{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 785,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.032,
      "grad_norm": 5.01998705221364,
      "learning_rate": 2.0253164556962026e-06,
      "loss": 0.7079,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.16147100925445557,
      "step": 5,
      "valid_targets_mean": 4128.0,
      "valid_targets_min": 467
    },
    {
      "epoch": 0.064,
      "grad_norm": 4.314878833539552,
      "learning_rate": 4.556962025316456e-06,
      "loss": 0.6765,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.19859322905540466,
      "step": 10,
      "valid_targets_mean": 5291.8,
      "valid_targets_min": 1014
    },
    {
      "epoch": 0.096,
      "grad_norm": 1.988814285885175,
      "learning_rate": 7.08860759493671e-06,
      "loss": 0.6292,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.14756502211093903,
      "step": 15,
      "valid_targets_mean": 4373.5,
      "valid_targets_min": 602
    },
    {
      "epoch": 0.128,
      "grad_norm": 0.9762754146389517,
      "learning_rate": 9.620253164556963e-06,
      "loss": 0.5817,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.14927241206169128,
      "step": 20,
      "valid_targets_mean": 3277.7,
      "valid_targets_min": 585
    },
    {
      "epoch": 0.16,
      "grad_norm": 0.7166931999784403,
      "learning_rate": 1.2151898734177216e-05,
      "loss": 0.5807,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.15850476920604706,
      "step": 25,
      "valid_targets_mean": 4558.8,
      "valid_targets_min": 797
    },
    {
      "epoch": 0.192,
      "grad_norm": 0.5499972361468907,
      "learning_rate": 1.468354430379747e-05,
      "loss": 0.5524,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1318606734275818,
      "step": 30,
      "valid_targets_mean": 4065.6,
      "valid_targets_min": 572
    },
    {
      "epoch": 0.224,
      "grad_norm": 0.49651576098629047,
      "learning_rate": 1.7215189873417723e-05,
      "loss": 0.5184,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.13220933079719543,
      "step": 35,
      "valid_targets_mean": 4484.5,
      "valid_targets_min": 694
    },
    {
      "epoch": 0.256,
      "grad_norm": 0.39684271457127246,
      "learning_rate": 1.974683544303798e-05,
      "loss": 0.4895,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10753373801708221,
      "step": 40,
      "valid_targets_mean": 4064.0,
      "valid_targets_min": 858
    },
    {
      "epoch": 0.288,
      "grad_norm": 0.3882272132605043,
      "learning_rate": 2.2278481012658228e-05,
      "loss": 0.4937,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.11416088044643402,
      "step": 45,
      "valid_targets_mean": 3499.6,
      "valid_targets_min": 746
    },
    {
      "epoch": 0.32,
      "grad_norm": 0.33239432708213396,
      "learning_rate": 2.481012658227848e-05,
      "loss": 0.496,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1379791796207428,
      "step": 50,
      "valid_targets_mean": 4982.7,
      "valid_targets_min": 565
    },
    {
      "epoch": 0.352,
      "grad_norm": 0.3000223370140001,
      "learning_rate": 2.7341772151898737e-05,
      "loss": 0.4724,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08760569989681244,
      "step": 55,
      "valid_targets_mean": 3359.5,
      "valid_targets_min": 928
    },
    {
      "epoch": 0.384,
      "grad_norm": 0.31541516199289,
      "learning_rate": 2.987341772151899e-05,
      "loss": 0.4683,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1303568333387375,
      "step": 60,
      "valid_targets_mean": 4005.9,
      "valid_targets_min": 738
    },
    {
      "epoch": 0.416,
      "grad_norm": 0.28380418904477406,
      "learning_rate": 3.240506329113924e-05,
      "loss": 0.4635,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.12035663425922394,
      "step": 65,
      "valid_targets_mean": 3336.9,
      "valid_targets_min": 556
    },
    {
      "epoch": 0.448,
      "grad_norm": 0.2849083178668719,
      "learning_rate": 3.49367088607595e-05,
      "loss": 0.4529,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.11541417241096497,
      "step": 70,
      "valid_targets_mean": 3875.5,
      "valid_targets_min": 707
    },
    {
      "epoch": 0.48,
      "grad_norm": 0.25334960810520835,
      "learning_rate": 3.746835443037975e-05,
      "loss": 0.4382,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09423618018627167,
      "step": 75,
      "valid_targets_mean": 4197.8,
      "valid_targets_min": 1264
    },
    {
      "epoch": 0.512,
      "grad_norm": 0.2916284172357648,
      "learning_rate": 4e-05,
      "loss": 0.4427,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09766705334186554,
      "step": 80,
      "valid_targets_mean": 3003.1,
      "valid_targets_min": 889
    },
    {
      "epoch": 0.544,
      "grad_norm": 0.2846800529452174,
      "learning_rate": 3.999504991751045e-05,
      "loss": 0.422,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10553984344005585,
      "step": 85,
      "valid_targets_mean": 4164.1,
      "valid_targets_min": 404
    },
    {
      "epoch": 0.576,
      "grad_norm": 0.33848408548782016,
      "learning_rate": 3.9980202120373464e-05,
      "loss": 0.4419,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1027446910738945,
      "step": 90,
      "valid_targets_mean": 3032.6,
      "valid_targets_min": 760
    },
    {
      "epoch": 0.608,
      "grad_norm": 0.2742060851985788,
      "learning_rate": 3.995546395837111e-05,
      "loss": 0.4173,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07320959866046906,
      "step": 95,
      "valid_targets_mean": 3033.0,
      "valid_targets_min": 649
    },
    {
      "epoch": 0.64,
      "grad_norm": 0.29391595941296217,
      "learning_rate": 3.992084767709763e-05,
      "loss": 0.4278,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10428269952535629,
      "step": 100,
      "valid_targets_mean": 3348.6,
      "valid_targets_min": 957
    },
    {
      "epoch": 0.672,
      "grad_norm": 0.30808847813195855,
      "learning_rate": 3.987637041189781e-05,
      "loss": 0.4298,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07673041522502899,
      "step": 105,
      "valid_targets_mean": 2722.7,
      "valid_targets_min": 523
    },
    {
      "epoch": 0.704,
      "grad_norm": 0.2381148496686615,
      "learning_rate": 3.982205417938482e-05,
      "loss": 0.4086,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1124047040939331,
      "step": 110,
      "valid_targets_mean": 6361.1,
      "valid_targets_min": 1895
    },
    {
      "epoch": 0.736,
      "grad_norm": 0.25769400258650194,
      "learning_rate": 3.975792586654179e-05,
      "loss": 0.4015,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10018768161535263,
      "step": 115,
      "valid_targets_mean": 4707.2,
      "valid_targets_min": 901
    },
    {
      "epoch": 0.768,
      "grad_norm": 0.2505290876236424,
      "learning_rate": 3.968401721741259e-05,
      "loss": 0.3993,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06726896017789841,
      "step": 120,
      "valid_targets_mean": 2663.2,
      "valid_targets_min": 712
    },
    {
      "epoch": 0.8,
      "grad_norm": 0.2678954863863751,
      "learning_rate": 3.960036481738819e-05,
      "loss": 0.4108,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0924043208360672,
      "step": 125,
      "valid_targets_mean": 3957.1,
      "valid_targets_min": 831
    },
    {
      "epoch": 0.832,
      "grad_norm": 0.6260136957606488,
      "learning_rate": 3.950701007509667e-05,
      "loss": 0.4088,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09773127734661102,
      "step": 130,
      "valid_targets_mean": 3864.2,
      "valid_targets_min": 682
    },
    {
      "epoch": 0.864,
      "grad_norm": 0.2880682097259892,
      "learning_rate": 3.940399920190552e-05,
      "loss": 0.4139,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09523665904998779,
      "step": 135,
      "valid_targets_mean": 3315.7,
      "valid_targets_min": 844
    },
    {
      "epoch": 0.896,
      "grad_norm": 0.2876325012182124,
      "learning_rate": 3.92913831890467e-05,
      "loss": 0.4188,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.12054994702339172,
      "step": 140,
      "valid_targets_mean": 4335.6,
      "valid_targets_min": 1006
    },
    {
      "epoch": 0.928,
      "grad_norm": 0.24778635395362159,
      "learning_rate": 3.916921778237556e-05,
      "loss": 0.3921,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09123274683952332,
      "step": 145,
      "valid_targets_mean": 4663.9,
      "valid_targets_min": 569
    },
    {
      "epoch": 0.96,
      "grad_norm": 0.284592064996954,
      "learning_rate": 3.903756345477612e-05,
      "loss": 0.4031,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.12018332630395889,
      "step": 150,
      "valid_targets_mean": 4130.1,
      "valid_targets_min": 1097
    },
    {
      "epoch": 0.992,
      "grad_norm": 0.29738961550404686,
      "learning_rate": 3.889648537622657e-05,
      "loss": 0.4064,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10315520316362381,
      "step": 155,
      "valid_targets_mean": 3502.9,
      "valid_targets_min": 979
    },
    {
      "epoch": 1.0192,
      "grad_norm": 0.2731735127054483,
      "learning_rate": 3.874605338153952e-05,
      "loss": 0.391,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10287749022245407,
      "step": 160,
      "valid_targets_mean": 4471.2,
      "valid_targets_min": 955
    },
    {
      "epoch": 1.0512,
      "grad_norm": 0.28765617426017454,
      "learning_rate": 3.8586341935793265e-05,
      "loss": 0.3894,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09611120074987411,
      "step": 165,
      "valid_targets_mean": 3267.9,
      "valid_targets_min": 757
    },
    {
      "epoch": 1.0832,
      "grad_norm": 0.31101265356864577,
      "learning_rate": 3.841743009747089e-05,
      "loss": 0.4083,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09512562304735184,
      "step": 170,
      "valid_targets_mean": 3343.1,
      "valid_targets_min": 944
    },
    {
      "epoch": 1.1152,
      "grad_norm": 0.2905907579966133,
      "learning_rate": 3.8239401479325714e-05,
      "loss": 0.3937,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08502990007400513,
      "step": 175,
      "valid_targets_mean": 2825.6,
      "valid_targets_min": 659
    },
    {
      "epoch": 1.1472,
      "grad_norm": 0.2832976829376036,
      "learning_rate": 3.8052344206992276e-05,
      "loss": 0.3969,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.13293340802192688,
      "step": 180,
      "valid_targets_mean": 4333.3,
      "valid_targets_min": 426
    },
    {
      "epoch": 1.1792,
      "grad_norm": 0.28463958184499244,
      "learning_rate": 3.7856350875363396e-05,
      "loss": 0.3874,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1205093115568161,
      "step": 185,
      "valid_targets_mean": 5400.4,
      "valid_targets_min": 983
    },
    {
      "epoch": 1.2112,
      "grad_norm": 0.2692051009946816,
      "learning_rate": 3.765151850275497e-05,
      "loss": 0.3768,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09737041592597961,
      "step": 190,
      "valid_targets_mean": 4397.6,
      "valid_targets_min": 789
    },
    {
      "epoch": 1.2432,
      "grad_norm": 0.2859575252537898,
      "learning_rate": 3.7437948482881104e-05,
      "loss": 0.4238,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10958155989646912,
      "step": 195,
      "valid_targets_mean": 4294.4,
      "valid_targets_min": 926
    },
    {
      "epoch": 1.2752,
      "grad_norm": 0.33354975765989114,
      "learning_rate": 3.721574653466336e-05,
      "loss": 0.4021,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09777568280696869,
      "step": 200,
      "valid_targets_mean": 3833.1,
      "valid_targets_min": 1095
    },
    {
      "epoch": 1.3072,
      "grad_norm": 0.29013275970765756,
      "learning_rate": 3.698502264989903e-05,
      "loss": 0.375,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08013059943914413,
      "step": 205,
      "valid_targets_mean": 3123.3,
      "valid_targets_min": 840
    },
    {
      "epoch": 1.3392,
      "grad_norm": 0.2446668013665798,
      "learning_rate": 3.674589103881432e-05,
      "loss": 0.383,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10853144526481628,
      "step": 210,
      "valid_targets_mean": 5498.0,
      "valid_targets_min": 969
    },
    {
      "epoch": 1.3712,
      "grad_norm": 0.31532330784965257,
      "learning_rate": 3.64984700735293e-05,
      "loss": 0.3884,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10882705450057983,
      "step": 215,
      "valid_targets_mean": 4198.4,
      "valid_targets_min": 596
    },
    {
      "epoch": 1.4032,
      "grad_norm": 0.2570799225968013,
      "learning_rate": 3.624288222946273e-05,
      "loss": 0.3983,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.11490095406770706,
      "step": 220,
      "valid_targets_mean": 5037.5,
      "valid_targets_min": 780
    },
    {
      "epoch": 1.4352,
      "grad_norm": 0.3364437735288334,
      "learning_rate": 3.597925402470578e-05,
      "loss": 0.3962,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07053709030151367,
      "step": 225,
      "valid_targets_mean": 1981.5,
      "valid_targets_min": 662
    },
    {
      "epoch": 1.4672,
      "grad_norm": 0.6771297145385203,
      "learning_rate": 3.570771595739445e-05,
      "loss": 0.3814,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08904281258583069,
      "step": 230,
      "valid_targets_mean": 4233.0,
      "valid_targets_min": 766
    },
    {
      "epoch": 1.4992,
      "grad_norm": 0.275707090430415,
      "learning_rate": 3.5428402441111964e-05,
      "loss": 0.3651,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.11852708458900452,
      "step": 235,
      "valid_targets_mean": 5108.5,
      "valid_targets_min": 816
    },
    {
      "epoch": 1.5312000000000001,
      "grad_norm": 0.29650227882941366,
      "learning_rate": 3.5141451738352936e-05,
      "loss": 0.3994,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10156121850013733,
      "step": 240,
      "valid_targets_mean": 3942.5,
      "valid_targets_min": 721
    },
    {
      "epoch": 1.5632000000000001,
      "grad_norm": 0.255641622829641,
      "learning_rate": 3.4847005892082266e-05,
      "loss": 0.3867,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07486311346292496,
      "step": 245,
      "valid_targets_mean": 4506.0,
      "valid_targets_min": 690
    },
    {
      "epoch": 1.5952,
      "grad_norm": 0.2663743138508868,
      "learning_rate": 3.454521065542273e-05,
      "loss": 0.3818,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07315202802419662,
      "step": 250,
      "valid_targets_mean": 2802.1,
      "valid_targets_min": 422
    },
    {
      "epoch": 1.6272,
      "grad_norm": 0.2830883354341115,
      "learning_rate": 3.423621541950597e-05,
      "loss": 0.3694,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09550144523382187,
      "step": 255,
      "valid_targets_mean": 3638.8,
      "valid_targets_min": 567
    },
    {
      "epoch": 1.6592,
      "grad_norm": 0.2664476249658465,
      "learning_rate": 3.3920173139522664e-05,
      "loss": 0.3925,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10125439614057541,
      "step": 260,
      "valid_targets_mean": 4288.8,
      "valid_targets_min": 1138
    },
    {
      "epoch": 1.6912,
      "grad_norm": 0.25499158613866524,
      "learning_rate": 3.35972402590084e-05,
      "loss": 0.3821,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08418112248182297,
      "step": 265,
      "valid_targets_mean": 5148.1,
      "valid_targets_min": 701
    },
    {
      "epoch": 1.7231999999999998,
      "grad_norm": 0.26192491388253997,
      "learning_rate": 3.326757663240291e-05,
      "loss": 0.4026,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.124482661485672,
      "step": 270,
      "valid_targets_mean": 5508.5,
      "valid_targets_min": 990
    },
    {
      "epoch": 1.7551999999999999,
      "grad_norm": 0.24884953430346898,
      "learning_rate": 3.293134544592073e-05,
      "loss": 0.3592,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09600067883729935,
      "step": 275,
      "valid_targets_mean": 4959.2,
      "valid_targets_min": 958
    },
    {
      "epoch": 1.7872,
      "grad_norm": 0.27099567040525147,
      "learning_rate": 3.258871313677274e-05,
      "loss": 0.3937,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09132636338472366,
      "step": 280,
      "valid_targets_mean": 4607.1,
      "valid_targets_min": 919
    },
    {
      "epoch": 1.8192,
      "grad_norm": 0.27997026775033934,
      "learning_rate": 3.2239849310778316e-05,
      "loss": 0.3942,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08237655460834503,
      "step": 285,
      "valid_targets_mean": 3027.5,
      "valid_targets_min": 697
    },
    {
      "epoch": 1.8512,
      "grad_norm": 0.30458982164036624,
      "learning_rate": 3.188492665840909e-05,
      "loss": 0.3745,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06693554669618607,
      "step": 290,
      "valid_targets_mean": 2766.1,
      "valid_targets_min": 570
    },
    {
      "epoch": 1.8832,
      "grad_norm": 0.2764200635911673,
      "learning_rate": 3.1524120869305726e-05,
      "loss": 0.3749,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1323719620704651,
      "step": 295,
      "valid_targets_mean": 5092.9,
      "valid_targets_min": 898
    },
    {
      "epoch": 1.9152,
      "grad_norm": 0.2559138552607389,
      "learning_rate": 3.11576105453101e-05,
      "loss": 0.3722,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06034687161445618,
      "step": 300,
      "valid_targets_mean": 2967.7,
      "valid_targets_min": 693
    },
    {
      "epoch": 1.9472,
      "grad_norm": 0.2485016007838173,
      "learning_rate": 3.0785577112055916e-05,
      "loss": 0.3944,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09485583752393723,
      "step": 305,
      "valid_targets_mean": 3780.9,
      "valid_targets_min": 484
    },
    {
      "epoch": 1.9792,
      "grad_norm": 0.2801394049130346,
      "learning_rate": 3.040820472916153e-05,
      "loss": 0.3733,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.13164201378822327,
      "step": 310,
      "valid_targets_mean": 4941.2,
      "valid_targets_min": 618
    },
    {
      "epoch": 2.0064,
      "grad_norm": 0.34020200928849137,
      "learning_rate": 3.002568019906939e-05,
      "loss": 0.3884,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0848844051361084,
      "step": 315,
      "valid_targets_mean": 2673.7,
      "valid_targets_min": 829
    },
    {
      "epoch": 2.0384,
      "grad_norm": 0.27337669075984855,
      "learning_rate": 2.963819287457733e-05,
      "loss": 0.3693,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0860731303691864,
      "step": 320,
      "valid_targets_mean": 3390.1,
      "valid_targets_min": 599
    },
    {
      "epoch": 2.0704,
      "grad_norm": 0.2783114002913622,
      "learning_rate": 2.924593456510733e-05,
      "loss": 0.3578,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.05866974592208862,
      "step": 325,
      "valid_targets_mean": 2847.2,
      "valid_targets_min": 816
    },
    {
      "epoch": 2.1024,
      "grad_norm": 0.28437495992172673,
      "learning_rate": 2.8849099441758306e-05,
      "loss": 0.3649,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08873450756072998,
      "step": 330,
      "valid_targets_mean": 4102.1,
      "valid_targets_min": 974
    },
    {
      "epoch": 2.1344,
      "grad_norm": 0.4327891714878026,
      "learning_rate": 2.844788394118979e-05,
      "loss": 0.3828,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07443979382514954,
      "step": 335,
      "valid_targets_mean": 3739.9,
      "valid_targets_min": 700
    },
    {
      "epoch": 2.1664,
      "grad_norm": 0.2773782860323178,
      "learning_rate": 2.8042486668384164e-05,
      "loss": 0.3711,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07605782151222229,
      "step": 340,
      "valid_targets_mean": 3293.5,
      "valid_targets_min": 647
    },
    {
      "epoch": 2.1984,
      "grad_norm": 0.32684812133097485,
      "learning_rate": 2.7633108298335582e-05,
      "loss": 0.3932,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.11338317394256592,
      "step": 345,
      "valid_targets_mean": 4493.5,
      "valid_targets_min": 551
    },
    {
      "epoch": 2.2304,
      "grad_norm": 0.2754448813377894,
      "learning_rate": 2.721995147671416e-05,
      "loss": 0.3689,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09111467748880386,
      "step": 350,
      "valid_targets_mean": 3898.8,
      "valid_targets_min": 943
    },
    {
      "epoch": 2.2624,
      "grad_norm": 0.2960034595879353,
      "learning_rate": 2.68032207195547e-05,
      "loss": 0.3656,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08844823390245438,
      "step": 355,
      "valid_targets_mean": 3074.0,
      "valid_targets_min": 819
    },
    {
      "epoch": 2.2944,
      "grad_norm": 0.2800287115680443,
      "learning_rate": 2.6383122312019604e-05,
      "loss": 0.3741,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.11491911113262177,
      "step": 360,
      "valid_targets_mean": 4926.6,
      "valid_targets_min": 923
    },
    {
      "epoch": 2.3264,
      "grad_norm": 0.2724348953446165,
      "learning_rate": 2.595986420628597e-05,
      "loss": 0.3622,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07493268698453903,
      "step": 365,
      "valid_targets_mean": 3832.2,
      "valid_targets_min": 815
    },
    {
      "epoch": 2.3584,
      "grad_norm": 0.30605393860518737,
      "learning_rate": 2.5533655918607573e-05,
      "loss": 0.3622,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07456633448600769,
      "step": 370,
      "valid_targets_mean": 2906.2,
      "valid_targets_min": 792
    },
    {
      "epoch": 2.3904,
      "grad_norm": 0.24341978273572606,
      "learning_rate": 2.510470842560259e-05,
      "loss": 0.3596,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06545670330524445,
      "step": 375,
      "valid_targets_mean": 3845.6,
      "valid_targets_min": 761
    },
    {
      "epoch": 2.4224,
      "grad_norm": 0.27946411898092166,
      "learning_rate": 2.467323405981841e-05,
      "loss": 0.3691,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.12025629729032516,
      "step": 380,
      "valid_targets_mean": 4938.5,
      "valid_targets_min": 807
    },
    {
      "epoch": 2.4544,
      "grad_norm": 0.29284916704415465,
      "learning_rate": 2.423944640462533e-05,
      "loss": 0.3713,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0910378098487854,
      "step": 385,
      "valid_targets_mean": 3405.7,
      "valid_targets_min": 686
    },
    {
      "epoch": 2.4864,
      "grad_norm": 0.2900319137834253,
      "learning_rate": 2.3803560188490968e-05,
      "loss": 0.3749,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10050083696842194,
      "step": 390,
      "valid_targets_mean": 3770.3,
      "valid_targets_min": 790
    },
    {
      "epoch": 2.5183999999999997,
      "grad_norm": 0.27605368296747607,
      "learning_rate": 2.336579117868789e-05,
      "loss": 0.3522,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.05563701316714287,
      "step": 395,
      "valid_targets_mean": 2379.9,
      "valid_targets_min": 512
    },
    {
      "epoch": 2.5504,
      "grad_norm": 0.24596698179499862,
      "learning_rate": 2.292635607448711e-05,
      "loss": 0.3686,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0977577418088913,
      "step": 400,
      "valid_targets_mean": 4720.2,
      "valid_targets_min": 982
    },
    {
      "epoch": 2.5824,
      "grad_norm": 0.3064085929419356,
      "learning_rate": 2.248547239989008e-05,
      "loss": 0.3742,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07541108876466751,
      "step": 405,
      "valid_targets_mean": 3049.4,
      "valid_targets_min": 673
    },
    {
      "epoch": 2.6144,
      "grad_norm": 0.2813136411331836,
      "learning_rate": 2.204335839595255e-05,
      "loss": 0.378,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.12282659858465195,
      "step": 410,
      "valid_targets_mean": 5723.6,
      "valid_targets_min": 688
    },
    {
      "epoch": 2.6464,
      "grad_norm": 0.2371573288855974,
      "learning_rate": 2.1600232912753452e-05,
      "loss": 0.3792,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06939087808132172,
      "step": 415,
      "valid_targets_mean": 4753.8,
      "valid_targets_min": 777
    },
    {
      "epoch": 2.6784,
      "grad_norm": 0.2999696661955818,
      "learning_rate": 2.1156315301062293e-05,
      "loss": 0.38,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10134214162826538,
      "step": 420,
      "valid_targets_mean": 4125.9,
      "valid_targets_min": 744
    },
    {
      "epoch": 2.7104,
      "grad_norm": 0.23881903955855627,
      "learning_rate": 2.0711825303758712e-05,
      "loss": 0.3588,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08206147700548172,
      "step": 425,
      "valid_targets_mean": 4427.5,
      "valid_targets_min": 676
    },
    {
      "epoch": 2.7424,
      "grad_norm": 0.23621642170072332,
      "learning_rate": 2.0266982947057962e-05,
      "loss": 0.3625,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09457995742559433,
      "step": 430,
      "valid_targets_mean": 5558.1,
      "valid_targets_min": 1035
    },
    {
      "epoch": 2.7744,
      "grad_norm": 0.2396557510203461,
      "learning_rate": 1.9822008431596083e-05,
      "loss": 0.3565,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07737955451011658,
      "step": 435,
      "valid_targets_mean": 4501.7,
      "valid_targets_min": 454
    },
    {
      "epoch": 2.8064,
      "grad_norm": 0.2996391770693032,
      "learning_rate": 1.937712202342881e-05,
      "loss": 0.3811,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08548533916473389,
      "step": 440,
      "valid_targets_mean": 2863.6,
      "valid_targets_min": 996
    },
    {
      "epoch": 2.8384,
      "grad_norm": 0.2798952151599294,
      "learning_rate": 1.8932543944998037e-05,
      "loss": 0.3752,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10243545472621918,
      "step": 445,
      "valid_targets_mean": 4691.2,
      "valid_targets_min": 939
    },
    {
      "epoch": 2.8704,
      "grad_norm": 0.2512026376529301,
      "learning_rate": 1.8488494266119877e-05,
      "loss": 0.3578,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09460596740245819,
      "step": 450,
      "valid_targets_mean": 4827.6,
      "valid_targets_min": 916
    },
    {
      "epoch": 2.9024,
      "grad_norm": 0.27140518289799176,
      "learning_rate": 1.804519279504834e-05,
      "loss": 0.358,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08733229339122772,
      "step": 455,
      "valid_targets_mean": 3709.1,
      "valid_targets_min": 1010
    },
    {
      "epoch": 2.9344,
      "grad_norm": 0.3157210826461013,
      "learning_rate": 1.7602858969668365e-05,
      "loss": 0.3762,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0807497650384903,
      "step": 460,
      "valid_targets_mean": 3692.1,
      "valid_targets_min": 427
    },
    {
      "epoch": 2.9664,
      "grad_norm": 0.2764925498447452,
      "learning_rate": 1.716171174887231e-05,
      "loss": 0.3546,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09397895634174347,
      "step": 465,
      "valid_targets_mean": 4732.0,
      "valid_targets_min": 680
    },
    {
      "epoch": 2.9984,
      "grad_norm": 0.2745536484205702,
      "learning_rate": 1.6721969504173484e-05,
      "loss": 0.3718,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0864214152097702,
      "step": 470,
      "valid_targets_mean": 3416.9,
      "valid_targets_min": 645
    },
    {
      "epoch": 3.0256,
      "grad_norm": 0.2769677881256928,
      "learning_rate": 1.628384991161041e-05,
      "loss": 0.3651,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09713249653577805,
      "step": 475,
      "valid_targets_mean": 5895.2,
      "valid_targets_min": 956
    },
    {
      "epoch": 3.0576,
      "grad_norm": 0.2490158683228284,
      "learning_rate": 1.5847569843995452e-05,
      "loss": 0.3594,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07401568442583084,
      "step": 480,
      "valid_targets_mean": 4233.2,
      "valid_targets_min": 881
    },
    {
      "epoch": 3.0896,
      "grad_norm": 0.3352405826884512,
      "learning_rate": 1.5413345263560922e-05,
      "loss": 0.3711,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10404595732688904,
      "step": 485,
      "valid_targets_mean": 4041.5,
      "valid_targets_min": 716
    },
    {
      "epoch": 3.1216,
      "grad_norm": 0.2757084776370529,
      "learning_rate": 1.4981391115056032e-05,
      "loss": 0.3565,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09845767915248871,
      "step": 490,
      "valid_targets_mean": 4260.2,
      "valid_targets_min": 927
    },
    {
      "epoch": 3.1536,
      "grad_norm": 0.40609821553023173,
      "learning_rate": 1.455192121934748e-05,
      "loss": 0.3532,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10451191663742065,
      "step": 495,
      "valid_targets_mean": 4472.2,
      "valid_targets_min": 1499
    },
    {
      "epoch": 3.1856,
      "grad_norm": 0.2831681404817536,
      "learning_rate": 1.4125148167576303e-05,
      "loss": 0.3653,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10405318439006805,
      "step": 500,
      "valid_targets_mean": 4023.4,
      "valid_targets_min": 740
    },
    {
      "epoch": 3.2176,
      "grad_norm": 0.284313975291648,
      "learning_rate": 1.3701283215923563e-05,
      "loss": 0.3631,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09694987535476685,
      "step": 505,
      "valid_targets_mean": 4803.8,
      "valid_targets_min": 771
    },
    {
      "epoch": 3.2496,
      "grad_norm": 0.28781470117417157,
      "learning_rate": 1.328053618103677e-05,
      "loss": 0.3716,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10028200596570969,
      "step": 510,
      "valid_targets_mean": 4057.0,
      "valid_targets_min": 803
    },
    {
      "epoch": 3.2816,
      "grad_norm": 0.2584710891473149,
      "learning_rate": 1.2863115336168916e-05,
      "loss": 0.3568,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07050131261348724,
      "step": 515,
      "valid_targets_mean": 3638.6,
      "valid_targets_min": 613
    },
    {
      "epoch": 3.3136,
      "grad_norm": 0.260137770246497,
      "learning_rate": 1.2449227308081509e-05,
      "loss": 0.3467,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07993818074464798,
      "step": 520,
      "valid_targets_mean": 4442.3,
      "valid_targets_min": 962
    },
    {
      "epoch": 3.3456,
      "grad_norm": 0.27786727440367376,
      "learning_rate": 1.2039076974762587e-05,
      "loss": 0.3564,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07751774042844772,
      "step": 525,
      "valid_targets_mean": 3960.4,
      "valid_targets_min": 585
    },
    {
      "epoch": 3.3776,
      "grad_norm": 0.28949739215719694,
      "learning_rate": 1.163286736401044e-05,
      "loss": 0.3517,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10838130861520767,
      "step": 530,
      "valid_targets_mean": 4890.1,
      "valid_targets_min": 1188
    },
    {
      "epoch": 3.4096,
      "grad_norm": 0.29180588640825234,
      "learning_rate": 1.123079955293322e-05,
      "loss": 0.3553,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.05709638446569443,
      "step": 535,
      "valid_targets_mean": 2373.1,
      "valid_targets_min": 892
    },
    {
      "epoch": 3.4416,
      "grad_norm": 0.2785168622660705,
      "learning_rate": 1.0833072568414037e-05,
      "loss": 0.3524,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08552928268909454,
      "step": 540,
      "valid_targets_mean": 4066.9,
      "valid_targets_min": 894
    },
    {
      "epoch": 3.4736000000000002,
      "grad_norm": 0.2666793684005187,
      "learning_rate": 1.0439883288591057e-05,
      "loss": 0.3523,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07097834348678589,
      "step": 545,
      "valid_targets_mean": 3522.5,
      "valid_targets_min": 948
    },
    {
      "epoch": 3.5056000000000003,
      "grad_norm": 0.2661159658712619,
      "learning_rate": 1.0051426345401202e-05,
      "loss": 0.3728,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07339584827423096,
      "step": 550,
      "valid_targets_mean": 3156.4,
      "valid_targets_min": 534
    },
    {
      "epoch": 3.5376,
      "grad_norm": 0.24501108014940887,
      "learning_rate": 9.667894028235704e-06,
      "loss": 0.346,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08953475952148438,
      "step": 555,
      "valid_targets_mean": 5984.2,
      "valid_targets_min": 711
    },
    {
      "epoch": 3.5696,
      "grad_norm": 0.2906586233564921,
      "learning_rate": 9.289476188755315e-06,
      "loss": 0.3673,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09603782743215561,
      "step": 560,
      "valid_targets_mean": 4052.5,
      "valid_targets_min": 855
    },
    {
      "epoch": 3.6016,
      "grad_norm": 0.272386870398156,
      "learning_rate": 8.916360146912122e-06,
      "loss": 0.3706,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10408547520637512,
      "step": 565,
      "valid_targets_mean": 5669.6,
      "valid_targets_min": 573
    },
    {
      "epoch": 3.6336,
      "grad_norm": 0.29167582987552243,
      "learning_rate": 8.548730598224646e-06,
      "loss": 0.3428,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10972733795642853,
      "step": 570,
      "valid_targets_mean": 4005.2,
      "valid_targets_min": 364
    },
    {
      "epoch": 3.6656,
      "grad_norm": 0.2506154175009061,
      "learning_rate": 8.186769522352053e-06,
      "loss": 0.3454,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08571542054414749,
      "step": 575,
      "valid_targets_mean": 4964.8,
      "valid_targets_min": 662
    },
    {
      "epoch": 3.6976,
      "grad_norm": 0.2833111429001337,
      "learning_rate": 7.830656093012714e-06,
      "loss": 0.368,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07560089230537415,
      "step": 580,
      "valid_targets_mean": 3193.0,
      "valid_targets_min": 831
    },
    {
      "epoch": 3.7296,
      "grad_norm": 0.25648445826160626,
      "learning_rate": 7.480566589291696e-06,
      "loss": 0.3535,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0924566239118576,
      "step": 585,
      "valid_targets_mean": 4515.9,
      "valid_targets_min": 819
    },
    {
      "epoch": 3.7616,
      "grad_norm": 0.24085429575690023,
      "learning_rate": 7.1366743083812285e-06,
      "loss": 0.358,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08820848912000656,
      "step": 590,
      "valid_targets_mean": 5276.7,
      "valid_targets_min": 792
    },
    {
      "epoch": 3.7936,
      "grad_norm": 0.3065040628125282,
      "learning_rate": 6.799149479797101e-06,
      "loss": 0.3438,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.11131012439727783,
      "step": 595,
      "valid_targets_mean": 4458.4,
      "valid_targets_min": 660
    },
    {
      "epoch": 3.8256,
      "grad_norm": 0.26576774517699353,
      "learning_rate": 6.4681591811137e-06,
      "loss": 0.3499,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07206078618764877,
      "step": 600,
      "valid_targets_mean": 3065.3,
      "valid_targets_min": 731
    },
    {
      "epoch": 3.8576,
      "grad_norm": 0.31902392164489723,
      "learning_rate": 6.143867255259197e-06,
      "loss": 0.3727,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09210214018821716,
      "step": 605,
      "valid_targets_mean": 3760.6,
      "valid_targets_min": 495
    },
    {
      "epoch": 3.8895999999999997,
      "grad_norm": 0.2675703617456016,
      "learning_rate": 5.8264342294119504e-06,
      "loss": 0.3494,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08684878796339035,
      "step": 610,
      "valid_targets_mean": 3022.6,
      "valid_targets_min": 740
    },
    {
      "epoch": 3.9215999999999998,
      "grad_norm": 0.2574492657127961,
      "learning_rate": 5.516017235538258e-06,
      "loss": 0.3649,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0930505245923996,
      "step": 615,
      "valid_targets_mean": 4483.4,
      "valid_targets_min": 860
    },
    {
      "epoch": 3.9536,
      "grad_norm": 0.2867195535050637,
      "learning_rate": 5.212769932610695e-06,
      "loss": 0.3522,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10067837685346603,
      "step": 620,
      "valid_targets_mean": 4845.6,
      "valid_targets_min": 1032
    },
    {
      "epoch": 3.9856,
      "grad_norm": 0.267049627621211,
      "learning_rate": 4.916842430545681e-06,
      "loss": 0.3562,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10272210091352463,
      "step": 625,
      "valid_targets_mean": 3688.7,
      "valid_targets_min": 1466
    },
    {
      "epoch": 4.0128,
      "grad_norm": 0.27458289977616085,
      "learning_rate": 4.628381215897837e-06,
      "loss": 0.3495,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06917114555835724,
      "step": 630,
      "valid_targets_mean": 3087.8,
      "valid_targets_min": 923
    },
    {
      "epoch": 4.0448,
      "grad_norm": 0.2517192137514772,
      "learning_rate": 4.347529079347914e-06,
      "loss": 0.3506,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06825089454650879,
      "step": 635,
      "valid_targets_mean": 3181.8,
      "valid_targets_min": 600
    },
    {
      "epoch": 4.0768,
      "grad_norm": 0.28660586312956327,
      "learning_rate": 4.074425045020247e-06,
      "loss": 0.3573,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08183202147483826,
      "step": 640,
      "valid_targets_mean": 3011.0,
      "valid_targets_min": 765
    },
    {
      "epoch": 4.1088,
      "grad_norm": 0.28463640630416887,
      "learning_rate": 3.8092043016646487e-06,
      "loss": 0.3552,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10132152587175369,
      "step": 645,
      "valid_targets_mean": 4639.2,
      "valid_targets_min": 986
    },
    {
      "epoch": 4.1408,
      "grad_norm": 0.2925059222811773,
      "learning_rate": 3.551998135736867e-06,
      "loss": 0.3643,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.11987960338592529,
      "step": 650,
      "valid_targets_mean": 4305.8,
      "valid_targets_min": 596
    },
    {
      "epoch": 4.1728,
      "grad_norm": 0.24703709684860067,
      "learning_rate": 3.3029338664107267e-06,
      "loss": 0.3588,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08891130983829498,
      "step": 655,
      "valid_targets_mean": 4974.1,
      "valid_targets_min": 866
    },
    {
      "epoch": 4.2048,
      "grad_norm": 0.2508952800570821,
      "learning_rate": 3.0621347825540625e-06,
      "loss": 0.3423,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1111423522233963,
      "step": 660,
      "valid_targets_mean": 5120.2,
      "valid_targets_min": 792
    },
    {
      "epoch": 4.2368,
      "grad_norm": 0.27565385811171317,
      "learning_rate": 2.8297200816997183e-06,
      "loss": 0.3495,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10503530502319336,
      "step": 665,
      "valid_targets_mean": 4422.1,
      "valid_targets_min": 929
    },
    {
      "epoch": 4.2688,
      "grad_norm": 0.2680017465041526,
      "learning_rate": 2.605804811041803e-06,
      "loss": 0.3471,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08640946447849274,
      "step": 670,
      "valid_targets_mean": 4343.9,
      "valid_targets_min": 1293
    },
    {
      "epoch": 4.3008,
      "grad_norm": 0.24533072291740976,
      "learning_rate": 2.390499810486351e-06,
      "loss": 0.3417,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06334862112998962,
      "step": 675,
      "valid_targets_mean": 3860.0,
      "valid_targets_min": 737
    },
    {
      "epoch": 4.3328,
      "grad_norm": 0.2914306867157244,
      "learning_rate": 2.183911657784685e-06,
      "loss": 0.3501,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.089777871966362,
      "step": 680,
      "valid_targets_mean": 4237.7,
      "valid_targets_min": 645
    },
    {
      "epoch": 4.3648,
      "grad_norm": 0.28694285402952496,
      "learning_rate": 1.986142615776532e-06,
      "loss": 0.3668,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06682998687028885,
      "step": 685,
      "valid_targets_mean": 2856.2,
      "valid_targets_min": 814
    },
    {
      "epoch": 4.3968,
      "grad_norm": 0.2802402740685262,
      "learning_rate": 1.7972905817690644e-06,
      "loss": 0.3599,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.06919320672750473,
      "step": 690,
      "valid_targets_mean": 2777.1,
      "valid_targets_min": 797
    },
    {
      "epoch": 4.4288,
      "grad_norm": 0.23764619085964592,
      "learning_rate": 1.617449039076955e-06,
      "loss": 0.3544,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09146711975336075,
      "step": 695,
      "valid_targets_mean": 5609.9,
      "valid_targets_min": 779
    },
    {
      "epoch": 4.4608,
      "grad_norm": 0.27218616013106633,
      "learning_rate": 1.4467070107473413e-06,
      "loss": 0.3537,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08877747505903244,
      "step": 700,
      "valid_targets_mean": 4865.7,
      "valid_targets_min": 781
    },
    {
      "epoch": 4.4928,
      "grad_norm": 0.24836689184527166,
      "learning_rate": 1.2851490154926816e-06,
      "loss": 0.3558,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.12025726586580276,
      "step": 705,
      "valid_targets_mean": 6002.2,
      "valid_targets_min": 511
    },
    {
      "epoch": 4.5248,
      "grad_norm": 0.2839955873051625,
      "learning_rate": 1.1328550258533211e-06,
      "loss": 0.35,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.05349263548851013,
      "step": 710,
      "valid_targets_mean": 2157.0,
      "valid_targets_min": 676
    },
    {
      "epoch": 4.5568,
      "grad_norm": 0.25339934073000153,
      "learning_rate": 9.899004286103953e-07,
      "loss": 0.3628,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0897410586476326,
      "step": 715,
      "valid_targets_mean": 4071.1,
      "valid_targets_min": 815
    },
    {
      "epoch": 4.5888,
      "grad_norm": 0.28765186751070365,
      "learning_rate": 8.5635598746876e-07,
      "loss": 0.3552,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.07775408029556274,
      "step": 720,
      "valid_targets_mean": 2341.9,
      "valid_targets_min": 901
    },
    {
      "epoch": 4.6208,
      "grad_norm": 0.23909899217629269,
      "learning_rate": 7.32287808028389e-07,
      "loss": 0.358,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09574270248413086,
      "step": 725,
      "valid_targets_mean": 5048.5,
      "valid_targets_min": 571
    },
    {
      "epoch": 4.6528,
      "grad_norm": 0.28852955206005965,
      "learning_rate": 6.177573050615327e-07,
      "loss": 0.353,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0888708233833313,
      "step": 730,
      "valid_targets_mean": 3223.6,
      "valid_targets_min": 752
    },
    {
      "epoch": 4.6848,
      "grad_norm": 0.2396949849023496,
      "learning_rate": 5.128211721119213e-07,
      "loss": 0.3548,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08730411529541016,
      "step": 735,
      "valid_targets_mean": 6840.9,
      "valid_targets_min": 574
    },
    {
      "epoch": 4.7168,
      "grad_norm": 0.26724615386539685,
      "learning_rate": 4.175313534309755e-07,
      "loss": 0.3513,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.065192312002182,
      "step": 740,
      "valid_targets_mean": 2788.9,
      "valid_targets_min": 810
    },
    {
      "epoch": 4.7488,
      "grad_norm": 0.3814064753977133,
      "learning_rate": 3.319350182649861e-07,
      "loss": 0.3426,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09376055747270584,
      "step": 745,
      "valid_targets_mean": 4400.3,
      "valid_targets_min": 795
    },
    {
      "epoch": 4.7808,
      "grad_norm": 0.26279204921073596,
      "learning_rate": 2.560745375059392e-07,
      "loss": 0.3345,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.10804213583469391,
      "step": 750,
      "valid_targets_mean": 4718.9,
      "valid_targets_min": 1056
    },
    {
      "epoch": 4.8128,
      "grad_norm": 0.31623967025773475,
      "learning_rate": 1.8998746271758016e-07,
      "loss": 0.3658,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08865915983915329,
      "step": 755,
      "valid_targets_mean": 3017.4,
      "valid_targets_min": 747
    },
    {
      "epoch": 4.8448,
      "grad_norm": 0.2729210569029963,
      "learning_rate": 1.337065075470778e-07,
      "loss": 0.3555,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09759851545095444,
      "step": 760,
      "valid_targets_mean": 3898.8,
      "valid_targets_min": 726
    },
    {
      "epoch": 4.8768,
      "grad_norm": 0.26101693014253413,
      "learning_rate": 8.725953153150279e-08,
      "loss": 0.3516,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09580200910568237,
      "step": 765,
      "valid_targets_mean": 4539.4,
      "valid_targets_min": 901
    },
    {
      "epoch": 4.9088,
      "grad_norm": 0.25225033997235635,
      "learning_rate": 5.066952630711886e-08,
      "loss": 0.3407,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.0713324025273323,
      "step": 770,
      "valid_targets_mean": 4291.9,
      "valid_targets_min": 844
    },
    {
      "epoch": 4.9408,
      "grad_norm": 0.29707751018015915,
      "learning_rate": 2.3954604228342283e-08,
      "loss": 0.3615,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.09581207484006882,
      "step": 775,
      "valid_targets_mean": 3682.2,
      "valid_targets_min": 624
    },
    {
      "epoch": 4.9728,
      "grad_norm": 0.25529371525702294,
      "learning_rate": 7.12798940197601e-09,
      "loss": 0.3442,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.08812999725341797,
      "step": 780,
      "valid_targets_mean": 5000.4,
      "valid_targets_min": 721
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.5107614747941043,
      "learning_rate": 1.9801114115480802e-10,
      "loss": 0.3493,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40154755115509033,
      "step": 785,
      "valid_targets_mean": 4463.1,
      "valid_targets_min": 880
    },
    {
      "epoch": 5.0,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40154755115509033,
      "step": 785,
      "total_flos": 1.0927339449708708e+18,
      "train_loss": 0.387720340224588,
      "train_runtime": 16761.1211,
      "train_samples_per_second": 2.983,
      "train_steps_per_second": 0.047,
      "valid_targets_mean": 4463.1,
      "valid_targets_min": 880
    }
  ],
  "logging_steps": 5,
  "max_steps": 785,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": false,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 1.0927339449708708e+18,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}