + "epoch": 0.42016806722689076, + "grad_norm": 0.918276243513663, + "learning_rate": 2.3928035982009e-05, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26278817653656006, + "step": 400, + "valid_targets_mean": 1540.9, + "valid_targets_min": 823 + }, + { + "epoch": 0.4254201680672269, + "grad_norm": 0.829017462865176, + "learning_rate": 2.4227886056971515e-05, + "loss": 0.2182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1897987425327301, + "step": 405, + "valid_targets_mean": 1442.4, + "valid_targets_min": 682 + }, + { + "epoch": 0.43067226890756305, + "grad_norm": 1.009337402964787, + "learning_rate": 2.4527736131934036e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25567498803138733, + "step": 410, + "valid_targets_mean": 1407.6, + "valid_targets_min": 693 + }, + { + "epoch": 0.43592436974789917, + "grad_norm": 1.102110069443316, + "learning_rate": 2.4827586206896553e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22340771555900574, + "step": 415, + "valid_targets_mean": 1197.8, + "valid_targets_min": 639 + }, + { + "epoch": 0.4411764705882353, + "grad_norm": 0.9439286600674722, + "learning_rate": 2.5127436281859074e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22316749393939972, + "step": 420, + "valid_targets_mean": 1423.3, + "valid_targets_min": 685 + }, + { + "epoch": 0.44642857142857145, + "grad_norm": 0.9288887022810997, + "learning_rate": 2.542728635682159e-05, + "loss": 0.2356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22205528616905212, + "step": 425, + "valid_targets_mean": 1331.5, + "valid_targets_min": 827 + }, + { + "epoch": 0.45168067226890757, + "grad_norm": 0.918938060661647, + "learning_rate": 2.572713643178411e-05, + "loss": 0.2436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24578994512557983, + "step": 430, + "valid_targets_mean": 1381.7, + "valid_targets_min": 686 + }, + { + "epoch": 0.4569327731092437, + "grad_norm": 0.8450512785959888, + "learning_rate": 2.602698650674663e-05, + "loss": 0.2512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25319570302963257, + "step": 435, + "valid_targets_mean": 1472.2, + "valid_targets_min": 713 + }, + { + "epoch": 0.46218487394957986, + "grad_norm": 1.3247180796200522, + "learning_rate": 2.632683658170915e-05, + "loss": 0.2979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4720996022224426, + "step": 440, + "valid_targets_mean": 2248.9, + "valid_targets_min": 920 + }, + { + "epoch": 0.46743697478991597, + "grad_norm": 0.7756817606153216, + "learning_rate": 2.6626686656671666e-05, + "loss": 0.4381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4191029667854309, + "step": 445, + "valid_targets_mean": 3664.4, + "valid_targets_min": 746 + }, + { + "epoch": 0.4726890756302521, + "grad_norm": 0.8141822973818857, + "learning_rate": 2.6926536731634184e-05, + "loss": 0.4066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.42180395126342773, + "step": 450, + "valid_targets_mean": 3591.1, + "valid_targets_min": 1635 + }, + { + "epoch": 0.47794117647058826, + "grad_norm": 0.6183205557943057, + "learning_rate": 2.7226386806596705e-05, + "loss": 0.399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37809157371520996, + "step": 455, + "valid_targets_mean": 4172.7, + "valid_targets_min": 777 + }, + { + "epoch": 0.4831932773109244, + "grad_norm": 0.6341645150900163, + "learning_rate": 2.752623688155922e-05, + "loss": 0.3807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3980065882205963, + "step": 460, + "valid_targets_mean": 4040.2, + "valid_targets_min": 2187 + }, + { + "epoch": 0.4884453781512605, + "grad_norm": 0.7232188026091549, + "learning_rate": 2.782608695652174e-05, + "loss": 0.3993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39658811688423157, + "step": 465, + "valid_targets_mean": 3156.3, + "valid_targets_min": 1432 + }, + { + "epoch": 0.49369747899159666, + "grad_norm": 0.5975850484936072, + "learning_rate": 2.812593703148426e-05, + "loss": 0.377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35548049211502075, + "step": 470, + "valid_targets_mean": 3529.7, + "valid_targets_min": 767 + }, + { + "epoch": 0.4989495798319328, + "grad_norm": 0.6544292649550012, + "learning_rate": 2.8425787106446782e-05, + "loss": 0.3589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36396756768226624, + "step": 475, + "valid_targets_mean": 3485.6, + "valid_targets_min": 1488 + }, + { + "epoch": 0.5042016806722689, + "grad_norm": 0.5784234889060588, + "learning_rate": 2.8725637181409296e-05, + "loss": 0.3769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3580145239830017, + "step": 480, + "valid_targets_mean": 3707.0, + "valid_targets_min": 578 + }, + { + "epoch": 0.509453781512605, + "grad_norm": 0.5809204879202081, + "learning_rate": 2.9025487256371818e-05, + "loss": 0.3991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37097570300102234, + "step": 485, + "valid_targets_mean": 4252.4, + "valid_targets_min": 1241 + }, + { + "epoch": 0.5147058823529411, + "grad_norm": 0.6518660166623274, + "learning_rate": 2.9325337331334335e-05, + "loss": 0.3791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39846816658973694, + "step": 490, + "valid_targets_mean": 4029.2, + "valid_targets_min": 807 + }, + { + "epoch": 0.5199579831932774, + "grad_norm": 0.5952710397208067, + "learning_rate": 2.9625187406296856e-05, + "loss": 0.3793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34132808446884155, + "step": 495, + "valid_targets_mean": 4507.0, + "valid_targets_min": 2011 + }, + { + "epoch": 0.5252100840336135, + "grad_norm": 0.5921750584078646, + "learning_rate": 2.992503748125937e-05, + "loss": 0.3773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3380280137062073, + "step": 500, + "valid_targets_mean": 4268.2, + "valid_targets_min": 1329 + }, + { + "epoch": 0.5304621848739496, + "grad_norm": 0.7847684848643338, + "learning_rate": 3.022488755622189e-05, + "loss": 0.3692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3614731431007385, + "step": 505, + "valid_targets_mean": 3858.9, + "valid_targets_min": 1366 + }, + { + "epoch": 0.5357142857142857, + "grad_norm": 0.6674187275757447, + "learning_rate": 3.052473763118441e-05, + "loss": 0.3796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36018460988998413, + "step": 510, + "valid_targets_mean": 3169.1, + "valid_targets_min": 1334 + }, + { + "epoch": 0.5409663865546218, + "grad_norm": 0.5809835935023818, + "learning_rate": 3.0824587706146934e-05, + "loss": 0.3895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34386080503463745, + "step": 515, + "valid_targets_mean": 4130.4, + "valid_targets_min": 1762 + }, + { + "epoch": 0.5462184873949579, + "grad_norm": 0.6856115085091843, + "learning_rate": 3.112443778110945e-05, + "loss": 0.3807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3728490471839905, + "step": 520, + "valid_targets_mean": 3285.0, + "valid_targets_min": 855 + }, + { + "epoch": 0.5514705882352942, + "grad_norm": 0.7152310015109019, + "learning_rate": 3.142428785607197e-05, + "loss": 0.3934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3993592858314514, + "step": 525, + "valid_targets_mean": 3298.2, + "valid_targets_min": 1269 + }, + { + "epoch": 0.5567226890756303, + "grad_norm": 0.7853116245952066, + "learning_rate": 3.172413793103448e-05, + "loss": 0.3929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3852170705795288, + "step": 530, + "valid_targets_mean": 2740.4, + "valid_targets_min": 1041 + }, + { + "epoch": 0.5619747899159664, + "grad_norm": 0.7253886613797418, + "learning_rate": 3.2023988005997004e-05, + "loss": 0.3777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3952094614505768, + "step": 535, + "valid_targets_mean": 2995.1, + "valid_targets_min": 1291 + }, + { + "epoch": 0.5672268907563025, + "grad_norm": 0.7947910205242297, + "learning_rate": 3.2323838080959525e-05, + "loss": 0.3671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3840707540512085, + "step": 540, + "valid_targets_mean": 2811.6, + "valid_targets_min": 1133 + }, + { + "epoch": 0.5724789915966386, + "grad_norm": 0.7101634109805368, + "learning_rate": 3.262368815592204e-05, + "loss": 0.3656, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3580155372619629, + "step": 545, + "valid_targets_mean": 3709.5, + "valid_targets_min": 865 + }, + { + "epoch": 0.5777310924369747, + "grad_norm": 0.7401801947692503, + "learning_rate": 3.292353823088456e-05, + "loss": 0.3732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37310606241226196, + "step": 550, + "valid_targets_mean": 2960.6, + "valid_targets_min": 958 + }, + { + "epoch": 0.582983193277311, + "grad_norm": 0.6819266322867479, + "learning_rate": 3.3223388305847075e-05, + "loss": 0.3727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.376257061958313, + "step": 555, + "valid_targets_mean": 3189.2, + "valid_targets_min": 771 + }, + { + "epoch": 0.5882352941176471, + "grad_norm": 0.5792785228231021, + "learning_rate": 3.3523238380809596e-05, + "loss": 0.3706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3450887203216553, + "step": 560, + "valid_targets_mean": 3823.1, + "valid_targets_min": 1631 + }, + { + "epoch": 0.5934873949579832, + "grad_norm": 0.706870647311464, + "learning_rate": 3.382308845577212e-05, + "loss": 0.3699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33471646904945374, + "step": 565, + "valid_targets_mean": 2639.4, + "valid_targets_min": 908 + }, + { + "epoch": 0.5987394957983193, + "grad_norm": 0.8079454260503596, + "learning_rate": 3.412293853073464e-05, + "loss": 0.3653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.39000123739242554, + "step": 570, + "valid_targets_mean": 2988.3, + "valid_targets_min": 1264 + }, + { + "epoch": 0.6039915966386554, + "grad_norm": 0.6685925501276144, + "learning_rate": 3.442278860569715e-05, + "loss": 0.3659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3652319312095642, + "step": 575, + "valid_targets_mean": 3563.1, + "valid_targets_min": 1889 + }, + { + "epoch": 0.6092436974789915, + "grad_norm": 0.6271409370176393, + "learning_rate": 3.4722638680659673e-05, + "loss": 0.3594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3371011018753052, + "step": 580, + "valid_targets_mean": 3737.9, + "valid_targets_min": 1075 + }, + { + "epoch": 0.6144957983193278, + "grad_norm": 0.7635609780961795, + "learning_rate": 3.5022488755622194e-05, + "loss": 0.3644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3491378426551819, + "step": 585, + "valid_targets_mean": 2745.8, + "valid_targets_min": 1036 + }, + { + "epoch": 0.6197478991596639, + "grad_norm": 0.9486847263065615, + "learning_rate": 3.5322338830584716e-05, + "loss": 0.3682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3554984927177429, + "step": 590, + "valid_targets_mean": 3516.4, + "valid_targets_min": 1162 + }, + { + "epoch": 0.625, + "grad_norm": 0.6979971886991938, + "learning_rate": 3.562218890554723e-05, + "loss": 0.3604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3572150468826294, + "step": 595, + "valid_targets_mean": 3279.4, + "valid_targets_min": 1281 + }, + { + "epoch": 0.6302521008403361, + "grad_norm": 0.628571339160864, + "learning_rate": 3.592203898050975e-05, + "loss": 0.3601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32577332854270935, + "step": 600, + "valid_targets_mean": 3180.4, + "valid_targets_min": 1816 + }, + { + "epoch": 0.6355042016806722, + "grad_norm": 0.7162106687544288, + "learning_rate": 3.6221889055472265e-05, + "loss": 0.3733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36203664541244507, + "step": 605, + "valid_targets_mean": 2939.2, + "valid_targets_min": 1629 + }, + { + "epoch": 0.6407563025210085, + "grad_norm": 0.7531458122055458, + "learning_rate": 3.6521739130434786e-05, + "loss": 0.3618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3718830943107605, + "step": 610, + "valid_targets_mean": 2755.7, + "valid_targets_min": 1254 + }, + { + "epoch": 0.6460084033613446, + "grad_norm": 0.7043721880651528, + "learning_rate": 3.68215892053973e-05, + "loss": 0.3629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3637351989746094, + "step": 615, + "valid_targets_mean": 3390.2, + "valid_targets_min": 1973 + }, + { + "epoch": 0.6512605042016807, + "grad_norm": 0.7411003182380056, + "learning_rate": 3.712143928035982e-05, + "loss": 0.3638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35509952902793884, + "step": 620, + "valid_targets_mean": 2960.4, + "valid_targets_min": 1195 + }, + { + "epoch": 0.6565126050420168, + "grad_norm": 0.6916717573319705, + "learning_rate": 3.742128935532234e-05, + "loss": 0.3592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37266242504119873, + "step": 625, + "valid_targets_mean": 3181.4, + "valid_targets_min": 1244 + }, + { + "epoch": 0.6617647058823529, + "grad_norm": 0.6405799101681395, + "learning_rate": 3.772113943028486e-05, + "loss": 0.3642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.351756751537323, + "step": 630, + "valid_targets_mean": 3464.2, + "valid_targets_min": 1421 + }, + { + "epoch": 0.667016806722689, + "grad_norm": 0.6665303943258725, + "learning_rate": 3.802098950524738e-05, + "loss": 0.3496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34305959939956665, + "step": 635, + "valid_targets_mean": 3443.1, + "valid_targets_min": 1097 + }, + { + "epoch": 0.6722689075630253, + "grad_norm": 0.6963251575164391, + "learning_rate": 3.83208395802099e-05, + "loss": 0.349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36686572432518005, + "step": 640, + "valid_targets_mean": 3891.4, + "valid_targets_min": 1168 + }, + { + "epoch": 0.6775210084033614, + "grad_norm": 0.6447119132301098, + "learning_rate": 3.862068965517242e-05, + "loss": 0.3547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35105079412460327, + "step": 645, + "valid_targets_mean": 3368.9, + "valid_targets_min": 1700 + }, + { + "epoch": 0.6827731092436975, + "grad_norm": 0.775468044490351, + "learning_rate": 3.8920539730134934e-05, + "loss": 0.3561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35495996475219727, + "step": 650, + "valid_targets_mean": 2497.1, + "valid_targets_min": 1261 + }, + { + "epoch": 0.6880252100840336, + "grad_norm": 0.6998673088388078, + "learning_rate": 3.9220389805097455e-05, + "loss": 0.37, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.373879611492157, + "step": 655, + "valid_targets_mean": 3391.2, + "valid_targets_min": 1124 + }, + { + "epoch": 0.6932773109243697, + "grad_norm": 0.6911940333174383, + "learning_rate": 3.9520239880059976e-05, + "loss": 0.3604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3743062913417816, + "step": 660, + "valid_targets_mean": 3003.8, + "valid_targets_min": 1361 + }, + { + "epoch": 0.6985294117647058, + "grad_norm": 0.6312467853415549, + "learning_rate": 3.982008995502249e-05, + "loss": 0.3406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3238492012023926, + "step": 665, + "valid_targets_mean": 3239.7, + "valid_targets_min": 1679 + }, + { + "epoch": 0.7037815126050421, + "grad_norm": 0.6929968801305512, + "learning_rate": 3.9999989022799437e-05, + "loss": 0.3559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.374039888381958, + "step": 670, + "valid_targets_mean": 3101.2, + "valid_targets_min": 1739 + }, + { + "epoch": 0.7090336134453782, + "grad_norm": 0.7069459947278421, + "learning_rate": 3.9999865529431466e-05, + "loss": 0.3481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3492530584335327, + "step": 675, + "valid_targets_mean": 2844.4, + "valid_targets_min": 1278 + }, + { + "epoch": 0.7142857142857143, + "grad_norm": 0.6222986155844251, + "learning_rate": 3.9999604822044886e-05, + "loss": 0.3406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.327778160572052, + "step": 680, + "valid_targets_mean": 3563.4, + "valid_targets_min": 1522 + }, + { + "epoch": 0.7195378151260504, + "grad_norm": 0.6470886363287454, + "learning_rate": 3.999920690242835e-05, + "loss": 0.3288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33931905031204224, + "step": 685, + "valid_targets_mean": 3589.0, + "valid_targets_min": 464 + }, + { + "epoch": 0.7247899159663865, + "grad_norm": 0.5841030308921137, + "learning_rate": 3.999867177331189e-05, + "loss": 0.362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3474075198173523, + "step": 690, + "valid_targets_mean": 3827.1, + "valid_targets_min": 1003 + }, + { + "epoch": 0.7300420168067226, + "grad_norm": 0.6944347153816217, + "learning_rate": 3.9997999438366895e-05, + "loss": 0.3424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36091533303260803, + "step": 695, + "valid_targets_mean": 3042.1, + "valid_targets_min": 1446 + }, + { + "epoch": 0.7352941176470589, + "grad_norm": 0.6583742822611474, + "learning_rate": 3.9997189902206065e-05, + "loss": 0.3494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3591790199279785, + "step": 700, + "valid_targets_mean": 3257.8, + "valid_targets_min": 1256 + }, + { + "epoch": 0.740546218487395, + "grad_norm": 0.6796220489697876, + "learning_rate": 3.999624317038344e-05, + "loss": 0.3623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3584475815296173, + "step": 705, + "valid_targets_mean": 3858.5, + "valid_targets_min": 1336 + }, + { + "epoch": 0.7457983193277311, + "grad_norm": 0.5779511153503359, + "learning_rate": 3.9995159249394303e-05, + "loss": 0.3632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3402530550956726, + "step": 710, + "valid_targets_mean": 3821.1, + "valid_targets_min": 1275 + }, + { + "epoch": 0.7510504201680672, + "grad_norm": 0.6920943343973615, + "learning_rate": 3.999393814667517e-05, + "loss": 0.3394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35493505001068115, + "step": 715, + "valid_targets_mean": 2914.6, + "valid_targets_min": 644 + }, + { + "epoch": 0.7563025210084033, + "grad_norm": 0.6484863168026516, + "learning_rate": 3.9992579870603695e-05, + "loss": 0.336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3329768478870392, + "step": 720, + "valid_targets_mean": 3083.4, + "valid_targets_min": 1397 + }, + { + "epoch": 0.7615546218487395, + "grad_norm": 0.7459309224625077, + "learning_rate": 3.999108443049869e-05, + "loss": 0.3408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35886648297309875, + "step": 725, + "valid_targets_mean": 2747.4, + "valid_targets_min": 1557 + }, + { + "epoch": 0.7668067226890757, + "grad_norm": 0.6643965239188359, + "learning_rate": 3.998945183661997e-05, + "loss": 0.3442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3370826840400696, + "step": 730, + "valid_targets_mean": 3405.2, + "valid_targets_min": 1405 + }, + { + "epoch": 0.7720588235294118, + "grad_norm": 0.6311941065520537, + "learning_rate": 3.9987682100168376e-05, + "loss": 0.3432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32585620880126953, + "step": 735, + "valid_targets_mean": 3507.1, + "valid_targets_min": 1042 + }, + { + "epoch": 0.7773109243697479, + "grad_norm": 0.6934056789740213, + "learning_rate": 3.99857752332856e-05, + "loss": 0.3318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3426210284233093, + "step": 740, + "valid_targets_mean": 3079.7, + "valid_targets_min": 858 + }, + { + "epoch": 0.782563025210084, + "grad_norm": 0.6683294329205226, + "learning_rate": 3.998373124905418e-05, + "loss": 0.3404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3405384421348572, + "step": 745, + "valid_targets_mean": 2934.3, + "valid_targets_min": 1288 + }, + { + "epoch": 0.7878151260504201, + "grad_norm": 0.6089870902503931, + "learning_rate": 3.9981550161497375e-05, + "loss": 0.36, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3385617136955261, + "step": 750, + "valid_targets_mean": 3428.2, + "valid_targets_min": 1915 + }, + { + "epoch": 0.7930672268907563, + "grad_norm": 0.6028162675947055, + "learning_rate": 3.9979231985579074e-05, + "loss": 0.3484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3657104969024658, + "step": 755, + "valid_targets_mean": 3764.9, + "valid_targets_min": 1244 + }, + { + "epoch": 0.7983193277310925, + "grad_norm": 0.7017817313652911, + "learning_rate": 3.99767767372037e-05, + "loss": 0.3304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33708494901657104, + "step": 760, + "valid_targets_mean": 3006.5, + "valid_targets_min": 1255 + }, + { + "epoch": 0.8035714285714286, + "grad_norm": 0.6788931537841902, + "learning_rate": 3.997418443321609e-05, + "loss": 0.3478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3829076290130615, + "step": 765, + "valid_targets_mean": 3326.2, + "valid_targets_min": 1586 + }, + { + "epoch": 0.8088235294117647, + "grad_norm": 0.6396470227937194, + "learning_rate": 3.997145509140138e-05, + "loss": 0.3451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3249973654747009, + "step": 770, + "valid_targets_mean": 3205.2, + "valid_targets_min": 1181 + }, + { + "epoch": 0.8140756302521008, + "grad_norm": 0.6959358945807129, + "learning_rate": 3.9968588730484896e-05, + "loss": 0.3302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35094955563545227, + "step": 775, + "valid_targets_mean": 2735.1, + "valid_targets_min": 974 + }, + { + "epoch": 0.819327731092437, + "grad_norm": 0.7121078872001974, + "learning_rate": 3.9965585370132013e-05, + "loss": 0.3355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3337578773498535, + "step": 780, + "valid_targets_mean": 3735.8, + "valid_targets_min": 1681 + }, + { + "epoch": 0.8245798319327731, + "grad_norm": 0.731579949310607, + "learning_rate": 3.996244503094804e-05, + "loss": 0.3497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37752842903137207, + "step": 785, + "valid_targets_mean": 2883.7, + "valid_targets_min": 1171 + }, + { + "epoch": 0.8298319327731093, + "grad_norm": 0.6650190290860145, + "learning_rate": 3.995916773447804e-05, + "loss": 0.3574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3751668930053711, + "step": 790, + "valid_targets_mean": 3324.1, + "valid_targets_min": 1455 + }, + { + "epoch": 0.8350840336134454, + "grad_norm": 0.6084187992006697, + "learning_rate": 3.995575350320671e-05, + "loss": 0.3363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3428555428981781, + "step": 795, + "valid_targets_mean": 3948.6, + "valid_targets_min": 1529 + }, + { + "epoch": 0.8403361344537815, + "grad_norm": 0.6734592462466594, + "learning_rate": 3.995220236055824e-05, + "loss": 0.3662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.36074283719062805, + "step": 800, + "valid_targets_mean": 3419.9, + "valid_targets_min": 1450 + }, + { + "epoch": 0.8455882352941176, + "grad_norm": 0.5774931076226504, + "learning_rate": 3.99485143308961e-05, + "loss": 0.3351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3051412105560303, + "step": 805, + "valid_targets_mean": 3465.6, + "valid_targets_min": 1431 + }, + { + "epoch": 0.8508403361344538, + "grad_norm": 0.8400200119736569, + "learning_rate": 3.994468943952296e-05, + "loss": 0.3546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35333842039108276, + "step": 810, + "valid_targets_mean": 3085.4, + "valid_targets_min": 802 + }, + { + "epoch": 0.8560924369747899, + "grad_norm": 0.6336110237839332, + "learning_rate": 3.994072771268041e-05, + "loss": 0.362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.347425639629364, + "step": 815, + "valid_targets_mean": 3230.6, + "valid_targets_min": 1433 + }, + { + "epoch": 0.8613445378151261, + "grad_norm": 0.6958422384825939, + "learning_rate": 3.993662917754885e-05, + "loss": 0.3478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.35881751775741577, + "step": 820, + "valid_targets_mean": 2908.7, + "valid_targets_min": 614 + }, + { + "epoch": 0.8665966386554622, + "grad_norm": 0.7411583695170622, + "learning_rate": 3.993239386224732e-05, + "loss": 0.3536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3721003532409668, + "step": 825, + "valid_targets_mean": 2748.8, + "valid_targets_min": 637 + }, + { + "epoch": 0.8718487394957983, + "grad_norm": 0.63511436606181, + "learning_rate": 3.992802179583322e-05, + "loss": 0.3513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3608887195587158, + "step": 830, + "valid_targets_mean": 3297.7, + "valid_targets_min": 1683 + }, + { + "epoch": 0.8771008403361344, + "grad_norm": 0.6522816701659483, + "learning_rate": 3.9923513008302204e-05, + "loss": 0.3484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32889869809150696, + "step": 835, + "valid_targets_mean": 3406.1, + "valid_targets_min": 1913 + }, + { + "epoch": 0.8823529411764706, + "grad_norm": 0.5886512960386365, + "learning_rate": 3.991886753058792e-05, + "loss": 0.3322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32120782136917114, + "step": 840, + "valid_targets_mean": 3848.1, + "valid_targets_min": 1659 + }, + { + "epoch": 0.8876050420168067, + "grad_norm": 0.6037677731592493, + "learning_rate": 3.991408539456182e-05, + "loss": 0.3476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31538859009742737, + "step": 845, + "valid_targets_mean": 3390.7, + "valid_targets_min": 1390 + }, + { + "epoch": 0.8928571428571429, + "grad_norm": 0.7282746261851238, + "learning_rate": 3.990916663303293e-05, + "loss": 0.351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.4035952389240265, + "step": 850, + "valid_targets_mean": 3338.5, + "valid_targets_min": 611 + }, + { + "epoch": 0.898109243697479, + "grad_norm": 0.8755920451444243, + "learning_rate": 3.990411127974762e-05, + "loss": 0.3411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3417074978351593, + "step": 855, + "valid_targets_mean": 3389.6, + "valid_targets_min": 1451 + }, + { + "epoch": 0.9033613445378151, + "grad_norm": 0.6595337996347566, + "learning_rate": 3.989891936938939e-05, + "loss": 0.349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31698334217071533, + "step": 860, + "valid_targets_mean": 3097.9, + "valid_targets_min": 1459 + }, + { + "epoch": 0.9086134453781513, + "grad_norm": 0.7205543746187569, + "learning_rate": 3.9893590937578634e-05, + "loss": 0.3356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37621456384658813, + "step": 865, + "valid_targets_mean": 2619.4, + "valid_targets_min": 999 + }, + { + "epoch": 0.9138655462184874, + "grad_norm": 0.5683063041911645, + "learning_rate": 3.9888126020872375e-05, + "loss": 0.3317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31596362590789795, + "step": 870, + "valid_targets_mean": 3791.2, + "valid_targets_min": 1564 + }, + { + "epoch": 0.9191176470588235, + "grad_norm": 1.070844790253151, + "learning_rate": 3.988252465676401e-05, + "loss": 0.3364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33733999729156494, + "step": 875, + "valid_targets_mean": 3223.3, + "valid_targets_min": 1233 + }, + { + "epoch": 0.9243697478991597, + "grad_norm": 0.637333593228743, + "learning_rate": 3.98767868836831e-05, + "loss": 0.3193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3216400146484375, + "step": 880, + "valid_targets_mean": 3439.2, + "valid_targets_min": 1524 + }, + { + "epoch": 0.9296218487394958, + "grad_norm": 0.5754820914455384, + "learning_rate": 3.987091274099504e-05, + "loss": 0.3239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3149118423461914, + "step": 885, + "valid_targets_mean": 3744.9, + "valid_targets_min": 1375 + }, + { + "epoch": 0.9348739495798319, + "grad_norm": 0.6651917315458717, + "learning_rate": 3.986490226900084e-05, + "loss": 0.3368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3212442398071289, + "step": 890, + "valid_targets_mean": 3036.2, + "valid_targets_min": 906 + }, + { + "epoch": 0.9401260504201681, + "grad_norm": 0.6067258438956236, + "learning_rate": 3.985875550893684e-05, + "loss": 0.3232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30532318353652954, + "step": 895, + "valid_targets_mean": 3766.5, + "valid_targets_min": 1246 + }, + { + "epoch": 0.9453781512605042, + "grad_norm": 0.6702408814712129, + "learning_rate": 3.9852472502974386e-05, + "loss": 0.3288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3562542498111725, + "step": 900, + "valid_targets_mean": 3225.9, + "valid_targets_min": 1333 + }, + { + "epoch": 0.9506302521008403, + "grad_norm": 0.5790810375186305, + "learning_rate": 3.984605329421961e-05, + "loss": 0.3279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32805150747299194, + "step": 905, + "valid_targets_mean": 3751.3, + "valid_targets_min": 2264 + }, + { + "epoch": 0.9558823529411765, + "grad_norm": 0.5978703507597435, + "learning_rate": 3.983949792671307e-05, + "loss": 0.3481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.347832053899765, + "step": 910, + "valid_targets_mean": 3505.2, + "valid_targets_min": 1454 + }, + { + "epoch": 0.9611344537815126, + "grad_norm": 0.624321623961909, + "learning_rate": 3.9832806445429486e-05, + "loss": 0.3425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3324778378009796, + "step": 915, + "valid_targets_mean": 3385.7, + "valid_targets_min": 1430 + }, + { + "epoch": 0.9663865546218487, + "grad_norm": 0.6245540796780895, + "learning_rate": 3.982597889627742e-05, + "loss": 0.3348, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.317277729511261, + "step": 920, + "valid_targets_mean": 3201.6, + "valid_targets_min": 1433 + }, + { + "epoch": 0.9716386554621849, + "grad_norm": 0.6175278753890704, + "learning_rate": 3.981901532609896e-05, + "loss": 0.3239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32414883375167847, + "step": 925, + "valid_targets_mean": 3362.1, + "valid_targets_min": 1512 + }, + { + "epoch": 0.976890756302521, + "grad_norm": 0.6165488317026186, + "learning_rate": 3.9811915782669406e-05, + "loss": 0.3298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31244921684265137, + "step": 930, + "valid_targets_mean": 3643.9, + "valid_targets_min": 2048 + }, + { + "epoch": 0.9821428571428571, + "grad_norm": 0.6227928286121143, + "learning_rate": 3.980468031469691e-05, + "loss": 0.3463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3477458953857422, + "step": 935, + "valid_targets_mean": 3107.1, + "valid_targets_min": 815 + }, + { + "epoch": 0.9873949579831933, + "grad_norm": 0.6175153136159267, + "learning_rate": 3.9797308971822206e-05, + "loss": 0.3446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3596673607826233, + "step": 940, + "valid_targets_mean": 3339.4, + "valid_targets_min": 1878 + }, + { + "epoch": 0.9926470588235294, + "grad_norm": 0.6173093585223963, + "learning_rate": 3.978980180461821e-05, + "loss": 0.3367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33542400598526, + "step": 945, + "valid_targets_mean": 3394.6, + "valid_targets_min": 1727 + }, + { + "epoch": 0.9978991596638656, + "grad_norm": 0.6354828168457491, + "learning_rate": 3.9782158864589696e-05, + "loss": 0.3261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3046625554561615, + "step": 950, + "valid_targets_mean": 2924.3, + "valid_targets_min": 1622 + }, + { + "epoch": 1.0031512605042017, + "grad_norm": 1.3016888810382736, + "learning_rate": 3.977438020417293e-05, + "loss": 0.3059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2962370216846466, + "step": 955, + "valid_targets_mean": 1587.9, + "valid_targets_min": 788 + }, + { + "epoch": 1.0084033613445378, + "grad_norm": 1.178432388446946, + "learning_rate": 3.9766465876735354e-05, + "loss": 0.2533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23576419055461884, + "step": 960, + "valid_targets_mean": 1591.2, + "valid_targets_min": 957 + }, + { + "epoch": 1.013655462184874, + "grad_norm": 0.9417299138332059, + "learning_rate": 3.9758415936575155e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23771171271800995, + "step": 965, + "valid_targets_mean": 1418.7, + "valid_targets_min": 796 + }, + { + "epoch": 1.01890756302521, + "grad_norm": 0.9391474883661355, + "learning_rate": 3.975023043892094e-05, + "loss": 0.2429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23540674149990082, + "step": 970, + "valid_targets_mean": 1389.0, + "valid_targets_min": 704 + }, + { + "epoch": 1.0241596638655461, + "grad_norm": 0.8489834118982603, + "learning_rate": 3.974190943993133e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.221901997923851, + "step": 975, + "valid_targets_mean": 1513.9, + "valid_targets_min": 638 + }, + { + "epoch": 1.0294117647058822, + "grad_norm": 0.830262362849459, + "learning_rate": 3.973345299669461e-05, + "loss": 0.2401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23195049166679382, + "step": 980, + "valid_targets_mean": 1482.3, + "valid_targets_min": 727 + }, + { + "epoch": 1.0346638655462186, + "grad_norm": 0.8915461834880904, + "learning_rate": 3.972486116722828e-05, + "loss": 0.2468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.208994060754776, + "step": 985, + "valid_targets_mean": 1189.1, + "valid_targets_min": 655 + }, + { + "epoch": 1.0399159663865547, + "grad_norm": 0.9988345647892146, + "learning_rate": 3.971613401047872e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2332269549369812, + "step": 990, + "valid_targets_mean": 1299.6, + "valid_targets_min": 714 + }, + { + "epoch": 1.0451680672268908, + "grad_norm": 0.8523206270419763, + "learning_rate": 3.970727158632075e-05, + "loss": 0.2491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23563072085380554, + "step": 995, + "valid_targets_mean": 1407.6, + "valid_targets_min": 741 + }, + { + "epoch": 1.050420168067227, + "grad_norm": 1.0576241067376564, + "learning_rate": 3.969827395555721e-05, + "loss": 0.235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2590252757072449, + "step": 1000, + "valid_targets_mean": 1544.1, + "valid_targets_min": 663 + }, + { + "epoch": 1.055672268907563, + "grad_norm": 1.1458471492556306, + "learning_rate": 3.968914117991857e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.230397567152977, + "step": 1005, + "valid_targets_mean": 1511.1, + "valid_targets_min": 632 + }, + { + "epoch": 1.0609243697478992, + "grad_norm": 0.9006777040659446, + "learning_rate": 3.967987332206249e-05, + "loss": 0.2396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2451692819595337, + "step": 1010, + "valid_targets_mean": 1638.0, + "valid_targets_min": 933 + }, + { + "epoch": 1.0661764705882353, + "grad_norm": 0.7626623690643839, + "learning_rate": 3.967047044557341e-05, + "loss": 0.2329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22060266137123108, + "step": 1015, + "valid_targets_mean": 1526.9, + "valid_targets_min": 717 + }, + { + "epoch": 1.0714285714285714, + "grad_norm": 0.8215685697177867, + "learning_rate": 3.966093261496208e-05, + "loss": 0.2217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2275817096233368, + "step": 1020, + "valid_targets_mean": 1387.4, + "valid_targets_min": 659 + }, + { + "epoch": 1.0766806722689075, + "grad_norm": 0.8698242331020952, + "learning_rate": 3.9651259895665146e-05, + "loss": 0.2212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2096673548221588, + "step": 1025, + "valid_targets_mean": 1382.9, + "valid_targets_min": 642 + }, + { + "epoch": 1.0819327731092436, + "grad_norm": 0.768616027752401, + "learning_rate": 3.9641452354044687e-05, + "loss": 0.2263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20485013723373413, + "step": 1030, + "valid_targets_mean": 1312.1, + "valid_targets_min": 765 + }, + { + "epoch": 1.0871848739495797, + "grad_norm": 0.8590772693317404, + "learning_rate": 3.9631510057387765e-05, + "loss": 0.2403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24823632836341858, + "step": 1035, + "valid_targets_mean": 1397.6, + "valid_targets_min": 631 + }, + { + "epoch": 1.092436974789916, + "grad_norm": 0.9234663166300711, + "learning_rate": 3.962143307390596e-05, + "loss": 0.236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2525438666343689, + "step": 1040, + "valid_targets_mean": 1419.4, + "valid_targets_min": 623 + }, + { + "epoch": 1.0976890756302522, + "grad_norm": 0.7428187089700554, + "learning_rate": 3.961122147273491e-05, + "loss": 0.2362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24657368659973145, + "step": 1045, + "valid_targets_mean": 1647.7, + "valid_targets_min": 818 + }, + { + "epoch": 1.1029411764705883, + "grad_norm": 0.7835107757413885, + "learning_rate": 3.9600875323933826e-05, + "loss": 0.2294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21424023807048798, + "step": 1050, + "valid_targets_mean": 1545.5, + "valid_targets_min": 776 + }, + { + "epoch": 1.1081932773109244, + "grad_norm": 0.8623930769993505, + "learning_rate": 3.959039469848502e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2390882670879364, + "step": 1055, + "valid_targets_mean": 1411.8, + "valid_targets_min": 623 + }, + { + "epoch": 1.1134453781512605, + "grad_norm": 0.8342187952076956, + "learning_rate": 3.9579779668293416e-05, + "loss": 0.2341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24478605389595032, + "step": 1060, + "valid_targets_mean": 1437.3, + "valid_targets_min": 742 + }, + { + "epoch": 1.1186974789915967, + "grad_norm": 0.8716802159762432, + "learning_rate": 3.956903030618605e-05, + "loss": 0.229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21243079006671906, + "step": 1065, + "valid_targets_mean": 1193.6, + "valid_targets_min": 611 + }, + { + "epoch": 1.1239495798319328, + "grad_norm": 0.8276094439043256, + "learning_rate": 3.955814668591156e-05, + "loss": 0.221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2292919009923935, + "step": 1070, + "valid_targets_mean": 1514.0, + "valid_targets_min": 793 + }, + { + "epoch": 1.129201680672269, + "grad_norm": 0.8269530258415205, + "learning_rate": 3.9547128882139735e-05, + "loss": 0.2218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2063109278678894, + "step": 1075, + "valid_targets_mean": 1379.1, + "valid_targets_min": 886 + }, + { + "epoch": 1.134453781512605, + "grad_norm": 0.7349504784818733, + "learning_rate": 3.9535976970460925e-05, + "loss": 0.222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2086534947156906, + "step": 1080, + "valid_targets_mean": 1455.5, + "valid_targets_min": 991 + }, + { + "epoch": 1.1397058823529411, + "grad_norm": 0.7910940364262357, + "learning_rate": 3.9524691027385585e-05, + "loss": 0.2244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22840695083141327, + "step": 1085, + "valid_targets_mean": 1525.2, + "valid_targets_min": 908 + }, + { + "epoch": 1.1449579831932772, + "grad_norm": 0.9125068091813882, + "learning_rate": 3.95132711303437e-05, + "loss": 0.234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23359861969947815, + "step": 1090, + "valid_targets_mean": 1476.6, + "valid_targets_min": 680 + }, + { + "epoch": 1.1502100840336134, + "grad_norm": 0.7517080972876787, + "learning_rate": 3.9501717357684315e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22301185131072998, + "step": 1095, + "valid_targets_mean": 1467.1, + "valid_targets_min": 803 + }, + { + "epoch": 1.1554621848739495, + "grad_norm": 0.8187151581180054, + "learning_rate": 3.9490029788674934e-05, + "loss": 0.224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22149021923542023, + "step": 1100, + "valid_targets_mean": 1341.3, + "valid_targets_min": 765 + }, + { + "epoch": 1.1607142857142858, + "grad_norm": 0.7811554150228613, + "learning_rate": 3.9478208503501e-05, + "loss": 0.2245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23048382997512817, + "step": 1105, + "valid_targets_mean": 1556.4, + "valid_targets_min": 804 + }, + { + "epoch": 1.165966386554622, + "grad_norm": 0.8232786153715305, + "learning_rate": 3.946625358326538e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24594677984714508, + "step": 1110, + "valid_targets_mean": 1509.7, + "valid_targets_min": 675 + }, + { + "epoch": 1.171218487394958, + "grad_norm": 0.7583404192175697, + "learning_rate": 3.945416510998775e-05, + "loss": 0.2165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.229123055934906, + "step": 1115, + "valid_targets_mean": 1408.4, + "valid_targets_min": 882 + }, + { + "epoch": 1.1764705882352942, + "grad_norm": 0.7798738937206717, + "learning_rate": 3.944194316660406e-05, + "loss": 0.216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20927229523658752, + "step": 1120, + "valid_targets_mean": 1449.8, + "valid_targets_min": 660 + }, + { + "epoch": 1.1817226890756303, + "grad_norm": 0.7782369720225292, + "learning_rate": 3.942958783696598e-05, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21136952936649323, + "step": 1125, + "valid_targets_mean": 1489.8, + "valid_targets_min": 678 + }, + { + "epoch": 1.1869747899159664, + "grad_norm": 0.8389164365291358, + "learning_rate": 3.94170992058403e-05, + "loss": 0.2194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2247747778892517, + "step": 1130, + "valid_targets_mean": 1379.1, + "valid_targets_min": 676 + }, + { + "epoch": 1.1922268907563025, + "grad_norm": 0.85218432627087, + "learning_rate": 3.9404477358908354e-05, + "loss": 0.2185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20599870383739471, + "step": 1135, + "valid_targets_mean": 1347.1, + "valid_targets_min": 794 + }, + { + "epoch": 1.1974789915966386, + "grad_norm": 0.7736857874589698, + "learning_rate": 3.9391722382765445e-05, + "loss": 0.2194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2081984579563141, + "step": 1140, + "valid_targets_mean": 1372.6, + "valid_targets_min": 607 + }, + { + "epoch": 1.2027310924369747, + "grad_norm": 0.8143324846381077, + "learning_rate": 3.937883436492025e-05, + "loss": 0.2181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.210773304104805, + "step": 1145, + "valid_targets_mean": 1323.6, + "valid_targets_min": 722 + }, + { + "epoch": 1.2079831932773109, + "grad_norm": 0.8168389117740881, + "learning_rate": 3.9365813393794186e-05, + "loss": 0.2306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2400819957256317, + "step": 1150, + "valid_targets_mean": 1597.7, + "valid_targets_min": 858 + }, + { + "epoch": 1.213235294117647, + "grad_norm": 0.798484953443434, + "learning_rate": 3.9352659558720836e-05, + "loss": 0.2018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19974064826965332, + "step": 1155, + "valid_targets_mean": 1270.8, + "valid_targets_min": 689 + }, + { + "epoch": 1.2184873949579833, + "grad_norm": 0.7597271913919195, + "learning_rate": 3.933937294994535e-05, + "loss": 0.2306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22933819890022278, + "step": 1160, + "valid_targets_mean": 1388.2, + "valid_targets_min": 692 + }, + { + "epoch": 1.2237394957983194, + "grad_norm": 0.7412896199433119, + "learning_rate": 3.932595365862379e-05, + "loss": 0.2194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2073134481906891, + "step": 1165, + "valid_targets_mean": 1447.2, + "valid_targets_min": 828 + }, + { + "epoch": 1.2289915966386555, + "grad_norm": 0.8029587259024663, + "learning_rate": 3.9312401776822504e-05, + "loss": 0.235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22155123949050903, + "step": 1170, + "valid_targets_mean": 1399.7, + "valid_targets_min": 618 + }, + { + "epoch": 1.2342436974789917, + "grad_norm": 0.753092266372571, + "learning_rate": 3.9298717397517546e-05, + "loss": 0.2104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20688486099243164, + "step": 1175, + "valid_targets_mean": 1454.3, + "valid_targets_min": 629 + }, + { + "epoch": 1.2394957983193278, + "grad_norm": 0.7552069088401491, + "learning_rate": 3.928490061459396e-05, + "loss": 0.2193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2513880133628845, + "step": 1180, + "valid_targets_mean": 1746.1, + "valid_targets_min": 1071 + }, + { + "epoch": 1.2447478991596639, + "grad_norm": 0.7724105286503384, + "learning_rate": 3.927095152284521e-05, + "loss": 0.2215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2342674434185028, + "step": 1185, + "valid_targets_mean": 1282.2, + "valid_targets_min": 677 + }, + { + "epoch": 1.25, + "grad_norm": 0.8255573649270425, + "learning_rate": 3.925687021797249e-05, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19916367530822754, + "step": 1190, + "valid_targets_mean": 1227.6, + "valid_targets_min": 679 + }, + { + "epoch": 1.2552521008403361, + "grad_norm": 0.7348173616246707, + "learning_rate": 3.924265679658407e-05, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20118603110313416, + "step": 1195, + "valid_targets_mean": 1319.1, + "valid_targets_min": 504 + }, + { + "epoch": 1.2605042016806722, + "grad_norm": 0.7536987070404887, + "learning_rate": 3.922831135619462e-05, + "loss": 0.2161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20859253406524658, + "step": 1200, + "valid_targets_mean": 1616.5, + "valid_targets_min": 1023 + }, + { + "epoch": 1.2657563025210083, + "grad_norm": 0.7577745081343713, + "learning_rate": 3.9213833995224605e-05, + "loss": 0.2203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2004815638065338, + "step": 1205, + "valid_targets_mean": 1453.0, + "valid_targets_min": 661 + }, + { + "epoch": 1.2710084033613445, + "grad_norm": 0.7784879343054987, + "learning_rate": 3.919922481299952e-05, + "loss": 0.222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21279674768447876, + "step": 1210, + "valid_targets_mean": 1442.0, + "valid_targets_min": 889 + }, + { + "epoch": 1.2762605042016806, + "grad_norm": 0.7318237431101263, + "learning_rate": 3.918448390974928e-05, + "loss": 0.2088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20114900171756744, + "step": 1215, + "valid_targets_mean": 1548.4, + "valid_targets_min": 723 + }, + { + "epoch": 1.2815126050420167, + "grad_norm": 0.8439347277266402, + "learning_rate": 3.9169611386607476e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20195883512496948, + "step": 1220, + "valid_targets_mean": 1359.8, + "valid_targets_min": 793 + }, + { + "epoch": 1.2867647058823528, + "grad_norm": 0.7840655198861911, + "learning_rate": 3.9154607345610746e-05, + "loss": 0.2145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22591176629066467, + "step": 1225, + "valid_targets_mean": 1478.6, + "valid_targets_min": 861 + }, + { + "epoch": 1.2920168067226891, + "grad_norm": 0.7776629349610283, + "learning_rate": 3.913947188969801e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19768911600112915, + "step": 1230, + "valid_targets_mean": 1332.7, + "valid_targets_min": 752 + }, + { + "epoch": 1.2972689075630253, + "grad_norm": 0.7729134790563235, + "learning_rate": 3.912420512270981e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2305484265089035, + "step": 1235, + "valid_targets_mean": 1622.7, + "valid_targets_min": 789 + }, + { + "epoch": 1.3025210084033614, + "grad_norm": 0.8616431181942282, + "learning_rate": 3.910880714938757e-05, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1932060420513153, + "step": 1240, + "valid_targets_mean": 1195.4, + "valid_targets_min": 567 + }, + { + "epoch": 1.3077731092436975, + "grad_norm": 0.7695069319629071, + "learning_rate": 3.909327807537288e-05, + "loss": 0.213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.232661634683609, + "step": 1245, + "valid_targets_mean": 1504.5, + "valid_targets_min": 605 + }, + { + "epoch": 1.3130252100840336, + "grad_norm": 0.7113969327868939, + "learning_rate": 3.90776180072068e-05, + "loss": 0.2116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20750564336776733, + "step": 1250, + "valid_targets_mean": 1580.1, + "valid_targets_min": 746 + }, + { + "epoch": 1.3182773109243697, + "grad_norm": 0.7945153800878063, + "learning_rate": 3.906182705232909e-05, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20195449888706207, + "step": 1255, + "valid_targets_mean": 1158.2, + "valid_targets_min": 723 + }, + { + "epoch": 1.3235294117647058, + "grad_norm": 0.7587646804913716, + "learning_rate": 3.904590531907751e-05, + "loss": 0.2159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21249453723430634, + "step": 1260, + "valid_targets_mean": 1430.4, + "valid_targets_min": 855 + }, + { + "epoch": 1.328781512605042, + "grad_norm": 0.7185755186164045, + "learning_rate": 3.9029852916687034e-05, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2006302922964096, + "step": 1265, + "valid_targets_mean": 1590.4, + "valid_targets_min": 902 + }, + { + "epoch": 1.334033613445378, + "grad_norm": 0.7257232800212272, + "learning_rate": 3.9013669955289136e-05, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20573708415031433, + "step": 1270, + "valid_targets_mean": 1521.5, + "valid_targets_min": 803 + }, + { + "epoch": 1.3392857142857144, + "grad_norm": 0.7974748283202712, + "learning_rate": 3.899735654591101e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22591374814510345, + "step": 1275, + "valid_targets_mean": 1455.2, + "valid_targets_min": 740 + }, + { + "epoch": 1.3445378151260505, + "grad_norm": 0.8433766016494005, + "learning_rate": 3.898091280047486e-05, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20434291660785675, + "step": 1280, + "valid_targets_mean": 1279.4, + "valid_targets_min": 244 + }, + { + "epoch": 1.3497899159663866, + "grad_norm": 0.7810238280537369, + "learning_rate": 3.896433883179703e-05, + "loss": 0.2212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22051532566547394, + "step": 1285, + "valid_targets_mean": 1366.9, + "valid_targets_min": 714 + }, + { + "epoch": 1.3550420168067228, + "grad_norm": 0.9021667080271751, + "learning_rate": 3.894763475358736e-05, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19154158234596252, + "step": 1290, + "valid_targets_mean": 1220.5, + "valid_targets_min": 585 + }, + { + "epoch": 1.3602941176470589, + "grad_norm": 0.8197315751887221, + "learning_rate": 3.8930800680448275e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.230904221534729, + "step": 1295, + "valid_targets_mean": 1596.4, + "valid_targets_min": 1041 + }, + { + "epoch": 1.365546218487395, + "grad_norm": 0.7275089472098615, + "learning_rate": 3.891383672787411e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20911553502082825, + "step": 1300, + "valid_targets_mean": 1401.3, + "valid_targets_min": 777 + }, + { + "epoch": 1.370798319327731, + "grad_norm": 0.7376318715306975, + "learning_rate": 3.889674301225025e-05, + "loss": 0.2162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19725364446640015, + "step": 1305, + "valid_targets_mean": 1546.8, + "valid_targets_min": 959 + }, + { + "epoch": 1.3760504201680672, + "grad_norm": 0.7408476899885955, + "learning_rate": 3.8879519650852356e-05, + "loss": 0.2053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18753978610038757, + "step": 1310, + "valid_targets_mean": 1284.4, + "valid_targets_min": 620 + }, + { + "epoch": 1.3813025210084033, + "grad_norm": 0.7266684151521309, + "learning_rate": 3.886216676184555e-05, + "loss": 0.214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20051538944244385, + "step": 1315, + "valid_targets_mean": 1356.6, + "valid_targets_min": 656 + }, + { + "epoch": 1.3865546218487395, + "grad_norm": 0.7791953600527686, + "learning_rate": 3.8844684464283614e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23191730678081512, + "step": 1320, + "valid_targets_mean": 1564.5, + "valid_targets_min": 697 + }, + { + "epoch": 1.3918067226890756, + "grad_norm": 0.6803022151794903, + "learning_rate": 3.882707287810817e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20292726159095764, + "step": 1325, + "valid_targets_mean": 1604.7, + "valid_targets_min": 682 + }, + { + "epoch": 1.3970588235294117, + "grad_norm": 0.7109648477839206, + "learning_rate": 3.880933212414786e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19687709212303162, + "step": 1330, + "valid_targets_mean": 1383.4, + "valid_targets_min": 730 + }, + { + "epoch": 1.4023109243697478, + "grad_norm": 0.7636435446308724, + "learning_rate": 3.87914623241175e-05, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19838300347328186, + "step": 1335, + "valid_targets_mean": 1482.5, + "valid_targets_min": 747 + }, + { + "epoch": 1.407563025210084, + "grad_norm": 0.7790600366170161, + "learning_rate": 3.877346360061728e-05, + "loss": 0.2213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20176908373832703, + "step": 1340, + "valid_targets_mean": 1603.6, + "valid_targets_min": 933 + }, + { + "epoch": 1.41281512605042, + "grad_norm": 0.7307120382472138, + "learning_rate": 3.8755336077131894e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19978278875350952, + "step": 1345, + "valid_targets_mean": 1517.9, + "valid_targets_min": 799 + }, + { + "epoch": 1.4180672268907564, + "grad_norm": 0.7009558460825884, + "learning_rate": 3.873707987802967e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20674484968185425, + "step": 1350, + "valid_targets_mean": 1516.2, + "valid_targets_min": 909 + }, + { + "epoch": 1.4233193277310925, + "grad_norm": 0.6935543353633367, + "learning_rate": 3.871869512856179e-05, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19707924127578735, + "step": 1355, + "valid_targets_mean": 1606.8, + "valid_targets_min": 741 + }, + { + "epoch": 1.4285714285714286, + "grad_norm": 0.7721082945840706, + "learning_rate": 3.870018195486138e-05, + "loss": 0.1903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20531484484672546, + "step": 1360, + "valid_targets_mean": 1328.7, + "valid_targets_min": 720 + }, + { + "epoch": 1.4338235294117647, + "grad_norm": 0.7736104466743927, + "learning_rate": 3.868154048394262e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.201510488986969, + "step": 1365, + "valid_targets_mean": 1396.2, + "valid_targets_min": 610 + }, + { + "epoch": 1.4390756302521008, + "grad_norm": 0.7550538776229306, + "learning_rate": 3.8662770843699944e-05, + "loss": 0.204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2158433496952057, + "step": 1370, + "valid_targets_mean": 1352.9, + "valid_targets_min": 684 + }, + { + "epoch": 1.444327731092437, + "grad_norm": 0.6820026840811728, + "learning_rate": 3.8643873162907086e-05, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19932374358177185, + "step": 1375, + "valid_targets_mean": 1479.4, + "valid_targets_min": 874 + }, + { + "epoch": 1.449579831932773, + "grad_norm": 0.6850687148845401, + "learning_rate": 3.862484757121627e-05, + "loss": 0.2008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19579848647117615, + "step": 1380, + "valid_targets_mean": 1501.2, + "valid_targets_min": 765 + }, + { + "epoch": 1.4548319327731092, + "grad_norm": 0.6990618001541059, + "learning_rate": 3.860569419915727e-05, + "loss": 0.2082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20401817560195923, + "step": 1385, + "valid_targets_mean": 1419.9, + "valid_targets_min": 589 + }, + { + "epoch": 1.4600840336134453, + "grad_norm": 0.699051470230191, + "learning_rate": 3.858641317813653e-05, + "loss": 0.2151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23287919163703918, + "step": 1390, + "valid_targets_mean": 1587.2, + "valid_targets_min": 918 + }, + { + "epoch": 1.4653361344537816, + "grad_norm": 0.5943455101511255, + "learning_rate": 3.8567004640436234e-05, + "loss": 0.3011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3295353651046753, + "step": 1395, + "valid_targets_mean": 3546.9, + "valid_targets_min": 702 + }, + { + "epoch": 1.4705882352941178, + "grad_norm": 0.570252231140088, + "learning_rate": 3.8547468719213476e-05, + "loss": 0.3267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33764392137527466, + "step": 1400, + "valid_targets_mean": 4000.6, + "valid_targets_min": 1005 + }, + { + "epoch": 1.4758403361344539, + "grad_norm": 0.5515303020894601, + "learning_rate": 3.852780554849925e-05, + "loss": 0.3055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3034968376159668, + "step": 1405, + "valid_targets_mean": 4063.9, + "valid_targets_min": 1296 + }, + { + "epoch": 1.48109243697479, + "grad_norm": 0.5595030961730294, + "learning_rate": 3.850801526319761e-05, + "loss": 0.3004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.288005530834198, + "step": 1410, + "valid_targets_mean": 3747.0, + "valid_targets_min": 1561 + }, + { + "epoch": 1.486344537815126, + "grad_norm": 0.5948008178131649, + "learning_rate": 3.84880979990847e-05, + "loss": 0.304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32669079303741455, + "step": 1415, + "valid_targets_mean": 3562.5, + "valid_targets_min": 1044 + }, + { + "epoch": 1.4915966386554622, + "grad_norm": 0.5216067222057376, + "learning_rate": 3.846805389280782e-05, + "loss": 0.3154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2934839725494385, + "step": 1420, + "valid_targets_mean": 3911.4, + "valid_targets_min": 920 + }, + { + "epoch": 1.4968487394957983, + "grad_norm": 0.4836038411370819, + "learning_rate": 3.844788308188452e-05, + "loss": 0.2861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3024061918258667, + "step": 1425, + "valid_targets_mean": 5753.8, + "valid_targets_min": 2060 + }, + { + "epoch": 1.5021008403361344, + "grad_norm": 0.5556606445880677, + "learning_rate": 3.8427585704701634e-05, + "loss": 0.3043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31254953145980835, + "step": 1430, + "valid_targets_mean": 3728.6, + "valid_targets_min": 467 + }, + { + "epoch": 1.5073529411764706, + "grad_norm": 0.5895498142754928, + "learning_rate": 3.840716190051433e-05, + "loss": 0.3141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32763177156448364, + "step": 1435, + "valid_targets_mean": 3554.0, + "valid_targets_min": 1689 + }, + { + "epoch": 1.5126050420168067, + "grad_norm": 0.54899724482139, + "learning_rate": 3.8386611809445155e-05, + "loss": 0.3091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30432096123695374, + "step": 1440, + "valid_targets_mean": 3708.1, + "valid_targets_min": 1697 + }, + { + "epoch": 1.5178571428571428, + "grad_norm": 0.6115854171093719, + "learning_rate": 3.8365935572483095e-05, + "loss": 0.3162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3384946882724762, + "step": 1445, + "valid_targets_mean": 3652.6, + "valid_targets_min": 1223 + }, + { + "epoch": 1.523109243697479, + "grad_norm": 0.5435695015200867, + "learning_rate": 3.834513333148258e-05, + "loss": 0.3085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3052658438682556, + "step": 1450, + "valid_targets_mean": 4036.1, + "valid_targets_min": 1552 + }, + { + "epoch": 1.528361344537815, + "grad_norm": 0.614280577128499, + "learning_rate": 3.832420522916252e-05, + "loss": 0.3084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.316119909286499, + "step": 1455, + "valid_targets_mean": 3543.5, + "valid_targets_min": 1439 + }, + { + "epoch": 1.5336134453781511, + "grad_norm": 0.5885577114747288, + "learning_rate": 3.830315140910534e-05, + "loss": 0.3058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30280590057373047, + "step": 1460, + "valid_targets_mean": 3287.0, + "valid_targets_min": 1402 + }, + { + "epoch": 1.5388655462184873, + "grad_norm": 0.6135025188561368, + "learning_rate": 3.8281972015755965e-05, + "loss": 0.3204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3206782341003418, + "step": 1465, + "valid_targets_mean": 3423.9, + "valid_targets_min": 1586 + }, + { + "epoch": 1.5441176470588234, + "grad_norm": 0.584621502686007, + "learning_rate": 3.826066719442086e-05, + "loss": 0.3117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3108097314834595, + "step": 1470, + "valid_targets_mean": 3406.0, + "valid_targets_min": 1538 + }, + { + "epoch": 1.5493697478991597, + "grad_norm": 0.6706962593993887, + "learning_rate": 3.823923709126701e-05, + "loss": 0.3158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31825387477874756, + "step": 1475, + "valid_targets_mean": 2954.0, + "valid_targets_min": 1058 + }, + { + "epoch": 1.5546218487394958, + "grad_norm": 0.651116388475916, + "learning_rate": 3.821768185332095e-05, + "loss": 0.3305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32802870869636536, + "step": 1480, + "valid_targets_mean": 3051.8, + "valid_targets_min": 1362 + }, + { + "epoch": 1.559873949579832, + "grad_norm": 0.5745329146566351, + "learning_rate": 3.81960016284677e-05, + "loss": 0.3145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3129153251647949, + "step": 1485, + "valid_targets_mean": 3620.5, + "valid_targets_min": 1392 + }, + { + "epoch": 1.565126050420168, + "grad_norm": 0.6580272364249068, + "learning_rate": 3.817419656544979e-05, + "loss": 0.3154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3198480010032654, + "step": 1490, + "valid_targets_mean": 3103.6, + "valid_targets_min": 1267 + }, + { + "epoch": 1.5703781512605042, + "grad_norm": 0.6033632847231352, + "learning_rate": 3.815226681386626e-05, + "loss": 0.3079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3069283664226532, + "step": 1495, + "valid_targets_mean": 3592.7, + "valid_targets_min": 1026 + }, + { + "epoch": 1.5756302521008403, + "grad_norm": 0.5732499269923438, + "learning_rate": 3.8130212524171576e-05, + "loss": 0.2966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30469539761543274, + "step": 1500, + "valid_targets_mean": 3632.1, + "valid_targets_min": 1785 + }, + { + "epoch": 1.5808823529411766, + "grad_norm": 0.5775861462441018, + "learning_rate": 3.810803384767465e-05, + "loss": 0.319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2886812388896942, + "step": 1505, + "valid_targets_mean": 3557.1, + "valid_targets_min": 1957 + }, + { + "epoch": 1.5861344537815127, + "grad_norm": 0.5694130410160565, + "learning_rate": 3.808573093653777e-05, + "loss": 0.315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2857256233692169, + "step": 1510, + "valid_targets_mean": 3230.6, + "valid_targets_min": 1570 + }, + { + "epoch": 1.5913865546218489, + "grad_norm": 0.6246265430681128, + "learning_rate": 3.806330394377556e-05, + "loss": 0.3086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30203184485435486, + "step": 1515, + "valid_targets_mean": 3023.7, + "valid_targets_min": 997 + }, + { + "epoch": 1.596638655462185, + "grad_norm": 0.658813587454356, + "learning_rate": 3.8040753023253956e-05, + "loss": 0.3007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33436664938926697, + "step": 1520, + "valid_targets_mean": 3056.6, + "valid_targets_min": 1293 + }, + { + "epoch": 1.601890756302521, + "grad_norm": 0.6211106779673433, + "learning_rate": 3.801807832968912e-05, + "loss": 0.3093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29863476753234863, + "step": 1525, + "valid_targets_mean": 2915.6, + "valid_targets_min": 1687 + }, + { + "epoch": 1.6071428571428572, + "grad_norm": 0.6368491555844873, + "learning_rate": 3.799528001864637e-05, + "loss": 0.3068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.33547699451446533, + "step": 1530, + "valid_targets_mean": 2985.1, + "valid_targets_min": 1108 + }, + { + "epoch": 1.6123949579831933, + "grad_norm": 0.5840908746370534, + "learning_rate": 3.797235824653918e-05, + "loss": 0.3048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3032509684562683, + "step": 1535, + "valid_targets_mean": 3195.0, + "valid_targets_min": 571 + }, + { + "epoch": 1.6176470588235294, + "grad_norm": 0.6352859314737174, + "learning_rate": 3.7949313170628006e-05, + "loss": 0.3143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3243550956249237, + "step": 1540, + "valid_targets_mean": 3106.2, + "valid_targets_min": 738 + }, + { + "epoch": 1.6228991596638656, + "grad_norm": 0.6224156356901303, + "learning_rate": 3.79261449490193e-05, + "loss": 0.2955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30039089918136597, + "step": 1545, + "valid_targets_mean": 3189.5, + "valid_targets_min": 1896 + }, + { + "epoch": 1.6281512605042017, + "grad_norm": 0.6278005675540494, + "learning_rate": 3.7902853740664356e-05, + "loss": 0.3128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3083696663379669, + "step": 1550, + "valid_targets_mean": 3278.9, + "valid_targets_min": 1177 + }, + { + "epoch": 1.6334033613445378, + "grad_norm": 0.5850412843704133, + "learning_rate": 3.7879439705358286e-05, + "loss": 0.3043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3127497732639313, + "step": 1555, + "valid_targets_mean": 3398.8, + "valid_targets_min": 1270 + }, + { + "epoch": 1.638655462184874, + "grad_norm": 0.5805366053879955, + "learning_rate": 3.785590300373884e-05, + "loss": 0.3063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29494237899780273, + "step": 1560, + "valid_targets_mean": 3829.2, + "valid_targets_min": 1668 + }, + { + "epoch": 1.64390756302521, + "grad_norm": 0.569934576020579, + "learning_rate": 3.7832243797285385e-05, + "loss": 0.3052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3094933032989502, + "step": 1565, + "valid_targets_mean": 3426.9, + "valid_targets_min": 1184 + }, + { + "epoch": 1.6491596638655461, + "grad_norm": 0.6159772681925453, + "learning_rate": 3.780846224831775e-05, + "loss": 0.3049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3124518394470215, + "step": 1570, + "valid_targets_mean": 3523.6, + "valid_targets_min": 1858 + }, + { + "epoch": 1.6544117647058822, + "grad_norm": 0.5971056987489094, + "learning_rate": 3.7784558519995135e-05, + "loss": 0.3075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28216129541397095, + "step": 1575, + "valid_targets_mean": 3161.2, + "valid_targets_min": 922 + }, + { + "epoch": 1.6596638655462184, + "grad_norm": 0.699651916524887, + "learning_rate": 3.776053277631496e-05, + "loss": 0.3079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.34362688660621643, + "step": 1580, + "valid_targets_mean": 2948.9, + "valid_targets_min": 1812 + }, + { + "epoch": 1.6649159663865545, + "grad_norm": 0.7182980448996313, + "learning_rate": 3.7736385182111785e-05, + "loss": 0.2979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2877700626850128, + "step": 1585, + "valid_targets_mean": 3107.7, + "valid_targets_min": 977 + }, + { + "epoch": 1.6701680672268906, + "grad_norm": 0.5675825049010571, + "learning_rate": 3.771211590305614e-05, + "loss": 0.2923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2754617929458618, + "step": 1590, + "valid_targets_mean": 3372.6, + "valid_targets_min": 810 + }, + { + "epoch": 1.675420168067227, + "grad_norm": 0.5802055167295502, + "learning_rate": 3.768772510565342e-05, + "loss": 0.2973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29426002502441406, + "step": 1595, + "valid_targets_mean": 3305.2, + "valid_targets_min": 1673 + }, + { + "epoch": 1.680672268907563, + "grad_norm": 0.5249464047322138, + "learning_rate": 3.76632129572427e-05, + "loss": 0.2977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2580610513687134, + "step": 1600, + "valid_targets_mean": 3933.0, + "valid_targets_min": 1869 + }, + { + "epoch": 1.6859243697478992, + "grad_norm": 1.426921636472864, + "learning_rate": 3.763857962599565e-05, + "loss": 0.3098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3049716353416443, + "step": 1605, + "valid_targets_mean": 3010.2, + "valid_targets_min": 1377 + }, + { + "epoch": 1.6911764705882353, + "grad_norm": 0.6454799620397031, + "learning_rate": 3.761382528091531e-05, + "loss": 0.3122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30134791135787964, + "step": 1610, + "valid_targets_mean": 3055.6, + "valid_targets_min": 1219 + }, + { + "epoch": 1.6964285714285714, + "grad_norm": 0.5717900298014066, + "learning_rate": 3.7588950091834986e-05, + "loss": 0.3049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2911268174648285, + "step": 1615, + "valid_targets_mean": 3442.4, + "valid_targets_min": 1168 + }, + { + "epoch": 1.7016806722689075, + "grad_norm": 0.645609709718482, + "learning_rate": 3.756395422941706e-05, + "loss": 0.2833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30699074268341064, + "step": 1620, + "valid_targets_mean": 3041.9, + "valid_targets_min": 1955 + }, + { + "epoch": 1.7069327731092439, + "grad_norm": 0.5866883370447175, + "learning_rate": 3.753883786515181e-05, + "loss": 0.2999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2968907654285431, + "step": 1625, + "valid_targets_mean": 3640.2, + "valid_targets_min": 1628 + }, + { + "epoch": 1.71218487394958, + "grad_norm": 0.5743154885711474, + "learning_rate": 3.751360117135628e-05, + "loss": 0.2881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2586438059806824, + "step": 1630, + "valid_targets_mean": 3358.4, + "valid_targets_min": 1482 + }, + { + "epoch": 1.717436974789916, + "grad_norm": 0.5881142433423517, + "learning_rate": 3.7488244321173025e-05, + "loss": 0.2883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26784512400627136, + "step": 1635, + "valid_targets_mean": 3460.4, + "valid_targets_min": 904 + }, + { + "epoch": 1.7226890756302522, + "grad_norm": 0.6168171706911441, + "learning_rate": 3.746276748856898e-05, + "loss": 0.2982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3478624224662781, + "step": 1640, + "valid_targets_mean": 3728.4, + "valid_targets_min": 1194 + }, + { + "epoch": 1.7279411764705883, + "grad_norm": 0.5755170388250846, + "learning_rate": 3.743717084833425e-05, + "loss": 0.3, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30867037177085876, + "step": 1645, + "valid_targets_mean": 3940.4, + "valid_targets_min": 1658 + }, + { + "epoch": 1.7331932773109244, + "grad_norm": 0.6224361276979423, + "learning_rate": 3.741145457608093e-05, + "loss": 0.2957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31917211413383484, + "step": 1650, + "valid_targets_mean": 3536.9, + "valid_targets_min": 1423 + }, + { + "epoch": 1.7384453781512605, + "grad_norm": 0.6144445149004428, + "learning_rate": 3.738561884824183e-05, + "loss": 0.2996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3049688935279846, + "step": 1655, + "valid_targets_mean": 3351.2, + "valid_targets_min": 2010 + }, + { + "epoch": 1.7436974789915967, + "grad_norm": 0.5954264479652966, + "learning_rate": 3.735966384206936e-05, + "loss": 0.3155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30082428455352783, + "step": 1660, + "valid_targets_mean": 3498.6, + "valid_targets_min": 1172 + }, + { + "epoch": 1.7489495798319328, + "grad_norm": 0.5715440496549611, + "learning_rate": 3.733358973563425e-05, + "loss": 0.2958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2998632788658142, + "step": 1665, + "valid_targets_mean": 3863.8, + "valid_targets_min": 1229 + }, + { + "epoch": 1.754201680672269, + "grad_norm": 0.6026903585406433, + "learning_rate": 3.730739670782435e-05, + "loss": 0.2896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2936575412750244, + "step": 1670, + "valid_targets_mean": 3600.5, + "valid_targets_min": 1523 + }, + { + "epoch": 1.759453781512605, + "grad_norm": 0.5960750461839239, + "learning_rate": 3.72810849383434e-05, + "loss": 0.2869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.286255419254303, + "step": 1675, + "valid_targets_mean": 3623.9, + "valid_targets_min": 997 + }, + { + "epoch": 1.7647058823529411, + "grad_norm": 0.5856872929601087, + "learning_rate": 3.725465460770978e-05, + "loss": 0.2922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29570281505584717, + "step": 1680, + "valid_targets_mean": 3535.2, + "valid_targets_min": 621 + }, + { + "epoch": 1.7699579831932772, + "grad_norm": 0.6652466131769478, + "learning_rate": 3.7228105897255324e-05, + "loss": 0.293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29390841722488403, + "step": 1685, + "valid_targets_mean": 3071.5, + "valid_targets_min": 1282 + }, + { + "epoch": 1.7752100840336134, + "grad_norm": 0.5607387594187455, + "learning_rate": 3.7201438989124e-05, + "loss": 0.2794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25984397530555725, + "step": 1690, + "valid_targets_mean": 3779.1, + "valid_targets_min": 1421 + }, + { + "epoch": 1.7804621848739495, + "grad_norm": 0.5923487985204593, + "learning_rate": 3.717465406627074e-05, + "loss": 0.288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27242133021354675, + "step": 1695, + "valid_targets_mean": 3330.2, + "valid_targets_min": 1853 + }, + { + "epoch": 1.7857142857142856, + "grad_norm": 0.5899947816496157, + "learning_rate": 3.714775131246011e-05, + "loss": 0.314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2966066002845764, + "step": 1700, + "valid_targets_mean": 3327.8, + "valid_targets_min": 1410 + }, + { + "epoch": 1.7909663865546217, + "grad_norm": 0.66060731387224, + "learning_rate": 3.71207309122651e-05, + "loss": 0.289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2911011576652527, + "step": 1705, + "valid_targets_mean": 2920.6, + "valid_targets_min": 1305 + }, + { + "epoch": 1.7962184873949578, + "grad_norm": 0.5843889526857856, + "learning_rate": 3.709359305106585e-05, + "loss": 0.2926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2868155837059021, + "step": 1710, + "valid_targets_mean": 3539.7, + "valid_targets_min": 996 + }, + { + "epoch": 1.8014705882352942, + "grad_norm": 0.586228171865445, + "learning_rate": 3.7066337915048354e-05, + "loss": 0.2895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27134719491004944, + "step": 1715, + "valid_targets_mean": 3357.1, + "valid_targets_min": 1349 + }, + { + "epoch": 1.8067226890756303, + "grad_norm": 0.606158452920606, + "learning_rate": 3.7038965691203205e-05, + "loss": 0.3019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31426700949668884, + "step": 1720, + "valid_targets_mean": 3613.5, + "valid_targets_min": 1248 + }, + { + "epoch": 1.8119747899159664, + "grad_norm": 0.5526990318942386, + "learning_rate": 3.701147656732431e-05, + "loss": 0.2812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24181315302848816, + "step": 1725, + "valid_targets_mean": 3327.4, + "valid_targets_min": 1109 + }, + { + "epoch": 1.8172268907563025, + "grad_norm": 0.6528173405719958, + "learning_rate": 3.6983870732007596e-05, + "loss": 0.2979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2989552617073059, + "step": 1730, + "valid_targets_mean": 2790.8, + "valid_targets_min": 1649 + }, + { + "epoch": 1.8224789915966386, + "grad_norm": 0.6835745034650174, + "learning_rate": 3.695614837464972e-05, + "loss": 0.2895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3165176808834076, + "step": 1735, + "valid_targets_mean": 2811.8, + "valid_targets_min": 1197 + }, + { + "epoch": 1.8277310924369747, + "grad_norm": 0.6755172073863958, + "learning_rate": 3.692830968544675e-05, + "loss": 0.306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3160332441329956, + "step": 1740, + "valid_targets_mean": 3420.2, + "valid_targets_min": 1720 + }, + { + "epoch": 1.832983193277311, + "grad_norm": 0.6087751854436854, + "learning_rate": 3.690035485539291e-05, + "loss": 0.2993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29514777660369873, + "step": 1745, + "valid_targets_mean": 3542.7, + "valid_targets_min": 1054 + }, + { + "epoch": 1.8382352941176472, + "grad_norm": 0.6241785789650026, + "learning_rate": 3.6872284076279205e-05, + "loss": 0.2966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2934738099575043, + "step": 1750, + "valid_targets_mean": 3049.7, + "valid_targets_min": 1643 + }, + { + "epoch": 1.8434873949579833, + "grad_norm": 0.5825825660256299, + "learning_rate": 3.684409754069215e-05, + "loss": 0.3039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30002132058143616, + "step": 1755, + "valid_targets_mean": 3757.8, + "valid_targets_min": 1135 + }, + { + "epoch": 1.8487394957983194, + "grad_norm": 0.6258913745185652, + "learning_rate": 3.681579544201244e-05, + "loss": 0.2952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.290172278881073, + "step": 1760, + "valid_targets_mean": 3715.1, + "valid_targets_min": 1174 + }, + { + "epoch": 1.8539915966386555, + "grad_norm": 0.6554971879023268, + "learning_rate": 3.6787377974413614e-05, + "loss": 0.3127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32472649216651917, + "step": 1765, + "valid_targets_mean": 3321.9, + "valid_targets_min": 992 + }, + { + "epoch": 1.8592436974789917, + "grad_norm": 0.6104106962565312, + "learning_rate": 3.6758845332860734e-05, + "loss": 0.3036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.320326030254364, + "step": 1770, + "valid_targets_mean": 3713.2, + "valid_targets_min": 1133 + }, + { + "epoch": 1.8644957983193278, + "grad_norm": 0.6602404199908332, + "learning_rate": 3.673019771310903e-05, + "loss": 0.3009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3192262649536133, + "step": 1775, + "valid_targets_mean": 2961.8, + "valid_targets_min": 1195 + }, + { + "epoch": 1.8697478991596639, + "grad_norm": 0.9226641425449463, + "learning_rate": 3.670143531170258e-05, + "loss": 0.3044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30954253673553467, + "step": 1780, + "valid_targets_mean": 3012.6, + "valid_targets_min": 484 + }, + { + "epoch": 1.875, + "grad_norm": 0.5947567961611435, + "learning_rate": 3.667255832597294e-05, + "loss": 0.2982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3080810010433197, + "step": 1785, + "valid_targets_mean": 3639.9, + "valid_targets_min": 1434 + }, + { + "epoch": 1.8802521008403361, + "grad_norm": 0.5635109826748741, + "learning_rate": 3.664356695403781e-05, + "loss": 0.2972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2792786955833435, + "step": 1790, + "valid_targets_mean": 4017.4, + "valid_targets_min": 1349 + }, + { + "epoch": 1.8855042016806722, + "grad_norm": 0.7060376087001079, + "learning_rate": 3.661446139479965e-05, + "loss": 0.2965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31975793838500977, + "step": 1795, + "valid_targets_mean": 3178.8, + "valid_targets_min": 1434 + }, + { + "epoch": 1.8907563025210083, + "grad_norm": 0.5936744481364934, + "learning_rate": 3.658524184794436e-05, + "loss": 0.2927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2860378324985504, + "step": 1800, + "valid_targets_mean": 3633.9, + "valid_targets_min": 521 + }, + { + "epoch": 1.8960084033613445, + "grad_norm": 0.5822754669183862, + "learning_rate": 3.655590851393984e-05, + "loss": 0.3049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3000732660293579, + "step": 1805, + "valid_targets_mean": 3981.0, + "valid_targets_min": 1464 + }, + { + "epoch": 1.9012605042016806, + "grad_norm": 0.6913451388646343, + "learning_rate": 3.652646159403468e-05, + "loss": 0.3136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3260645866394043, + "step": 1810, + "valid_targets_mean": 2612.0, + "valid_targets_min": 1239 + }, + { + "epoch": 1.9065126050420167, + "grad_norm": 0.6287538437173945, + "learning_rate": 3.649690129025675e-05, + "loss": 0.2841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29579633474349976, + "step": 1815, + "valid_targets_mean": 3489.8, + "valid_targets_min": 1457 + }, + { + "epoch": 1.9117647058823528, + "grad_norm": 0.6767664513173561, + "learning_rate": 3.6467227805411824e-05, + "loss": 0.2916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29462265968322754, + "step": 1820, + "valid_targets_mean": 3059.1, + "valid_targets_min": 1460 + }, + { + "epoch": 1.917016806722689, + "grad_norm": 0.6395995499615521, + "learning_rate": 3.6437441343082174e-05, + "loss": 0.2944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3023369312286377, + "step": 1825, + "valid_targets_mean": 3259.2, + "valid_targets_min": 1084 + }, + { + "epoch": 1.9222689075630253, + "grad_norm": 0.5856128524517397, + "learning_rate": 3.6407542107625167e-05, + "loss": 0.2836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27946245670318604, + "step": 1830, + "valid_targets_mean": 3337.9, + "valid_targets_min": 1035 + }, + { + "epoch": 1.9275210084033614, + "grad_norm": 0.5516337340536542, + "learning_rate": 3.637753030417192e-05, + "loss": 0.282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2830444574356079, + "step": 1835, + "valid_targets_mean": 4091.9, + "valid_targets_min": 1876 + }, + { + "epoch": 1.9327731092436975, + "grad_norm": 0.6188213113954284, + "learning_rate": 3.6347406138625805e-05, + "loss": 0.2869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3005865812301636, + "step": 1840, + "valid_targets_mean": 3705.8, + "valid_targets_min": 1062 + }, + { + "epoch": 1.9380252100840336, + "grad_norm": 0.6342838492371049, + "learning_rate": 3.631716981766111e-05, + "loss": 0.2778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2753984034061432, + "step": 1845, + "valid_targets_mean": 3030.8, + "valid_targets_min": 993 + }, + { + "epoch": 1.9432773109243697, + "grad_norm": 0.5880238709579537, + "learning_rate": 3.6286821548721594e-05, + "loss": 0.2822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28602135181427, + "step": 1850, + "valid_targets_mean": 3366.9, + "valid_targets_min": 873 + }, + { + "epoch": 1.9485294117647058, + "grad_norm": 0.5720082036013604, + "learning_rate": 3.625636154001904e-05, + "loss": 0.2821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2676365375518799, + "step": 1855, + "valid_targets_mean": 3651.0, + "valid_targets_min": 1295 + }, + { + "epoch": 1.9537815126050422, + "grad_norm": 0.585699773613234, + "learning_rate": 3.622579000053186e-05, + "loss": 0.2902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29407864809036255, + "step": 1860, + "valid_targets_mean": 3583.6, + "valid_targets_min": 1598 + }, + { + "epoch": 1.9590336134453783, + "grad_norm": 0.6205691001395918, + "learning_rate": 3.619510714000368e-05, + "loss": 0.3088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27674800157546997, + "step": 1865, + "valid_targets_mean": 2903.2, + "valid_targets_min": 1123 + }, + { + "epoch": 1.9642857142857144, + "grad_norm": 0.7438366690160714, + "learning_rate": 3.616431316894181e-05, + "loss": 0.2938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31191760301589966, + "step": 1870, + "valid_targets_mean": 3116.1, + "valid_targets_min": 1621 + }, + { + "epoch": 1.9695378151260505, + "grad_norm": 0.6496526680316776, + "learning_rate": 3.613340829861592e-05, + "loss": 0.2777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2977924346923828, + "step": 1875, + "valid_targets_mean": 3081.4, + "valid_targets_min": 1858 + }, + { + "epoch": 1.9747899159663866, + "grad_norm": 0.5707764114436218, + "learning_rate": 3.61023927410565e-05, + "loss": 0.2855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25262773036956787, + "step": 1880, + "valid_targets_mean": 3470.7, + "valid_targets_min": 420 + }, + { + "epoch": 1.9800420168067228, + "grad_norm": 0.601493220440057, + "learning_rate": 3.607126670905346e-05, + "loss": 0.2966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2920111417770386, + "step": 1885, + "valid_targets_mean": 3584.2, + "valid_targets_min": 757 + }, + { + "epoch": 1.9852941176470589, + "grad_norm": 0.6874129379616689, + "learning_rate": 3.60400304161546e-05, + "loss": 0.3002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2971467673778534, + "step": 1890, + "valid_targets_mean": 2940.8, + "valid_targets_min": 732 + }, + { + "epoch": 1.990546218487395, + "grad_norm": 0.6483438175679863, + "learning_rate": 3.600868407666426e-05, + "loss": 0.2939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29474449157714844, + "step": 1895, + "valid_targets_mean": 3248.2, + "valid_targets_min": 929 + }, + { + "epoch": 1.995798319327731, + "grad_norm": 0.5918173246249905, + "learning_rate": 3.597722790564172e-05, + "loss": 0.2929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2911064624786377, + "step": 1900, + "valid_targets_mean": 3728.8, + "valid_targets_min": 1217 + }, + { + "epoch": 2.0010504201680672, + "grad_norm": 1.269479871339695, + "learning_rate": 3.5945662118899824e-05, + "loss": 0.2774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21561074256896973, + "step": 1905, + "valid_targets_mean": 1523.7, + "valid_targets_min": 544 + }, + { + "epoch": 2.0063025210084033, + "grad_norm": 0.834502760942854, + "learning_rate": 3.5913986933003455e-05, + "loss": 0.2186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1978900134563446, + "step": 1910, + "valid_targets_mean": 1311.9, + "valid_targets_min": 632 + }, + { + "epoch": 2.0115546218487395, + "grad_norm": 0.8472702704612042, + "learning_rate": 3.5882202565268046e-05, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19883087277412415, + "step": 1915, + "valid_targets_mean": 1391.8, + "valid_targets_min": 514 + }, + { + "epoch": 2.0168067226890756, + "grad_norm": 0.8558019853903324, + "learning_rate": 3.58503092337581e-05, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2164035588502884, + "step": 1920, + "valid_targets_mean": 1409.8, + "valid_targets_min": 701 + }, + { + "epoch": 2.0220588235294117, + "grad_norm": 0.755196603700863, + "learning_rate": 3.5818307157285705e-05, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20270323753356934, + "step": 1925, + "valid_targets_mean": 1448.8, + "valid_targets_min": 742 + }, + { + "epoch": 2.027310924369748, + "grad_norm": 0.7296703034732365, + "learning_rate": 3.5786196555409006e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18208900094032288, + "step": 1930, + "valid_targets_mean": 1219.5, + "valid_targets_min": 746 + }, + { + "epoch": 2.032563025210084, + "grad_norm": 0.7684329269994562, + "learning_rate": 3.575397764843073e-05, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19005586206912994, + "step": 1935, + "valid_targets_mean": 1416.7, + "valid_targets_min": 774 + }, + { + "epoch": 2.03781512605042, + "grad_norm": 0.7346120823737189, + "learning_rate": 3.572165065739664e-05, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.173833429813385, + "step": 1940, + "valid_targets_mean": 1294.7, + "valid_targets_min": 648 + }, + { + "epoch": 2.043067226890756, + "grad_norm": 0.7872472114882302, + "learning_rate": 3.5689215804094053e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19836294651031494, + "step": 1945, + "valid_targets_mean": 1322.9, + "valid_targets_min": 708 + }, + { + "epoch": 2.0483193277310923, + "grad_norm": 0.7139156271445127, + "learning_rate": 3.5656673311050285e-05, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.166924387216568, + "step": 1950, + "valid_targets_mean": 1500.9, + "valid_targets_min": 788 + }, + { + "epoch": 2.0535714285714284, + "grad_norm": 0.7546720109875962, + "learning_rate": 3.562402340153115e-05, + "loss": 0.1969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1813475340604782, + "step": 1955, + "valid_targets_mean": 1277.2, + "valid_targets_min": 615 + }, + { + "epoch": 2.0588235294117645, + "grad_norm": 0.8007930214332662, + "learning_rate": 3.559126629953942e-05, + "loss": 0.1966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19298464059829712, + "step": 1960, + "valid_targets_mean": 1355.1, + "valid_targets_min": 627 + }, + { + "epoch": 2.064075630252101, + "grad_norm": 1.0715850452767404, + "learning_rate": 3.5558402229813287e-05, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19311118125915527, + "step": 1965, + "valid_targets_mean": 1323.2, + "valid_targets_min": 753 + }, + { + "epoch": 2.069327731092437, + "grad_norm": 0.7534257104768941, + "learning_rate": 3.552543141782483e-05, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19203367829322815, + "step": 1970, + "valid_targets_mean": 1427.2, + "valid_targets_min": 740 + }, + { + "epoch": 2.0745798319327733, + "grad_norm": 0.7434395907521474, + "learning_rate": 3.549235408977843e-05, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19772931933403015, + "step": 1975, + "valid_targets_mean": 1519.6, + "valid_targets_min": 642 + }, + { + "epoch": 2.0798319327731094, + "grad_norm": 0.7265121475274179, + "learning_rate": 3.545917047260928e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16659091413021088, + "step": 1980, + "valid_targets_mean": 1209.2, + "valid_targets_min": 518 + }, + { + "epoch": 2.0850840336134455, + "grad_norm": 0.7590657870156368, + "learning_rate": 3.542588079398179e-05, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20139284431934357, + "step": 1985, + "valid_targets_mean": 1439.6, + "valid_targets_min": 721 + }, + { + "epoch": 2.0903361344537816, + "grad_norm": 0.7606478486879998, + "learning_rate": 3.539248528228801e-05, + "loss": 0.1994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1940763294696808, + "step": 1990, + "valid_targets_mean": 1469.9, + "valid_targets_min": 727 + }, + { + "epoch": 2.0955882352941178, + "grad_norm": 0.7061929545295577, + "learning_rate": 3.535898416664611e-05, + "loss": 0.2012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19255903363227844, + "step": 1995, + "valid_targets_mean": 1412.1, + "valid_targets_min": 803 + }, + { + "epoch": 2.100840336134454, + "grad_norm": 0.7887719103581653, + "learning_rate": 3.5325377676898766e-05, + "loss": 0.1958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2186465561389923, + "step": 2000, + "valid_targets_mean": 1546.6, + "valid_targets_min": 698 + }, + { + "epoch": 2.10609243697479, + "grad_norm": 0.8070718344561886, + "learning_rate": 3.529166604361161e-05, + "loss": 0.1912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2157689332962036, + "step": 2005, + "valid_targets_mean": 1728.5, + "valid_targets_min": 832 + }, + { + "epoch": 2.111344537815126, + "grad_norm": 0.6799684604036553, + "learning_rate": 3.525784949807162e-05, + "loss": 0.1942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18615804612636566, + "step": 2010, + "valid_targets_mean": 1575.1, + "valid_targets_min": 891 + }, + { + "epoch": 2.116596638655462, + "grad_norm": 0.7859292131535734, + "learning_rate": 3.522392827228557e-05, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20224624872207642, + "step": 2015, + "valid_targets_mean": 1346.5, + "valid_targets_min": 621 + }, + { + "epoch": 2.1218487394957983, + "grad_norm": 0.7769917237320048, + "learning_rate": 3.518990259897841e-05, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16991014778614044, + "step": 2020, + "valid_targets_mean": 1285.5, + "valid_targets_min": 697 + }, + { + "epoch": 2.1271008403361344, + "grad_norm": 0.7202399948858199, + "learning_rate": 3.515577271159167e-05, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1941305696964264, + "step": 2025, + "valid_targets_mean": 1565.5, + "valid_targets_min": 824 + }, + { + "epoch": 2.1323529411764706, + "grad_norm": 0.7845226485201681, + "learning_rate": 3.512153884428188e-05, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17622260749340057, + "step": 2030, + "valid_targets_mean": 1331.2, + "valid_targets_min": 584 + }, + { + "epoch": 2.1376050420168067, + "grad_norm": 0.7398989600657419, + "learning_rate": 3.5087201231918944e-05, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1894277185201645, + "step": 2035, + "valid_targets_mean": 1568.8, + "valid_targets_min": 1033 + }, + { + "epoch": 2.142857142857143, + "grad_norm": 0.7770180042406308, + "learning_rate": 3.505276011008454e-05, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2185641974210739, + "step": 2040, + "valid_targets_mean": 1584.9, + "valid_targets_min": 1007 + }, + { + "epoch": 2.148109243697479, + "grad_norm": 0.78686344769378, + "learning_rate": 3.501821571507048e-05, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1753498911857605, + "step": 2045, + "valid_targets_mean": 1267.1, + "valid_targets_min": 613 + }, + { + "epoch": 2.153361344537815, + "grad_norm": 1.063699580344548, + "learning_rate": 3.498356828387714e-05, + "loss": 0.189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1909262090921402, + "step": 2050, + "valid_targets_mean": 1532.2, + "valid_targets_min": 768 + }, + { + "epoch": 2.158613445378151, + "grad_norm": 0.7643365744929539, + "learning_rate": 3.494881805421176e-05, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16610810160636902, + "step": 2055, + "valid_targets_mean": 1112.7, + "valid_targets_min": 643 + }, + { + "epoch": 2.1638655462184873, + "grad_norm": 0.7926765771940254, + "learning_rate": 3.4913965264486894e-05, + "loss": 0.1948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18852682411670685, + "step": 2060, + "valid_targets_mean": 1402.3, + "valid_targets_min": 841 + }, + { + "epoch": 2.1691176470588234, + "grad_norm": 0.729414923035659, + "learning_rate": 3.48790101538187e-05, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18665212392807007, + "step": 2065, + "valid_targets_mean": 1577.6, + "valid_targets_min": 705 + }, + { + "epoch": 2.1743697478991595, + "grad_norm": 0.8405323016094074, + "learning_rate": 3.484395296202537e-05, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1933499127626419, + "step": 2070, + "valid_targets_mean": 1190.7, + "valid_targets_min": 716 + }, + { + "epoch": 2.1796218487394956, + "grad_norm": 0.7875851862101669, + "learning_rate": 3.480879392962541e-05, + "loss": 0.1741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17440925538539886, + "step": 2075, + "valid_targets_mean": 1209.2, + "valid_targets_min": 627 + }, + { + "epoch": 2.184873949579832, + "grad_norm": 0.7185518361219825, + "learning_rate": 3.477353329783606e-05, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16545218229293823, + "step": 2080, + "valid_targets_mean": 1253.7, + "valid_targets_min": 619 + }, + { + "epoch": 2.190126050420168, + "grad_norm": 0.7600617986037965, + "learning_rate": 3.4738171308571595e-05, + "loss": 0.1863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19123047590255737, + "step": 2085, + "valid_targets_mean": 1403.3, + "valid_targets_min": 734 + }, + { + "epoch": 2.1953781512605044, + "grad_norm": 0.7730815354374817, + "learning_rate": 3.470270820444168e-05, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18116678297519684, + "step": 2090, + "valid_targets_mean": 1356.3, + "valid_targets_min": 700 + }, + { + "epoch": 2.2006302521008405, + "grad_norm": 0.7542365322548513, + "learning_rate": 3.4667144228749697e-05, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18537303805351257, + "step": 2095, + "valid_targets_mean": 1550.9, + "valid_targets_min": 892 + }, + { + "epoch": 2.2058823529411766, + "grad_norm": 0.7438888508924026, + "learning_rate": 3.463147962549109e-05, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16940736770629883, + "step": 2100, + "valid_targets_mean": 1295.9, + "valid_targets_min": 711 + }, + { + "epoch": 2.2111344537815127, + "grad_norm": 0.7547714805297574, + "learning_rate": 3.459571463935168e-05, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16632413864135742, + "step": 2105, + "valid_targets_mean": 1380.6, + "valid_targets_min": 694 + }, + { + "epoch": 2.216386554621849, + "grad_norm": 0.7974469156906152, + "learning_rate": 3.455984951570598e-05, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20498281717300415, + "step": 2110, + "valid_targets_mean": 1487.6, + "valid_targets_min": 818 + }, + { + "epoch": 2.221638655462185, + "grad_norm": 0.7460498917875708, + "learning_rate": 3.452388450061556e-05, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1839742511510849, + "step": 2115, + "valid_targets_mean": 1640.1, + "valid_targets_min": 1011 + }, + { + "epoch": 2.226890756302521, + "grad_norm": 0.7976082882204165, + "learning_rate": 3.448781984082727e-05, + "loss": 0.1967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1880989819765091, + "step": 2120, + "valid_targets_mean": 1356.1, + "valid_targets_min": 748 + }, + { + "epoch": 2.232142857142857, + "grad_norm": 0.8198726303330894, + "learning_rate": 3.4451655783771625e-05, + "loss": 0.1822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19785097241401672, + "step": 2125, + "valid_targets_mean": 1494.3, + "valid_targets_min": 699 + }, + { + "epoch": 2.2373949579831933, + "grad_norm": 0.7959727926112404, + "learning_rate": 3.4415392577561064e-05, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17913001775741577, + "step": 2130, + "valid_targets_mean": 1201.1, + "valid_targets_min": 534 + }, + { + "epoch": 2.2426470588235294, + "grad_norm": 0.7163813248237905, + "learning_rate": 3.4379030470988285e-05, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1842835247516632, + "step": 2135, + "valid_targets_mean": 1393.6, + "valid_targets_min": 687 + }, + { + "epoch": 2.2478991596638656, + "grad_norm": 0.7075959634728527, + "learning_rate": 3.4342569713524496e-05, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18208453059196472, + "step": 2140, + "valid_targets_mean": 1363.2, + "valid_targets_min": 549 + }, + { + "epoch": 2.2531512605042017, + "grad_norm": 0.7211274630211372, + "learning_rate": 3.4306010555317735e-05, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18160000443458557, + "step": 2145, + "valid_targets_mean": 1349.1, + "valid_targets_min": 725 + }, + { + "epoch": 2.258403361344538, + "grad_norm": 0.7402311941740003, + "learning_rate": 3.4269353247191135e-05, + "loss": 0.1802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1999184489250183, + "step": 2150, + "valid_targets_mean": 1651.7, + "valid_targets_min": 852 + }, + { + "epoch": 2.263655462184874, + "grad_norm": 0.6795776738854672, + "learning_rate": 3.423259804064122e-05, + "loss": 0.1914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18849042057991028, + "step": 2155, + "valid_targets_mean": 1559.5, + "valid_targets_min": 1010 + }, + { + "epoch": 2.26890756302521, + "grad_norm": 0.7308651601236005, + "learning_rate": 3.419574518783616e-05, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17530426383018494, + "step": 2160, + "valid_targets_mean": 1387.0, + "valid_targets_min": 545 + }, + { + "epoch": 2.274159663865546, + "grad_norm": 0.8144354216387479, + "learning_rate": 3.4158794941614055e-05, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19270774722099304, + "step": 2165, + "valid_targets_mean": 1385.4, + "valid_targets_min": 843 + }, + { + "epoch": 2.2794117647058822, + "grad_norm": 0.631924641323973, + "learning_rate": 3.4121747555481216e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19547529518604279, + "step": 2170, + "valid_targets_mean": 1716.2, + "valid_targets_min": 802 + }, + { + "epoch": 2.2846638655462184, + "grad_norm": 0.7544602510753329, + "learning_rate": 3.4084603283610365e-05, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18898774683475494, + "step": 2175, + "valid_targets_mean": 1409.9, + "valid_targets_min": 612 + }, + { + "epoch": 2.2899159663865545, + "grad_norm": 0.7250456296463277, + "learning_rate": 3.404736238083897e-05, + "loss": 0.1814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17501644790172577, + "step": 2180, + "valid_targets_mean": 1415.8, + "valid_targets_min": 673 + }, + { + "epoch": 2.2951680672268906, + "grad_norm": 0.8001168867925627, + "learning_rate": 3.401002510266744e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18818825483322144, + "step": 2185, + "valid_targets_mean": 1332.9, + "valid_targets_min": 705 + }, + { + "epoch": 2.3004201680672267, + "grad_norm": 0.6987572245775185, + "learning_rate": 3.397259170525739e-05, + "loss": 0.1747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16560634970664978, + "step": 2190, + "valid_targets_mean": 1402.1, + "valid_targets_min": 822 + }, + { + "epoch": 2.3056722689075633, + "grad_norm": 0.7531921047773884, + "learning_rate": 3.393506244542991e-05, + "loss": 0.1702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16486376523971558, + "step": 2195, + "valid_targets_mean": 1272.8, + "valid_targets_min": 604 + }, + { + "epoch": 2.310924369747899, + "grad_norm": 0.6999344056695282, + "learning_rate": 3.389743758066374e-05, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17392639815807343, + "step": 2200, + "valid_targets_mean": 1460.0, + "valid_targets_min": 647 + }, + { + "epoch": 2.3161764705882355, + "grad_norm": 0.7036250917228805, + "learning_rate": 3.385971736909357e-05, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1617729663848877, + "step": 2205, + "valid_targets_mean": 1283.6, + "valid_targets_min": 613 + }, + { + "epoch": 2.3214285714285716, + "grad_norm": 0.7068382427421205, + "learning_rate": 3.382190206950822e-05, + "loss": 0.1801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17291636765003204, + "step": 2210, + "valid_targets_mean": 1288.4, + "valid_targets_min": 713 + }, + { + "epoch": 2.3266806722689077, + "grad_norm": 0.663523616107997, + "learning_rate": 3.3783991941348916e-05, + "loss": 0.1823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16455985605716705, + "step": 2215, + "valid_targets_mean": 1533.8, + "valid_targets_min": 892 + }, + { + "epoch": 2.331932773109244, + "grad_norm": 0.7721343986047551, + "learning_rate": 3.374598724470745e-05, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19193603098392487, + "step": 2220, + "valid_targets_mean": 1345.8, + "valid_targets_min": 772 + }, + { + "epoch": 2.33718487394958, + "grad_norm": 0.7240487637559474, + "learning_rate": 3.370788824032441e-05, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1901538372039795, + "step": 2225, + "valid_targets_mean": 1601.2, + "valid_targets_min": 636 + }, + { + "epoch": 2.342436974789916, + "grad_norm": 0.7718537058087979, + "learning_rate": 3.366969518958746e-05, + "loss": 0.1781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1778993457555771, + "step": 2230, + "valid_targets_mean": 1366.1, + "valid_targets_min": 641 + }, + { + "epoch": 2.347689075630252, + "grad_norm": 0.7977490601269798, + "learning_rate": 3.363140835452945e-05, + "loss": 0.1874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20514068007469177, + "step": 2235, + "valid_targets_mean": 1482.2, + "valid_targets_min": 818 + }, + { + "epoch": 2.3529411764705883, + "grad_norm": 0.7668591077227411, + "learning_rate": 3.3593027997826654e-05, + "loss": 0.1832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1879502534866333, + "step": 2240, + "valid_targets_mean": 1420.5, + "valid_targets_min": 729 + }, + { + "epoch": 2.3581932773109244, + "grad_norm": 0.7855984838086749, + "learning_rate": 3.355455438279702e-05, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21655318140983582, + "step": 2245, + "valid_targets_mean": 1723.6, + "valid_targets_min": 1001 + }, + { + "epoch": 2.3634453781512605, + "grad_norm": 0.7364382070700145, + "learning_rate": 3.351598777339827e-05, + "loss": 0.1837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1900310218334198, + "step": 2250, + "valid_targets_mean": 1626.4, + "valid_targets_min": 1175 + }, + { + "epoch": 2.3686974789915967, + "grad_norm": 0.7437939379036828, + "learning_rate": 3.3477328434226164e-05, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19543638825416565, + "step": 2255, + "valid_targets_mean": 1434.9, + "valid_targets_min": 779 + }, + { + "epoch": 2.3739495798319328, + "grad_norm": 0.7246181422661423, + "learning_rate": 3.343857663051264e-05, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1804441660642624, + "step": 2260, + "valid_targets_mean": 1413.4, + "valid_targets_min": 730 + }, + { + "epoch": 2.379201680672269, + "grad_norm": 0.73443894447988, + "learning_rate": 3.339973262812403e-05, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17590978741645813, + "step": 2265, + "valid_targets_mean": 1360.3, + "valid_targets_min": 564 + }, + { + "epoch": 2.384453781512605, + "grad_norm": 0.8777730335212939, + "learning_rate": 3.336079669355921e-05, + "loss": 0.1786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18321183323860168, + "step": 2270, + "valid_targets_mean": 1550.0, + "valid_targets_min": 651 + }, + { + "epoch": 2.389705882352941, + "grad_norm": 0.7244538791290572, + "learning_rate": 3.332176909394777e-05, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1737801432609558, + "step": 2275, + "valid_targets_mean": 1342.3, + "valid_targets_min": 716 + }, + { + "epoch": 2.3949579831932772, + "grad_norm": 0.7245107985964735, + "learning_rate": 3.3282650097048206e-05, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1800583004951477, + "step": 2280, + "valid_targets_mean": 1339.6, + "valid_targets_min": 707 + }, + { + "epoch": 2.4002100840336134, + "grad_norm": 0.7235945312664076, + "learning_rate": 3.324343997124606e-05, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15614992380142212, + "step": 2285, + "valid_targets_mean": 1459.4, + "valid_targets_min": 964 + }, + { + "epoch": 2.4054621848739495, + "grad_norm": 0.7567614980512993, + "learning_rate": 3.3204138985552094e-05, + "loss": 0.1841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20014092326164246, + "step": 2290, + "valid_targets_mean": 1562.7, + "valid_targets_min": 793 + }, + { + "epoch": 2.4107142857142856, + "grad_norm": 0.7418561305374557, + "learning_rate": 3.316474740960041e-05, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17074398696422577, + "step": 2295, + "valid_targets_mean": 1367.2, + "valid_targets_min": 727 + }, + { + "epoch": 2.4159663865546217, + "grad_norm": 0.7089430053862654, + "learning_rate": 3.312526551364667e-05, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17129768431186676, + "step": 2300, + "valid_targets_mean": 1451.4, + "valid_targets_min": 734 + }, + { + "epoch": 2.421218487394958, + "grad_norm": 0.5820555364713835, + "learning_rate": 3.308569356856616e-05, + "loss": 0.1706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1320435255765915, + "step": 2305, + "valid_targets_mean": 2171.5, + "valid_targets_min": 978 + }, + { + "epoch": 2.426470588235294, + "grad_norm": 0.6553889260546893, + "learning_rate": 3.3046031845851993e-05, + "loss": 0.1644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16868937015533447, + "step": 2310, + "valid_targets_mean": 1519.2, + "valid_targets_min": 595 + }, + { + "epoch": 2.43172268907563, + "grad_norm": 0.7087358397507307, + "learning_rate": 3.3006280617613214e-05, + "loss": 0.1747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17215979099273682, + "step": 2315, + "valid_targets_mean": 1396.3, + "valid_targets_min": 570 + }, + { + "epoch": 2.4369747899159666, + "grad_norm": 0.7369066907552546, + "learning_rate": 3.296644015657295e-05, + "loss": 0.1754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18438240885734558, + "step": 2320, + "valid_targets_mean": 1494.4, + "valid_targets_min": 914 + }, + { + "epoch": 2.4422268907563023, + "grad_norm": 0.69010875765003, + "learning_rate": 3.292651073606652e-05, + "loss": 0.1719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16703712940216064, + "step": 2325, + "valid_targets_mean": 1541.1, + "valid_targets_min": 1070 + }, + { + "epoch": 2.447478991596639, + "grad_norm": 0.7340635719059896, + "learning_rate": 3.288649263003958e-05, + "loss": 0.1748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18820658326148987, + "step": 2330, + "valid_targets_mean": 1509.0, + "valid_targets_min": 626 + }, + { + "epoch": 2.452731092436975, + "grad_norm": 0.7197216030974956, + "learning_rate": 3.284638611304623e-05, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17071965336799622, + "step": 2335, + "valid_targets_mean": 1484.4, + "valid_targets_min": 718 + }, + { + "epoch": 2.457983193277311, + "grad_norm": 0.7517975842539306, + "learning_rate": 3.280619146024714e-05, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19978466629981995, + "step": 2340, + "valid_targets_mean": 1505.0, + "valid_targets_min": 762 + }, + { + "epoch": 2.463235294117647, + "grad_norm": 0.7266865301196809, + "learning_rate": 3.2765908947407665e-05, + "loss": 0.2243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2888498306274414, + "step": 2345, + "valid_targets_mean": 3602.9, + "valid_targets_min": 1698 + }, + { + "epoch": 2.4684873949579833, + "grad_norm": 0.5649119554586547, + "learning_rate": 3.2725538850895933e-05, + "loss": 0.2777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25794461369514465, + "step": 2350, + "valid_targets_mean": 3718.5, + "valid_targets_min": 533 + }, + { + "epoch": 2.4737394957983194, + "grad_norm": 0.7045726718503901, + "learning_rate": 3.268508144768096e-05, + "loss": 0.2725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2772102653980255, + "step": 2355, + "valid_targets_mean": 3299.2, + "valid_targets_min": 835 + }, + { + "epoch": 2.4789915966386555, + "grad_norm": 0.564312940590048, + "learning_rate": 3.264453701533075e-05, + "loss": 0.2685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2812797725200653, + "step": 2360, + "valid_targets_mean": 4040.6, + "valid_targets_min": 1201 + }, + { + "epoch": 2.4842436974789917, + "grad_norm": 0.6103845308794094, + "learning_rate": 3.260390583201041e-05, + "loss": 0.2546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26815706491470337, + "step": 2365, + "valid_targets_mean": 3995.7, + "valid_targets_min": 980 + }, + { + "epoch": 2.4894957983193278, + "grad_norm": 0.5643937322946125, + "learning_rate": 3.2563188176480217e-05, + "loss": 0.2817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31056275963783264, + "step": 2370, + "valid_targets_mean": 4070.5, + "valid_targets_min": 1744 + }, + { + "epoch": 2.494747899159664, + "grad_norm": 0.5232363392589175, + "learning_rate": 3.252238432809368e-05, + "loss": 0.2435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22355665266513824, + "step": 2375, + "valid_targets_mean": 3831.8, + "valid_targets_min": 1734 + }, + { + "epoch": 2.5, + "grad_norm": 0.666569390983386, + "learning_rate": 3.2481494566795716e-05, + "loss": 0.2587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2557009756565094, + "step": 2380, + "valid_targets_mean": 4593.9, + "valid_targets_min": 1381 + }, + { + "epoch": 2.505252100840336, + "grad_norm": 0.5634090394701392, + "learning_rate": 3.244051917312061e-05, + "loss": 0.2671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2774540185928345, + "step": 2385, + "valid_targets_mean": 3523.4, + "valid_targets_min": 1024 + }, + { + "epoch": 2.5105042016806722, + "grad_norm": 0.5827012399453538, + "learning_rate": 3.239945842819018e-05, + "loss": 0.2779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27029716968536377, + "step": 2390, + "valid_targets_mean": 3686.2, + "valid_targets_min": 2107 + }, + { + "epoch": 2.5157563025210083, + "grad_norm": 0.6119604303971371, + "learning_rate": 3.2358312613711804e-05, + "loss": 0.2674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2673264741897583, + "step": 2395, + "valid_targets_mean": 3195.1, + "valid_targets_min": 802 + }, + { + "epoch": 2.5210084033613445, + "grad_norm": 0.5309909333341549, + "learning_rate": 3.231708201197653e-05, + "loss": 0.2697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27537453174591064, + "step": 2400, + "valid_targets_mean": 4344.2, + "valid_targets_min": 1597 + }, + { + "epoch": 2.5262605042016806, + "grad_norm": 0.6196309832297628, + "learning_rate": 3.227576690585706e-05, + "loss": 0.2717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27964839339256287, + "step": 2405, + "valid_targets_mean": 3039.4, + "valid_targets_min": 1535 + }, + { + "epoch": 2.5315126050420167, + "grad_norm": 0.6105328048520959, + "learning_rate": 3.223436757880591e-05, + "loss": 0.2655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2900930643081665, + "step": 2410, + "valid_targets_mean": 3732.8, + "valid_targets_min": 1385 + }, + { + "epoch": 2.536764705882353, + "grad_norm": 0.6121657286750687, + "learning_rate": 3.219288431485337e-05, + "loss": 0.271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.296478807926178, + "step": 2415, + "valid_targets_mean": 3802.3, + "valid_targets_min": 1451 + }, + { + "epoch": 2.542016806722689, + "grad_norm": 0.6645007327168074, + "learning_rate": 3.2151317398605646e-05, + "loss": 0.2782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29341888427734375, + "step": 2420, + "valid_targets_mean": 2909.6, + "valid_targets_min": 1269 + }, + { + "epoch": 2.5472689075630255, + "grad_norm": 0.6452736492861937, + "learning_rate": 3.2109667115242815e-05, + "loss": 0.2678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2756231725215912, + "step": 2425, + "valid_targets_mean": 2893.6, + "valid_targets_min": 935 + }, + { + "epoch": 2.552521008403361, + "grad_norm": 0.6459381577493258, + "learning_rate": 3.206793375051694e-05, + "loss": 0.2821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2834113836288452, + "step": 2430, + "valid_targets_mean": 3271.2, + "valid_targets_min": 1695 + }, + { + "epoch": 2.5577731092436977, + "grad_norm": 0.6599512718640789, + "learning_rate": 3.2026117590750086e-05, + "loss": 0.2787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2733537554740906, + "step": 2435, + "valid_targets_mean": 3117.0, + "valid_targets_min": 671 + }, + { + "epoch": 2.5630252100840334, + "grad_norm": 0.567668736976835, + "learning_rate": 3.198421892283235e-05, + "loss": 0.2722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2664722502231598, + "step": 2440, + "valid_targets_mean": 3738.3, + "valid_targets_min": 1020 + }, + { + "epoch": 2.56827731092437, + "grad_norm": 0.6231370164675432, + "learning_rate": 3.1942238034219885e-05, + "loss": 0.2675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28766465187072754, + "step": 2445, + "valid_targets_mean": 3198.6, + "valid_targets_min": 1889 + }, + { + "epoch": 2.5735294117647056, + "grad_norm": 0.6316291488754031, + "learning_rate": 3.190017521293296e-05, + "loss": 0.2649, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2555858790874481, + "step": 2450, + "valid_targets_mean": 3085.4, + "valid_targets_min": 1385 + }, + { + "epoch": 2.578781512605042, + "grad_norm": 0.6263309654818119, + "learning_rate": 3.185803074755395e-05, + "loss": 0.2695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2710253596305847, + "step": 2455, + "valid_targets_mean": 3538.3, + "valid_targets_min": 1548 + }, + { + "epoch": 2.5840336134453783, + "grad_norm": 0.6117849557050007, + "learning_rate": 3.1815804927225363e-05, + "loss": 0.2718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26848098635673523, + "step": 2460, + "valid_targets_mean": 3289.2, + "valid_targets_min": 699 + }, + { + "epoch": 2.5892857142857144, + "grad_norm": 0.6124680799238794, + "learning_rate": 3.1773498041647876e-05, + "loss": 0.2706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27038100361824036, + "step": 2465, + "valid_targets_mean": 3072.8, + "valid_targets_min": 648 + }, + { + "epoch": 2.5945378151260505, + "grad_norm": 0.5501431568044479, + "learning_rate": 3.173111038107834e-05, + "loss": 0.2625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2502066195011139, + "step": 2470, + "valid_targets_mean": 4004.5, + "valid_targets_min": 1851 + }, + { + "epoch": 2.5997899159663866, + "grad_norm": 0.5757740316879347, + "learning_rate": 3.168864223632776e-05, + "loss": 0.2665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26100954413414, + "step": 2475, + "valid_targets_mean": 3751.6, + "valid_targets_min": 1540 + }, + { + "epoch": 2.6050420168067228, + "grad_norm": 0.623122866098537, + "learning_rate": 3.1646093898759334e-05, + "loss": 0.2675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.269726425409317, + "step": 2480, + "valid_targets_mean": 3402.6, + "valid_targets_min": 1410 + }, + { + "epoch": 2.610294117647059, + "grad_norm": 0.6140867953660799, + "learning_rate": 3.160346566028645e-05, + "loss": 0.2671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28402426838874817, + "step": 2485, + "valid_targets_mean": 3822.9, + "valid_targets_min": 2040 + }, + { + "epoch": 2.615546218487395, + "grad_norm": 0.6458360451166494, + "learning_rate": 3.156075781337068e-05, + "loss": 0.2722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2804952561855316, + "step": 2490, + "valid_targets_mean": 2824.4, + "valid_targets_min": 1458 + }, + { + "epoch": 2.620798319327731, + "grad_norm": 0.614578628604973, + "learning_rate": 3.1517970651019735e-05, + "loss": 0.2673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.267007976770401, + "step": 2495, + "valid_targets_mean": 3497.6, + "valid_targets_min": 1589 + }, + { + "epoch": 2.6260504201680672, + "grad_norm": 0.6550756431874655, + "learning_rate": 3.1475104466785536e-05, + "loss": 0.2628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2752663195133209, + "step": 2500, + "valid_targets_mean": 2745.1, + "valid_targets_min": 513 + }, + { + "epoch": 2.6313025210084033, + "grad_norm": 0.7285449233093796, + "learning_rate": 3.143215955476211e-05, + "loss": 0.2695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30785495042800903, + "step": 2505, + "valid_targets_mean": 3143.5, + "valid_targets_min": 1636 + }, + { + "epoch": 2.6365546218487395, + "grad_norm": 0.5763401907930161, + "learning_rate": 3.138913620958365e-05, + "loss": 0.266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26492440700531006, + "step": 2510, + "valid_targets_mean": 3906.8, + "valid_targets_min": 2383 + }, + { + "epoch": 2.6418067226890756, + "grad_norm": 0.5946439618794739, + "learning_rate": 3.134603472642244e-05, + "loss": 0.2672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25987619161605835, + "step": 2515, + "valid_targets_mean": 3180.3, + "valid_targets_min": 1372 + }, + { + "epoch": 2.6470588235294117, + "grad_norm": 0.7017146713713763, + "learning_rate": 3.1302855400986866e-05, + "loss": 0.2652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25516122579574585, + "step": 2520, + "valid_targets_mean": 2634.1, + "valid_targets_min": 1401 + }, + { + "epoch": 2.652310924369748, + "grad_norm": 0.7063131822296524, + "learning_rate": 3.1259598529519336e-05, + "loss": 0.27, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2827465236186981, + "step": 2525, + "valid_targets_mean": 2937.3, + "valid_targets_min": 1504 + }, + { + "epoch": 2.657563025210084, + "grad_norm": 0.6322826497339568, + "learning_rate": 3.121626440879432e-05, + "loss": 0.2532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24802082777023315, + "step": 2530, + "valid_targets_mean": 3097.8, + "valid_targets_min": 1648 + }, + { + "epoch": 2.66281512605042, + "grad_norm": 0.6405774977690727, + "learning_rate": 3.117285333611625e-05, + "loss": 0.2679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.247537761926651, + "step": 2535, + "valid_targets_mean": 3982.1, + "valid_targets_min": 959 + }, + { + "epoch": 2.668067226890756, + "grad_norm": 0.5925204140545449, + "learning_rate": 3.112936560931752e-05, + "loss": 0.2564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24100324511528015, + "step": 2540, + "valid_targets_mean": 3329.1, + "valid_targets_min": 1261 + }, + { + "epoch": 2.6733193277310923, + "grad_norm": 0.6277943771927688, + "learning_rate": 3.108580152675642e-05, + "loss": 0.2575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2357967495918274, + "step": 2545, + "valid_targets_mean": 3185.8, + "valid_targets_min": 1294 + }, + { + "epoch": 2.678571428571429, + "grad_norm": 0.6292055198908869, + "learning_rate": 3.104216138731511e-05, + "loss": 0.2676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2757669687271118, + "step": 2550, + "valid_targets_mean": 3340.1, + "valid_targets_min": 1432 + }, + { + "epoch": 2.6838235294117645, + "grad_norm": 0.6031199057951037, + "learning_rate": 3.0998445490397546e-05, + "loss": 0.2601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26977115869522095, + "step": 2555, + "valid_targets_mean": 3379.6, + "valid_targets_min": 742 + }, + { + "epoch": 2.689075630252101, + "grad_norm": 0.6624342295310539, + "learning_rate": 3.095465413592743e-05, + "loss": 0.276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27833864092826843, + "step": 2560, + "valid_targets_mean": 2969.1, + "valid_targets_min": 988 + }, + { + "epoch": 2.6943277310924367, + "grad_norm": 0.6361231679633825, + "learning_rate": 3.091078762434616e-05, + "loss": 0.265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2749124765396118, + "step": 2565, + "valid_targets_mean": 3232.6, + "valid_targets_min": 654 + }, + { + "epoch": 2.6995798319327733, + "grad_norm": 0.616757107135465, + "learning_rate": 3.086684625661081e-05, + "loss": 0.2497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2631353735923767, + "step": 2570, + "valid_targets_mean": 3125.7, + "valid_targets_min": 1328 + }, + { + "epoch": 2.7048319327731094, + "grad_norm": 0.6893720258364816, + "learning_rate": 3.082283033419197e-05, + "loss": 0.2647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2686452567577362, + "step": 2575, + "valid_targets_mean": 3166.4, + "valid_targets_min": 1302 + }, + { + "epoch": 2.7100840336134455, + "grad_norm": 0.6092903528556366, + "learning_rate": 3.077874015907174e-05, + "loss": 0.2529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2558513283729553, + "step": 2580, + "valid_targets_mean": 4101.6, + "valid_targets_min": 1701 + }, + { + "epoch": 2.7153361344537816, + "grad_norm": 0.5889526951400187, + "learning_rate": 3.073457603374167e-05, + "loss": 0.2533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25581663846969604, + "step": 2585, + "valid_targets_mean": 3677.3, + "valid_targets_min": 1319 + }, + { + "epoch": 2.7205882352941178, + "grad_norm": 0.642935413766165, + "learning_rate": 3.069033826120066e-05, + "loss": 0.2416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.255086213350296, + "step": 2590, + "valid_targets_mean": 3259.8, + "valid_targets_min": 1194 + }, + { + "epoch": 2.725840336134454, + "grad_norm": 0.6235865519519191, + "learning_rate": 3.064602714495286e-05, + "loss": 0.2711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25447195768356323, + "step": 2595, + "valid_targets_mean": 2886.7, + "valid_targets_min": 1220 + }, + { + "epoch": 2.73109243697479, + "grad_norm": 0.5840881578880576, + "learning_rate": 3.060164298900565e-05, + "loss": 0.2541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25606703758239746, + "step": 2600, + "valid_targets_mean": 3948.4, + "valid_targets_min": 1742 + }, + { + "epoch": 2.736344537815126, + "grad_norm": 0.70617682044821, + "learning_rate": 3.055718609786749e-05, + "loss": 0.2641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2716052234172821, + "step": 2605, + "valid_targets_mean": 2877.3, + "valid_targets_min": 1234 + }, + { + "epoch": 2.741596638655462, + "grad_norm": 0.6377305035799299, + "learning_rate": 3.051265677654585e-05, + "loss": 0.2687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2713176906108856, + "step": 2610, + "valid_targets_mean": 3126.6, + "valid_targets_min": 1686 + }, + { + "epoch": 2.7468487394957983, + "grad_norm": 0.5626998875518994, + "learning_rate": 3.0468055330545162e-05, + "loss": 0.2651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23960676789283752, + "step": 2615, + "valid_targets_mean": 3957.4, + "valid_targets_min": 618 + }, + { + "epoch": 2.7521008403361344, + "grad_norm": 0.6556128296668272, + "learning_rate": 3.0423382065864653e-05, + "loss": 0.2577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25945764780044556, + "step": 2620, + "valid_targets_mean": 2844.3, + "valid_targets_min": 1364 + }, + { + "epoch": 2.7573529411764706, + "grad_norm": 0.6666237064634677, + "learning_rate": 3.0378637288996304e-05, + "loss": 0.2505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26987403631210327, + "step": 2625, + "valid_targets_mean": 3537.1, + "valid_targets_min": 1514 + }, + { + "epoch": 2.7626050420168067, + "grad_norm": 0.6590248047133942, + "learning_rate": 3.033382130692269e-05, + "loss": 0.2504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26807478070259094, + "step": 2630, + "valid_targets_mean": 2863.1, + "valid_targets_min": 1073 + }, + { + "epoch": 2.767857142857143, + "grad_norm": 0.6761466657105407, + "learning_rate": 3.0288934427114955e-05, + "loss": 0.2553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27322080731391907, + "step": 2635, + "valid_targets_mean": 2907.8, + "valid_targets_min": 1026 + }, + { + "epoch": 2.773109243697479, + "grad_norm": 0.6353889677418061, + "learning_rate": 3.0243976957530602e-05, + "loss": 0.2463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23086395859718323, + "step": 2640, + "valid_targets_mean": 2695.9, + "valid_targets_min": 1254 + }, + { + "epoch": 2.778361344537815, + "grad_norm": 0.7083565681570431, + "learning_rate": 3.0198949206611475e-05, + "loss": 0.2493, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27140820026397705, + "step": 2645, + "valid_targets_mean": 3112.3, + "valid_targets_min": 720 + }, + { + "epoch": 2.783613445378151, + "grad_norm": 0.6445873412413864, + "learning_rate": 3.0153851483281582e-05, + "loss": 0.2574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29137080907821655, + "step": 2650, + "valid_targets_mean": 3687.8, + "valid_targets_min": 1817 + }, + { + "epoch": 2.7888655462184873, + "grad_norm": 0.6275537197106373, + "learning_rate": 3.010868409694499e-05, + "loss": 0.2617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2466498464345932, + "step": 2655, + "valid_targets_mean": 3197.7, + "valid_targets_min": 1411 + }, + { + "epoch": 2.7941176470588234, + "grad_norm": 0.5902973797283267, + "learning_rate": 3.0063447357483712e-05, + "loss": 0.2595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2413676530122757, + "step": 2660, + "valid_targets_mean": 3804.1, + "valid_targets_min": 2253 + }, + { + "epoch": 2.79936974789916, + "grad_norm": 0.6686261268710058, + "learning_rate": 3.0018141575255587e-05, + "loss": 0.2557, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2801203727722168, + "step": 2665, + "valid_targets_mean": 3269.8, + "valid_targets_min": 1860 + }, + { + "epoch": 2.8046218487394956, + "grad_norm": 0.5958481124800947, + "learning_rate": 2.9972767061092107e-05, + "loss": 0.2523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24206840991973877, + "step": 2670, + "valid_targets_mean": 3550.5, + "valid_targets_min": 1805 + }, + { + "epoch": 2.809873949579832, + "grad_norm": 0.5964898157229672, + "learning_rate": 2.9927324126296352e-05, + "loss": 0.2584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24410061538219452, + "step": 2675, + "valid_targets_mean": 3646.2, + "valid_targets_min": 1145 + }, + { + "epoch": 2.815126050420168, + "grad_norm": 0.6446057186222791, + "learning_rate": 2.98818130826408e-05, + "loss": 0.252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2676333785057068, + "step": 2680, + "valid_targets_mean": 3167.9, + "valid_targets_min": 1722 + }, + { + "epoch": 2.8203781512605044, + "grad_norm": 0.7582205523508831, + "learning_rate": 2.9836234242365204e-05, + "loss": 0.2476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25383812189102173, + "step": 2685, + "valid_targets_mean": 2996.0, + "valid_targets_min": 1423 + }, + { + "epoch": 2.82563025210084, + "grad_norm": 0.6301709788550426, + "learning_rate": 2.979058791817447e-05, + "loss": 0.2661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2749696969985962, + "step": 2690, + "valid_targets_mean": 3538.0, + "valid_targets_min": 1426 + }, + { + "epoch": 2.8308823529411766, + "grad_norm": 0.6480505666026394, + "learning_rate": 2.974487442323647e-05, + "loss": 0.2685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2599114179611206, + "step": 2695, + "valid_targets_mean": 3115.1, + "valid_targets_min": 538 + }, + { + "epoch": 2.8361344537815127, + "grad_norm": 0.6882139342020044, + "learning_rate": 2.9699094071179938e-05, + "loss": 0.2552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2708072364330292, + "step": 2700, + "valid_targets_mean": 2551.9, + "valid_targets_min": 537 + }, + { + "epoch": 2.841386554621849, + "grad_norm": 0.6582888441321656, + "learning_rate": 2.9653247176092286e-05, + "loss": 0.2684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2539324164390564, + "step": 2705, + "valid_targets_mean": 3079.0, + "valid_targets_min": 845 + }, + { + "epoch": 2.846638655462185, + "grad_norm": 0.638833559343612, + "learning_rate": 2.9607334052517466e-05, + "loss": 0.2569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.28745412826538086, + "step": 2710, + "valid_targets_mean": 3822.1, + "valid_targets_min": 1625 + }, + { + "epoch": 2.851890756302521, + "grad_norm": 0.7422015502102081, + "learning_rate": 2.9561355015453803e-05, + "loss": 0.2732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29670873284339905, + "step": 2715, + "valid_targets_mean": 2883.9, + "valid_targets_min": 1400 + }, + { + "epoch": 2.857142857142857, + "grad_norm": 0.6490748235477826, + "learning_rate": 2.9515310380351847e-05, + "loss": 0.2608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24093633890151978, + "step": 2720, + "valid_targets_mean": 3222.6, + "valid_targets_min": 1868 + }, + { + "epoch": 2.8623949579831933, + "grad_norm": 0.7372791077034592, + "learning_rate": 2.9469200463112186e-05, + "loss": 0.2659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25444963574409485, + "step": 2725, + "valid_targets_mean": 2862.1, + "valid_targets_min": 648 + }, + { + "epoch": 2.8676470588235294, + "grad_norm": 0.7192710572965154, + "learning_rate": 2.9423025580083298e-05, + "loss": 0.2623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.245719313621521, + "step": 2730, + "valid_targets_mean": 3024.6, + "valid_targets_min": 1404 + }, + { + "epoch": 2.8728991596638656, + "grad_norm": 0.6647970803562331, + "learning_rate": 2.937678604805938e-05, + "loss": 0.268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2540484070777893, + "step": 2735, + "valid_targets_mean": 3429.3, + "valid_targets_min": 814 + }, + { + "epoch": 2.8781512605042017, + "grad_norm": 0.612522922361621, + "learning_rate": 2.9330482184278168e-05, + "loss": 0.2619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2477385699748993, + "step": 2740, + "valid_targets_mean": 3270.8, + "valid_targets_min": 953 + }, + { + "epoch": 2.883403361344538, + "grad_norm": 0.7016630866988153, + "learning_rate": 2.9284114306418758e-05, + "loss": 0.2545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2649661898612976, + "step": 2745, + "valid_targets_mean": 3123.0, + "valid_targets_min": 1410 + }, + { + "epoch": 2.888655462184874, + "grad_norm": 0.6661752773535051, + "learning_rate": 2.923768273259942e-05, + "loss": 0.2639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.263710081577301, + "step": 2750, + "valid_targets_mean": 3085.5, + "valid_targets_min": 530 + }, + { + "epoch": 2.89390756302521, + "grad_norm": 0.6197849509680783, + "learning_rate": 2.919118778137546e-05, + "loss": 0.2659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25989899039268494, + "step": 2755, + "valid_targets_mean": 3491.7, + "valid_targets_min": 1066 + }, + { + "epoch": 2.899159663865546, + "grad_norm": 0.594389488012544, + "learning_rate": 2.9144629771736954e-05, + "loss": 0.2689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2792191207408905, + "step": 2760, + "valid_targets_mean": 3904.9, + "valid_targets_min": 1613 + }, + { + "epoch": 2.9044117647058822, + "grad_norm": 0.5752638283362668, + "learning_rate": 2.9098009023106645e-05, + "loss": 0.2538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23264659941196442, + "step": 2765, + "valid_targets_mean": 3731.6, + "valid_targets_min": 952 + }, + { + "epoch": 2.9096638655462184, + "grad_norm": 0.6490696888490648, + "learning_rate": 2.9051325855337684e-05, + "loss": 0.2605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26240965723991394, + "step": 2770, + "valid_targets_mean": 2823.0, + "valid_targets_min": 1025 + }, + { + "epoch": 2.9149159663865545, + "grad_norm": 0.6126363252048349, + "learning_rate": 2.90045805887115e-05, + "loss": 0.2539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27359646558761597, + "step": 2775, + "valid_targets_mean": 4032.0, + "valid_targets_min": 1397 + }, + { + "epoch": 2.9201680672268906, + "grad_norm": 0.6068839428847373, + "learning_rate": 2.8957773543935518e-05, + "loss": 0.2559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2474924623966217, + "step": 2780, + "valid_targets_mean": 3534.0, + "valid_targets_min": 1920 + }, + { + "epoch": 2.9254201680672267, + "grad_norm": 0.6329791757885024, + "learning_rate": 2.8910905042141046e-05, + "loss": 0.2452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25704294443130493, + "step": 2785, + "valid_targets_mean": 3527.5, + "valid_targets_min": 1687 + }, + { + "epoch": 2.9306722689075633, + "grad_norm": 0.6948103907388832, + "learning_rate": 2.8863975404881022e-05, + "loss": 0.2574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30331847071647644, + "step": 2790, + "valid_targets_mean": 3213.0, + "valid_targets_min": 1564 + }, + { + "epoch": 2.935924369747899, + "grad_norm": 0.6352221742702212, + "learning_rate": 2.881698495412781e-05, + "loss": 0.2477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2414446771144867, + "step": 2795, + "valid_targets_mean": 3689.7, + "valid_targets_min": 1947 + }, + { + "epoch": 2.9411764705882355, + "grad_norm": 0.6629581821365779, + "learning_rate": 2.8769934012271004e-05, + "loss": 0.2462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25725820660591125, + "step": 2800, + "valid_targets_mean": 2978.2, + "valid_targets_min": 1447 + }, + { + "epoch": 2.946428571428571, + "grad_norm": 0.6673202572073865, + "learning_rate": 2.8722822902115228e-05, + "loss": 0.2448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25360405445098877, + "step": 2805, + "valid_targets_mean": 3016.5, + "valid_targets_min": 1243 + }, + { + "epoch": 2.9516806722689077, + "grad_norm": 0.7063504788496237, + "learning_rate": 2.8675651946877884e-05, + "loss": 0.2506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27201586961746216, + "step": 2810, + "valid_targets_mean": 3199.0, + "valid_targets_min": 1679 + }, + { + "epoch": 2.956932773109244, + "grad_norm": 0.6749918994768315, + "learning_rate": 2.862842147018696e-05, + "loss": 0.2667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2709447741508484, + "step": 2815, + "valid_targets_mean": 3001.2, + "valid_targets_min": 1695 + }, + { + "epoch": 2.96218487394958, + "grad_norm": 0.6389022545015189, + "learning_rate": 2.858113179607881e-05, + "loss": 0.2539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24959295988082886, + "step": 2820, + "valid_targets_mean": 3370.9, + "valid_targets_min": 1469 + }, + { + "epoch": 2.967436974789916, + "grad_norm": 0.7014173062929902, + "learning_rate": 2.8533783248995938e-05, + "loss": 0.2515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24700427055358887, + "step": 2825, + "valid_targets_mean": 3245.9, + "valid_targets_min": 1593 + }, + { + "epoch": 2.972689075630252, + "grad_norm": 0.6469956378099957, + "learning_rate": 2.848637615378475e-05, + "loss": 0.2509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27041953802108765, + "step": 2830, + "valid_targets_mean": 3130.5, + "valid_targets_min": 1664 + }, + { + "epoch": 2.9779411764705883, + "grad_norm": 0.6169242449395833, + "learning_rate": 2.8438910835693314e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22996433079242706, + "step": 2835, + "valid_targets_mean": 3460.2, + "valid_targets_min": 1605 + }, + { + "epoch": 2.9831932773109244, + "grad_norm": 0.6394181377322952, + "learning_rate": 2.8391387620369192e-05, + "loss": 0.2729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2803104519844055, + "step": 2840, + "valid_targets_mean": 3288.6, + "valid_targets_min": 842 + }, + { + "epoch": 2.9884453781512605, + "grad_norm": 0.6854150282623844, + "learning_rate": 2.834380683385714e-05, + "loss": 0.265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2902924418449402, + "step": 2845, + "valid_targets_mean": 3601.4, + "valid_targets_min": 809 + }, + { + "epoch": 2.9936974789915967, + "grad_norm": 0.6051689785920974, + "learning_rate": 2.8296168802596892e-05, + "loss": 0.2543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2509153187274933, + "step": 2850, + "valid_targets_mean": 3752.4, + "valid_targets_min": 2026 + }, + { + "epoch": 2.9989495798319328, + "grad_norm": 0.6083481456927304, + "learning_rate": 2.8248473853420933e-05, + "loss": 0.2533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2585148811340332, + "step": 2855, + "valid_targets_mean": 3430.5, + "valid_targets_min": 1214 + }, + { + "epoch": 3.004201680672269, + "grad_norm": 0.7319423401810515, + "learning_rate": 2.8200722313552247e-05, + "loss": 0.2164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2003440260887146, + "step": 2860, + "valid_targets_mean": 1489.3, + "valid_targets_min": 722 + }, + { + "epoch": 3.009453781512605, + "grad_norm": 0.717771718263267, + "learning_rate": 2.8152914510602073e-05, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17707990109920502, + "step": 2865, + "valid_targets_mean": 1515.6, + "valid_targets_min": 725 + }, + { + "epoch": 3.014705882352941, + "grad_norm": 0.8226879595885981, + "learning_rate": 2.810505077256765e-05, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18063870072364807, + "step": 2870, + "valid_targets_mean": 1443.1, + "valid_targets_min": 874 + }, + { + "epoch": 3.0199579831932772, + "grad_norm": 0.8338357187658497, + "learning_rate": 2.8057131427829983e-05, + "loss": 0.1696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17110946774482727, + "step": 2875, + "valid_targets_mean": 1316.8, + "valid_targets_min": 615 + }, + { + "epoch": 3.0252100840336134, + "grad_norm": 0.7376091557217379, + "learning_rate": 2.800915680515157e-05, + "loss": 0.1698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17430976033210754, + "step": 2880, + "valid_targets_mean": 1540.8, + "valid_targets_min": 715 + }, + { + "epoch": 3.0304621848739495, + "grad_norm": 0.7831042239119287, + "learning_rate": 2.7961127233674167e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2140425145626068, + "step": 2885, + "valid_targets_mean": 1648.9, + "valid_targets_min": 625 + }, + { + "epoch": 3.0357142857142856, + "grad_norm": 0.7415542108375209, + "learning_rate": 2.7913043042916513e-05, + "loss": 0.1653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1582803726196289, + "step": 2890, + "valid_targets_mean": 1354.9, + "valid_targets_min": 761 + }, + { + "epoch": 3.0409663865546217, + "grad_norm": 0.715142653421594, + "learning_rate": 2.786490456277208e-05, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25299903750419617, + "step": 2895, + "valid_targets_mean": 1589.9, + "valid_targets_min": 703 + }, + { + "epoch": 3.046218487394958, + "grad_norm": 0.7873361806067022, + "learning_rate": 2.7816712123506796e-05, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1710663139820099, + "step": 2900, + "valid_targets_mean": 1485.4, + "valid_targets_min": 733 + }, + { + "epoch": 3.051470588235294, + "grad_norm": 0.7770722242990561, + "learning_rate": 2.776846605575681e-05, + "loss": 0.1731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1824086308479309, + "step": 2905, + "valid_targets_mean": 1712.8, + "valid_targets_min": 964 + }, + { + "epoch": 3.05672268907563, + "grad_norm": 0.7616729738480414, + "learning_rate": 2.772016669052618e-05, + "loss": 0.164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1748737096786499, + "step": 2910, + "valid_targets_mean": 1498.4, + "valid_targets_min": 815 + }, + { + "epoch": 3.0619747899159666, + "grad_norm": 0.7583507582473836, + "learning_rate": 2.767181435918464e-05, + "loss": 0.1724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1768401861190796, + "step": 2915, + "valid_targets_mean": 1677.5, + "valid_targets_min": 1027 + }, + { + "epoch": 3.0672268907563027, + "grad_norm": 0.7278307085282013, + "learning_rate": 2.7623409393465298e-05, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1581869125366211, + "step": 2920, + "valid_targets_mean": 1325.3, + "valid_targets_min": 570 + }, + { + "epoch": 3.072478991596639, + "grad_norm": 0.7617121869975262, + "learning_rate": 2.7574952125462386e-05, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16011205315589905, + "step": 2925, + "valid_targets_mean": 1324.7, + "valid_targets_min": 548 + }, + { + "epoch": 3.077731092436975, + "grad_norm": 0.7554042058432768, + "learning_rate": 2.7526442887628962e-05, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15193405747413635, + "step": 2930, + "valid_targets_mean": 1288.1, + "valid_targets_min": 682 + }, + { + "epoch": 3.082983193277311, + "grad_norm": 0.737075733110425, + "learning_rate": 2.7477882012774633e-05, + "loss": 0.162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16672229766845703, + "step": 2935, + "valid_targets_mean": 1563.0, + "valid_targets_min": 911 + }, + { + "epoch": 3.088235294117647, + "grad_norm": 0.8359898452206559, + "learning_rate": 2.742926983406328e-05, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16956086456775665, + "step": 2940, + "valid_targets_mean": 1278.3, + "valid_targets_min": 597 + }, + { + "epoch": 3.0934873949579833, + "grad_norm": 0.7553756145118757, + "learning_rate": 2.7380606685010767e-05, + "loss": 0.1735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17766933143138885, + "step": 2945, + "valid_targets_mean": 1552.1, + "valid_targets_min": 772 + }, + { + "epoch": 3.0987394957983194, + "grad_norm": 0.7472526084646638, + "learning_rate": 2.7331892899482654e-05, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1692330241203308, + "step": 2950, + "valid_targets_mean": 1548.2, + "valid_targets_min": 940 + }, + { + "epoch": 3.1039915966386555, + "grad_norm": 0.6799508713485111, + "learning_rate": 2.7283128811691893e-05, + "loss": 0.1642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.146545872092247, + "step": 2955, + "valid_targets_mean": 1334.6, + "valid_targets_min": 785 + }, + { + "epoch": 3.1092436974789917, + "grad_norm": 0.7171813544493029, + "learning_rate": 2.7234314756196564e-05, + "loss": 0.1689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15770323574543, + "step": 2960, + "valid_targets_mean": 1404.9, + "valid_targets_min": 617 + }, + { + "epoch": 3.1144957983193278, + "grad_norm": 0.8566587780828107, + "learning_rate": 2.718545106789755e-05, + "loss": 0.1698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17677560448646545, + "step": 2965, + "valid_targets_mean": 1266.8, + "valid_targets_min": 648 + }, + { + "epoch": 3.119747899159664, + "grad_norm": 0.8193397073508751, + "learning_rate": 2.713653808203626e-05, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15627950429916382, + "step": 2970, + "valid_targets_mean": 1427.1, + "valid_targets_min": 634 + }, + { + "epoch": 3.125, + "grad_norm": 0.8274451736674825, + "learning_rate": 2.7087576134192305e-05, + "loss": 0.1606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16775844991207123, + "step": 2975, + "valid_targets_mean": 1421.0, + "valid_targets_min": 774 + }, + { + "epoch": 3.130252100840336, + "grad_norm": 0.6953309030381479, + "learning_rate": 2.7038565560281252e-05, + "loss": 0.1607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16650184988975525, + "step": 2980, + "valid_targets_mean": 1685.9, + "valid_targets_min": 563 + }, + { + "epoch": 3.1355042016806722, + "grad_norm": 0.8099890208928163, + "learning_rate": 2.6989506696552224e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16752278804779053, + "step": 2985, + "valid_targets_mean": 1378.0, + "valid_targets_min": 837 + }, + { + "epoch": 3.1407563025210083, + "grad_norm": 0.7795749997325356, + "learning_rate": 2.6940399879585685e-05, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16560722887516022, + "step": 2990, + "valid_targets_mean": 1599.2, + "valid_targets_min": 803 + }, + { + "epoch": 3.1460084033613445, + "grad_norm": 0.7336389778027822, + "learning_rate": 2.6891245446291093e-05, + "loss": 0.1674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14416930079460144, + "step": 2995, + "valid_targets_mean": 1380.4, + "valid_targets_min": 797 + }, + { + "epoch": 3.1512605042016806, + "grad_norm": 0.7490034715901619, + "learning_rate": 2.6842043733904578e-05, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15650300681591034, + "step": 3000, + "valid_targets_mean": 1571.5, + "valid_targets_min": 646 + }, + { + "epoch": 3.1565126050420167, + "grad_norm": 0.8401911427116212, + "learning_rate": 2.6792795079986634e-05, + "loss": 0.1679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17290136218070984, + "step": 3005, + "valid_targets_mean": 1447.8, + "valid_targets_min": 591 + }, + { + "epoch": 3.161764705882353, + "grad_norm": 0.8101203980462212, + "learning_rate": 2.6743499822419833e-05, + "loss": 0.1663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18794336915016174, + "step": 3010, + "valid_targets_mean": 1512.7, + "valid_targets_min": 796 + }, + { + "epoch": 3.167016806722689, + "grad_norm": 0.7549079816000475, + "learning_rate": 2.669415829940646e-05, + "loss": 0.1578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.148887038230896, + "step": 3015, + "valid_targets_mean": 1333.1, + "valid_targets_min": 848 + }, + { + "epoch": 3.172268907563025, + "grad_norm": 0.8655198786602623, + "learning_rate": 2.6644770849466226e-05, + "loss": 0.1575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14870238304138184, + "step": 3020, + "valid_targets_mean": 1233.6, + "valid_targets_min": 665 + }, + { + "epoch": 3.177521008403361, + "grad_norm": 0.6869481556598374, + "learning_rate": 2.6595337811433925e-05, + "loss": 0.1573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1471511870622635, + "step": 3025, + "valid_targets_mean": 1408.1, + "valid_targets_min": 652 + }, + { + "epoch": 3.1827731092436973, + "grad_norm": 0.8476070443290739, + "learning_rate": 2.654585952445714e-05, + "loss": 0.1563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17854884266853333, + "step": 3030, + "valid_targets_mean": 1573.8, + "valid_targets_min": 662 + }, + { + "epoch": 3.1880252100840334, + "grad_norm": 0.759943747176002, + "learning_rate": 2.6496336327993882e-05, + "loss": 0.1566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17004242539405823, + "step": 3035, + "valid_targets_mean": 1410.9, + "valid_targets_min": 770 + }, + { + "epoch": 3.19327731092437, + "grad_norm": 0.7914392244697611, + "learning_rate": 2.6446768561810264e-05, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16128523647785187, + "step": 3040, + "valid_targets_mean": 1338.0, + "valid_targets_min": 842 + }, + { + "epoch": 3.198529411764706, + "grad_norm": 0.7573632096372179, + "learning_rate": 2.6397156565978198e-05, + "loss": 0.1623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16719350218772888, + "step": 3045, + "valid_targets_mean": 1522.6, + "valid_targets_min": 773 + }, + { + "epoch": 3.203781512605042, + "grad_norm": 0.7823159419268372, + "learning_rate": 2.6347500680873043e-05, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.170121967792511, + "step": 3050, + "valid_targets_mean": 1530.2, + "valid_targets_min": 807 + }, + { + "epoch": 3.2090336134453783, + "grad_norm": 0.7715992991548748, + "learning_rate": 2.6297801247171258e-05, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14779695868492126, + "step": 3055, + "valid_targets_mean": 1266.6, + "valid_targets_min": 782 + }, + { + "epoch": 3.2142857142857144, + "grad_norm": 0.7213196908664066, + "learning_rate": 2.6248058605848097e-05, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17193026840686798, + "step": 3060, + "valid_targets_mean": 1624.2, + "valid_targets_min": 634 + }, + { + "epoch": 3.2195378151260505, + "grad_norm": 0.7460107195452109, + "learning_rate": 2.6198273098175234e-05, + "loss": 0.1684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1723858267068863, + "step": 3065, + "valid_targets_mean": 1508.1, + "valid_targets_min": 573 + }, + { + "epoch": 3.2247899159663866, + "grad_norm": 0.7033456921737298, + "learning_rate": 2.614844506571844e-05, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16065514087677002, + "step": 3070, + "valid_targets_mean": 1645.1, + "valid_targets_min": 718 + }, + { + "epoch": 3.2300420168067228, + "grad_norm": 0.700948054604266, + "learning_rate": 2.609857485033525e-05, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13682524859905243, + "step": 3075, + "valid_targets_mean": 1385.9, + "valid_targets_min": 685 + }, + { + "epoch": 3.235294117647059, + "grad_norm": 0.8365081550151657, + "learning_rate": 2.6048662794172595e-05, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1537787914276123, + "step": 3080, + "valid_targets_mean": 1489.9, + "valid_targets_min": 744 + }, + { + "epoch": 3.240546218487395, + "grad_norm": 0.9620567262629217, + "learning_rate": 2.599870923966447e-05, + "loss": 0.1613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15811693668365479, + "step": 3085, + "valid_targets_mean": 1253.3, + "valid_targets_min": 650 + }, + { + "epoch": 3.245798319327731, + "grad_norm": 0.7129512442205985, + "learning_rate": 2.5948714529529565e-05, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16663914918899536, + "step": 3090, + "valid_targets_mean": 1337.3, + "valid_targets_min": 780 + }, + { + "epoch": 3.2510504201680672, + "grad_norm": 0.728601631393252, + "learning_rate": 2.5898679006768958e-05, + "loss": 0.1507, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14995308220386505, + "step": 3095, + "valid_targets_mean": 1344.5, + "valid_targets_min": 677 + }, + { + "epoch": 3.2563025210084033, + "grad_norm": 0.7807161695418345, + "learning_rate": 2.5848603014663702e-05, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15245237946510315, + "step": 3100, + "valid_targets_mean": 1302.3, + "valid_targets_min": 727 + }, + { + "epoch": 3.2615546218487395, + "grad_norm": 0.7833067221288974, + "learning_rate": 2.579848689677252e-05, + "loss": 0.16, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.164436474442482, + "step": 3105, + "valid_targets_mean": 1489.2, + "valid_targets_min": 726 + }, + { + "epoch": 3.2668067226890756, + "grad_norm": 0.7582720046388419, + "learning_rate": 2.5748330996929414e-05, + "loss": 0.1641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17622435092926025, + "step": 3110, + "valid_targets_mean": 1497.3, + "valid_targets_min": 669 + }, + { + "epoch": 3.2720588235294117, + "grad_norm": 0.6761627015093946, + "learning_rate": 2.5698135659241334e-05, + "loss": 0.1564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14206218719482422, + "step": 3115, + "valid_targets_mean": 1376.9, + "valid_targets_min": 750 + }, + { + "epoch": 3.277310924369748, + "grad_norm": 0.7569519780492563, + "learning_rate": 2.564790122808579e-05, + "loss": 0.1572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15797209739685059, + "step": 3120, + "valid_targets_mean": 1475.9, + "valid_targets_min": 736 + }, + { + "epoch": 3.282563025210084, + "grad_norm": 0.7724507253369616, + "learning_rate": 2.5597628048108504e-05, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1651463806629181, + "step": 3125, + "valid_targets_mean": 1426.2, + "valid_targets_min": 694 + }, + { + "epoch": 3.28781512605042, + "grad_norm": 0.7583229755048851, + "learning_rate": 2.5547316464221058e-05, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1635044515132904, + "step": 3130, + "valid_targets_mean": 1547.6, + "valid_targets_min": 753 + }, + { + "epoch": 3.293067226890756, + "grad_norm": 0.7492941228082307, + "learning_rate": 2.5496966821598493e-05, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14868690073490143, + "step": 3135, + "valid_targets_mean": 1277.1, + "valid_targets_min": 809 + }, + { + "epoch": 3.2983193277310923, + "grad_norm": 0.7918115514025105, + "learning_rate": 2.5446579465676977e-05, + "loss": 0.1565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14087459444999695, + "step": 3140, + "valid_targets_mean": 1292.3, + "valid_targets_min": 698 + }, + { + "epoch": 3.3035714285714284, + "grad_norm": 0.7674670047525459, + "learning_rate": 2.5396154742151407e-05, + "loss": 0.1457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15545767545700073, + "step": 3145, + "valid_targets_mean": 1448.1, + "valid_targets_min": 750 + }, + { + "epoch": 3.3088235294117645, + "grad_norm": 0.8361476210011558, + "learning_rate": 2.5345692996973075e-05, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17045646905899048, + "step": 3150, + "valid_targets_mean": 1472.6, + "valid_targets_min": 733 + }, + { + "epoch": 3.314075630252101, + "grad_norm": 0.7425956748582699, + "learning_rate": 2.5295194576347233e-05, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15749000012874603, + "step": 3155, + "valid_targets_mean": 1597.3, + "valid_targets_min": 907 + }, + { + "epoch": 3.3193277310924367, + "grad_norm": 0.778056511779034, + "learning_rate": 2.5244659826730774e-05, + "loss": 0.1519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17436806857585907, + "step": 3160, + "valid_targets_mean": 1564.9, + "valid_targets_min": 898 + }, + { + "epoch": 3.3245798319327733, + "grad_norm": 0.7457317148773122, + "learning_rate": 2.519408909482984e-05, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16804683208465576, + "step": 3165, + "valid_targets_mean": 1660.6, + "valid_targets_min": 1119 + }, + { + "epoch": 3.3298319327731094, + "grad_norm": 0.8015187816073955, + "learning_rate": 2.5143482727597427e-05, + "loss": 0.1572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16357378661632538, + "step": 3170, + "valid_targets_mean": 1392.3, + "valid_targets_min": 700 + }, + { + "epoch": 3.3350840336134455, + "grad_norm": 0.7927587699859597, + "learning_rate": 2.509284107223102e-05, + "loss": 0.155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15347766876220703, + "step": 3175, + "valid_targets_mean": 1407.8, + "valid_targets_min": 798 + }, + { + "epoch": 3.3403361344537816, + "grad_norm": 0.8291776111318878, + "learning_rate": 2.50421644761702e-05, + "loss": 0.155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14594411849975586, + "step": 3180, + "valid_targets_mean": 1237.5, + "valid_targets_min": 589 + }, + { + "epoch": 3.3455882352941178, + "grad_norm": 0.786383044943235, + "learning_rate": 2.499145328709429e-05, + "loss": 0.1552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1620325744152069, + "step": 3185, + "valid_targets_mean": 1363.2, + "valid_targets_min": 673 + }, + { + "epoch": 3.350840336134454, + "grad_norm": 0.8314225202905355, + "learning_rate": 2.4940707852919924e-05, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1637428104877472, + "step": 3190, + "valid_targets_mean": 1297.5, + "valid_targets_min": 641 + }, + { + "epoch": 3.35609243697479, + "grad_norm": 0.8247848576495689, + "learning_rate": 2.488992852179868e-05, + "loss": 0.1534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1613052636384964, + "step": 3195, + "valid_targets_mean": 1466.3, + "valid_targets_min": 701 + }, + { + "epoch": 3.361344537815126, + "grad_norm": 0.7169780960903507, + "learning_rate": 2.483911564211472e-05, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1574411392211914, + "step": 3200, + "valid_targets_mean": 1801.2, + "valid_targets_min": 1202 + }, + { + "epoch": 3.366596638655462, + "grad_norm": 0.6906945355495036, + "learning_rate": 2.4788269562482355e-05, + "loss": 0.1575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15213462710380554, + "step": 3205, + "valid_targets_mean": 1441.4, + "valid_targets_min": 657 + }, + { + "epoch": 3.3718487394957983, + "grad_norm": 0.7855731357671363, + "learning_rate": 2.4737390631743675e-05, + "loss": 0.1601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15283282101154327, + "step": 3210, + "valid_targets_mean": 1442.9, + "valid_targets_min": 672 + }, + { + "epoch": 3.3771008403361344, + "grad_norm": 0.7999940593483418, + "learning_rate": 2.4686479198966146e-05, + "loss": 0.1561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16701379418373108, + "step": 3215, + "valid_targets_mean": 1305.1, + "valid_targets_min": 723 + }, + { + "epoch": 3.3823529411764706, + "grad_norm": 0.7802222544911187, + "learning_rate": 2.463553561344024e-05, + "loss": 0.1492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1388361155986786, + "step": 3220, + "valid_targets_mean": 1236.6, + "valid_targets_min": 732 + }, + { + "epoch": 3.3876050420168067, + "grad_norm": 0.7997658424367834, + "learning_rate": 2.4584560224677013e-05, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14376091957092285, + "step": 3225, + "valid_targets_mean": 1261.1, + "valid_targets_min": 634 + }, + { + "epoch": 3.392857142857143, + "grad_norm": 0.7589895889200633, + "learning_rate": 2.4533553382405702e-05, + "loss": 0.1524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16269132494926453, + "step": 3230, + "valid_targets_mean": 1451.1, + "valid_targets_min": 905 + }, + { + "epoch": 3.398109243697479, + "grad_norm": 0.9088457131074477, + "learning_rate": 2.4482515436571365e-05, + "loss": 0.149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1583980917930603, + "step": 3235, + "valid_targets_mean": 1466.4, + "valid_targets_min": 596 + }, + { + "epoch": 3.403361344537815, + "grad_norm": 0.8253891514753098, + "learning_rate": 2.4431446737332437e-05, + "loss": 0.1514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16583885252475739, + "step": 3240, + "valid_targets_mean": 1435.1, + "valid_targets_min": 756 + }, + { + "epoch": 3.408613445378151, + "grad_norm": 0.7126823695261846, + "learning_rate": 2.438034763505833e-05, + "loss": 0.1617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14703774452209473, + "step": 3245, + "valid_targets_mean": 1411.8, + "valid_targets_min": 567 + }, + { + "epoch": 3.4138655462184873, + "grad_norm": 0.7389656217215542, + "learning_rate": 2.432921848032707e-05, + "loss": 0.1507, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14308568835258484, + "step": 3250, + "valid_targets_mean": 1495.9, + "valid_targets_min": 810 + }, + { + "epoch": 3.4191176470588234, + "grad_norm": 0.7063481251451478, + "learning_rate": 2.427805962392286e-05, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15941093862056732, + "step": 3255, + "valid_targets_mean": 1643.2, + "valid_targets_min": 868 + }, + { + "epoch": 3.4243697478991595, + "grad_norm": 0.7292997254366875, + "learning_rate": 2.4226871416833683e-05, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.146925687789917, + "step": 3260, + "valid_targets_mean": 1451.4, + "valid_targets_min": 824 + }, + { + "epoch": 3.4296218487394956, + "grad_norm": 0.7542406085374198, + "learning_rate": 2.4175654210248848e-05, + "loss": 0.1404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.161025732755661, + "step": 3265, + "valid_targets_mean": 1615.9, + "valid_targets_min": 797 + }, + { + "epoch": 3.434873949579832, + "grad_norm": 0.6966112598828675, + "learning_rate": 2.4124408355556677e-05, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1535557061433792, + "step": 3270, + "valid_targets_mean": 1596.5, + "valid_targets_min": 1043 + }, + { + "epoch": 3.440126050420168, + "grad_norm": 0.7555059854029672, + "learning_rate": 2.4073134204342014e-05, + "loss": 0.1523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16238391399383545, + "step": 3275, + "valid_targets_mean": 1445.5, + "valid_targets_min": 920 + }, + { + "epoch": 3.4453781512605044, + "grad_norm": 0.8208149984721342, + "learning_rate": 2.4021832108383837e-05, + "loss": 0.1462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15757635235786438, + "step": 3280, + "valid_targets_mean": 1437.9, + "valid_targets_min": 847 + }, + { + "epoch": 3.4506302521008405, + "grad_norm": 0.7056998799882124, + "learning_rate": 2.3970502419652854e-05, + "loss": 0.15, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15004190802574158, + "step": 3285, + "valid_targets_mean": 1559.1, + "valid_targets_min": 557 + }, + { + "epoch": 3.4558823529411766, + "grad_norm": 0.7958553721684484, + "learning_rate": 2.3919145490309076e-05, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15160951018333435, + "step": 3290, + "valid_targets_mean": 1341.6, + "valid_targets_min": 594 + }, + { + "epoch": 3.4611344537815127, + "grad_norm": 0.703390230979893, + "learning_rate": 2.38677616726994e-05, + "loss": 0.1563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1494627594947815, + "step": 3295, + "valid_targets_mean": 1747.5, + "valid_targets_min": 760 + }, + { + "epoch": 3.466386554621849, + "grad_norm": 0.6829115937221362, + "learning_rate": 2.381635131935521e-05, + "loss": 0.2567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27713343501091003, + "step": 3300, + "valid_targets_mean": 3387.2, + "valid_targets_min": 770 + }, + { + "epoch": 3.471638655462185, + "grad_norm": 0.5039365129794023, + "learning_rate": 2.3764914782989926e-05, + "loss": 0.2376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21012486517429352, + "step": 3305, + "valid_targets_mean": 4538.6, + "valid_targets_min": 1799 + }, + { + "epoch": 3.476890756302521, + "grad_norm": 0.5334008441824496, + "learning_rate": 2.3713452416496625e-05, + "loss": 0.2417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2371271550655365, + "step": 3310, + "valid_targets_mean": 4182.4, + "valid_targets_min": 1560 + }, + { + "epoch": 3.482142857142857, + "grad_norm": 0.5252301039129134, + "learning_rate": 2.366196457294558e-05, + "loss": 0.2243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20833566784858704, + "step": 3315, + "valid_targets_mean": 4361.0, + "valid_targets_min": 1681 + }, + { + "epoch": 3.4873949579831933, + "grad_norm": 0.617549050257743, + "learning_rate": 2.3610451605581858e-05, + "loss": 0.2411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2543012499809265, + "step": 3320, + "valid_targets_mean": 3627.6, + "valid_targets_min": 1374 + }, + { + "epoch": 3.4926470588235294, + "grad_norm": 0.6304506600012845, + "learning_rate": 2.3558913867822905e-05, + "loss": 0.2335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22949200868606567, + "step": 3325, + "valid_targets_mean": 3204.2, + "valid_targets_min": 1089 + }, + { + "epoch": 3.4978991596638656, + "grad_norm": 0.5586583847778026, + "learning_rate": 2.3507351713256105e-05, + "loss": 0.2229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23757925629615784, + "step": 3330, + "valid_targets_mean": 4203.4, + "valid_targets_min": 627 + }, + { + "epoch": 3.5031512605042017, + "grad_norm": 0.5579364042848073, + "learning_rate": 2.3455765495636353e-05, + "loss": 0.2318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23478400707244873, + "step": 3335, + "valid_targets_mean": 3778.9, + "valid_targets_min": 1574 + }, + { + "epoch": 3.508403361344538, + "grad_norm": 0.6191817742700788, + "learning_rate": 2.3404155568883643e-05, + "loss": 0.2425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25709354877471924, + "step": 3340, + "valid_targets_mean": 3704.9, + "valid_targets_min": 1921 + }, + { + "epoch": 3.513655462184874, + "grad_norm": 0.587021701284441, + "learning_rate": 2.3352522287080625e-05, + "loss": 0.2314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23616047203540802, + "step": 3345, + "valid_targets_mean": 3960.9, + "valid_targets_min": 1477 + }, + { + "epoch": 3.51890756302521, + "grad_norm": 0.6254607818726631, + "learning_rate": 2.3300866004470182e-05, + "loss": 0.2414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24480389058589935, + "step": 3350, + "valid_targets_mean": 3504.5, + "valid_targets_min": 1499 + }, + { + "epoch": 3.524159663865546, + "grad_norm": 0.5925836663255958, + "learning_rate": 2.324918707545302e-05, + "loss": 0.2396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26201432943344116, + "step": 3355, + "valid_targets_mean": 3803.8, + "valid_targets_min": 1013 + }, + { + "epoch": 3.5294117647058822, + "grad_norm": 0.5642203624546177, + "learning_rate": 2.3197485854585183e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2236984670162201, + "step": 3360, + "valid_targets_mean": 4146.8, + "valid_targets_min": 1543 + }, + { + "epoch": 3.5346638655462184, + "grad_norm": 0.6221598960605759, + "learning_rate": 2.3145762696575675e-05, + "loss": 0.2386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24771781265735626, + "step": 3365, + "valid_targets_mean": 3705.0, + "valid_targets_min": 2050 + }, + { + "epoch": 3.5399159663865545, + "grad_norm": 0.667308661071586, + "learning_rate": 2.3094017956283995e-05, + "loss": 0.2444, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2562797963619232, + "step": 3370, + "valid_targets_mean": 3287.6, + "valid_targets_min": 746 + }, + { + "epoch": 3.5451680672268906, + "grad_norm": 0.6591430962273881, + "learning_rate": 2.3042251988717728e-05, + "loss": 0.234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24168621003627777, + "step": 3375, + "valid_targets_mean": 3219.6, + "valid_targets_min": 1210 + }, + { + "epoch": 3.5504201680672267, + "grad_norm": 0.5861565049376661, + "learning_rate": 2.2990465149030077e-05, + "loss": 0.2408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2576037347316742, + "step": 3380, + "valid_targets_mean": 3420.2, + "valid_targets_min": 1064 + }, + { + "epoch": 3.5556722689075633, + "grad_norm": 0.676375370367313, + "learning_rate": 2.2938657792517446e-05, + "loss": 0.2489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24123093485832214, + "step": 3385, + "valid_targets_mean": 2978.2, + "valid_targets_min": 1173 + }, + { + "epoch": 3.560924369747899, + "grad_norm": 0.7090066415466851, + "learning_rate": 2.2886830274617017e-05, + "loss": 0.2381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2423102855682373, + "step": 3390, + "valid_targets_mean": 3106.8, + "valid_targets_min": 1165 + }, + { + "epoch": 3.5661764705882355, + "grad_norm": 0.6215177878656555, + "learning_rate": 2.2834982950904267e-05, + "loss": 0.2309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20194599032402039, + "step": 3395, + "valid_targets_mean": 3800.3, + "valid_targets_min": 1365 + }, + { + "epoch": 3.571428571428571, + "grad_norm": 0.6568484423092628, + "learning_rate": 2.2783116177090573e-05, + "loss": 0.2369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2180924117565155, + "step": 3400, + "valid_targets_mean": 3020.4, + "valid_targets_min": 1051 + }, + { + "epoch": 3.5766806722689077, + "grad_norm": 0.72432526478877, + "learning_rate": 2.2731230309020762e-05, + "loss": 0.2335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2589854300022125, + "step": 3405, + "valid_targets_mean": 2882.1, + "valid_targets_min": 1345 + }, + { + "epoch": 3.581932773109244, + "grad_norm": 0.647347676568134, + "learning_rate": 2.2679325702670642e-05, + "loss": 0.2387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23592257499694824, + "step": 3410, + "valid_targets_mean": 3123.4, + "valid_targets_min": 1747 + }, + { + "epoch": 3.58718487394958, + "grad_norm": 0.6321313310777158, + "learning_rate": 2.2627402714144586e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23397618532180786, + "step": 3415, + "valid_targets_mean": 3041.9, + "valid_targets_min": 1166 + }, + { + "epoch": 3.592436974789916, + "grad_norm": 0.6163425879015834, + "learning_rate": 2.2575461699673085e-05, + "loss": 0.2353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24911564588546753, + "step": 3420, + "valid_targets_mean": 3629.4, + "valid_targets_min": 1511 + }, + { + "epoch": 3.597689075630252, + "grad_norm": 0.8662060186406194, + "learning_rate": 2.2523503015610303e-05, + "loss": 0.2217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.219038188457489, + "step": 3425, + "valid_targets_mean": 3007.8, + "valid_targets_min": 869 + }, + { + "epoch": 3.6029411764705883, + "grad_norm": 0.6505678950823406, + "learning_rate": 2.2471527018431637e-05, + "loss": 0.2332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2250857800245285, + "step": 3430, + "valid_targets_mean": 3147.4, + "valid_targets_min": 926 + }, + { + "epoch": 3.6081932773109244, + "grad_norm": 0.6624584977841851, + "learning_rate": 2.2419534064731238e-05, + "loss": 0.2359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22663256525993347, + "step": 3435, + "valid_targets_mean": 3047.1, + "valid_targets_min": 1016 + }, + { + "epoch": 3.6134453781512605, + "grad_norm": 0.68180528409829, + "learning_rate": 2.236752451121963e-05, + "loss": 0.2388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2614358067512512, + "step": 3440, + "valid_targets_mean": 3288.3, + "valid_targets_min": 1367 + }, + { + "epoch": 3.6186974789915967, + "grad_norm": 0.5943225731704131, + "learning_rate": 2.2315498714721198e-05, + "loss": 0.2327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2073296308517456, + "step": 3445, + "valid_targets_mean": 3621.2, + "valid_targets_min": 761 + }, + { + "epoch": 3.6239495798319328, + "grad_norm": 0.7143384166342921, + "learning_rate": 2.2263457032171762e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22933131456375122, + "step": 3450, + "valid_targets_mean": 2664.7, + "valid_targets_min": 1480 + }, + { + "epoch": 3.629201680672269, + "grad_norm": 0.6637253904207627, + "learning_rate": 2.2211399820616154e-05, + "loss": 0.232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21584278345108032, + "step": 3455, + "valid_targets_mean": 3043.8, + "valid_targets_min": 1229 + }, + { + "epoch": 3.634453781512605, + "grad_norm": 0.6381313624483769, + "learning_rate": 2.2159327437205733e-05, + "loss": 0.2364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23317265510559082, + "step": 3460, + "valid_targets_mean": 3533.5, + "valid_targets_min": 1727 + }, + { + "epoch": 3.639705882352941, + "grad_norm": 0.6079366483063177, + "learning_rate": 2.210724023919595e-05, + "loss": 0.2337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23725667595863342, + "step": 3465, + "valid_targets_mean": 3763.2, + "valid_targets_min": 1860 + }, + { + "epoch": 3.6449579831932772, + "grad_norm": 0.6617853538852045, + "learning_rate": 2.205513858394389e-05, + "loss": 0.233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23891595005989075, + "step": 3470, + "valid_targets_mean": 3145.0, + "valid_targets_min": 1145 + }, + { + "epoch": 3.6502100840336134, + "grad_norm": 0.6931454910102819, + "learning_rate": 2.2003022828905832e-05, + "loss": 0.2342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24554722011089325, + "step": 3475, + "valid_targets_mean": 3148.6, + "valid_targets_min": 1363 + }, + { + "epoch": 3.6554621848739495, + "grad_norm": 0.6830097683535337, + "learning_rate": 2.19508933316348e-05, + "loss": 0.2227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2083757519721985, + "step": 3480, + "valid_targets_mean": 3347.4, + "valid_targets_min": 1087 + }, + { + "epoch": 3.6607142857142856, + "grad_norm": 0.6284986422875785, + "learning_rate": 2.189875044977808e-05, + "loss": 0.2357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22864796221256256, + "step": 3485, + "valid_targets_mean": 4141.5, + "valid_targets_min": 730 + }, + { + "epoch": 3.6659663865546217, + "grad_norm": 0.6424551917347495, + "learning_rate": 2.1846594541074796e-05, + "loss": 0.2283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21687257289886475, + "step": 3490, + "valid_targets_mean": 3053.2, + "valid_targets_min": 1406 + }, + { + "epoch": 3.671218487394958, + "grad_norm": 0.6118476480500536, + "learning_rate": 2.179442596335345e-05, + "loss": 0.2217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21437662839889526, + "step": 3495, + "valid_targets_mean": 3413.8, + "valid_targets_min": 1833 + }, + { + "epoch": 3.6764705882352944, + "grad_norm": 0.6834904631426817, + "learning_rate": 2.174224507452945e-05, + "loss": 0.2307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23827575147151947, + "step": 3500, + "valid_targets_mean": 3292.0, + "valid_targets_min": 1182 + }, + { + "epoch": 3.68172268907563, + "grad_norm": 0.6540573235032019, + "learning_rate": 2.1690052232602677e-05, + "loss": 0.2306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2561336159706116, + "step": 3505, + "valid_targets_mean": 3815.1, + "valid_targets_min": 1607 + }, + { + "epoch": 3.6869747899159666, + "grad_norm": 0.7038556545538145, + "learning_rate": 2.1637847795655022e-05, + "loss": 0.2357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23643259704113007, + "step": 3510, + "valid_targets_mean": 2850.3, + "valid_targets_min": 1358 + }, + { + "epoch": 3.6922268907563023, + "grad_norm": 0.6142418362337423, + "learning_rate": 2.1585632121847925e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2268371880054474, + "step": 3515, + "valid_targets_mean": 3897.3, + "valid_targets_min": 1757 + }, + { + "epoch": 3.697478991596639, + "grad_norm": 0.6018277229359115, + "learning_rate": 2.153340556941991e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19471925497055054, + "step": 3520, + "valid_targets_mean": 3365.7, + "valid_targets_min": 1170 + }, + { + "epoch": 3.7027310924369745, + "grad_norm": 0.7817444103413864, + "learning_rate": 2.1481168496684145e-05, + "loss": 0.224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24332907795906067, + "step": 3525, + "valid_targets_mean": 2908.3, + "valid_targets_min": 474 + }, + { + "epoch": 3.707983193277311, + "grad_norm": 0.6733856863461328, + "learning_rate": 2.142892126202597e-05, + "loss": 0.2271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22530926764011383, + "step": 3530, + "valid_targets_mean": 3249.9, + "valid_targets_min": 1140 + }, + { + "epoch": 3.713235294117647, + "grad_norm": 0.6772781803626257, + "learning_rate": 2.137666422390045e-05, + "loss": 0.2233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2468833178281784, + "step": 3535, + "valid_targets_mean": 3621.2, + "valid_targets_min": 2266 + }, + { + "epoch": 3.7184873949579833, + "grad_norm": 0.596320718539212, + "learning_rate": 2.132439774082991e-05, + "loss": 0.2084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19030636548995972, + "step": 3540, + "valid_targets_mean": 3192.4, + "valid_targets_min": 809 + }, + { + "epoch": 3.7237394957983194, + "grad_norm": 0.6683616408438122, + "learning_rate": 2.1272122171401467e-05, + "loss": 0.24, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2368716150522232, + "step": 3545, + "valid_targets_mean": 3176.4, + "valid_targets_min": 1059 + }, + { + "epoch": 3.7289915966386555, + "grad_norm": 0.5993575332277289, + "learning_rate": 2.1219837874264573e-05, + "loss": 0.2217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19735181331634521, + "step": 3550, + "valid_targets_mean": 3426.4, + "valid_targets_min": 964 + }, + { + "epoch": 3.7342436974789917, + "grad_norm": 0.6105324798756142, + "learning_rate": 2.116754520812857e-05, + "loss": 0.2313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22294853627681732, + "step": 3555, + "valid_targets_mean": 3946.7, + "valid_targets_min": 1105 + }, + { + "epoch": 3.7394957983193278, + "grad_norm": 0.7317392218199291, + "learning_rate": 2.111524453176022e-05, + "loss": 0.2324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23324161767959595, + "step": 3560, + "valid_targets_mean": 2534.7, + "valid_targets_min": 1460 + }, + { + "epoch": 3.744747899159664, + "grad_norm": 0.679940437130019, + "learning_rate": 2.1062936203981227e-05, + "loss": 0.243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23880356550216675, + "step": 3565, + "valid_targets_mean": 3673.0, + "valid_targets_min": 1091 + }, + { + "epoch": 3.75, + "grad_norm": 0.6386888401169865, + "learning_rate": 2.10106205836658e-05, + "loss": 0.2229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21099084615707397, + "step": 3570, + "valid_targets_mean": 3355.7, + "valid_targets_min": 1771 + }, + { + "epoch": 3.755252100840336, + "grad_norm": 0.609185135460291, + "learning_rate": 2.095829802973817e-05, + "loss": 0.2207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20622901618480682, + "step": 3575, + "valid_targets_mean": 3403.8, + "valid_targets_min": 1625 + }, + { + "epoch": 3.7605042016806722, + "grad_norm": 0.6933994920811911, + "learning_rate": 2.090596890117016e-05, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20166133344173431, + "step": 3580, + "valid_targets_mean": 3379.6, + "valid_targets_min": 1180 + }, + { + "epoch": 3.7657563025210083, + "grad_norm": 0.7001758773355745, + "learning_rate": 2.0853633556978674e-05, + "loss": 0.2214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2164371758699417, + "step": 3585, + "valid_targets_mean": 3085.8, + "valid_targets_min": 1460 + }, + { + "epoch": 3.7710084033613445, + "grad_norm": 0.6982893325743658, + "learning_rate": 2.0801292356223265e-05, + "loss": 0.2245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22557231783866882, + "step": 3590, + "valid_targets_mean": 3090.9, + "valid_targets_min": 1079 + }, + { + "epoch": 3.7762605042016806, + "grad_norm": 0.6862429302576437, + "learning_rate": 2.0748945658003695e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21564340591430664, + "step": 3595, + "valid_targets_mean": 2751.7, + "valid_targets_min": 1143 + }, + { + "epoch": 3.7815126050420167, + "grad_norm": 0.7486537584683461, + "learning_rate": 2.0696593821457392e-05, + "loss": 0.222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2403099536895752, + "step": 3600, + "valid_targets_mean": 3262.4, + "valid_targets_min": 914 + }, + { + "epoch": 3.786764705882353, + "grad_norm": 0.6292490113807976, + "learning_rate": 2.064423720575709e-05, + "loss": 0.2365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21284765005111694, + "step": 3605, + "valid_targets_mean": 3452.1, + "valid_targets_min": 1312 + }, + { + "epoch": 3.792016806722689, + "grad_norm": 0.6731713864780883, + "learning_rate": 2.0591876170108257e-05, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23124480247497559, + "step": 3610, + "valid_targets_mean": 2805.6, + "valid_targets_min": 987 + }, + { + "epoch": 3.7972689075630255, + "grad_norm": 0.570026508038886, + "learning_rate": 2.0539511073746743e-05, + "loss": 0.2224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21886953711509705, + "step": 3615, + "valid_targets_mean": 4512.7, + "valid_targets_min": 1501 + }, + { + "epoch": 3.802521008403361, + "grad_norm": 0.6587414192570736, + "learning_rate": 2.0487142275936213e-05, + "loss": 0.2218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21692374348640442, + "step": 3620, + "valid_targets_mean": 3188.2, + "valid_targets_min": 1388 + }, + { + "epoch": 3.8077731092436977, + "grad_norm": 0.6975757464948207, + "learning_rate": 2.0434770135965743e-05, + "loss": 0.238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23684664070606232, + "step": 3625, + "valid_targets_mean": 3106.9, + "valid_targets_min": 1986 + }, + { + "epoch": 3.8130252100840334, + "grad_norm": 0.6390716602559315, + "learning_rate": 2.0382395013147347e-05, + "loss": 0.213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23830687999725342, + "step": 3630, + "valid_targets_mean": 3893.1, + "valid_targets_min": 1671 + }, + { + "epoch": 3.81827731092437, + "grad_norm": 0.6520611936803277, + "learning_rate": 2.033001726681349e-05, + "loss": 0.2184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19424161314964294, + "step": 3635, + "valid_targets_mean": 3246.8, + "valid_targets_min": 512 + }, + { + "epoch": 3.8235294117647056, + "grad_norm": 0.653877417451473, + "learning_rate": 2.027763725631463e-05, + "loss": 0.2257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21053674817085266, + "step": 3640, + "valid_targets_mean": 3285.3, + "valid_targets_min": 1202 + }, + { + "epoch": 3.828781512605042, + "grad_norm": 0.6519721595841632, + "learning_rate": 2.0225255341016794e-05, + "loss": 0.2395, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22552523016929626, + "step": 3645, + "valid_targets_mean": 4012.8, + "valid_targets_min": 1645 + }, + { + "epoch": 3.8340336134453783, + "grad_norm": 0.6008225197326998, + "learning_rate": 2.017287188029904e-05, + "loss": 0.226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20129550993442535, + "step": 3650, + "valid_targets_mean": 3525.8, + "valid_targets_min": 1434 + }, + { + "epoch": 3.8392857142857144, + "grad_norm": 0.7071580358913306, + "learning_rate": 2.0120487233551035e-05, + "loss": 0.2369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2568947672843933, + "step": 3655, + "valid_targets_mean": 3044.6, + "valid_targets_min": 1055 + }, + { + "epoch": 3.8445378151260505, + "grad_norm": 0.7176723110707379, + "learning_rate": 2.006810176017059e-05, + "loss": 0.2279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23161205649375916, + "step": 3660, + "valid_targets_mean": 3179.8, + "valid_targets_min": 1625 + }, + { + "epoch": 3.8497899159663866, + "grad_norm": 0.6943499018659017, + "learning_rate": 2.0015715819561205e-05, + "loss": 0.2328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24318306148052216, + "step": 3665, + "valid_targets_mean": 3399.9, + "valid_targets_min": 1147 + }, + { + "epoch": 3.8550420168067228, + "grad_norm": 0.6973815528618685, + "learning_rate": 1.9963329771129558e-05, + "loss": 0.2401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24660438299179077, + "step": 3670, + "valid_targets_mean": 2866.4, + "valid_targets_min": 998 + }, + { + "epoch": 3.860294117647059, + "grad_norm": 0.6342688826905563, + "learning_rate": 1.9910943974283066e-05, + "loss": 0.2331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23309576511383057, + "step": 3675, + "valid_targets_mean": 3480.4, + "valid_targets_min": 1078 + }, + { + "epoch": 3.865546218487395, + "grad_norm": 0.735979206541467, + "learning_rate": 1.9858558788427447e-05, + "loss": 0.2294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22612996399402618, + "step": 3680, + "valid_targets_mean": 2920.8, + "valid_targets_min": 1005 + }, + { + "epoch": 3.870798319327731, + "grad_norm": 0.6974844461519691, + "learning_rate": 1.9806174572964205e-05, + "loss": 0.2349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22539496421813965, + "step": 3685, + "valid_targets_mean": 2983.9, + "valid_targets_min": 719 + }, + { + "epoch": 3.8760504201680672, + "grad_norm": 0.7806658133125973, + "learning_rate": 1.975379168728819e-05, + "loss": 0.2367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25455334782600403, + "step": 3690, + "valid_targets_mean": 2842.1, + "valid_targets_min": 617 + }, + { + "epoch": 3.8813025210084033, + "grad_norm": 0.653382148721513, + "learning_rate": 1.9701410490785128e-05, + "loss": 0.2215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22338590025901794, + "step": 3695, + "valid_targets_mean": 3210.6, + "valid_targets_min": 1665 + }, + { + "epoch": 3.8865546218487395, + "grad_norm": 0.6977766979229648, + "learning_rate": 1.9649031342829178e-05, + "loss": 0.2357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24201029539108276, + "step": 3700, + "valid_targets_mean": 3020.0, + "valid_targets_min": 1076 + }, + { + "epoch": 3.8918067226890756, + "grad_norm": 0.6441901232419098, + "learning_rate": 1.9596654602780398e-05, + "loss": 0.2217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2265818864107132, + "step": 3705, + "valid_targets_mean": 3548.8, + "valid_targets_min": 790 + }, + { + "epoch": 3.8970588235294117, + "grad_norm": 0.715226965706479, + "learning_rate": 1.9544280629982364e-05, + "loss": 0.2428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25313618779182434, + "step": 3710, + "valid_targets_mean": 3038.7, + "valid_targets_min": 937 + }, + { + "epoch": 3.902310924369748, + "grad_norm": 0.6360683211422588, + "learning_rate": 1.949190978375966e-05, + "loss": 0.2366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22172778844833374, + "step": 3715, + "valid_targets_mean": 3616.2, + "valid_targets_min": 981 + }, + { + "epoch": 3.907563025210084, + "grad_norm": 0.6393776390642489, + "learning_rate": 1.9439542423415413e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.205665722489357, + "step": 3720, + "valid_targets_mean": 3152.5, + "valid_targets_min": 1351 + }, + { + "epoch": 3.91281512605042, + "grad_norm": 0.6380081375418487, + "learning_rate": 1.9387178908228838e-05, + "loss": 0.2292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23773056268692017, + "step": 3725, + "valid_targets_mean": 4153.7, + "valid_targets_min": 1307 + }, + { + "epoch": 3.918067226890756, + "grad_norm": 0.5502610030856153, + "learning_rate": 1.933481959745276e-05, + "loss": 0.225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2143445611000061, + "step": 3730, + "valid_targets_mean": 5039.1, + "valid_targets_min": 1587 + }, + { + "epoch": 3.9233193277310923, + "grad_norm": 0.5822473871522699, + "learning_rate": 1.9282464850311184e-05, + "loss": 0.2179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20977917313575745, + "step": 3735, + "valid_targets_mean": 4101.8, + "valid_targets_min": 1741 + }, + { + "epoch": 3.928571428571429, + "grad_norm": 0.5793820699573822, + "learning_rate": 1.9230115025996764e-05, + "loss": 0.2199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20586253702640533, + "step": 3740, + "valid_targets_mean": 3830.2, + "valid_targets_min": 856 + }, + { + "epoch": 3.9338235294117645, + "grad_norm": 0.7167408375172564, + "learning_rate": 1.9177770483668416e-05, + "loss": 0.2302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22084660828113556, + "step": 3745, + "valid_targets_mean": 3468.1, + "valid_targets_min": 1886 + }, + { + "epoch": 3.939075630252101, + "grad_norm": 0.6404425612951213, + "learning_rate": 1.912543158244881e-05, + "loss": 0.2174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2383722960948944, + "step": 3750, + "valid_targets_mean": 3603.1, + "valid_targets_min": 832 + }, + { + "epoch": 3.9443277310924367, + "grad_norm": 0.6506827435964047, + "learning_rate": 1.9073098681421895e-05, + "loss": 0.2069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19391639530658722, + "step": 3755, + "valid_targets_mean": 3134.8, + "valid_targets_min": 1484 + }, + { + "epoch": 3.9495798319327733, + "grad_norm": 0.6963957310682597, + "learning_rate": 1.902077213963048e-05, + "loss": 0.2203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21799013018608093, + "step": 3760, + "valid_targets_mean": 3377.2, + "valid_targets_min": 1152 + }, + { + "epoch": 3.9548319327731094, + "grad_norm": 0.7783983083913816, + "learning_rate": 1.896845231607372e-05, + "loss": 0.238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.27412140369415283, + "step": 3765, + "valid_targets_mean": 2750.6, + "valid_targets_min": 1128 + }, + { + "epoch": 3.9600840336134455, + "grad_norm": 0.6595123518563015, + "learning_rate": 1.8916139569704704e-05, + "loss": 0.2292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2191227525472641, + "step": 3770, + "valid_targets_mean": 3052.4, + "valid_targets_min": 1156 + }, + { + "epoch": 3.9653361344537816, + "grad_norm": 0.6850818119709559, + "learning_rate": 1.886383425942795e-05, + "loss": 0.2229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21025024354457855, + "step": 3775, + "valid_targets_mean": 3002.1, + "valid_targets_min": 993 + }, + { + "epoch": 3.9705882352941178, + "grad_norm": 0.7391622979048171, + "learning_rate": 1.8811536744096956e-05, + "loss": 0.2154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22044730186462402, + "step": 3780, + "valid_targets_mean": 2914.0, + "valid_targets_min": 1275 + }, + { + "epoch": 3.975840336134454, + "grad_norm": 0.7238761932845064, + "learning_rate": 1.8759247382511748e-05, + "loss": 0.2242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2314056158065796, + "step": 3785, + "valid_targets_mean": 3362.9, + "valid_targets_min": 1669 + }, + { + "epoch": 3.98109243697479, + "grad_norm": 0.6629448508084417, + "learning_rate": 1.8706966533416413e-05, + "loss": 0.2281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2214796096086502, + "step": 3790, + "valid_targets_mean": 3065.4, + "valid_targets_min": 1233 + }, + { + "epoch": 3.986344537815126, + "grad_norm": 0.957526256499955, + "learning_rate": 1.8654694555496624e-05, + "loss": 0.2292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21655909717082977, + "step": 3795, + "valid_targets_mean": 3591.1, + "valid_targets_min": 1685 + }, + { + "epoch": 3.991596638655462, + "grad_norm": 0.6385611154020212, + "learning_rate": 1.860243180737721e-05, + "loss": 0.2315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23146796226501465, + "step": 3800, + "valid_targets_mean": 3571.6, + "valid_targets_min": 1793 + }, + { + "epoch": 3.9968487394957983, + "grad_norm": 0.6361743923463639, + "learning_rate": 1.8550178647619664e-05, + "loss": 0.2344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23826557397842407, + "step": 3805, + "valid_targets_mean": 3868.6, + "valid_targets_min": 1976 + }, + { + "epoch": 4.0021008403361344, + "grad_norm": 0.9276215598529717, + "learning_rate": 1.84979354347197e-05, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17189082503318787, + "step": 3810, + "valid_targets_mean": 1553.1, + "valid_targets_min": 777 + }, + { + "epoch": 4.007352941176471, + "grad_norm": 0.866049721179808, + "learning_rate": 1.8445702527104782e-05, + "loss": 0.1654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1560104489326477, + "step": 3815, + "valid_targets_mean": 1315.5, + "valid_targets_min": 705 + }, + { + "epoch": 4.012605042016807, + "grad_norm": 0.7548114470500479, + "learning_rate": 1.8393480283131677e-05, + "loss": 0.1496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1479974240064621, + "step": 3820, + "valid_targets_mean": 1626.6, + "valid_targets_min": 906 + }, + { + "epoch": 4.017857142857143, + "grad_norm": 0.8337315927989255, + "learning_rate": 1.8341269061083996e-05, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12990880012512207, + "step": 3825, + "valid_targets_mean": 1316.2, + "valid_targets_min": 680 + }, + { + "epoch": 4.023109243697479, + "grad_norm": 0.8291733641726188, + "learning_rate": 1.8289069219169717e-05, + "loss": 0.1484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15963760018348694, + "step": 3830, + "valid_targets_mean": 1752.4, + "valid_targets_min": 819 + }, + { + "epoch": 4.0283613445378155, + "grad_norm": 0.8304874985400824, + "learning_rate": 1.8236881115518766e-05, + "loss": 0.1498, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14934542775154114, + "step": 3835, + "valid_targets_mean": 1383.9, + "valid_targets_min": 734 + }, + { + "epoch": 4.033613445378151, + "grad_norm": 0.8052129046929378, + "learning_rate": 1.8184705108180516e-05, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15371397137641907, + "step": 3840, + "valid_targets_mean": 1573.9, + "valid_targets_min": 930 + }, + { + "epoch": 4.038865546218488, + "grad_norm": 0.8085904112301928, + "learning_rate": 1.8132541555121353e-05, + "loss": 0.1364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14829567074775696, + "step": 3845, + "valid_targets_mean": 1383.1, + "valid_targets_min": 737 + }, + { + "epoch": 4.044117647058823, + "grad_norm": 0.8018131771688088, + "learning_rate": 1.808039081422223e-05, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1333727240562439, + "step": 3850, + "valid_targets_mean": 1518.7, + "valid_targets_min": 759 + }, + { + "epoch": 4.04936974789916, + "grad_norm": 0.8491004069838234, + "learning_rate": 1.802825324327618e-05, + "loss": 0.1447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15288317203521729, + "step": 3855, + "valid_targets_mean": 1489.7, + "valid_targets_min": 753 + }, + { + "epoch": 4.054621848739496, + "grad_norm": 0.8165835733627114, + "learning_rate": 1.7976129199985886e-05, + "loss": 0.147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1399635225534439, + "step": 3860, + "valid_targets_mean": 1304.5, + "valid_targets_min": 572 + }, + { + "epoch": 4.059873949579832, + "grad_norm": 0.7748737501704929, + "learning_rate": 1.7924019041961228e-05, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14403750002384186, + "step": 3865, + "valid_targets_mean": 1409.2, + "valid_targets_min": 679 + }, + { + "epoch": 4.065126050420168, + "grad_norm": 0.8803098411860152, + "learning_rate": 1.7871923126716827e-05, + "loss": 0.1501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15918989479541779, + "step": 3870, + "valid_targets_mean": 1331.1, + "valid_targets_min": 771 + }, + { + "epoch": 4.070378151260504, + "grad_norm": 0.753689435569324, + "learning_rate": 1.7819841811669573e-05, + "loss": 0.1342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13107150793075562, + "step": 3875, + "valid_targets_mean": 1473.2, + "valid_targets_min": 842 + }, + { + "epoch": 4.07563025210084, + "grad_norm": 0.7637352369796083, + "learning_rate": 1.7767775454136194e-05, + "loss": 0.1367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12565407156944275, + "step": 3880, + "valid_targets_mean": 1367.2, + "valid_targets_min": 822 + }, + { + "epoch": 4.080882352941177, + "grad_norm": 0.8931529994254264, + "learning_rate": 1.7715724411330806e-05, + "loss": 0.1381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1554366648197174, + "step": 3885, + "valid_targets_mean": 1507.5, + "valid_targets_min": 675 + }, + { + "epoch": 4.086134453781512, + "grad_norm": 0.8058581428686146, + "learning_rate": 1.7663689040362446e-05, + "loss": 0.1444, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14042530953884125, + "step": 3890, + "valid_targets_mean": 1363.7, + "valid_targets_min": 892 + }, + { + "epoch": 4.091386554621849, + "grad_norm": 0.8507202166895215, + "learning_rate": 1.761166969823262e-05, + "loss": 0.1453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13831481337547302, + "step": 3895, + "valid_targets_mean": 1315.4, + "valid_targets_min": 656 + }, + { + "epoch": 4.0966386554621845, + "grad_norm": 0.7947736774793486, + "learning_rate": 1.7559666741832896e-05, + "loss": 0.1474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12960849702358246, + "step": 3900, + "valid_targets_mean": 1341.5, + "valid_targets_min": 681 + }, + { + "epoch": 4.101890756302521, + "grad_norm": 0.8251862770713435, + "learning_rate": 1.750768052794239e-05, + "loss": 0.1514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15292903780937195, + "step": 3905, + "valid_targets_mean": 1436.5, + "valid_targets_min": 548 + }, + { + "epoch": 4.107142857142857, + "grad_norm": 0.799156965792982, + "learning_rate": 1.7455711413225372e-05, + "loss": 0.142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13684728741645813, + "step": 3910, + "valid_targets_mean": 1338.8, + "valid_targets_min": 658 + }, + { + "epoch": 4.112394957983193, + "grad_norm": 0.8117329071283277, + "learning_rate": 1.7403759754228794e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13306397199630737, + "step": 3915, + "valid_targets_mean": 1343.0, + "valid_targets_min": 650 + }, + { + "epoch": 4.117647058823529, + "grad_norm": 0.8678892755082293, + "learning_rate": 1.7351825907379837e-05, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1487090289592743, + "step": 3920, + "valid_targets_mean": 1398.7, + "valid_targets_min": 734 + }, + { + "epoch": 4.1228991596638656, + "grad_norm": 0.8732015715863755, + "learning_rate": 1.7299910228983513e-05, + "loss": 0.1317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14107802510261536, + "step": 3925, + "valid_targets_mean": 1339.1, + "valid_targets_min": 704 + }, + { + "epoch": 4.128151260504202, + "grad_norm": 0.8358116761683432, + "learning_rate": 1.7248013075220128e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13086049258708954, + "step": 3930, + "valid_targets_mean": 1377.6, + "valid_targets_min": 686 + }, + { + "epoch": 4.133403361344538, + "grad_norm": 0.882675069386114, + "learning_rate": 1.7196134802142945e-05, + "loss": 0.1393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1546640843153, + "step": 3935, + "valid_targets_mean": 1540.8, + "valid_targets_min": 697 + }, + { + "epoch": 4.138655462184874, + "grad_norm": 0.7673275063053835, + "learning_rate": 1.7144275765675673e-05, + "loss": 0.1388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13223353028297424, + "step": 3940, + "valid_targets_mean": 1435.7, + "valid_targets_min": 779 + }, + { + "epoch": 4.14390756302521, + "grad_norm": 0.8161750428827426, + "learning_rate": 1.7092436321610042e-05, + "loss": 0.1482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.147655189037323, + "step": 3945, + "valid_targets_mean": 1520.5, + "valid_targets_min": 729 + }, + { + "epoch": 4.149159663865547, + "grad_norm": 0.8361768379372263, + "learning_rate": 1.704061682560336e-05, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13745343685150146, + "step": 3950, + "valid_targets_mean": 1408.2, + "valid_targets_min": 661 + }, + { + "epoch": 4.154411764705882, + "grad_norm": 0.8345898800935504, + "learning_rate": 1.698881763317609e-05, + "loss": 0.1438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13992251455783844, + "step": 3955, + "valid_targets_mean": 1531.4, + "valid_targets_min": 847 + }, + { + "epoch": 4.159663865546219, + "grad_norm": 0.7875946022995587, + "learning_rate": 1.69370390997094e-05, + "loss": 0.1387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13176432251930237, + "step": 3960, + "valid_targets_mean": 1391.6, + "valid_targets_min": 758 + }, + { + "epoch": 4.1649159663865545, + "grad_norm": 0.7017178626610292, + "learning_rate": 1.6885281580442677e-05, + "loss": 0.142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.114842988550663, + "step": 3965, + "valid_targets_mean": 1312.7, + "valid_targets_min": 800 + }, + { + "epoch": 4.170168067226891, + "grad_norm": 0.7974602402640565, + "learning_rate": 1.6833545430471194e-05, + "loss": 0.1374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1272060126066208, + "step": 3970, + "valid_targets_mean": 1347.1, + "valid_targets_min": 803 + }, + { + "epoch": 4.175420168067227, + "grad_norm": 0.8100956763693268, + "learning_rate": 1.678183100474359e-05, + "loss": 0.1378, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13280612230300903, + "step": 3975, + "valid_targets_mean": 1394.2, + "valid_targets_min": 749 + }, + { + "epoch": 4.180672268907563, + "grad_norm": 0.8839233586552578, + "learning_rate": 1.6730138658059448e-05, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13839851319789886, + "step": 3980, + "valid_targets_mean": 1372.1, + "valid_targets_min": 837 + }, + { + "epoch": 4.185924369747899, + "grad_norm": 0.7770716555066293, + "learning_rate": 1.667846874506689e-05, + "loss": 0.1345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12787684798240662, + "step": 3985, + "valid_targets_mean": 1425.2, + "valid_targets_min": 719 + }, + { + "epoch": 4.1911764705882355, + "grad_norm": 0.8263514960899874, + "learning_rate": 1.662682162026012e-05, + "loss": 0.1384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14100591838359833, + "step": 3990, + "valid_targets_mean": 1423.1, + "valid_targets_min": 765 + }, + { + "epoch": 4.196428571428571, + "grad_norm": 0.8174867184649295, + "learning_rate": 1.6575197637976995e-05, + "loss": 0.1377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14689315855503082, + "step": 3995, + "valid_targets_mean": 1509.8, + "valid_targets_min": 875 + }, + { + "epoch": 4.201680672268908, + "grad_norm": 0.8589762302363031, + "learning_rate": 1.652359715239661e-05, + "loss": 0.1368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13010036945343018, + "step": 4000, + "valid_targets_mean": 1292.4, + "valid_targets_min": 655 + }, + { + "epoch": 4.206932773109243, + "grad_norm": 0.8425688419243291, + "learning_rate": 1.6472020517536828e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1463192105293274, + "step": 4005, + "valid_targets_mean": 1474.6, + "valid_targets_min": 669 + }, + { + "epoch": 4.21218487394958, + "grad_norm": 0.8171247963907109, + "learning_rate": 1.6420468087251907e-05, + "loss": 0.1321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13386958837509155, + "step": 4010, + "valid_targets_mean": 1377.2, + "valid_targets_min": 1046 + }, + { + "epoch": 4.217436974789916, + "grad_norm": 0.8221524193856137, + "learning_rate": 1.6368940215230026e-05, + "loss": 0.1398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14860522747039795, + "step": 4015, + "valid_targets_mean": 1693.8, + "valid_targets_min": 703 + }, + { + "epoch": 4.222689075630252, + "grad_norm": 0.8080758516493216, + "learning_rate": 1.6317437254990875e-05, + "loss": 0.1413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13878750801086426, + "step": 4020, + "valid_targets_mean": 1473.4, + "valid_targets_min": 846 + }, + { + "epoch": 4.227941176470588, + "grad_norm": 0.8329372613650491, + "learning_rate": 1.626595955988325e-05, + "loss": 0.1445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14070574939250946, + "step": 4025, + "valid_targets_mean": 1439.1, + "valid_targets_min": 764 + }, + { + "epoch": 4.233193277310924, + "grad_norm": 0.7819540997359885, + "learning_rate": 1.6214507483082593e-05, + "loss": 0.1315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13133744895458221, + "step": 4030, + "valid_targets_mean": 1577.8, + "valid_targets_min": 822 + }, + { + "epoch": 4.23844537815126, + "grad_norm": 0.7658805060914474, + "learning_rate": 1.6163081377588588e-05, + "loss": 0.131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13184303045272827, + "step": 4035, + "valid_targets_mean": 1500.7, + "valid_targets_min": 982 + }, + { + "epoch": 4.243697478991597, + "grad_norm": 0.7854703396401594, + "learning_rate": 1.6111681596222732e-05, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13113921880722046, + "step": 4040, + "valid_targets_mean": 1402.2, + "valid_targets_min": 713 + }, + { + "epoch": 4.248949579831933, + "grad_norm": 0.765545221207553, + "learning_rate": 1.606030849162593e-05, + "loss": 0.1336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1270998865365982, + "step": 4045, + "valid_targets_mean": 1522.0, + "valid_targets_min": 721 + }, + { + "epoch": 4.254201680672269, + "grad_norm": 0.80112535205176, + "learning_rate": 1.600896241625605e-05, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13291269540786743, + "step": 4050, + "valid_targets_mean": 1373.8, + "valid_targets_min": 741 + }, + { + "epoch": 4.2594537815126055, + "grad_norm": 0.8177224842495399, + "learning_rate": 1.5957643722385532e-05, + "loss": 0.132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14079031348228455, + "step": 4055, + "valid_targets_mean": 1517.9, + "valid_targets_min": 734 + }, + { + "epoch": 4.264705882352941, + "grad_norm": 0.795808535657422, + "learning_rate": 1.590635276209896e-05, + "loss": 0.1396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13821426033973694, + "step": 4060, + "valid_targets_mean": 1521.1, + "valid_targets_min": 744 + }, + { + "epoch": 4.269957983193278, + "grad_norm": 0.8258679101265551, + "learning_rate": 1.5855089887290634e-05, + "loss": 0.1391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14685073494911194, + "step": 4065, + "valid_targets_mean": 1565.1, + "valid_targets_min": 620 + }, + { + "epoch": 4.275210084033613, + "grad_norm": 0.8117779327857177, + "learning_rate": 1.5803855449662175e-05, + "loss": 0.1323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13573217391967773, + "step": 4070, + "valid_targets_mean": 1411.9, + "valid_targets_min": 622 + }, + { + "epoch": 4.28046218487395, + "grad_norm": 0.7103133306723244, + "learning_rate": 1.5752649800720096e-05, + "loss": 0.1355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14503194391727448, + "step": 4075, + "valid_targets_mean": 1797.2, + "valid_targets_min": 740 + }, + { + "epoch": 4.285714285714286, + "grad_norm": 0.7877525685043663, + "learning_rate": 1.5701473291773395e-05, + "loss": 0.1303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12483812868595123, + "step": 4080, + "valid_targets_mean": 1316.6, + "valid_targets_min": 694 + }, + { + "epoch": 4.290966386554622, + "grad_norm": 0.7744871393889592, + "learning_rate": 1.5650326273931166e-05, + "loss": 0.137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13510701060295105, + "step": 4085, + "valid_targets_mean": 1644.8, + "valid_targets_min": 902 + }, + { + "epoch": 4.296218487394958, + "grad_norm": 0.7786406479142675, + "learning_rate": 1.559920909810016e-05, + "loss": 0.1311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13013191521167755, + "step": 4090, + "valid_targets_mean": 1388.4, + "valid_targets_min": 674 + }, + { + "epoch": 4.301470588235294, + "grad_norm": 0.7764860246380487, + "learning_rate": 1.5548122114982393e-05, + "loss": 0.1272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12222108244895935, + "step": 4095, + "valid_targets_mean": 1277.5, + "valid_targets_min": 661 + }, + { + "epoch": 4.30672268907563, + "grad_norm": 0.8106042463016419, + "learning_rate": 1.5497065675072728e-05, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14016015827655792, + "step": 4100, + "valid_targets_mean": 1502.6, + "valid_targets_min": 912 + }, + { + "epoch": 4.311974789915967, + "grad_norm": 0.8543220451468436, + "learning_rate": 1.5446040128656485e-05, + "loss": 0.1403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14713945984840393, + "step": 4105, + "valid_targets_mean": 1561.8, + "valid_targets_min": 1181 + }, + { + "epoch": 4.317226890756302, + "grad_norm": 0.7953069524968457, + "learning_rate": 1.539504582580704e-05, + "loss": 0.1292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12721598148345947, + "step": 4110, + "valid_targets_mean": 1307.5, + "valid_targets_min": 773 + }, + { + "epoch": 4.322478991596639, + "grad_norm": 0.8097826390836019, + "learning_rate": 1.5344083116383395e-05, + "loss": 0.1339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12812337279319763, + "step": 4115, + "valid_targets_mean": 1381.6, + "valid_targets_min": 679 + }, + { + "epoch": 4.3277310924369745, + "grad_norm": 0.7958890094736837, + "learning_rate": 1.529315235002781e-05, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14337149262428284, + "step": 4120, + "valid_targets_mean": 1599.2, + "valid_targets_min": 818 + }, + { + "epoch": 4.332983193277311, + "grad_norm": 0.783528991087372, + "learning_rate": 1.5242253876163399e-05, + "loss": 0.1351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12217546254396439, + "step": 4125, + "valid_targets_mean": 1385.0, + "valid_targets_min": 635 + }, + { + "epoch": 4.338235294117647, + "grad_norm": 0.8168471451852063, + "learning_rate": 1.5191388043991712e-05, + "loss": 0.1322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12848719954490662, + "step": 4130, + "valid_targets_mean": 1420.3, + "valid_targets_min": 592 + }, + { + "epoch": 4.343487394957983, + "grad_norm": 0.8687088982404517, + "learning_rate": 1.5140555202490359e-05, + "loss": 0.1335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1381555199623108, + "step": 4135, + "valid_targets_mean": 1435.8, + "valid_targets_min": 662 + }, + { + "epoch": 4.348739495798319, + "grad_norm": 0.824701387692817, + "learning_rate": 1.5089755700410602e-05, + "loss": 0.1358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13314849138259888, + "step": 4140, + "valid_targets_mean": 1477.9, + "valid_targets_min": 622 + }, + { + "epoch": 4.3539915966386555, + "grad_norm": 1.1187004100398168, + "learning_rate": 1.5038989886274992e-05, + "loss": 0.1372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1633179783821106, + "step": 4145, + "valid_targets_mean": 1382.3, + "valid_targets_min": 625 + }, + { + "epoch": 4.359243697478991, + "grad_norm": 0.843059495998902, + "learning_rate": 1.498825810837492e-05, + "loss": 0.1354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12570440769195557, + "step": 4150, + "valid_targets_mean": 1461.6, + "valid_targets_min": 888 + }, + { + "epoch": 4.364495798319328, + "grad_norm": 0.816554096771243, + "learning_rate": 1.4937560714768283e-05, + "loss": 0.1385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1403723806142807, + "step": 4155, + "valid_targets_mean": 1461.2, + "valid_targets_min": 635 + }, + { + "epoch": 4.369747899159664, + "grad_norm": 0.8797104989098917, + "learning_rate": 1.4886898053277086e-05, + "loss": 0.1366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15383698046207428, + "step": 4160, + "valid_targets_mean": 1542.3, + "valid_targets_min": 522 + }, + { + "epoch": 4.375, + "grad_norm": 0.7898626242305901, + "learning_rate": 1.4836270471485029e-05, + "loss": 0.1326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12376511842012405, + "step": 4165, + "valid_targets_mean": 1244.7, + "valid_targets_min": 697 + }, + { + "epoch": 4.380252100840336, + "grad_norm": 0.8780197054651814, + "learning_rate": 1.4785678316735144e-05, + "loss": 0.1308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14664515852928162, + "step": 4170, + "valid_targets_mean": 1673.9, + "valid_targets_min": 730 + }, + { + "epoch": 4.385504201680672, + "grad_norm": 0.9346525477732902, + "learning_rate": 1.4735121936127406e-05, + "loss": 0.1321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15194876492023468, + "step": 4175, + "valid_targets_mean": 1695.9, + "valid_targets_min": 719 + }, + { + "epoch": 4.390756302521009, + "grad_norm": 0.9211785858456039, + "learning_rate": 1.4684601676516366e-05, + "loss": 0.1318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.138489231467247, + "step": 4180, + "valid_targets_mean": 1400.1, + "valid_targets_min": 745 + }, + { + "epoch": 4.3960084033613445, + "grad_norm": 0.8373085168906266, + "learning_rate": 1.4634117884508738e-05, + "loss": 0.1274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12678782641887665, + "step": 4185, + "valid_targets_mean": 1299.4, + "valid_targets_min": 680 + }, + { + "epoch": 4.401260504201681, + "grad_norm": 0.8325212171741512, + "learning_rate": 1.458367090646105e-05, + "loss": 0.1278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1405733823776245, + "step": 4190, + "valid_targets_mean": 1521.1, + "valid_targets_min": 778 + }, + { + "epoch": 4.406512605042017, + "grad_norm": 0.9073092150371975, + "learning_rate": 1.4533261088477257e-05, + "loss": 0.141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16051650047302246, + "step": 4195, + "valid_targets_mean": 1617.8, + "valid_targets_min": 851 + }, + { + "epoch": 4.411764705882353, + "grad_norm": 0.7397162950148408, + "learning_rate": 1.448288877640637e-05, + "loss": 0.1297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11534975469112396, + "step": 4200, + "valid_targets_mean": 1434.4, + "valid_targets_min": 692 + }, + { + "epoch": 4.417016806722689, + "grad_norm": 0.7733942072643732, + "learning_rate": 1.4432554315840082e-05, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12870749831199646, + "step": 4205, + "valid_targets_mean": 1500.7, + "valid_targets_min": 741 + }, + { + "epoch": 4.4222689075630255, + "grad_norm": 0.8287058084711577, + "learning_rate": 1.438225805211039e-05, + "loss": 0.1252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12352833896875381, + "step": 4210, + "valid_targets_mean": 1306.1, + "valid_targets_min": 639 + }, + { + "epoch": 4.427521008403361, + "grad_norm": 0.8460146278274804, + "learning_rate": 1.4332000330287248e-05, + "loss": 0.1163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11612292379140854, + "step": 4215, + "valid_targets_mean": 1381.9, + "valid_targets_min": 772 + }, + { + "epoch": 4.432773109243698, + "grad_norm": 0.8428338799366506, + "learning_rate": 1.428178149517617e-05, + "loss": 0.1303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1271061897277832, + "step": 4220, + "valid_targets_mean": 1532.2, + "valid_targets_min": 765 + }, + { + "epoch": 4.438025210084033, + "grad_norm": 0.7764971458397468, + "learning_rate": 1.4231601891315876e-05, + "loss": 0.1255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1206718236207962, + "step": 4225, + "valid_targets_mean": 1425.2, + "valid_targets_min": 695 + }, + { + "epoch": 4.44327731092437, + "grad_norm": 0.8017610552892988, + "learning_rate": 1.418146186297594e-05, + "loss": 0.1257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12809637188911438, + "step": 4230, + "valid_targets_mean": 1575.6, + "valid_targets_min": 807 + }, + { + "epoch": 4.448529411764706, + "grad_norm": 0.8204445283311675, + "learning_rate": 1.413136175415441e-05, + "loss": 0.1298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1299220323562622, + "step": 4235, + "valid_targets_mean": 1367.1, + "valid_targets_min": 873 + }, + { + "epoch": 4.453781512605042, + "grad_norm": 0.8229673951166021, + "learning_rate": 1.4081301908575457e-05, + "loss": 0.1265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13332301378250122, + "step": 4240, + "valid_targets_mean": 1369.4, + "valid_targets_min": 591 + }, + { + "epoch": 4.459033613445378, + "grad_norm": 0.8025930909766332, + "learning_rate": 1.4031282669687035e-05, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11734200268983841, + "step": 4245, + "valid_targets_mean": 1319.9, + "valid_targets_min": 646 + }, + { + "epoch": 4.464285714285714, + "grad_norm": 0.9239894606873152, + "learning_rate": 1.3981304380658474e-05, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21401086449623108, + "step": 4250, + "valid_targets_mean": 4488.5, + "valid_targets_min": 1323 + }, + { + "epoch": 4.46953781512605, + "grad_norm": 0.6598274422549769, + "learning_rate": 1.3931367384378183e-05, + "loss": 0.2242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2256534993648529, + "step": 4255, + "valid_targets_mean": 4344.8, + "valid_targets_min": 1631 + }, + { + "epoch": 4.474789915966387, + "grad_norm": 0.5777142536216642, + "learning_rate": 1.3881472023451262e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21953260898590088, + "step": 4260, + "valid_targets_mean": 4358.4, + "valid_targets_min": 1509 + }, + { + "epoch": 4.480042016806722, + "grad_norm": 0.5644265671396247, + "learning_rate": 1.3831618640197153e-05, + "loss": 0.211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19617916643619537, + "step": 4265, + "valid_targets_mean": 4067.4, + "valid_targets_min": 2689 + }, + { + "epoch": 4.485294117647059, + "grad_norm": 0.6800329773093203, + "learning_rate": 1.3781807576647315e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19588659703731537, + "step": 4270, + "valid_targets_mean": 3231.8, + "valid_targets_min": 1184 + }, + { + "epoch": 4.4905462184873945, + "grad_norm": 0.583998108779037, + "learning_rate": 1.373203917454285e-05, + "loss": 0.2187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18826688826084137, + "step": 4275, + "valid_targets_mean": 4015.0, + "valid_targets_min": 1413 + }, + { + "epoch": 4.495798319327731, + "grad_norm": 0.5993018924234834, + "learning_rate": 1.3682313775332186e-05, + "loss": 0.1913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20078253746032715, + "step": 4280, + "valid_targets_mean": 4218.1, + "valid_targets_min": 2160 + }, + { + "epoch": 4.501050420168067, + "grad_norm": 0.6213345974417059, + "learning_rate": 1.3632631720168705e-05, + "loss": 0.2102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21006864309310913, + "step": 4285, + "valid_targets_mean": 3998.2, + "valid_targets_min": 1654 + }, + { + "epoch": 4.506302521008403, + "grad_norm": 0.6454438419555589, + "learning_rate": 1.358299334990842e-05, + "loss": 0.2107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21891412138938904, + "step": 4290, + "valid_targets_mean": 3480.9, + "valid_targets_min": 1462 + }, + { + "epoch": 4.51155462184874, + "grad_norm": 0.5793855112485845, + "learning_rate": 1.3533399005107635e-05, + "loss": 0.214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19543211162090302, + "step": 4295, + "valid_targets_mean": 4008.8, + "valid_targets_min": 874 + }, + { + "epoch": 4.516806722689076, + "grad_norm": 0.5886149747788205, + "learning_rate": 1.3483849026020603e-05, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19601313769817352, + "step": 4300, + "valid_targets_mean": 3658.9, + "valid_targets_min": 953 + }, + { + "epoch": 4.522058823529412, + "grad_norm": 0.6162622988474424, + "learning_rate": 1.3434343752597199e-05, + "loss": 0.2143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2158222198486328, + "step": 4305, + "valid_targets_mean": 4291.2, + "valid_targets_min": 1265 + }, + { + "epoch": 4.527310924369748, + "grad_norm": 0.5225855536952907, + "learning_rate": 1.3384883524480576e-05, + "loss": 0.2084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1962510198354721, + "step": 4310, + "valid_targets_mean": 4939.0, + "valid_targets_min": 1596 + }, + { + "epoch": 4.532563025210084, + "grad_norm": 0.639285317945504, + "learning_rate": 1.333546868100486e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20231905579566956, + "step": 4315, + "valid_targets_mean": 3340.8, + "valid_targets_min": 1759 + }, + { + "epoch": 4.53781512605042, + "grad_norm": 0.6692843580892279, + "learning_rate": 1.3286099561192784e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21312329173088074, + "step": 4320, + "valid_targets_mean": 3084.3, + "valid_targets_min": 1353 + }, + { + "epoch": 4.543067226890757, + "grad_norm": 0.6120205294157653, + "learning_rate": 1.3236776503753397e-05, + "loss": 0.2112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19397076964378357, + "step": 4325, + "valid_targets_mean": 3763.4, + "valid_targets_min": 1643 + }, + { + "epoch": 4.548319327731092, + "grad_norm": 0.668479121847745, + "learning_rate": 1.3187499847079713e-05, + "loss": 0.2093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2041492760181427, + "step": 4330, + "valid_targets_mean": 3227.9, + "valid_targets_min": 1269 + }, + { + "epoch": 4.553571428571429, + "grad_norm": 0.6752786859520122, + "learning_rate": 1.3138269929246427e-05, + "loss": 0.2217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22046181559562683, + "step": 4335, + "valid_targets_mean": 3324.2, + "valid_targets_min": 1681 + }, + { + "epoch": 4.5588235294117645, + "grad_norm": 0.6548319871768743, + "learning_rate": 1.308908708800753e-05, + "loss": 0.2125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2117079198360443, + "step": 4340, + "valid_targets_mean": 3792.3, + "valid_targets_min": 1477 + }, + { + "epoch": 4.564075630252101, + "grad_norm": 0.6879232525076725, + "learning_rate": 1.3039951660794063e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20829437673091888, + "step": 4345, + "valid_targets_mean": 3502.8, + "valid_targets_min": 662 + }, + { + "epoch": 4.569327731092437, + "grad_norm": 0.6897322297757514, + "learning_rate": 1.2990863984711773e-05, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21518868207931519, + "step": 4350, + "valid_targets_mean": 3320.8, + "valid_targets_min": 2077 + }, + { + "epoch": 4.574579831932773, + "grad_norm": 0.688837675359442, + "learning_rate": 1.294182439653878e-05, + "loss": 0.2003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18814504146575928, + "step": 4355, + "valid_targets_mean": 2756.4, + "valid_targets_min": 1505 + }, + { + "epoch": 4.579831932773109, + "grad_norm": 0.7330677533755521, + "learning_rate": 1.2892833232723297e-05, + "loss": 0.2183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24092328548431396, + "step": 4360, + "valid_targets_mean": 3447.9, + "valid_targets_min": 1978 + }, + { + "epoch": 4.5850840336134455, + "grad_norm": 0.7241849761685287, + "learning_rate": 1.2843890829381303e-05, + "loss": 0.2165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26030996441841125, + "step": 4365, + "valid_targets_mean": 4194.2, + "valid_targets_min": 1490 + }, + { + "epoch": 4.590336134453781, + "grad_norm": 0.7026396424067808, + "learning_rate": 1.2794997522294263e-05, + "loss": 0.2035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21944516897201538, + "step": 4370, + "valid_targets_mean": 2970.4, + "valid_targets_min": 1422 + }, + { + "epoch": 4.595588235294118, + "grad_norm": 0.6259185795346857, + "learning_rate": 1.2746153646906766e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1922265887260437, + "step": 4375, + "valid_targets_mean": 3396.0, + "valid_targets_min": 1603 + }, + { + "epoch": 4.600840336134453, + "grad_norm": 0.7438486179833006, + "learning_rate": 1.2697359538324303e-05, + "loss": 0.2121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21673396229743958, + "step": 4380, + "valid_targets_mean": 2728.2, + "valid_targets_min": 863 + }, + { + "epoch": 4.60609243697479, + "grad_norm": 0.6370907634088678, + "learning_rate": 1.2648615531310907e-05, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19770187139511108, + "step": 4385, + "valid_targets_mean": 3677.7, + "valid_targets_min": 1208 + }, + { + "epoch": 4.6113445378151265, + "grad_norm": 0.68482682086709, + "learning_rate": 1.259992196028688e-05, + "loss": 0.2102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2116718888282776, + "step": 4390, + "valid_targets_mean": 3316.8, + "valid_targets_min": 1501 + }, + { + "epoch": 4.616596638655462, + "grad_norm": 0.7052208338762722, + "learning_rate": 1.2551279159326495e-05, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2252679467201233, + "step": 4395, + "valid_targets_mean": 3425.5, + "valid_targets_min": 1828 + }, + { + "epoch": 4.621848739495798, + "grad_norm": 0.6124764309205445, + "learning_rate": 1.2502687462155709e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1903487741947174, + "step": 4400, + "valid_targets_mean": 3237.6, + "valid_targets_min": 1110 + }, + { + "epoch": 4.6271008403361344, + "grad_norm": 0.8771050749310795, + "learning_rate": 1.2454147202149865e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21824470162391663, + "step": 4405, + "valid_targets_mean": 2931.8, + "valid_targets_min": 1283 + }, + { + "epoch": 4.632352941176471, + "grad_norm": 0.681928418139891, + "learning_rate": 1.2405658712331409e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20178645849227905, + "step": 4410, + "valid_targets_mean": 3311.4, + "valid_targets_min": 1714 + }, + { + "epoch": 4.637605042016807, + "grad_norm": 0.7118807798201505, + "learning_rate": 1.2357222325367604e-05, + "loss": 0.2084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21912041306495667, + "step": 4415, + "valid_targets_mean": 3572.0, + "valid_targets_min": 1216 + }, + { + "epoch": 4.642857142857143, + "grad_norm": 0.6823054558978693, + "learning_rate": 1.2308838373568249e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20421810448169708, + "step": 4420, + "valid_targets_mean": 3451.4, + "valid_targets_min": 924 + }, + { + "epoch": 4.648109243697479, + "grad_norm": 0.6938843637670672, + "learning_rate": 1.2260507188883395e-05, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19850008189678192, + "step": 4425, + "valid_targets_mean": 3175.6, + "valid_targets_min": 1292 + }, + { + "epoch": 4.6533613445378155, + "grad_norm": 0.7077142011650852, + "learning_rate": 1.2212229102901077e-05, + "loss": 0.2091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2047484815120697, + "step": 4430, + "valid_targets_mean": 3533.4, + "valid_targets_min": 1265 + }, + { + "epoch": 4.658613445378151, + "grad_norm": 0.7159319249656814, + "learning_rate": 1.2164004446845037e-05, + "loss": 0.1944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20691466331481934, + "step": 4435, + "valid_targets_mean": 3325.9, + "valid_targets_min": 1394 + }, + { + "epoch": 4.663865546218488, + "grad_norm": 0.7559645592765224, + "learning_rate": 1.2115833551572438e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2295243889093399, + "step": 4440, + "valid_targets_mean": 3590.2, + "valid_targets_min": 1844 + }, + { + "epoch": 4.669117647058823, + "grad_norm": 0.7415198687787329, + "learning_rate": 1.2067716747571616e-05, + "loss": 0.1999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23479296267032623, + "step": 4445, + "valid_targets_mean": 3564.9, + "valid_targets_min": 1482 + }, + { + "epoch": 4.67436974789916, + "grad_norm": 0.688853324893974, + "learning_rate": 1.201965436495978e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21714499592781067, + "step": 4450, + "valid_targets_mean": 3340.8, + "valid_targets_min": 1654 + }, + { + "epoch": 4.679621848739496, + "grad_norm": 0.7016909162355258, + "learning_rate": 1.1971646733480784e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20046481490135193, + "step": 4455, + "valid_targets_mean": 2951.6, + "valid_targets_min": 584 + }, + { + "epoch": 4.684873949579832, + "grad_norm": 0.6941971218245336, + "learning_rate": 1.1923694182502848e-05, + "loss": 0.2035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2116384208202362, + "step": 4460, + "valid_targets_mean": 3478.9, + "valid_targets_min": 1049 + }, + { + "epoch": 4.690126050420168, + "grad_norm": 0.6893303247496907, + "learning_rate": 1.1875797041016286e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2102683186531067, + "step": 4465, + "valid_targets_mean": 3287.8, + "valid_targets_min": 1350 + }, + { + "epoch": 4.695378151260504, + "grad_norm": 0.7350721707173831, + "learning_rate": 1.1827955637631283e-05, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21726731956005096, + "step": 4470, + "valid_targets_mean": 3590.0, + "valid_targets_min": 1624 + }, + { + "epoch": 4.70063025210084, + "grad_norm": 0.6938494286562441, + "learning_rate": 1.1780170300575602e-05, + "loss": 0.1884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18289947509765625, + "step": 4475, + "valid_targets_mean": 2807.6, + "valid_targets_min": 916 + }, + { + "epoch": 4.705882352941177, + "grad_norm": 0.6412962232296662, + "learning_rate": 1.1732441357692353e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18760257959365845, + "step": 4480, + "valid_targets_mean": 3658.0, + "valid_targets_min": 2032 + }, + { + "epoch": 4.711134453781512, + "grad_norm": 0.7583818317813796, + "learning_rate": 1.1684769136437742e-05, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21241483092308044, + "step": 4485, + "valid_targets_mean": 3141.9, + "valid_targets_min": 1259 + }, + { + "epoch": 4.716386554621849, + "grad_norm": 0.6326913726737966, + "learning_rate": 1.1637153963878815e-05, + "loss": 0.195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19029675424098969, + "step": 4490, + "valid_targets_mean": 3575.4, + "valid_targets_min": 1183 + }, + { + "epoch": 4.7216386554621845, + "grad_norm": 0.6290790387214149, + "learning_rate": 1.1589596166691231e-05, + "loss": 0.192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21391956508159637, + "step": 4495, + "valid_targets_mean": 4193.9, + "valid_targets_min": 1101 + }, + { + "epoch": 4.726890756302521, + "grad_norm": 0.6475834579736413, + "learning_rate": 1.1542096071157012e-05, + "loss": 0.2114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20298393070697784, + "step": 4500, + "valid_targets_mean": 3821.9, + "valid_targets_min": 1726 + }, + { + "epoch": 4.732142857142857, + "grad_norm": 0.6979196464072809, + "learning_rate": 1.1494654003162285e-05, + "loss": 0.1993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20892876386642456, + "step": 4505, + "valid_targets_mean": 3171.0, + "valid_targets_min": 1669 + }, + { + "epoch": 4.737394957983193, + "grad_norm": 0.6535989262032701, + "learning_rate": 1.1447270288195089e-05, + "loss": 0.208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20477300882339478, + "step": 4510, + "valid_targets_mean": 3491.4, + "valid_targets_min": 1185 + }, + { + "epoch": 4.742647058823529, + "grad_norm": 0.7434999336674271, + "learning_rate": 1.1399945251343114e-05, + "loss": 0.215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23692408204078674, + "step": 4515, + "valid_targets_mean": 3283.9, + "valid_targets_min": 1091 + }, + { + "epoch": 4.7478991596638656, + "grad_norm": 0.7222774058425186, + "learning_rate": 1.1352679217291457e-05, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2167915254831314, + "step": 4520, + "valid_targets_mean": 3654.3, + "valid_targets_min": 1352 + }, + { + "epoch": 4.753151260504202, + "grad_norm": 0.5928834886740821, + "learning_rate": 1.1305472510320419e-05, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1938672959804535, + "step": 4525, + "valid_targets_mean": 4265.6, + "valid_targets_min": 1535 + }, + { + "epoch": 4.758403361344538, + "grad_norm": 0.7282043659742619, + "learning_rate": 1.1258325454303286e-05, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1851435899734497, + "step": 4530, + "valid_targets_mean": 2997.0, + "valid_targets_min": 854 + }, + { + "epoch": 4.7636554621848735, + "grad_norm": 0.7451730107148029, + "learning_rate": 1.1211238372704073e-05, + "loss": 0.1919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18663758039474487, + "step": 4535, + "valid_targets_mean": 2795.1, + "valid_targets_min": 1179 + }, + { + "epoch": 4.76890756302521, + "grad_norm": 0.7184545671700696, + "learning_rate": 1.1164211588575339e-05, + "loss": 0.1969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19145233929157257, + "step": 4540, + "valid_targets_mean": 3652.7, + "valid_targets_min": 1444 + }, + { + "epoch": 4.774159663865547, + "grad_norm": 0.7397780134283227, + "learning_rate": 1.1117245424555967e-05, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1895849108695984, + "step": 4545, + "valid_targets_mean": 2500.1, + "valid_targets_min": 642 + }, + { + "epoch": 4.779411764705882, + "grad_norm": 0.6713856995898463, + "learning_rate": 1.1070340202868915e-05, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1881069391965866, + "step": 4550, + "valid_targets_mean": 3590.1, + "valid_targets_min": 1637 + }, + { + "epoch": 4.784663865546219, + "grad_norm": 0.8087445793498248, + "learning_rate": 1.1023496245319056e-05, + "loss": 0.2125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24480175971984863, + "step": 4555, + "valid_targets_mean": 3466.0, + "valid_targets_min": 1175 + }, + { + "epoch": 4.7899159663865545, + "grad_norm": 0.7068565649556457, + "learning_rate": 1.0976713873290925e-05, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20139086246490479, + "step": 4560, + "valid_targets_mean": 3205.6, + "valid_targets_min": 1612 + }, + { + "epoch": 4.795168067226891, + "grad_norm": 0.6620926828840262, + "learning_rate": 1.0929993407746543e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19388242065906525, + "step": 4565, + "valid_targets_mean": 3423.2, + "valid_targets_min": 1743 + }, + { + "epoch": 4.800420168067227, + "grad_norm": 0.6727413718092184, + "learning_rate": 1.0883335169223212e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1882367581129074, + "step": 4570, + "valid_targets_mean": 3168.8, + "valid_targets_min": 1398 + }, + { + "epoch": 4.805672268907563, + "grad_norm": 0.6558248286205179, + "learning_rate": 1.0836739477831297e-05, + "loss": 0.1994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20704635977745056, + "step": 4575, + "valid_targets_mean": 3489.9, + "valid_targets_min": 1558 + }, + { + "epoch": 4.810924369747899, + "grad_norm": 0.6847057248913088, + "learning_rate": 1.0790206653252055e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20168985426425934, + "step": 4580, + "valid_targets_mean": 3278.1, + "valid_targets_min": 1445 + }, + { + "epoch": 4.8161764705882355, + "grad_norm": 0.6533366829855569, + "learning_rate": 1.0743737014735434e-05, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18918371200561523, + "step": 4585, + "valid_targets_mean": 3682.9, + "valid_targets_min": 2000 + }, + { + "epoch": 4.821428571428571, + "grad_norm": 0.7708553362949551, + "learning_rate": 1.0697330881097857e-05, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19962045550346375, + "step": 4590, + "valid_targets_mean": 3065.2, + "valid_targets_min": 1243 + }, + { + "epoch": 4.826680672268908, + "grad_norm": 0.7089419295191447, + "learning_rate": 1.0650988570720077e-05, + "loss": 0.2091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20890700817108154, + "step": 4595, + "valid_targets_mean": 3286.4, + "valid_targets_min": 1223 + }, + { + "epoch": 4.831932773109243, + "grad_norm": 0.7147247615274656, + "learning_rate": 1.0604710401544963e-05, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19596712291240692, + "step": 4600, + "valid_targets_mean": 3027.9, + "valid_targets_min": 869 + }, + { + "epoch": 4.83718487394958, + "grad_norm": 0.7304597580233416, + "learning_rate": 1.0558496691075318e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20904818177223206, + "step": 4605, + "valid_targets_mean": 3077.8, + "valid_targets_min": 1603 + }, + { + "epoch": 4.842436974789916, + "grad_norm": 0.7224217938478468, + "learning_rate": 1.0512347756371726e-05, + "loss": 0.206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18744680285453796, + "step": 4610, + "valid_targets_mean": 2883.4, + "valid_targets_min": 721 + }, + { + "epoch": 4.847689075630252, + "grad_norm": 0.7107738441028865, + "learning_rate": 1.0466263914050362e-05, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2116435468196869, + "step": 4615, + "valid_targets_mean": 3873.2, + "valid_targets_min": 1994 + }, + { + "epoch": 4.852941176470588, + "grad_norm": 0.7236551319922022, + "learning_rate": 1.0420245480280804e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18254084885120392, + "step": 4620, + "valid_targets_mean": 3090.9, + "valid_targets_min": 1066 + }, + { + "epoch": 4.858193277310924, + "grad_norm": 0.6924268052728789, + "learning_rate": 1.0374292770783891e-05, + "loss": 0.2099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21132606267929077, + "step": 4625, + "valid_targets_mean": 3609.6, + "valid_targets_min": 1648 + }, + { + "epoch": 4.86344537815126, + "grad_norm": 0.710132639844148, + "learning_rate": 1.0328406100829542e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19806388020515442, + "step": 4630, + "valid_targets_mean": 3135.9, + "valid_targets_min": 1889 + }, + { + "epoch": 4.868697478991597, + "grad_norm": 8.223593196499644, + "learning_rate": 1.0282585785234578e-05, + "loss": 0.2083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21035614609718323, + "step": 4635, + "valid_targets_mean": 2491.6, + "valid_targets_min": 885 + }, + { + "epoch": 4.873949579831933, + "grad_norm": 0.6950690926035392, + "learning_rate": 1.023683213836061e-05, + "loss": 0.2101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20301932096481323, + "step": 4640, + "valid_targets_mean": 3529.4, + "valid_targets_min": 1257 + }, + { + "epoch": 4.879201680672269, + "grad_norm": 0.7338702013416679, + "learning_rate": 1.0191145474111823e-05, + "loss": 0.2063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20622602105140686, + "step": 4645, + "valid_targets_mean": 3178.1, + "valid_targets_min": 1372 + }, + { + "epoch": 4.884453781512605, + "grad_norm": 0.6717591588502287, + "learning_rate": 1.0145526105932872e-05, + "loss": 0.2012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2017562985420227, + "step": 4650, + "valid_targets_mean": 3422.6, + "valid_targets_min": 1598 + }, + { + "epoch": 4.889705882352941, + "grad_norm": 0.76533252784604, + "learning_rate": 1.0099974346806714e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1944223940372467, + "step": 4655, + "valid_targets_mean": 2925.1, + "valid_targets_min": 1074 + }, + { + "epoch": 4.894957983193278, + "grad_norm": 0.6637317114328515, + "learning_rate": 1.0054490509252423e-05, + "loss": 0.2099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19309523701667786, + "step": 4660, + "valid_targets_mean": 3366.2, + "valid_targets_min": 980 + }, + { + "epoch": 4.900210084033613, + "grad_norm": 0.6982820833530574, + "learning_rate": 1.0009074905323118e-05, + "loss": 0.2175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21207424998283386, + "step": 4665, + "valid_targets_mean": 3100.1, + "valid_targets_min": 793 + }, + { + "epoch": 4.90546218487395, + "grad_norm": 0.6432617078880017, + "learning_rate": 9.963727846603784e-06, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21252809464931488, + "step": 4670, + "valid_targets_mean": 3674.5, + "valid_targets_min": 852 + }, + { + "epoch": 4.910714285714286, + "grad_norm": 0.7841146093401851, + "learning_rate": 9.918449644209087e-06, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18188413977622986, + "step": 4675, + "valid_targets_mean": 2951.1, + "valid_targets_min": 1107 + }, + { + "epoch": 4.915966386554622, + "grad_norm": 0.7711023446249636, + "learning_rate": 9.873240608781341e-06, + "loss": 0.2054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20018140971660614, + "step": 4680, + "valid_targets_mean": 2839.9, + "valid_targets_min": 1298 + }, + { + "epoch": 4.921218487394958, + "grad_norm": 0.6720914422866899, + "learning_rate": 9.828101050488308e-06, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19184359908103943, + "step": 4685, + "valid_targets_mean": 3508.3, + "valid_targets_min": 1428 + }, + { + "epoch": 4.926470588235294, + "grad_norm": 0.6979997171728765, + "learning_rate": 9.783031279021063e-06, + "loss": 0.1949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1979558765888214, + "step": 4690, + "valid_targets_mean": 3186.6, + "valid_targets_min": 1193 + }, + { + "epoch": 4.93172268907563, + "grad_norm": 0.6375536515405493, + "learning_rate": 9.738031603591926e-06, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18585914373397827, + "step": 4695, + "valid_targets_mean": 3687.8, + "valid_targets_min": 1210 + }, + { + "epoch": 4.936974789915967, + "grad_norm": 0.6740300095798715, + "learning_rate": 9.69310233293227e-06, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1872563362121582, + "step": 4700, + "valid_targets_mean": 3261.6, + "valid_targets_min": 974 + }, + { + "epoch": 4.942226890756302, + "grad_norm": 0.6728949238979616, + "learning_rate": 9.648243775290476e-06, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1736118197441101, + "step": 4705, + "valid_targets_mean": 2934.8, + "valid_targets_min": 1575 + }, + { + "epoch": 4.947478991596639, + "grad_norm": 0.7317387069659124, + "learning_rate": 9.60345623842974e-06, + "loss": 0.1944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19196751713752747, + "step": 4710, + "valid_targets_mean": 3090.5, + "valid_targets_min": 1697 + }, + { + "epoch": 4.9527310924369745, + "grad_norm": 0.7423800247961655, + "learning_rate": 9.558740029626046e-06, + "loss": 0.1994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19240882992744446, + "step": 4715, + "valid_targets_mean": 3297.7, + "valid_targets_min": 897 + }, + { + "epoch": 4.957983193277311, + "grad_norm": 0.79806625760805, + "learning_rate": 9.51409545566597e-06, + "loss": 0.2198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22060896456241608, + "step": 4720, + "valid_targets_mean": 3071.2, + "valid_targets_min": 887 + }, + { + "epoch": 4.963235294117647, + "grad_norm": 0.7615537115059738, + "learning_rate": 9.469522822844663e-06, + "loss": 0.1948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20150801539421082, + "step": 4725, + "valid_targets_mean": 2714.7, + "valid_targets_min": 1279 + }, + { + "epoch": 4.968487394957983, + "grad_norm": 0.7230728317576487, + "learning_rate": 9.425022436963664e-06, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17812782526016235, + "step": 4730, + "valid_targets_mean": 3112.9, + "valid_targets_min": 1021 + }, + { + "epoch": 4.973739495798319, + "grad_norm": 0.8017286637803975, + "learning_rate": 9.380594603328875e-06, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21535563468933105, + "step": 4735, + "valid_targets_mean": 3526.3, + "valid_targets_min": 654 + }, + { + "epoch": 4.9789915966386555, + "grad_norm": 0.7962974276395127, + "learning_rate": 9.336239626748432e-06, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24022144079208374, + "step": 4740, + "valid_targets_mean": 2941.2, + "valid_targets_min": 1039 + }, + { + "epoch": 4.984243697478991, + "grad_norm": 0.6987990811270663, + "learning_rate": 9.291957811530602e-06, + "loss": 0.208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20734629034996033, + "step": 4745, + "valid_targets_mean": 3804.6, + "valid_targets_min": 1737 + }, + { + "epoch": 4.989495798319328, + "grad_norm": 0.6320507915977177, + "learning_rate": 9.247749461481712e-06, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17486436665058136, + "step": 4750, + "valid_targets_mean": 3503.6, + "valid_targets_min": 1276 + }, + { + "epoch": 4.994747899159664, + "grad_norm": 0.6565835428886582, + "learning_rate": 9.203614879904083e-06, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21443158388137817, + "step": 4755, + "valid_targets_mean": 4137.7, + "valid_targets_min": 1572 + }, + { + "epoch": 5.0, + "grad_norm": 0.7198533786254716, + "learning_rate": 9.159554369593901e-06, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2270594984292984, + "step": 4760, + "valid_targets_mean": 3715.1, + "valid_targets_min": 791 + }, + { + "epoch": 5.005252100840337, + "grad_norm": 0.7571959238315181, + "learning_rate": 9.115568232839193e-06, + "loss": 0.1534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13461460173130035, + "step": 4765, + "valid_targets_mean": 1423.4, + "valid_targets_min": 715 + }, + { + "epoch": 5.010504201680672, + "grad_norm": 0.8301602733718247, + "learning_rate": 9.071656771417721e-06, + "loss": 0.1363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12977659702301025, + "step": 4770, + "valid_targets_mean": 1407.8, + "valid_targets_min": 612 + }, + { + "epoch": 5.015756302521009, + "grad_norm": 0.8166485961728386, + "learning_rate": 9.027820286594902e-06, + "loss": 0.1336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13986307382583618, + "step": 4775, + "valid_targets_mean": 1413.5, + "valid_targets_min": 714 + }, + { + "epoch": 5.0210084033613445, + "grad_norm": 0.8634652135372101, + "learning_rate": 8.984059079121785e-06, + "loss": 0.1281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12289293855428696, + "step": 4780, + "valid_targets_mean": 1450.1, + "valid_targets_min": 754 + }, + { + "epoch": 5.026260504201681, + "grad_norm": 1.0868991923247484, + "learning_rate": 8.940373449232935e-06, + "loss": 0.1387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1558552235364914, + "step": 4785, + "valid_targets_mean": 1481.1, + "valid_targets_min": 718 + }, + { + "epoch": 5.031512605042017, + "grad_norm": 0.8280248464569533, + "learning_rate": 8.896763696644405e-06, + "loss": 0.1378, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13719496130943298, + "step": 4790, + "valid_targets_mean": 1530.8, + "valid_targets_min": 731 + }, + { + "epoch": 5.036764705882353, + "grad_norm": 0.8223621736586295, + "learning_rate": 8.853230120551693e-06, + "loss": 0.1238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12621982395648956, + "step": 4795, + "valid_targets_mean": 1416.6, + "valid_targets_min": 725 + }, + { + "epoch": 5.042016806722689, + "grad_norm": 0.870763805985631, + "learning_rate": 8.809773019627635e-06, + "loss": 0.1404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13102170825004578, + "step": 4800, + "valid_targets_mean": 1371.5, + "valid_targets_min": 610 + }, + { + "epoch": 5.0472689075630255, + "grad_norm": 0.8397454624147834, + "learning_rate": 8.766392692020413e-06, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12862995266914368, + "step": 4805, + "valid_targets_mean": 1484.7, + "valid_targets_min": 720 + }, + { + "epoch": 5.052521008403361, + "grad_norm": 0.7882246072954778, + "learning_rate": 8.723089435351497e-06, + "loss": 0.1306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11443109065294266, + "step": 4810, + "valid_targets_mean": 1365.5, + "valid_targets_min": 765 + }, + { + "epoch": 5.057773109243698, + "grad_norm": 0.8742920757552072, + "learning_rate": 8.679863546713559e-06, + "loss": 0.1256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.127573624253273, + "step": 4815, + "valid_targets_mean": 1234.7, + "valid_targets_min": 593 + }, + { + "epoch": 5.063025210084033, + "grad_norm": 0.7920380834226816, + "learning_rate": 8.636715322668498e-06, + "loss": 0.1276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11717408895492554, + "step": 4820, + "valid_targets_mean": 1540.4, + "valid_targets_min": 704 + }, + { + "epoch": 5.06827731092437, + "grad_norm": 0.780341581121388, + "learning_rate": 8.593645059245365e-06, + "loss": 0.1243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11017392575740814, + "step": 4825, + "valid_targets_mean": 1329.9, + "valid_targets_min": 722 + }, + { + "epoch": 5.073529411764706, + "grad_norm": 0.7603227971883492, + "learning_rate": 8.550653051938333e-06, + "loss": 0.1181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11557639390230179, + "step": 4830, + "valid_targets_mean": 1549.6, + "valid_targets_min": 865 + }, + { + "epoch": 5.078781512605042, + "grad_norm": 0.9030555834076215, + "learning_rate": 8.507739595704695e-06, + "loss": 0.1192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13490742444992065, + "step": 4835, + "valid_targets_mean": 1507.4, + "valid_targets_min": 797 + }, + { + "epoch": 5.084033613445378, + "grad_norm": 0.954349785672474, + "learning_rate": 8.464904984962832e-06, + "loss": 0.1242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14801670610904694, + "step": 4840, + "valid_targets_mean": 1542.4, + "valid_targets_min": 728 + }, + { + "epoch": 5.089285714285714, + "grad_norm": 0.9125020805445457, + "learning_rate": 8.422149513590151e-06, + "loss": 0.1271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13509975373744965, + "step": 4845, + "valid_targets_mean": 1595.2, + "valid_targets_min": 683 + }, + { + "epoch": 5.09453781512605, + "grad_norm": 0.871199900697412, + "learning_rate": 8.37947347492115e-06, + "loss": 0.1303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13447391986846924, + "step": 4850, + "valid_targets_mean": 1401.6, + "valid_targets_min": 706 + }, + { + "epoch": 5.099789915966387, + "grad_norm": 0.8629277026545269, + "learning_rate": 8.33687716174532e-06, + "loss": 0.1242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11550725251436234, + "step": 4855, + "valid_targets_mean": 1298.4, + "valid_targets_min": 625 + }, + { + "epoch": 5.105042016806722, + "grad_norm": 0.8030715253201013, + "learning_rate": 8.294360866305192e-06, + "loss": 0.1272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1309053897857666, + "step": 4860, + "valid_targets_mean": 1659.7, + "valid_targets_min": 790 + }, + { + "epoch": 5.110294117647059, + "grad_norm": 0.9265947353354664, + "learning_rate": 8.251924880294317e-06, + "loss": 0.1264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13342975080013275, + "step": 4865, + "valid_targets_mean": 1600.4, + "valid_targets_min": 684 + }, + { + "epoch": 5.1155462184873945, + "grad_norm": 0.8041955738679034, + "learning_rate": 8.20956949485527e-06, + "loss": 0.1196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1035323292016983, + "step": 4870, + "valid_targets_mean": 1338.3, + "valid_targets_min": 739 + }, + { + "epoch": 5.120798319327731, + "grad_norm": 0.8494495064085696, + "learning_rate": 8.167295000577622e-06, + "loss": 0.1197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11942795664072037, + "step": 4875, + "valid_targets_mean": 1456.2, + "valid_targets_min": 657 + }, + { + "epoch": 5.126050420168067, + "grad_norm": 1.0365198065360532, + "learning_rate": 8.125101687496e-06, + "loss": 0.1181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11641368269920349, + "step": 4880, + "valid_targets_mean": 1340.2, + "valid_targets_min": 783 + }, + { + "epoch": 5.131302521008403, + "grad_norm": 0.8635148283698016, + "learning_rate": 8.082989845088038e-06, + "loss": 0.1203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11869832128286362, + "step": 4885, + "valid_targets_mean": 1394.4, + "valid_targets_min": 734 + }, + { + "epoch": 5.13655462184874, + "grad_norm": 0.877173265352079, + "learning_rate": 8.040959762272441e-06, + "loss": 0.1232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12606994807720184, + "step": 4890, + "valid_targets_mean": 1501.8, + "valid_targets_min": 678 + }, + { + "epoch": 5.141806722689076, + "grad_norm": 0.9671178369518909, + "learning_rate": 7.999011727406995e-06, + "loss": 0.1224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12263111025094986, + "step": 4895, + "valid_targets_mean": 1651.7, + "valid_targets_min": 843 + }, + { + "epoch": 5.147058823529412, + "grad_norm": 0.8799343051970719, + "learning_rate": 7.957146028286524e-06, + "loss": 0.1285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13233694434165955, + "step": 4900, + "valid_targets_mean": 1632.1, + "valid_targets_min": 782 + }, + { + "epoch": 5.152310924369748, + "grad_norm": 0.8281302903839795, + "learning_rate": 7.915362952141017e-06, + "loss": 0.1215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13629840314388275, + "step": 4905, + "valid_targets_mean": 1615.3, + "valid_targets_min": 957 + }, + { + "epoch": 5.157563025210084, + "grad_norm": 0.9542718681795296, + "learning_rate": 7.873662785633594e-06, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14337027072906494, + "step": 4910, + "valid_targets_mean": 1482.9, + "valid_targets_min": 713 + }, + { + "epoch": 5.16281512605042, + "grad_norm": 1.3637598590452737, + "learning_rate": 7.832045814858538e-06, + "loss": 0.1235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12161692976951599, + "step": 4915, + "valid_targets_mean": 1552.8, + "valid_targets_min": 726 + }, + { + "epoch": 5.168067226890757, + "grad_norm": 0.821599075602171, + "learning_rate": 7.790512325339367e-06, + "loss": 0.1169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12724722921848297, + "step": 4920, + "valid_targets_mean": 1590.4, + "valid_targets_min": 1147 + }, + { + "epoch": 5.173319327731092, + "grad_norm": 0.8859043037264651, + "learning_rate": 7.74906260202685e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1232469454407692, + "step": 4925, + "valid_targets_mean": 1417.6, + "valid_targets_min": 700 + }, + { + "epoch": 5.178571428571429, + "grad_norm": 0.8207302664180218, + "learning_rate": 7.707696929297053e-06, + "loss": 0.1116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10313037782907486, + "step": 4930, + "valid_targets_mean": 1255.4, + "valid_targets_min": 640 + }, + { + "epoch": 5.1838235294117645, + "grad_norm": 0.8196199367733106, + "learning_rate": 7.666415590949382e-06, + "loss": 0.1214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12141343951225281, + "step": 4935, + "valid_targets_mean": 1566.6, + "valid_targets_min": 817 + }, + { + "epoch": 5.189075630252101, + "grad_norm": 0.8397722105470354, + "learning_rate": 7.625218870204676e-06, + "loss": 0.1157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11369533091783524, + "step": 4940, + "valid_targets_mean": 1440.9, + "valid_targets_min": 724 + }, + { + "epoch": 5.194327731092437, + "grad_norm": 0.84723032993192, + "learning_rate": 7.584107049703195e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1220942959189415, + "step": 4945, + "valid_targets_mean": 1399.1, + "valid_targets_min": 657 + }, + { + "epoch": 5.199579831932773, + "grad_norm": 1.1583713663548252, + "learning_rate": 7.5430804115027615e-06, + "loss": 0.1222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1239510029554367, + "step": 4950, + "valid_targets_mean": 1334.3, + "valid_targets_min": 746 + }, + { + "epoch": 5.204831932773109, + "grad_norm": 0.940619146388919, + "learning_rate": 7.502139237076744e-06, + "loss": 0.1203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13509805500507355, + "step": 4955, + "valid_targets_mean": 1452.8, + "valid_targets_min": 692 + }, + { + "epoch": 5.2100840336134455, + "grad_norm": 0.9031648217368532, + "learning_rate": 7.461283807312194e-06, + "loss": 0.117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10953864455223083, + "step": 4960, + "valid_targets_mean": 1338.4, + "valid_targets_min": 682 + }, + { + "epoch": 5.215336134453781, + "grad_norm": 0.7885774798170768, + "learning_rate": 7.420514402507886e-06, + "loss": 0.1134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11579865217208862, + "step": 4965, + "valid_targets_mean": 1360.2, + "valid_targets_min": 703 + }, + { + "epoch": 5.220588235294118, + "grad_norm": 0.8567753502507811, + "learning_rate": 7.379831302372389e-06, + "loss": 0.1264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12191884219646454, + "step": 4970, + "valid_targets_mean": 1520.9, + "valid_targets_min": 864 + }, + { + "epoch": 5.225840336134453, + "grad_norm": 0.9494851235486876, + "learning_rate": 7.3392347860221556e-06, + "loss": 0.1282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16232752799987793, + "step": 4975, + "valid_targets_mean": 1814.6, + "valid_targets_min": 734 + }, + { + "epoch": 5.23109243697479, + "grad_norm": 0.8077534069393267, + "learning_rate": 7.298725131979629e-06, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11486963927745819, + "step": 4980, + "valid_targets_mean": 1478.3, + "valid_targets_min": 607 + }, + { + "epoch": 5.236344537815126, + "grad_norm": 0.8359577317112136, + "learning_rate": 7.258302618171287e-06, + "loss": 0.1153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11313410103321075, + "step": 4985, + "valid_targets_mean": 1497.9, + "valid_targets_min": 551 + }, + { + "epoch": 5.241596638655462, + "grad_norm": 0.9503036196925028, + "learning_rate": 7.217967521925775e-06, + "loss": 0.1215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12508943676948547, + "step": 4990, + "valid_targets_mean": 1550.5, + "valid_targets_min": 676 + }, + { + "epoch": 5.246848739495798, + "grad_norm": 0.8718205924513632, + "learning_rate": 7.177720119971998e-06, + "loss": 0.1179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11915310472249985, + "step": 4995, + "valid_targets_mean": 1372.5, + "valid_targets_min": 904 + }, + { + "epoch": 5.2521008403361344, + "grad_norm": 0.9288193265136786, + "learning_rate": 7.137560688437184e-06, + "loss": 0.1122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12416621297597885, + "step": 5000, + "valid_targets_mean": 1479.0, + "valid_targets_min": 859 + }, + { + "epoch": 5.257352941176471, + "grad_norm": 0.8635518493795323, + "learning_rate": 7.097489502845047e-06, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09854140877723694, + "step": 5005, + "valid_targets_mean": 1234.8, + "valid_targets_min": 737 + }, + { + "epoch": 5.262605042016807, + "grad_norm": 0.9381227734247657, + "learning_rate": 7.0575068381138525e-06, + "loss": 0.1246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12833383679389954, + "step": 5010, + "valid_targets_mean": 1363.2, + "valid_targets_min": 791 + }, + { + "epoch": 5.267857142857143, + "grad_norm": 0.8806803188401461, + "learning_rate": 7.0176129685545414e-06, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12506265938282013, + "step": 5015, + "valid_targets_mean": 1532.1, + "valid_targets_min": 786 + }, + { + "epoch": 5.273109243697479, + "grad_norm": 0.8640788078067893, + "learning_rate": 6.977808167868867e-06, + "loss": 0.1139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1157342940568924, + "step": 5020, + "valid_targets_mean": 1471.0, + "valid_targets_min": 944 + }, + { + "epoch": 5.2783613445378155, + "grad_norm": 0.8116063638815398, + "learning_rate": 6.9380927091475085e-06, + "loss": 0.1159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10994266718626022, + "step": 5025, + "valid_targets_mean": 1494.1, + "valid_targets_min": 1075 + }, + { + "epoch": 5.283613445378151, + "grad_norm": 0.8409841937184768, + "learning_rate": 6.898466864868165e-06, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1092015877366066, + "step": 5030, + "valid_targets_mean": 1340.9, + "valid_targets_min": 740 + }, + { + "epoch": 5.288865546218488, + "grad_norm": 0.8012571565338946, + "learning_rate": 6.858930906893751e-06, + "loss": 0.119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12151382863521576, + "step": 5035, + "valid_targets_mean": 1530.9, + "valid_targets_min": 727 + }, + { + "epoch": 5.294117647058823, + "grad_norm": 1.0099731756987655, + "learning_rate": 6.819485106470454e-06, + "loss": 0.1119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.115475133061409, + "step": 5040, + "valid_targets_mean": 1233.0, + "valid_targets_min": 636 + }, + { + "epoch": 5.29936974789916, + "grad_norm": 0.8353570829541407, + "learning_rate": 6.780129734225949e-06, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1095833033323288, + "step": 5045, + "valid_targets_mean": 1246.8, + "valid_targets_min": 684 + }, + { + "epoch": 5.304621848739496, + "grad_norm": 0.8983751034814326, + "learning_rate": 6.740865060167483e-06, + "loss": 0.1079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1118505597114563, + "step": 5050, + "valid_targets_mean": 1392.5, + "valid_targets_min": 798 + }, + { + "epoch": 5.309873949579832, + "grad_norm": 0.8427967821093256, + "learning_rate": 6.701691353680049e-06, + "loss": 0.1195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1080687940120697, + "step": 5055, + "valid_targets_mean": 1319.9, + "valid_targets_min": 661 + }, + { + "epoch": 5.315126050420168, + "grad_norm": 0.8079776828047187, + "learning_rate": 6.662608883524542e-06, + "loss": 0.118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11414103209972382, + "step": 5060, + "valid_targets_mean": 1554.4, + "valid_targets_min": 665 + }, + { + "epoch": 5.320378151260504, + "grad_norm": 0.9308438137878955, + "learning_rate": 6.6236179178359115e-06, + "loss": 0.1152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12251292169094086, + "step": 5065, + "valid_targets_mean": 1417.8, + "valid_targets_min": 692 + }, + { + "epoch": 5.32563025210084, + "grad_norm": 1.0492334793929392, + "learning_rate": 6.5847187241213e-06, + "loss": 0.1171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11902189999818802, + "step": 5070, + "valid_targets_mean": 1223.8, + "valid_targets_min": 632 + }, + { + "epoch": 5.330882352941177, + "grad_norm": 0.8741745375157363, + "learning_rate": 6.5459115692582386e-06, + "loss": 0.1184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1177992969751358, + "step": 5075, + "valid_targets_mean": 1550.6, + "valid_targets_min": 922 + }, + { + "epoch": 5.336134453781512, + "grad_norm": 0.8231983027778437, + "learning_rate": 6.507196719492819e-06, + "loss": 0.115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11274613440036774, + "step": 5080, + "valid_targets_mean": 1542.2, + "valid_targets_min": 710 + }, + { + "epoch": 5.341386554621849, + "grad_norm": 0.838222313849615, + "learning_rate": 6.468574440437801e-06, + "loss": 0.1162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11383190006017685, + "step": 5085, + "valid_targets_mean": 1390.6, + "valid_targets_min": 750 + }, + { + "epoch": 5.3466386554621845, + "grad_norm": 0.8704926336132325, + "learning_rate": 6.43004499707089e-06, + "loss": 0.1154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12034046649932861, + "step": 5090, + "valid_targets_mean": 1515.5, + "valid_targets_min": 766 + }, + { + "epoch": 5.351890756302521, + "grad_norm": 0.7942910330002284, + "learning_rate": 6.3916086537328545e-06, + "loss": 0.1151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09943848103284836, + "step": 5095, + "valid_targets_mean": 1333.3, + "valid_targets_min": 748 + }, + { + "epoch": 5.357142857142857, + "grad_norm": 0.8272207055546795, + "learning_rate": 6.3532656741257145e-06, + "loss": 0.1196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11656372249126434, + "step": 5100, + "valid_targets_mean": 1395.2, + "valid_targets_min": 682 + }, + { + "epoch": 5.362394957983193, + "grad_norm": 0.8835103895559686, + "learning_rate": 6.315016321310965e-06, + "loss": 0.1223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11481855809688568, + "step": 5105, + "valid_targets_mean": 1515.6, + "valid_targets_min": 738 + }, + { + "epoch": 5.367647058823529, + "grad_norm": 0.8627265594469545, + "learning_rate": 6.276860857707727e-06, + "loss": 0.1153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11496169120073318, + "step": 5110, + "valid_targets_mean": 1312.8, + "valid_targets_min": 731 + }, + { + "epoch": 5.3728991596638656, + "grad_norm": 0.909694468472037, + "learning_rate": 6.238799545090994e-06, + "loss": 0.1215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12938551604747772, + "step": 5115, + "valid_targets_mean": 1492.9, + "valid_targets_min": 646 + }, + { + "epoch": 5.378151260504202, + "grad_norm": 0.8310216699675694, + "learning_rate": 6.20083264458981e-06, + "loss": 0.1099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10411947965621948, + "step": 5120, + "valid_targets_mean": 1369.1, + "valid_targets_min": 780 + }, + { + "epoch": 5.383403361344538, + "grad_norm": 0.8711155777854998, + "learning_rate": 6.162960416685469e-06, + "loss": 0.1109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11428333818912506, + "step": 5125, + "valid_targets_mean": 1521.9, + "valid_targets_min": 676 + }, + { + "epoch": 5.388655462184874, + "grad_norm": 0.870482484994608, + "learning_rate": 6.125183121209741e-06, + "loss": 0.1191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10633750259876251, + "step": 5130, + "valid_targets_mean": 1277.9, + "valid_targets_min": 687 + }, + { + "epoch": 5.39390756302521, + "grad_norm": 0.8795820034570103, + "learning_rate": 6.087501017343107e-06, + "loss": 0.1118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10446271300315857, + "step": 5135, + "valid_targets_mean": 1336.5, + "valid_targets_min": 703 + }, + { + "epoch": 5.399159663865547, + "grad_norm": 0.8575277820951928, + "learning_rate": 6.0499143636129384e-06, + "loss": 0.1105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10649226605892181, + "step": 5140, + "valid_targets_mean": 1288.1, + "valid_targets_min": 522 + }, + { + "epoch": 5.404411764705882, + "grad_norm": 0.8701860188774736, + "learning_rate": 6.012423417891766e-06, + "loss": 0.116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11557996273040771, + "step": 5145, + "valid_targets_mean": 1526.9, + "valid_targets_min": 764 + }, + { + "epoch": 5.409663865546219, + "grad_norm": 0.9973788403003421, + "learning_rate": 5.975028437395493e-06, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1302410513162613, + "step": 5150, + "valid_targets_mean": 1372.4, + "valid_targets_min": 548 + }, + { + "epoch": 5.4149159663865545, + "grad_norm": 0.8939645783701128, + "learning_rate": 5.937729678681612e-06, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11539514362812042, + "step": 5155, + "valid_targets_mean": 1356.8, + "valid_targets_min": 694 + }, + { + "epoch": 5.420168067226891, + "grad_norm": 0.8246273501078926, + "learning_rate": 5.9005273976474684e-06, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11548338830471039, + "step": 5160, + "valid_targets_mean": 1540.9, + "valid_targets_min": 823 + }, + { + "epoch": 5.425420168067227, + "grad_norm": 0.7236132877010878, + "learning_rate": 5.8634218495285126e-06, + "loss": 0.0987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08354189991950989, + "step": 5165, + "valid_targets_mean": 1442.4, + "valid_targets_min": 682 + }, + { + "epoch": 5.430672268907563, + "grad_norm": 0.8988049282372673, + "learning_rate": 5.8264132888965155e-06, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11418993771076202, + "step": 5170, + "valid_targets_mean": 1407.6, + "valid_targets_min": 693 + }, + { + "epoch": 5.435924369747899, + "grad_norm": 0.8655375216766177, + "learning_rate": 5.789501969657853e-06, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10003848373889923, + "step": 5175, + "valid_targets_mean": 1197.8, + "valid_targets_min": 639 + }, + { + "epoch": 5.4411764705882355, + "grad_norm": 0.8362323911596797, + "learning_rate": 5.752688145051761e-06, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09275621920824051, + "step": 5180, + "valid_targets_mean": 1423.3, + "valid_targets_min": 685 + }, + { + "epoch": 5.446428571428571, + "grad_norm": 0.8296465425650306, + "learning_rate": 5.715972067648561e-06, + "loss": 0.1078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10052791237831116, + "step": 5185, + "valid_targets_mean": 1331.5, + "valid_targets_min": 827 + }, + { + "epoch": 5.451680672268908, + "grad_norm": 0.8321903503583684, + "learning_rate": 5.679353989347989e-06, + "loss": 0.1124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10639987140893936, + "step": 5190, + "valid_targets_mean": 1381.7, + "valid_targets_min": 686 + }, + { + "epoch": 5.456932773109243, + "grad_norm": 0.8907023908538945, + "learning_rate": 5.642834161377408e-06, + "loss": 0.1107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11636456847190857, + "step": 5195, + "valid_targets_mean": 1472.2, + "valid_targets_min": 713 + }, + { + "epoch": 5.46218487394958, + "grad_norm": 1.304588509837649, + "learning_rate": 5.606412834290121e-06, + "loss": 0.1349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2201894223690033, + "step": 5200, + "valid_targets_mean": 2248.9, + "valid_targets_min": 920 + }, + { + "epoch": 5.467436974789916, + "grad_norm": 1.075014973641161, + "learning_rate": 5.570090257963645e-06, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2059841752052307, + "step": 5205, + "valid_targets_mean": 3664.4, + "valid_targets_min": 746 + }, + { + "epoch": 5.472689075630252, + "grad_norm": 0.8446914601975626, + "learning_rate": 5.533866681597981e-06, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20752739906311035, + "step": 5210, + "valid_targets_mean": 3591.1, + "valid_targets_min": 1635 + }, + { + "epoch": 5.477941176470588, + "grad_norm": 0.6615718547643281, + "learning_rate": 5.497742353713928e-06, + "loss": 0.1996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1898113489151001, + "step": 5215, + "valid_targets_mean": 4172.7, + "valid_targets_min": 777 + }, + { + "epoch": 5.483193277310924, + "grad_norm": 0.6331051011117288, + "learning_rate": 5.461717522151371e-06, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2003975212574005, + "step": 5220, + "valid_targets_mean": 4040.2, + "valid_targets_min": 2187 + }, + { + "epoch": 5.48844537815126, + "grad_norm": 0.6969541078527951, + "learning_rate": 5.4257924340675535e-06, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18878155946731567, + "step": 5225, + "valid_targets_mean": 3156.3, + "valid_targets_min": 1432 + }, + { + "epoch": 5.493697478991597, + "grad_norm": 0.5916400175296245, + "learning_rate": 5.389967335935429e-06, + "loss": 0.189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17551878094673157, + "step": 5230, + "valid_targets_mean": 3529.7, + "valid_targets_min": 767 + }, + { + "epoch": 5.498949579831933, + "grad_norm": 0.624324534174277, + "learning_rate": 5.3542424735419306e-06, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17499442398548126, + "step": 5235, + "valid_targets_mean": 3485.6, + "valid_targets_min": 1488 + }, + { + "epoch": 5.504201680672269, + "grad_norm": 0.6222898231435748, + "learning_rate": 5.318618091986294e-06, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17875801026821136, + "step": 5240, + "valid_targets_mean": 3707.0, + "valid_targets_min": 578 + }, + { + "epoch": 5.509453781512605, + "grad_norm": 0.6245877514119683, + "learning_rate": 5.283094435678398e-06, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19261011481285095, + "step": 5245, + "valid_targets_mean": 4252.4, + "valid_targets_min": 1241 + }, + { + "epoch": 5.514705882352941, + "grad_norm": 0.6826116098814055, + "learning_rate": 5.2476717483370685e-06, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21187934279441833, + "step": 5250, + "valid_targets_mean": 4029.2, + "valid_targets_min": 807 + }, + { + "epoch": 5.519957983193278, + "grad_norm": 0.5667810450047195, + "learning_rate": 5.212350272988394e-06, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17751820385456085, + "step": 5255, + "valid_targets_mean": 4507.0, + "valid_targets_min": 2011 + }, + { + "epoch": 5.525210084033613, + "grad_norm": 0.6114125208911464, + "learning_rate": 5.177130251964088e-06, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17235472798347473, + "step": 5260, + "valid_targets_mean": 4268.2, + "valid_targets_min": 1329 + }, + { + "epoch": 5.53046218487395, + "grad_norm": 0.6625902480426628, + "learning_rate": 5.1420119268998085e-06, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18950673937797546, + "step": 5265, + "valid_targets_mean": 3858.9, + "valid_targets_min": 1366 + }, + { + "epoch": 5.535714285714286, + "grad_norm": 0.6651401874085916, + "learning_rate": 5.106995538733488e-06, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17550374567508698, + "step": 5270, + "valid_targets_mean": 3169.1, + "valid_targets_min": 1334 + }, + { + "epoch": 5.540966386554622, + "grad_norm": 0.6231114187474077, + "learning_rate": 5.07208132770372e-06, + "loss": 0.1999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18051046133041382, + "step": 5275, + "valid_targets_mean": 4130.4, + "valid_targets_min": 1762 + }, + { + "epoch": 5.546218487394958, + "grad_norm": 0.6708862310223077, + "learning_rate": 5.037269533348075e-06, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19089899957180023, + "step": 5280, + "valid_targets_mean": 3285.0, + "valid_targets_min": 855 + }, + { + "epoch": 5.551470588235294, + "grad_norm": 0.7236660627737345, + "learning_rate": 5.002560394501463e-06, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2103574275970459, + "step": 5285, + "valid_targets_mean": 3298.2, + "valid_targets_min": 1269 + }, + { + "epoch": 5.55672268907563, + "grad_norm": 0.7530081580597202, + "learning_rate": 4.9679541492945135e-06, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1874314844608307, + "step": 5290, + "valid_targets_mean": 2740.4, + "valid_targets_min": 1041 + }, + { + "epoch": 5.561974789915967, + "grad_norm": 0.7386136479775732, + "learning_rate": 4.933451035151913e-06, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19821321964263916, + "step": 5295, + "valid_targets_mean": 2995.1, + "valid_targets_min": 1291 + }, + { + "epoch": 5.567226890756302, + "grad_norm": 0.72395057035861, + "learning_rate": 4.899051288790806e-06, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18655264377593994, + "step": 5300, + "valid_targets_mean": 2811.6, + "valid_targets_min": 1133 + }, + { + "epoch": 5.572478991596639, + "grad_norm": 0.7470437002909036, + "learning_rate": 4.86475514621916e-06, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19498442113399506, + "step": 5305, + "valid_targets_mean": 3709.5, + "valid_targets_min": 865 + }, + { + "epoch": 5.5777310924369745, + "grad_norm": 0.7302797839356747, + "learning_rate": 4.830562842734112e-06, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18496988713741302, + "step": 5310, + "valid_targets_mean": 2960.6, + "valid_targets_min": 958 + }, + { + "epoch": 5.582983193277311, + "grad_norm": 0.7028513017526096, + "learning_rate": 4.796474612920421e-06, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19564983248710632, + "step": 5315, + "valid_targets_mean": 3189.2, + "valid_targets_min": 771 + }, + { + "epoch": 5.588235294117647, + "grad_norm": 0.6143075180595077, + "learning_rate": 4.762490690648813e-06, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1765946000814438, + "step": 5320, + "valid_targets_mean": 3823.1, + "valid_targets_min": 1631 + }, + { + "epoch": 5.593487394957983, + "grad_norm": 0.7331719157365119, + "learning_rate": 4.728611309074374e-06, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15561848878860474, + "step": 5325, + "valid_targets_mean": 2639.4, + "valid_targets_min": 908 + }, + { + "epoch": 5.598739495798319, + "grad_norm": 0.799889226199126, + "learning_rate": 4.694836700634975e-06, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20055752992630005, + "step": 5330, + "valid_targets_mean": 2988.3, + "valid_targets_min": 1264 + }, + { + "epoch": 5.6039915966386555, + "grad_norm": 0.7723682554917705, + "learning_rate": 4.661167097049668e-06, + "loss": 0.1886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19730721414089203, + "step": 5335, + "valid_targets_mean": 3563.1, + "valid_targets_min": 1889 + }, + { + "epoch": 5.609243697478991, + "grad_norm": 0.6899888581366523, + "learning_rate": 4.62760272931708e-06, + "loss": 0.1879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17451027035713196, + "step": 5340, + "valid_targets_mean": 3737.9, + "valid_targets_min": 1075 + }, + { + "epoch": 5.614495798319328, + "grad_norm": 0.7422252794353547, + "learning_rate": 4.594143827713844e-06, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17348533868789673, + "step": 5345, + "valid_targets_mean": 2745.8, + "valid_targets_min": 1036 + }, + { + "epoch": 5.619747899159664, + "grad_norm": 0.7524666990737761, + "learning_rate": 4.56079062179303e-06, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1979708969593048, + "step": 5350, + "valid_targets_mean": 3516.4, + "valid_targets_min": 1162 + }, + { + "epoch": 5.625, + "grad_norm": 0.6922592034045233, + "learning_rate": 4.527543340382538e-06, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17850226163864136, + "step": 5355, + "valid_targets_mean": 3279.4, + "valid_targets_min": 1281 + }, + { + "epoch": 5.630252100840336, + "grad_norm": 0.6631764263997426, + "learning_rate": 4.494402211583569e-06, + "loss": 0.1855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16947326064109802, + "step": 5360, + "valid_targets_mean": 3180.4, + "valid_targets_min": 1816 + }, + { + "epoch": 5.635504201680672, + "grad_norm": 0.7291414927160216, + "learning_rate": 4.461367462769009e-06, + "loss": 0.1959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18259739875793457, + "step": 5365, + "valid_targets_mean": 2939.2, + "valid_targets_min": 1629 + }, + { + "epoch": 5.640756302521009, + "grad_norm": 0.7503072179394285, + "learning_rate": 4.428439320581923e-06, + "loss": 0.1934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1907494068145752, + "step": 5370, + "valid_targets_mean": 2755.7, + "valid_targets_min": 1254 + }, + { + "epoch": 5.6460084033613445, + "grad_norm": 0.7463064226667883, + "learning_rate": 4.3956180109339705e-06, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19494910538196564, + "step": 5375, + "valid_targets_mean": 3390.2, + "valid_targets_min": 1973 + }, + { + "epoch": 5.651260504201681, + "grad_norm": 0.7574766718662052, + "learning_rate": 4.362903759003849e-06, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.175164133310318, + "step": 5380, + "valid_targets_mean": 2960.4, + "valid_targets_min": 1195 + }, + { + "epoch": 5.656512605042017, + "grad_norm": 0.7408855986734103, + "learning_rate": 4.33029678923576e-06, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19168353080749512, + "step": 5385, + "valid_targets_mean": 3181.4, + "valid_targets_min": 1244 + }, + { + "epoch": 5.661764705882353, + "grad_norm": 0.7242394613336414, + "learning_rate": 4.297797325337889e-06, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1881379634141922, + "step": 5390, + "valid_targets_mean": 3464.2, + "valid_targets_min": 1421 + }, + { + "epoch": 5.667016806722689, + "grad_norm": 0.729199912715331, + "learning_rate": 4.265405590280822e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18175436556339264, + "step": 5395, + "valid_targets_mean": 3443.1, + "valid_targets_min": 1097 + }, + { + "epoch": 5.6722689075630255, + "grad_norm": 0.7780886091437504, + "learning_rate": 4.233121806296072e-06, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20643776655197144, + "step": 5400, + "valid_targets_mean": 3891.4, + "valid_targets_min": 1168 + }, + { + "epoch": 5.677521008403361, + "grad_norm": 2.0082601362553083, + "learning_rate": 4.200946194874514e-06, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19502270221710205, + "step": 5405, + "valid_targets_mean": 3368.9, + "valid_targets_min": 1700 + }, + { + "epoch": 5.682773109243698, + "grad_norm": 0.7542156394780143, + "learning_rate": 4.168878976764872e-06, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17632341384887695, + "step": 5410, + "valid_targets_mean": 2497.1, + "valid_targets_min": 1261 + }, + { + "epoch": 5.688025210084033, + "grad_norm": 0.7204058422362926, + "learning_rate": 4.136920371972228e-06, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1968623399734497, + "step": 5415, + "valid_targets_mean": 3391.2, + "valid_targets_min": 1124 + }, + { + "epoch": 5.69327731092437, + "grad_norm": 0.7514367050409141, + "learning_rate": 4.105070599756482e-06, + "loss": 0.1906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1881876438856125, + "step": 5420, + "valid_targets_mean": 3003.8, + "valid_targets_min": 1361 + }, + { + "epoch": 5.698529411764706, + "grad_norm": 0.671835209360093, + "learning_rate": 4.073329878630856e-06, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16645438969135284, + "step": 5425, + "valid_targets_mean": 3239.7, + "valid_targets_min": 1679 + }, + { + "epoch": 5.703781512605042, + "grad_norm": 0.7151451061634367, + "learning_rate": 4.041698426360425e-06, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.190008282661438, + "step": 5430, + "valid_targets_mean": 3101.2, + "valid_targets_min": 1739 + }, + { + "epoch": 5.709033613445378, + "grad_norm": 0.8510471746451064, + "learning_rate": 4.010176459960569e-06, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.172542542219162, + "step": 5435, + "valid_targets_mean": 2844.4, + "valid_targets_min": 1278 + }, + { + "epoch": 5.714285714285714, + "grad_norm": 0.6595419486913193, + "learning_rate": 3.978764195695528e-06, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1725081354379654, + "step": 5440, + "valid_targets_mean": 3563.4, + "valid_targets_min": 1522 + }, + { + "epoch": 5.71953781512605, + "grad_norm": 0.7513983043013762, + "learning_rate": 3.947461849076912e-06, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1882949024438858, + "step": 5445, + "valid_targets_mean": 3589.0, + "valid_targets_min": 464 + }, + { + "epoch": 5.724789915966387, + "grad_norm": 0.6781411607065908, + "learning_rate": 3.916269634862193e-06, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1876165270805359, + "step": 5450, + "valid_targets_mean": 3827.1, + "valid_targets_min": 1003 + }, + { + "epoch": 5.730042016806722, + "grad_norm": 0.7521498102078784, + "learning_rate": 3.885187767053269e-06, + "loss": 0.1826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18409094214439392, + "step": 5455, + "valid_targets_mean": 3042.1, + "valid_targets_min": 1446 + }, + { + "epoch": 5.735294117647059, + "grad_norm": 0.6989125002761396, + "learning_rate": 3.854216458894973e-06, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18674632906913757, + "step": 5460, + "valid_targets_mean": 3257.8, + "valid_targets_min": 1256 + }, + { + "epoch": 5.740546218487395, + "grad_norm": 0.7363983943406658, + "learning_rate": 3.823355922873606e-06, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20742374658584595, + "step": 5465, + "valid_targets_mean": 3858.5, + "valid_targets_min": 1336 + }, + { + "epoch": 5.745798319327731, + "grad_norm": 0.6688145340655683, + "learning_rate": 3.792606370715508e-06, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18492421507835388, + "step": 5470, + "valid_targets_mean": 3821.1, + "valid_targets_min": 1275 + }, + { + "epoch": 5.751050420168067, + "grad_norm": 0.7198758184286466, + "learning_rate": 3.761968013385584e-06, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1842348277568817, + "step": 5475, + "valid_targets_mean": 2914.6, + "valid_targets_min": 644 + }, + { + "epoch": 5.756302521008403, + "grad_norm": 0.71314811892582, + "learning_rate": 3.731441061085843e-06, + "loss": 0.1799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17556104063987732, + "step": 5480, + "valid_targets_mean": 3083.4, + "valid_targets_min": 1397 + }, + { + "epoch": 5.76155462184874, + "grad_norm": 0.7631003279463335, + "learning_rate": 3.701025723253988e-06, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17353874444961548, + "step": 5485, + "valid_targets_mean": 2747.4, + "valid_targets_min": 1557 + }, + { + "epoch": 5.766806722689076, + "grad_norm": 0.7552536091187607, + "learning_rate": 3.67072220856197e-06, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18228024244308472, + "step": 5490, + "valid_targets_mean": 3405.2, + "valid_targets_min": 1405 + }, + { + "epoch": 5.772058823529412, + "grad_norm": 0.7052066859512669, + "learning_rate": 3.640530724914517e-06, + "loss": 0.1817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17114409804344177, + "step": 5495, + "valid_targets_mean": 3507.1, + "valid_targets_min": 1042 + }, + { + "epoch": 5.777310924369748, + "grad_norm": 0.7275066681746295, + "learning_rate": 3.61045147944777e-06, + "loss": 0.1661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1737072765827179, + "step": 5500, + "valid_targets_mean": 3079.7, + "valid_targets_min": 858 + }, + { + "epoch": 5.782563025210084, + "grad_norm": 0.738557032225339, + "learning_rate": 3.5804846785278315e-06, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17162227630615234, + "step": 5505, + "valid_targets_mean": 2934.3, + "valid_targets_min": 1288 + }, + { + "epoch": 5.78781512605042, + "grad_norm": 0.6881422147874104, + "learning_rate": 3.5506305277493326e-06, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17860713601112366, + "step": 5510, + "valid_targets_mean": 3428.2, + "valid_targets_min": 1915 + }, + { + "epoch": 5.793067226890757, + "grad_norm": 0.8082399213846237, + "learning_rate": 3.520889231934057e-06, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2075139284133911, + "step": 5515, + "valid_targets_mean": 3764.9, + "valid_targets_min": 1244 + }, + { + "epoch": 5.798319327731092, + "grad_norm": 0.7470662206853251, + "learning_rate": 3.4912609951295084e-06, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18600235879421234, + "step": 5520, + "valid_targets_mean": 3006.5, + "valid_targets_min": 1255 + }, + { + "epoch": 5.803571428571429, + "grad_norm": 0.7396369320045814, + "learning_rate": 3.461746020607526e-06, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21041221916675568, + "step": 5525, + "valid_targets_mean": 3326.2, + "valid_targets_min": 1586 + }, + { + "epoch": 5.8088235294117645, + "grad_norm": 0.6985035686259948, + "learning_rate": 3.432344510862895e-06, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16554896533489227, + "step": 5530, + "valid_targets_mean": 3205.2, + "valid_targets_min": 1181 + }, + { + "epoch": 5.814075630252101, + "grad_norm": 0.7863118868734756, + "learning_rate": 3.403056667611917e-06, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17784982919692993, + "step": 5535, + "valid_targets_mean": 2735.1, + "valid_targets_min": 974 + }, + { + "epoch": 5.819327731092437, + "grad_norm": 0.740543140779203, + "learning_rate": 3.373882691791088e-06, + "loss": 0.1821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19035407900810242, + "step": 5540, + "valid_targets_mean": 3735.8, + "valid_targets_min": 1681 + }, + { + "epoch": 5.824579831932773, + "grad_norm": 0.774595013152137, + "learning_rate": 3.344822783555679e-06, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20152834057807922, + "step": 5545, + "valid_targets_mean": 2883.7, + "valid_targets_min": 1171 + }, + { + "epoch": 5.829831932773109, + "grad_norm": 0.807774445390228, + "learning_rate": 3.315877142278363e-06, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21530133485794067, + "step": 5550, + "valid_targets_mean": 3324.1, + "valid_targets_min": 1455 + }, + { + "epoch": 5.8350840336134455, + "grad_norm": 0.6932108713568098, + "learning_rate": 3.287045966547866e-06, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19992104172706604, + "step": 5555, + "valid_targets_mean": 3948.6, + "valid_targets_min": 1529 + }, + { + "epoch": 5.840336134453781, + "grad_norm": 0.7588685281732017, + "learning_rate": 3.258329454167599e-06, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20404741168022156, + "step": 5560, + "valid_targets_mean": 3419.9, + "valid_targets_min": 1450 + }, + { + "epoch": 5.845588235294118, + "grad_norm": 0.6654008948279083, + "learning_rate": 3.2297278021542833e-06, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1653444766998291, + "step": 5565, + "valid_targets_mean": 3465.6, + "valid_targets_min": 1431 + }, + { + "epoch": 5.850840336134453, + "grad_norm": 0.7617245678374055, + "learning_rate": 3.201241206736614e-06, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19317778944969177, + "step": 5570, + "valid_targets_mean": 3085.4, + "valid_targets_min": 802 + }, + { + "epoch": 5.85609243697479, + "grad_norm": 0.7601710209752651, + "learning_rate": 3.1728698633539266e-06, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1923319697380066, + "step": 5575, + "valid_targets_mean": 3230.6, + "valid_targets_min": 1433 + }, + { + "epoch": 5.8613445378151265, + "grad_norm": 0.8026621042498756, + "learning_rate": 3.1446139666548212e-06, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19206663966178894, + "step": 5580, + "valid_targets_mean": 2908.7, + "valid_targets_min": 614 + }, + { + "epoch": 5.866596638655462, + "grad_norm": 0.8143821059911558, + "learning_rate": 3.1164737104958707e-06, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2039254605770111, + "step": 5585, + "valid_targets_mean": 2748.8, + "valid_targets_min": 637 + }, + { + "epoch": 5.871848739495798, + "grad_norm": 0.7917201537427566, + "learning_rate": 3.0884492879402452e-06, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2050473690032959, + "step": 5590, + "valid_targets_mean": 3297.7, + "valid_targets_min": 1683 + }, + { + "epoch": 5.8771008403361344, + "grad_norm": 0.7656092656857109, + "learning_rate": 3.0605408912564184e-06, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18506518006324768, + "step": 5595, + "valid_targets_mean": 3406.1, + "valid_targets_min": 1913 + }, + { + "epoch": 5.882352941176471, + "grad_norm": 0.7130368872222834, + "learning_rate": 3.032748711916851e-06, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18892902135849, + "step": 5600, + "valid_targets_mean": 3848.1, + "valid_targets_min": 1659 + }, + { + "epoch": 5.887605042016807, + "grad_norm": 0.7127010208647905, + "learning_rate": 3.0050729405966495e-06, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17573906481266022, + "step": 5605, + "valid_targets_mean": 3390.7, + "valid_targets_min": 1390 + }, + { + "epoch": 5.892857142857143, + "grad_norm": 0.7730291184647539, + "learning_rate": 2.9775137671722754e-06, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2364846169948578, + "step": 5610, + "valid_targets_mean": 3338.5, + "valid_targets_min": 611 + }, + { + "epoch": 5.898109243697479, + "grad_norm": 0.7437425304974528, + "learning_rate": 2.9500713807202565e-06, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1972287893295288, + "step": 5615, + "valid_targets_mean": 3389.6, + "valid_targets_min": 1451 + }, + { + "epoch": 5.9033613445378155, + "grad_norm": 0.7338868027074469, + "learning_rate": 2.9227459695158543e-06, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.165279358625412, + "step": 5620, + "valid_targets_mean": 3097.9, + "valid_targets_min": 1459 + }, + { + "epoch": 5.908613445378151, + "grad_norm": 0.8398618395284945, + "learning_rate": 2.895537721031809e-06, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20163550972938538, + "step": 5625, + "valid_targets_mean": 2619.4, + "valid_targets_min": 999 + }, + { + "epoch": 5.913865546218488, + "grad_norm": 0.6872190021050335, + "learning_rate": 2.8684468219370364e-06, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17330724000930786, + "step": 5630, + "valid_targets_mean": 3791.2, + "valid_targets_min": 1564 + }, + { + "epoch": 5.919117647058823, + "grad_norm": 0.7915970239951657, + "learning_rate": 2.841473458095334e-06, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19606292247772217, + "step": 5635, + "valid_targets_mean": 3223.3, + "valid_targets_min": 1233 + }, + { + "epoch": 5.92436974789916, + "grad_norm": 0.7371089384298193, + "learning_rate": 2.8146178145641355e-06, + "loss": 0.178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18127654492855072, + "step": 5640, + "valid_targets_mean": 3439.2, + "valid_targets_min": 1524 + }, + { + "epoch": 5.929621848739496, + "grad_norm": 0.7111215478308646, + "learning_rate": 2.787880075593217e-06, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18391185998916626, + "step": 5645, + "valid_targets_mean": 3744.9, + "valid_targets_min": 1375 + }, + { + "epoch": 5.934873949579832, + "grad_norm": 0.7280994223942271, + "learning_rate": 2.7612604246234355e-06, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17705269157886505, + "step": 5650, + "valid_targets_mean": 3036.2, + "valid_targets_min": 906 + }, + { + "epoch": 5.940126050420168, + "grad_norm": 0.6242246207880907, + "learning_rate": 2.7347590442854867e-06, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1647983193397522, + "step": 5655, + "valid_targets_mean": 3766.5, + "valid_targets_min": 1246 + }, + { + "epoch": 5.945378151260504, + "grad_norm": 0.7653759814637277, + "learning_rate": 2.7083761163986433e-06, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20250293612480164, + "step": 5660, + "valid_targets_mean": 3225.9, + "valid_targets_min": 1333 + }, + { + "epoch": 5.95063025210084, + "grad_norm": 0.7475006568229906, + "learning_rate": 2.682111821969493e-06, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19375839829444885, + "step": 5665, + "valid_targets_mean": 3751.3, + "valid_targets_min": 2264 + }, + { + "epoch": 5.955882352941177, + "grad_norm": 0.737881904148586, + "learning_rate": 2.6559663411907233e-06, + "loss": 0.2019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1940421611070633, + "step": 5670, + "valid_targets_mean": 3505.2, + "valid_targets_min": 1454 + }, + { + "epoch": 5.961134453781512, + "grad_norm": 0.7664181534172059, + "learning_rate": 2.629939853439856e-06, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18167239427566528, + "step": 5675, + "valid_targets_mean": 3385.7, + "valid_targets_min": 1430 + }, + { + "epoch": 5.966386554621849, + "grad_norm": 0.7497306698153924, + "learning_rate": 2.604032537278052e-06, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1727236807346344, + "step": 5680, + "valid_targets_mean": 3201.6, + "valid_targets_min": 1433 + }, + { + "epoch": 5.9716386554621845, + "grad_norm": 0.7405178705998576, + "learning_rate": 2.5782445704488447e-06, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1837659776210785, + "step": 5685, + "valid_targets_mean": 3362.1, + "valid_targets_min": 1512 + }, + { + "epoch": 5.976890756302521, + "grad_norm": 0.697865915003433, + "learning_rate": 2.5525761298769515e-06, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18182769417762756, + "step": 5690, + "valid_targets_mean": 3643.9, + "valid_targets_min": 2048 + }, + { + "epoch": 5.982142857142857, + "grad_norm": 0.7426309465732394, + "learning_rate": 2.527027391667054e-06, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1931290328502655, + "step": 5695, + "valid_targets_mean": 3107.1, + "valid_targets_min": 815 + }, + { + "epoch": 5.987394957983193, + "grad_norm": 0.7413852866874329, + "learning_rate": 2.501598531102587e-06, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20602191984653473, + "step": 5700, + "valid_targets_mean": 3339.4, + "valid_targets_min": 1878 + }, + { + "epoch": 5.992647058823529, + "grad_norm": 0.8082149132119589, + "learning_rate": 2.47628972264452e-06, + "loss": 0.1972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21446332335472107, + "step": 5705, + "valid_targets_mean": 3394.6, + "valid_targets_min": 1727 + }, + { + "epoch": 5.9978991596638656, + "grad_norm": 0.673717887910672, + "learning_rate": 2.4511011399301944e-06, + "loss": 0.1884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1594567894935608, + "step": 5710, + "valid_targets_mean": 2924.3, + "valid_targets_min": 1622 + }, + { + "epoch": 6.003151260504202, + "grad_norm": 0.888726685758785, + "learning_rate": 2.4260329557721107e-06, + "loss": 0.1686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1586988866329193, + "step": 5715, + "valid_targets_mean": 1587.9, + "valid_targets_min": 788 + }, + { + "epoch": 6.008403361344538, + "grad_norm": 0.7892723724374661, + "learning_rate": 2.4010853421567237e-06, + "loss": 0.131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12804922461509705, + "step": 5720, + "valid_targets_mean": 1591.2, + "valid_targets_min": 957 + }, + { + "epoch": 6.013655462184874, + "grad_norm": 0.7624099131843867, + "learning_rate": 2.376258470243309e-06, + "loss": 0.1218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11783352494239807, + "step": 5725, + "valid_targets_mean": 1418.7, + "valid_targets_min": 796 + }, + { + "epoch": 6.01890756302521, + "grad_norm": 0.7792462436940873, + "learning_rate": 2.351552510362758e-06, + "loss": 0.1228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11473619192838669, + "step": 5730, + "valid_targets_mean": 1389.0, + "valid_targets_min": 704 + }, + { + "epoch": 6.024159663865547, + "grad_norm": 0.8703079476348818, + "learning_rate": 2.326967632016406e-06, + "loss": 0.1226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11579575389623642, + "step": 5735, + "valid_targets_mean": 1513.9, + "valid_targets_min": 638 + }, + { + "epoch": 6.029411764705882, + "grad_norm": 0.8915876673258326, + "learning_rate": 2.302504003874888e-06, + "loss": 0.1241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12122771143913269, + "step": 5740, + "valid_targets_mean": 1482.3, + "valid_targets_min": 727 + }, + { + "epoch": 6.034663865546219, + "grad_norm": 0.8177570400824066, + "learning_rate": 2.2781617937769675e-06, + "loss": 0.1256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10171643644571304, + "step": 5745, + "valid_targets_mean": 1189.1, + "valid_targets_min": 655 + }, + { + "epoch": 6.0399159663865545, + "grad_norm": 1.025573849186844, + "learning_rate": 2.2539411687283907e-06, + "loss": 0.1108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1116051971912384, + "step": 5750, + "valid_targets_mean": 1299.6, + "valid_targets_min": 714 + }, + { + "epoch": 6.045168067226891, + "grad_norm": 0.873608647534603, + "learning_rate": 2.229842294900744e-06, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1157689318060875, + "step": 5755, + "valid_targets_mean": 1407.6, + "valid_targets_min": 741 + }, + { + "epoch": 6.050420168067227, + "grad_norm": 0.9798749449474746, + "learning_rate": 2.205865337630302e-06, + "loss": 0.1219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13768497109413147, + "step": 5760, + "valid_targets_mean": 1544.1, + "valid_targets_min": 663 + }, + { + "epoch": 6.055672268907563, + "grad_norm": 0.916099508349051, + "learning_rate": 2.182010461416899e-06, + "loss": 0.115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11846692860126495, + "step": 5765, + "valid_targets_mean": 1511.1, + "valid_targets_min": 632 + }, + { + "epoch": 6.060924369747899, + "grad_norm": 0.8853677750337303, + "learning_rate": 2.158277829922817e-06, + "loss": 0.1186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12952449917793274, + "step": 5770, + "valid_targets_mean": 1638.0, + "valid_targets_min": 933 + }, + { + "epoch": 6.0661764705882355, + "grad_norm": 0.7694796846656141, + "learning_rate": 2.1346676059716253e-06, + "loss": 0.1188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11292778700590134, + "step": 5775, + "valid_targets_mean": 1526.9, + "valid_targets_min": 717 + }, + { + "epoch": 6.071428571428571, + "grad_norm": 0.8108257454301182, + "learning_rate": 2.1111799515471066e-06, + "loss": 0.1079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1118139773607254, + "step": 5780, + "valid_targets_mean": 1387.4, + "valid_targets_min": 659 + }, + { + "epoch": 6.076680672268908, + "grad_norm": 0.816088021557387, + "learning_rate": 2.0878150277921195e-06, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1065862700343132, + "step": 5785, + "valid_targets_mean": 1382.9, + "valid_targets_min": 642 + }, + { + "epoch": 6.081932773109243, + "grad_norm": 0.8469714058115316, + "learning_rate": 2.064572995007492e-06, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09756132960319519, + "step": 5790, + "valid_targets_mean": 1312.1, + "valid_targets_min": 765 + }, + { + "epoch": 6.08718487394958, + "grad_norm": 0.9292126584824716, + "learning_rate": 2.0414540126509274e-06, + "loss": 0.1195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11742901057004929, + "step": 5795, + "valid_targets_mean": 1397.6, + "valid_targets_min": 631 + }, + { + "epoch": 6.092436974789916, + "grad_norm": 0.9989157343383366, + "learning_rate": 2.0184582393359253e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12957970798015594, + "step": 5800, + "valid_targets_mean": 1419.4, + "valid_targets_min": 623 + }, + { + "epoch": 6.097689075630252, + "grad_norm": 0.8687109853494948, + "learning_rate": 1.99558583283066e-06, + "loss": 0.118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12820440530776978, + "step": 5805, + "valid_targets_mean": 1647.7, + "valid_targets_min": 818 + }, + { + "epoch": 6.102941176470588, + "grad_norm": 0.8429889291483497, + "learning_rate": 1.9728369500569333e-06, + "loss": 0.1183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11039917171001434, + "step": 5810, + "valid_targets_mean": 1545.5, + "valid_targets_min": 776 + }, + { + "epoch": 6.108193277310924, + "grad_norm": 0.9401593816021946, + "learning_rate": 1.950211747089077e-06, + "loss": 0.1152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11282265186309814, + "step": 5815, + "valid_targets_mean": 1411.8, + "valid_targets_min": 623 + }, + { + "epoch": 6.11344537815126, + "grad_norm": 0.8579915255635823, + "learning_rate": 1.927710379152881e-06, + "loss": 0.1104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11633343994617462, + "step": 5820, + "valid_targets_mean": 1437.3, + "valid_targets_min": 742 + }, + { + "epoch": 6.118697478991597, + "grad_norm": 0.8470196779150363, + "learning_rate": 1.9053330006245452e-06, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0939580649137497, + "step": 5825, + "valid_targets_mean": 1193.6, + "valid_targets_min": 611 + }, + { + "epoch": 6.123949579831933, + "grad_norm": 0.8951203579512895, + "learning_rate": 1.8830797650296006e-06, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11357836425304413, + "step": 5830, + "valid_targets_mean": 1514.0, + "valid_targets_min": 793 + }, + { + "epoch": 6.129201680672269, + "grad_norm": 0.8204950266453027, + "learning_rate": 1.8609508250418628e-06, + "loss": 0.1089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10361876338720322, + "step": 5835, + "valid_targets_mean": 1379.1, + "valid_targets_min": 886 + }, + { + "epoch": 6.1344537815126055, + "grad_norm": 0.827334438729143, + "learning_rate": 1.8389463324823986e-06, + "loss": 0.1121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10381718724966049, + "step": 5840, + "valid_targets_mean": 1455.5, + "valid_targets_min": 991 + }, + { + "epoch": 6.139705882352941, + "grad_norm": 1.0670876809595125, + "learning_rate": 1.81706643831846e-06, + "loss": 0.113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11400048434734344, + "step": 5845, + "valid_targets_mean": 1525.2, + "valid_targets_min": 908 + }, + { + "epoch": 6.144957983193278, + "grad_norm": 0.9783834172435171, + "learning_rate": 1.7953112926624715e-06, + "loss": 0.119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11489921063184738, + "step": 5850, + "valid_targets_mean": 1476.6, + "valid_targets_min": 680 + }, + { + "epoch": 6.150210084033613, + "grad_norm": 0.8766319864539988, + "learning_rate": 1.7736810447709852e-06, + "loss": 0.1092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11097388714551926, + "step": 5855, + "valid_targets_mean": 1467.1, + "valid_targets_min": 803 + }, + { + "epoch": 6.15546218487395, + "grad_norm": 0.8785524728820334, + "learning_rate": 1.7521758430436553e-06, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10426779836416245, + "step": 5860, + "valid_targets_mean": 1341.3, + "valid_targets_min": 765 + }, + { + "epoch": 6.160714285714286, + "grad_norm": 0.8829153848682123, + "learning_rate": 1.7307958350222366e-06, + "loss": 0.1139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11685581505298615, + "step": 5865, + "valid_targets_mean": 1556.4, + "valid_targets_min": 804 + }, + { + "epoch": 6.165966386554622, + "grad_norm": 0.8180165865392631, + "learning_rate": 1.7095411673895545e-06, + "loss": 0.1147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11777027696371078, + "step": 5870, + "valid_targets_mean": 1509.7, + "valid_targets_min": 675 + }, + { + "epoch": 6.171218487394958, + "grad_norm": 0.8375114382633897, + "learning_rate": 1.6884119859685034e-06, + "loss": 0.1098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11995431780815125, + "step": 5875, + "valid_targets_mean": 1408.4, + "valid_targets_min": 882 + }, + { + "epoch": 6.176470588235294, + "grad_norm": 0.7972330265214028, + "learning_rate": 1.6674084357210562e-06, + "loss": 0.1071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10255067050457001, + "step": 5880, + "valid_targets_mean": 1449.8, + "valid_targets_min": 660 + }, + { + "epoch": 6.18172268907563, + "grad_norm": 0.8858809578498907, + "learning_rate": 1.6465306607472608e-06, + "loss": 0.102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10549542307853699, + "step": 5885, + "valid_targets_mean": 1489.8, + "valid_targets_min": 678 + }, + { + "epoch": 6.186974789915967, + "grad_norm": 0.9235725158063691, + "learning_rate": 1.625778804284246e-06, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10902103036642075, + "step": 5890, + "valid_targets_mean": 1379.1, + "valid_targets_min": 676 + }, + { + "epoch": 6.192226890756302, + "grad_norm": 0.8610604938191373, + "learning_rate": 1.6051530087052558e-06, + "loss": 0.1107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10405048727989197, + "step": 5895, + "valid_targets_mean": 1347.1, + "valid_targets_min": 794 + }, + { + "epoch": 6.197478991596639, + "grad_norm": 0.844841243204075, + "learning_rate": 1.5846534155186511e-06, + "loss": 0.1103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1075349748134613, + "step": 5900, + "valid_targets_mean": 1372.6, + "valid_targets_min": 607 + }, + { + "epoch": 6.2027310924369745, + "grad_norm": 0.9299916077128926, + "learning_rate": 1.564280165366956e-06, + "loss": 0.1104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11143974959850311, + "step": 5905, + "valid_targets_mean": 1323.6, + "valid_targets_min": 722 + }, + { + "epoch": 6.207983193277311, + "grad_norm": 0.9520206844500675, + "learning_rate": 1.544033398025886e-06, + "loss": 0.1164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12437736988067627, + "step": 5910, + "valid_targets_mean": 1597.7, + "valid_targets_min": 858 + }, + { + "epoch": 6.213235294117647, + "grad_norm": 0.8264419715144223, + "learning_rate": 1.5239132524033951e-06, + "loss": 0.0983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09098273515701294, + "step": 5915, + "valid_targets_mean": 1270.8, + "valid_targets_min": 689 + }, + { + "epoch": 6.218487394957983, + "grad_norm": 0.9121347316675537, + "learning_rate": 1.50391986653871e-06, + "loss": 0.1151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10781864076852798, + "step": 5920, + "valid_targets_mean": 1388.2, + "valid_targets_min": 692 + }, + { + "epoch": 6.223739495798319, + "grad_norm": 0.8183103322870855, + "learning_rate": 1.4840533776013954e-06, + "loss": 0.1116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0995546281337738, + "step": 5925, + "valid_targets_mean": 1447.2, + "valid_targets_min": 828 + }, + { + "epoch": 6.2289915966386555, + "grad_norm": 0.8695709319336882, + "learning_rate": 1.4643139218904035e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1060028076171875, + "step": 5930, + "valid_targets_mean": 1399.7, + "valid_targets_min": 618 + }, + { + "epoch": 6.234243697478991, + "grad_norm": 0.7936160826191869, + "learning_rate": 1.4447016348331477e-06, + "loss": 0.1041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10275650769472122, + "step": 5935, + "valid_targets_mean": 1454.3, + "valid_targets_min": 629 + }, + { + "epoch": 6.239495798319328, + "grad_norm": 1.0003364025533468, + "learning_rate": 1.4252166509845777e-06, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1305723935365677, + "step": 5940, + "valid_targets_mean": 1746.1, + "valid_targets_min": 1071 + }, + { + "epoch": 6.244747899159663, + "grad_norm": 0.9039823804020426, + "learning_rate": 1.4058591040262259e-06, + "loss": 0.1079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1077783852815628, + "step": 5945, + "valid_targets_mean": 1282.2, + "valid_targets_min": 677 + }, + { + "epoch": 6.25, + "grad_norm": 0.8985676998446693, + "learning_rate": 1.3866291267653309e-06, + "loss": 0.1031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09621434658765793, + "step": 5950, + "valid_targets_mean": 1227.6, + "valid_targets_min": 679 + }, + { + "epoch": 6.255252100840336, + "grad_norm": 0.8526105872349939, + "learning_rate": 1.3675268511339067e-06, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09855210781097412, + "step": 5955, + "valid_targets_mean": 1319.1, + "valid_targets_min": 504 + }, + { + "epoch": 6.260504201680672, + "grad_norm": 0.8620937247837483, + "learning_rate": 1.3485524081878288e-06, + "loss": 0.1069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1124173030257225, + "step": 5960, + "valid_targets_mean": 1616.5, + "valid_targets_min": 1023 + }, + { + "epoch": 6.265756302521009, + "grad_norm": 0.8269435472596893, + "learning_rate": 1.3297059281059533e-06, + "loss": 0.1092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10087428987026215, + "step": 5965, + "valid_targets_mean": 1453.0, + "valid_targets_min": 661 + }, + { + "epoch": 6.2710084033613445, + "grad_norm": 0.8464773093003739, + "learning_rate": 1.3109875401892146e-06, + "loss": 0.111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10120785981416702, + "step": 5970, + "valid_targets_mean": 1442.0, + "valid_targets_min": 889 + }, + { + "epoch": 6.276260504201681, + "grad_norm": 0.809843624015962, + "learning_rate": 1.2923973728597373e-06, + "loss": 0.1037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10031883418560028, + "step": 5975, + "valid_targets_mean": 1548.4, + "valid_targets_min": 723 + }, + { + "epoch": 6.281512605042017, + "grad_norm": 0.8939758957054761, + "learning_rate": 1.2739355536599508e-06, + "loss": 0.1095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10138392448425293, + "step": 5980, + "valid_targets_mean": 1359.8, + "valid_targets_min": 793 + }, + { + "epoch": 6.286764705882353, + "grad_norm": 0.9990516754583353, + "learning_rate": 1.2556022092517362e-06, + "loss": 0.1055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11516615748405457, + "step": 5985, + "valid_targets_mean": 1478.6, + "valid_targets_min": 861 + }, + { + "epoch": 6.292016806722689, + "grad_norm": 0.8483028409836717, + "learning_rate": 1.237397465415524e-06, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0986555963754654, + "step": 5990, + "valid_targets_mean": 1332.7, + "valid_targets_min": 752 + }, + { + "epoch": 6.2972689075630255, + "grad_norm": 0.9848150205283159, + "learning_rate": 1.2193214470494664e-06, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11558929085731506, + "step": 5995, + "valid_targets_mean": 1622.7, + "valid_targets_min": 789 + }, + { + "epoch": 6.302521008403361, + "grad_norm": 0.9238690287611016, + "learning_rate": 1.201374278168541e-06, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08862060308456421, + "step": 6000, + "valid_targets_mean": 1195.4, + "valid_targets_min": 567 + }, + { + "epoch": 6.307773109243698, + "grad_norm": 0.9550625507498628, + "learning_rate": 1.1835560819037429e-06, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12365525960922241, + "step": 6005, + "valid_targets_mean": 1504.5, + "valid_targets_min": 605 + }, + { + "epoch": 6.313025210084033, + "grad_norm": 0.8977013451179156, + "learning_rate": 1.1658669805012112e-06, + "loss": 0.1088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10444406419992447, + "step": 6010, + "valid_targets_mean": 1580.1, + "valid_targets_min": 746 + }, + { + "epoch": 6.31827731092437, + "grad_norm": 0.9626620928600009, + "learning_rate": 1.1483070953213904e-06, + "loss": 0.1017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09486856311559677, + "step": 6015, + "valid_targets_mean": 1158.2, + "valid_targets_min": 723 + }, + { + "epoch": 6.323529411764706, + "grad_norm": 0.857918445866971, + "learning_rate": 1.130876546838211e-06, + "loss": 0.1092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.107955202460289, + "step": 6020, + "valid_targets_mean": 1430.4, + "valid_targets_min": 855 + }, + { + "epoch": 6.328781512605042, + "grad_norm": 0.8446463752304134, + "learning_rate": 1.1135754546382627e-06, + "loss": 0.111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10878735780715942, + "step": 6025, + "valid_targets_mean": 1590.4, + "valid_targets_min": 902 + }, + { + "epoch": 6.334033613445378, + "grad_norm": 0.8359256112732846, + "learning_rate": 1.0964039374199564e-06, + "loss": 0.108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10430100560188293, + "step": 6030, + "valid_targets_mean": 1521.5, + "valid_targets_min": 803 + }, + { + "epoch": 6.339285714285714, + "grad_norm": 0.9402998650985804, + "learning_rate": 1.07936211299273e-06, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11771460622549057, + "step": 6035, + "valid_targets_mean": 1455.2, + "valid_targets_min": 740 + }, + { + "epoch": 6.34453781512605, + "grad_norm": 0.8950360049257687, + "learning_rate": 1.0624500982762376e-06, + "loss": 0.1031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09641949832439423, + "step": 6040, + "valid_targets_mean": 1279.4, + "valid_targets_min": 244 + }, + { + "epoch": 6.349789915966387, + "grad_norm": 0.8797683867630403, + "learning_rate": 1.045668009299523e-06, + "loss": 0.1104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10380536317825317, + "step": 6045, + "valid_targets_mean": 1366.9, + "valid_targets_min": 714 + }, + { + "epoch": 6.355042016806722, + "grad_norm": 0.8500779068020419, + "learning_rate": 1.029015961200266e-06, + "loss": 0.1055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08908183872699738, + "step": 6050, + "valid_targets_mean": 1220.5, + "valid_targets_min": 585 + }, + { + "epoch": 6.360294117647059, + "grad_norm": 0.963766954956786, + "learning_rate": 1.012494068223946e-06, + "loss": 0.1142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11748844385147095, + "step": 6055, + "valid_targets_mean": 1596.4, + "valid_targets_min": 1041 + }, + { + "epoch": 6.3655462184873945, + "grad_norm": 1.013786349356977, + "learning_rate": 9.961024437230948e-07, + "loss": 0.1081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10118795186281204, + "step": 6060, + "valid_targets_mean": 1401.3, + "valid_targets_min": 777 + }, + { + "epoch": 6.370798319327731, + "grad_norm": 0.8907088069055336, + "learning_rate": 9.798412001565038e-07, + "loss": 0.1095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10110174119472504, + "step": 6065, + "valid_targets_mean": 1546.8, + "valid_targets_min": 959 + }, + { + "epoch": 6.376050420168067, + "grad_norm": 0.836878249917925, + "learning_rate": 9.63710449088453e-07, + "loss": 0.1052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0972297191619873, + "step": 6070, + "valid_targets_mean": 1284.4, + "valid_targets_min": 620 + }, + { + "epoch": 6.381302521008403, + "grad_norm": 0.8342602387448971, + "learning_rate": 9.477103011879408e-07, + "loss": 0.1035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0927862748503685, + "step": 6075, + "valid_targets_mean": 1356.6, + "valid_targets_min": 656 + }, + { + "epoch": 6.38655462184874, + "grad_norm": 0.9521526310876444, + "learning_rate": 9.318408662279399e-07, + "loss": 0.1113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12353593111038208, + "step": 6080, + "valid_targets_mean": 1564.5, + "valid_targets_min": 697 + }, + { + "epoch": 6.391806722689076, + "grad_norm": 0.8338141493940605, + "learning_rate": 9.161022530846253e-07, + "loss": 0.0993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09638993442058563, + "step": 6085, + "valid_targets_mean": 1604.7, + "valid_targets_min": 682 + }, + { + "epoch": 6.397058823529412, + "grad_norm": 0.9780093813312944, + "learning_rate": 9.004945697366452e-07, + "loss": 0.1, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09383905678987503, + "step": 6090, + "valid_targets_mean": 1383.4, + "valid_targets_min": 730 + }, + { + "epoch": 6.402310924369748, + "grad_norm": 0.891231287791507, + "learning_rate": 8.850179232643641e-07, + "loss": 0.1047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10458773374557495, + "step": 6095, + "valid_targets_mean": 1482.5, + "valid_targets_min": 747 + }, + { + "epoch": 6.407563025210084, + "grad_norm": 0.8092854746476409, + "learning_rate": 8.696724198491369e-07, + "loss": 0.1152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10784599184989929, + "step": 6100, + "valid_targets_mean": 1603.6, + "valid_targets_min": 933 + }, + { + "epoch": 6.41281512605042, + "grad_norm": 0.9405797225301739, + "learning_rate": 8.544581647725825e-07, + "loss": 0.1025, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10436847805976868, + "step": 6105, + "valid_targets_mean": 1517.9, + "valid_targets_min": 799 + }, + { + "epoch": 6.418067226890757, + "grad_norm": 0.8958196229325652, + "learning_rate": 8.393752624158603e-07, + "loss": 0.102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10442647337913513, + "step": 6110, + "valid_targets_mean": 1516.2, + "valid_targets_min": 909 + }, + { + "epoch": 6.423319327731092, + "grad_norm": 0.8319689597681014, + "learning_rate": 8.244238162589435e-07, + "loss": 0.0991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10206621140241623, + "step": 6115, + "valid_targets_mean": 1606.8, + "valid_targets_min": 741 + }, + { + "epoch": 6.428571428571429, + "grad_norm": 0.8710279467847112, + "learning_rate": 8.096039288799251e-07, + "loss": 0.0907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09631064534187317, + "step": 6120, + "valid_targets_mean": 1328.7, + "valid_targets_min": 720 + }, + { + "epoch": 6.4338235294117645, + "grad_norm": 0.8396789652735168, + "learning_rate": 7.949157019543064e-07, + "loss": 0.1038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10037371516227722, + "step": 6125, + "valid_targets_mean": 1396.2, + "valid_targets_min": 610 + }, + { + "epoch": 6.439075630252101, + "grad_norm": 0.8987178018129869, + "learning_rate": 7.803592362542911e-07, + "loss": 0.0992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10167519003152847, + "step": 6130, + "valid_targets_mean": 1352.9, + "valid_targets_min": 684 + }, + { + "epoch": 6.444327731092437, + "grad_norm": 0.8464981713812435, + "learning_rate": 7.659346316481086e-07, + "loss": 0.0978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09975661337375641, + "step": 6135, + "valid_targets_mean": 1479.4, + "valid_targets_min": 874 + }, + { + "epoch": 6.449579831932773, + "grad_norm": 0.8500419492524814, + "learning_rate": 7.516419870993296e-07, + "loss": 0.1026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09866391867399216, + "step": 6140, + "valid_targets_mean": 1501.2, + "valid_targets_min": 765 + }, + { + "epoch": 6.454831932773109, + "grad_norm": 0.8846021286748634, + "learning_rate": 7.374814006661667e-07, + "loss": 0.1004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0995045006275177, + "step": 6145, + "valid_targets_mean": 1419.9, + "valid_targets_min": 589 + }, + { + "epoch": 6.4600840336134455, + "grad_norm": 0.9586129744458028, + "learning_rate": 7.234529695008241e-07, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1108023077249527, + "step": 6150, + "valid_targets_mean": 1587.2, + "valid_targets_min": 918 + }, + { + "epoch": 6.465336134453781, + "grad_norm": 1.136739780639734, + "learning_rate": 7.095567898488175e-07, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20650501549243927, + "step": 6155, + "valid_targets_mean": 3546.9, + "valid_targets_min": 702 + }, + { + "epoch": 6.470588235294118, + "grad_norm": 1.156568540868847, + "learning_rate": 6.957929570483224e-07, + "loss": 0.2101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2196894884109497, + "step": 6160, + "valid_targets_mean": 4000.6, + "valid_targets_min": 1005 + }, + { + "epoch": 6.475840336134453, + "grad_norm": 1.0910237455501957, + "learning_rate": 6.821615655295155e-07, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19706688821315765, + "step": 6165, + "valid_targets_mean": 4063.9, + "valid_targets_min": 1296 + }, + { + "epoch": 6.48109243697479, + "grad_norm": 0.9433508198462144, + "learning_rate": 6.686627088139253e-07, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18134765326976776, + "step": 6170, + "valid_targets_mean": 3747.0, + "valid_targets_min": 1561 + }, + { + "epoch": 6.486344537815126, + "grad_norm": 0.9443013244173882, + "learning_rate": 6.552964795137895e-07, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.206732839345932, + "step": 6175, + "valid_targets_mean": 3562.5, + "valid_targets_min": 1044 + }, + { + "epoch": 6.491596638655462, + "grad_norm": 0.7981181274477949, + "learning_rate": 6.420629693314317e-07, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1745770275592804, + "step": 6180, + "valid_targets_mean": 3911.4, + "valid_targets_min": 920 + }, + { + "epoch": 6.496848739495798, + "grad_norm": 1.0053288601599724, + "learning_rate": 6.289622690586151e-07, + "loss": 0.1802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21677300333976746, + "step": 6185, + "valid_targets_mean": 5753.8, + "valid_targets_min": 2060 + }, + { + "epoch": 6.5021008403361344, + "grad_norm": 0.8097609883366772, + "learning_rate": 6.159944685759289e-07, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19428494572639465, + "step": 6190, + "valid_targets_mean": 3728.6, + "valid_targets_min": 467 + }, + { + "epoch": 6.507352941176471, + "grad_norm": 0.8064900139004815, + "learning_rate": 6.031596568521792e-07, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20241563022136688, + "step": 6195, + "valid_targets_mean": 3554.0, + "valid_targets_min": 1689 + }, + { + "epoch": 6.512605042016807, + "grad_norm": 0.776184144360134, + "learning_rate": 5.904579219437567e-07, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1855916827917099, + "step": 6200, + "valid_targets_mean": 3708.1, + "valid_targets_min": 1697 + }, + { + "epoch": 6.517857142857143, + "grad_norm": 0.7831624701562231, + "learning_rate": 5.778893509940497e-07, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2027127742767334, + "step": 6205, + "valid_targets_mean": 3652.6, + "valid_targets_min": 1223 + }, + { + "epoch": 6.523109243697479, + "grad_norm": 0.7592030323419526, + "learning_rate": 5.654540302328526e-07, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18956655263900757, + "step": 6210, + "valid_targets_mean": 4036.1, + "valid_targets_min": 1552 + }, + { + "epoch": 6.5283613445378155, + "grad_norm": 0.7395667543125498, + "learning_rate": 5.531520449757465e-07, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1876063048839569, + "step": 6215, + "valid_targets_mean": 3543.5, + "valid_targets_min": 1439 + }, + { + "epoch": 6.533613445378151, + "grad_norm": 0.7046765053549935, + "learning_rate": 5.409834796235447e-07, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1793154776096344, + "step": 6220, + "valid_targets_mean": 3287.0, + "valid_targets_min": 1402 + }, + { + "epoch": 6.538865546218488, + "grad_norm": 0.7352367532176454, + "learning_rate": 5.28948417661701e-07, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19139184057712555, + "step": 6225, + "valid_targets_mean": 3423.9, + "valid_targets_min": 1586 + }, + { + "epoch": 6.544117647058823, + "grad_norm": 0.7217587370221558, + "learning_rate": 5.170469416597223e-07, + "loss": 0.1874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18540339171886444, + "step": 6230, + "valid_targets_mean": 3406.0, + "valid_targets_min": 1538 + }, + { + "epoch": 6.54936974789916, + "grad_norm": 0.8448265208591422, + "learning_rate": 5.05279133270633e-07, + "loss": 0.1863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18816813826560974, + "step": 6235, + "valid_targets_mean": 2954.0, + "valid_targets_min": 1058 + }, + { + "epoch": 6.554621848739496, + "grad_norm": 0.7796897965938916, + "learning_rate": 4.936450732303866e-07, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.198177307844162, + "step": 6240, + "valid_targets_mean": 3051.8, + "valid_targets_min": 1362 + }, + { + "epoch": 6.559873949579832, + "grad_norm": 0.6964707697678552, + "learning_rate": 4.821448413573237e-07, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1926005631685257, + "step": 6245, + "valid_targets_mean": 3620.5, + "valid_targets_min": 1392 + }, + { + "epoch": 6.565126050420168, + "grad_norm": 0.7685525826201484, + "learning_rate": 4.70778516551631e-07, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18368004262447357, + "step": 6250, + "valid_targets_mean": 3103.6, + "valid_targets_min": 1267 + }, + { + "epoch": 6.570378151260504, + "grad_norm": 0.7299698999315323, + "learning_rate": 4.595461767947829e-07, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19365397095680237, + "step": 6255, + "valid_targets_mean": 3592.7, + "valid_targets_min": 1026 + }, + { + "epoch": 6.57563025210084, + "grad_norm": 0.7134849644818172, + "learning_rate": 4.484478991490249e-07, + "loss": 0.1767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18238605558872223, + "step": 6260, + "valid_targets_mean": 3632.1, + "valid_targets_min": 1785 + }, + { + "epoch": 6.580882352941177, + "grad_norm": 0.7566638135613766, + "learning_rate": 4.374837597568338e-07, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17180000245571136, + "step": 6265, + "valid_targets_mean": 3557.1, + "valid_targets_min": 1957 + }, + { + "epoch": 6.586134453781512, + "grad_norm": 0.7212160746237427, + "learning_rate": 4.266538338403936e-07, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17052122950553894, + "step": 6270, + "valid_targets_mean": 3230.6, + "valid_targets_min": 1570 + }, + { + "epoch": 6.591386554621849, + "grad_norm": 0.7590336258795505, + "learning_rate": 4.159581957010894e-07, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18137861788272858, + "step": 6275, + "valid_targets_mean": 3023.7, + "valid_targets_min": 997 + }, + { + "epoch": 6.5966386554621845, + "grad_norm": 0.7939812548685019, + "learning_rate": 4.0539691871898545e-07, + "loss": 0.1811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20064428448677063, + "step": 6280, + "valid_targets_mean": 3056.6, + "valid_targets_min": 1293 + }, + { + "epoch": 6.601890756302521, + "grad_norm": 0.7459390024198103, + "learning_rate": 3.949700753523344e-07, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16911663115024567, + "step": 6285, + "valid_targets_mean": 2915.6, + "valid_targets_min": 1687 + }, + { + "epoch": 6.607142857142857, + "grad_norm": 0.8123891604766348, + "learning_rate": 3.8467773713707134e-07, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19759421050548553, + "step": 6290, + "valid_targets_mean": 2985.1, + "valid_targets_min": 1108 + }, + { + "epoch": 6.612394957983193, + "grad_norm": 0.7820274649796405, + "learning_rate": 3.7451997468632486e-07, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17913171648979187, + "step": 6295, + "valid_targets_mean": 3195.0, + "valid_targets_min": 571 + }, + { + "epoch": 6.617647058823529, + "grad_norm": 0.7958593118852751, + "learning_rate": 3.6449685768993327e-07, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19205614924430847, + "step": 6300, + "valid_targets_mean": 3106.2, + "valid_targets_min": 738 + }, + { + "epoch": 6.6228991596638656, + "grad_norm": 0.727974701275133, + "learning_rate": 3.5460845491396944e-07, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1842992901802063, + "step": 6305, + "valid_targets_mean": 3189.5, + "valid_targets_min": 1896 + }, + { + "epoch": 6.628151260504202, + "grad_norm": 0.737246932634588, + "learning_rate": 3.448548342002589e-07, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18911173939704895, + "step": 6310, + "valid_targets_mean": 3278.9, + "valid_targets_min": 1177 + }, + { + "epoch": 6.633403361344538, + "grad_norm": 0.7200101677641911, + "learning_rate": 3.352360624659312e-07, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19001314043998718, + "step": 6315, + "valid_targets_mean": 3398.8, + "valid_targets_min": 1270 + }, + { + "epoch": 6.6386554621848735, + "grad_norm": 0.6991330740624144, + "learning_rate": 3.2575220570294276e-07, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1835232377052307, + "step": 6320, + "valid_targets_mean": 3829.2, + "valid_targets_min": 1668 + }, + { + "epoch": 6.64390756302521, + "grad_norm": 0.6942560490970551, + "learning_rate": 3.164033289776369e-07, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18814723193645477, + "step": 6325, + "valid_targets_mean": 3426.9, + "valid_targets_min": 1184 + }, + { + "epoch": 6.649159663865547, + "grad_norm": 0.7240553846645934, + "learning_rate": 3.071894964302935e-07, + "loss": 0.1851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19533832371234894, + "step": 6330, + "valid_targets_mean": 3523.6, + "valid_targets_min": 1858 + }, + { + "epoch": 6.654411764705882, + "grad_norm": 0.6784877207352158, + "learning_rate": 2.981107712746867e-07, + "loss": 0.181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16024400293827057, + "step": 6335, + "valid_targets_mean": 3161.2, + "valid_targets_min": 922 + }, + { + "epoch": 6.659663865546219, + "grad_norm": 0.8123686861228415, + "learning_rate": 2.891672157976522e-07, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.212624192237854, + "step": 6340, + "valid_targets_mean": 2948.9, + "valid_targets_min": 1812 + }, + { + "epoch": 6.6649159663865545, + "grad_norm": 0.7276516066486569, + "learning_rate": 2.803588913586608e-07, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17187952995300293, + "step": 6345, + "valid_targets_mean": 3107.7, + "valid_targets_min": 977 + }, + { + "epoch": 6.670168067226891, + "grad_norm": 0.6887997469688529, + "learning_rate": 2.716858583894033e-07, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16475774347782135, + "step": 6350, + "valid_targets_mean": 3372.6, + "valid_targets_min": 810 + }, + { + "epoch": 6.675420168067227, + "grad_norm": 0.7520402794788272, + "learning_rate": 2.6314817639335964e-07, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17598888278007507, + "step": 6355, + "valid_targets_mean": 3305.2, + "valid_targets_min": 1673 + }, + { + "epoch": 6.680672268907563, + "grad_norm": 0.6523534201108536, + "learning_rate": 2.547459039454103e-07, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15471762418746948, + "step": 6360, + "valid_targets_mean": 3933.0, + "valid_targets_min": 1869 + }, + { + "epoch": 6.685924369747899, + "grad_norm": 0.8552212655149624, + "learning_rate": 2.4647909869142117e-07, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20678521692752838, + "step": 6365, + "valid_targets_mean": 3010.2, + "valid_targets_min": 1377 + }, + { + "epoch": 6.6911764705882355, + "grad_norm": 0.7322859160819369, + "learning_rate": 2.3834781734784817e-07, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18106794357299805, + "step": 6370, + "valid_targets_mean": 3055.6, + "valid_targets_min": 1219 + }, + { + "epoch": 6.696428571428571, + "grad_norm": 0.738490804680364, + "learning_rate": 2.3035211570135995e-07, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17908601462841034, + "step": 6375, + "valid_targets_mean": 3442.4, + "valid_targets_min": 1168 + }, + { + "epoch": 6.701680672268908, + "grad_norm": 0.7971361125915069, + "learning_rate": 2.224920486084403e-07, + "loss": 0.1707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18598723411560059, + "step": 6380, + "valid_targets_mean": 3041.9, + "valid_targets_min": 1955 + }, + { + "epoch": 6.706932773109243, + "grad_norm": 0.7900277569565785, + "learning_rate": 2.1476766999502408e-07, + "loss": 0.1855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18570013344287872, + "step": 6385, + "valid_targets_mean": 3640.2, + "valid_targets_min": 1628 + }, + { + "epoch": 6.71218487394958, + "grad_norm": 0.6556232945874763, + "learning_rate": 2.071790328561152e-07, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1550453007221222, + "step": 6390, + "valid_targets_mean": 3358.4, + "valid_targets_min": 1482 + }, + { + "epoch": 6.717436974789916, + "grad_norm": 0.6812790984777259, + "learning_rate": 1.997261892554403e-07, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15719173848628998, + "step": 6395, + "valid_targets_mean": 3460.4, + "valid_targets_min": 904 + }, + { + "epoch": 6.722689075630252, + "grad_norm": 0.7769101571262454, + "learning_rate": 1.9240919032506688e-07, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2275974154472351, + "step": 6400, + "valid_targets_mean": 3728.4, + "valid_targets_min": 1194 + }, + { + "epoch": 6.727941176470588, + "grad_norm": 0.7389737083348429, + "learning_rate": 1.8522808626507683e-07, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1957985758781433, + "step": 6405, + "valid_targets_mean": 3940.4, + "valid_targets_min": 1658 + }, + { + "epoch": 6.733193277310924, + "grad_norm": 0.77498392264914, + "learning_rate": 1.781829263432111e-07, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1999933421611786, + "step": 6410, + "valid_targets_mean": 3536.9, + "valid_targets_min": 1423 + }, + { + "epoch": 6.73844537815126, + "grad_norm": 0.7005856165044073, + "learning_rate": 1.7127375889452569e-07, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1850917786359787, + "step": 6415, + "valid_targets_mean": 3351.2, + "valid_targets_min": 2010 + }, + { + "epoch": 6.743697478991597, + "grad_norm": 0.7167520426399059, + "learning_rate": 1.6450063132107396e-07, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18986815214157104, + "step": 6420, + "valid_targets_mean": 3498.6, + "valid_targets_min": 1172 + }, + { + "epoch": 6.748949579831933, + "grad_norm": 0.6942759631181366, + "learning_rate": 1.5786359009156928e-07, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18716512620449066, + "step": 6425, + "valid_targets_mean": 3863.8, + "valid_targets_min": 1229 + }, + { + "epoch": 6.754201680672269, + "grad_norm": 0.6936745579008864, + "learning_rate": 1.5136268074107398e-07, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1890535056591034, + "step": 6430, + "valid_targets_mean": 3600.5, + "valid_targets_min": 1523 + }, + { + "epoch": 6.759453781512605, + "grad_norm": 0.7070845571059466, + "learning_rate": 1.44997947870682e-07, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17694947123527527, + "step": 6435, + "valid_targets_mean": 3623.9, + "valid_targets_min": 997 + }, + { + "epoch": 6.764705882352941, + "grad_norm": 0.7375315587768612, + "learning_rate": 1.3876943514721465e-07, + "loss": 0.1731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17876756191253662, + "step": 6440, + "valid_targets_mean": 3535.2, + "valid_targets_min": 621 + }, + { + "epoch": 6.769957983193278, + "grad_norm": 0.7430227920142868, + "learning_rate": 1.3267718530292296e-07, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18011704087257385, + "step": 6445, + "valid_targets_mean": 3071.5, + "valid_targets_min": 1282 + }, + { + "epoch": 6.775210084033613, + "grad_norm": 0.6530774473575953, + "learning_rate": 1.267212401351925e-07, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1584375500679016, + "step": 6450, + "valid_targets_mean": 3779.1, + "valid_targets_min": 1421 + }, + { + "epoch": 6.78046218487395, + "grad_norm": 0.6882124626090157, + "learning_rate": 1.2090164050625907e-07, + "loss": 0.1722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1654224693775177, + "step": 6455, + "valid_targets_mean": 3330.2, + "valid_targets_min": 1853 + }, + { + "epoch": 6.785714285714286, + "grad_norm": 0.7044440655396256, + "learning_rate": 1.1521842634292013e-07, + "loss": 0.1967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17785203456878662, + "step": 6460, + "valid_targets_mean": 3327.8, + "valid_targets_min": 1410 + }, + { + "epoch": 6.790966386554622, + "grad_norm": 0.8200049965639139, + "learning_rate": 1.0967163663627044e-07, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18418024480342865, + "step": 6465, + "valid_targets_mean": 2920.6, + "valid_targets_min": 1305 + }, + { + "epoch": 6.796218487394958, + "grad_norm": 0.7154441982579451, + "learning_rate": 1.0426130944143353e-07, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1741981953382492, + "step": 6470, + "valid_targets_mean": 3539.7, + "valid_targets_min": 996 + }, + { + "epoch": 6.801470588235294, + "grad_norm": 0.6654603332753054, + "learning_rate": 9.898748187729513e-08, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16577930748462677, + "step": 6475, + "valid_targets_mean": 3357.1, + "valid_targets_min": 1349 + }, + { + "epoch": 6.80672268907563, + "grad_norm": 0.8172505155020807, + "learning_rate": 9.385019012625007e-08, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21172448992729187, + "step": 6480, + "valid_targets_mean": 3613.5, + "valid_targets_min": 1248 + }, + { + "epoch": 6.811974789915967, + "grad_norm": 0.6891423541136645, + "learning_rate": 8.884946943395811e-08, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14347678422927856, + "step": 6485, + "valid_targets_mean": 3327.4, + "valid_targets_min": 1109 + }, + { + "epoch": 6.817226890756302, + "grad_norm": 0.7516599357315075, + "learning_rate": 8.398535410910402e-08, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18097805976867676, + "step": 6490, + "valid_targets_mean": 2790.8, + "valid_targets_min": 1649 + }, + { + "epoch": 6.822478991596639, + "grad_norm": 0.7616811039107437, + "learning_rate": 7.925787752314674e-08, + "loss": 0.1793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18834054470062256, + "step": 6495, + "valid_targets_mean": 2811.8, + "valid_targets_min": 1197 + }, + { + "epoch": 6.8277310924369745, + "grad_norm": 0.7317487761588832, + "learning_rate": 7.466707211010838e-08, + "loss": 0.1919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20096731185913086, + "step": 6500, + "valid_targets_mean": 3420.2, + "valid_targets_min": 1720 + }, + { + "epoch": 6.832983193277311, + "grad_norm": 0.6976502515107287, + "learning_rate": 7.02129693663478e-08, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18510772287845612, + "step": 6505, + "valid_targets_mean": 3542.7, + "valid_targets_min": 1054 + }, + { + "epoch": 6.838235294117647, + "grad_norm": 0.8023438462902227, + "learning_rate": 6.589559985033189e-08, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17661021649837494, + "step": 6510, + "valid_targets_mean": 3049.7, + "valid_targets_min": 1643 + }, + { + "epoch": 6.843487394957983, + "grad_norm": 0.6905962109077173, + "learning_rate": 6.171499318244234e-08, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19038131833076477, + "step": 6515, + "valid_targets_mean": 3757.8, + "valid_targets_min": 1135 + }, + { + "epoch": 6.848739495798319, + "grad_norm": 0.7819298312536493, + "learning_rate": 5.767117804476696e-08, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.188393235206604, + "step": 6520, + "valid_targets_mean": 3715.1, + "valid_targets_min": 1174 + }, + { + "epoch": 6.8539915966386555, + "grad_norm": 0.7603239023118682, + "learning_rate": 5.376418218089541e-08, + "loss": 0.1934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20249368250370026, + "step": 6525, + "valid_targets_mean": 3321.9, + "valid_targets_min": 992 + }, + { + "epoch": 6.859243697478991, + "grad_norm": 0.8027496504534014, + "learning_rate": 4.99940323957393e-08, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2095927596092224, + "step": 6530, + "valid_targets_mean": 3713.2, + "valid_targets_min": 1133 + }, + { + "epoch": 6.864495798319328, + "grad_norm": 0.7827471840659329, + "learning_rate": 4.63607545553435e-08, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19427041709423065, + "step": 6535, + "valid_targets_mean": 2961.8, + "valid_targets_min": 1195 + }, + { + "epoch": 6.869747899159664, + "grad_norm": 0.8495916082471087, + "learning_rate": 4.2864373586706254e-08, + "loss": 0.1944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20868608355522156, + "step": 6540, + "valid_targets_mean": 3012.6, + "valid_targets_min": 484 + }, + { + "epoch": 6.875, + "grad_norm": 0.7058473937236638, + "learning_rate": 3.950491347761487e-08, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19130313396453857, + "step": 6545, + "valid_targets_mean": 3639.9, + "valid_targets_min": 1434 + }, + { + "epoch": 6.880252100840336, + "grad_norm": 0.8058046651510254, + "learning_rate": 3.628239727647254e-08, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17615005373954773, + "step": 6550, + "valid_targets_mean": 4017.4, + "valid_targets_min": 1349 + }, + { + "epoch": 6.885504201680672, + "grad_norm": 0.7986547158784268, + "learning_rate": 3.319684709215176e-08, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20986217260360718, + "step": 6555, + "valid_targets_mean": 3178.8, + "valid_targets_min": 1434 + }, + { + "epoch": 6.890756302521009, + "grad_norm": 0.7069871681601839, + "learning_rate": 3.024828409383007e-08, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18144480884075165, + "step": 6560, + "valid_targets_mean": 3633.9, + "valid_targets_min": 521 + }, + { + "epoch": 6.8960084033613445, + "grad_norm": 0.747775716498079, + "learning_rate": 2.743672851085233e-08, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19807741045951843, + "step": 6565, + "valid_targets_mean": 3981.0, + "valid_targets_min": 1464 + }, + { + "epoch": 6.901260504201681, + "grad_norm": 1.0660860736808098, + "learning_rate": 2.4762199632588634e-08, + "loss": 0.1992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1948835551738739, + "step": 6570, + "valid_targets_mean": 2612.0, + "valid_targets_min": 1239 + }, + { + "epoch": 6.906512605042017, + "grad_norm": 0.8171298625094956, + "learning_rate": 2.2224715808309983e-08, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18982771039009094, + "step": 6575, + "valid_targets_mean": 3489.8, + "valid_targets_min": 1457 + }, + { + "epoch": 6.911764705882353, + "grad_norm": 0.7615110154926487, + "learning_rate": 1.9824294447043923e-08, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.186547189950943, + "step": 6580, + "valid_targets_mean": 3059.1, + "valid_targets_min": 1460 + }, + { + "epoch": 6.917016806722689, + "grad_norm": 0.7778984161564974, + "learning_rate": 1.7560952017481313e-08, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19061462581157684, + "step": 6585, + "valid_targets_mean": 3259.2, + "valid_targets_min": 1084 + }, + { + "epoch": 6.9222689075630255, + "grad_norm": 0.7069596850632259, + "learning_rate": 1.5434704047836424e-08, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17042481899261475, + "step": 6590, + "valid_targets_mean": 3337.9, + "valid_targets_min": 1035 + }, + { + "epoch": 6.927521008403361, + "grad_norm": 0.7072879579761938, + "learning_rate": 1.344556512576256e-08, + "loss": 0.1826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19188648462295532, + "step": 6595, + "valid_targets_mean": 4091.9, + "valid_targets_min": 1876 + }, + { + "epoch": 6.932773109243698, + "grad_norm": 0.7505076999505291, + "learning_rate": 1.1593548898236606e-08, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1926942616701126, + "step": 6600, + "valid_targets_mean": 3705.8, + "valid_targets_min": 1062 + }, + { + "epoch": 6.938025210084033, + "grad_norm": 0.831906921551297, + "learning_rate": 9.878668071474639e-09, + "loss": 0.1745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16769427061080933, + "step": 6605, + "valid_targets_mean": 3030.8, + "valid_targets_min": 993 + }, + { + "epoch": 6.94327731092437, + "grad_norm": 0.7424388421322633, + "learning_rate": 8.3009344108409e-09, + "loss": 0.1753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17962056398391724, + "step": 6610, + "valid_targets_mean": 3366.9, + "valid_targets_min": 873 + }, + { + "epoch": 6.948529411764706, + "grad_norm": 0.676120936221246, + "learning_rate": 6.860358740763406e-09, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17056873440742493, + "step": 6615, + "valid_targets_mean": 3651.0, + "valid_targets_min": 1295 + }, + { + "epoch": 6.953781512605042, + "grad_norm": 0.7276545536651401, + "learning_rate": 5.55695094467179e-09, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19623544812202454, + "step": 6620, + "valid_targets_mean": 3583.6, + "valid_targets_min": 1598 + }, + { + "epoch": 6.959033613445378, + "grad_norm": 0.7364736521570804, + "learning_rate": 4.3907199649151355e-09, + "loss": 0.198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17404237389564514, + "step": 6625, + "valid_targets_mean": 2903.2, + "valid_targets_min": 1123 + }, + { + "epoch": 6.964285714285714, + "grad_norm": 0.8283192384968932, + "learning_rate": 3.361673802708687e-09, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2022954821586609, + "step": 6630, + "valid_targets_mean": 3116.1, + "valid_targets_min": 1621 + }, + { + "epoch": 6.96953781512605, + "grad_norm": 0.853057291277257, + "learning_rate": 2.469819518080563e-09, + "loss": 0.1754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19089040160179138, + "step": 6635, + "valid_targets_mean": 3081.4, + "valid_targets_min": 1858 + }, + { + "epoch": 6.974789915966387, + "grad_norm": 0.6873113252451049, + "learning_rate": 1.7151632298140209e-09, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15922731161117554, + "step": 6640, + "valid_targets_mean": 3470.7, + "valid_targets_min": 420 + }, + { + "epoch": 6.980042016806722, + "grad_norm": 0.754724576359931, + "learning_rate": 1.0977101154163727e-09, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19214347004890442, + "step": 6645, + "valid_targets_mean": 3584.2, + "valid_targets_min": 757 + }, + { + "epoch": 6.985294117647059, + "grad_norm": 0.8029047273305313, + "learning_rate": 6.174644110767958e-10, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17756588757038116, + "step": 6650, + "valid_targets_mean": 2940.8, + "valid_targets_min": 732 + }, + { + "epoch": 6.990546218487395, + "grad_norm": 0.7749949104741897, + "learning_rate": 2.744294116419077e-10, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18697890639305115, + "step": 6655, + "valid_targets_mean": 3248.2, + "valid_targets_min": 929 + }, + { + "epoch": 6.995798319327731, + "grad_norm": 0.7329373554307617, + "learning_rate": 6.860747058468064e-11, + "loss": 0.1959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18864449858665466, + "step": 6660, + "valid_targets_mean": 3728.8, + "valid_targets_min": 1217 + }, + { + "epoch": 7.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2101745903491974, + "step": 6664, + "total_flos": 1189003618287616.0, + "train_loss": 0.21699104300674413, + "train_runtime": 24543.5183, + "train_samples_per_second": 4.344, + "train_steps_per_second": 0.272, + "valid_targets_mean": 3715.1, + "valid_targets_min": 791 + } + ], + "logging_steps": 5, + "max_steps": 6664, + "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": 1189003618287616.0, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}