| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 7.0, |
| "eval_steps": 500, |
| "global_step": 4375, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.008, |
| "grad_norm": 12.374207826720879, |
| "learning_rate": 3.6529680365296803e-07, |
| "loss": 0.673, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6549139022827148, |
| "step": 5, |
| "valid_targets_mean": 3321.2, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 0.016, |
| "grad_norm": 11.43057305648331, |
| "learning_rate": 8.219178082191781e-07, |
| "loss": 0.6633, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6954354643821716, |
| "step": 10, |
| "valid_targets_mean": 2652.1, |
| "valid_targets_min": 682 |
| }, |
| { |
| "epoch": 0.024, |
| "grad_norm": 9.570235812016593, |
| "learning_rate": 1.278538812785388e-06, |
| "loss": 0.6271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6770514249801636, |
| "step": 15, |
| "valid_targets_mean": 3358.0, |
| "valid_targets_min": 739 |
| }, |
| { |
| "epoch": 0.032, |
| "grad_norm": 8.170405598762851, |
| "learning_rate": 1.7351598173515982e-06, |
| "loss": 0.6202, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5899783372879028, |
| "step": 20, |
| "valid_targets_mean": 4322.4, |
| "valid_targets_min": 743 |
| }, |
| { |
| "epoch": 0.04, |
| "grad_norm": 3.6087283340453133, |
| "learning_rate": 2.191780821917808e-06, |
| "loss": 0.5571, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4666231572628021, |
| "step": 25, |
| "valid_targets_mean": 5854.6, |
| "valid_targets_min": 718 |
| }, |
| { |
| "epoch": 0.048, |
| "grad_norm": 2.6527603461614553, |
| "learning_rate": 2.6484018264840183e-06, |
| "loss": 0.5838, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5448369979858398, |
| "step": 30, |
| "valid_targets_mean": 4481.5, |
| "valid_targets_min": 920 |
| }, |
| { |
| "epoch": 0.056, |
| "grad_norm": 2.027006540805368, |
| "learning_rate": 3.1050228310502285e-06, |
| "loss": 0.5264, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5310207605361938, |
| "step": 35, |
| "valid_targets_mean": 2902.4, |
| "valid_targets_min": 828 |
| }, |
| { |
| "epoch": 0.064, |
| "grad_norm": 1.8084836411086191, |
| "learning_rate": 3.5616438356164386e-06, |
| "loss": 0.5147, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5529466867446899, |
| "step": 40, |
| "valid_targets_mean": 2200.1, |
| "valid_targets_min": 732 |
| }, |
| { |
| "epoch": 0.072, |
| "grad_norm": 1.0819999335224122, |
| "learning_rate": 4.018264840182649e-06, |
| "loss": 0.4949, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45418739318847656, |
| "step": 45, |
| "valid_targets_mean": 5406.0, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 0.08, |
| "grad_norm": 0.9962749166883843, |
| "learning_rate": 4.4748858447488585e-06, |
| "loss": 0.5261, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5496567487716675, |
| "step": 50, |
| "valid_targets_mean": 3678.8, |
| "valid_targets_min": 902 |
| }, |
| { |
| "epoch": 0.088, |
| "grad_norm": 0.8013005822835737, |
| "learning_rate": 4.931506849315069e-06, |
| "loss": 0.4898, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5123369693756104, |
| "step": 55, |
| "valid_targets_mean": 4873.1, |
| "valid_targets_min": 950 |
| }, |
| { |
| "epoch": 0.096, |
| "grad_norm": 0.6575479902865903, |
| "learning_rate": 5.388127853881279e-06, |
| "loss": 0.4628, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41041916608810425, |
| "step": 60, |
| "valid_targets_mean": 4923.4, |
| "valid_targets_min": 887 |
| }, |
| { |
| "epoch": 0.104, |
| "grad_norm": 0.7006327660151911, |
| "learning_rate": 5.8447488584474885e-06, |
| "loss": 0.476, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4419603943824768, |
| "step": 65, |
| "valid_targets_mean": 3550.9, |
| "valid_targets_min": 954 |
| }, |
| { |
| "epoch": 0.112, |
| "grad_norm": 0.6222761889218557, |
| "learning_rate": 6.301369863013699e-06, |
| "loss": 0.4573, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45139503479003906, |
| "step": 70, |
| "valid_targets_mean": 3806.9, |
| "valid_targets_min": 779 |
| }, |
| { |
| "epoch": 0.12, |
| "grad_norm": 0.5620156952847125, |
| "learning_rate": 6.757990867579909e-06, |
| "loss": 0.4339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44665512442588806, |
| "step": 75, |
| "valid_targets_mean": 4951.6, |
| "valid_targets_min": 994 |
| }, |
| { |
| "epoch": 0.128, |
| "grad_norm": 0.49883881664127766, |
| "learning_rate": 7.214611872146119e-06, |
| "loss": 0.4414, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42222359776496887, |
| "step": 80, |
| "valid_targets_mean": 5721.9, |
| "valid_targets_min": 909 |
| }, |
| { |
| "epoch": 0.136, |
| "grad_norm": 0.5538426247192075, |
| "learning_rate": 7.671232876712329e-06, |
| "loss": 0.4276, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4493791460990906, |
| "step": 85, |
| "valid_targets_mean": 4876.8, |
| "valid_targets_min": 1047 |
| }, |
| { |
| "epoch": 0.144, |
| "grad_norm": 0.5791700438161751, |
| "learning_rate": 8.127853881278539e-06, |
| "loss": 0.4333, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4152202606201172, |
| "step": 90, |
| "valid_targets_mean": 3532.4, |
| "valid_targets_min": 959 |
| }, |
| { |
| "epoch": 0.152, |
| "grad_norm": 0.5477499942337496, |
| "learning_rate": 8.584474885844748e-06, |
| "loss": 0.4003, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37116068601608276, |
| "step": 95, |
| "valid_targets_mean": 3519.7, |
| "valid_targets_min": 1179 |
| }, |
| { |
| "epoch": 0.16, |
| "grad_norm": 0.5775964980586745, |
| "learning_rate": 9.04109589041096e-06, |
| "loss": 0.4339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40842100977897644, |
| "step": 100, |
| "valid_targets_mean": 3587.7, |
| "valid_targets_min": 875 |
| }, |
| { |
| "epoch": 0.168, |
| "grad_norm": 0.6430856727704666, |
| "learning_rate": 9.49771689497717e-06, |
| "loss": 0.4225, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47565746307373047, |
| "step": 105, |
| "valid_targets_mean": 3647.5, |
| "valid_targets_min": 887 |
| }, |
| { |
| "epoch": 0.176, |
| "grad_norm": 0.716966298704087, |
| "learning_rate": 9.95433789954338e-06, |
| "loss": 0.3843, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42896223068237305, |
| "step": 110, |
| "valid_targets_mean": 2842.8, |
| "valid_targets_min": 921 |
| }, |
| { |
| "epoch": 0.184, |
| "grad_norm": 0.5301806019840736, |
| "learning_rate": 1.0410958904109589e-05, |
| "loss": 0.3547, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3746205270290375, |
| "step": 115, |
| "valid_targets_mean": 3999.1, |
| "valid_targets_min": 1007 |
| }, |
| { |
| "epoch": 0.192, |
| "grad_norm": 0.48241364109719725, |
| "learning_rate": 1.08675799086758e-05, |
| "loss": 0.3868, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33790087699890137, |
| "step": 120, |
| "valid_targets_mean": 5485.6, |
| "valid_targets_min": 709 |
| }, |
| { |
| "epoch": 0.2, |
| "grad_norm": 0.38455061918249317, |
| "learning_rate": 1.132420091324201e-05, |
| "loss": 0.3826, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3305814862251282, |
| "step": 125, |
| "valid_targets_mean": 7719.6, |
| "valid_targets_min": 865 |
| }, |
| { |
| "epoch": 0.208, |
| "grad_norm": 0.4907444754258652, |
| "learning_rate": 1.178082191780822e-05, |
| "loss": 0.3677, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.401222288608551, |
| "step": 130, |
| "valid_targets_mean": 4818.7, |
| "valid_targets_min": 961 |
| }, |
| { |
| "epoch": 0.216, |
| "grad_norm": 0.45743331436009144, |
| "learning_rate": 1.223744292237443e-05, |
| "loss": 0.3565, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3574488162994385, |
| "step": 135, |
| "valid_targets_mean": 5390.3, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 0.224, |
| "grad_norm": 0.511011558834825, |
| "learning_rate": 1.2694063926940641e-05, |
| "loss": 0.3764, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3379029631614685, |
| "step": 140, |
| "valid_targets_mean": 5142.8, |
| "valid_targets_min": 893 |
| }, |
| { |
| "epoch": 0.232, |
| "grad_norm": 0.5822963240467418, |
| "learning_rate": 1.3150684931506849e-05, |
| "loss": 0.3659, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34744590520858765, |
| "step": 145, |
| "valid_targets_mean": 5302.1, |
| "valid_targets_min": 880 |
| }, |
| { |
| "epoch": 0.24, |
| "grad_norm": 0.6216847277162149, |
| "learning_rate": 1.360730593607306e-05, |
| "loss": 0.3759, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3552771210670471, |
| "step": 150, |
| "valid_targets_mean": 4948.8, |
| "valid_targets_min": 814 |
| }, |
| { |
| "epoch": 0.248, |
| "grad_norm": 0.574367767908384, |
| "learning_rate": 1.406392694063927e-05, |
| "loss": 0.3622, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3638104498386383, |
| "step": 155, |
| "valid_targets_mean": 3668.3, |
| "valid_targets_min": 763 |
| }, |
| { |
| "epoch": 0.256, |
| "grad_norm": 0.5264337459661648, |
| "learning_rate": 1.4520547945205482e-05, |
| "loss": 0.3625, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3273163437843323, |
| "step": 160, |
| "valid_targets_mean": 4477.4, |
| "valid_targets_min": 926 |
| }, |
| { |
| "epoch": 0.264, |
| "grad_norm": 0.49616555200759516, |
| "learning_rate": 1.497716894977169e-05, |
| "loss": 0.3732, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3494306802749634, |
| "step": 165, |
| "valid_targets_mean": 4903.5, |
| "valid_targets_min": 766 |
| }, |
| { |
| "epoch": 0.272, |
| "grad_norm": 0.5183306288028923, |
| "learning_rate": 1.54337899543379e-05, |
| "loss": 0.3495, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32361409068107605, |
| "step": 170, |
| "valid_targets_mean": 4151.4, |
| "valid_targets_min": 979 |
| }, |
| { |
| "epoch": 0.28, |
| "grad_norm": 0.6625899937361988, |
| "learning_rate": 1.589041095890411e-05, |
| "loss": 0.3497, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3405327796936035, |
| "step": 175, |
| "valid_targets_mean": 2944.8, |
| "valid_targets_min": 843 |
| }, |
| { |
| "epoch": 0.288, |
| "grad_norm": 0.521897584371646, |
| "learning_rate": 1.634703196347032e-05, |
| "loss": 0.3227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3742603063583374, |
| "step": 180, |
| "valid_targets_mean": 4572.4, |
| "valid_targets_min": 872 |
| }, |
| { |
| "epoch": 0.296, |
| "grad_norm": 0.49053163267667843, |
| "learning_rate": 1.680365296803653e-05, |
| "loss": 0.3486, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30405259132385254, |
| "step": 185, |
| "valid_targets_mean": 5678.8, |
| "valid_targets_min": 629 |
| }, |
| { |
| "epoch": 0.304, |
| "grad_norm": 0.9074873265259921, |
| "learning_rate": 1.726027397260274e-05, |
| "loss": 0.3689, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5085344314575195, |
| "step": 190, |
| "valid_targets_mean": 2015.6, |
| "valid_targets_min": 976 |
| }, |
| { |
| "epoch": 0.312, |
| "grad_norm": 0.7231403812681834, |
| "learning_rate": 1.771689497716895e-05, |
| "loss": 0.3531, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4101080000400543, |
| "step": 195, |
| "valid_targets_mean": 3050.3, |
| "valid_targets_min": 869 |
| }, |
| { |
| "epoch": 0.32, |
| "grad_norm": 0.4901983123951643, |
| "learning_rate": 1.8173515981735163e-05, |
| "loss": 0.3366, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3015870749950409, |
| "step": 200, |
| "valid_targets_mean": 6477.4, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 0.328, |
| "grad_norm": 0.7091813307576407, |
| "learning_rate": 1.863013698630137e-05, |
| "loss": 0.3278, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36794477701187134, |
| "step": 205, |
| "valid_targets_mean": 2819.5, |
| "valid_targets_min": 648 |
| }, |
| { |
| "epoch": 0.336, |
| "grad_norm": 0.6242515345896389, |
| "learning_rate": 1.9086757990867582e-05, |
| "loss": 0.3378, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38683176040649414, |
| "step": 210, |
| "valid_targets_mean": 3260.6, |
| "valid_targets_min": 822 |
| }, |
| { |
| "epoch": 0.344, |
| "grad_norm": 0.5367951319766895, |
| "learning_rate": 1.954337899543379e-05, |
| "loss": 0.3027, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31571871042251587, |
| "step": 215, |
| "valid_targets_mean": 4805.6, |
| "valid_targets_min": 1114 |
| }, |
| { |
| "epoch": 0.352, |
| "grad_norm": 0.7441017340177637, |
| "learning_rate": 2e-05, |
| "loss": 0.3545, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35355275869369507, |
| "step": 220, |
| "valid_targets_mean": 2396.6, |
| "valid_targets_min": 889 |
| }, |
| { |
| "epoch": 0.36, |
| "grad_norm": 0.5463780408548322, |
| "learning_rate": 2.045662100456621e-05, |
| "loss": 0.3252, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2769712805747986, |
| "step": 225, |
| "valid_targets_mean": 4953.7, |
| "valid_targets_min": 938 |
| }, |
| { |
| "epoch": 0.368, |
| "grad_norm": 0.5117792360088306, |
| "learning_rate": 2.0913242009132424e-05, |
| "loss": 0.334, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3192422389984131, |
| "step": 230, |
| "valid_targets_mean": 5398.4, |
| "valid_targets_min": 917 |
| }, |
| { |
| "epoch": 0.376, |
| "grad_norm": 0.6081591338406857, |
| "learning_rate": 2.1369863013698632e-05, |
| "loss": 0.3315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3515390157699585, |
| "step": 235, |
| "valid_targets_mean": 3282.2, |
| "valid_targets_min": 1082 |
| }, |
| { |
| "epoch": 0.384, |
| "grad_norm": 0.5361233007075954, |
| "learning_rate": 2.182648401826484e-05, |
| "loss": 0.327, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33629757165908813, |
| "step": 240, |
| "valid_targets_mean": 5654.5, |
| "valid_targets_min": 794 |
| }, |
| { |
| "epoch": 0.392, |
| "grad_norm": 0.4879945295289328, |
| "learning_rate": 2.2283105022831052e-05, |
| "loss": 0.3186, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2991241216659546, |
| "step": 245, |
| "valid_targets_mean": 6910.0, |
| "valid_targets_min": 693 |
| }, |
| { |
| "epoch": 0.4, |
| "grad_norm": 0.5818121087256863, |
| "learning_rate": 2.2739726027397263e-05, |
| "loss": 0.3211, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2698156237602234, |
| "step": 250, |
| "valid_targets_mean": 5727.4, |
| "valid_targets_min": 837 |
| }, |
| { |
| "epoch": 0.408, |
| "grad_norm": 0.5484701967234875, |
| "learning_rate": 2.3196347031963475e-05, |
| "loss": 0.3167, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2877916991710663, |
| "step": 255, |
| "valid_targets_mean": 3935.9, |
| "valid_targets_min": 908 |
| }, |
| { |
| "epoch": 0.416, |
| "grad_norm": 0.7598295746278937, |
| "learning_rate": 2.3652968036529683e-05, |
| "loss": 0.3256, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.379886269569397, |
| "step": 260, |
| "valid_targets_mean": 3017.8, |
| "valid_targets_min": 929 |
| }, |
| { |
| "epoch": 0.424, |
| "grad_norm": 0.6292837735711766, |
| "learning_rate": 2.410958904109589e-05, |
| "loss": 0.3291, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3254546821117401, |
| "step": 265, |
| "valid_targets_mean": 3521.9, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 0.432, |
| "grad_norm": 0.507874716177431, |
| "learning_rate": 2.4566210045662106e-05, |
| "loss": 0.2991, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2930123805999756, |
| "step": 270, |
| "valid_targets_mean": 4279.1, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 0.44, |
| "grad_norm": 0.4869811841188768, |
| "learning_rate": 2.5022831050228314e-05, |
| "loss": 0.3159, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.295903742313385, |
| "step": 275, |
| "valid_targets_mean": 5994.6, |
| "valid_targets_min": 807 |
| }, |
| { |
| "epoch": 0.448, |
| "grad_norm": 0.618514883028784, |
| "learning_rate": 2.547945205479452e-05, |
| "loss": 0.3316, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32905834913253784, |
| "step": 280, |
| "valid_targets_mean": 4779.1, |
| "valid_targets_min": 1148 |
| }, |
| { |
| "epoch": 0.456, |
| "grad_norm": 0.6790060720865024, |
| "learning_rate": 2.593607305936073e-05, |
| "loss": 0.3311, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3841496706008911, |
| "step": 285, |
| "valid_targets_mean": 3403.9, |
| "valid_targets_min": 723 |
| }, |
| { |
| "epoch": 0.464, |
| "grad_norm": 1.3404239047851452, |
| "learning_rate": 2.6392694063926944e-05, |
| "loss": 0.3167, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29200658202171326, |
| "step": 290, |
| "valid_targets_mean": 4245.9, |
| "valid_targets_min": 657 |
| }, |
| { |
| "epoch": 0.472, |
| "grad_norm": 0.5708709470409948, |
| "learning_rate": 2.6849315068493153e-05, |
| "loss": 0.3095, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30335745215415955, |
| "step": 295, |
| "valid_targets_mean": 4115.2, |
| "valid_targets_min": 1058 |
| }, |
| { |
| "epoch": 0.48, |
| "grad_norm": 0.601194220459844, |
| "learning_rate": 2.7305936073059364e-05, |
| "loss": 0.3138, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4048035740852356, |
| "step": 300, |
| "valid_targets_mean": 5644.8, |
| "valid_targets_min": 1055 |
| }, |
| { |
| "epoch": 0.488, |
| "grad_norm": 0.5494984385782556, |
| "learning_rate": 2.7762557077625572e-05, |
| "loss": 0.2913, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29575061798095703, |
| "step": 305, |
| "valid_targets_mean": 5261.0, |
| "valid_targets_min": 734 |
| }, |
| { |
| "epoch": 0.496, |
| "grad_norm": 0.5098156256599904, |
| "learning_rate": 2.8219178082191783e-05, |
| "loss": 0.3329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3082813024520874, |
| "step": 310, |
| "valid_targets_mean": 6082.6, |
| "valid_targets_min": 826 |
| }, |
| { |
| "epoch": 0.504, |
| "grad_norm": 0.5159281940057138, |
| "learning_rate": 2.8675799086757995e-05, |
| "loss": 0.3099, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3101263642311096, |
| "step": 315, |
| "valid_targets_mean": 5008.8, |
| "valid_targets_min": 806 |
| }, |
| { |
| "epoch": 0.512, |
| "grad_norm": 0.6804451743074025, |
| "learning_rate": 2.9132420091324203e-05, |
| "loss": 0.3051, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29619288444519043, |
| "step": 320, |
| "valid_targets_mean": 4585.7, |
| "valid_targets_min": 1016 |
| }, |
| { |
| "epoch": 0.52, |
| "grad_norm": 0.5850935505815877, |
| "learning_rate": 2.958904109589041e-05, |
| "loss": 0.3008, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28627079725265503, |
| "step": 325, |
| "valid_targets_mean": 3302.2, |
| "valid_targets_min": 966 |
| }, |
| { |
| "epoch": 0.528, |
| "grad_norm": 0.965007545709362, |
| "learning_rate": 3.0045662100456626e-05, |
| "loss": 0.3114, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35180073976516724, |
| "step": 330, |
| "valid_targets_mean": 3250.3, |
| "valid_targets_min": 829 |
| }, |
| { |
| "epoch": 0.536, |
| "grad_norm": 0.5381764485997723, |
| "learning_rate": 3.0502283105022834e-05, |
| "loss": 0.3031, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2976144552230835, |
| "step": 335, |
| "valid_targets_mean": 4887.6, |
| "valid_targets_min": 827 |
| }, |
| { |
| "epoch": 0.544, |
| "grad_norm": 0.648507415849347, |
| "learning_rate": 3.0958904109589045e-05, |
| "loss": 0.319, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3534129858016968, |
| "step": 340, |
| "valid_targets_mean": 4146.6, |
| "valid_targets_min": 1034 |
| }, |
| { |
| "epoch": 0.552, |
| "grad_norm": 0.5712482499204495, |
| "learning_rate": 3.141552511415525e-05, |
| "loss": 0.2799, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25627678632736206, |
| "step": 345, |
| "valid_targets_mean": 4196.8, |
| "valid_targets_min": 917 |
| }, |
| { |
| "epoch": 0.56, |
| "grad_norm": 0.44280064147997866, |
| "learning_rate": 3.187214611872147e-05, |
| "loss": 0.2947, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29131901264190674, |
| "step": 350, |
| "valid_targets_mean": 6720.6, |
| "valid_targets_min": 1054 |
| }, |
| { |
| "epoch": 0.568, |
| "grad_norm": 0.5998227591824579, |
| "learning_rate": 3.2328767123287676e-05, |
| "loss": 0.294, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2890249192714691, |
| "step": 355, |
| "valid_targets_mean": 3408.9, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 0.576, |
| "grad_norm": 0.7483944088873048, |
| "learning_rate": 3.2785388127853884e-05, |
| "loss": 0.3097, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3844132721424103, |
| "step": 360, |
| "valid_targets_mean": 4073.4, |
| "valid_targets_min": 893 |
| }, |
| { |
| "epoch": 0.584, |
| "grad_norm": 0.5805802611531444, |
| "learning_rate": 3.324200913242009e-05, |
| "loss": 0.2985, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31355297565460205, |
| "step": 365, |
| "valid_targets_mean": 3970.4, |
| "valid_targets_min": 796 |
| }, |
| { |
| "epoch": 0.592, |
| "grad_norm": 0.6042096610158246, |
| "learning_rate": 3.369863013698631e-05, |
| "loss": 0.3084, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2948343753814697, |
| "step": 370, |
| "valid_targets_mean": 3084.6, |
| "valid_targets_min": 677 |
| }, |
| { |
| "epoch": 0.6, |
| "grad_norm": 0.6461668885744746, |
| "learning_rate": 3.4155251141552515e-05, |
| "loss": 0.2984, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2847023010253906, |
| "step": 375, |
| "valid_targets_mean": 3152.9, |
| "valid_targets_min": 827 |
| }, |
| { |
| "epoch": 0.608, |
| "grad_norm": 0.6252813642439993, |
| "learning_rate": 3.461187214611872e-05, |
| "loss": 0.3071, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31125932931900024, |
| "step": 380, |
| "valid_targets_mean": 3398.4, |
| "valid_targets_min": 714 |
| }, |
| { |
| "epoch": 0.616, |
| "grad_norm": 0.5513266421188043, |
| "learning_rate": 3.506849315068493e-05, |
| "loss": 0.302, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3118017315864563, |
| "step": 385, |
| "valid_targets_mean": 4706.7, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 0.624, |
| "grad_norm": 0.5782519741863913, |
| "learning_rate": 3.5525114155251146e-05, |
| "loss": 0.3088, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.272940993309021, |
| "step": 390, |
| "valid_targets_mean": 3489.2, |
| "valid_targets_min": 855 |
| }, |
| { |
| "epoch": 0.632, |
| "grad_norm": 0.4835190489595145, |
| "learning_rate": 3.5981735159817354e-05, |
| "loss": 0.2907, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24451082944869995, |
| "step": 395, |
| "valid_targets_mean": 6700.7, |
| "valid_targets_min": 865 |
| }, |
| { |
| "epoch": 0.64, |
| "grad_norm": 0.5678926388355849, |
| "learning_rate": 3.643835616438356e-05, |
| "loss": 0.3022, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27626603841781616, |
| "step": 400, |
| "valid_targets_mean": 3785.8, |
| "valid_targets_min": 662 |
| }, |
| { |
| "epoch": 0.648, |
| "grad_norm": 0.5159242200400144, |
| "learning_rate": 3.689497716894977e-05, |
| "loss": 0.3, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26561927795410156, |
| "step": 405, |
| "valid_targets_mean": 4753.9, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 0.656, |
| "grad_norm": 0.6417729066274248, |
| "learning_rate": 3.7351598173515985e-05, |
| "loss": 0.3159, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36154353618621826, |
| "step": 410, |
| "valid_targets_mean": 3772.2, |
| "valid_targets_min": 1060 |
| }, |
| { |
| "epoch": 0.664, |
| "grad_norm": 0.5052939522988528, |
| "learning_rate": 3.780821917808219e-05, |
| "loss": 0.2794, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2737759053707123, |
| "step": 415, |
| "valid_targets_mean": 6162.9, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 0.672, |
| "grad_norm": 0.41783612659228964, |
| "learning_rate": 3.82648401826484e-05, |
| "loss": 0.2938, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2618584930896759, |
| "step": 420, |
| "valid_targets_mean": 9499.0, |
| "valid_targets_min": 1090 |
| }, |
| { |
| "epoch": 0.68, |
| "grad_norm": 1.0340429120195325, |
| "learning_rate": 3.8721461187214615e-05, |
| "loss": 0.311, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3576170802116394, |
| "step": 425, |
| "valid_targets_mean": 3117.8, |
| "valid_targets_min": 1019 |
| }, |
| { |
| "epoch": 0.688, |
| "grad_norm": 0.6432463334293257, |
| "learning_rate": 3.9178082191780823e-05, |
| "loss": 0.2938, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31897398829460144, |
| "step": 430, |
| "valid_targets_mean": 2963.1, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 0.696, |
| "grad_norm": 0.6510917417085519, |
| "learning_rate": 3.963470319634704e-05, |
| "loss": 0.3034, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3051530718803406, |
| "step": 435, |
| "valid_targets_mean": 3365.6, |
| "valid_targets_min": 974 |
| }, |
| { |
| "epoch": 0.704, |
| "grad_norm": 0.5340165987215613, |
| "learning_rate": 3.99999936325009e-05, |
| "loss": 0.3129, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3342752456665039, |
| "step": 440, |
| "valid_targets_mean": 4853.4, |
| "valid_targets_min": 1068 |
| }, |
| { |
| "epoch": 0.712, |
| "grad_norm": 0.5677728076782123, |
| "learning_rate": 3.9999770770457856e-05, |
| "loss": 0.3128, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2853566110134125, |
| "step": 445, |
| "valid_targets_mean": 4938.3, |
| "valid_targets_min": 910 |
| }, |
| { |
| "epoch": 0.72, |
| "grad_norm": 0.566650322028894, |
| "learning_rate": 3.9999229537513936e-05, |
| "loss": 0.2989, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27975791692733765, |
| "step": 450, |
| "valid_targets_mean": 3907.8, |
| "valid_targets_min": 793 |
| }, |
| { |
| "epoch": 0.728, |
| "grad_norm": 0.512996855287856, |
| "learning_rate": 3.999836994228487e-05, |
| "loss": 0.2823, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2654213011264801, |
| "step": 455, |
| "valid_targets_mean": 3657.9, |
| "valid_targets_min": 976 |
| }, |
| { |
| "epoch": 0.736, |
| "grad_norm": 0.5712508862988912, |
| "learning_rate": 3.999719199845432e-05, |
| "loss": 0.2871, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28030380606651306, |
| "step": 460, |
| "valid_targets_mean": 3826.2, |
| "valid_targets_min": 791 |
| }, |
| { |
| "epoch": 0.744, |
| "grad_norm": 0.550763618650829, |
| "learning_rate": 3.999569572477366e-05, |
| "loss": 0.3016, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2925485074520111, |
| "step": 465, |
| "valid_targets_mean": 3767.2, |
| "valid_targets_min": 1074 |
| }, |
| { |
| "epoch": 0.752, |
| "grad_norm": 0.4844778100988879, |
| "learning_rate": 3.999388114506166e-05, |
| "loss": 0.2779, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28359127044677734, |
| "step": 470, |
| "valid_targets_mean": 5441.2, |
| "valid_targets_min": 1040 |
| }, |
| { |
| "epoch": 0.76, |
| "grad_norm": 0.5223336482223219, |
| "learning_rate": 3.999174828820413e-05, |
| "loss": 0.3165, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2604113817214966, |
| "step": 475, |
| "valid_targets_mean": 4254.1, |
| "valid_targets_min": 662 |
| }, |
| { |
| "epoch": 0.768, |
| "grad_norm": 0.7085323476959958, |
| "learning_rate": 3.998929718815341e-05, |
| "loss": 0.3046, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.366795152425766, |
| "step": 480, |
| "valid_targets_mean": 3101.3, |
| "valid_targets_min": 941 |
| }, |
| { |
| "epoch": 0.776, |
| "grad_norm": 0.5280273815145139, |
| "learning_rate": 3.998652788392792e-05, |
| "loss": 0.2805, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2906476855278015, |
| "step": 485, |
| "valid_targets_mean": 4255.8, |
| "valid_targets_min": 968 |
| }, |
| { |
| "epoch": 0.784, |
| "grad_norm": 0.45090956227053314, |
| "learning_rate": 3.9983440419611445e-05, |
| "loss": 0.2803, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23735715448856354, |
| "step": 490, |
| "valid_targets_mean": 4642.8, |
| "valid_targets_min": 868 |
| }, |
| { |
| "epoch": 0.792, |
| "grad_norm": 0.4683162495835124, |
| "learning_rate": 3.9980034844352494e-05, |
| "loss": 0.3064, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.268685519695282, |
| "step": 495, |
| "valid_targets_mean": 6424.1, |
| "valid_targets_min": 1279 |
| }, |
| { |
| "epoch": 0.8, |
| "grad_norm": 0.517351165125786, |
| "learning_rate": 3.9976311212363495e-05, |
| "loss": 0.2838, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3032742142677307, |
| "step": 500, |
| "valid_targets_mean": 4640.5, |
| "valid_targets_min": 738 |
| }, |
| { |
| "epoch": 0.808, |
| "grad_norm": 0.5071613397785332, |
| "learning_rate": 3.997226958291992e-05, |
| "loss": 0.3027, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26866069436073303, |
| "step": 505, |
| "valid_targets_mean": 5655.1, |
| "valid_targets_min": 905 |
| }, |
| { |
| "epoch": 0.816, |
| "grad_norm": 0.6466486162647562, |
| "learning_rate": 3.996791002035937e-05, |
| "loss": 0.3158, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4299527406692505, |
| "step": 510, |
| "valid_targets_mean": 3593.1, |
| "valid_targets_min": 859 |
| }, |
| { |
| "epoch": 0.824, |
| "grad_norm": 0.5443624306552559, |
| "learning_rate": 3.996323259408055e-05, |
| "loss": 0.3062, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2630675435066223, |
| "step": 515, |
| "valid_targets_mean": 7815.6, |
| "valid_targets_min": 532 |
| }, |
| { |
| "epoch": 0.832, |
| "grad_norm": 0.6243451089128813, |
| "learning_rate": 3.995823737854211e-05, |
| "loss": 0.2925, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3227846622467041, |
| "step": 520, |
| "valid_targets_mean": 3138.6, |
| "valid_targets_min": 757 |
| }, |
| { |
| "epoch": 0.84, |
| "grad_norm": 0.6022810456897381, |
| "learning_rate": 3.9952924453261534e-05, |
| "loss": 0.3032, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3212684988975525, |
| "step": 525, |
| "valid_targets_mean": 3576.4, |
| "valid_targets_min": 914 |
| }, |
| { |
| "epoch": 0.848, |
| "grad_norm": 0.5723242679917651, |
| "learning_rate": 3.994729390281384e-05, |
| "loss": 0.2769, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2988251745700836, |
| "step": 530, |
| "valid_targets_mean": 4023.6, |
| "valid_targets_min": 722 |
| }, |
| { |
| "epoch": 0.856, |
| "grad_norm": 0.5462336841915116, |
| "learning_rate": 3.994134581683021e-05, |
| "loss": 0.2911, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2816080152988434, |
| "step": 535, |
| "valid_targets_mean": 3982.4, |
| "valid_targets_min": 1090 |
| }, |
| { |
| "epoch": 0.864, |
| "grad_norm": 0.4918215735498511, |
| "learning_rate": 3.9935080289996626e-05, |
| "loss": 0.298, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26696616411209106, |
| "step": 540, |
| "valid_targets_mean": 5292.1, |
| "valid_targets_min": 839 |
| }, |
| { |
| "epoch": 0.872, |
| "grad_norm": 0.6349909083508697, |
| "learning_rate": 3.992849742205228e-05, |
| "loss": 0.281, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32703348994255066, |
| "step": 545, |
| "valid_targets_mean": 3155.8, |
| "valid_targets_min": 712 |
| }, |
| { |
| "epoch": 0.88, |
| "grad_norm": 0.455853550279736, |
| "learning_rate": 3.9921597317788065e-05, |
| "loss": 0.276, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2557429075241089, |
| "step": 550, |
| "valid_targets_mean": 4919.1, |
| "valid_targets_min": 767 |
| }, |
| { |
| "epoch": 0.888, |
| "grad_norm": 0.6086037931430966, |
| "learning_rate": 3.991438008704486e-05, |
| "loss": 0.281, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33778220415115356, |
| "step": 555, |
| "valid_targets_mean": 3085.8, |
| "valid_targets_min": 820 |
| }, |
| { |
| "epoch": 0.896, |
| "grad_norm": 0.5213582280989516, |
| "learning_rate": 3.990684584471179e-05, |
| "loss": 0.2946, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29489654302597046, |
| "step": 560, |
| "valid_targets_mean": 4436.3, |
| "valid_targets_min": 779 |
| }, |
| { |
| "epoch": 0.904, |
| "grad_norm": 0.5843714438083748, |
| "learning_rate": 3.989899471072441e-05, |
| "loss": 0.2991, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26970839500427246, |
| "step": 565, |
| "valid_targets_mean": 2767.6, |
| "valid_targets_min": 835 |
| }, |
| { |
| "epoch": 0.912, |
| "grad_norm": 0.541008479057164, |
| "learning_rate": 3.9890826810062784e-05, |
| "loss": 0.2861, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25051432847976685, |
| "step": 570, |
| "valid_targets_mean": 3267.8, |
| "valid_targets_min": 901 |
| }, |
| { |
| "epoch": 0.92, |
| "grad_norm": 0.44648718657516523, |
| "learning_rate": 3.988234227274949e-05, |
| "loss": 0.2902, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28003740310668945, |
| "step": 575, |
| "valid_targets_mean": 5206.3, |
| "valid_targets_min": 616 |
| }, |
| { |
| "epoch": 0.928, |
| "grad_norm": 0.5963913155091053, |
| "learning_rate": 3.987354123384757e-05, |
| "loss": 0.2813, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27831757068634033, |
| "step": 580, |
| "valid_targets_mean": 3726.2, |
| "valid_targets_min": 1063 |
| }, |
| { |
| "epoch": 0.936, |
| "grad_norm": 0.5529862065194503, |
| "learning_rate": 3.9864423833458364e-05, |
| "loss": 0.2803, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33845236897468567, |
| "step": 585, |
| "valid_targets_mean": 3352.5, |
| "valid_targets_min": 1010 |
| }, |
| { |
| "epoch": 0.944, |
| "grad_norm": 0.5276685630923462, |
| "learning_rate": 3.9854990216719285e-05, |
| "loss": 0.2734, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30608147382736206, |
| "step": 590, |
| "valid_targets_mean": 3963.6, |
| "valid_targets_min": 960 |
| }, |
| { |
| "epoch": 0.952, |
| "grad_norm": 0.5761410220991169, |
| "learning_rate": 3.98452405338015e-05, |
| "loss": 0.2941, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3611443042755127, |
| "step": 595, |
| "valid_targets_mean": 3870.4, |
| "valid_targets_min": 670 |
| }, |
| { |
| "epoch": 0.96, |
| "grad_norm": 0.43161899888992733, |
| "learning_rate": 3.983517493990756e-05, |
| "loss": 0.3008, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2567683458328247, |
| "step": 600, |
| "valid_targets_mean": 5753.8, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 0.968, |
| "grad_norm": 0.5174699444690888, |
| "learning_rate": 3.982479359526892e-05, |
| "loss": 0.2846, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31545543670654297, |
| "step": 605, |
| "valid_targets_mean": 4297.7, |
| "valid_targets_min": 1026 |
| }, |
| { |
| "epoch": 0.976, |
| "grad_norm": 0.5582649393151906, |
| "learning_rate": 3.981409666514336e-05, |
| "loss": 0.287, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28723669052124023, |
| "step": 610, |
| "valid_targets_mean": 3499.9, |
| "valid_targets_min": 730 |
| }, |
| { |
| "epoch": 0.984, |
| "grad_norm": 0.4569144879632991, |
| "learning_rate": 3.98030843198124e-05, |
| "loss": 0.2793, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26559823751449585, |
| "step": 615, |
| "valid_targets_mean": 3593.3, |
| "valid_targets_min": 764 |
| }, |
| { |
| "epoch": 0.992, |
| "grad_norm": 0.5364468262792286, |
| "learning_rate": 3.979175673457858e-05, |
| "loss": 0.2979, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2658482491970062, |
| "step": 620, |
| "valid_targets_mean": 3494.4, |
| "valid_targets_min": 670 |
| }, |
| { |
| "epoch": 1.0, |
| "grad_norm": 0.589932569473527, |
| "learning_rate": 3.9780114089762616e-05, |
| "loss": 0.2845, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2865822911262512, |
| "step": 625, |
| "valid_targets_mean": 3167.1, |
| "valid_targets_min": 924 |
| }, |
| { |
| "epoch": 1.008, |
| "grad_norm": 0.4223733533410392, |
| "learning_rate": 3.976815657070062e-05, |
| "loss": 0.2863, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24453572928905487, |
| "step": 630, |
| "valid_targets_mean": 5350.5, |
| "valid_targets_min": 909 |
| }, |
| { |
| "epoch": 1.016, |
| "grad_norm": 0.5617173103201819, |
| "learning_rate": 3.975588436774107e-05, |
| "loss": 0.2779, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2646633982658386, |
| "step": 635, |
| "valid_targets_mean": 3123.4, |
| "valid_targets_min": 1007 |
| }, |
| { |
| "epoch": 1.024, |
| "grad_norm": 0.5490235878628557, |
| "learning_rate": 3.9743297676241826e-05, |
| "loss": 0.2662, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2878129184246063, |
| "step": 640, |
| "valid_targets_mean": 4515.7, |
| "valid_targets_min": 784 |
| }, |
| { |
| "epoch": 1.032, |
| "grad_norm": 0.8694303583221871, |
| "learning_rate": 3.9730396696566994e-05, |
| "loss": 0.299, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39216941595077515, |
| "step": 645, |
| "valid_targets_mean": 1941.7, |
| "valid_targets_min": 1014 |
| }, |
| { |
| "epoch": 1.04, |
| "grad_norm": 0.5176313613959357, |
| "learning_rate": 3.971718163408375e-05, |
| "loss": 0.2728, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2930067777633667, |
| "step": 650, |
| "valid_targets_mean": 4818.8, |
| "valid_targets_min": 1106 |
| }, |
| { |
| "epoch": 1.048, |
| "grad_norm": 0.44358594950793484, |
| "learning_rate": 3.9703652699159093e-05, |
| "loss": 0.2825, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30579259991645813, |
| "step": 655, |
| "valid_targets_mean": 5172.5, |
| "valid_targets_min": 871 |
| }, |
| { |
| "epoch": 1.056, |
| "grad_norm": 0.5740468844331686, |
| "learning_rate": 3.9689810107156425e-05, |
| "loss": 0.2827, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3474939167499542, |
| "step": 660, |
| "valid_targets_mean": 4049.9, |
| "valid_targets_min": 821 |
| }, |
| { |
| "epoch": 1.064, |
| "grad_norm": 0.5556529187566539, |
| "learning_rate": 3.967565407843222e-05, |
| "loss": 0.2977, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29836633801460266, |
| "step": 665, |
| "valid_targets_mean": 4204.5, |
| "valid_targets_min": 1348 |
| }, |
| { |
| "epoch": 1.072, |
| "grad_norm": 0.5024698593720891, |
| "learning_rate": 3.966118483833242e-05, |
| "loss": 0.2738, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26958394050598145, |
| "step": 670, |
| "valid_targets_mean": 4268.8, |
| "valid_targets_min": 1033 |
| }, |
| { |
| "epoch": 1.08, |
| "grad_norm": 0.4842800380286951, |
| "learning_rate": 3.964640261718893e-05, |
| "loss": 0.2664, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26958194375038147, |
| "step": 675, |
| "valid_targets_mean": 3595.6, |
| "valid_targets_min": 862 |
| }, |
| { |
| "epoch": 1.088, |
| "grad_norm": 0.5273262657978705, |
| "learning_rate": 3.963130765031589e-05, |
| "loss": 0.2808, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2689908444881439, |
| "step": 680, |
| "valid_targets_mean": 3801.2, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 1.096, |
| "grad_norm": 0.5751221396165925, |
| "learning_rate": 3.961590017800598e-05, |
| "loss": 0.2775, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2882736027240753, |
| "step": 685, |
| "valid_targets_mean": 3101.3, |
| "valid_targets_min": 664 |
| }, |
| { |
| "epoch": 1.104, |
| "grad_norm": 0.5471508339834354, |
| "learning_rate": 3.960018044552653e-05, |
| "loss": 0.2994, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31060564517974854, |
| "step": 690, |
| "valid_targets_mean": 3763.3, |
| "valid_targets_min": 846 |
| }, |
| { |
| "epoch": 1.112, |
| "grad_norm": 0.5792478778548963, |
| "learning_rate": 3.9584148703115704e-05, |
| "loss": 0.2829, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2743430435657501, |
| "step": 695, |
| "valid_targets_mean": 2966.4, |
| "valid_targets_min": 1021 |
| }, |
| { |
| "epoch": 1.12, |
| "grad_norm": 0.5701845098447167, |
| "learning_rate": 3.956780520597842e-05, |
| "loss": 0.2719, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31023383140563965, |
| "step": 700, |
| "valid_targets_mean": 3593.6, |
| "valid_targets_min": 748 |
| }, |
| { |
| "epoch": 1.1280000000000001, |
| "grad_norm": 0.4243313433223685, |
| "learning_rate": 3.955115021428236e-05, |
| "loss": 0.2655, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2508736550807953, |
| "step": 705, |
| "valid_targets_mean": 5902.2, |
| "valid_targets_min": 912 |
| }, |
| { |
| "epoch": 1.1360000000000001, |
| "grad_norm": 0.4790109609572266, |
| "learning_rate": 3.95341839931538e-05, |
| "loss": 0.2804, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2982293367385864, |
| "step": 710, |
| "valid_targets_mean": 5286.7, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 1.144, |
| "grad_norm": 0.5585199616419411, |
| "learning_rate": 3.95169068126734e-05, |
| "loss": 0.2816, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2889254093170166, |
| "step": 715, |
| "valid_targets_mean": 2680.4, |
| "valid_targets_min": 822 |
| }, |
| { |
| "epoch": 1.152, |
| "grad_norm": 0.393666834541853, |
| "learning_rate": 3.949931894787187e-05, |
| "loss": 0.243, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22171318531036377, |
| "step": 720, |
| "valid_targets_mean": 6140.5, |
| "valid_targets_min": 913 |
| }, |
| { |
| "epoch": 1.16, |
| "grad_norm": 0.4554560378266813, |
| "learning_rate": 3.948142067872565e-05, |
| "loss": 0.2743, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.276266485452652, |
| "step": 725, |
| "valid_targets_mean": 4809.1, |
| "valid_targets_min": 1074 |
| }, |
| { |
| "epoch": 1.168, |
| "grad_norm": 0.4943688708217394, |
| "learning_rate": 3.946321229015241e-05, |
| "loss": 0.2716, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2835407853126526, |
| "step": 730, |
| "valid_targets_mean": 4429.9, |
| "valid_targets_min": 1050 |
| }, |
| { |
| "epoch": 1.176, |
| "grad_norm": 0.4477093216825828, |
| "learning_rate": 3.944469407200652e-05, |
| "loss": 0.2788, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23854275047779083, |
| "step": 735, |
| "valid_targets_mean": 4396.1, |
| "valid_targets_min": 855 |
| }, |
| { |
| "epoch": 1.184, |
| "grad_norm": 0.5896727157094116, |
| "learning_rate": 3.942586631907444e-05, |
| "loss": 0.2705, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2517150044441223, |
| "step": 740, |
| "valid_targets_mean": 3115.4, |
| "valid_targets_min": 850 |
| }, |
| { |
| "epoch": 1.192, |
| "grad_norm": 0.5289038060630349, |
| "learning_rate": 3.9406729331070054e-05, |
| "loss": 0.2763, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2541031539440155, |
| "step": 745, |
| "valid_targets_mean": 3815.3, |
| "valid_targets_min": 962 |
| }, |
| { |
| "epoch": 1.2, |
| "grad_norm": 0.44733647620635336, |
| "learning_rate": 3.938728341262985e-05, |
| "loss": 0.291, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2868531048297882, |
| "step": 750, |
| "valid_targets_mean": 4895.9, |
| "valid_targets_min": 623 |
| }, |
| { |
| "epoch": 1.208, |
| "grad_norm": 0.7202988131820695, |
| "learning_rate": 3.936752887330812e-05, |
| "loss": 0.289, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3202763497829437, |
| "step": 755, |
| "valid_targets_mean": 2302.6, |
| "valid_targets_min": 929 |
| }, |
| { |
| "epoch": 1.216, |
| "grad_norm": 0.5140036991263218, |
| "learning_rate": 3.9347466027571975e-05, |
| "loss": 0.2653, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2578458786010742, |
| "step": 760, |
| "valid_targets_mean": 4010.8, |
| "valid_targets_min": 832 |
| }, |
| { |
| "epoch": 1.224, |
| "grad_norm": 0.5537418130011722, |
| "learning_rate": 3.932709519479639e-05, |
| "loss": 0.2653, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2952079176902771, |
| "step": 765, |
| "valid_targets_mean": 3206.0, |
| "valid_targets_min": 800 |
| }, |
| { |
| "epoch": 1.232, |
| "grad_norm": 0.49323600062823847, |
| "learning_rate": 3.930641669925911e-05, |
| "loss": 0.2759, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.264598548412323, |
| "step": 770, |
| "valid_targets_mean": 3837.5, |
| "valid_targets_min": 878 |
| }, |
| { |
| "epoch": 1.24, |
| "grad_norm": 0.46711652952383553, |
| "learning_rate": 3.928543087013546e-05, |
| "loss": 0.2778, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27722471952438354, |
| "step": 775, |
| "valid_targets_mean": 4181.8, |
| "valid_targets_min": 877 |
| }, |
| { |
| "epoch": 1.248, |
| "grad_norm": 0.43556151715805724, |
| "learning_rate": 3.926413804149315e-05, |
| "loss": 0.2631, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21522939205169678, |
| "step": 780, |
| "valid_targets_mean": 4802.7, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 1.256, |
| "grad_norm": 0.6366361986050708, |
| "learning_rate": 3.9242538552286894e-05, |
| "loss": 0.2921, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32596951723098755, |
| "step": 785, |
| "valid_targets_mean": 2695.1, |
| "valid_targets_min": 707 |
| }, |
| { |
| "epoch": 1.264, |
| "grad_norm": 0.5512190834498045, |
| "learning_rate": 3.9220632746353096e-05, |
| "loss": 0.2994, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3354596495628357, |
| "step": 790, |
| "valid_targets_mean": 3169.8, |
| "valid_targets_min": 714 |
| }, |
| { |
| "epoch": 1.272, |
| "grad_norm": 0.4994124792250944, |
| "learning_rate": 3.91984209724043e-05, |
| "loss": 0.253, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24309590458869934, |
| "step": 795, |
| "valid_targets_mean": 3187.5, |
| "valid_targets_min": 717 |
| }, |
| { |
| "epoch": 1.28, |
| "grad_norm": 0.5058818429479062, |
| "learning_rate": 3.917590358402369e-05, |
| "loss": 0.2748, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2896243929862976, |
| "step": 800, |
| "valid_targets_mean": 4368.1, |
| "valid_targets_min": 662 |
| }, |
| { |
| "epoch": 1.288, |
| "grad_norm": 0.5741606754229182, |
| "learning_rate": 3.915308093965943e-05, |
| "loss": 0.2857, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31103113293647766, |
| "step": 805, |
| "valid_targets_mean": 3463.2, |
| "valid_targets_min": 1093 |
| }, |
| { |
| "epoch": 1.296, |
| "grad_norm": 0.5269169994703332, |
| "learning_rate": 3.9129953402618976e-05, |
| "loss": 0.2764, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28437525033950806, |
| "step": 810, |
| "valid_targets_mean": 3127.5, |
| "valid_targets_min": 768 |
| }, |
| { |
| "epoch": 1.304, |
| "grad_norm": 0.5105272259938182, |
| "learning_rate": 3.91065213410633e-05, |
| "loss": 0.2647, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27728912234306335, |
| "step": 815, |
| "valid_targets_mean": 3164.8, |
| "valid_targets_min": 954 |
| }, |
| { |
| "epoch": 1.312, |
| "grad_norm": 0.7068657863624991, |
| "learning_rate": 3.908278512800098e-05, |
| "loss": 0.2749, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30154961347579956, |
| "step": 820, |
| "valid_targets_mean": 2021.1, |
| "valid_targets_min": 771 |
| }, |
| { |
| "epoch": 1.32, |
| "grad_norm": 0.43262404009468025, |
| "learning_rate": 3.905874514128235e-05, |
| "loss": 0.2574, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22265702486038208, |
| "step": 825, |
| "valid_targets_mean": 5069.6, |
| "valid_targets_min": 1046 |
| }, |
| { |
| "epoch": 1.328, |
| "grad_norm": 0.4833800812123599, |
| "learning_rate": 3.903440176359338e-05, |
| "loss": 0.2652, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2597518265247345, |
| "step": 830, |
| "valid_targets_mean": 4273.6, |
| "valid_targets_min": 642 |
| }, |
| { |
| "epoch": 1.336, |
| "grad_norm": 0.495986902645752, |
| "learning_rate": 3.90097553824497e-05, |
| "loss": 0.2649, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26148658990859985, |
| "step": 835, |
| "valid_targets_mean": 3473.2, |
| "valid_targets_min": 558 |
| }, |
| { |
| "epoch": 1.3439999999999999, |
| "grad_norm": 0.51848462208148, |
| "learning_rate": 3.8984806390190304e-05, |
| "loss": 0.2636, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30521005392074585, |
| "step": 840, |
| "valid_targets_mean": 3706.1, |
| "valid_targets_min": 1051 |
| }, |
| { |
| "epoch": 1.3519999999999999, |
| "grad_norm": 0.557804275665896, |
| "learning_rate": 3.895955518397141e-05, |
| "loss": 0.264, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2697344124317169, |
| "step": 845, |
| "valid_targets_mean": 2746.5, |
| "valid_targets_min": 976 |
| }, |
| { |
| "epoch": 1.3599999999999999, |
| "grad_norm": 0.4866612431488525, |
| "learning_rate": 3.893400216576011e-05, |
| "loss": 0.2739, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24589639902114868, |
| "step": 850, |
| "valid_targets_mean": 3582.2, |
| "valid_targets_min": 765 |
| }, |
| { |
| "epoch": 1.3679999999999999, |
| "grad_norm": 0.46312511696658815, |
| "learning_rate": 3.89081477423279e-05, |
| "loss": 0.2593, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25814926624298096, |
| "step": 855, |
| "valid_targets_mean": 4407.3, |
| "valid_targets_min": 820 |
| }, |
| { |
| "epoch": 1.376, |
| "grad_norm": 0.4559151598342351, |
| "learning_rate": 3.888199232524434e-05, |
| "loss": 0.2737, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27635207772254944, |
| "step": 860, |
| "valid_targets_mean": 4211.1, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 1.384, |
| "grad_norm": 0.42662571052348236, |
| "learning_rate": 3.8855536330870354e-05, |
| "loss": 0.262, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27323490381240845, |
| "step": 865, |
| "valid_targets_mean": 5150.4, |
| "valid_targets_min": 851 |
| }, |
| { |
| "epoch": 1.392, |
| "grad_norm": 0.43724344437427815, |
| "learning_rate": 3.882878018035173e-05, |
| "loss": 0.2682, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.298825204372406, |
| "step": 870, |
| "valid_targets_mean": 4823.4, |
| "valid_targets_min": 648 |
| }, |
| { |
| "epoch": 1.4, |
| "grad_norm": 0.4513041544448632, |
| "learning_rate": 3.880172429961232e-05, |
| "loss": 0.2641, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2363213747739792, |
| "step": 875, |
| "valid_targets_mean": 4356.6, |
| "valid_targets_min": 676 |
| }, |
| { |
| "epoch": 1.408, |
| "grad_norm": 0.4023862024488909, |
| "learning_rate": 3.877436911934733e-05, |
| "loss": 0.2825, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26997482776641846, |
| "step": 880, |
| "valid_targets_mean": 5533.5, |
| "valid_targets_min": 904 |
| }, |
| { |
| "epoch": 1.416, |
| "grad_norm": 0.7080890847938507, |
| "learning_rate": 3.874671507501641e-05, |
| "loss": 0.2796, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3377278745174408, |
| "step": 885, |
| "valid_targets_mean": 2200.1, |
| "valid_targets_min": 912 |
| }, |
| { |
| "epoch": 1.424, |
| "grad_norm": 0.5267587721617124, |
| "learning_rate": 3.871876260683677e-05, |
| "loss": 0.2809, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2512984275817871, |
| "step": 890, |
| "valid_targets_mean": 3473.6, |
| "valid_targets_min": 810 |
| }, |
| { |
| "epoch": 1.432, |
| "grad_norm": 0.5790695758130487, |
| "learning_rate": 3.869051215977612e-05, |
| "loss": 0.2884, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32350438833236694, |
| "step": 895, |
| "valid_targets_mean": 3247.3, |
| "valid_targets_min": 822 |
| }, |
| { |
| "epoch": 1.44, |
| "grad_norm": 0.4803005137644847, |
| "learning_rate": 3.8661964183545634e-05, |
| "loss": 0.2766, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27841246128082275, |
| "step": 900, |
| "valid_targets_mean": 4475.9, |
| "valid_targets_min": 948 |
| }, |
| { |
| "epoch": 1.448, |
| "grad_norm": 0.475901655735222, |
| "learning_rate": 3.863311913259276e-05, |
| "loss": 0.2797, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33764156699180603, |
| "step": 905, |
| "valid_targets_mean": 5242.4, |
| "valid_targets_min": 826 |
| }, |
| { |
| "epoch": 1.456, |
| "grad_norm": 0.748149672688827, |
| "learning_rate": 3.860397746609402e-05, |
| "loss": 0.306, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31807559728622437, |
| "step": 910, |
| "valid_targets_mean": 2950.9, |
| "valid_targets_min": 816 |
| }, |
| { |
| "epoch": 1.464, |
| "grad_norm": 0.37580443450106377, |
| "learning_rate": 3.857453964794764e-05, |
| "loss": 0.2834, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2712574005126953, |
| "step": 915, |
| "valid_targets_mean": 6658.7, |
| "valid_targets_min": 1186 |
| }, |
| { |
| "epoch": 1.472, |
| "grad_norm": 0.45030105000964843, |
| "learning_rate": 3.854480614676624e-05, |
| "loss": 0.2789, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.272754043340683, |
| "step": 920, |
| "valid_targets_mean": 3925.5, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 1.48, |
| "grad_norm": 0.34923108926716045, |
| "learning_rate": 3.851477743586932e-05, |
| "loss": 0.2515, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2357773333787918, |
| "step": 925, |
| "valid_targets_mean": 6845.4, |
| "valid_targets_min": 998 |
| }, |
| { |
| "epoch": 1.488, |
| "grad_norm": 0.5115811479757503, |
| "learning_rate": 3.8484453993275746e-05, |
| "loss": 0.2768, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23984485864639282, |
| "step": 930, |
| "valid_targets_mean": 3133.7, |
| "valid_targets_min": 774 |
| }, |
| { |
| "epoch": 1.496, |
| "grad_norm": 0.48452858019317047, |
| "learning_rate": 3.8453836301696134e-05, |
| "loss": 0.2732, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2921665608882904, |
| "step": 935, |
| "valid_targets_mean": 3872.1, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 1.504, |
| "grad_norm": 0.5141883495932179, |
| "learning_rate": 3.842292484852518e-05, |
| "loss": 0.2772, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2862679958343506, |
| "step": 940, |
| "valid_targets_mean": 4214.1, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 1.512, |
| "grad_norm": 0.46283592158032705, |
| "learning_rate": 3.8391720125833875e-05, |
| "loss": 0.282, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2581455707550049, |
| "step": 945, |
| "valid_targets_mean": 3914.6, |
| "valid_targets_min": 987 |
| }, |
| { |
| "epoch": 1.52, |
| "grad_norm": 0.48706765716515343, |
| "learning_rate": 3.83602226303617e-05, |
| "loss": 0.2863, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2788928747177124, |
| "step": 950, |
| "valid_targets_mean": 4213.4, |
| "valid_targets_min": 839 |
| }, |
| { |
| "epoch": 1.528, |
| "grad_norm": 0.585464765294358, |
| "learning_rate": 3.83284328635087e-05, |
| "loss": 0.2678, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2879747152328491, |
| "step": 955, |
| "valid_targets_mean": 3057.9, |
| "valid_targets_min": 932 |
| }, |
| { |
| "epoch": 1.536, |
| "grad_norm": 0.5315853015293113, |
| "learning_rate": 3.829635133132751e-05, |
| "loss": 0.2663, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2610795199871063, |
| "step": 960, |
| "valid_targets_mean": 3284.8, |
| "valid_targets_min": 733 |
| }, |
| { |
| "epoch": 1.544, |
| "grad_norm": 0.5296944336561442, |
| "learning_rate": 3.8263978544515304e-05, |
| "loss": 0.2836, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26374924182891846, |
| "step": 965, |
| "valid_targets_mean": 3174.6, |
| "valid_targets_min": 824 |
| }, |
| { |
| "epoch": 1.552, |
| "grad_norm": 0.4659744673324122, |
| "learning_rate": 3.823131501840565e-05, |
| "loss": 0.2616, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27271947264671326, |
| "step": 970, |
| "valid_targets_mean": 3992.8, |
| "valid_targets_min": 786 |
| }, |
| { |
| "epoch": 1.56, |
| "grad_norm": 0.41130937508304954, |
| "learning_rate": 3.819836127296032e-05, |
| "loss": 0.2874, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26125800609588623, |
| "step": 975, |
| "valid_targets_mean": 4572.8, |
| "valid_targets_min": 587 |
| }, |
| { |
| "epoch": 1.568, |
| "grad_norm": 0.5146658247796757, |
| "learning_rate": 3.8165117832761016e-05, |
| "loss": 0.2861, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3052479326725006, |
| "step": 980, |
| "valid_targets_mean": 3267.8, |
| "valid_targets_min": 963 |
| }, |
| { |
| "epoch": 1.576, |
| "grad_norm": 0.5707618643836874, |
| "learning_rate": 3.813158522700098e-05, |
| "loss": 0.2592, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3040449619293213, |
| "step": 985, |
| "valid_targets_mean": 3492.4, |
| "valid_targets_min": 995 |
| }, |
| { |
| "epoch": 1.584, |
| "grad_norm": 0.45611771016792524, |
| "learning_rate": 3.809776398947665e-05, |
| "loss": 0.2721, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2837957441806793, |
| "step": 990, |
| "valid_targets_mean": 4701.9, |
| "valid_targets_min": 1074 |
| }, |
| { |
| "epoch": 1.592, |
| "grad_norm": 0.4630680235085982, |
| "learning_rate": 3.806365465857908e-05, |
| "loss": 0.2701, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2922239601612091, |
| "step": 995, |
| "valid_targets_mean": 3976.1, |
| "valid_targets_min": 918 |
| }, |
| { |
| "epoch": 1.6, |
| "grad_norm": 0.399096117534781, |
| "learning_rate": 3.802925777728541e-05, |
| "loss": 0.2633, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2673601806163788, |
| "step": 1000, |
| "valid_targets_mean": 5512.1, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 1.608, |
| "grad_norm": 0.41627336969619316, |
| "learning_rate": 3.799457389315023e-05, |
| "loss": 0.2574, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2412080317735672, |
| "step": 1005, |
| "valid_targets_mean": 4379.9, |
| "valid_targets_min": 1023 |
| }, |
| { |
| "epoch": 1.616, |
| "grad_norm": 0.440295852006296, |
| "learning_rate": 3.795960355829683e-05, |
| "loss": 0.2708, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2329067587852478, |
| "step": 1010, |
| "valid_targets_mean": 4429.0, |
| "valid_targets_min": 1099 |
| }, |
| { |
| "epoch": 1.624, |
| "grad_norm": 0.5318475478532778, |
| "learning_rate": 3.7924347329408444e-05, |
| "loss": 0.2674, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2775961756706238, |
| "step": 1015, |
| "valid_targets_mean": 2742.1, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 1.6320000000000001, |
| "grad_norm": 0.4161727759675333, |
| "learning_rate": 3.788880576771937e-05, |
| "loss": 0.2617, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24313268065452576, |
| "step": 1020, |
| "valid_targets_mean": 5492.8, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 1.6400000000000001, |
| "grad_norm": 0.7999546972605147, |
| "learning_rate": 3.785297943900605e-05, |
| "loss": 0.2627, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28334081172943115, |
| "step": 1025, |
| "valid_targets_mean": 5762.0, |
| "valid_targets_min": 852 |
| }, |
| { |
| "epoch": 1.6480000000000001, |
| "grad_norm": 0.48331284388571666, |
| "learning_rate": 3.7816868913578044e-05, |
| "loss": 0.2624, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2545091509819031, |
| "step": 1030, |
| "valid_targets_mean": 3178.3, |
| "valid_targets_min": 728 |
| }, |
| { |
| "epoch": 1.6560000000000001, |
| "grad_norm": 0.45991402061064773, |
| "learning_rate": 3.778047476626897e-05, |
| "loss": 0.2936, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25272148847579956, |
| "step": 1035, |
| "valid_targets_mean": 4338.1, |
| "valid_targets_min": 1049 |
| }, |
| { |
| "epoch": 1.6640000000000001, |
| "grad_norm": 0.3451764939977074, |
| "learning_rate": 3.7743797576427335e-05, |
| "loss": 0.2611, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2401534467935562, |
| "step": 1040, |
| "valid_targets_mean": 6253.1, |
| "valid_targets_min": 712 |
| }, |
| { |
| "epoch": 1.6720000000000002, |
| "grad_norm": 0.4111940706574221, |
| "learning_rate": 3.770683792790733e-05, |
| "loss": 0.2651, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2934751510620117, |
| "step": 1045, |
| "valid_targets_mean": 4744.1, |
| "valid_targets_min": 1059 |
| }, |
| { |
| "epoch": 1.6800000000000002, |
| "grad_norm": 0.4087263278002059, |
| "learning_rate": 3.766959640905954e-05, |
| "loss": 0.2562, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2707015872001648, |
| "step": 1050, |
| "valid_targets_mean": 5591.9, |
| "valid_targets_min": 710 |
| }, |
| { |
| "epoch": 1.688, |
| "grad_norm": 0.5089485422227342, |
| "learning_rate": 3.763207361272153e-05, |
| "loss": 0.2569, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27094948291778564, |
| "step": 1055, |
| "valid_targets_mean": 3115.1, |
| "valid_targets_min": 787 |
| }, |
| { |
| "epoch": 1.696, |
| "grad_norm": 0.39668167250416864, |
| "learning_rate": 3.759427013620849e-05, |
| "loss": 0.2525, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23788487911224365, |
| "step": 1060, |
| "valid_targets_mean": 4781.2, |
| "valid_targets_min": 1059 |
| }, |
| { |
| "epoch": 1.704, |
| "grad_norm": 0.4307058651279445, |
| "learning_rate": 3.755618658130366e-05, |
| "loss": 0.2426, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2379121482372284, |
| "step": 1065, |
| "valid_targets_mean": 3575.8, |
| "valid_targets_min": 730 |
| }, |
| { |
| "epoch": 1.712, |
| "grad_norm": 0.6248314289543657, |
| "learning_rate": 3.751782355424877e-05, |
| "loss": 0.3129, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33632132411003113, |
| "step": 1070, |
| "valid_targets_mean": 2413.9, |
| "valid_targets_min": 919 |
| }, |
| { |
| "epoch": 1.72, |
| "grad_norm": 0.3832687956684591, |
| "learning_rate": 3.7479181665734395e-05, |
| "loss": 0.2513, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27428677678108215, |
| "step": 1075, |
| "valid_targets_mean": 5823.8, |
| "valid_targets_min": 795 |
| }, |
| { |
| "epoch": 1.728, |
| "grad_norm": 0.4146423424953804, |
| "learning_rate": 3.7440261530890213e-05, |
| "loss": 0.2665, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24523381888866425, |
| "step": 1080, |
| "valid_targets_mean": 5357.3, |
| "valid_targets_min": 1045 |
| }, |
| { |
| "epoch": 1.736, |
| "grad_norm": 0.4881475237391791, |
| "learning_rate": 3.740106376927527e-05, |
| "loss": 0.2525, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25967293977737427, |
| "step": 1085, |
| "valid_targets_mean": 3166.9, |
| "valid_targets_min": 756 |
| }, |
| { |
| "epoch": 1.744, |
| "grad_norm": 0.4001288366939507, |
| "learning_rate": 3.7361589004868035e-05, |
| "loss": 0.2772, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2417476922273636, |
| "step": 1090, |
| "valid_targets_mean": 4840.1, |
| "valid_targets_min": 693 |
| }, |
| { |
| "epoch": 1.752, |
| "grad_norm": 0.4845852844999974, |
| "learning_rate": 3.7321837866056535e-05, |
| "loss": 0.2736, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28427213430404663, |
| "step": 1095, |
| "valid_targets_mean": 3742.1, |
| "valid_targets_min": 993 |
| }, |
| { |
| "epoch": 1.76, |
| "grad_norm": 0.44706609607667686, |
| "learning_rate": 3.728181098562831e-05, |
| "loss": 0.2547, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23905465006828308, |
| "step": 1100, |
| "valid_targets_mean": 4081.2, |
| "valid_targets_min": 888 |
| }, |
| { |
| "epoch": 1.768, |
| "grad_norm": 0.4606642272123344, |
| "learning_rate": 3.7241509000760355e-05, |
| "loss": 0.2919, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2809835374355316, |
| "step": 1105, |
| "valid_targets_mean": 4572.8, |
| "valid_targets_min": 790 |
| }, |
| { |
| "epoch": 1.776, |
| "grad_norm": 0.44374131291191704, |
| "learning_rate": 3.720093255300899e-05, |
| "loss": 0.2622, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23165518045425415, |
| "step": 1110, |
| "valid_targets_mean": 5182.4, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 1.784, |
| "grad_norm": 0.39542261654137617, |
| "learning_rate": 3.7160082288299645e-05, |
| "loss": 0.2495, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26467621326446533, |
| "step": 1115, |
| "valid_targets_mean": 6130.7, |
| "valid_targets_min": 1011 |
| }, |
| { |
| "epoch": 1.792, |
| "grad_norm": 0.40912132064625895, |
| "learning_rate": 3.7118958856916534e-05, |
| "loss": 0.2694, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25465548038482666, |
| "step": 1120, |
| "valid_targets_mean": 6377.2, |
| "valid_targets_min": 1032 |
| }, |
| { |
| "epoch": 1.8, |
| "grad_norm": 0.39755902551807576, |
| "learning_rate": 3.707756291349237e-05, |
| "loss": 0.2684, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2527580261230469, |
| "step": 1125, |
| "valid_targets_mean": 5014.4, |
| "valid_targets_min": 1094 |
| }, |
| { |
| "epoch": 1.808, |
| "grad_norm": 0.514457815285411, |
| "learning_rate": 3.703589511699787e-05, |
| "loss": 0.265, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2838781476020813, |
| "step": 1130, |
| "valid_targets_mean": 3433.5, |
| "valid_targets_min": 975 |
| }, |
| { |
| "epoch": 1.8159999999999998, |
| "grad_norm": 0.44550270673299985, |
| "learning_rate": 3.6993956130731355e-05, |
| "loss": 0.2586, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2567026615142822, |
| "step": 1135, |
| "valid_targets_mean": 3949.9, |
| "valid_targets_min": 976 |
| }, |
| { |
| "epoch": 1.8239999999999998, |
| "grad_norm": 0.5422815583872076, |
| "learning_rate": 3.6951746622308106e-05, |
| "loss": 0.2691, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2764151692390442, |
| "step": 1140, |
| "valid_targets_mean": 2868.9, |
| "valid_targets_min": 1072 |
| }, |
| { |
| "epoch": 1.8319999999999999, |
| "grad_norm": 0.41869577320813534, |
| "learning_rate": 3.69092672636498e-05, |
| "loss": 0.2932, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28278565406799316, |
| "step": 1145, |
| "valid_targets_mean": 5837.7, |
| "valid_targets_min": 1035 |
| }, |
| { |
| "epoch": 1.8399999999999999, |
| "grad_norm": 0.3757713903865787, |
| "learning_rate": 3.686651873097375e-05, |
| "loss": 0.2985, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23112526535987854, |
| "step": 1150, |
| "valid_targets_mean": 6685.0, |
| "valid_targets_min": 932 |
| }, |
| { |
| "epoch": 1.8479999999999999, |
| "grad_norm": 0.4409326357854815, |
| "learning_rate": 3.682350170478223e-05, |
| "loss": 0.2569, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23602645099163055, |
| "step": 1155, |
| "valid_targets_mean": 5068.7, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 1.8559999999999999, |
| "grad_norm": 0.40528589174751617, |
| "learning_rate": 3.678021686985153e-05, |
| "loss": 0.2737, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2743171453475952, |
| "step": 1160, |
| "valid_targets_mean": 4716.9, |
| "valid_targets_min": 1127 |
| }, |
| { |
| "epoch": 1.8639999999999999, |
| "grad_norm": 0.41676048150188805, |
| "learning_rate": 3.6736664915221144e-05, |
| "loss": 0.2609, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23870986700057983, |
| "step": 1165, |
| "valid_targets_mean": 4469.6, |
| "valid_targets_min": 843 |
| }, |
| { |
| "epoch": 1.8719999999999999, |
| "grad_norm": 0.5284619471071941, |
| "learning_rate": 3.669284653418278e-05, |
| "loss": 0.2659, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29566067457199097, |
| "step": 1170, |
| "valid_targets_mean": 3079.8, |
| "valid_targets_min": 710 |
| }, |
| { |
| "epoch": 1.88, |
| "grad_norm": 0.47907263465195665, |
| "learning_rate": 3.6648762424269306e-05, |
| "loss": 0.2626, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2716890275478363, |
| "step": 1175, |
| "valid_targets_mean": 3397.4, |
| "valid_targets_min": 884 |
| }, |
| { |
| "epoch": 1.888, |
| "grad_norm": 0.41124099809929116, |
| "learning_rate": 3.660441328724365e-05, |
| "loss": 0.2859, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26922544836997986, |
| "step": 1180, |
| "valid_targets_mean": 5981.5, |
| "valid_targets_min": 1078 |
| }, |
| { |
| "epoch": 1.896, |
| "grad_norm": 0.46088125238526284, |
| "learning_rate": 3.655979982908764e-05, |
| "loss": 0.2667, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25594204664230347, |
| "step": 1185, |
| "valid_targets_mean": 3616.8, |
| "valid_targets_min": 719 |
| }, |
| { |
| "epoch": 1.904, |
| "grad_norm": 0.4543017272146403, |
| "learning_rate": 3.6514922759990756e-05, |
| "loss": 0.2744, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26674598455429077, |
| "step": 1190, |
| "valid_targets_mean": 3739.4, |
| "valid_targets_min": 1079 |
| }, |
| { |
| "epoch": 1.912, |
| "grad_norm": 0.4192846645781652, |
| "learning_rate": 3.646978279433883e-05, |
| "loss": 0.2673, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2927936911582947, |
| "step": 1195, |
| "valid_targets_mean": 5007.6, |
| "valid_targets_min": 938 |
| }, |
| { |
| "epoch": 1.92, |
| "grad_norm": 0.4436459924254246, |
| "learning_rate": 3.6424380650702685e-05, |
| "loss": 0.2573, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26739734411239624, |
| "step": 1200, |
| "valid_targets_mean": 4250.5, |
| "valid_targets_min": 721 |
| }, |
| { |
| "epoch": 1.928, |
| "grad_norm": 0.4240018692969433, |
| "learning_rate": 3.637871705182667e-05, |
| "loss": 0.2787, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24744366109371185, |
| "step": 1205, |
| "valid_targets_mean": 5123.5, |
| "valid_targets_min": 968 |
| }, |
| { |
| "epoch": 1.936, |
| "grad_norm": 0.36924402152142755, |
| "learning_rate": 3.633279272461717e-05, |
| "loss": 0.2657, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23109649121761322, |
| "step": 1210, |
| "valid_targets_mean": 5121.6, |
| "valid_targets_min": 710 |
| }, |
| { |
| "epoch": 1.944, |
| "grad_norm": 0.3870106636312311, |
| "learning_rate": 3.628660840013102e-05, |
| "loss": 0.2727, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24079151451587677, |
| "step": 1215, |
| "valid_targets_mean": 5143.0, |
| "valid_targets_min": 916 |
| }, |
| { |
| "epoch": 1.952, |
| "grad_norm": 0.47018664339123123, |
| "learning_rate": 3.624016481356392e-05, |
| "loss": 0.2725, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28172844648361206, |
| "step": 1220, |
| "valid_targets_mean": 3450.4, |
| "valid_targets_min": 616 |
| }, |
| { |
| "epoch": 1.96, |
| "grad_norm": 0.46397233622758055, |
| "learning_rate": 3.619346270423866e-05, |
| "loss": 0.2552, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2602039575576782, |
| "step": 1225, |
| "valid_targets_mean": 3486.5, |
| "valid_targets_min": 815 |
| }, |
| { |
| "epoch": 1.968, |
| "grad_norm": 0.43784857356480683, |
| "learning_rate": 3.6146502815593384e-05, |
| "loss": 0.2578, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26954376697540283, |
| "step": 1230, |
| "valid_targets_mean": 4244.4, |
| "valid_targets_min": 897 |
| }, |
| { |
| "epoch": 1.976, |
| "grad_norm": 0.3574426307276786, |
| "learning_rate": 3.609928589516977e-05, |
| "loss": 0.2739, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2643759846687317, |
| "step": 1235, |
| "valid_targets_mean": 5850.1, |
| "valid_targets_min": 1066 |
| }, |
| { |
| "epoch": 1.984, |
| "grad_norm": 0.38025251874050797, |
| "learning_rate": 3.6051812694601114e-05, |
| "loss": 0.2714, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2867104411125183, |
| "step": 1240, |
| "valid_targets_mean": 6052.6, |
| "valid_targets_min": 796 |
| }, |
| { |
| "epoch": 1.992, |
| "grad_norm": 0.4749876308252262, |
| "learning_rate": 3.6004083969600346e-05, |
| "loss": 0.2653, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2491377741098404, |
| "step": 1245, |
| "valid_targets_mean": 3096.4, |
| "valid_targets_min": 881 |
| }, |
| { |
| "epoch": 2.0, |
| "grad_norm": 0.39416119445325465, |
| "learning_rate": 3.595610047994804e-05, |
| "loss": 0.2393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21893836557865143, |
| "step": 1250, |
| "valid_targets_mean": 4368.7, |
| "valid_targets_min": 685 |
| }, |
| { |
| "epoch": 2.008, |
| "grad_norm": 0.4997159356103167, |
| "learning_rate": 3.5907862989480285e-05, |
| "loss": 0.2607, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24703270196914673, |
| "step": 1255, |
| "valid_targets_mean": 3216.5, |
| "valid_targets_min": 943 |
| }, |
| { |
| "epoch": 2.016, |
| "grad_norm": 0.537153879112119, |
| "learning_rate": 3.585937226607656e-05, |
| "loss": 0.2395, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24673984944820404, |
| "step": 1260, |
| "valid_targets_mean": 3123.9, |
| "valid_targets_min": 626 |
| }, |
| { |
| "epoch": 2.024, |
| "grad_norm": 0.5137006043744009, |
| "learning_rate": 3.5810629081647476e-05, |
| "loss": 0.2712, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25879064202308655, |
| "step": 1265, |
| "valid_targets_mean": 2930.7, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 2.032, |
| "grad_norm": 0.39071865921634885, |
| "learning_rate": 3.576163421212249e-05, |
| "loss": 0.2366, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21474619209766388, |
| "step": 1270, |
| "valid_targets_mean": 5110.3, |
| "valid_targets_min": 879 |
| }, |
| { |
| "epoch": 2.04, |
| "grad_norm": 0.42220001994538664, |
| "learning_rate": 3.5712388437437576e-05, |
| "loss": 0.2688, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2964683473110199, |
| "step": 1275, |
| "valid_targets_mean": 5167.3, |
| "valid_targets_min": 918 |
| }, |
| { |
| "epoch": 2.048, |
| "grad_norm": 0.49759776720417487, |
| "learning_rate": 3.566289254152283e-05, |
| "loss": 0.2405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25483906269073486, |
| "step": 1280, |
| "valid_targets_mean": 3432.9, |
| "valid_targets_min": 850 |
| }, |
| { |
| "epoch": 2.056, |
| "grad_norm": 0.43733797744170083, |
| "learning_rate": 3.56131473122899e-05, |
| "loss": 0.2627, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2655543386936188, |
| "step": 1285, |
| "valid_targets_mean": 4495.1, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 2.064, |
| "grad_norm": 0.4421965923645115, |
| "learning_rate": 3.556315354161955e-05, |
| "loss": 0.2412, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23384560644626617, |
| "step": 1290, |
| "valid_targets_mean": 4117.4, |
| "valid_targets_min": 828 |
| }, |
| { |
| "epoch": 2.072, |
| "grad_norm": 0.3725666055720055, |
| "learning_rate": 3.551291202534899e-05, |
| "loss": 0.2543, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2299247682094574, |
| "step": 1295, |
| "valid_targets_mean": 5426.9, |
| "valid_targets_min": 1115 |
| }, |
| { |
| "epoch": 2.08, |
| "grad_norm": 0.6713941175950688, |
| "learning_rate": 3.546242356325922e-05, |
| "loss": 0.2612, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29278409481048584, |
| "step": 1300, |
| "valid_targets_mean": 2013.2, |
| "valid_targets_min": 782 |
| }, |
| { |
| "epoch": 2.088, |
| "grad_norm": 0.557278656484483, |
| "learning_rate": 3.5411688959062323e-05, |
| "loss": 0.2458, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2544538080692291, |
| "step": 1305, |
| "valid_targets_mean": 3579.6, |
| "valid_targets_min": 762 |
| }, |
| { |
| "epoch": 2.096, |
| "grad_norm": 0.3687842744740229, |
| "learning_rate": 3.5360709020388625e-05, |
| "loss": 0.2418, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25263142585754395, |
| "step": 1310, |
| "valid_targets_mean": 6351.1, |
| "valid_targets_min": 905 |
| }, |
| { |
| "epoch": 2.104, |
| "grad_norm": 0.3793263690857251, |
| "learning_rate": 3.530948455877388e-05, |
| "loss": 0.2659, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24786485731601715, |
| "step": 1315, |
| "valid_targets_mean": 5880.1, |
| "valid_targets_min": 910 |
| }, |
| { |
| "epoch": 2.112, |
| "grad_norm": 0.47916929111479617, |
| "learning_rate": 3.525801638964634e-05, |
| "loss": 0.2435, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2757444679737091, |
| "step": 1320, |
| "valid_targets_mean": 3739.0, |
| "valid_targets_min": 685 |
| }, |
| { |
| "epoch": 2.12, |
| "grad_norm": 0.5444709991468958, |
| "learning_rate": 3.520630533231376e-05, |
| "loss": 0.2636, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3086904287338257, |
| "step": 1325, |
| "valid_targets_mean": 3999.2, |
| "valid_targets_min": 974 |
| }, |
| { |
| "epoch": 2.128, |
| "grad_norm": 0.48962862421008274, |
| "learning_rate": 3.5154352209950376e-05, |
| "loss": 0.2585, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27979689836502075, |
| "step": 1330, |
| "valid_targets_mean": 3676.9, |
| "valid_targets_min": 825 |
| }, |
| { |
| "epoch": 2.136, |
| "grad_norm": 0.45083380728163236, |
| "learning_rate": 3.510215784958376e-05, |
| "loss": 0.2491, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2559383511543274, |
| "step": 1335, |
| "valid_targets_mean": 4485.2, |
| "valid_targets_min": 779 |
| }, |
| { |
| "epoch": 2.144, |
| "grad_norm": 0.33484746429649287, |
| "learning_rate": 3.5049723082081755e-05, |
| "loss": 0.244, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2397380918264389, |
| "step": 1340, |
| "valid_targets_mean": 6896.4, |
| "valid_targets_min": 629 |
| }, |
| { |
| "epoch": 2.152, |
| "grad_norm": 0.5512883869808385, |
| "learning_rate": 3.49970487421391e-05, |
| "loss": 0.252, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2993601858615875, |
| "step": 1345, |
| "valid_targets_mean": 3205.4, |
| "valid_targets_min": 988 |
| }, |
| { |
| "epoch": 2.16, |
| "grad_norm": 0.42190102820629805, |
| "learning_rate": 3.494413566826427e-05, |
| "loss": 0.2541, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2174837589263916, |
| "step": 1350, |
| "valid_targets_mean": 4768.2, |
| "valid_targets_min": 972 |
| }, |
| { |
| "epoch": 2.168, |
| "grad_norm": 0.5081581527596065, |
| "learning_rate": 3.489098470276608e-05, |
| "loss": 0.2626, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23775967955589294, |
| "step": 1355, |
| "valid_targets_mean": 2642.8, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 2.176, |
| "grad_norm": 0.40826096131920026, |
| "learning_rate": 3.483759669174024e-05, |
| "loss": 0.2624, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27968931198120117, |
| "step": 1360, |
| "valid_targets_mean": 4708.7, |
| "valid_targets_min": 797 |
| }, |
| { |
| "epoch": 2.184, |
| "grad_norm": 0.4485915411385824, |
| "learning_rate": 3.478397248505598e-05, |
| "loss": 0.2486, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2608921527862549, |
| "step": 1365, |
| "valid_targets_mean": 4368.9, |
| "valid_targets_min": 1052 |
| }, |
| { |
| "epoch": 2.192, |
| "grad_norm": 0.5512019565130977, |
| "learning_rate": 3.473011293634241e-05, |
| "loss": 0.2669, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2932088375091553, |
| "step": 1370, |
| "valid_targets_mean": 3872.3, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 2.2, |
| "grad_norm": 0.42038314886774936, |
| "learning_rate": 3.467601890297502e-05, |
| "loss": 0.2451, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23290091753005981, |
| "step": 1375, |
| "valid_targets_mean": 3820.8, |
| "valid_targets_min": 862 |
| }, |
| { |
| "epoch": 2.208, |
| "grad_norm": 0.486361040619312, |
| "learning_rate": 3.4621691246061976e-05, |
| "loss": 0.2465, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2654815912246704, |
| "step": 1380, |
| "valid_targets_mean": 2967.8, |
| "valid_targets_min": 748 |
| }, |
| { |
| "epoch": 2.216, |
| "grad_norm": 0.3689995882810626, |
| "learning_rate": 3.456713083043046e-05, |
| "loss": 0.2575, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2481921762228012, |
| "step": 1385, |
| "valid_targets_mean": 6700.3, |
| "valid_targets_min": 766 |
| }, |
| { |
| "epoch": 2.224, |
| "grad_norm": 0.3741844661161578, |
| "learning_rate": 3.451233852461285e-05, |
| "loss": 0.253, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2775578498840332, |
| "step": 1390, |
| "valid_targets_mean": 6166.1, |
| "valid_targets_min": 1012 |
| }, |
| { |
| "epoch": 2.232, |
| "grad_norm": 0.4690171509050791, |
| "learning_rate": 3.4457315200832935e-05, |
| "loss": 0.2537, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27017349004745483, |
| "step": 1395, |
| "valid_targets_mean": 3758.4, |
| "valid_targets_min": 916 |
| }, |
| { |
| "epoch": 2.24, |
| "grad_norm": 0.46502707551837297, |
| "learning_rate": 3.440206173499201e-05, |
| "loss": 0.2647, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24808195233345032, |
| "step": 1400, |
| "valid_targets_mean": 3311.0, |
| "valid_targets_min": 919 |
| }, |
| { |
| "epoch": 2.248, |
| "grad_norm": 0.5358873234497143, |
| "learning_rate": 3.4346579006654945e-05, |
| "loss": 0.2398, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28968346118927, |
| "step": 1405, |
| "valid_targets_mean": 3115.9, |
| "valid_targets_min": 556 |
| }, |
| { |
| "epoch": 2.2560000000000002, |
| "grad_norm": 0.42084243548194467, |
| "learning_rate": 3.4290867899036166e-05, |
| "loss": 0.2563, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22804833948612213, |
| "step": 1410, |
| "valid_targets_mean": 3692.8, |
| "valid_targets_min": 753 |
| }, |
| { |
| "epoch": 2.2640000000000002, |
| "grad_norm": 0.4001618784605517, |
| "learning_rate": 3.4234929298985614e-05, |
| "loss": 0.2656, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23770037293434143, |
| "step": 1415, |
| "valid_targets_mean": 5107.0, |
| "valid_targets_min": 1186 |
| }, |
| { |
| "epoch": 2.2720000000000002, |
| "grad_norm": 0.5080158679760888, |
| "learning_rate": 3.417876409697463e-05, |
| "loss": 0.2382, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20453549921512604, |
| "step": 1420, |
| "valid_targets_mean": 2835.4, |
| "valid_targets_min": 690 |
| }, |
| { |
| "epoch": 2.2800000000000002, |
| "grad_norm": 0.4168243110819609, |
| "learning_rate": 3.412237318708175e-05, |
| "loss": 0.2374, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23663721978664398, |
| "step": 1425, |
| "valid_targets_mean": 4147.2, |
| "valid_targets_min": 957 |
| }, |
| { |
| "epoch": 2.288, |
| "grad_norm": 0.45954399327365, |
| "learning_rate": 3.4065757466978504e-05, |
| "loss": 0.2596, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2507503926753998, |
| "step": 1430, |
| "valid_targets_mean": 3781.1, |
| "valid_targets_min": 808 |
| }, |
| { |
| "epoch": 2.296, |
| "grad_norm": 0.4702539653734292, |
| "learning_rate": 3.400891783791511e-05, |
| "loss": 0.2671, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27482253313064575, |
| "step": 1435, |
| "valid_targets_mean": 3705.0, |
| "valid_targets_min": 1032 |
| }, |
| { |
| "epoch": 2.304, |
| "grad_norm": 0.40026868215420613, |
| "learning_rate": 3.395185520470614e-05, |
| "loss": 0.2681, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24075943231582642, |
| "step": 1440, |
| "valid_targets_mean": 5237.6, |
| "valid_targets_min": 760 |
| }, |
| { |
| "epoch": 2.312, |
| "grad_norm": 0.3905513051922005, |
| "learning_rate": 3.38945704757161e-05, |
| "loss": 0.2522, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22074247896671295, |
| "step": 1445, |
| "valid_targets_mean": 4402.3, |
| "valid_targets_min": 756 |
| }, |
| { |
| "epoch": 2.32, |
| "grad_norm": 0.4882416307667647, |
| "learning_rate": 3.383706456284498e-05, |
| "loss": 0.2442, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2848014533519745, |
| "step": 1450, |
| "valid_targets_mean": 3364.2, |
| "valid_targets_min": 901 |
| }, |
| { |
| "epoch": 2.328, |
| "grad_norm": 0.45445708291864306, |
| "learning_rate": 3.377933838151374e-05, |
| "loss": 0.2603, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27066075801849365, |
| "step": 1455, |
| "valid_targets_mean": 3707.1, |
| "valid_targets_min": 1017 |
| }, |
| { |
| "epoch": 2.336, |
| "grad_norm": 0.6608392796917465, |
| "learning_rate": 3.3721392850649714e-05, |
| "loss": 0.2744, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33999860286712646, |
| "step": 1460, |
| "valid_targets_mean": 2394.0, |
| "valid_targets_min": 1063 |
| }, |
| { |
| "epoch": 2.344, |
| "grad_norm": 0.47360622747404885, |
| "learning_rate": 3.3663228892672034e-05, |
| "loss": 0.2536, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2844492197036743, |
| "step": 1465, |
| "valid_targets_mean": 4209.1, |
| "valid_targets_min": 1020 |
| }, |
| { |
| "epoch": 2.352, |
| "grad_norm": 0.37445807515235896, |
| "learning_rate": 3.36048474334769e-05, |
| "loss": 0.2717, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21594926714897156, |
| "step": 1470, |
| "valid_targets_mean": 4527.0, |
| "valid_targets_min": 1016 |
| }, |
| { |
| "epoch": 2.36, |
| "grad_norm": 0.4066613926869203, |
| "learning_rate": 3.3546249402422834e-05, |
| "loss": 0.2637, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2440742552280426, |
| "step": 1475, |
| "valid_targets_mean": 4281.1, |
| "valid_targets_min": 884 |
| }, |
| { |
| "epoch": 2.368, |
| "grad_norm": 0.450872690024147, |
| "learning_rate": 3.3487435732315944e-05, |
| "loss": 0.25, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23967817425727844, |
| "step": 1480, |
| "valid_targets_mean": 3973.4, |
| "valid_targets_min": 815 |
| }, |
| { |
| "epoch": 2.376, |
| "grad_norm": 0.4373336501281426, |
| "learning_rate": 3.342840735939501e-05, |
| "loss": 0.2588, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28060564398765564, |
| "step": 1485, |
| "valid_targets_mean": 3881.0, |
| "valid_targets_min": 1006 |
| }, |
| { |
| "epoch": 2.384, |
| "grad_norm": 0.41616555215578727, |
| "learning_rate": 3.33691652233166e-05, |
| "loss": 0.2397, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23200851678848267, |
| "step": 1490, |
| "valid_targets_mean": 4087.1, |
| "valid_targets_min": 682 |
| }, |
| { |
| "epoch": 2.392, |
| "grad_norm": 0.4741632763884423, |
| "learning_rate": 3.330971026714016e-05, |
| "loss": 0.2632, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2454630434513092, |
| "step": 1495, |
| "valid_targets_mean": 3751.9, |
| "valid_targets_min": 1012 |
| }, |
| { |
| "epoch": 2.4, |
| "grad_norm": 0.44777751707599694, |
| "learning_rate": 3.325004343731292e-05, |
| "loss": 0.267, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2585783004760742, |
| "step": 1500, |
| "valid_targets_mean": 4000.0, |
| "valid_targets_min": 897 |
| }, |
| { |
| "epoch": 2.408, |
| "grad_norm": 0.4838887067897237, |
| "learning_rate": 3.3190165683654885e-05, |
| "loss": 0.2516, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27823567390441895, |
| "step": 1505, |
| "valid_targets_mean": 3297.0, |
| "valid_targets_min": 1109 |
| }, |
| { |
| "epoch": 2.416, |
| "grad_norm": 0.42450262258661076, |
| "learning_rate": 3.31300779593437e-05, |
| "loss": 0.2528, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25310438871383667, |
| "step": 1510, |
| "valid_targets_mean": 4192.0, |
| "valid_targets_min": 1057 |
| }, |
| { |
| "epoch": 2.424, |
| "grad_norm": 0.4114223292664404, |
| "learning_rate": 3.306978122089948e-05, |
| "loss": 0.2499, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23483526706695557, |
| "step": 1515, |
| "valid_targets_mean": 4129.9, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 2.432, |
| "grad_norm": 0.4382457699711022, |
| "learning_rate": 3.300927642816957e-05, |
| "loss": 0.2611, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25855690240859985, |
| "step": 1520, |
| "valid_targets_mean": 3935.2, |
| "valid_targets_min": 765 |
| }, |
| { |
| "epoch": 2.44, |
| "grad_norm": 0.37454574276197367, |
| "learning_rate": 3.294856454431328e-05, |
| "loss": 0.2628, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2558472752571106, |
| "step": 1525, |
| "valid_targets_mean": 4478.7, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 2.448, |
| "grad_norm": 0.6185855420387542, |
| "learning_rate": 3.288764653578653e-05, |
| "loss": 0.2724, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3244587182998657, |
| "step": 1530, |
| "valid_targets_mean": 2447.6, |
| "valid_targets_min": 946 |
| }, |
| { |
| "epoch": 2.456, |
| "grad_norm": 0.41449538190334706, |
| "learning_rate": 3.2826523372326516e-05, |
| "loss": 0.2464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22788169980049133, |
| "step": 1535, |
| "valid_targets_mean": 4480.9, |
| "valid_targets_min": 930 |
| }, |
| { |
| "epoch": 2.464, |
| "grad_norm": 0.3844395691586439, |
| "learning_rate": 3.276519602693621e-05, |
| "loss": 0.2502, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2364940345287323, |
| "step": 1540, |
| "valid_targets_mean": 4955.6, |
| "valid_targets_min": 1039 |
| }, |
| { |
| "epoch": 2.472, |
| "grad_norm": 0.37710049485157016, |
| "learning_rate": 3.270366547586892e-05, |
| "loss": 0.2606, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24367119371891022, |
| "step": 1545, |
| "valid_targets_mean": 5767.8, |
| "valid_targets_min": 1061 |
| }, |
| { |
| "epoch": 2.48, |
| "grad_norm": 0.38894525582252293, |
| "learning_rate": 3.2641932698612715e-05, |
| "loss": 0.2541, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23194950819015503, |
| "step": 1550, |
| "valid_targets_mean": 4335.3, |
| "valid_targets_min": 726 |
| }, |
| { |
| "epoch": 2.488, |
| "grad_norm": 0.3982220790804564, |
| "learning_rate": 3.2579998677874855e-05, |
| "loss": 0.2463, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23282426595687866, |
| "step": 1555, |
| "valid_targets_mean": 4442.9, |
| "valid_targets_min": 890 |
| }, |
| { |
| "epoch": 2.496, |
| "grad_norm": 0.3920266130118298, |
| "learning_rate": 3.251786439956614e-05, |
| "loss": 0.2535, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22500693798065186, |
| "step": 1560, |
| "valid_targets_mean": 3975.3, |
| "valid_targets_min": 638 |
| }, |
| { |
| "epoch": 2.504, |
| "grad_norm": 0.5374661523331783, |
| "learning_rate": 3.2455530852785206e-05, |
| "loss": 0.2727, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3037063479423523, |
| "step": 1565, |
| "valid_targets_mean": 3909.3, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 2.512, |
| "grad_norm": 0.42133581754576677, |
| "learning_rate": 3.239299902980281e-05, |
| "loss": 0.2591, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27413496375083923, |
| "step": 1570, |
| "valid_targets_mean": 4875.9, |
| "valid_targets_min": 1016 |
| }, |
| { |
| "epoch": 2.52, |
| "grad_norm": 0.4540812695017237, |
| "learning_rate": 3.2330269926046e-05, |
| "loss": 0.2616, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3003402054309845, |
| "step": 1575, |
| "valid_targets_mean": 4698.6, |
| "valid_targets_min": 1056 |
| }, |
| { |
| "epoch": 2.528, |
| "grad_norm": 0.48721669960386377, |
| "learning_rate": 3.2267344540082284e-05, |
| "loss": 0.2773, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2769826650619507, |
| "step": 1580, |
| "valid_targets_mean": 3544.6, |
| "valid_targets_min": 959 |
| }, |
| { |
| "epoch": 2.536, |
| "grad_norm": 0.44779354854153475, |
| "learning_rate": 3.220422387360373e-05, |
| "loss": 0.2413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23960761725902557, |
| "step": 1585, |
| "valid_targets_mean": 4160.4, |
| "valid_targets_min": 1206 |
| }, |
| { |
| "epoch": 2.544, |
| "grad_norm": 0.36624305776785404, |
| "learning_rate": 3.2140908931411026e-05, |
| "loss": 0.2491, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24884971976280212, |
| "step": 1590, |
| "valid_targets_mean": 5886.5, |
| "valid_targets_min": 777 |
| }, |
| { |
| "epoch": 2.552, |
| "grad_norm": 0.4205442975727069, |
| "learning_rate": 3.207740072139748e-05, |
| "loss": 0.2641, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2376871556043625, |
| "step": 1595, |
| "valid_targets_mean": 4554.3, |
| "valid_targets_min": 807 |
| }, |
| { |
| "epoch": 2.56, |
| "grad_norm": 0.4390164861743431, |
| "learning_rate": 3.2013700254532996e-05, |
| "loss": 0.2626, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2751956582069397, |
| "step": 1600, |
| "valid_targets_mean": 3900.7, |
| "valid_targets_min": 722 |
| }, |
| { |
| "epoch": 2.568, |
| "grad_norm": 0.45059040166393205, |
| "learning_rate": 3.194980854484794e-05, |
| "loss": 0.2556, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24145694077014923, |
| "step": 1605, |
| "valid_targets_mean": 3108.7, |
| "valid_targets_min": 1069 |
| }, |
| { |
| "epoch": 2.576, |
| "grad_norm": 0.38498385041200234, |
| "learning_rate": 3.188572660941702e-05, |
| "loss": 0.2393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25530481338500977, |
| "step": 1610, |
| "valid_targets_mean": 4882.4, |
| "valid_targets_min": 630 |
| }, |
| { |
| "epoch": 2.584, |
| "grad_norm": 0.45078053611552554, |
| "learning_rate": 3.182145546834311e-05, |
| "loss": 0.2521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23659245669841766, |
| "step": 1615, |
| "valid_targets_mean": 3979.6, |
| "valid_targets_min": 933 |
| }, |
| { |
| "epoch": 2.592, |
| "grad_norm": 0.4873505068621663, |
| "learning_rate": 3.1756996144740994e-05, |
| "loss": 0.2731, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24679982662200928, |
| "step": 1620, |
| "valid_targets_mean": 2844.8, |
| "valid_targets_min": 728 |
| }, |
| { |
| "epoch": 2.6, |
| "grad_norm": 0.375889816303558, |
| "learning_rate": 3.1692349664721074e-05, |
| "loss": 0.2752, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25449809432029724, |
| "step": 1625, |
| "valid_targets_mean": 5291.6, |
| "valid_targets_min": 1192 |
| }, |
| { |
| "epoch": 2.608, |
| "grad_norm": 0.43710999498519076, |
| "learning_rate": 3.1627517057373046e-05, |
| "loss": 0.259, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2557065486907959, |
| "step": 1630, |
| "valid_targets_mean": 4127.9, |
| "valid_targets_min": 906 |
| }, |
| { |
| "epoch": 2.616, |
| "grad_norm": 0.47479745259245676, |
| "learning_rate": 3.156249935474953e-05, |
| "loss": 0.2684, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2876622974872589, |
| "step": 1635, |
| "valid_targets_mean": 4105.4, |
| "valid_targets_min": 739 |
| }, |
| { |
| "epoch": 2.624, |
| "grad_norm": 0.38433518760476604, |
| "learning_rate": 3.1497297591849614e-05, |
| "loss": 0.2629, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25629955530166626, |
| "step": 1640, |
| "valid_targets_mean": 5574.8, |
| "valid_targets_min": 858 |
| }, |
| { |
| "epoch": 2.632, |
| "grad_norm": 0.43299417778940696, |
| "learning_rate": 3.143191280660238e-05, |
| "loss": 0.2445, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23174050450325012, |
| "step": 1645, |
| "valid_targets_mean": 3752.2, |
| "valid_targets_min": 1012 |
| }, |
| { |
| "epoch": 2.64, |
| "grad_norm": 0.4042914030109211, |
| "learning_rate": 3.1366346039850424e-05, |
| "loss": 0.2468, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2770325541496277, |
| "step": 1650, |
| "valid_targets_mean": 4572.9, |
| "valid_targets_min": 955 |
| }, |
| { |
| "epoch": 2.648, |
| "grad_norm": 0.4000943004214781, |
| "learning_rate": 3.130059833533323e-05, |
| "loss": 0.2506, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2539942264556885, |
| "step": 1655, |
| "valid_targets_mean": 4367.9, |
| "valid_targets_min": 851 |
| }, |
| { |
| "epoch": 2.656, |
| "grad_norm": 0.5246478390502145, |
| "learning_rate": 3.123467073967059e-05, |
| "loss": 0.2646, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25003015995025635, |
| "step": 1660, |
| "valid_targets_mean": 2662.1, |
| "valid_targets_min": 935 |
| }, |
| { |
| "epoch": 2.664, |
| "grad_norm": 0.48721529047895606, |
| "learning_rate": 3.116856430234594e-05, |
| "loss": 0.2472, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24807119369506836, |
| "step": 1665, |
| "valid_targets_mean": 2776.6, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 2.672, |
| "grad_norm": 0.4011766685019322, |
| "learning_rate": 3.110228007568963e-05, |
| "loss": 0.2532, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25203073024749756, |
| "step": 1670, |
| "valid_targets_mean": 4795.0, |
| "valid_targets_min": 1091 |
| }, |
| { |
| "epoch": 2.68, |
| "grad_norm": 0.4108384417120924, |
| "learning_rate": 3.103581911486221e-05, |
| "loss": 0.244, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2413668930530548, |
| "step": 1675, |
| "valid_targets_mean": 4554.4, |
| "valid_targets_min": 807 |
| }, |
| { |
| "epoch": 2.6879999999999997, |
| "grad_norm": 0.4318489284689886, |
| "learning_rate": 3.0969182477837604e-05, |
| "loss": 0.2747, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2699677348136902, |
| "step": 1680, |
| "valid_targets_mean": 4299.4, |
| "valid_targets_min": 951 |
| }, |
| { |
| "epoch": 2.6959999999999997, |
| "grad_norm": 0.46195534616572037, |
| "learning_rate": 3.090237122538628e-05, |
| "loss": 0.2711, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2878229022026062, |
| "step": 1685, |
| "valid_targets_mean": 3651.4, |
| "valid_targets_min": 882 |
| }, |
| { |
| "epoch": 2.7039999999999997, |
| "grad_norm": 0.4064409277711358, |
| "learning_rate": 3.0835386421058345e-05, |
| "loss": 0.2432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.252693772315979, |
| "step": 1690, |
| "valid_targets_mean": 4166.9, |
| "valid_targets_min": 1026 |
| }, |
| { |
| "epoch": 2.7119999999999997, |
| "grad_norm": 0.3315855719749269, |
| "learning_rate": 3.0768229131166664e-05, |
| "loss": 0.261, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22979766130447388, |
| "step": 1695, |
| "valid_targets_mean": 6155.2, |
| "valid_targets_min": 971 |
| }, |
| { |
| "epoch": 2.7199999999999998, |
| "grad_norm": 0.46914934540874326, |
| "learning_rate": 3.070090042476983e-05, |
| "loss": 0.2458, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23643366992473602, |
| "step": 1700, |
| "valid_targets_mean": 3444.9, |
| "valid_targets_min": 850 |
| }, |
| { |
| "epoch": 2.7279999999999998, |
| "grad_norm": 0.6284719316531368, |
| "learning_rate": 3.063340137365517e-05, |
| "loss": 0.27, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2869764566421509, |
| "step": 1705, |
| "valid_targets_mean": 2083.1, |
| "valid_targets_min": 837 |
| }, |
| { |
| "epoch": 2.7359999999999998, |
| "grad_norm": 0.38536559734095, |
| "learning_rate": 3.0565733052321674e-05, |
| "loss": 0.2549, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25470173358917236, |
| "step": 1710, |
| "valid_targets_mean": 5245.9, |
| "valid_targets_min": 1287 |
| }, |
| { |
| "epoch": 2.7439999999999998, |
| "grad_norm": 0.4111284973330911, |
| "learning_rate": 3.0497896537962924e-05, |
| "loss": 0.2464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2471093237400055, |
| "step": 1715, |
| "valid_targets_mean": 4554.9, |
| "valid_targets_min": 863 |
| }, |
| { |
| "epoch": 2.752, |
| "grad_norm": 0.340479711041136, |
| "learning_rate": 3.042989291044991e-05, |
| "loss": 0.2501, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.255589097738266, |
| "step": 1720, |
| "valid_targets_mean": 8342.3, |
| "valid_targets_min": 843 |
| }, |
| { |
| "epoch": 2.76, |
| "grad_norm": 0.4997516270386313, |
| "learning_rate": 3.036172325231383e-05, |
| "loss": 0.2614, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26695573329925537, |
| "step": 1725, |
| "valid_targets_mean": 2913.8, |
| "valid_targets_min": 928 |
| }, |
| { |
| "epoch": 2.768, |
| "grad_norm": 0.4160259337644636, |
| "learning_rate": 3.0293388648728908e-05, |
| "loss": 0.2608, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23652175068855286, |
| "step": 1730, |
| "valid_targets_mean": 4476.8, |
| "valid_targets_min": 982 |
| }, |
| { |
| "epoch": 2.776, |
| "grad_norm": 0.39927985153665424, |
| "learning_rate": 3.022489018749508e-05, |
| "loss": 0.246, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24088355898857117, |
| "step": 1735, |
| "valid_targets_mean": 3952.5, |
| "valid_targets_min": 1032 |
| }, |
| { |
| "epoch": 2.784, |
| "grad_norm": 0.4550887207976117, |
| "learning_rate": 3.015622895902068e-05, |
| "loss": 0.2423, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2190483808517456, |
| "step": 1740, |
| "valid_targets_mean": 3110.4, |
| "valid_targets_min": 1039 |
| }, |
| { |
| "epoch": 2.792, |
| "grad_norm": 0.33861046318017596, |
| "learning_rate": 3.008740605630508e-05, |
| "loss": 0.2541, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24156677722930908, |
| "step": 1745, |
| "valid_targets_mean": 6048.4, |
| "valid_targets_min": 748 |
| }, |
| { |
| "epoch": 2.8, |
| "grad_norm": 0.36617432315449255, |
| "learning_rate": 3.0018422574921337e-05, |
| "loss": 0.2392, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24149316549301147, |
| "step": 1750, |
| "valid_targets_mean": 5241.0, |
| "valid_targets_min": 1079 |
| }, |
| { |
| "epoch": 2.808, |
| "grad_norm": 0.3464726951761244, |
| "learning_rate": 2.9949279612998673e-05, |
| "loss": 0.2423, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2470303475856781, |
| "step": 1755, |
| "valid_targets_mean": 6875.5, |
| "valid_targets_min": 1210 |
| }, |
| { |
| "epoch": 2.816, |
| "grad_norm": 0.3450155466545645, |
| "learning_rate": 2.9879978271205064e-05, |
| "loss": 0.2374, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2277037501335144, |
| "step": 1760, |
| "valid_targets_mean": 5934.5, |
| "valid_targets_min": 912 |
| }, |
| { |
| "epoch": 2.824, |
| "grad_norm": 0.4016568882747598, |
| "learning_rate": 2.9810519652729692e-05, |
| "loss": 0.2793, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2241845428943634, |
| "step": 1765, |
| "valid_targets_mean": 4558.4, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 2.832, |
| "grad_norm": 0.41748302286864053, |
| "learning_rate": 2.9740904863265378e-05, |
| "loss": 0.2462, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.263439804315567, |
| "step": 1770, |
| "valid_targets_mean": 4341.7, |
| "valid_targets_min": 1283 |
| }, |
| { |
| "epoch": 2.84, |
| "grad_norm": 0.48850050096135555, |
| "learning_rate": 2.967113501099097e-05, |
| "loss": 0.2458, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23303407430648804, |
| "step": 1775, |
| "valid_targets_mean": 2772.5, |
| "valid_targets_min": 787 |
| }, |
| { |
| "epoch": 2.848, |
| "grad_norm": 0.47900204436253896, |
| "learning_rate": 2.9601211206553745e-05, |
| "loss": 0.28, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32549360394477844, |
| "step": 1780, |
| "valid_targets_mean": 3978.5, |
| "valid_targets_min": 851 |
| }, |
| { |
| "epoch": 2.856, |
| "grad_norm": 0.39202466712624395, |
| "learning_rate": 2.9531134563051686e-05, |
| "loss": 0.2521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24673870205879211, |
| "step": 1785, |
| "valid_targets_mean": 4397.2, |
| "valid_targets_min": 789 |
| }, |
| { |
| "epoch": 2.864, |
| "grad_norm": 0.454753156477103, |
| "learning_rate": 2.946090619601579e-05, |
| "loss": 0.255, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2228916883468628, |
| "step": 1790, |
| "valid_targets_mean": 3641.2, |
| "valid_targets_min": 801 |
| }, |
| { |
| "epoch": 2.872, |
| "grad_norm": 0.5239133569031957, |
| "learning_rate": 2.9390527223392292e-05, |
| "loss": 0.2483, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26477718353271484, |
| "step": 1795, |
| "valid_targets_mean": 2819.3, |
| "valid_targets_min": 764 |
| }, |
| { |
| "epoch": 2.88, |
| "grad_norm": 0.43565748091817924, |
| "learning_rate": 2.931999876552488e-05, |
| "loss": 0.2698, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2930182218551636, |
| "step": 1800, |
| "valid_targets_mean": 4048.8, |
| "valid_targets_min": 1123 |
| }, |
| { |
| "epoch": 2.888, |
| "grad_norm": 0.4126016381922162, |
| "learning_rate": 2.9249321945136854e-05, |
| "loss": 0.2456, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26448121666908264, |
| "step": 1805, |
| "valid_targets_mean": 4604.4, |
| "valid_targets_min": 948 |
| }, |
| { |
| "epoch": 2.896, |
| "grad_norm": 0.4216543598725713, |
| "learning_rate": 2.9178497887313257e-05, |
| "loss": 0.2589, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27048546075820923, |
| "step": 1810, |
| "valid_targets_mean": 4440.7, |
| "valid_targets_min": 926 |
| }, |
| { |
| "epoch": 2.904, |
| "grad_norm": 0.339972934761611, |
| "learning_rate": 2.9107527719482968e-05, |
| "loss": 0.2497, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24039791524410248, |
| "step": 1815, |
| "valid_targets_mean": 5725.7, |
| "valid_targets_min": 1038 |
| }, |
| { |
| "epoch": 2.912, |
| "grad_norm": 0.4284574874009162, |
| "learning_rate": 2.9036412571400747e-05, |
| "loss": 0.2348, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25202876329421997, |
| "step": 1820, |
| "valid_targets_mean": 3956.6, |
| "valid_targets_min": 841 |
| }, |
| { |
| "epoch": 2.92, |
| "grad_norm": 0.3887923193278144, |
| "learning_rate": 2.8965153575129255e-05, |
| "loss": 0.2383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26022517681121826, |
| "step": 1825, |
| "valid_targets_mean": 5180.9, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 2.928, |
| "grad_norm": 0.42207586149145204, |
| "learning_rate": 2.8893751865021044e-05, |
| "loss": 0.2454, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24122312664985657, |
| "step": 1830, |
| "valid_targets_mean": 3972.8, |
| "valid_targets_min": 946 |
| }, |
| { |
| "epoch": 2.936, |
| "grad_norm": 0.38545070788876795, |
| "learning_rate": 2.8822208577700473e-05, |
| "loss": 0.2401, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2604748606681824, |
| "step": 1835, |
| "valid_targets_mean": 4398.5, |
| "valid_targets_min": 782 |
| }, |
| { |
| "epoch": 2.944, |
| "grad_norm": 0.5068190010044846, |
| "learning_rate": 2.8750524852045642e-05, |
| "loss": 0.2548, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26114439964294434, |
| "step": 1840, |
| "valid_targets_mean": 2930.1, |
| "valid_targets_min": 790 |
| }, |
| { |
| "epoch": 2.952, |
| "grad_norm": 0.38331449055182676, |
| "learning_rate": 2.867870182917024e-05, |
| "loss": 0.244, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23407770693302155, |
| "step": 1845, |
| "valid_targets_mean": 4040.6, |
| "valid_targets_min": 758 |
| }, |
| { |
| "epoch": 2.96, |
| "grad_norm": 0.40079080583130816, |
| "learning_rate": 2.8606740652405394e-05, |
| "loss": 0.2779, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2599574327468872, |
| "step": 1850, |
| "valid_targets_mean": 4798.5, |
| "valid_targets_min": 934 |
| }, |
| { |
| "epoch": 2.968, |
| "grad_norm": 0.3479930057967736, |
| "learning_rate": 2.853464246728147e-05, |
| "loss": 0.2447, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23751023411750793, |
| "step": 1855, |
| "valid_targets_mean": 5672.2, |
| "valid_targets_min": 832 |
| }, |
| { |
| "epoch": 2.976, |
| "grad_norm": 0.3369578047927962, |
| "learning_rate": 2.846240842150984e-05, |
| "loss": 0.2385, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20889130234718323, |
| "step": 1860, |
| "valid_targets_mean": 5308.2, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 2.984, |
| "grad_norm": 0.4291039444647585, |
| "learning_rate": 2.839003966496458e-05, |
| "loss": 0.2477, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2270471304655075, |
| "step": 1865, |
| "valid_targets_mean": 3958.4, |
| "valid_targets_min": 897 |
| }, |
| { |
| "epoch": 2.992, |
| "grad_norm": 0.43281841130812354, |
| "learning_rate": 2.8317537349664215e-05, |
| "loss": 0.2807, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2587190866470337, |
| "step": 1870, |
| "valid_targets_mean": 4151.9, |
| "valid_targets_min": 950 |
| }, |
| { |
| "epoch": 3.0, |
| "grad_norm": 0.3978270265787385, |
| "learning_rate": 2.824490262975334e-05, |
| "loss": 0.2441, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2422841191291809, |
| "step": 1875, |
| "valid_targets_mean": 4473.6, |
| "valid_targets_min": 931 |
| }, |
| { |
| "epoch": 3.008, |
| "grad_norm": 0.3686194912968534, |
| "learning_rate": 2.817213666148427e-05, |
| "loss": 0.2345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20717495679855347, |
| "step": 1880, |
| "valid_targets_mean": 4454.7, |
| "valid_targets_min": 1031 |
| }, |
| { |
| "epoch": 3.016, |
| "grad_norm": 0.4391451892820445, |
| "learning_rate": 2.809924060319862e-05, |
| "loss": 0.2248, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2463517189025879, |
| "step": 1885, |
| "valid_targets_mean": 4455.8, |
| "valid_targets_min": 850 |
| }, |
| { |
| "epoch": 3.024, |
| "grad_norm": 0.37945227439492357, |
| "learning_rate": 2.802621561530888e-05, |
| "loss": 0.2377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23665887117385864, |
| "step": 1890, |
| "valid_targets_mean": 4614.1, |
| "valid_targets_min": 801 |
| }, |
| { |
| "epoch": 3.032, |
| "grad_norm": 0.5047817941563999, |
| "learning_rate": 2.7953062860279937e-05, |
| "loss": 0.2438, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24539366364479065, |
| "step": 1895, |
| "valid_targets_mean": 2838.2, |
| "valid_targets_min": 945 |
| }, |
| { |
| "epoch": 3.04, |
| "grad_norm": 0.640395323678212, |
| "learning_rate": 2.7879783502610557e-05, |
| "loss": 0.2587, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33286240696907043, |
| "step": 1900, |
| "valid_targets_mean": 2273.1, |
| "valid_targets_min": 937 |
| }, |
| { |
| "epoch": 3.048, |
| "grad_norm": 0.42850604085756194, |
| "learning_rate": 2.7806378708814875e-05, |
| "loss": 0.2476, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24500463902950287, |
| "step": 1905, |
| "valid_targets_mean": 4063.4, |
| "valid_targets_min": 1026 |
| }, |
| { |
| "epoch": 3.056, |
| "grad_norm": 0.4303493596218602, |
| "learning_rate": 2.773284964740379e-05, |
| "loss": 0.2396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23514248430728912, |
| "step": 1910, |
| "valid_targets_mean": 3806.9, |
| "valid_targets_min": 1057 |
| }, |
| { |
| "epoch": 3.064, |
| "grad_norm": 0.5376670517442778, |
| "learning_rate": 2.7659197488866403e-05, |
| "loss": 0.2436, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24571284651756287, |
| "step": 1915, |
| "valid_targets_mean": 2978.4, |
| "valid_targets_min": 835 |
| }, |
| { |
| "epoch": 3.072, |
| "grad_norm": 0.4856549540142223, |
| "learning_rate": 2.7585423405651347e-05, |
| "loss": 0.2459, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2590499818325043, |
| "step": 1920, |
| "valid_targets_mean": 3158.0, |
| "valid_targets_min": 817 |
| }, |
| { |
| "epoch": 3.08, |
| "grad_norm": 0.41265678997745975, |
| "learning_rate": 2.7511528572148153e-05, |
| "loss": 0.2285, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24581827223300934, |
| "step": 1925, |
| "valid_targets_mean": 3845.9, |
| "valid_targets_min": 779 |
| }, |
| { |
| "epoch": 3.088, |
| "grad_norm": 0.39148774769841904, |
| "learning_rate": 2.7437514164668536e-05, |
| "loss": 0.2211, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2122671902179718, |
| "step": 1930, |
| "valid_targets_mean": 4021.6, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 3.096, |
| "grad_norm": 0.42384683209642465, |
| "learning_rate": 2.7363381361427692e-05, |
| "loss": 0.2509, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2753984332084656, |
| "step": 1935, |
| "valid_targets_mean": 4518.8, |
| "valid_targets_min": 996 |
| }, |
| { |
| "epoch": 3.104, |
| "grad_norm": 0.5156696168531341, |
| "learning_rate": 2.72891313425255e-05, |
| "loss": 0.2448, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2671349346637726, |
| "step": 1940, |
| "valid_targets_mean": 3111.5, |
| "valid_targets_min": 856 |
| }, |
| { |
| "epoch": 3.112, |
| "grad_norm": 0.40374314761107644, |
| "learning_rate": 2.7214765289927777e-05, |
| "loss": 0.2362, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22792913019657135, |
| "step": 1945, |
| "valid_targets_mean": 4236.8, |
| "valid_targets_min": 1042 |
| }, |
| { |
| "epoch": 3.12, |
| "grad_norm": 0.41400421213473976, |
| "learning_rate": 2.714028438744746e-05, |
| "loss": 0.2624, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28113529086112976, |
| "step": 1950, |
| "valid_targets_mean": 4941.1, |
| "valid_targets_min": 945 |
| }, |
| { |
| "epoch": 3.128, |
| "grad_norm": 0.3753847058592898, |
| "learning_rate": 2.706568982072573e-05, |
| "loss": 0.2403, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22099116444587708, |
| "step": 1955, |
| "valid_targets_mean": 4709.8, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 3.136, |
| "grad_norm": 0.4414960074804987, |
| "learning_rate": 2.6990982777213174e-05, |
| "loss": 0.2336, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24348290264606476, |
| "step": 1960, |
| "valid_targets_mean": 4199.9, |
| "valid_targets_min": 833 |
| }, |
| { |
| "epoch": 3.144, |
| "grad_norm": 0.39031860273125923, |
| "learning_rate": 2.691616444615085e-05, |
| "loss": 0.2565, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2442023903131485, |
| "step": 1965, |
| "valid_targets_mean": 4948.6, |
| "valid_targets_min": 662 |
| }, |
| { |
| "epoch": 3.152, |
| "grad_norm": 0.42869791654686545, |
| "learning_rate": 2.6841236018551402e-05, |
| "loss": 0.2524, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24929338693618774, |
| "step": 1970, |
| "valid_targets_mean": 4449.2, |
| "valid_targets_min": 1184 |
| }, |
| { |
| "epoch": 3.16, |
| "grad_norm": 0.3724581951765585, |
| "learning_rate": 2.6766198687180028e-05, |
| "loss": 0.2418, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2496282011270523, |
| "step": 1975, |
| "valid_targets_mean": 5392.3, |
| "valid_targets_min": 909 |
| }, |
| { |
| "epoch": 3.168, |
| "grad_norm": 0.3179361729676042, |
| "learning_rate": 2.6691053646535564e-05, |
| "loss": 0.2281, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20636862516403198, |
| "step": 1980, |
| "valid_targets_mean": 6938.2, |
| "valid_targets_min": 787 |
| }, |
| { |
| "epoch": 3.176, |
| "grad_norm": 0.46664739794355814, |
| "learning_rate": 2.6615802092831446e-05, |
| "loss": 0.2712, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22112616896629333, |
| "step": 1985, |
| "valid_targets_mean": 2945.5, |
| "valid_targets_min": 884 |
| }, |
| { |
| "epoch": 3.184, |
| "grad_norm": 0.4339184924410943, |
| "learning_rate": 2.6540445223976637e-05, |
| "loss": 0.2509, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2532888352870941, |
| "step": 1990, |
| "valid_targets_mean": 4107.5, |
| "valid_targets_min": 1002 |
| }, |
| { |
| "epoch": 3.192, |
| "grad_norm": 0.3729189208408062, |
| "learning_rate": 2.6464984239556602e-05, |
| "loss": 0.2458, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24716854095458984, |
| "step": 1995, |
| "valid_targets_mean": 5853.9, |
| "valid_targets_min": 932 |
| }, |
| { |
| "epoch": 3.2, |
| "grad_norm": 0.3663312807394943, |
| "learning_rate": 2.63894203408142e-05, |
| "loss": 0.2449, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2260250449180603, |
| "step": 2000, |
| "valid_targets_mean": 5386.6, |
| "valid_targets_min": 760 |
| }, |
| { |
| "epoch": 3.208, |
| "grad_norm": 0.43597269696745533, |
| "learning_rate": 2.6313754730630528e-05, |
| "loss": 0.2339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23965974152088165, |
| "step": 2005, |
| "valid_targets_mean": 3921.7, |
| "valid_targets_min": 1011 |
| }, |
| { |
| "epoch": 3.216, |
| "grad_norm": 0.38031686943560017, |
| "learning_rate": 2.623798861350582e-05, |
| "loss": 0.2377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2457212209701538, |
| "step": 2010, |
| "valid_targets_mean": 5434.2, |
| "valid_targets_min": 793 |
| }, |
| { |
| "epoch": 3.224, |
| "grad_norm": 0.37852046127420746, |
| "learning_rate": 2.6162123195540247e-05, |
| "loss": 0.2391, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24134644865989685, |
| "step": 2015, |
| "valid_targets_mean": 4507.3, |
| "valid_targets_min": 965 |
| }, |
| { |
| "epoch": 3.232, |
| "grad_norm": 0.3774539132405667, |
| "learning_rate": 2.6086159684414726e-05, |
| "loss": 0.2626, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20981399714946747, |
| "step": 2020, |
| "valid_targets_mean": 5689.2, |
| "valid_targets_min": 983 |
| }, |
| { |
| "epoch": 3.24, |
| "grad_norm": 0.5876396358898935, |
| "learning_rate": 2.6010099289371694e-05, |
| "loss": 0.2517, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24734702706336975, |
| "step": 2025, |
| "valid_targets_mean": 3616.8, |
| "valid_targets_min": 962 |
| }, |
| { |
| "epoch": 3.248, |
| "grad_norm": 0.41025472682008307, |
| "learning_rate": 2.5933943221195844e-05, |
| "loss": 0.2369, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2263985127210617, |
| "step": 2030, |
| "valid_targets_mean": 4079.0, |
| "valid_targets_min": 976 |
| }, |
| { |
| "epoch": 3.2560000000000002, |
| "grad_norm": 0.3585820544990285, |
| "learning_rate": 2.5857692692194884e-05, |
| "loss": 0.2358, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24530109763145447, |
| "step": 2035, |
| "valid_targets_mean": 5752.3, |
| "valid_targets_min": 902 |
| }, |
| { |
| "epoch": 3.2640000000000002, |
| "grad_norm": 0.4016499873533317, |
| "learning_rate": 2.5781348916180195e-05, |
| "loss": 0.2618, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26491880416870117, |
| "step": 2040, |
| "valid_targets_mean": 4694.8, |
| "valid_targets_min": 918 |
| }, |
| { |
| "epoch": 3.2720000000000002, |
| "grad_norm": 0.40307560980296747, |
| "learning_rate": 2.570491310844755e-05, |
| "loss": 0.2409, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2063388228416443, |
| "step": 2045, |
| "valid_targets_mean": 3911.2, |
| "valid_targets_min": 688 |
| }, |
| { |
| "epoch": 3.2800000000000002, |
| "grad_norm": 0.44852837331808115, |
| "learning_rate": 2.562838648575774e-05, |
| "loss": 0.2376, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21985195577144623, |
| "step": 2050, |
| "valid_targets_mean": 3346.2, |
| "valid_targets_min": 963 |
| }, |
| { |
| "epoch": 3.288, |
| "grad_norm": 0.4285370827323687, |
| "learning_rate": 2.5551770266317224e-05, |
| "loss": 0.2427, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24515582621097565, |
| "step": 2055, |
| "valid_targets_mean": 4142.2, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 3.296, |
| "grad_norm": 0.3469454440141706, |
| "learning_rate": 2.5475065669758713e-05, |
| "loss": 0.2292, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2313218116760254, |
| "step": 2060, |
| "valid_targets_mean": 6451.2, |
| "valid_targets_min": 1134 |
| }, |
| { |
| "epoch": 3.304, |
| "grad_norm": 0.4660419550497371, |
| "learning_rate": 2.5398273917121786e-05, |
| "loss": 0.2366, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23948527872562408, |
| "step": 2065, |
| "valid_targets_mean": 3834.3, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 3.312, |
| "grad_norm": 0.46020726031423426, |
| "learning_rate": 2.532139623083342e-05, |
| "loss": 0.2305, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22267141938209534, |
| "step": 2070, |
| "valid_targets_mean": 3291.5, |
| "valid_targets_min": 709 |
| }, |
| { |
| "epoch": 3.32, |
| "grad_norm": 0.4066475864337174, |
| "learning_rate": 2.5244433834688552e-05, |
| "loss": 0.2259, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24701744318008423, |
| "step": 2075, |
| "valid_targets_mean": 4166.8, |
| "valid_targets_min": 1174 |
| }, |
| { |
| "epoch": 3.328, |
| "grad_norm": 0.3389260313456441, |
| "learning_rate": 2.5167387953830602e-05, |
| "loss": 0.2393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2435072362422943, |
| "step": 2080, |
| "valid_targets_mean": 6527.8, |
| "valid_targets_min": 881 |
| }, |
| { |
| "epoch": 3.336, |
| "grad_norm": 0.44452524791300885, |
| "learning_rate": 2.5090259814731946e-05, |
| "loss": 0.2564, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23047269880771637, |
| "step": 2085, |
| "valid_targets_mean": 4016.4, |
| "valid_targets_min": 882 |
| }, |
| { |
| "epoch": 3.344, |
| "grad_norm": 0.5236654293524927, |
| "learning_rate": 2.5013050645174414e-05, |
| "loss": 0.2606, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25171908736228943, |
| "step": 2090, |
| "valid_targets_mean": 2928.3, |
| "valid_targets_min": 945 |
| }, |
| { |
| "epoch": 3.352, |
| "grad_norm": 0.46686168548498463, |
| "learning_rate": 2.4935761674229735e-05, |
| "loss": 0.2738, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25355401635169983, |
| "step": 2095, |
| "valid_targets_mean": 3444.8, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 3.36, |
| "grad_norm": 0.4658436781904886, |
| "learning_rate": 2.4858394132239982e-05, |
| "loss": 0.2506, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24720919132232666, |
| "step": 2100, |
| "valid_targets_mean": 3369.1, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 3.368, |
| "grad_norm": 0.41610717299525096, |
| "learning_rate": 2.4780949250797964e-05, |
| "loss": 0.2536, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25919485092163086, |
| "step": 2105, |
| "valid_targets_mean": 4384.9, |
| "valid_targets_min": 895 |
| }, |
| { |
| "epoch": 3.376, |
| "grad_norm": 0.3776558528507944, |
| "learning_rate": 2.4703428262727656e-05, |
| "loss": 0.2243, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2172096073627472, |
| "step": 2110, |
| "valid_targets_mean": 4115.6, |
| "valid_targets_min": 758 |
| }, |
| { |
| "epoch": 3.384, |
| "grad_norm": 0.4332689527083962, |
| "learning_rate": 2.4625832402064525e-05, |
| "loss": 0.2438, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24866342544555664, |
| "step": 2115, |
| "valid_targets_mean": 4320.1, |
| "valid_targets_min": 908 |
| }, |
| { |
| "epoch": 3.392, |
| "grad_norm": 0.34298883302266486, |
| "learning_rate": 2.454816290403595e-05, |
| "loss": 0.2395, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2340373992919922, |
| "step": 2120, |
| "valid_targets_mean": 5510.9, |
| "valid_targets_min": 914 |
| }, |
| { |
| "epoch": 3.4, |
| "grad_norm": 0.39746900839974997, |
| "learning_rate": 2.4470421005041492e-05, |
| "loss": 0.2437, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27413320541381836, |
| "step": 2125, |
| "valid_targets_mean": 4856.7, |
| "valid_targets_min": 1080 |
| }, |
| { |
| "epoch": 3.408, |
| "grad_norm": 0.3693795848606551, |
| "learning_rate": 2.4392607942633263e-05, |
| "loss": 0.2315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20374570786952972, |
| "step": 2130, |
| "valid_targets_mean": 4384.4, |
| "valid_targets_min": 682 |
| }, |
| { |
| "epoch": 3.416, |
| "grad_norm": 0.6248740105863385, |
| "learning_rate": 2.43147249554962e-05, |
| "loss": 0.2438, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27008944749832153, |
| "step": 2135, |
| "valid_targets_mean": 2089.4, |
| "valid_targets_min": 819 |
| }, |
| { |
| "epoch": 3.424, |
| "grad_norm": 0.4780329930048178, |
| "learning_rate": 2.423677328342835e-05, |
| "loss": 0.2344, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22337397933006287, |
| "step": 2140, |
| "valid_targets_mean": 3094.6, |
| "valid_targets_min": 780 |
| }, |
| { |
| "epoch": 3.432, |
| "grad_norm": 0.3654477042526744, |
| "learning_rate": 2.415875416732113e-05, |
| "loss": 0.2525, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2606598734855652, |
| "step": 2145, |
| "valid_targets_mean": 5555.8, |
| "valid_targets_min": 774 |
| }, |
| { |
| "epoch": 3.44, |
| "grad_norm": 0.36116589317728864, |
| "learning_rate": 2.4080668849139603e-05, |
| "loss": 0.2493, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23566177487373352, |
| "step": 2150, |
| "valid_targets_mean": 5559.3, |
| "valid_targets_min": 682 |
| }, |
| { |
| "epoch": 3.448, |
| "grad_norm": 0.4284496251086486, |
| "learning_rate": 2.4002518571902665e-05, |
| "loss": 0.2391, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27861011028289795, |
| "step": 2155, |
| "valid_targets_mean": 5020.9, |
| "valid_targets_min": 1002 |
| }, |
| { |
| "epoch": 3.456, |
| "grad_norm": 0.3785079045493894, |
| "learning_rate": 2.392430457966328e-05, |
| "loss": 0.2495, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21359242498874664, |
| "step": 2160, |
| "valid_targets_mean": 4796.9, |
| "valid_targets_min": 778 |
| }, |
| { |
| "epoch": 3.464, |
| "grad_norm": 0.3289768949204519, |
| "learning_rate": 2.3846028117488686e-05, |
| "loss": 0.2373, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23382601141929626, |
| "step": 2165, |
| "valid_targets_mean": 6570.4, |
| "valid_targets_min": 1014 |
| }, |
| { |
| "epoch": 3.472, |
| "grad_norm": 0.3147676115980224, |
| "learning_rate": 2.3767690431440533e-05, |
| "loss": 0.2402, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2314678281545639, |
| "step": 2170, |
| "valid_targets_mean": 7406.5, |
| "valid_targets_min": 1083 |
| }, |
| { |
| "epoch": 3.48, |
| "grad_norm": 0.3671848999318846, |
| "learning_rate": 2.368929276855512e-05, |
| "loss": 0.2266, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24486009776592255, |
| "step": 2175, |
| "valid_targets_mean": 4844.6, |
| "valid_targets_min": 952 |
| }, |
| { |
| "epoch": 3.488, |
| "grad_norm": 0.39604902765788624, |
| "learning_rate": 2.361083637682347e-05, |
| "loss": 0.2222, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23706820607185364, |
| "step": 2180, |
| "valid_targets_mean": 4694.9, |
| "valid_targets_min": 744 |
| }, |
| { |
| "epoch": 3.496, |
| "grad_norm": 0.5655798668270807, |
| "learning_rate": 2.3532322505171502e-05, |
| "loss": 0.2616, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27846089005470276, |
| "step": 2185, |
| "valid_targets_mean": 2487.4, |
| "valid_targets_min": 670 |
| }, |
| { |
| "epoch": 3.504, |
| "grad_norm": 0.37720888972750033, |
| "learning_rate": 2.3453752403440147e-05, |
| "loss": 0.237, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2432319074869156, |
| "step": 2190, |
| "valid_targets_mean": 4674.2, |
| "valid_targets_min": 1020 |
| }, |
| { |
| "epoch": 3.512, |
| "grad_norm": 0.38338363036309026, |
| "learning_rate": 2.337512732236545e-05, |
| "loss": 0.2537, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2563523054122925, |
| "step": 2195, |
| "valid_targets_mean": 4757.8, |
| "valid_targets_min": 859 |
| }, |
| { |
| "epoch": 3.52, |
| "grad_norm": 0.3383403543256627, |
| "learning_rate": 2.3296448513558628e-05, |
| "loss": 0.2329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.230048269033432, |
| "step": 2200, |
| "valid_targets_mean": 6312.8, |
| "valid_targets_min": 906 |
| }, |
| { |
| "epoch": 3.528, |
| "grad_norm": 0.42174030998567286, |
| "learning_rate": 2.321771722948622e-05, |
| "loss": 0.2442, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23159737884998322, |
| "step": 2205, |
| "valid_targets_mean": 3522.6, |
| "valid_targets_min": 777 |
| }, |
| { |
| "epoch": 3.536, |
| "grad_norm": 0.3441421244493924, |
| "learning_rate": 2.3138934723450074e-05, |
| "loss": 0.243, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23105758428573608, |
| "step": 2210, |
| "valid_targets_mean": 5247.6, |
| "valid_targets_min": 639 |
| }, |
| { |
| "epoch": 3.544, |
| "grad_norm": 0.4237690704094179, |
| "learning_rate": 2.306010224956744e-05, |
| "loss": 0.2323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20057953894138336, |
| "step": 2215, |
| "valid_targets_mean": 3344.2, |
| "valid_targets_min": 777 |
| }, |
| { |
| "epoch": 3.552, |
| "grad_norm": 0.40314828900451494, |
| "learning_rate": 2.2981221062750986e-05, |
| "loss": 0.2432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23921722173690796, |
| "step": 2220, |
| "valid_targets_mean": 4463.3, |
| "valid_targets_min": 897 |
| }, |
| { |
| "epoch": 3.56, |
| "grad_norm": 0.35913421571712756, |
| "learning_rate": 2.290229241868882e-05, |
| "loss": 0.2336, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22847603261470795, |
| "step": 2225, |
| "valid_targets_mean": 4817.8, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 3.568, |
| "grad_norm": 0.4300249539977357, |
| "learning_rate": 2.282331757382454e-05, |
| "loss": 0.2462, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2206239104270935, |
| "step": 2230, |
| "valid_targets_mean": 3468.5, |
| "valid_targets_min": 913 |
| }, |
| { |
| "epoch": 3.576, |
| "grad_norm": 0.44969351210373343, |
| "learning_rate": 2.2744297785337155e-05, |
| "loss": 0.2287, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21605494618415833, |
| "step": 2235, |
| "valid_targets_mean": 3286.3, |
| "valid_targets_min": 842 |
| }, |
| { |
| "epoch": 3.584, |
| "grad_norm": 0.42301687473892546, |
| "learning_rate": 2.2665234311121155e-05, |
| "loss": 0.264, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23820188641548157, |
| "step": 2240, |
| "valid_targets_mean": 3650.9, |
| "valid_targets_min": 587 |
| }, |
| { |
| "epoch": 3.592, |
| "grad_norm": 0.392543464599928, |
| "learning_rate": 2.258612840976645e-05, |
| "loss": 0.2346, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24136680364608765, |
| "step": 2245, |
| "valid_targets_mean": 4852.1, |
| "valid_targets_min": 1183 |
| }, |
| { |
| "epoch": 3.6, |
| "grad_norm": 0.367395547542587, |
| "learning_rate": 2.2506981340538315e-05, |
| "loss": 0.2661, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2347022294998169, |
| "step": 2250, |
| "valid_targets_mean": 5126.2, |
| "valid_targets_min": 765 |
| }, |
| { |
| "epoch": 3.608, |
| "grad_norm": 0.3205045384740095, |
| "learning_rate": 2.2427794363357384e-05, |
| "loss": 0.2351, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2343018651008606, |
| "step": 2255, |
| "valid_targets_mean": 6760.8, |
| "valid_targets_min": 957 |
| }, |
| { |
| "epoch": 3.616, |
| "grad_norm": 0.4272334380835925, |
| "learning_rate": 2.2348568738779566e-05, |
| "loss": 0.2501, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23038670420646667, |
| "step": 2260, |
| "valid_targets_mean": 4056.7, |
| "valid_targets_min": 877 |
| }, |
| { |
| "epoch": 3.624, |
| "grad_norm": 0.40730070006221614, |
| "learning_rate": 2.2269305727975993e-05, |
| "loss": 0.2498, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26035311818122864, |
| "step": 2265, |
| "valid_targets_mean": 4897.9, |
| "valid_targets_min": 770 |
| }, |
| { |
| "epoch": 3.632, |
| "grad_norm": 0.39008805670033025, |
| "learning_rate": 2.2190006592712927e-05, |
| "loss": 0.2422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2496114820241928, |
| "step": 2270, |
| "valid_targets_mean": 4502.3, |
| "valid_targets_min": 956 |
| }, |
| { |
| "epoch": 3.64, |
| "grad_norm": 0.4098580625905869, |
| "learning_rate": 2.2110672595331698e-05, |
| "loss": 0.2327, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2622841000556946, |
| "step": 2275, |
| "valid_targets_mean": 4436.5, |
| "valid_targets_min": 1137 |
| }, |
| { |
| "epoch": 3.648, |
| "grad_norm": 0.42996882915875295, |
| "learning_rate": 2.2031304998728587e-05, |
| "loss": 0.2452, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2525962293148041, |
| "step": 2280, |
| "valid_targets_mean": 4247.2, |
| "valid_targets_min": 711 |
| }, |
| { |
| "epoch": 3.656, |
| "grad_norm": 0.3824484509543543, |
| "learning_rate": 2.1951905066334737e-05, |
| "loss": 0.2218, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23063942790031433, |
| "step": 2285, |
| "valid_targets_mean": 4258.1, |
| "valid_targets_min": 959 |
| }, |
| { |
| "epoch": 3.664, |
| "grad_norm": 0.40350526206019344, |
| "learning_rate": 2.1872474062096046e-05, |
| "loss": 0.2241, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23316706717014313, |
| "step": 2290, |
| "valid_targets_mean": 4255.7, |
| "valid_targets_min": 698 |
| }, |
| { |
| "epoch": 3.672, |
| "grad_norm": 0.44596553563121094, |
| "learning_rate": 2.179301325045301e-05, |
| "loss": 0.2716, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26697400212287903, |
| "step": 2295, |
| "valid_targets_mean": 3889.1, |
| "valid_targets_min": 764 |
| }, |
| { |
| "epoch": 3.68, |
| "grad_norm": 0.3479970961360567, |
| "learning_rate": 2.1713523896320647e-05, |
| "loss": 0.2378, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21883633732795715, |
| "step": 2300, |
| "valid_targets_mean": 4919.6, |
| "valid_targets_min": 1008 |
| }, |
| { |
| "epoch": 3.6879999999999997, |
| "grad_norm": 0.35571201991914386, |
| "learning_rate": 2.163400726506832e-05, |
| "loss": 0.2542, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23191869258880615, |
| "step": 2305, |
| "valid_targets_mean": 5307.3, |
| "valid_targets_min": 1096 |
| }, |
| { |
| "epoch": 3.6959999999999997, |
| "grad_norm": 0.3604714079928107, |
| "learning_rate": 2.155446462249961e-05, |
| "loss": 0.247, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2559710144996643, |
| "step": 2310, |
| "valid_targets_mean": 5540.6, |
| "valid_targets_min": 728 |
| }, |
| { |
| "epoch": 3.7039999999999997, |
| "grad_norm": 0.32612870359060286, |
| "learning_rate": 2.147489723483217e-05, |
| "loss": 0.2389, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2424880862236023, |
| "step": 2315, |
| "valid_targets_mean": 6959.5, |
| "valid_targets_min": 1094 |
| }, |
| { |
| "epoch": 3.7119999999999997, |
| "grad_norm": 0.41493912008225486, |
| "learning_rate": 2.139530636867757e-05, |
| "loss": 0.2396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24148619174957275, |
| "step": 2320, |
| "valid_targets_mean": 4001.1, |
| "valid_targets_min": 1041 |
| }, |
| { |
| "epoch": 3.7199999999999998, |
| "grad_norm": 0.4110902151701686, |
| "learning_rate": 2.1315693291021114e-05, |
| "loss": 0.2473, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22682532668113708, |
| "step": 2325, |
| "valid_targets_mean": 3857.5, |
| "valid_targets_min": 532 |
| }, |
| { |
| "epoch": 3.7279999999999998, |
| "grad_norm": 0.3771443937450319, |
| "learning_rate": 2.1236059269201686e-05, |
| "loss": 0.2264, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2078881859779358, |
| "step": 2330, |
| "valid_targets_mean": 3960.9, |
| "valid_targets_min": 1015 |
| }, |
| { |
| "epoch": 3.7359999999999998, |
| "grad_norm": 0.4361587523350003, |
| "learning_rate": 2.1156405570891584e-05, |
| "loss": 0.2493, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.240316241979599, |
| "step": 2335, |
| "valid_targets_mean": 3572.1, |
| "valid_targets_min": 996 |
| }, |
| { |
| "epoch": 3.7439999999999998, |
| "grad_norm": 0.4089677660324536, |
| "learning_rate": 2.1076733464076322e-05, |
| "loss": 0.2405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22675375640392303, |
| "step": 2340, |
| "valid_targets_mean": 3776.9, |
| "valid_targets_min": 874 |
| }, |
| { |
| "epoch": 3.752, |
| "grad_norm": 0.42692556018849953, |
| "learning_rate": 2.0997044217034462e-05, |
| "loss": 0.2487, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2749752402305603, |
| "step": 2345, |
| "valid_targets_mean": 4011.2, |
| "valid_targets_min": 1087 |
| }, |
| { |
| "epoch": 3.76, |
| "grad_norm": 0.3607657231192886, |
| "learning_rate": 2.0917339098317405e-05, |
| "loss": 0.2291, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23275761306285858, |
| "step": 2350, |
| "valid_targets_mean": 5583.9, |
| "valid_targets_min": 887 |
| }, |
| { |
| "epoch": 3.768, |
| "grad_norm": 0.53112047115276, |
| "learning_rate": 2.083761937672922e-05, |
| "loss": 0.2346, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23606562614440918, |
| "step": 2355, |
| "valid_targets_mean": 2382.6, |
| "valid_targets_min": 941 |
| }, |
| { |
| "epoch": 3.776, |
| "grad_norm": 0.44795554427006484, |
| "learning_rate": 2.0757886321306433e-05, |
| "loss": 0.2545, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30794838070869446, |
| "step": 2360, |
| "valid_targets_mean": 4058.8, |
| "valid_targets_min": 709 |
| }, |
| { |
| "epoch": 3.784, |
| "grad_norm": 0.38096232285290094, |
| "learning_rate": 2.0678141201297827e-05, |
| "loss": 0.2339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2388550341129303, |
| "step": 2365, |
| "valid_targets_mean": 4521.2, |
| "valid_targets_min": 976 |
| }, |
| { |
| "epoch": 3.792, |
| "grad_norm": 0.3677287419628772, |
| "learning_rate": 2.059838528614423e-05, |
| "loss": 0.2483, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2527107894420624, |
| "step": 2370, |
| "valid_targets_mean": 4927.9, |
| "valid_targets_min": 717 |
| }, |
| { |
| "epoch": 3.8, |
| "grad_norm": 0.4603649045230554, |
| "learning_rate": 2.0518619845458322e-05, |
| "loss": 0.2357, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2524024546146393, |
| "step": 2375, |
| "valid_targets_mean": 3487.9, |
| "valid_targets_min": 900 |
| }, |
| { |
| "epoch": 3.808, |
| "grad_norm": 0.49340867750085365, |
| "learning_rate": 2.0438846149004426e-05, |
| "loss": 0.2266, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2245899885892868, |
| "step": 2380, |
| "valid_targets_mean": 3973.9, |
| "valid_targets_min": 1047 |
| }, |
| { |
| "epoch": 3.816, |
| "grad_norm": 0.3391628456601482, |
| "learning_rate": 2.0359065466678268e-05, |
| "loss": 0.234, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22324374318122864, |
| "step": 2385, |
| "valid_targets_mean": 5603.8, |
| "valid_targets_min": 1077 |
| }, |
| { |
| "epoch": 3.824, |
| "grad_norm": 0.4064978135894322, |
| "learning_rate": 2.0279279068486795e-05, |
| "loss": 0.2499, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24098357558250427, |
| "step": 2390, |
| "valid_targets_mean": 4030.3, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 3.832, |
| "grad_norm": 0.45574421310954694, |
| "learning_rate": 2.019948822452794e-05, |
| "loss": 0.2584, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2887115776538849, |
| "step": 2395, |
| "valid_targets_mean": 3871.5, |
| "valid_targets_min": 593 |
| }, |
| { |
| "epoch": 3.84, |
| "grad_norm": 0.42912425327908477, |
| "learning_rate": 2.0119694204970393e-05, |
| "loss": 0.2347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2643599510192871, |
| "step": 2400, |
| "valid_targets_mean": 3953.9, |
| "valid_targets_min": 752 |
| }, |
| { |
| "epoch": 3.848, |
| "grad_norm": 0.5452664048943783, |
| "learning_rate": 2.0039898280033414e-05, |
| "loss": 0.2447, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25383543968200684, |
| "step": 2405, |
| "valid_targets_mean": 2670.8, |
| "valid_targets_min": 879 |
| }, |
| { |
| "epoch": 3.856, |
| "grad_norm": 0.35547888248847065, |
| "learning_rate": 1.9960101719966592e-05, |
| "loss": 0.2536, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25882688164711, |
| "step": 2410, |
| "valid_targets_mean": 6267.4, |
| "valid_targets_min": 1061 |
| }, |
| { |
| "epoch": 3.864, |
| "grad_norm": 0.4236770102857227, |
| "learning_rate": 1.9880305795029617e-05, |
| "loss": 0.2552, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23290708661079407, |
| "step": 2415, |
| "valid_targets_mean": 3531.1, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 3.872, |
| "grad_norm": 0.5053989332284283, |
| "learning_rate": 1.980051177547207e-05, |
| "loss": 0.2442, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2855120897293091, |
| "step": 2420, |
| "valid_targets_mean": 3139.5, |
| "valid_targets_min": 883 |
| }, |
| { |
| "epoch": 3.88, |
| "grad_norm": 0.40363417243530514, |
| "learning_rate": 1.9720720931513212e-05, |
| "loss": 0.2504, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2508298456668854, |
| "step": 2425, |
| "valid_targets_mean": 4059.1, |
| "valid_targets_min": 614 |
| }, |
| { |
| "epoch": 3.888, |
| "grad_norm": 0.43924888297351233, |
| "learning_rate": 1.9640934533321735e-05, |
| "loss": 0.2542, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2625318765640259, |
| "step": 2430, |
| "valid_targets_mean": 3693.1, |
| "valid_targets_min": 948 |
| }, |
| { |
| "epoch": 3.896, |
| "grad_norm": 0.3419350657709466, |
| "learning_rate": 1.9561153850995577e-05, |
| "loss": 0.2407, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21997395157814026, |
| "step": 2435, |
| "valid_targets_mean": 5387.0, |
| "valid_targets_min": 979 |
| }, |
| { |
| "epoch": 3.904, |
| "grad_norm": 0.4260255800186257, |
| "learning_rate": 1.948138015454168e-05, |
| "loss": 0.2381, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23484918475151062, |
| "step": 2440, |
| "valid_targets_mean": 3805.9, |
| "valid_targets_min": 1012 |
| }, |
| { |
| "epoch": 3.912, |
| "grad_norm": 0.4706145084246985, |
| "learning_rate": 1.9401614713855775e-05, |
| "loss": 0.2499, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30133742094039917, |
| "step": 2445, |
| "valid_targets_mean": 3473.2, |
| "valid_targets_min": 666 |
| }, |
| { |
| "epoch": 3.92, |
| "grad_norm": 0.42155693267392574, |
| "learning_rate": 1.932185879870218e-05, |
| "loss": 0.2392, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23585715889930725, |
| "step": 2450, |
| "valid_targets_mean": 4632.4, |
| "valid_targets_min": 946 |
| }, |
| { |
| "epoch": 3.928, |
| "grad_norm": 0.42216004508529464, |
| "learning_rate": 1.924211367869357e-05, |
| "loss": 0.2731, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25942063331604004, |
| "step": 2455, |
| "valid_targets_mean": 4232.3, |
| "valid_targets_min": 1077 |
| }, |
| { |
| "epoch": 3.936, |
| "grad_norm": 0.43356255890123696, |
| "learning_rate": 1.9162380623270783e-05, |
| "loss": 0.2306, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24051374197006226, |
| "step": 2460, |
| "valid_targets_mean": 3791.2, |
| "valid_targets_min": 655 |
| }, |
| { |
| "epoch": 3.944, |
| "grad_norm": 0.41613078686108457, |
| "learning_rate": 1.90826609016826e-05, |
| "loss": 0.2323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2255561798810959, |
| "step": 2465, |
| "valid_targets_mean": 3539.2, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 3.952, |
| "grad_norm": 0.38294039222353565, |
| "learning_rate": 1.9002955782965548e-05, |
| "loss": 0.2417, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2199365645647049, |
| "step": 2470, |
| "valid_targets_mean": 4334.4, |
| "valid_targets_min": 791 |
| }, |
| { |
| "epoch": 3.96, |
| "grad_norm": 0.45878262976054157, |
| "learning_rate": 1.8923266535923688e-05, |
| "loss": 0.2528, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27829837799072266, |
| "step": 2475, |
| "valid_targets_mean": 3980.3, |
| "valid_targets_min": 1038 |
| }, |
| { |
| "epoch": 3.968, |
| "grad_norm": 0.42297288882981043, |
| "learning_rate": 1.8843594429108426e-05, |
| "loss": 0.2469, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29415255784988403, |
| "step": 2480, |
| "valid_targets_mean": 4948.5, |
| "valid_targets_min": 932 |
| }, |
| { |
| "epoch": 3.976, |
| "grad_norm": 0.4638883460903558, |
| "learning_rate": 1.8763940730798324e-05, |
| "loss": 0.2375, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2466743290424347, |
| "step": 2485, |
| "valid_targets_mean": 3243.7, |
| "valid_targets_min": 712 |
| }, |
| { |
| "epoch": 3.984, |
| "grad_norm": 0.4897364741478554, |
| "learning_rate": 1.8684306708978896e-05, |
| "loss": 0.2458, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25233909487724304, |
| "step": 2490, |
| "valid_targets_mean": 3177.6, |
| "valid_targets_min": 850 |
| }, |
| { |
| "epoch": 3.992, |
| "grad_norm": 0.46252614216792853, |
| "learning_rate": 1.8604693631322433e-05, |
| "loss": 0.2521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23966452479362488, |
| "step": 2495, |
| "valid_targets_mean": 3006.2, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 4.0, |
| "grad_norm": 0.31710051876751844, |
| "learning_rate": 1.852510276516783e-05, |
| "loss": 0.2467, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20360557734966278, |
| "step": 2500, |
| "valid_targets_mean": 5554.4, |
| "valid_targets_min": 762 |
| }, |
| { |
| "epoch": 4.008, |
| "grad_norm": 0.4182463285312466, |
| "learning_rate": 1.8445535377500393e-05, |
| "loss": 0.2262, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2686900496482849, |
| "step": 2505, |
| "valid_targets_mean": 4283.1, |
| "valid_targets_min": 780 |
| }, |
| { |
| "epoch": 4.016, |
| "grad_norm": 0.34215268944260163, |
| "learning_rate": 1.8365992734931686e-05, |
| "loss": 0.2386, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24253295361995697, |
| "step": 2510, |
| "valid_targets_mean": 6879.1, |
| "valid_targets_min": 993 |
| }, |
| { |
| "epoch": 4.024, |
| "grad_norm": 0.38956003361720304, |
| "learning_rate": 1.8286476103679356e-05, |
| "loss": 0.2283, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21489191055297852, |
| "step": 2515, |
| "valid_targets_mean": 4114.4, |
| "valid_targets_min": 1045 |
| }, |
| { |
| "epoch": 4.032, |
| "grad_norm": 0.3507035259915641, |
| "learning_rate": 1.8206986749546992e-05, |
| "loss": 0.234, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22584864497184753, |
| "step": 2520, |
| "valid_targets_mean": 5080.9, |
| "valid_targets_min": 841 |
| }, |
| { |
| "epoch": 4.04, |
| "grad_norm": 0.3812064932824899, |
| "learning_rate": 1.8127525937903957e-05, |
| "loss": 0.2355, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23593032360076904, |
| "step": 2525, |
| "valid_targets_mean": 4911.8, |
| "valid_targets_min": 760 |
| }, |
| { |
| "epoch": 4.048, |
| "grad_norm": 0.5267876933020568, |
| "learning_rate": 1.8048094933665262e-05, |
| "loss": 0.2375, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2608511745929718, |
| "step": 2530, |
| "valid_targets_mean": 2954.8, |
| "valid_targets_min": 969 |
| }, |
| { |
| "epoch": 4.056, |
| "grad_norm": 0.3449942876056951, |
| "learning_rate": 1.7968695001271416e-05, |
| "loss": 0.2466, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2336217761039734, |
| "step": 2535, |
| "valid_targets_mean": 6292.1, |
| "valid_targets_min": 790 |
| }, |
| { |
| "epoch": 4.064, |
| "grad_norm": 0.4355115464048802, |
| "learning_rate": 1.7889327404668316e-05, |
| "loss": 0.2195, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2266024351119995, |
| "step": 2540, |
| "valid_targets_mean": 3884.9, |
| "valid_targets_min": 879 |
| }, |
| { |
| "epoch": 4.072, |
| "grad_norm": 0.3081848481333973, |
| "learning_rate": 1.7809993407287083e-05, |
| "loss": 0.2544, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22978955507278442, |
| "step": 2545, |
| "valid_targets_mean": 7871.2, |
| "valid_targets_min": 797 |
| }, |
| { |
| "epoch": 4.08, |
| "grad_norm": 0.5481512251233728, |
| "learning_rate": 1.7730694272024018e-05, |
| "loss": 0.2451, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30004870891571045, |
| "step": 2550, |
| "valid_targets_mean": 3161.5, |
| "valid_targets_min": 804 |
| }, |
| { |
| "epoch": 4.088, |
| "grad_norm": 0.48046326523783284, |
| "learning_rate": 1.765143126122044e-05, |
| "loss": 0.2554, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2633433938026428, |
| "step": 2555, |
| "valid_targets_mean": 3137.6, |
| "valid_targets_min": 913 |
| }, |
| { |
| "epoch": 4.096, |
| "grad_norm": 0.45306035014956936, |
| "learning_rate": 1.7572205636642622e-05, |
| "loss": 0.2404, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27019399404525757, |
| "step": 2560, |
| "valid_targets_mean": 4373.4, |
| "valid_targets_min": 793 |
| }, |
| { |
| "epoch": 4.104, |
| "grad_norm": 0.3583844187285849, |
| "learning_rate": 1.749301865946169e-05, |
| "loss": 0.2278, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21709087491035461, |
| "step": 2565, |
| "valid_targets_mean": 5196.9, |
| "valid_targets_min": 823 |
| }, |
| { |
| "epoch": 4.112, |
| "grad_norm": 0.4402520016943769, |
| "learning_rate": 1.7413871590233557e-05, |
| "loss": 0.2571, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26503556966781616, |
| "step": 2570, |
| "valid_targets_mean": 4493.4, |
| "valid_targets_min": 809 |
| }, |
| { |
| "epoch": 4.12, |
| "grad_norm": 0.3555706054314649, |
| "learning_rate": 1.7334765688878848e-05, |
| "loss": 0.2263, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2207963764667511, |
| "step": 2575, |
| "valid_targets_mean": 5484.2, |
| "valid_targets_min": 690 |
| }, |
| { |
| "epoch": 4.128, |
| "grad_norm": 0.4611646375770736, |
| "learning_rate": 1.7255702214662852e-05, |
| "loss": 0.2404, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24676471948623657, |
| "step": 2580, |
| "valid_targets_mean": 3392.4, |
| "valid_targets_min": 784 |
| }, |
| { |
| "epoch": 4.136, |
| "grad_norm": 0.47243039431807954, |
| "learning_rate": 1.7176682426175468e-05, |
| "loss": 0.2566, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2714911103248596, |
| "step": 2585, |
| "valid_targets_mean": 3790.6, |
| "valid_targets_min": 996 |
| }, |
| { |
| "epoch": 4.144, |
| "grad_norm": 0.3937112234840623, |
| "learning_rate": 1.709770758131118e-05, |
| "loss": 0.2487, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25096097588539124, |
| "step": 2590, |
| "valid_targets_mean": 4663.3, |
| "valid_targets_min": 653 |
| }, |
| { |
| "epoch": 4.152, |
| "grad_norm": 0.4070587446512345, |
| "learning_rate": 1.7018778937249017e-05, |
| "loss": 0.2175, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2075790911912918, |
| "step": 2595, |
| "valid_targets_mean": 3767.4, |
| "valid_targets_min": 858 |
| }, |
| { |
| "epoch": 4.16, |
| "grad_norm": 0.36994325328075317, |
| "learning_rate": 1.6939897750432562e-05, |
| "loss": 0.2354, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22925224900245667, |
| "step": 2600, |
| "valid_targets_mean": 5491.9, |
| "valid_targets_min": 719 |
| }, |
| { |
| "epoch": 4.168, |
| "grad_norm": 0.45041384359435777, |
| "learning_rate": 1.6861065276549933e-05, |
| "loss": 0.238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24161088466644287, |
| "step": 2605, |
| "valid_targets_mean": 4071.9, |
| "valid_targets_min": 855 |
| }, |
| { |
| "epoch": 4.176, |
| "grad_norm": 0.4581954398994809, |
| "learning_rate": 1.6782282770513788e-05, |
| "loss": 0.2421, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2478228509426117, |
| "step": 2610, |
| "valid_targets_mean": 3686.6, |
| "valid_targets_min": 709 |
| }, |
| { |
| "epoch": 4.184, |
| "grad_norm": 0.4007037738747935, |
| "learning_rate": 1.6703551486441382e-05, |
| "loss": 0.2359, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2544166147708893, |
| "step": 2615, |
| "valid_targets_mean": 4708.0, |
| "valid_targets_min": 835 |
| }, |
| { |
| "epoch": 4.192, |
| "grad_norm": 0.35558467476339845, |
| "learning_rate": 1.6624872677634565e-05, |
| "loss": 0.2191, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22993795573711395, |
| "step": 2620, |
| "valid_targets_mean": 5626.6, |
| "valid_targets_min": 1024 |
| }, |
| { |
| "epoch": 4.2, |
| "grad_norm": 0.4362176145931464, |
| "learning_rate": 1.654624759655986e-05, |
| "loss": 0.2327, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23500782251358032, |
| "step": 2625, |
| "valid_targets_mean": 4308.1, |
| "valid_targets_min": 908 |
| }, |
| { |
| "epoch": 4.208, |
| "grad_norm": 0.43269722980811254, |
| "learning_rate": 1.64676774948285e-05, |
| "loss": 0.2531, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25792813301086426, |
| "step": 2630, |
| "valid_targets_mean": 4154.8, |
| "valid_targets_min": 1168 |
| }, |
| { |
| "epoch": 4.216, |
| "grad_norm": 0.362438370317793, |
| "learning_rate": 1.6389163623176536e-05, |
| "loss": 0.2188, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23212498426437378, |
| "step": 2635, |
| "valid_targets_mean": 5189.7, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 4.224, |
| "grad_norm": 0.6495031296425801, |
| "learning_rate": 1.6310707231444884e-05, |
| "loss": 0.2496, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33409446477890015, |
| "step": 2640, |
| "valid_targets_mean": 2262.6, |
| "valid_targets_min": 923 |
| }, |
| { |
| "epoch": 4.232, |
| "grad_norm": 0.3044482725544302, |
| "learning_rate": 1.623230956855947e-05, |
| "loss": 0.2257, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1921289712190628, |
| "step": 2645, |
| "valid_targets_mean": 6717.5, |
| "valid_targets_min": 1050 |
| }, |
| { |
| "epoch": 4.24, |
| "grad_norm": 0.4021251156140196, |
| "learning_rate": 1.6153971882511324e-05, |
| "loss": 0.2388, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23349186778068542, |
| "step": 2650, |
| "valid_targets_mean": 4582.3, |
| "valid_targets_min": 722 |
| }, |
| { |
| "epoch": 4.248, |
| "grad_norm": 0.3445640898802388, |
| "learning_rate": 1.6075695420336724e-05, |
| "loss": 0.2186, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24022488296031952, |
| "step": 2655, |
| "valid_targets_mean": 5738.2, |
| "valid_targets_min": 859 |
| }, |
| { |
| "epoch": 4.256, |
| "grad_norm": 0.3634851638301243, |
| "learning_rate": 1.5997481428097338e-05, |
| "loss": 0.2211, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18648147583007812, |
| "step": 2660, |
| "valid_targets_mean": 5074.1, |
| "valid_targets_min": 879 |
| }, |
| { |
| "epoch": 4.264, |
| "grad_norm": 0.4641862374466239, |
| "learning_rate": 1.5919331150860396e-05, |
| "loss": 0.2343, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23830462992191315, |
| "step": 2665, |
| "valid_targets_mean": 3713.1, |
| "valid_targets_min": 823 |
| }, |
| { |
| "epoch": 4.272, |
| "grad_norm": 0.5391929747868811, |
| "learning_rate": 1.5841245832678873e-05, |
| "loss": 0.2384, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2308681607246399, |
| "step": 2670, |
| "valid_targets_mean": 3575.1, |
| "valid_targets_min": 748 |
| }, |
| { |
| "epoch": 4.28, |
| "grad_norm": 0.35324812073512407, |
| "learning_rate": 1.576322671657166e-05, |
| "loss": 0.2296, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22632098197937012, |
| "step": 2675, |
| "valid_targets_mean": 5186.4, |
| "valid_targets_min": 764 |
| }, |
| { |
| "epoch": 4.288, |
| "grad_norm": 0.36772896338506605, |
| "learning_rate": 1.5685275044503804e-05, |
| "loss": 0.2296, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20238947868347168, |
| "step": 2680, |
| "valid_targets_mean": 4375.9, |
| "valid_targets_min": 998 |
| }, |
| { |
| "epoch": 4.296, |
| "grad_norm": 0.5917343917305425, |
| "learning_rate": 1.560739205736674e-05, |
| "loss": 0.2441, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2685689628124237, |
| "step": 2685, |
| "valid_targets_mean": 2465.2, |
| "valid_targets_min": 745 |
| }, |
| { |
| "epoch": 4.304, |
| "grad_norm": 0.38800770329149276, |
| "learning_rate": 1.552957899495851e-05, |
| "loss": 0.2386, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22670996189117432, |
| "step": 2690, |
| "valid_targets_mean": 4851.9, |
| "valid_targets_min": 1051 |
| }, |
| { |
| "epoch": 4.312, |
| "grad_norm": 0.4348192390077996, |
| "learning_rate": 1.5451837095964054e-05, |
| "loss": 0.2312, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23623718321323395, |
| "step": 2695, |
| "valid_targets_mean": 4257.4, |
| "valid_targets_min": 741 |
| }, |
| { |
| "epoch": 4.32, |
| "grad_norm": 0.3150520299968283, |
| "learning_rate": 1.5374167597935478e-05, |
| "loss": 0.2113, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18955495953559875, |
| "step": 2700, |
| "valid_targets_mean": 5891.4, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 4.328, |
| "grad_norm": 0.4228257855789774, |
| "learning_rate": 1.5296571737272354e-05, |
| "loss": 0.2296, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22553138434886932, |
| "step": 2705, |
| "valid_targets_mean": 3918.9, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 4.336, |
| "grad_norm": 0.4441377948845305, |
| "learning_rate": 1.5219050749202037e-05, |
| "loss": 0.221, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22206555306911469, |
| "step": 2710, |
| "valid_targets_mean": 3918.2, |
| "valid_targets_min": 916 |
| }, |
| { |
| "epoch": 4.344, |
| "grad_norm": 0.5003922378983456, |
| "learning_rate": 1.5141605867760021e-05, |
| "loss": 0.2539, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2991781234741211, |
| "step": 2715, |
| "valid_targets_mean": 3582.6, |
| "valid_targets_min": 670 |
| }, |
| { |
| "epoch": 4.352, |
| "grad_norm": 0.3967867101584086, |
| "learning_rate": 1.5064238325770267e-05, |
| "loss": 0.2226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24074117839336395, |
| "step": 2720, |
| "valid_targets_mean": 4480.3, |
| "valid_targets_min": 945 |
| }, |
| { |
| "epoch": 4.36, |
| "grad_norm": 0.40832709295548075, |
| "learning_rate": 1.498694935482559e-05, |
| "loss": 0.236, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2141822874546051, |
| "step": 2725, |
| "valid_targets_mean": 3890.5, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 4.368, |
| "grad_norm": 0.5353662529083422, |
| "learning_rate": 1.4909740185268056e-05, |
| "loss": 0.2433, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2768443822860718, |
| "step": 2730, |
| "valid_targets_mean": 3256.1, |
| "valid_targets_min": 732 |
| }, |
| { |
| "epoch": 4.376, |
| "grad_norm": 0.3903407215377192, |
| "learning_rate": 1.4832612046169408e-05, |
| "loss": 0.2425, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25160282850265503, |
| "step": 2735, |
| "valid_targets_mean": 5186.7, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 4.384, |
| "grad_norm": 0.3610608999407683, |
| "learning_rate": 1.4755566165311455e-05, |
| "loss": 0.2223, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21602243185043335, |
| "step": 2740, |
| "valid_targets_mean": 5826.9, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 4.392, |
| "grad_norm": 0.3705361464347009, |
| "learning_rate": 1.4678603769166591e-05, |
| "loss": 0.2323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2093980610370636, |
| "step": 2745, |
| "valid_targets_mean": 4775.3, |
| "valid_targets_min": 685 |
| }, |
| { |
| "epoch": 4.4, |
| "grad_norm": 0.5672264522039926, |
| "learning_rate": 1.4601726082878226e-05, |
| "loss": 0.2265, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23798605799674988, |
| "step": 2750, |
| "valid_targets_mean": 2565.5, |
| "valid_targets_min": 851 |
| }, |
| { |
| "epoch": 4.408, |
| "grad_norm": 0.3839516251209785, |
| "learning_rate": 1.4524934330241292e-05, |
| "loss": 0.2155, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20655933022499084, |
| "step": 2755, |
| "valid_targets_mean": 4338.2, |
| "valid_targets_min": 710 |
| }, |
| { |
| "epoch": 4.416, |
| "grad_norm": 0.6594074981447611, |
| "learning_rate": 1.4448229733682784e-05, |
| "loss": 0.235, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24626955389976501, |
| "step": 2760, |
| "valid_targets_mean": 2719.0, |
| "valid_targets_min": 717 |
| }, |
| { |
| "epoch": 4.424, |
| "grad_norm": 0.5159254418985769, |
| "learning_rate": 1.4371613514242264e-05, |
| "loss": 0.2307, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24901220202445984, |
| "step": 2765, |
| "valid_targets_mean": 3019.8, |
| "valid_targets_min": 881 |
| }, |
| { |
| "epoch": 4.432, |
| "grad_norm": 0.3594039502636306, |
| "learning_rate": 1.4295086891552457e-05, |
| "loss": 0.2197, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2654702663421631, |
| "step": 2770, |
| "valid_targets_mean": 6056.7, |
| "valid_targets_min": 1036 |
| }, |
| { |
| "epoch": 4.44, |
| "grad_norm": 0.34691201685012074, |
| "learning_rate": 1.4218651083819811e-05, |
| "loss": 0.2217, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21709834039211273, |
| "step": 2775, |
| "valid_targets_mean": 5528.9, |
| "valid_targets_min": 1091 |
| }, |
| { |
| "epoch": 4.448, |
| "grad_norm": 0.4454510808301901, |
| "learning_rate": 1.4142307307805125e-05, |
| "loss": 0.2361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2332070767879486, |
| "step": 2780, |
| "valid_targets_mean": 3425.8, |
| "valid_targets_min": 626 |
| }, |
| { |
| "epoch": 4.456, |
| "grad_norm": 0.33114532349140574, |
| "learning_rate": 1.406605677880416e-05, |
| "loss": 0.225, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.216552734375, |
| "step": 2785, |
| "valid_targets_mean": 5557.9, |
| "valid_targets_min": 686 |
| }, |
| { |
| "epoch": 4.464, |
| "grad_norm": 0.36774664936117085, |
| "learning_rate": 1.3989900710628313e-05, |
| "loss": 0.239, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21505959331989288, |
| "step": 2790, |
| "valid_targets_mean": 4693.8, |
| "valid_targets_min": 990 |
| }, |
| { |
| "epoch": 4.4719999999999995, |
| "grad_norm": 0.3545900301698533, |
| "learning_rate": 1.3913840315585279e-05, |
| "loss": 0.238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21788690984249115, |
| "step": 2795, |
| "valid_targets_mean": 5270.3, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 4.48, |
| "grad_norm": 0.4302379594475467, |
| "learning_rate": 1.3837876804459765e-05, |
| "loss": 0.2277, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24874816834926605, |
| "step": 2800, |
| "valid_targets_mean": 4287.2, |
| "valid_targets_min": 1103 |
| }, |
| { |
| "epoch": 4.4879999999999995, |
| "grad_norm": 0.367392918189934, |
| "learning_rate": 1.3762011386494191e-05, |
| "loss": 0.2332, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22634848952293396, |
| "step": 2805, |
| "valid_targets_mean": 4934.0, |
| "valid_targets_min": 917 |
| }, |
| { |
| "epoch": 4.496, |
| "grad_norm": 0.3355554630083534, |
| "learning_rate": 1.3686245269369485e-05, |
| "loss": 0.2392, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21311712265014648, |
| "step": 2810, |
| "valid_targets_mean": 5740.8, |
| "valid_targets_min": 991 |
| }, |
| { |
| "epoch": 4.504, |
| "grad_norm": 0.3839903505349243, |
| "learning_rate": 1.3610579659185809e-05, |
| "loss": 0.2291, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22502079606056213, |
| "step": 2815, |
| "valid_targets_mean": 4504.8, |
| "valid_targets_min": 955 |
| }, |
| { |
| "epoch": 4.5120000000000005, |
| "grad_norm": 0.3433561120934192, |
| "learning_rate": 1.35350157604434e-05, |
| "loss": 0.2361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23318302631378174, |
| "step": 2820, |
| "valid_targets_mean": 6099.9, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 4.52, |
| "grad_norm": 0.3924173308347554, |
| "learning_rate": 1.345955477602337e-05, |
| "loss": 0.232, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22661542892456055, |
| "step": 2825, |
| "valid_targets_mean": 4413.6, |
| "valid_targets_min": 1228 |
| }, |
| { |
| "epoch": 4.5280000000000005, |
| "grad_norm": 0.3671450930398099, |
| "learning_rate": 1.3384197907168561e-05, |
| "loss": 0.2323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2506842613220215, |
| "step": 2830, |
| "valid_targets_mean": 5014.2, |
| "valid_targets_min": 865 |
| }, |
| { |
| "epoch": 4.536, |
| "grad_norm": 0.4471447410885578, |
| "learning_rate": 1.3308946353464438e-05, |
| "loss": 0.2395, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25687381625175476, |
| "step": 2835, |
| "valid_targets_mean": 3734.6, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 4.5440000000000005, |
| "grad_norm": 0.3665444810496819, |
| "learning_rate": 1.3233801312819979e-05, |
| "loss": 0.2364, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24433283507823944, |
| "step": 2840, |
| "valid_targets_mean": 5373.4, |
| "valid_targets_min": 967 |
| }, |
| { |
| "epoch": 4.552, |
| "grad_norm": 0.38498330932928676, |
| "learning_rate": 1.3158763981448606e-05, |
| "loss": 0.2491, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23016983270645142, |
| "step": 2845, |
| "valid_targets_mean": 4454.9, |
| "valid_targets_min": 1088 |
| }, |
| { |
| "epoch": 4.5600000000000005, |
| "grad_norm": 0.3993486182470705, |
| "learning_rate": 1.3083835553849148e-05, |
| "loss": 0.2384, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24869374930858612, |
| "step": 2850, |
| "valid_targets_mean": 4435.8, |
| "valid_targets_min": 739 |
| }, |
| { |
| "epoch": 4.568, |
| "grad_norm": 0.4530152443374673, |
| "learning_rate": 1.3009017222786828e-05, |
| "loss": 0.2405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2325194925069809, |
| "step": 2855, |
| "valid_targets_mean": 3650.1, |
| "valid_targets_min": 1149 |
| }, |
| { |
| "epoch": 4.576, |
| "grad_norm": 0.45145189985347955, |
| "learning_rate": 1.2934310179274269e-05, |
| "loss": 0.2342, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24084368348121643, |
| "step": 2860, |
| "valid_targets_mean": 3644.5, |
| "valid_targets_min": 704 |
| }, |
| { |
| "epoch": 4.584, |
| "grad_norm": 0.4475782725112683, |
| "learning_rate": 1.2859715612552541e-05, |
| "loss": 0.2372, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21897542476654053, |
| "step": 2865, |
| "valid_targets_mean": 3397.4, |
| "valid_targets_min": 1067 |
| }, |
| { |
| "epoch": 4.592, |
| "grad_norm": 0.4612371616890256, |
| "learning_rate": 1.278523471007223e-05, |
| "loss": 0.2474, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22148343920707703, |
| "step": 2870, |
| "valid_targets_mean": 2955.4, |
| "valid_targets_min": 777 |
| }, |
| { |
| "epoch": 4.6, |
| "grad_norm": 0.3387602746131469, |
| "learning_rate": 1.271086865747451e-05, |
| "loss": 0.2464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24698293209075928, |
| "step": 2875, |
| "valid_targets_mean": 6360.4, |
| "valid_targets_min": 1383 |
| }, |
| { |
| "epoch": 4.608, |
| "grad_norm": 0.3839373637303783, |
| "learning_rate": 1.2636618638572316e-05, |
| "loss": 0.2216, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2379951924085617, |
| "step": 2880, |
| "valid_targets_mean": 5022.4, |
| "valid_targets_min": 829 |
| }, |
| { |
| "epoch": 4.616, |
| "grad_norm": 0.3197611013771196, |
| "learning_rate": 1.2562485835331466e-05, |
| "loss": 0.2396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23118649423122406, |
| "step": 2885, |
| "valid_targets_mean": 7057.1, |
| "valid_targets_min": 1038 |
| }, |
| { |
| "epoch": 4.624, |
| "grad_norm": 0.43290245672745614, |
| "learning_rate": 1.2488471427851852e-05, |
| "loss": 0.2271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23651964962482452, |
| "step": 2890, |
| "valid_targets_mean": 3611.8, |
| "valid_targets_min": 851 |
| }, |
| { |
| "epoch": 4.632, |
| "grad_norm": 0.4331683426323707, |
| "learning_rate": 1.241457659434866e-05, |
| "loss": 0.2371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22483578324317932, |
| "step": 2895, |
| "valid_targets_mean": 3447.9, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 4.64, |
| "grad_norm": 0.3988096882137287, |
| "learning_rate": 1.2340802511133605e-05, |
| "loss": 0.2381, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2243075966835022, |
| "step": 2900, |
| "valid_targets_mean": 4385.4, |
| "valid_targets_min": 863 |
| }, |
| { |
| "epoch": 4.648, |
| "grad_norm": 0.40557367557155005, |
| "learning_rate": 1.2267150352596216e-05, |
| "loss": 0.2268, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22742396593093872, |
| "step": 2905, |
| "valid_targets_mean": 4335.8, |
| "valid_targets_min": 941 |
| }, |
| { |
| "epoch": 4.656, |
| "grad_norm": 0.41903497071079515, |
| "learning_rate": 1.2193621291185132e-05, |
| "loss": 0.2272, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23160028457641602, |
| "step": 2910, |
| "valid_targets_mean": 3808.8, |
| "valid_targets_min": 916 |
| }, |
| { |
| "epoch": 4.664, |
| "grad_norm": 0.3333342358026764, |
| "learning_rate": 1.2120216497389446e-05, |
| "loss": 0.2259, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21926115453243256, |
| "step": 2915, |
| "valid_targets_mean": 5564.4, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 4.672, |
| "grad_norm": 0.35498213383205174, |
| "learning_rate": 1.2046937139720068e-05, |
| "loss": 0.2202, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23465262353420258, |
| "step": 2920, |
| "valid_targets_mean": 5612.9, |
| "valid_targets_min": 901 |
| }, |
| { |
| "epoch": 4.68, |
| "grad_norm": 0.43676624123660573, |
| "learning_rate": 1.1973784384691121e-05, |
| "loss": 0.2355, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22778140008449554, |
| "step": 2925, |
| "valid_targets_mean": 3652.6, |
| "valid_targets_min": 744 |
| }, |
| { |
| "epoch": 4.688, |
| "grad_norm": 0.43153283817401733, |
| "learning_rate": 1.1900759396801382e-05, |
| "loss": 0.2533, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23726201057434082, |
| "step": 2930, |
| "valid_targets_mean": 3769.8, |
| "valid_targets_min": 1104 |
| }, |
| { |
| "epoch": 4.696, |
| "grad_norm": 0.4041357939241981, |
| "learning_rate": 1.1827863338515741e-05, |
| "loss": 0.2352, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23281404376029968, |
| "step": 2935, |
| "valid_targets_mean": 4432.1, |
| "valid_targets_min": 897 |
| }, |
| { |
| "epoch": 4.704, |
| "grad_norm": 0.3514377881831837, |
| "learning_rate": 1.1755097370246669e-05, |
| "loss": 0.224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.220592200756073, |
| "step": 2940, |
| "valid_targets_mean": 5495.7, |
| "valid_targets_min": 793 |
| }, |
| { |
| "epoch": 4.712, |
| "grad_norm": 0.4214613254086311, |
| "learning_rate": 1.1682462650335791e-05, |
| "loss": 0.2244, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2115774154663086, |
| "step": 2945, |
| "valid_targets_mean": 3903.4, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 4.72, |
| "grad_norm": 0.5099284948600926, |
| "learning_rate": 1.1609960335035423e-05, |
| "loss": 0.2211, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22030048072338104, |
| "step": 2950, |
| "valid_targets_mean": 2719.6, |
| "valid_targets_min": 771 |
| }, |
| { |
| "epoch": 4.728, |
| "grad_norm": 0.3655229242181005, |
| "learning_rate": 1.1537591578490165e-05, |
| "loss": 0.237, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20158278942108154, |
| "step": 2955, |
| "valid_targets_mean": 4298.5, |
| "valid_targets_min": 793 |
| }, |
| { |
| "epoch": 4.736, |
| "grad_norm": 0.46139845446107813, |
| "learning_rate": 1.146535753271853e-05, |
| "loss": 0.2321, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24375522136688232, |
| "step": 2960, |
| "valid_targets_mean": 3361.3, |
| "valid_targets_min": 902 |
| }, |
| { |
| "epoch": 4.744, |
| "grad_norm": 0.4688790022851686, |
| "learning_rate": 1.139325934759461e-05, |
| "loss": 0.2485, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25852593779563904, |
| "step": 2965, |
| "valid_targets_mean": 3596.0, |
| "valid_targets_min": 968 |
| }, |
| { |
| "epoch": 4.752, |
| "grad_norm": 0.5227422410545948, |
| "learning_rate": 1.1321298170829768e-05, |
| "loss": 0.2668, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24697157740592957, |
| "step": 2970, |
| "valid_targets_mean": 2877.5, |
| "valid_targets_min": 887 |
| }, |
| { |
| "epoch": 4.76, |
| "grad_norm": 0.4038881666049419, |
| "learning_rate": 1.1249475147954363e-05, |
| "loss": 0.2329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21838301420211792, |
| "step": 2975, |
| "valid_targets_mean": 4123.2, |
| "valid_targets_min": 777 |
| }, |
| { |
| "epoch": 4.768, |
| "grad_norm": 0.5184660299912265, |
| "learning_rate": 1.1177791422299528e-05, |
| "loss": 0.2531, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27220386266708374, |
| "step": 2980, |
| "valid_targets_mean": 3015.9, |
| "valid_targets_min": 630 |
| }, |
| { |
| "epoch": 4.776, |
| "grad_norm": 0.3654819987069819, |
| "learning_rate": 1.1106248134978959e-05, |
| "loss": 0.2352, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22408969700336456, |
| "step": 2985, |
| "valid_targets_mean": 5349.6, |
| "valid_targets_min": 1141 |
| }, |
| { |
| "epoch": 4.784, |
| "grad_norm": 0.3660126606227132, |
| "learning_rate": 1.1034846424870744e-05, |
| "loss": 0.2417, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2300996333360672, |
| "step": 2990, |
| "valid_targets_mean": 4971.9, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 4.792, |
| "grad_norm": 0.3756476057343205, |
| "learning_rate": 1.0963587428599256e-05, |
| "loss": 0.2314, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2119663953781128, |
| "step": 2995, |
| "valid_targets_mean": 4842.7, |
| "valid_targets_min": 732 |
| }, |
| { |
| "epoch": 4.8, |
| "grad_norm": 0.4613837764914878, |
| "learning_rate": 1.089247228051704e-05, |
| "loss": 0.2306, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24079769849777222, |
| "step": 3000, |
| "valid_targets_mean": 3716.3, |
| "valid_targets_min": 759 |
| }, |
| { |
| "epoch": 4.808, |
| "grad_norm": 0.28943767912823337, |
| "learning_rate": 1.0821502112686753e-05, |
| "loss": 0.2298, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19776251912117004, |
| "step": 3005, |
| "valid_targets_mean": 6197.8, |
| "valid_targets_min": 782 |
| }, |
| { |
| "epoch": 4.816, |
| "grad_norm": 0.4161037969654231, |
| "learning_rate": 1.0750678054863158e-05, |
| "loss": 0.2321, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22658786177635193, |
| "step": 3010, |
| "valid_targets_mean": 3665.1, |
| "valid_targets_min": 752 |
| }, |
| { |
| "epoch": 4.824, |
| "grad_norm": 0.31379216309346025, |
| "learning_rate": 1.0680001234475127e-05, |
| "loss": 0.2212, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20140352845191956, |
| "step": 3015, |
| "valid_targets_mean": 6177.3, |
| "valid_targets_min": 887 |
| }, |
| { |
| "epoch": 4.832, |
| "grad_norm": 0.3780126118328765, |
| "learning_rate": 1.0609472776607715e-05, |
| "loss": 0.2296, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24119669198989868, |
| "step": 3020, |
| "valid_targets_mean": 6278.8, |
| "valid_targets_min": 1106 |
| }, |
| { |
| "epoch": 4.84, |
| "grad_norm": 0.3373262706812363, |
| "learning_rate": 1.0539093803984217e-05, |
| "loss": 0.2142, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20110008120536804, |
| "step": 3025, |
| "valid_targets_mean": 5751.3, |
| "valid_targets_min": 1033 |
| }, |
| { |
| "epoch": 4.848, |
| "grad_norm": 0.37649695603656197, |
| "learning_rate": 1.046886543694832e-05, |
| "loss": 0.2338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2516307532787323, |
| "step": 3030, |
| "valid_targets_mean": 5042.4, |
| "valid_targets_min": 1139 |
| }, |
| { |
| "epoch": 4.856, |
| "grad_norm": 0.38263029301997636, |
| "learning_rate": 1.0398788793446263e-05, |
| "loss": 0.2327, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.207616925239563, |
| "step": 3035, |
| "valid_targets_mean": 4312.8, |
| "valid_targets_min": 1036 |
| }, |
| { |
| "epoch": 4.864, |
| "grad_norm": 0.42458541451939813, |
| "learning_rate": 1.0328864989009037e-05, |
| "loss": 0.229, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22665390372276306, |
| "step": 3040, |
| "valid_targets_mean": 3687.1, |
| "valid_targets_min": 969 |
| }, |
| { |
| "epoch": 4.872, |
| "grad_norm": 0.3892606340058019, |
| "learning_rate": 1.0259095136734634e-05, |
| "loss": 0.2207, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23070894181728363, |
| "step": 3045, |
| "valid_targets_mean": 4354.3, |
| "valid_targets_min": 918 |
| }, |
| { |
| "epoch": 4.88, |
| "grad_norm": 0.40230384154567744, |
| "learning_rate": 1.0189480347270311e-05, |
| "loss": 0.232, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24141496419906616, |
| "step": 3050, |
| "valid_targets_mean": 4812.8, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 4.888, |
| "grad_norm": 0.4084264186589575, |
| "learning_rate": 1.0120021728794938e-05, |
| "loss": 0.2204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21485230326652527, |
| "step": 3055, |
| "valid_targets_mean": 4172.5, |
| "valid_targets_min": 969 |
| }, |
| { |
| "epoch": 4.896, |
| "grad_norm": 0.5082884887305914, |
| "learning_rate": 1.0050720387001334e-05, |
| "loss": 0.2296, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2505829930305481, |
| "step": 3060, |
| "valid_targets_mean": 2977.1, |
| "valid_targets_min": 690 |
| }, |
| { |
| "epoch": 4.904, |
| "grad_norm": 0.38678833449761596, |
| "learning_rate": 9.981577425078672e-06, |
| "loss": 0.2343, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2366374433040619, |
| "step": 3065, |
| "valid_targets_mean": 4762.4, |
| "valid_targets_min": 1002 |
| }, |
| { |
| "epoch": 4.912, |
| "grad_norm": 0.38418961481494723, |
| "learning_rate": 9.912593943694924e-06, |
| "loss": 0.2319, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1991523802280426, |
| "step": 3070, |
| "valid_targets_mean": 4543.4, |
| "valid_targets_min": 1056 |
| }, |
| { |
| "epoch": 4.92, |
| "grad_norm": 0.4832708684343786, |
| "learning_rate": 9.843771040979328e-06, |
| "loss": 0.2498, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24161699414253235, |
| "step": 3075, |
| "valid_targets_mean": 3462.6, |
| "valid_targets_min": 763 |
| }, |
| { |
| "epoch": 4.928, |
| "grad_norm": 0.4309087603014352, |
| "learning_rate": 9.775109812504922e-06, |
| "loss": 0.2606, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20923660695552826, |
| "step": 3080, |
| "valid_targets_mean": 3484.4, |
| "valid_targets_min": 664 |
| }, |
| { |
| "epoch": 4.936, |
| "grad_norm": 0.3788722471104869, |
| "learning_rate": 9.706611351271088e-06, |
| "loss": 0.2319, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22105582058429718, |
| "step": 3085, |
| "valid_targets_mean": 4319.1, |
| "valid_targets_min": 721 |
| }, |
| { |
| "epoch": 4.944, |
| "grad_norm": 0.4355369088708729, |
| "learning_rate": 9.638276747686169e-06, |
| "loss": 0.2428, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24673102796077728, |
| "step": 3090, |
| "valid_targets_mean": 3661.9, |
| "valid_targets_min": 881 |
| }, |
| { |
| "epoch": 4.952, |
| "grad_norm": 0.40667695740023535, |
| "learning_rate": 9.570107089550091e-06, |
| "loss": 0.2351, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23325340449810028, |
| "step": 3095, |
| "valid_targets_mean": 4411.1, |
| "valid_targets_min": 902 |
| }, |
| { |
| "epoch": 4.96, |
| "grad_norm": 0.3942359450360996, |
| "learning_rate": 9.502103462037074e-06, |
| "loss": 0.2253, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2507432699203491, |
| "step": 3100, |
| "valid_targets_mean": 4678.1, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 4.968, |
| "grad_norm": 0.395308435273449, |
| "learning_rate": 9.434266947678326e-06, |
| "loss": 0.2292, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22862330079078674, |
| "step": 3105, |
| "valid_targets_mean": 4401.1, |
| "valid_targets_min": 1048 |
| }, |
| { |
| "epoch": 4.976, |
| "grad_norm": 0.35534927093483687, |
| "learning_rate": 9.366598626344836e-06, |
| "loss": 0.2347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24637477099895477, |
| "step": 3110, |
| "valid_targets_mean": 5403.4, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 4.984, |
| "grad_norm": 0.3598184673475776, |
| "learning_rate": 9.299099575230172e-06, |
| "loss": 0.2528, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23107300698757172, |
| "step": 3115, |
| "valid_targets_mean": 5155.9, |
| "valid_targets_min": 1085 |
| }, |
| { |
| "epoch": 4.992, |
| "grad_norm": 0.5088527769647312, |
| "learning_rate": 9.231770868833334e-06, |
| "loss": 0.2464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26669660210609436, |
| "step": 3120, |
| "valid_targets_mean": 3198.4, |
| "valid_targets_min": 982 |
| }, |
| { |
| "epoch": 5.0, |
| "grad_norm": 0.34198183838948926, |
| "learning_rate": 9.164613578941652e-06, |
| "loss": 0.2254, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21536295115947723, |
| "step": 3125, |
| "valid_targets_mean": 5529.9, |
| "valid_targets_min": 896 |
| }, |
| { |
| "epoch": 5.008, |
| "grad_norm": 0.4912006951999863, |
| "learning_rate": 9.097628774613732e-06, |
| "loss": 0.2251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2081344723701477, |
| "step": 3130, |
| "valid_targets_mean": 2727.4, |
| "valid_targets_min": 743 |
| }, |
| { |
| "epoch": 5.016, |
| "grad_norm": 0.4084241000168698, |
| "learning_rate": 9.030817522162403e-06, |
| "loss": 0.23, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2399434745311737, |
| "step": 3135, |
| "valid_targets_mean": 4014.1, |
| "valid_targets_min": 664 |
| }, |
| { |
| "epoch": 5.024, |
| "grad_norm": 0.43709779953742833, |
| "learning_rate": 8.964180885137797e-06, |
| "loss": 0.2182, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22896772623062134, |
| "step": 3140, |
| "valid_targets_mean": 4365.4, |
| "valid_targets_min": 1046 |
| }, |
| { |
| "epoch": 5.032, |
| "grad_norm": 0.305518599208397, |
| "learning_rate": 8.897719924310375e-06, |
| "loss": 0.2348, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20471172034740448, |
| "step": 3145, |
| "valid_targets_mean": 6402.6, |
| "valid_targets_min": 931 |
| }, |
| { |
| "epoch": 5.04, |
| "grad_norm": 0.37990073340945496, |
| "learning_rate": 8.831435697654068e-06, |
| "loss": 0.2202, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19448421895503998, |
| "step": 3150, |
| "valid_targets_mean": 4079.8, |
| "valid_targets_min": 902 |
| }, |
| { |
| "epoch": 5.048, |
| "grad_norm": 0.31479459501943424, |
| "learning_rate": 8.765329260329413e-06, |
| "loss": 0.2219, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19626212120056152, |
| "step": 3155, |
| "valid_targets_mean": 6214.5, |
| "valid_targets_min": 951 |
| }, |
| { |
| "epoch": 5.056, |
| "grad_norm": 0.3740116945023727, |
| "learning_rate": 8.699401664666774e-06, |
| "loss": 0.2321, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19557811319828033, |
| "step": 3160, |
| "valid_targets_mean": 4366.0, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 5.064, |
| "grad_norm": 0.4192074460438228, |
| "learning_rate": 8.633653960149579e-06, |
| "loss": 0.2312, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26347237825393677, |
| "step": 3165, |
| "valid_targets_mean": 4294.9, |
| "valid_targets_min": 869 |
| }, |
| { |
| "epoch": 5.072, |
| "grad_norm": 0.3170004260395546, |
| "learning_rate": 8.56808719339762e-06, |
| "loss": 0.2366, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20620077848434448, |
| "step": 3170, |
| "valid_targets_mean": 6649.5, |
| "valid_targets_min": 888 |
| }, |
| { |
| "epoch": 5.08, |
| "grad_norm": 0.3841314180170002, |
| "learning_rate": 8.502702408150391e-06, |
| "loss": 0.2197, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24174794554710388, |
| "step": 3175, |
| "valid_targets_mean": 4925.1, |
| "valid_targets_min": 909 |
| }, |
| { |
| "epoch": 5.088, |
| "grad_norm": 0.34188326135593766, |
| "learning_rate": 8.43750064525047e-06, |
| "loss": 0.2337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19666939973831177, |
| "step": 3180, |
| "valid_targets_mean": 5542.1, |
| "valid_targets_min": 1072 |
| }, |
| { |
| "epoch": 5.096, |
| "grad_norm": 0.48888208213056195, |
| "learning_rate": 8.372482942626952e-06, |
| "loss": 0.2289, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25927528738975525, |
| "step": 3185, |
| "valid_targets_mean": 3291.2, |
| "valid_targets_min": 950 |
| }, |
| { |
| "epoch": 5.104, |
| "grad_norm": 0.5266810833672644, |
| "learning_rate": 8.307650335278927e-06, |
| "loss": 0.2403, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2526220381259918, |
| "step": 3190, |
| "valid_targets_mean": 3052.0, |
| "valid_targets_min": 842 |
| }, |
| { |
| "epoch": 5.112, |
| "grad_norm": 0.3489078895076499, |
| "learning_rate": 8.243003855259015e-06, |
| "loss": 0.2394, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23310494422912598, |
| "step": 3195, |
| "valid_targets_mean": 6052.8, |
| "valid_targets_min": 1104 |
| }, |
| { |
| "epoch": 5.12, |
| "grad_norm": 0.4273931588420199, |
| "learning_rate": 8.178544531656897e-06, |
| "loss": 0.2196, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2163240909576416, |
| "step": 3200, |
| "valid_targets_mean": 4048.2, |
| "valid_targets_min": 747 |
| }, |
| { |
| "epoch": 5.128, |
| "grad_norm": 0.4287189766428896, |
| "learning_rate": 8.11427339058299e-06, |
| "loss": 0.2492, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25983962416648865, |
| "step": 3205, |
| "valid_targets_mean": 4372.1, |
| "valid_targets_min": 1021 |
| }, |
| { |
| "epoch": 5.136, |
| "grad_norm": 0.4697741992206112, |
| "learning_rate": 8.050191455152072e-06, |
| "loss": 0.2228, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23713898658752441, |
| "step": 3210, |
| "valid_targets_mean": 3780.0, |
| "valid_targets_min": 833 |
| }, |
| { |
| "epoch": 5.144, |
| "grad_norm": 0.3426204752362461, |
| "learning_rate": 7.986299745467013e-06, |
| "loss": 0.2124, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2151990830898285, |
| "step": 3215, |
| "valid_targets_mean": 5340.3, |
| "valid_targets_min": 1017 |
| }, |
| { |
| "epoch": 5.152, |
| "grad_norm": 0.3476898053916093, |
| "learning_rate": 7.922599278602524e-06, |
| "loss": 0.2271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2065230756998062, |
| "step": 3220, |
| "valid_targets_mean": 5802.0, |
| "valid_targets_min": 987 |
| }, |
| { |
| "epoch": 5.16, |
| "grad_norm": 0.34997024370706975, |
| "learning_rate": 7.859091068588987e-06, |
| "loss": 0.223, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22452686727046967, |
| "step": 3225, |
| "valid_targets_mean": 5912.6, |
| "valid_targets_min": 711 |
| }, |
| { |
| "epoch": 5.168, |
| "grad_norm": 0.5145349413640553, |
| "learning_rate": 7.795776126396284e-06, |
| "loss": 0.2294, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26052239537239075, |
| "step": 3230, |
| "valid_targets_mean": 3026.5, |
| "valid_targets_min": 938 |
| }, |
| { |
| "epoch": 5.176, |
| "grad_norm": 0.44391793492379567, |
| "learning_rate": 7.732655459917726e-06, |
| "loss": 0.2162, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23127031326293945, |
| "step": 3235, |
| "valid_targets_mean": 3623.2, |
| "valid_targets_min": 646 |
| }, |
| { |
| "epoch": 5.184, |
| "grad_norm": 0.3630611195177078, |
| "learning_rate": 7.669730073954005e-06, |
| "loss": 0.2373, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21552309393882751, |
| "step": 3240, |
| "valid_targets_mean": 5270.8, |
| "valid_targets_min": 901 |
| }, |
| { |
| "epoch": 5.192, |
| "grad_norm": 0.4102333340497864, |
| "learning_rate": 7.607000970197194e-06, |
| "loss": 0.2317, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24429859220981598, |
| "step": 3245, |
| "valid_targets_mean": 4244.9, |
| "valid_targets_min": 891 |
| }, |
| { |
| "epoch": 5.2, |
| "grad_norm": 0.36913473264732094, |
| "learning_rate": 7.544469147214797e-06, |
| "loss": 0.2254, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22578775882720947, |
| "step": 3250, |
| "valid_targets_mean": 5226.1, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 5.208, |
| "grad_norm": 0.40820878543241174, |
| "learning_rate": 7.482135600433868e-06, |
| "loss": 0.239, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25584137439727783, |
| "step": 3255, |
| "valid_targets_mean": 4932.1, |
| "valid_targets_min": 1008 |
| }, |
| { |
| "epoch": 5.216, |
| "grad_norm": 0.4684722429375052, |
| "learning_rate": 7.420001322125156e-06, |
| "loss": 0.2441, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24265044927597046, |
| "step": 3260, |
| "valid_targets_mean": 3623.2, |
| "valid_targets_min": 874 |
| }, |
| { |
| "epoch": 5.224, |
| "grad_norm": 0.44692307822276434, |
| "learning_rate": 7.3580673013872946e-06, |
| "loss": 0.2271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20505103468894958, |
| "step": 3265, |
| "valid_targets_mean": 3615.4, |
| "valid_targets_min": 750 |
| }, |
| { |
| "epoch": 5.232, |
| "grad_norm": 0.3654288352288813, |
| "learning_rate": 7.2963345241310904e-06, |
| "loss": 0.2253, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19597743451595306, |
| "step": 3270, |
| "valid_targets_mean": 4878.7, |
| "valid_targets_min": 867 |
| }, |
| { |
| "epoch": 5.24, |
| "grad_norm": 0.3915593147251127, |
| "learning_rate": 7.234803973063797e-06, |
| "loss": 0.226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2212725579738617, |
| "step": 3275, |
| "valid_targets_mean": 4959.9, |
| "valid_targets_min": 723 |
| }, |
| { |
| "epoch": 5.248, |
| "grad_norm": 0.4748719949243943, |
| "learning_rate": 7.173476627673492e-06, |
| "loss": 0.2375, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2473698854446411, |
| "step": 3280, |
| "valid_targets_mean": 3323.0, |
| "valid_targets_min": 1015 |
| }, |
| { |
| "epoch": 5.256, |
| "grad_norm": 0.3588612065200158, |
| "learning_rate": 7.112353464213477e-06, |
| "loss": 0.2362, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2438439577817917, |
| "step": 3285, |
| "valid_targets_mean": 6511.6, |
| "valid_targets_min": 770 |
| }, |
| { |
| "epoch": 5.264, |
| "grad_norm": 0.4663585945976981, |
| "learning_rate": 7.051435455686735e-06, |
| "loss": 0.2226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23258927464485168, |
| "step": 3290, |
| "valid_targets_mean": 3390.1, |
| "valid_targets_min": 782 |
| }, |
| { |
| "epoch": 5.272, |
| "grad_norm": 0.4205026890280819, |
| "learning_rate": 6.990723571830438e-06, |
| "loss": 0.2073, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1906987726688385, |
| "step": 3295, |
| "valid_targets_mean": 3564.7, |
| "valid_targets_min": 630 |
| }, |
| { |
| "epoch": 5.28, |
| "grad_norm": 0.426317166162649, |
| "learning_rate": 6.93021877910052e-06, |
| "loss": 0.2271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20793795585632324, |
| "step": 3300, |
| "valid_targets_mean": 3830.4, |
| "valid_targets_min": 945 |
| }, |
| { |
| "epoch": 5.288, |
| "grad_norm": 0.42620206171289865, |
| "learning_rate": 6.8699220406562985e-06, |
| "loss": 0.2272, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2456003874540329, |
| "step": 3305, |
| "valid_targets_mean": 4204.2, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 5.296, |
| "grad_norm": 0.3345250802407866, |
| "learning_rate": 6.809834316345117e-06, |
| "loss": 0.22, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20420578122138977, |
| "step": 3310, |
| "valid_targets_mean": 5562.4, |
| "valid_targets_min": 842 |
| }, |
| { |
| "epoch": 5.304, |
| "grad_norm": 0.43008399094177285, |
| "learning_rate": 6.749956562687083e-06, |
| "loss": 0.2405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19601237773895264, |
| "step": 3315, |
| "valid_targets_mean": 3619.1, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 5.312, |
| "grad_norm": 0.44306372764221147, |
| "learning_rate": 6.690289732859841e-06, |
| "loss": 0.2356, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22992250323295593, |
| "step": 3320, |
| "valid_targets_mean": 3904.2, |
| "valid_targets_min": 887 |
| }, |
| { |
| "epoch": 5.32, |
| "grad_norm": 0.349409240727027, |
| "learning_rate": 6.630834776683403e-06, |
| "loss": 0.2284, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20814573764801025, |
| "step": 3325, |
| "valid_targets_mean": 5271.5, |
| "valid_targets_min": 835 |
| }, |
| { |
| "epoch": 5.328, |
| "grad_norm": 0.4377279904884754, |
| "learning_rate": 6.571592640605e-06, |
| "loss": 0.2308, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24027813971042633, |
| "step": 3330, |
| "valid_targets_mean": 4167.5, |
| "valid_targets_min": 1091 |
| }, |
| { |
| "epoch": 5.336, |
| "grad_norm": 0.40618186734048567, |
| "learning_rate": 6.512564267684061e-06, |
| "loss": 0.2156, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21228620409965515, |
| "step": 3335, |
| "valid_targets_mean": 4030.2, |
| "valid_targets_min": 951 |
| }, |
| { |
| "epoch": 5.344, |
| "grad_norm": 0.30357992280217233, |
| "learning_rate": 6.453750597577167e-06, |
| "loss": 0.2388, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2015833854675293, |
| "step": 3340, |
| "valid_targets_mean": 6969.1, |
| "valid_targets_min": 971 |
| }, |
| { |
| "epoch": 5.352, |
| "grad_norm": 0.773767098694673, |
| "learning_rate": 6.395152566523106e-06, |
| "loss": 0.2418, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27393215894699097, |
| "step": 3345, |
| "valid_targets_mean": 1623.4, |
| "valid_targets_min": 897 |
| }, |
| { |
| "epoch": 5.36, |
| "grad_norm": 0.33428819684027417, |
| "learning_rate": 6.336771107327966e-06, |
| "loss": 0.2175, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19410398602485657, |
| "step": 3350, |
| "valid_targets_mean": 5683.6, |
| "valid_targets_min": 797 |
| }, |
| { |
| "epoch": 5.368, |
| "grad_norm": 0.45328327875390717, |
| "learning_rate": 6.278607149350289e-06, |
| "loss": 0.2259, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23492465913295746, |
| "step": 3355, |
| "valid_targets_mean": 3600.2, |
| "valid_targets_min": 841 |
| }, |
| { |
| "epoch": 5.376, |
| "grad_norm": 0.42792886228681065, |
| "learning_rate": 6.220661618486268e-06, |
| "loss": 0.2189, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21100068092346191, |
| "step": 3360, |
| "valid_targets_mean": 4031.2, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 5.384, |
| "grad_norm": 0.3557291761519541, |
| "learning_rate": 6.162935437155024e-06, |
| "loss": 0.2156, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19593030214309692, |
| "step": 3365, |
| "valid_targets_mean": 5009.3, |
| "valid_targets_min": 932 |
| }, |
| { |
| "epoch": 5.392, |
| "grad_norm": 0.4833225610526371, |
| "learning_rate": 6.105429524283901e-06, |
| "loss": 0.231, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.218770831823349, |
| "step": 3370, |
| "valid_targets_mean": 3252.6, |
| "valid_targets_min": 1030 |
| }, |
| { |
| "epoch": 5.4, |
| "grad_norm": 0.39045313632453454, |
| "learning_rate": 6.04814479529386e-06, |
| "loss": 0.228, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23024097084999084, |
| "step": 3375, |
| "valid_targets_mean": 5329.8, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 5.408, |
| "grad_norm": 0.45340949213189335, |
| "learning_rate": 5.991082162084889e-06, |
| "loss": 0.2385, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2215794175863266, |
| "step": 3380, |
| "valid_targets_mean": 3607.6, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 5.416, |
| "grad_norm": 0.4468511461282984, |
| "learning_rate": 5.934242533021499e-06, |
| "loss": 0.2203, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23574653267860413, |
| "step": 3385, |
| "valid_targets_mean": 3929.5, |
| "valid_targets_min": 850 |
| }, |
| { |
| "epoch": 5.424, |
| "grad_norm": 0.45802290494827136, |
| "learning_rate": 5.877626812918258e-06, |
| "loss": 0.2204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2201964259147644, |
| "step": 3390, |
| "valid_targets_mean": 3413.6, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 5.432, |
| "grad_norm": 0.3481392608973216, |
| "learning_rate": 5.821235903025378e-06, |
| "loss": 0.2336, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18317848443984985, |
| "step": 3395, |
| "valid_targets_mean": 4914.8, |
| "valid_targets_min": 774 |
| }, |
| { |
| "epoch": 5.44, |
| "grad_norm": 0.46295264230395117, |
| "learning_rate": 5.765070701014391e-06, |
| "loss": 0.229, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27391642332077026, |
| "step": 3400, |
| "valid_targets_mean": 3862.6, |
| "valid_targets_min": 952 |
| }, |
| { |
| "epoch": 5.448, |
| "grad_norm": 0.44931670150343195, |
| "learning_rate": 5.709132100963841e-06, |
| "loss": 0.2213, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21906086802482605, |
| "step": 3405, |
| "valid_targets_mean": 3500.1, |
| "valid_targets_min": 987 |
| }, |
| { |
| "epoch": 5.456, |
| "grad_norm": 0.34292041400051865, |
| "learning_rate": 5.653420993345062e-06, |
| "loss": 0.2155, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21244516968727112, |
| "step": 3410, |
| "valid_targets_mean": 5683.7, |
| "valid_targets_min": 887 |
| }, |
| { |
| "epoch": 5.464, |
| "grad_norm": 0.5428548488630459, |
| "learning_rate": 5.597938265007994e-06, |
| "loss": 0.2312, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23990648984909058, |
| "step": 3415, |
| "valid_targets_mean": 2892.4, |
| "valid_targets_min": 951 |
| }, |
| { |
| "epoch": 5.4719999999999995, |
| "grad_norm": 0.4706595198308062, |
| "learning_rate": 5.542684799167069e-06, |
| "loss": 0.235, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.249759703874588, |
| "step": 3420, |
| "valid_targets_mean": 3540.8, |
| "valid_targets_min": 758 |
| }, |
| { |
| "epoch": 5.48, |
| "grad_norm": 0.33282202476267, |
| "learning_rate": 5.487661475387152e-06, |
| "loss": 0.2242, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19662153720855713, |
| "step": 3425, |
| "valid_targets_mean": 5487.8, |
| "valid_targets_min": 666 |
| }, |
| { |
| "epoch": 5.4879999999999995, |
| "grad_norm": 0.4006934508283925, |
| "learning_rate": 5.432869169569541e-06, |
| "loss": 0.2188, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20843669772148132, |
| "step": 3430, |
| "valid_targets_mean": 4598.9, |
| "valid_targets_min": 997 |
| }, |
| { |
| "epoch": 5.496, |
| "grad_norm": 0.44192500022237174, |
| "learning_rate": 5.378308753938024e-06, |
| "loss": 0.2173, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25620633363723755, |
| "step": 3435, |
| "valid_targets_mean": 3980.2, |
| "valid_targets_min": 815 |
| }, |
| { |
| "epoch": 5.504, |
| "grad_norm": 0.3681471863242707, |
| "learning_rate": 5.323981097024986e-06, |
| "loss": 0.2311, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22333069145679474, |
| "step": 3440, |
| "valid_targets_mean": 4963.6, |
| "valid_targets_min": 879 |
| }, |
| { |
| "epoch": 5.5120000000000005, |
| "grad_norm": 0.356019133477824, |
| "learning_rate": 5.269887063657595e-06, |
| "loss": 0.2241, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2231229841709137, |
| "step": 3445, |
| "valid_targets_mean": 5360.4, |
| "valid_targets_min": 810 |
| }, |
| { |
| "epoch": 5.52, |
| "grad_norm": 0.36879369128157946, |
| "learning_rate": 5.216027514944027e-06, |
| "loss": 0.2192, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18978220224380493, |
| "step": 3450, |
| "valid_targets_mean": 4393.9, |
| "valid_targets_min": 674 |
| }, |
| { |
| "epoch": 5.5280000000000005, |
| "grad_norm": 0.3875347538251794, |
| "learning_rate": 5.162403308259767e-06, |
| "loss": 0.2251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.289705753326416, |
| "step": 3455, |
| "valid_targets_mean": 5434.2, |
| "valid_targets_min": 1041 |
| }, |
| { |
| "epoch": 5.536, |
| "grad_norm": 0.4285607405336305, |
| "learning_rate": 5.109015297233935e-06, |
| "loss": 0.2226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21771425008773804, |
| "step": 3460, |
| "valid_targets_mean": 3872.1, |
| "valid_targets_min": 534 |
| }, |
| { |
| "epoch": 5.5440000000000005, |
| "grad_norm": 0.32801945020364764, |
| "learning_rate": 5.055864331735736e-06, |
| "loss": 0.2627, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2591525912284851, |
| "step": 3465, |
| "valid_targets_mean": 6929.5, |
| "valid_targets_min": 1136 |
| }, |
| { |
| "epoch": 5.552, |
| "grad_norm": 0.41166443102851746, |
| "learning_rate": 5.002951257860909e-06, |
| "loss": 0.2319, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24956223368644714, |
| "step": 3470, |
| "valid_targets_mean": 4287.4, |
| "valid_targets_min": 888 |
| }, |
| { |
| "epoch": 5.5600000000000005, |
| "grad_norm": 0.4554295252372924, |
| "learning_rate": 4.950276917918256e-06, |
| "loss": 0.2328, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.256850004196167, |
| "step": 3475, |
| "valid_targets_mean": 4103.8, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 5.568, |
| "grad_norm": 0.3506746288784145, |
| "learning_rate": 4.8978421504162385e-06, |
| "loss": 0.2297, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2253202199935913, |
| "step": 3480, |
| "valid_targets_mean": 5456.3, |
| "valid_targets_min": 745 |
| }, |
| { |
| "epoch": 5.576, |
| "grad_norm": 0.33485422936124876, |
| "learning_rate": 4.845647790049634e-06, |
| "loss": 0.2278, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21637138724327087, |
| "step": 3485, |
| "valid_targets_mean": 6199.0, |
| "valid_targets_min": 982 |
| }, |
| { |
| "epoch": 5.584, |
| "grad_norm": 0.35162902696087944, |
| "learning_rate": 4.793694667686244e-06, |
| "loss": 0.2388, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24226173758506775, |
| "step": 3490, |
| "valid_targets_mean": 5727.0, |
| "valid_targets_min": 1047 |
| }, |
| { |
| "epoch": 5.592, |
| "grad_norm": 0.37080133618018396, |
| "learning_rate": 4.741983610353664e-06, |
| "loss": 0.2093, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1949070692062378, |
| "step": 3495, |
| "valid_targets_mean": 4252.1, |
| "valid_targets_min": 919 |
| }, |
| { |
| "epoch": 5.6, |
| "grad_norm": 0.3999253776520656, |
| "learning_rate": 4.690515441226122e-06, |
| "loss": 0.2123, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1985628753900528, |
| "step": 3500, |
| "valid_targets_mean": 3772.6, |
| "valid_targets_min": 733 |
| }, |
| { |
| "epoch": 5.608, |
| "grad_norm": 0.4528901320681747, |
| "learning_rate": 4.639290979611379e-06, |
| "loss": 0.2291, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23670107126235962, |
| "step": 3505, |
| "valid_targets_mean": 3763.6, |
| "valid_targets_min": 771 |
| }, |
| { |
| "epoch": 5.616, |
| "grad_norm": 0.44302870753176904, |
| "learning_rate": 4.588311040937683e-06, |
| "loss": 0.2271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24087044596672058, |
| "step": 3510, |
| "valid_targets_mean": 3990.2, |
| "valid_targets_min": 932 |
| }, |
| { |
| "epoch": 5.624, |
| "grad_norm": 0.3201486197160802, |
| "learning_rate": 4.537576436740783e-06, |
| "loss": 0.2246, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2149900496006012, |
| "step": 3515, |
| "valid_targets_mean": 6031.9, |
| "valid_targets_min": 817 |
| }, |
| { |
| "epoch": 5.632, |
| "grad_norm": 0.46436580014316425, |
| "learning_rate": 4.487087974651016e-06, |
| "loss": 0.2267, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25911879539489746, |
| "step": 3520, |
| "valid_targets_mean": 4031.8, |
| "valid_targets_min": 809 |
| }, |
| { |
| "epoch": 5.64, |
| "grad_norm": 0.525866298269229, |
| "learning_rate": 4.436846458380455e-06, |
| "loss": 0.2166, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21722733974456787, |
| "step": 3525, |
| "valid_targets_mean": 4505.7, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 5.648, |
| "grad_norm": 0.43572904873432394, |
| "learning_rate": 4.386852687710104e-06, |
| "loss": 0.2361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23047277331352234, |
| "step": 3530, |
| "valid_targets_mean": 3991.1, |
| "valid_targets_min": 790 |
| }, |
| { |
| "epoch": 5.656, |
| "grad_norm": 0.44158709967770554, |
| "learning_rate": 4.337107458477177e-06, |
| "loss": 0.2352, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21699798107147217, |
| "step": 3535, |
| "valid_targets_mean": 4153.1, |
| "valid_targets_min": 850 |
| }, |
| { |
| "epoch": 5.664, |
| "grad_norm": 0.3868164489125616, |
| "learning_rate": 4.287611562562422e-06, |
| "loss": 0.226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2065228521823883, |
| "step": 3540, |
| "valid_targets_mean": 4349.3, |
| "valid_targets_min": 587 |
| }, |
| { |
| "epoch": 5.672, |
| "grad_norm": 0.3324271889745773, |
| "learning_rate": 4.238365787877516e-06, |
| "loss": 0.2109, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2171509563922882, |
| "step": 3545, |
| "valid_targets_mean": 6169.1, |
| "valid_targets_min": 662 |
| }, |
| { |
| "epoch": 5.68, |
| "grad_norm": 0.35264056082201267, |
| "learning_rate": 4.189370918352531e-06, |
| "loss": 0.2445, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2608416676521301, |
| "step": 3550, |
| "valid_targets_mean": 5970.1, |
| "valid_targets_min": 799 |
| }, |
| { |
| "epoch": 5.688, |
| "grad_norm": 0.35571995306346826, |
| "learning_rate": 4.140627733923439e-06, |
| "loss": 0.2235, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21136212348937988, |
| "step": 3555, |
| "valid_targets_mean": 5257.9, |
| "valid_targets_min": 1041 |
| }, |
| { |
| "epoch": 5.696, |
| "grad_norm": 0.5361387745629723, |
| "learning_rate": 4.092137010519712e-06, |
| "loss": 0.2284, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27192410826683044, |
| "step": 3560, |
| "valid_targets_mean": 2914.6, |
| "valid_targets_min": 778 |
| }, |
| { |
| "epoch": 5.704, |
| "grad_norm": 0.3505859733083266, |
| "learning_rate": 4.043899520051964e-06, |
| "loss": 0.2171, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2221640646457672, |
| "step": 3565, |
| "valid_targets_mean": 5925.2, |
| "valid_targets_min": 1229 |
| }, |
| { |
| "epoch": 5.712, |
| "grad_norm": 0.49758496384659806, |
| "learning_rate": 3.995916030399658e-06, |
| "loss": 0.2272, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27344635128974915, |
| "step": 3570, |
| "valid_targets_mean": 3491.2, |
| "valid_targets_min": 991 |
| }, |
| { |
| "epoch": 5.72, |
| "grad_norm": 0.45141540037088984, |
| "learning_rate": 3.948187305398892e-06, |
| "loss": 0.2227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21061620116233826, |
| "step": 3575, |
| "valid_targets_mean": 3213.4, |
| "valid_targets_min": 744 |
| }, |
| { |
| "epoch": 5.728, |
| "grad_norm": 0.4562178096296842, |
| "learning_rate": 3.90071410483023e-06, |
| "loss": 0.2238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2696354389190674, |
| "step": 3580, |
| "valid_targets_mean": 4130.9, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 5.736, |
| "grad_norm": 0.5457573577288195, |
| "learning_rate": 3.853497184406623e-06, |
| "loss": 0.2356, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2726416289806366, |
| "step": 3585, |
| "valid_targets_mean": 2820.8, |
| "valid_targets_min": 960 |
| }, |
| { |
| "epoch": 5.744, |
| "grad_norm": 0.38517110136124666, |
| "learning_rate": 3.80653729576135e-06, |
| "loss": 0.212, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2264452874660492, |
| "step": 3590, |
| "valid_targets_mean": 4501.1, |
| "valid_targets_min": 704 |
| }, |
| { |
| "epoch": 5.752, |
| "grad_norm": 0.3357747814802782, |
| "learning_rate": 3.7598351864360872e-06, |
| "loss": 0.2315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23266029357910156, |
| "step": 3595, |
| "valid_targets_mean": 6337.8, |
| "valid_targets_min": 714 |
| }, |
| { |
| "epoch": 5.76, |
| "grad_norm": 0.5020746707062533, |
| "learning_rate": 3.713391599868985e-06, |
| "loss": 0.2274, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2000608742237091, |
| "step": 3600, |
| "valid_targets_mean": 2661.7, |
| "valid_targets_min": 791 |
| }, |
| { |
| "epoch": 5.768, |
| "grad_norm": 0.3763969776767038, |
| "learning_rate": 3.6672072753828424e-06, |
| "loss": 0.2275, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21968242526054382, |
| "step": 3605, |
| "valid_targets_mean": 5032.1, |
| "valid_targets_min": 1116 |
| }, |
| { |
| "epoch": 5.776, |
| "grad_norm": 0.41289477559493815, |
| "learning_rate": 3.6212829481733368e-06, |
| "loss": 0.2142, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2321629524230957, |
| "step": 3610, |
| "valid_targets_mean": 4038.8, |
| "valid_targets_min": 805 |
| }, |
| { |
| "epoch": 5.784, |
| "grad_norm": 0.5445568651515015, |
| "learning_rate": 3.575619349297317e-06, |
| "loss": 0.237, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25765252113342285, |
| "step": 3615, |
| "valid_targets_mean": 2809.0, |
| "valid_targets_min": 739 |
| }, |
| { |
| "epoch": 5.792, |
| "grad_norm": 0.5472578411820961, |
| "learning_rate": 3.5302172056611682e-06, |
| "loss": 0.2371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25557756423950195, |
| "step": 3620, |
| "valid_targets_mean": 2783.1, |
| "valid_targets_min": 883 |
| }, |
| { |
| "epoch": 5.8, |
| "grad_norm": 0.38982773078778155, |
| "learning_rate": 3.485077240009247e-06, |
| "loss": 0.2176, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22403252124786377, |
| "step": 3625, |
| "valid_targets_mean": 4643.8, |
| "valid_targets_min": 939 |
| }, |
| { |
| "epoch": 5.808, |
| "grad_norm": 0.5233043026636898, |
| "learning_rate": 3.4402001709123643e-06, |
| "loss": 0.2175, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24592408537864685, |
| "step": 3630, |
| "valid_targets_mean": 2974.1, |
| "valid_targets_min": 979 |
| }, |
| { |
| "epoch": 5.816, |
| "grad_norm": 0.45922405516006465, |
| "learning_rate": 3.3955867127563515e-06, |
| "loss": 0.227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24016621708869934, |
| "step": 3635, |
| "valid_targets_mean": 3544.6, |
| "valid_targets_min": 832 |
| }, |
| { |
| "epoch": 5.824, |
| "grad_norm": 0.4404664235271227, |
| "learning_rate": 3.351237575730695e-06, |
| "loss": 0.2175, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19508153200149536, |
| "step": 3640, |
| "valid_targets_mean": 3297.2, |
| "valid_targets_min": 922 |
| }, |
| { |
| "epoch": 5.832, |
| "grad_norm": 0.4375315952651967, |
| "learning_rate": 3.307153465817219e-06, |
| "loss": 0.2313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24130797386169434, |
| "step": 3645, |
| "valid_targets_mean": 4176.2, |
| "valid_targets_min": 1140 |
| }, |
| { |
| "epoch": 5.84, |
| "grad_norm": 0.342391410159785, |
| "learning_rate": 3.263335084778856e-06, |
| "loss": 0.2123, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21955269575119019, |
| "step": 3650, |
| "valid_targets_mean": 5674.6, |
| "valid_targets_min": 916 |
| }, |
| { |
| "epoch": 5.848, |
| "grad_norm": 0.45569732049327233, |
| "learning_rate": 3.2197831301484816e-06, |
| "loss": 0.2237, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23995280265808105, |
| "step": 3655, |
| "valid_targets_mean": 3858.6, |
| "valid_targets_min": 1042 |
| }, |
| { |
| "epoch": 5.856, |
| "grad_norm": 0.43878393630957296, |
| "learning_rate": 3.1764982952177805e-06, |
| "loss": 0.2281, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2358689308166504, |
| "step": 3660, |
| "valid_targets_mean": 4056.5, |
| "valid_targets_min": 815 |
| }, |
| { |
| "epoch": 5.864, |
| "grad_norm": 0.4303963408010303, |
| "learning_rate": 3.1334812690262507e-06, |
| "loss": 0.223, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.247379869222641, |
| "step": 3665, |
| "valid_targets_mean": 4625.3, |
| "valid_targets_min": 1003 |
| }, |
| { |
| "epoch": 5.872, |
| "grad_norm": 0.45709638886257675, |
| "learning_rate": 3.0907327363502084e-06, |
| "loss": 0.2256, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21598784625530243, |
| "step": 3670, |
| "valid_targets_mean": 3449.9, |
| "valid_targets_min": 1054 |
| }, |
| { |
| "epoch": 5.88, |
| "grad_norm": 0.4280225439458205, |
| "learning_rate": 3.0482533776918987e-06, |
| "loss": 0.2411, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25824448466300964, |
| "step": 3675, |
| "valid_targets_mean": 4414.6, |
| "valid_targets_min": 875 |
| }, |
| { |
| "epoch": 5.888, |
| "grad_norm": 0.3077675965104222, |
| "learning_rate": 3.0060438692686533e-06, |
| "loss": 0.2277, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1975744068622589, |
| "step": 3680, |
| "valid_targets_mean": 5987.2, |
| "valid_targets_min": 793 |
| }, |
| { |
| "epoch": 5.896, |
| "grad_norm": 0.3403603704565209, |
| "learning_rate": 2.964104883002139e-06, |
| "loss": 0.2161, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21167896687984467, |
| "step": 3685, |
| "valid_targets_mean": 5536.9, |
| "valid_targets_min": 871 |
| }, |
| { |
| "epoch": 5.904, |
| "grad_norm": 0.4054315157161449, |
| "learning_rate": 2.9224370865076457e-06, |
| "loss": 0.2213, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25502288341522217, |
| "step": 3690, |
| "valid_targets_mean": 4805.2, |
| "valid_targets_min": 822 |
| }, |
| { |
| "epoch": 5.912, |
| "grad_norm": 0.35789980590099835, |
| "learning_rate": 2.8810411430834716e-06, |
| "loss": 0.229, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19632747769355774, |
| "step": 3695, |
| "valid_targets_mean": 5478.2, |
| "valid_targets_min": 712 |
| }, |
| { |
| "epoch": 5.92, |
| "grad_norm": 0.6164021351090805, |
| "learning_rate": 2.8399177117003595e-06, |
| "loss": 0.2297, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2608773410320282, |
| "step": 3700, |
| "valid_targets_mean": 2676.1, |
| "valid_targets_min": 944 |
| }, |
| { |
| "epoch": 5.928, |
| "grad_norm": 0.45745995980330256, |
| "learning_rate": 2.7990674469910085e-06, |
| "loss": 0.2337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21105098724365234, |
| "step": 3705, |
| "valid_targets_mean": 3218.9, |
| "valid_targets_min": 639 |
| }, |
| { |
| "epoch": 5.936, |
| "grad_norm": 0.3710919492928431, |
| "learning_rate": 2.7584909992396515e-06, |
| "loss": 0.2286, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25090497732162476, |
| "step": 3710, |
| "valid_targets_mean": 5523.6, |
| "valid_targets_min": 952 |
| }, |
| { |
| "epoch": 5.944, |
| "grad_norm": 0.3323094252823238, |
| "learning_rate": 2.7181890143716995e-06, |
| "loss": 0.2288, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20140644907951355, |
| "step": 3715, |
| "valid_targets_mean": 6081.9, |
| "valid_targets_min": 1116 |
| }, |
| { |
| "epoch": 5.952, |
| "grad_norm": 0.4718363509817398, |
| "learning_rate": 2.6781621339434717e-06, |
| "loss": 0.234, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20775261521339417, |
| "step": 3720, |
| "valid_targets_mean": 3344.9, |
| "valid_targets_min": 937 |
| }, |
| { |
| "epoch": 5.96, |
| "grad_norm": 0.3964056160212036, |
| "learning_rate": 2.638410995131966e-06, |
| "loss": 0.2396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2471739947795868, |
| "step": 3725, |
| "valid_targets_mean": 4901.1, |
| "valid_targets_min": 930 |
| }, |
| { |
| "epoch": 5.968, |
| "grad_norm": 0.3725995303239783, |
| "learning_rate": 2.5989362307247313e-06, |
| "loss": 0.2127, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20092034339904785, |
| "step": 3730, |
| "valid_targets_mean": 4411.7, |
| "valid_targets_min": 770 |
| }, |
| { |
| "epoch": 5.976, |
| "grad_norm": 0.38039589975141935, |
| "learning_rate": 2.5597384691097847e-06, |
| "loss": 0.2314, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21905504167079926, |
| "step": 3735, |
| "valid_targets_mean": 4735.8, |
| "valid_targets_min": 717 |
| }, |
| { |
| "epoch": 5.984, |
| "grad_norm": 0.5842176629456968, |
| "learning_rate": 2.520818334265611e-06, |
| "loss": 0.2326, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2322233021259308, |
| "step": 3740, |
| "valid_targets_mean": 2426.8, |
| "valid_targets_min": 762 |
| }, |
| { |
| "epoch": 5.992, |
| "grad_norm": 0.3974036678792113, |
| "learning_rate": 2.482176445751232e-06, |
| "loss": 0.2361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23819592595100403, |
| "step": 3745, |
| "valid_targets_mean": 5103.4, |
| "valid_targets_min": 806 |
| }, |
| { |
| "epoch": 6.0, |
| "grad_norm": 0.5999847162525445, |
| "learning_rate": 2.4438134186963415e-06, |
| "loss": 0.2451, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26903235912323, |
| "step": 3750, |
| "valid_targets_mean": 5103.5, |
| "valid_targets_min": 786 |
| }, |
| { |
| "epoch": 6.008, |
| "grad_norm": 0.3349477545342433, |
| "learning_rate": 2.4057298637915105e-06, |
| "loss": 0.2223, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21318253874778748, |
| "step": 3755, |
| "valid_targets_mean": 6269.4, |
| "valid_targets_min": 896 |
| }, |
| { |
| "epoch": 6.016, |
| "grad_norm": 0.4317583641741853, |
| "learning_rate": 2.3679263872784717e-06, |
| "loss": 0.2476, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26914650201797485, |
| "step": 3760, |
| "valid_targets_mean": 4306.1, |
| "valid_targets_min": 810 |
| }, |
| { |
| "epoch": 6.024, |
| "grad_norm": 0.41472298145715747, |
| "learning_rate": 2.330403590940471e-06, |
| "loss": 0.2305, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24200578033924103, |
| "step": 3765, |
| "valid_targets_mean": 4969.2, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 6.032, |
| "grad_norm": 0.40539356972798346, |
| "learning_rate": 2.2931620720926717e-06, |
| "loss": 0.2365, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2599494457244873, |
| "step": 3770, |
| "valid_targets_mean": 5092.1, |
| "valid_targets_min": 1093 |
| }, |
| { |
| "epoch": 6.04, |
| "grad_norm": 0.5024307919648837, |
| "learning_rate": 2.256202423572669e-06, |
| "loss": 0.228, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24911504983901978, |
| "step": 3775, |
| "valid_targets_mean": 3418.2, |
| "valid_targets_min": 819 |
| }, |
| { |
| "epoch": 6.048, |
| "grad_norm": 0.4439148698555944, |
| "learning_rate": 2.219525233731035e-06, |
| "loss": 0.2339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22976058721542358, |
| "step": 3780, |
| "valid_targets_mean": 3652.3, |
| "valid_targets_min": 878 |
| }, |
| { |
| "epoch": 6.056, |
| "grad_norm": 0.4025990629518132, |
| "learning_rate": 2.183131086421961e-06, |
| "loss": 0.2193, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2216932326555252, |
| "step": 3785, |
| "valid_targets_mean": 4467.1, |
| "valid_targets_min": 651 |
| }, |
| { |
| "epoch": 6.064, |
| "grad_norm": 0.34119261650968813, |
| "learning_rate": 2.1470205609939533e-06, |
| "loss": 0.2269, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22316013276576996, |
| "step": 3790, |
| "valid_targets_mean": 5872.5, |
| "valid_targets_min": 743 |
| }, |
| { |
| "epoch": 6.072, |
| "grad_norm": 0.44619413022846166, |
| "learning_rate": 2.1111942322806335e-06, |
| "loss": 0.2155, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22817836701869965, |
| "step": 3795, |
| "valid_targets_mean": 3575.1, |
| "valid_targets_min": 782 |
| }, |
| { |
| "epoch": 6.08, |
| "grad_norm": 0.3574078861297622, |
| "learning_rate": 2.0756526705915635e-06, |
| "loss": 0.2151, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24410027265548706, |
| "step": 3800, |
| "valid_targets_mean": 5989.6, |
| "valid_targets_min": 1000 |
| }, |
| { |
| "epoch": 6.088, |
| "grad_norm": 0.6009272336278385, |
| "learning_rate": 2.0403964417031764e-06, |
| "loss": 0.233, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23332129418849945, |
| "step": 3805, |
| "valid_targets_mean": 2330.4, |
| "valid_targets_min": 616 |
| }, |
| { |
| "epoch": 6.096, |
| "grad_norm": 0.36144075081920424, |
| "learning_rate": 2.0054261068497773e-06, |
| "loss": 0.2198, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20843176543712616, |
| "step": 3810, |
| "valid_targets_mean": 5203.1, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 6.104, |
| "grad_norm": 0.38552202621236137, |
| "learning_rate": 1.9707422227145922e-06, |
| "loss": 0.233, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22240760922431946, |
| "step": 3815, |
| "valid_targets_mean": 4687.2, |
| "valid_targets_min": 1028 |
| }, |
| { |
| "epoch": 6.112, |
| "grad_norm": 0.41879689644483564, |
| "learning_rate": 1.936345341420924e-06, |
| "loss": 0.24, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2170860469341278, |
| "step": 3820, |
| "valid_targets_mean": 4666.2, |
| "valid_targets_min": 1176 |
| }, |
| { |
| "epoch": 6.12, |
| "grad_norm": 0.6687266732348552, |
| "learning_rate": 1.9022360105233507e-06, |
| "loss": 0.2228, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2556125521659851, |
| "step": 3825, |
| "valid_targets_mean": 2104.1, |
| "valid_targets_min": 1037 |
| }, |
| { |
| "epoch": 6.128, |
| "grad_norm": 0.34492206038582207, |
| "learning_rate": 1.8684147729990188e-06, |
| "loss": 0.2382, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2455054074525833, |
| "step": 3830, |
| "valid_targets_mean": 6336.3, |
| "valid_targets_min": 1050 |
| }, |
| { |
| "epoch": 6.136, |
| "grad_norm": 0.4159150879186752, |
| "learning_rate": 1.8348821672389893e-06, |
| "loss": 0.2252, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.225693017244339, |
| "step": 3835, |
| "valid_targets_mean": 4592.3, |
| "valid_targets_min": 1017 |
| }, |
| { |
| "epoch": 6.144, |
| "grad_norm": 0.38859249778164073, |
| "learning_rate": 1.8016387270396784e-06, |
| "loss": 0.2413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23034977912902832, |
| "step": 3840, |
| "valid_targets_mean": 5167.3, |
| "valid_targets_min": 881 |
| }, |
| { |
| "epoch": 6.152, |
| "grad_norm": 0.3203760106705756, |
| "learning_rate": 1.7686849815943486e-06, |
| "loss": 0.2198, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21037934720516205, |
| "step": 3845, |
| "valid_targets_mean": 6393.3, |
| "valid_targets_min": 919 |
| }, |
| { |
| "epoch": 6.16, |
| "grad_norm": 0.3983227918706389, |
| "learning_rate": 1.7360214554847e-06, |
| "loss": 0.2147, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2141357660293579, |
| "step": 3850, |
| "valid_targets_mean": 4325.4, |
| "valid_targets_min": 920 |
| }, |
| { |
| "epoch": 6.168, |
| "grad_norm": 0.3289919821089591, |
| "learning_rate": 1.703648668672495e-06, |
| "loss": 0.2274, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21512682735919952, |
| "step": 3855, |
| "valid_targets_mean": 6087.9, |
| "valid_targets_min": 903 |
| }, |
| { |
| "epoch": 6.176, |
| "grad_norm": 0.43853553740948986, |
| "learning_rate": 1.6715671364913077e-06, |
| "loss": 0.2452, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21846427023410797, |
| "step": 3860, |
| "valid_targets_mean": 3917.2, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 6.184, |
| "grad_norm": 0.38127578812735524, |
| "learning_rate": 1.6397773696383091e-06, |
| "loss": 0.2136, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21105417609214783, |
| "step": 3865, |
| "valid_targets_mean": 4652.4, |
| "valid_targets_min": 787 |
| }, |
| { |
| "epoch": 6.192, |
| "grad_norm": 0.3396436270940679, |
| "learning_rate": 1.6082798741661321e-06, |
| "loss": 0.2313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20317316055297852, |
| "step": 3870, |
| "valid_targets_mean": 5430.1, |
| "valid_targets_min": 893 |
| }, |
| { |
| "epoch": 6.2, |
| "grad_norm": 0.38228587817606946, |
| "learning_rate": 1.5770751514748273e-06, |
| "loss": 0.2181, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21367615461349487, |
| "step": 3875, |
| "valid_targets_mean": 4672.6, |
| "valid_targets_min": 1012 |
| }, |
| { |
| "epoch": 6.208, |
| "grad_norm": 0.47820322237336815, |
| "learning_rate": 1.5461636983038686e-06, |
| "loss": 0.2155, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21911242604255676, |
| "step": 3880, |
| "valid_targets_mean": 3371.2, |
| "valid_targets_min": 997 |
| }, |
| { |
| "epoch": 6.216, |
| "grad_norm": 0.4399583273624092, |
| "learning_rate": 1.5155460067242578e-06, |
| "loss": 0.2418, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23615899682044983, |
| "step": 3885, |
| "valid_targets_mean": 4043.5, |
| "valid_targets_min": 745 |
| }, |
| { |
| "epoch": 6.224, |
| "grad_norm": 0.3398925492873628, |
| "learning_rate": 1.4852225641306816e-06, |
| "loss": 0.2255, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2137679159641266, |
| "step": 3890, |
| "valid_targets_mean": 5939.5, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 6.232, |
| "grad_norm": 0.4733714407702554, |
| "learning_rate": 1.4551938532337607e-06, |
| "loss": 0.2363, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2328024059534073, |
| "step": 3895, |
| "valid_targets_mean": 3431.8, |
| "valid_targets_min": 972 |
| }, |
| { |
| "epoch": 6.24, |
| "grad_norm": 0.561178279992971, |
| "learning_rate": 1.4254603520523614e-06, |
| "loss": 0.2294, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21961773931980133, |
| "step": 3900, |
| "valid_targets_mean": 2525.5, |
| "valid_targets_min": 969 |
| }, |
| { |
| "epoch": 6.248, |
| "grad_norm": 0.4741052980704381, |
| "learning_rate": 1.3960225339059875e-06, |
| "loss": 0.2233, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23817236721515656, |
| "step": 3905, |
| "valid_targets_mean": 4482.2, |
| "valid_targets_min": 693 |
| }, |
| { |
| "epoch": 6.256, |
| "grad_norm": 0.4411532668736603, |
| "learning_rate": 1.3668808674072409e-06, |
| "loss": 0.2242, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2441355437040329, |
| "step": 3910, |
| "valid_targets_mean": 3981.3, |
| "valid_targets_min": 984 |
| }, |
| { |
| "epoch": 6.264, |
| "grad_norm": 0.4003229501442681, |
| "learning_rate": 1.338035816454375e-06, |
| "loss": 0.2119, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21280062198638916, |
| "step": 3915, |
| "valid_targets_mean": 4549.2, |
| "valid_targets_min": 1135 |
| }, |
| { |
| "epoch": 6.272, |
| "grad_norm": 0.4571069597331237, |
| "learning_rate": 1.3094878402238887e-06, |
| "loss": 0.2265, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2197241187095642, |
| "step": 3920, |
| "valid_targets_mean": 3564.8, |
| "valid_targets_min": 762 |
| }, |
| { |
| "epoch": 6.28, |
| "grad_norm": 0.36182902012131324, |
| "learning_rate": 1.2812373931632371e-06, |
| "loss": 0.2133, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23633159697055817, |
| "step": 3925, |
| "valid_targets_mean": 5213.5, |
| "valid_targets_min": 938 |
| }, |
| { |
| "epoch": 6.288, |
| "grad_norm": 0.34785431888006046, |
| "learning_rate": 1.2532849249835932e-06, |
| "loss": 0.2183, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20579376816749573, |
| "step": 3930, |
| "valid_targets_mean": 5596.8, |
| "valid_targets_min": 776 |
| }, |
| { |
| "epoch": 6.296, |
| "grad_norm": 0.4114868588415092, |
| "learning_rate": 1.2256308806526774e-06, |
| "loss": 0.2244, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22689789533615112, |
| "step": 3935, |
| "valid_targets_mean": 4300.3, |
| "valid_targets_min": 813 |
| }, |
| { |
| "epoch": 6.304, |
| "grad_norm": 0.39604646648569675, |
| "learning_rate": 1.1982757003876855e-06, |
| "loss": 0.2337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24964919686317444, |
| "step": 3940, |
| "valid_targets_mean": 5153.9, |
| "valid_targets_min": 711 |
| }, |
| { |
| "epoch": 6.312, |
| "grad_norm": 0.3849772647568709, |
| "learning_rate": 1.1712198196482793e-06, |
| "loss": 0.224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26009365916252136, |
| "step": 3945, |
| "valid_targets_mean": 5058.6, |
| "valid_targets_min": 980 |
| }, |
| { |
| "epoch": 6.32, |
| "grad_norm": 0.48211470656475425, |
| "learning_rate": 1.1444636691296518e-06, |
| "loss": 0.2226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21652325987815857, |
| "step": 3950, |
| "valid_targets_mean": 3130.9, |
| "valid_targets_min": 798 |
| }, |
| { |
| "epoch": 6.328, |
| "grad_norm": 0.5644365548848527, |
| "learning_rate": 1.11800767475567e-06, |
| "loss": 0.246, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27526530623435974, |
| "step": 3955, |
| "valid_targets_mean": 2978.4, |
| "valid_targets_min": 849 |
| }, |
| { |
| "epoch": 6.336, |
| "grad_norm": 0.40333014592575334, |
| "learning_rate": 1.0918522576721014e-06, |
| "loss": 0.228, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27383342385292053, |
| "step": 3960, |
| "valid_targets_mean": 5273.9, |
| "valid_targets_min": 630 |
| }, |
| { |
| "epoch": 6.344, |
| "grad_norm": 0.4321759599915463, |
| "learning_rate": 1.0659978342399003e-06, |
| "loss": 0.2148, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24329343438148499, |
| "step": 3965, |
| "valid_targets_mean": 4204.4, |
| "valid_targets_min": 827 |
| }, |
| { |
| "epoch": 6.352, |
| "grad_norm": 0.4489831786268062, |
| "learning_rate": 1.0404448160285897e-06, |
| "loss": 0.2281, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20862233638763428, |
| "step": 3970, |
| "valid_targets_mean": 3446.6, |
| "valid_targets_min": 936 |
| }, |
| { |
| "epoch": 6.36, |
| "grad_norm": 0.6568484958919774, |
| "learning_rate": 1.0151936098097015e-06, |
| "loss": 0.2307, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2893434762954712, |
| "step": 3975, |
| "valid_targets_mean": 2431.0, |
| "valid_targets_min": 865 |
| }, |
| { |
| "epoch": 6.368, |
| "grad_norm": 0.436806643140305, |
| "learning_rate": 9.902446175503089e-07, |
| "loss": 0.2075, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21049389243125916, |
| "step": 3980, |
| "valid_targets_mean": 3591.6, |
| "valid_targets_min": 558 |
| }, |
| { |
| "epoch": 6.376, |
| "grad_norm": 0.3733718504908191, |
| "learning_rate": 9.655982364066197e-07, |
| "loss": 0.2208, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19698584079742432, |
| "step": 3985, |
| "valid_targets_mean": 4745.1, |
| "valid_targets_min": 1147 |
| }, |
| { |
| "epoch": 6.384, |
| "grad_norm": 0.56021488352525, |
| "learning_rate": 9.412548587176595e-07, |
| "loss": 0.2313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21981269121170044, |
| "step": 3990, |
| "valid_targets_mean": 2582.5, |
| "valid_targets_min": 679 |
| }, |
| { |
| "epoch": 6.392, |
| "grad_norm": 0.4255541904342452, |
| "learning_rate": 9.172148719990237e-07, |
| "loss": 0.2134, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23671914637088776, |
| "step": 3995, |
| "valid_targets_mean": 4313.3, |
| "valid_targets_min": 752 |
| }, |
| { |
| "epoch": 6.4, |
| "grad_norm": 0.4102539193396227, |
| "learning_rate": 8.934786589367106e-07, |
| "loss": 0.2198, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2218114733695984, |
| "step": 4000, |
| "valid_targets_mean": 4082.2, |
| "valid_targets_min": 787 |
| }, |
| { |
| "epoch": 6.408, |
| "grad_norm": 0.34989430636633106, |
| "learning_rate": 8.700465973810246e-07, |
| "loss": 0.2215, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1899658888578415, |
| "step": 4005, |
| "valid_targets_mean": 4683.2, |
| "valid_targets_min": 963 |
| }, |
| { |
| "epoch": 6.416, |
| "grad_norm": 0.5084082139951963, |
| "learning_rate": 8.469190603405719e-07, |
| "loss": 0.2362, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22666466236114502, |
| "step": 4010, |
| "valid_targets_mean": 2856.0, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 6.424, |
| "grad_norm": 0.38334389778158073, |
| "learning_rate": 8.240964159763121e-07, |
| "loss": 0.2276, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20574325323104858, |
| "step": 4015, |
| "valid_targets_mean": 4797.1, |
| "valid_targets_min": 989 |
| }, |
| { |
| "epoch": 6.432, |
| "grad_norm": 0.43778396216540555, |
| "learning_rate": 8.015790275957003e-07, |
| "loss": 0.2379, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2721363604068756, |
| "step": 4020, |
| "valid_targets_mean": 4351.1, |
| "valid_targets_min": 1141 |
| }, |
| { |
| "epoch": 6.44, |
| "grad_norm": 0.42691802245050525, |
| "learning_rate": 7.793672536469077e-07, |
| "loss": 0.2179, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.236424520611763, |
| "step": 4025, |
| "valid_targets_mean": 3813.6, |
| "valid_targets_min": 820 |
| }, |
| { |
| "epoch": 6.448, |
| "grad_norm": 0.36072477649308493, |
| "learning_rate": 7.574614477131081e-07, |
| "loss": 0.2267, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18866382539272308, |
| "step": 4030, |
| "valid_targets_mean": 4367.6, |
| "valid_targets_min": 843 |
| }, |
| { |
| "epoch": 6.456, |
| "grad_norm": 0.41123091937206413, |
| "learning_rate": 7.358619585068583e-07, |
| "loss": 0.2313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23328976333141327, |
| "step": 4035, |
| "valid_targets_mean": 4417.2, |
| "valid_targets_min": 939 |
| }, |
| { |
| "epoch": 6.464, |
| "grad_norm": 0.4254203784811884, |
| "learning_rate": 7.145691298645419e-07, |
| "loss": 0.2186, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2179802656173706, |
| "step": 4040, |
| "valid_targets_mean": 3939.2, |
| "valid_targets_min": 946 |
| }, |
| { |
| "epoch": 6.4719999999999995, |
| "grad_norm": 0.39330349529070346, |
| "learning_rate": 6.935833007408965e-07, |
| "loss": 0.2158, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22284376621246338, |
| "step": 4045, |
| "valid_targets_mean": 4508.6, |
| "valid_targets_min": 623 |
| }, |
| { |
| "epoch": 6.48, |
| "grad_norm": 0.44865937037871895, |
| "learning_rate": 6.729048052036136e-07, |
| "loss": 0.2339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21400213241577148, |
| "step": 4050, |
| "valid_targets_mean": 3500.1, |
| "valid_targets_min": 858 |
| }, |
| { |
| "epoch": 6.4879999999999995, |
| "grad_norm": 0.48021795524832595, |
| "learning_rate": 6.52533972428031e-07, |
| "loss": 0.2178, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24885594844818115, |
| "step": 4055, |
| "valid_targets_mean": 3483.8, |
| "valid_targets_min": 971 |
| }, |
| { |
| "epoch": 6.496, |
| "grad_norm": 0.34882215560284985, |
| "learning_rate": 6.324711266918826e-07, |
| "loss": 0.2136, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21410886943340302, |
| "step": 4060, |
| "valid_targets_mean": 5634.1, |
| "valid_targets_min": 991 |
| }, |
| { |
| "epoch": 6.504, |
| "grad_norm": 0.5126446488160206, |
| "learning_rate": 6.127165873701457e-07, |
| "loss": 0.2081, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21240456402301788, |
| "step": 4065, |
| "valid_targets_mean": 2864.9, |
| "valid_targets_min": 985 |
| }, |
| { |
| "epoch": 6.5120000000000005, |
| "grad_norm": 0.3140688580026172, |
| "learning_rate": 5.932706689299461e-07, |
| "loss": 0.2239, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21024207770824432, |
| "step": 4070, |
| "valid_targets_mean": 6588.2, |
| "valid_targets_min": 1101 |
| }, |
| { |
| "epoch": 6.52, |
| "grad_norm": 0.4986798403758234, |
| "learning_rate": 5.741336809255615e-07, |
| "loss": 0.2246, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22286343574523926, |
| "step": 4075, |
| "valid_targets_mean": 3069.7, |
| "valid_targets_min": 1026 |
| }, |
| { |
| "epoch": 6.5280000000000005, |
| "grad_norm": 0.3921933421532105, |
| "learning_rate": 5.553059279934902e-07, |
| "loss": 0.2241, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2469090223312378, |
| "step": 4080, |
| "valid_targets_mean": 4846.6, |
| "valid_targets_min": 741 |
| }, |
| { |
| "epoch": 6.536, |
| "grad_norm": 1.4197403594404054, |
| "learning_rate": 5.36787709847597e-07, |
| "loss": 0.218, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21249955892562866, |
| "step": 4085, |
| "valid_targets_mean": 3518.8, |
| "valid_targets_min": 947 |
| }, |
| { |
| "epoch": 6.5440000000000005, |
| "grad_norm": 0.39790011129897357, |
| "learning_rate": 5.185793212743529e-07, |
| "loss": 0.225, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2455723136663437, |
| "step": 4090, |
| "valid_targets_mean": 4974.9, |
| "valid_targets_min": 833 |
| }, |
| { |
| "epoch": 6.552, |
| "grad_norm": 0.506846713782852, |
| "learning_rate": 5.006810521281335e-07, |
| "loss": 0.2219, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22944439947605133, |
| "step": 4095, |
| "valid_targets_mean": 3162.4, |
| "valid_targets_min": 889 |
| }, |
| { |
| "epoch": 6.5600000000000005, |
| "grad_norm": 0.5304911335607064, |
| "learning_rate": 4.830931873266065e-07, |
| "loss": 0.2261, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.228239044547081, |
| "step": 4100, |
| "valid_targets_mean": 2786.3, |
| "valid_targets_min": 655 |
| }, |
| { |
| "epoch": 6.568, |
| "grad_norm": 0.42229465032601937, |
| "learning_rate": 4.658160068462025e-07, |
| "loss": 0.2342, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22529050707817078, |
| "step": 4105, |
| "valid_targets_mean": 4120.8, |
| "valid_targets_min": 924 |
| }, |
| { |
| "epoch": 6.576, |
| "grad_norm": 0.32368728463004987, |
| "learning_rate": 4.488497857176466e-07, |
| "loss": 0.2132, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21758395433425903, |
| "step": 4110, |
| "valid_targets_mean": 6052.4, |
| "valid_targets_min": 890 |
| }, |
| { |
| "epoch": 6.584, |
| "grad_norm": 0.3876726339635795, |
| "learning_rate": 4.321947940215898e-07, |
| "loss": 0.2271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27703386545181274, |
| "step": 4115, |
| "valid_targets_mean": 5807.9, |
| "valid_targets_min": 955 |
| }, |
| { |
| "epoch": 6.592, |
| "grad_norm": 0.5109155715461828, |
| "learning_rate": 4.1585129688430425e-07, |
| "loss": 0.2481, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2829819321632385, |
| "step": 4120, |
| "valid_targets_mean": 3333.1, |
| "valid_targets_min": 867 |
| }, |
| { |
| "epoch": 6.6, |
| "grad_norm": 0.4470302028073575, |
| "learning_rate": 3.998195544734706e-07, |
| "loss": 0.2337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24119625985622406, |
| "step": 4125, |
| "valid_targets_mean": 4098.7, |
| "valid_targets_min": 1055 |
| }, |
| { |
| "epoch": 6.608, |
| "grad_norm": 0.3879001559853112, |
| "learning_rate": 3.840998219940284e-07, |
| "loss": 0.2266, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23125238716602325, |
| "step": 4130, |
| "valid_targets_mean": 4793.9, |
| "valid_targets_min": 995 |
| }, |
| { |
| "epoch": 6.616, |
| "grad_norm": 0.3390815157643552, |
| "learning_rate": 3.6869234968411214e-07, |
| "loss": 0.2191, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21983827650547028, |
| "step": 4135, |
| "valid_targets_mean": 5686.7, |
| "valid_targets_min": 931 |
| }, |
| { |
| "epoch": 6.624, |
| "grad_norm": 0.48742691434984764, |
| "learning_rate": 3.5359738281107504e-07, |
| "loss": 0.2315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23688867688179016, |
| "step": 4140, |
| "valid_targets_mean": 3576.5, |
| "valid_targets_min": 877 |
| }, |
| { |
| "epoch": 6.632, |
| "grad_norm": 0.5825891651795995, |
| "learning_rate": 3.38815161667585e-07, |
| "loss": 0.2227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2347678393125534, |
| "step": 4145, |
| "valid_targets_mean": 2693.6, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 6.64, |
| "grad_norm": 0.33240965337871686, |
| "learning_rate": 3.24345921567788e-07, |
| "loss": 0.2247, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22942331433296204, |
| "step": 4150, |
| "valid_targets_mean": 6286.4, |
| "valid_targets_min": 1070 |
| }, |
| { |
| "epoch": 6.648, |
| "grad_norm": 0.47442097754914475, |
| "learning_rate": 3.101898928435754e-07, |
| "loss": 0.2112, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2143860161304474, |
| "step": 4155, |
| "valid_targets_mean": 2954.4, |
| "valid_targets_min": 1049 |
| }, |
| { |
| "epoch": 6.656, |
| "grad_norm": 0.47947322659482045, |
| "learning_rate": 2.9634730084091343e-07, |
| "loss": 0.2222, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22327253222465515, |
| "step": 4160, |
| "valid_targets_mean": 3591.6, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 6.664, |
| "grad_norm": 0.4454312428655186, |
| "learning_rate": 2.8281836591624865e-07, |
| "loss": 0.2137, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22166194021701813, |
| "step": 4165, |
| "valid_targets_mean": 3687.9, |
| "valid_targets_min": 710 |
| }, |
| { |
| "epoch": 6.672, |
| "grad_norm": 0.4171522266163402, |
| "learning_rate": 2.6960330343301033e-07, |
| "loss": 0.2166, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23376452922821045, |
| "step": 4170, |
| "valid_targets_mean": 4033.3, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 6.68, |
| "grad_norm": 0.4021870631095099, |
| "learning_rate": 2.5670232375817784e-07, |
| "loss": 0.2115, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19755598902702332, |
| "step": 4175, |
| "valid_targets_mean": 4376.8, |
| "valid_targets_min": 690 |
| }, |
| { |
| "epoch": 6.688, |
| "grad_norm": 0.4432006157292327, |
| "learning_rate": 2.441156322589322e-07, |
| "loss": 0.2094, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2194088101387024, |
| "step": 4180, |
| "valid_targets_mean": 3696.2, |
| "valid_targets_min": 937 |
| }, |
| { |
| "epoch": 6.696, |
| "grad_norm": 0.3533442609966585, |
| "learning_rate": 2.318434292993832e-07, |
| "loss": 0.2187, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22602006793022156, |
| "step": 4185, |
| "valid_targets_mean": 5390.4, |
| "valid_targets_min": 952 |
| }, |
| { |
| "epoch": 6.704, |
| "grad_norm": 0.5451719228606399, |
| "learning_rate": 2.1988591023738514e-07, |
| "loss": 0.2148, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23980574309825897, |
| "step": 4190, |
| "valid_targets_mean": 2911.3, |
| "valid_targets_min": 756 |
| }, |
| { |
| "epoch": 6.712, |
| "grad_norm": 0.3356047849602916, |
| "learning_rate": 2.0824326542142835e-07, |
| "loss": 0.213, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18571820855140686, |
| "step": 4195, |
| "valid_targets_mean": 5152.4, |
| "valid_targets_min": 1057 |
| }, |
| { |
| "epoch": 6.72, |
| "grad_norm": 0.5229315558444865, |
| "learning_rate": 1.9691568018759931e-07, |
| "loss": 0.233, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.277045339345932, |
| "step": 4200, |
| "valid_targets_mean": 3369.2, |
| "valid_targets_min": 805 |
| }, |
| { |
| "epoch": 6.728, |
| "grad_norm": 0.3325315865260218, |
| "learning_rate": 1.8590333485664525e-07, |
| "loss": 0.2141, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2092788815498352, |
| "step": 4205, |
| "valid_targets_mean": 5736.8, |
| "valid_targets_min": 972 |
| }, |
| { |
| "epoch": 6.736, |
| "grad_norm": 0.467439774714686, |
| "learning_rate": 1.752064047310853e-07, |
| "loss": 0.2197, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23705033957958221, |
| "step": 4210, |
| "valid_targets_mean": 3690.5, |
| "valid_targets_min": 852 |
| }, |
| { |
| "epoch": 6.744, |
| "grad_norm": 0.6183840075760885, |
| "learning_rate": 1.6482506009243949e-07, |
| "loss": 0.2336, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22905892133712769, |
| "step": 4215, |
| "valid_targets_mean": 2088.2, |
| "valid_targets_min": 810 |
| }, |
| { |
| "epoch": 6.752, |
| "grad_norm": 0.3342422145961255, |
| "learning_rate": 1.5475946619850192e-07, |
| "loss": 0.2206, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20969925820827484, |
| "step": 4220, |
| "valid_targets_mean": 5836.8, |
| "valid_targets_min": 1020 |
| }, |
| { |
| "epoch": 6.76, |
| "grad_norm": 0.34624668633463707, |
| "learning_rate": 1.4500978328071845e-07, |
| "loss": 0.2122, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22905424237251282, |
| "step": 4225, |
| "valid_targets_mean": 5463.2, |
| "valid_targets_min": 732 |
| }, |
| { |
| "epoch": 6.768, |
| "grad_norm": 0.4420870260003288, |
| "learning_rate": 1.3557616654163775e-07, |
| "loss": 0.2302, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.223080113530159, |
| "step": 4230, |
| "valid_targets_mean": 3664.9, |
| "valid_targets_min": 826 |
| }, |
| { |
| "epoch": 6.776, |
| "grad_norm": 0.5518275111575168, |
| "learning_rate": 1.264587661524308e-07, |
| "loss": 0.2196, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22578661143779755, |
| "step": 4235, |
| "valid_targets_mean": 2675.6, |
| "valid_targets_min": 966 |
| }, |
| { |
| "epoch": 6.784, |
| "grad_norm": 0.4049368287917327, |
| "learning_rate": 1.1765772725051084e-07, |
| "loss": 0.2131, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21148250997066498, |
| "step": 4240, |
| "valid_targets_mean": 4096.6, |
| "valid_targets_min": 877 |
| }, |
| { |
| "epoch": 6.792, |
| "grad_norm": 0.4040442168660199, |
| "learning_rate": 1.0917318993721726e-07, |
| "loss": 0.2192, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21757608652114868, |
| "step": 4245, |
| "valid_targets_mean": 4106.6, |
| "valid_targets_min": 657 |
| }, |
| { |
| "epoch": 6.8, |
| "grad_norm": 0.4656238469376599, |
| "learning_rate": 1.0100528927558861e-07, |
| "loss": 0.2405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20843830704689026, |
| "step": 4250, |
| "valid_targets_mean": 3208.7, |
| "valid_targets_min": 952 |
| }, |
| { |
| "epoch": 6.808, |
| "grad_norm": 0.42442223312996796, |
| "learning_rate": 9.31541552882087e-08, |
| "loss": 0.2299, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25054430961608887, |
| "step": 4255, |
| "valid_targets_mean": 4431.4, |
| "valid_targets_min": 932 |
| }, |
| { |
| "epoch": 6.816, |
| "grad_norm": 0.5254292888858535, |
| "learning_rate": 8.561991295514161e-08, |
| "loss": 0.2288, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23713023960590363, |
| "step": 4260, |
| "valid_targets_mean": 3043.6, |
| "valid_targets_min": 793 |
| }, |
| { |
| "epoch": 6.824, |
| "grad_norm": 0.4371339395015146, |
| "learning_rate": 7.840268221193548e-08, |
| "loss": 0.2259, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23466908931732178, |
| "step": 4265, |
| "valid_targets_mean": 3946.0, |
| "valid_targets_min": 874 |
| }, |
| { |
| "epoch": 6.832, |
| "grad_norm": 0.35973979810814993, |
| "learning_rate": 7.150257794772186e-08, |
| "loss": 0.2068, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21603873372077942, |
| "step": 4270, |
| "valid_targets_mean": 5114.9, |
| "valid_targets_min": 1057 |
| }, |
| { |
| "epoch": 6.84, |
| "grad_norm": 0.3711213623529721, |
| "learning_rate": 6.491971000337938e-08, |
| "loss": 0.2251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21824629604816437, |
| "step": 4275, |
| "valid_targets_mean": 4986.2, |
| "valid_targets_min": 934 |
| }, |
| { |
| "epoch": 6.848, |
| "grad_norm": 0.4906868870974589, |
| "learning_rate": 5.8654183169788435e-08, |
| "loss": 0.2244, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23453839123249054, |
| "step": 4280, |
| "valid_targets_mean": 3274.4, |
| "valid_targets_min": 1037 |
| }, |
| { |
| "epoch": 6.856, |
| "grad_norm": 0.5003819660470424, |
| "learning_rate": 5.270609718616593e-08, |
| "loss": 0.2243, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23117989301681519, |
| "step": 4285, |
| "valid_targets_mean": 3046.3, |
| "valid_targets_min": 887 |
| }, |
| { |
| "epoch": 6.864, |
| "grad_norm": 0.5538836489378799, |
| "learning_rate": 4.70755467384687e-08, |
| "loss": 0.2238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22990688681602478, |
| "step": 4290, |
| "valid_targets_mean": 2669.9, |
| "valid_targets_min": 722 |
| }, |
| { |
| "epoch": 6.872, |
| "grad_norm": 0.3339220377327871, |
| "learning_rate": 4.176262145789478e-08, |
| "loss": 0.2123, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19378703832626343, |
| "step": 4295, |
| "valid_targets_mean": 5727.6, |
| "valid_targets_min": 1068 |
| }, |
| { |
| "epoch": 6.88, |
| "grad_norm": 0.36779708232161507, |
| "learning_rate": 3.676740591945782e-08, |
| "loss": 0.2218, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2007201462984085, |
| "step": 4300, |
| "valid_targets_mean": 4660.1, |
| "valid_targets_min": 587 |
| }, |
| { |
| "epoch": 6.888, |
| "grad_norm": 0.416473109709361, |
| "learning_rate": 3.208997964062821e-08, |
| "loss": 0.2258, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23714400827884674, |
| "step": 4305, |
| "valid_targets_mean": 4501.4, |
| "valid_targets_min": 1129 |
| }, |
| { |
| "epoch": 6.896, |
| "grad_norm": 0.3816242071018597, |
| "learning_rate": 2.773041708008295e-08, |
| "loss": 0.2202, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2064952254295349, |
| "step": 4310, |
| "valid_targets_mean": 4667.4, |
| "valid_targets_min": 712 |
| }, |
| { |
| "epoch": 6.904, |
| "grad_norm": 0.3988685178845027, |
| "learning_rate": 2.3688787636511057e-08, |
| "loss": 0.219, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2150478959083557, |
| "step": 4315, |
| "valid_targets_mean": 3976.9, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 6.912, |
| "grad_norm": 0.4016474848255541, |
| "learning_rate": 1.9965155647507782e-08, |
| "loss": 0.2295, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2176404893398285, |
| "step": 4320, |
| "valid_targets_mean": 4303.1, |
| "valid_targets_min": 921 |
| }, |
| { |
| "epoch": 6.92, |
| "grad_norm": 0.37224492037629603, |
| "learning_rate": 1.655958038855765e-08, |
| "loss": 0.2221, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19105198979377747, |
| "step": 4325, |
| "valid_targets_mean": 4631.8, |
| "valid_targets_min": 1050 |
| }, |
| { |
| "epoch": 6.928, |
| "grad_norm": 0.4229429441256482, |
| "learning_rate": 1.3472116072084096e-08, |
| "loss": 0.221, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22886265814304352, |
| "step": 4330, |
| "valid_targets_mean": 4466.5, |
| "valid_targets_min": 1187 |
| }, |
| { |
| "epoch": 6.936, |
| "grad_norm": 0.3238044486735016, |
| "learning_rate": 1.0702811846590167e-08, |
| "loss": 0.2135, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20526066422462463, |
| "step": 4335, |
| "valid_targets_mean": 6032.8, |
| "valid_targets_min": 770 |
| }, |
| { |
| "epoch": 6.944, |
| "grad_norm": 0.405352461983432, |
| "learning_rate": 8.251711795876916e-09, |
| "loss": 0.2273, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20454303920269012, |
| "step": 4340, |
| "valid_targets_mean": 4101.2, |
| "valid_targets_min": 1081 |
| }, |
| { |
| "epoch": 6.952, |
| "grad_norm": 0.5254957752257481, |
| "learning_rate": 6.1188549383373044e-09, |
| "loss": 0.2125, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20966967940330505, |
| "step": 4345, |
| "valid_targets_mean": 2603.4, |
| "valid_targets_min": 856 |
| }, |
| { |
| "epoch": 6.96, |
| "grad_norm": 0.4987226166041217, |
| "learning_rate": 4.304275226338916e-09, |
| "loss": 0.2181, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20376208424568176, |
| "step": 4350, |
| "valid_targets_mean": 3057.9, |
| "valid_targets_min": 852 |
| }, |
| { |
| "epoch": 6.968, |
| "grad_norm": 0.4147630934098896, |
| "learning_rate": 2.8080015456799503e-09, |
| "loss": 0.2128, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22926411032676697, |
| "step": 4355, |
| "valid_targets_mean": 4151.9, |
| "valid_targets_min": 662 |
| }, |
| { |
| "epoch": 6.976, |
| "grad_norm": 0.4327182545700865, |
| "learning_rate": 1.6300577151340257e-09, |
| "loss": 0.2168, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21291124820709229, |
| "step": 4360, |
| "valid_targets_mean": 3751.6, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 6.984, |
| "grad_norm": 0.3961055234395137, |
| "learning_rate": 7.70462486070489e-10, |
| "loss": 0.2178, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21084696054458618, |
| "step": 4365, |
| "valid_targets_mean": 4619.8, |
| "valid_targets_min": 988 |
| }, |
| { |
| "epoch": 6.992, |
| "grad_norm": 0.4419544854529956, |
| "learning_rate": 2.2922954214799065e-10, |
| "loss": 0.2288, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23672346770763397, |
| "step": 4370, |
| "valid_targets_mean": 3912.4, |
| "valid_targets_min": 955 |
| }, |
| { |
| "epoch": 7.0, |
| "grad_norm": 0.4194586140813968, |
| "learning_rate": 6.367499107984288e-12, |
| "loss": 0.2104, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24523112177848816, |
| "step": 4375, |
| "valid_targets_mean": 4153.5, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 7.0, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24523112177848816, |
| "step": 4375, |
| "total_flos": 785288333426688.0, |
| "train_loss": 0.2574095404761178, |
| "train_runtime": 17825.8409, |
| "train_samples_per_second": 3.926, |
| "train_steps_per_second": 0.245, |
| "valid_targets_mean": 4153.5, |
| "valid_targets_min": 836 |
| } |
| ], |
| "logging_steps": 5, |
| "max_steps": 4375, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 7, |
| "save_steps": 1500, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": true |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 785288333426688.0, |
| "train_batch_size": 1, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
