{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 1565,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.016,
      "grad_norm": 7.2823718978969225,
      "learning_rate": 1.0191082802547772e-06,
      "loss": 0.9132,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.44200366735458374,
      "step": 5,
      "valid_targets_mean": 2732.6,
      "valid_targets_min": 1358
    },
    {
      "epoch": 0.032,
      "grad_norm": 6.257095873405941,
      "learning_rate": 2.2929936305732485e-06,
      "loss": 0.9252,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.46359115839004517,
      "step": 10,
      "valid_targets_mean": 3118.3,
      "valid_targets_min": 698
    },
    {
      "epoch": 0.048,
      "grad_norm": 4.411856550169914,
      "learning_rate": 3.56687898089172e-06,
      "loss": 0.8594,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.478965699672699,
      "step": 15,
      "valid_targets_mean": 3442.2,
      "valid_targets_min": 1957
    },
    {
      "epoch": 0.064,
      "grad_norm": 2.5700848407864276,
      "learning_rate": 4.840764331210192e-06,
      "loss": 0.8379,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3646432161331177,
      "step": 20,
      "valid_targets_mean": 2799.1,
      "valid_targets_min": 979
    },
    {
      "epoch": 0.08,
      "grad_norm": 1.5191408406019358,
      "learning_rate": 6.114649681528663e-06,
      "loss": 0.7846,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3868666887283325,
      "step": 25,
      "valid_targets_mean": 3145.7,
      "valid_targets_min": 1079
    },
    {
      "epoch": 0.096,
      "grad_norm": 1.0745656000903365,
      "learning_rate": 7.388535031847134e-06,
      "loss": 0.7665,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3939126133918762,
      "step": 30,
      "valid_targets_mean": 2892.2,
      "valid_targets_min": 1503
    },
    {
      "epoch": 0.112,
      "grad_norm": 0.8892773246940794,
      "learning_rate": 8.662420382165606e-06,
      "loss": 0.7494,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.37320828437805176,
      "step": 35,
      "valid_targets_mean": 3300.9,
      "valid_targets_min": 1886
    },
    {
      "epoch": 0.128,
      "grad_norm": 0.7556196584442465,
      "learning_rate": 9.936305732484078e-06,
      "loss": 0.7265,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.355090469121933,
      "step": 40,
      "valid_targets_mean": 3008.4,
      "valid_targets_min": 958
    },
    {
      "epoch": 0.144,
      "grad_norm": 0.585319744877401,
      "learning_rate": 1.1210191082802548e-05,
      "loss": 0.7227,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.33015894889831543,
      "step": 45,
      "valid_targets_mean": 3461.2,
      "valid_targets_min": 1283
    },
    {
      "epoch": 0.16,
      "grad_norm": 0.60549778657053,
      "learning_rate": 1.248407643312102e-05,
      "loss": 0.6643,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3755006790161133,
      "step": 50,
      "valid_targets_mean": 3302.6,
      "valid_targets_min": 1783
    },
    {
      "epoch": 0.176,
      "grad_norm": 0.5913196880977777,
      "learning_rate": 1.375796178343949e-05,
      "loss": 0.6731,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3550243377685547,
      "step": 55,
      "valid_targets_mean": 2947.4,
      "valid_targets_min": 1163
    },
    {
      "epoch": 0.192,
      "grad_norm": 0.48189240583131276,
      "learning_rate": 1.5031847133757964e-05,
      "loss": 0.6445,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24084284901618958,
      "step": 60,
      "valid_targets_mean": 3061.4,
      "valid_targets_min": 904
    },
    {
      "epoch": 0.208,
      "grad_norm": 0.5115869977478013,
      "learning_rate": 1.6305732484076436e-05,
      "loss": 0.6592,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.34680789709091187,
      "step": 65,
      "valid_targets_mean": 3287.8,
      "valid_targets_min": 2333
    },
    {
      "epoch": 0.224,
      "grad_norm": 0.49192101799682225,
      "learning_rate": 1.7579617834394907e-05,
      "loss": 0.6546,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3077095150947571,
      "step": 70,
      "valid_targets_mean": 2793.4,
      "valid_targets_min": 496
    },
    {
      "epoch": 0.24,
      "grad_norm": 0.43541915794559505,
      "learning_rate": 1.8853503184713376e-05,
      "loss": 0.6365,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.29219257831573486,
      "step": 75,
      "valid_targets_mean": 3412.2,
      "valid_targets_min": 1599
    },
    {
      "epoch": 0.256,
      "grad_norm": 0.489452472152694,
      "learning_rate": 2.0127388535031848e-05,
      "loss": 0.6261,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3591298460960388,
      "step": 80,
      "valid_targets_mean": 3012.2,
      "valid_targets_min": 1222
    },
    {
      "epoch": 0.272,
      "grad_norm": 0.5019386951535992,
      "learning_rate": 2.140127388535032e-05,
      "loss": 0.6193,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3157939612865448,
      "step": 85,
      "valid_targets_mean": 2860.2,
      "valid_targets_min": 1524
    },
    {
      "epoch": 0.288,
      "grad_norm": 0.445620357117832,
      "learning_rate": 2.267515923566879e-05,
      "loss": 0.6056,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2639375329017639,
      "step": 90,
      "valid_targets_mean": 3043.7,
      "valid_targets_min": 1624
    },
    {
      "epoch": 0.304,
      "grad_norm": 0.4702040139646772,
      "learning_rate": 2.3949044585987263e-05,
      "loss": 0.593,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3165432810783386,
      "step": 95,
      "valid_targets_mean": 3198.7,
      "valid_targets_min": 1785
    },
    {
      "epoch": 0.32,
      "grad_norm": 0.5070089852236436,
      "learning_rate": 2.5222929936305732e-05,
      "loss": 0.6156,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2870636582374573,
      "step": 100,
      "valid_targets_mean": 2585.8,
      "valid_targets_min": 828
    },
    {
      "epoch": 0.336,
      "grad_norm": 0.4954031096812358,
      "learning_rate": 2.6496815286624204e-05,
      "loss": 0.606,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.30892717838287354,
      "step": 105,
      "valid_targets_mean": 2920.3,
      "valid_targets_min": 1672
    },
    {
      "epoch": 0.352,
      "grad_norm": 0.501722029021901,
      "learning_rate": 2.7770700636942676e-05,
      "loss": 0.5943,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3055849075317383,
      "step": 110,
      "valid_targets_mean": 3229.6,
      "valid_targets_min": 1538
    },
    {
      "epoch": 0.368,
      "grad_norm": 0.4739878082204572,
      "learning_rate": 2.9044585987261148e-05,
      "loss": 0.5894,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24536016583442688,
      "step": 115,
      "valid_targets_mean": 2898.1,
      "valid_targets_min": 1205
    },
    {
      "epoch": 0.384,
      "grad_norm": 0.5550911619314564,
      "learning_rate": 3.0318471337579623e-05,
      "loss": 0.6026,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2870926856994629,
      "step": 120,
      "valid_targets_mean": 2879.2,
      "valid_targets_min": 1154
    },
    {
      "epoch": 0.4,
      "grad_norm": 0.4893227434987029,
      "learning_rate": 3.1592356687898095e-05,
      "loss": 0.585,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26536059379577637,
      "step": 125,
      "valid_targets_mean": 2657.9,
      "valid_targets_min": 1265
    },
    {
      "epoch": 0.416,
      "grad_norm": 0.4851334401577497,
      "learning_rate": 3.286624203821656e-05,
      "loss": 0.5725,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26549556851387024,
      "step": 130,
      "valid_targets_mean": 2905.2,
      "valid_targets_min": 1474
    },
    {
      "epoch": 0.432,
      "grad_norm": 0.46210757965881916,
      "learning_rate": 3.414012738853504e-05,
      "loss": 0.5531,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25114014744758606,
      "step": 135,
      "valid_targets_mean": 2985.4,
      "valid_targets_min": 408
    },
    {
      "epoch": 0.448,
      "grad_norm": 0.4967800770084451,
      "learning_rate": 3.541401273885351e-05,
      "loss": 0.5611,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2707679569721222,
      "step": 140,
      "valid_targets_mean": 3059.8,
      "valid_targets_min": 1204
    },
    {
      "epoch": 0.464,
      "grad_norm": 0.5351660385430722,
      "learning_rate": 3.6687898089171976e-05,
      "loss": 0.5799,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2705070972442627,
      "step": 145,
      "valid_targets_mean": 2721.8,
      "valid_targets_min": 957
    },
    {
      "epoch": 0.48,
      "grad_norm": 0.4785944656217512,
      "learning_rate": 3.796178343949045e-05,
      "loss": 0.5591,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26655346155166626,
      "step": 150,
      "valid_targets_mean": 2875.8,
      "valid_targets_min": 1844
    },
    {
      "epoch": 0.496,
      "grad_norm": 0.4805531707541487,
      "learning_rate": 3.923566878980892e-05,
      "loss": 0.5505,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2606361508369446,
      "step": 155,
      "valid_targets_mean": 3232.9,
      "valid_targets_min": 2114
    },
    {
      "epoch": 0.512,
      "grad_norm": 0.4682743359582423,
      "learning_rate": 3.999980086219931e-05,
      "loss": 0.569,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3145064413547516,
      "step": 160,
      "valid_targets_mean": 3246.5,
      "valid_targets_min": 747
    },
    {
      "epoch": 0.528,
      "grad_norm": 0.49393836582356365,
      "learning_rate": 3.9997560607483595e-05,
      "loss": 0.5474,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2361571043729782,
      "step": 165,
      "valid_targets_mean": 2631.5,
      "valid_targets_min": 775
    },
    {
      "epoch": 0.544,
      "grad_norm": 0.5146422635244974,
      "learning_rate": 3.999283145555291e-05,
      "loss": 0.5428,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2787424921989441,
      "step": 170,
      "valid_targets_mean": 3305.2,
      "valid_targets_min": 1540
    },
    {
      "epoch": 0.56,
      "grad_norm": 0.4912681156590772,
      "learning_rate": 3.998561399499772e-05,
      "loss": 0.5593,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24612675607204437,
      "step": 175,
      "valid_targets_mean": 3071.1,
      "valid_targets_min": 1315
    },
    {
      "epoch": 0.576,
      "grad_norm": 0.4698358070989407,
      "learning_rate": 3.997590912410345e-05,
      "loss": 0.5554,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2255033552646637,
      "step": 180,
      "valid_targets_mean": 3202.2,
      "valid_targets_min": 1394
    },
    {
      "epoch": 0.592,
      "grad_norm": 0.532009785657136,
      "learning_rate": 3.996371805073874e-05,
      "loss": 0.5552,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3014117181301117,
      "step": 185,
      "valid_targets_mean": 3159.7,
      "valid_targets_min": 1578
    },
    {
      "epoch": 0.608,
      "grad_norm": 0.552397732560268,
      "learning_rate": 3.994904229220507e-05,
      "loss": 0.5629,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3167943060398102,
      "step": 190,
      "valid_targets_mean": 3193.0,
      "valid_targets_min": 1288
    },
    {
      "epoch": 0.624,
      "grad_norm": 0.5695077158852391,
      "learning_rate": 3.9931883675047966e-05,
      "loss": 0.5757,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3574255406856537,
      "step": 195,
      "valid_targets_mean": 3208.6,
      "valid_targets_min": 1117
    },
    {
      "epoch": 0.64,
      "grad_norm": 0.5250356068856997,
      "learning_rate": 3.991224433482961e-05,
      "loss": 0.5672,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.27982521057128906,
      "step": 200,
      "valid_targets_mean": 2803.5,
      "valid_targets_min": 1333
    },
    {
      "epoch": 0.656,
      "grad_norm": 0.5191123720236969,
      "learning_rate": 3.98901267158631e-05,
      "loss": 0.5657,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3286457657814026,
      "step": 205,
      "valid_targets_mean": 3622.6,
      "valid_targets_min": 1755
    },
    {
      "epoch": 0.672,
      "grad_norm": 0.46894421435932004,
      "learning_rate": 3.98655335709082e-05,
      "loss": 0.5471,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2534596920013428,
      "step": 210,
      "valid_targets_mean": 3157.9,
      "valid_targets_min": 1070
    },
    {
      "epoch": 0.688,
      "grad_norm": 0.5511548565391978,
      "learning_rate": 3.9838467960828745e-05,
      "loss": 0.5581,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24683982133865356,
      "step": 215,
      "valid_targets_mean": 2883.2,
      "valid_targets_min": 1104
    },
    {
      "epoch": 0.704,
      "grad_norm": 0.4361229714007437,
      "learning_rate": 3.9808933254211665e-05,
      "loss": 0.5488,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2557350993156433,
      "step": 220,
      "valid_targets_mean": 3469.2,
      "valid_targets_min": 1040
    },
    {
      "epoch": 0.72,
      "grad_norm": 0.49040635805208715,
      "learning_rate": 3.977693312694778e-05,
      "loss": 0.559,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2932126820087433,
      "step": 225,
      "valid_targets_mean": 3349.6,
      "valid_targets_min": 1923
    },
    {
      "epoch": 0.736,
      "grad_norm": 0.5358635858571923,
      "learning_rate": 3.974247156177423e-05,
      "loss": 0.5719,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2997221350669861,
      "step": 230,
      "valid_targets_mean": 2815.5,
      "valid_targets_min": 1484
    },
    {
      "epoch": 0.752,
      "grad_norm": 0.4643809623394787,
      "learning_rate": 3.970555284777883e-05,
      "loss": 0.5625,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2824307382106781,
      "step": 235,
      "valid_targets_mean": 3399.6,
      "valid_targets_min": 1375
    },
    {
      "epoch": 0.768,
      "grad_norm": 0.46528415620681496,
      "learning_rate": 3.9666181579866244e-05,
      "loss": 0.5291,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24767976999282837,
      "step": 240,
      "valid_targets_mean": 3110.1,
      "valid_targets_min": 1705
    },
    {
      "epoch": 0.784,
      "grad_norm": 0.9577061650888293,
      "learning_rate": 3.962436265818611e-05,
      "loss": 0.55,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.27534106373786926,
      "step": 245,
      "valid_targets_mean": 3277.2,
      "valid_targets_min": 2021
    },
    {
      "epoch": 0.8,
      "grad_norm": 0.5583793522983518,
      "learning_rate": 3.9580101287523105e-05,
      "loss": 0.5752,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.29335466027259827,
      "step": 250,
      "valid_targets_mean": 2528.5,
      "valid_targets_min": 800
    },
    {
      "epoch": 0.816,
      "grad_norm": 0.4663328652814683,
      "learning_rate": 3.953340297664928e-05,
      "loss": 0.5627,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26815780997276306,
      "step": 255,
      "valid_targets_mean": 3442.6,
      "valid_targets_min": 927
    },
    {
      "epoch": 0.832,
      "grad_norm": 0.4508138830740144,
      "learning_rate": 3.948427353763829e-05,
      "loss": 0.5436,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2782423794269562,
      "step": 260,
      "valid_targets_mean": 3546.3,
      "valid_targets_min": 992
    },
    {
      "epoch": 0.848,
      "grad_norm": 0.46589746239536295,
      "learning_rate": 3.943271908514216e-05,
      "loss": 0.537,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.28761547803878784,
      "step": 265,
      "valid_targets_mean": 3277.4,
      "valid_targets_min": 1505
    },
    {
      "epoch": 0.864,
      "grad_norm": 0.4690148588224579,
      "learning_rate": 3.937874603563015e-05,
      "loss": 0.5646,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2736252546310425,
      "step": 270,
      "valid_targets_mean": 3230.6,
      "valid_targets_min": 1585
    },
    {
      "epoch": 0.88,
      "grad_norm": 0.46883250413065714,
      "learning_rate": 3.932236110659023e-05,
      "loss": 0.5411,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.267468124628067,
      "step": 275,
      "valid_targets_mean": 3123.0,
      "valid_targets_min": 1295
    },
    {
      "epoch": 0.896,
      "grad_norm": 0.4359333014470195,
      "learning_rate": 3.9263571315692976e-05,
      "loss": 0.5289,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2379305064678192,
      "step": 280,
      "valid_targets_mean": 3212.1,
      "valid_targets_min": 1729
    },
    {
      "epoch": 0.912,
      "grad_norm": 0.46436272764415976,
      "learning_rate": 3.920238397991818e-05,
      "loss": 0.5435,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.275387704372406,
      "step": 285,
      "valid_targets_mean": 3243.3,
      "valid_targets_min": 1626
    },
    {
      "epoch": 0.928,
      "grad_norm": 0.448182212151269,
      "learning_rate": 3.913880671464418e-05,
      "loss": 0.5399,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24673345685005188,
      "step": 290,
      "valid_targets_mean": 3588.0,
      "valid_targets_min": 1478
    },
    {
      "epoch": 0.944,
      "grad_norm": 0.4982702796643091,
      "learning_rate": 3.907284743270001e-05,
      "loss": 0.5478,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.260049045085907,
      "step": 295,
      "valid_targets_mean": 2819.7,
      "valid_targets_min": 1732
    },
    {
      "epoch": 0.96,
      "grad_norm": 0.5212191794351084,
      "learning_rate": 3.900451434338062e-05,
      "loss": 0.5497,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2794175446033478,
      "step": 300,
      "valid_targets_mean": 3049.4,
      "valid_targets_min": 1476
    },
    {
      "epoch": 0.976,
      "grad_norm": 0.4494652664686231,
      "learning_rate": 3.893381595142511e-05,
      "loss": 0.5379,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26482099294662476,
      "step": 305,
      "valid_targets_mean": 3424.4,
      "valid_targets_min": 1810
    },
    {
      "epoch": 0.992,
      "grad_norm": 0.4473537869984785,
      "learning_rate": 3.886076105595825e-05,
      "loss": 0.537,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24100354313850403,
      "step": 310,
      "valid_targets_mean": 2968.4,
      "valid_targets_min": 1734
    },
    {
      "epoch": 1.0064,
      "grad_norm": 0.49277910872631514,
      "learning_rate": 3.878535874939532e-05,
      "loss": 0.5475,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2600662410259247,
      "step": 315,
      "valid_targets_mean": 2880.3,
      "valid_targets_min": 1373
    },
    {
      "epoch": 1.0224,
      "grad_norm": 0.4962924059392822,
      "learning_rate": 3.870761841631051e-05,
      "loss": 0.5325,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2709118127822876,
      "step": 320,
      "valid_targets_mean": 3084.9,
      "valid_targets_min": 1101
    },
    {
      "epoch": 1.0384,
      "grad_norm": 0.4765894000972697,
      "learning_rate": 3.862754973226887e-05,
      "loss": 0.5105,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2291407436132431,
      "step": 325,
      "valid_targets_mean": 2385.7,
      "valid_targets_min": 1193
    },
    {
      "epoch": 1.0544,
      "grad_norm": 0.44554031986434506,
      "learning_rate": 3.85451626626221e-05,
      "loss": 0.5333,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25396525859832764,
      "step": 330,
      "valid_targets_mean": 3322.8,
      "valid_targets_min": 1730
    },
    {
      "epoch": 1.0704,
      "grad_norm": 0.4203337288517843,
      "learning_rate": 3.846046746126827e-05,
      "loss": 0.512,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2684890627861023,
      "step": 335,
      "valid_targets_mean": 3894.3,
      "valid_targets_min": 1782
    },
    {
      "epoch": 1.0864,
      "grad_norm": 0.475098411041102,
      "learning_rate": 3.837347466937562e-05,
      "loss": 0.4937,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24573341012001038,
      "step": 340,
      "valid_targets_mean": 3154.6,
      "valid_targets_min": 1349
    },
    {
      "epoch": 1.1024,
      "grad_norm": 0.5827683965056836,
      "learning_rate": 3.828419511407062e-05,
      "loss": 0.5493,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2953265607357025,
      "step": 345,
      "valid_targets_mean": 2745.6,
      "valid_targets_min": 1101
    },
    {
      "epoch": 1.1184,
      "grad_norm": 0.5124538925155401,
      "learning_rate": 3.819263990709037e-05,
      "loss": 0.511,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.27043575048446655,
      "step": 350,
      "valid_targets_mean": 3445.8,
      "valid_targets_min": 702
    },
    {
      "epoch": 1.1344,
      "grad_norm": 0.47607071622443325,
      "learning_rate": 3.809882044339971e-05,
      "loss": 0.5367,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24829918146133423,
      "step": 355,
      "valid_targets_mean": 2760.6,
      "valid_targets_min": 1474
    },
    {
      "epoch": 1.1504,
      "grad_norm": 0.5891156574165097,
      "learning_rate": 3.800274839977293e-05,
      "loss": 0.5211,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24416720867156982,
      "step": 360,
      "valid_targets_mean": 2987.4,
      "valid_targets_min": 1684
    },
    {
      "epoch": 1.1663999999999999,
      "grad_norm": 0.4383115726025833,
      "learning_rate": 3.790443573334055e-05,
      "loss": 0.5167,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25110000371932983,
      "step": 365,
      "valid_targets_mean": 2802.4,
      "valid_targets_min": 1491
    },
    {
      "epoch": 1.1824,
      "grad_norm": 0.4659780007043588,
      "learning_rate": 3.780389468010106e-05,
      "loss": 0.5442,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2537109851837158,
      "step": 370,
      "valid_targets_mean": 3175.0,
      "valid_targets_min": 1648
    },
    {
      "epoch": 1.1984,
      "grad_norm": 0.41971726808284665,
      "learning_rate": 3.7701137753398075e-05,
      "loss": 0.5322,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2624451816082001,
      "step": 375,
      "valid_targets_mean": 4199.4,
      "valid_targets_min": 1896
    },
    {
      "epoch": 1.2144,
      "grad_norm": 0.45669985830428217,
      "learning_rate": 3.759617774236292e-05,
      "loss": 0.5159,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25252872705459595,
      "step": 380,
      "valid_targets_mean": 3041.3,
      "valid_targets_min": 957
    },
    {
      "epoch": 1.2304,
      "grad_norm": 0.5219353962349923,
      "learning_rate": 3.748902771032288e-05,
      "loss": 0.5244,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2761985659599304,
      "step": 385,
      "valid_targets_mean": 3167.9,
      "valid_targets_min": 1403
    },
    {
      "epoch": 1.2464,
      "grad_norm": 0.4599125403809234,
      "learning_rate": 3.737970099317535e-05,
      "loss": 0.5291,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25756001472473145,
      "step": 390,
      "valid_targets_mean": 3301.4,
      "valid_targets_min": 1168
    },
    {
      "epoch": 1.2624,
      "grad_norm": 0.5264391888700657,
      "learning_rate": 3.726821119772803e-05,
      "loss": 0.5153,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25718802213668823,
      "step": 395,
      "valid_targets_mean": 2716.5,
      "valid_targets_min": 1755
    },
    {
      "epoch": 1.2784,
      "grad_norm": 0.5265175609805931,
      "learning_rate": 3.7154572200005446e-05,
      "loss": 0.513,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2895033359527588,
      "step": 400,
      "valid_targets_mean": 2866.0,
      "valid_targets_min": 1130
    },
    {
      "epoch": 1.2944,
      "grad_norm": 0.519454299106432,
      "learning_rate": 3.703879814352193e-05,
      "loss": 0.5057,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2757456600666046,
      "step": 405,
      "valid_targets_mean": 2979.6,
      "valid_targets_min": 957
    },
    {
      "epoch": 1.3104,
      "grad_norm": 0.4815016307644221,
      "learning_rate": 3.6920903437521305e-05,
      "loss": 0.5227,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2794150412082672,
      "step": 410,
      "valid_targets_mean": 3101.7,
      "valid_targets_min": 1352
    },
    {
      "epoch": 1.3264,
      "grad_norm": 0.5023423296160917,
      "learning_rate": 3.680090275518352e-05,
      "loss": 0.5242,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.29537904262542725,
      "step": 415,
      "valid_targets_mean": 2789.9,
      "valid_targets_min": 1205
    },
    {
      "epoch": 1.3424,
      "grad_norm": 0.4281900261017907,
      "learning_rate": 3.667881103179844e-05,
      "loss": 0.5068,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23545397818088531,
      "step": 420,
      "valid_targets_mean": 3231.1,
      "valid_targets_min": 1014
    },
    {
      "epoch": 1.3584,
      "grad_norm": 0.43510100417753445,
      "learning_rate": 3.655464346290697e-05,
      "loss": 0.5173,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26157039403915405,
      "step": 425,
      "valid_targets_mean": 3339.4,
      "valid_targets_min": 1406
    },
    {
      "epoch": 1.3744,
      "grad_norm": 0.4450583053213791,
      "learning_rate": 3.642841550240983e-05,
      "loss": 0.4996,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2654106020927429,
      "step": 430,
      "valid_targets_mean": 3275.6,
      "valid_targets_min": 1338
    },
    {
      "epoch": 1.3904,
      "grad_norm": 0.49405570189398096,
      "learning_rate": 3.630014286064419e-05,
      "loss": 0.5212,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25802505016326904,
      "step": 435,
      "valid_targets_mean": 3087.0,
      "valid_targets_min": 1392
    },
    {
      "epoch": 1.4064,
      "grad_norm": 0.520492861665326,
      "learning_rate": 3.6169841502428285e-05,
      "loss": 0.5057,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2416737824678421,
      "step": 440,
      "valid_targets_mean": 2791.6,
      "valid_targets_min": 1323
    },
    {
      "epoch": 1.4224,
      "grad_norm": 0.4559395788149373,
      "learning_rate": 3.603752764507454e-05,
      "loss": 0.5261,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2236054390668869,
      "step": 445,
      "valid_targets_mean": 3045.8,
      "valid_targets_min": 975
    },
    {
      "epoch": 1.4384000000000001,
      "grad_norm": 0.487958859838727,
      "learning_rate": 3.5903217756371066e-05,
      "loss": 0.5189,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.256151407957077,
      "step": 450,
      "valid_targets_mean": 3211.8,
      "valid_targets_min": 1425
    },
    {
      "epoch": 1.4544000000000001,
      "grad_norm": 6.026861812432653,
      "learning_rate": 3.576692855253213e-05,
      "loss": 0.5059,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.233452707529068,
      "step": 455,
      "valid_targets_mean": 3006.8,
      "valid_targets_min": 693
    },
    {
      "epoch": 1.4704,
      "grad_norm": 0.44925860522978966,
      "learning_rate": 3.562867699611764e-05,
      "loss": 0.5316,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26866137981414795,
      "step": 460,
      "valid_targets_mean": 3145.0,
      "valid_targets_min": 1974
    },
    {
      "epoch": 1.4864,
      "grad_norm": 0.43748519005117936,
      "learning_rate": 3.5488480293922e-05,
      "loss": 0.5127,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2456567883491516,
      "step": 465,
      "valid_targets_mean": 3107.4,
      "valid_targets_min": 1723
    },
    {
      "epoch": 1.5024,
      "grad_norm": 0.5485416488660548,
      "learning_rate": 3.5346355894832515e-05,
      "loss": 0.5263,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2427314817905426,
      "step": 470,
      "valid_targets_mean": 2881.4,
      "valid_targets_min": 1071
    },
    {
      "epoch": 1.5184,
      "grad_norm": 0.5492193241718348,
      "learning_rate": 3.520232148765774e-05,
      "loss": 0.5185,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24047909677028656,
      "step": 475,
      "valid_targets_mean": 3304.2,
      "valid_targets_min": 628
    },
    {
      "epoch": 1.5344,
      "grad_norm": 0.509030601412652,
      "learning_rate": 3.505639499892591e-05,
      "loss": 0.5235,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2926231026649475,
      "step": 480,
      "valid_targets_mean": 3097.2,
      "valid_targets_min": 1785
    },
    {
      "epoch": 1.5504,
      "grad_norm": 0.5556023004390762,
      "learning_rate": 3.490859459065382e-05,
      "loss": 0.4994,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24574291706085205,
      "step": 485,
      "valid_targets_mean": 2863.5,
      "valid_targets_min": 938
    },
    {
      "epoch": 1.5664,
      "grad_norm": 0.4355933250890639,
      "learning_rate": 3.475893865808633e-05,
      "loss": 0.5007,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21669138967990875,
      "step": 490,
      "valid_targets_mean": 2690.6,
      "valid_targets_min": 880
    },
    {
      "epoch": 1.5824,
      "grad_norm": 0.5435140058892103,
      "learning_rate": 3.4607445827406984e-05,
      "loss": 0.5201,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2567306160926819,
      "step": 495,
      "valid_targets_mean": 2844.2,
      "valid_targets_min": 1578
    },
    {
      "epoch": 1.5984,
      "grad_norm": 0.5099231583497758,
      "learning_rate": 3.445413495341971e-05,
      "loss": 0.514,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.259105920791626,
      "step": 500,
      "valid_targets_mean": 3156.5,
      "valid_targets_min": 1873
    },
    {
      "epoch": 1.6143999999999998,
      "grad_norm": 0.4737224278917,
      "learning_rate": 3.429902511720216e-05,
      "loss": 0.5176,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.27495771646499634,
      "step": 505,
      "valid_targets_mean": 2925.5,
      "valid_targets_min": 1313
    },
    {
      "epoch": 1.6303999999999998,
      "grad_norm": 0.4904495643158072,
      "learning_rate": 3.4142135623730954e-05,
      "loss": 0.5115,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2774983048439026,
      "step": 510,
      "valid_targets_mean": 2726.0,
      "valid_targets_min": 909
    },
    {
      "epoch": 1.6463999999999999,
      "grad_norm": 0.42906224936035436,
      "learning_rate": 3.398348599947888e-05,
      "loss": 0.4951,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2206537127494812,
      "step": 515,
      "valid_targets_mean": 3334.7,
      "valid_targets_min": 1466
    },
    {
      "epoch": 1.6623999999999999,
      "grad_norm": 0.504508105224736,
      "learning_rate": 3.3823095989984697e-05,
      "loss": 0.53,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3056814670562744,
      "step": 520,
      "valid_targets_mean": 2991.4,
      "valid_targets_min": 1933
    },
    {
      "epoch": 1.6784,
      "grad_norm": 0.5351202359792724,
      "learning_rate": 3.366098555739557e-05,
      "loss": 0.5188,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26875633001327515,
      "step": 525,
      "valid_targets_mean": 3094.2,
      "valid_targets_min": 1153
    },
    {
      "epoch": 1.6944,
      "grad_norm": 0.4573333991642385,
      "learning_rate": 3.349717487798261e-05,
      "loss": 0.5158,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22667232155799866,
      "step": 530,
      "valid_targets_mean": 2724.5,
      "valid_targets_min": 977
    },
    {
      "epoch": 1.7104,
      "grad_norm": 0.4661590895939061,
      "learning_rate": 3.3331684339629706e-05,
      "loss": 0.5164,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2817540466785431,
      "step": 535,
      "valid_targets_mean": 3025.5,
      "valid_targets_min": 1911
    },
    {
      "epoch": 1.7264,
      "grad_norm": 0.5316604683020004,
      "learning_rate": 3.3164534539296056e-05,
      "loss": 0.4904,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24950800836086273,
      "step": 540,
      "valid_targets_mean": 2558.2,
      "valid_targets_min": 1381
    },
    {
      "epoch": 1.7424,
      "grad_norm": 0.49226701109131127,
      "learning_rate": 3.299574628045269e-05,
      "loss": 0.5237,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2845689654350281,
      "step": 545,
      "valid_targets_mean": 3183.1,
      "valid_targets_min": 1391
    },
    {
      "epoch": 1.7584,
      "grad_norm": 0.4935962099638135,
      "learning_rate": 3.282534057049322e-05,
      "loss": 0.5111,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.27177274227142334,
      "step": 550,
      "valid_targets_mean": 2734.2,
      "valid_targets_min": 1313
    },
    {
      "epoch": 1.7744,
      "grad_norm": 0.5019233751841419,
      "learning_rate": 3.265333861811933e-05,
      "loss": 0.5237,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.29612475633621216,
      "step": 555,
      "valid_targets_mean": 3097.1,
      "valid_targets_min": 2044
    },
    {
      "epoch": 1.7904,
      "grad_norm": 0.4577395132839231,
      "learning_rate": 3.2479761830701075e-05,
      "loss": 0.5109,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2313130497932434,
      "step": 560,
      "valid_targets_mean": 2743.2,
      "valid_targets_min": 1329
    },
    {
      "epoch": 1.8064,
      "grad_norm": 0.46288866982137783,
      "learning_rate": 3.230463181161254e-05,
      "loss": 0.5006,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2379888892173767,
      "step": 565,
      "valid_targets_mean": 3225.7,
      "valid_targets_min": 1041
    },
    {
      "epoch": 1.8224,
      "grad_norm": 0.5980963250881742,
      "learning_rate": 3.212797035754311e-05,
      "loss": 0.5095,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23481547832489014,
      "step": 570,
      "valid_targets_mean": 2868.9,
      "valid_targets_min": 667
    },
    {
      "epoch": 1.8384,
      "grad_norm": 0.45869991131611926,
      "learning_rate": 3.194979945578461e-05,
      "loss": 0.5157,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22095608711242676,
      "step": 575,
      "valid_targets_mean": 2974.4,
      "valid_targets_min": 884
    },
    {
      "epoch": 1.8544,
      "grad_norm": 0.4781367450330346,
      "learning_rate": 3.177014128149479e-05,
      "loss": 0.5269,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2355835884809494,
      "step": 580,
      "valid_targets_mean": 2904.4,
      "valid_targets_min": 1593
    },
    {
      "epoch": 1.8704,
      "grad_norm": 0.5218205842284931,
      "learning_rate": 3.158901819493742e-05,
      "loss": 0.5335,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.30380451679229736,
      "step": 585,
      "valid_targets_mean": 2723.4,
      "valid_targets_min": 1488
    },
    {
      "epoch": 1.8864,
      "grad_norm": 0.49357175630318795,
      "learning_rate": 3.1406452738699284e-05,
      "loss": 0.5215,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26988309621810913,
      "step": 590,
      "valid_targets_mean": 2617.8,
      "valid_targets_min": 1034
    },
    {
      "epoch": 1.9024,
      "grad_norm": 0.44827271739585833,
      "learning_rate": 3.122246763488457e-05,
      "loss": 0.4997,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22802028059959412,
      "step": 595,
      "valid_targets_mean": 2838.8,
      "valid_targets_min": 1163
    },
    {
      "epoch": 1.9184,
      "grad_norm": 0.4632124014095409,
      "learning_rate": 3.103708578228686e-05,
      "loss": 0.4991,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24168995022773743,
      "step": 600,
      "valid_targets_mean": 2863.8,
      "valid_targets_min": 1428
    },
    {
      "epoch": 1.9344000000000001,
      "grad_norm": 0.7371275123922867,
      "learning_rate": 3.085033025353915e-05,
      "loss": 0.5124,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2760087251663208,
      "step": 605,
      "valid_targets_mean": 3405.6,
      "valid_targets_min": 2083
    },
    {
      "epoch": 1.9504000000000001,
      "grad_norm": 0.4756614197293939,
      "learning_rate": 3.066222429224221e-05,
      "loss": 0.5145,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24378544092178345,
      "step": 610,
      "valid_targets_mean": 2794.1,
      "valid_targets_min": 1296
    },
    {
      "epoch": 1.9664000000000001,
      "grad_norm": 0.42778554593881357,
      "learning_rate": 3.047279131007173e-05,
      "loss": 0.5025,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2605063021183014,
      "step": 615,
      "valid_targets_mean": 3072.7,
      "valid_targets_min": 1641
    },
    {
      "epoch": 1.9824000000000002,
      "grad_norm": 0.48736282270910075,
      "learning_rate": 3.0282054883864434e-05,
      "loss": 0.4881,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23605525493621826,
      "step": 620,
      "valid_targets_mean": 3135.9,
      "valid_targets_min": 1347
    },
    {
      "epoch": 1.9984,
      "grad_norm": 0.4800464610927692,
      "learning_rate": 3.009003875268379e-05,
      "loss": 0.4943,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2850508391857147,
      "step": 625,
      "valid_targets_mean": 3301.7,
      "valid_targets_min": 1878
    },
    {
      "epoch": 2.0128,
      "grad_norm": 0.5007940198017256,
      "learning_rate": 2.9896766814865355e-05,
      "loss": 0.4868,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23529338836669922,
      "step": 630,
      "valid_targets_mean": 2643.2,
      "valid_targets_min": 1284
    },
    {
      "epoch": 2.0288,
      "grad_norm": 0.5157074650090006,
      "learning_rate": 2.970226312504246e-05,
      "loss": 0.4948,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.27678927779197693,
      "step": 635,
      "valid_targets_mean": 2965.7,
      "valid_targets_min": 1415
    },
    {
      "epoch": 2.0448,
      "grad_norm": 0.5101562609902904,
      "learning_rate": 2.9506551891152334e-05,
      "loss": 0.4968,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25116124749183655,
      "step": 640,
      "valid_targets_mean": 2771.2,
      "valid_targets_min": 689
    },
    {
      "epoch": 2.0608,
      "grad_norm": 0.4529563499288533,
      "learning_rate": 2.930965747142319e-05,
      "loss": 0.4813,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20527660846710205,
      "step": 645,
      "valid_targets_mean": 2966.9,
      "valid_targets_min": 1239
    },
    {
      "epoch": 2.0768,
      "grad_norm": 0.45826327058683025,
      "learning_rate": 2.9111604371342593e-05,
      "loss": 0.4964,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22757063806056976,
      "step": 650,
      "valid_targets_mean": 2937.9,
      "valid_targets_min": 1403
    },
    {
      "epoch": 2.0928,
      "grad_norm": 0.4572314921940725,
      "learning_rate": 2.891241724060752e-05,
      "loss": 0.4649,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24460506439208984,
      "step": 655,
      "valid_targets_mean": 3235.1,
      "valid_targets_min": 1409
    },
    {
      "epoch": 2.1088,
      "grad_norm": 0.5057949551121218,
      "learning_rate": 2.8712120870056455e-05,
      "loss": 0.5008,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.27103084325790405,
      "step": 660,
      "valid_targets_mean": 2827.2,
      "valid_targets_min": 1838
    },
    {
      "epoch": 2.1248,
      "grad_norm": 0.434456854498136,
      "learning_rate": 2.851074018858389e-05,
      "loss": 0.4846,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23950263857841492,
      "step": 665,
      "valid_targets_mean": 3572.2,
      "valid_targets_min": 1233
    },
    {
      "epoch": 2.1408,
      "grad_norm": 0.48614035866392213,
      "learning_rate": 2.8308300260037734e-05,
      "loss": 0.5032,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23101502656936646,
      "step": 670,
      "valid_targets_mean": 2830.4,
      "valid_targets_min": 1506
    },
    {
      "epoch": 2.1568,
      "grad_norm": 0.44626305596576143,
      "learning_rate": 2.8104826280099796e-05,
      "loss": 0.4932,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25110024213790894,
      "step": 675,
      "valid_targets_mean": 3202.5,
      "valid_targets_min": 1848
    },
    {
      "epoch": 2.1728,
      "grad_norm": 0.4472844696645451,
      "learning_rate": 2.7900343573150003e-05,
      "loss": 0.486,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24017497897148132,
      "step": 680,
      "valid_targets_mean": 3204.7,
      "valid_targets_min": 1108
    },
    {
      "epoch": 2.1888,
      "grad_norm": 0.5343225728081022,
      "learning_rate": 2.7694877589114442e-05,
      "loss": 0.4923,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24907417595386505,
      "step": 685,
      "valid_targets_mean": 2942.4,
      "valid_targets_min": 1725
    },
    {
      "epoch": 2.2048,
      "grad_norm": 0.47468755794982304,
      "learning_rate": 2.748845390029794e-05,
      "loss": 0.4857,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22810614109039307,
      "step": 690,
      "valid_targets_mean": 2807.8,
      "valid_targets_min": 800
    },
    {
      "epoch": 2.2208,
      "grad_norm": 0.4826995803470774,
      "learning_rate": 2.728109819820129e-05,
      "loss": 0.5068,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2800004780292511,
      "step": 695,
      "valid_targets_mean": 3410.8,
      "valid_targets_min": 1240
    },
    {
      "epoch": 2.2368,
      "grad_norm": 0.4592512476126212,
      "learning_rate": 2.7072836290323698e-05,
      "loss": 0.4836,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2372513711452484,
      "step": 700,
      "valid_targets_mean": 3405.4,
      "valid_targets_min": 1309
    },
    {
      "epoch": 2.2528,
      "grad_norm": 0.4123298917676354,
      "learning_rate": 2.6863694096950763e-05,
      "loss": 0.4604,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2027531862258911,
      "step": 705,
      "valid_targets_mean": 3237.2,
      "valid_targets_min": 1600
    },
    {
      "epoch": 2.2688,
      "grad_norm": 0.4543061236567951,
      "learning_rate": 2.6653697647928485e-05,
      "loss": 0.4926,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2540218234062195,
      "step": 710,
      "valid_targets_mean": 2959.7,
      "valid_targets_min": 1461
    },
    {
      "epoch": 2.2848,
      "grad_norm": 0.5076008394558799,
      "learning_rate": 2.644287307942352e-05,
      "loss": 0.4793,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24240398406982422,
      "step": 715,
      "valid_targets_mean": 2681.1,
      "valid_targets_min": 1099
    },
    {
      "epoch": 2.3008,
      "grad_norm": 0.4089060269005043,
      "learning_rate": 2.623124663067034e-05,
      "loss": 0.4856,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23559370636940002,
      "step": 720,
      "valid_targets_mean": 3506.2,
      "valid_targets_min": 1550
    },
    {
      "epoch": 2.3168,
      "grad_norm": 0.4551212609304956,
      "learning_rate": 2.6018844640705448e-05,
      "loss": 0.4827,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23695677518844604,
      "step": 725,
      "valid_targets_mean": 3120.4,
      "valid_targets_min": 1065
    },
    {
      "epoch": 2.3327999999999998,
      "grad_norm": 0.4841751536620542,
      "learning_rate": 2.580569354508925e-05,
      "loss": 0.4825,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20382042229175568,
      "step": 730,
      "valid_targets_mean": 2650.9,
      "valid_targets_min": 881
    },
    {
      "epoch": 2.3487999999999998,
      "grad_norm": 0.4581740603290819,
      "learning_rate": 2.5591819872615856e-05,
      "loss": 0.4635,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23988109827041626,
      "step": 735,
      "valid_targets_mean": 2943.5,
      "valid_targets_min": 975
    },
    {
      "epoch": 2.3648,
      "grad_norm": 0.4378179137522697,
      "learning_rate": 2.5377250242011338e-05,
      "loss": 0.488,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2582421898841858,
      "step": 740,
      "valid_targets_mean": 3500.6,
      "valid_targets_min": 1040
    },
    {
      "epoch": 2.3808,
      "grad_norm": 0.47224573930300273,
      "learning_rate": 2.516201135862073e-05,
      "loss": 0.4754,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22435212135314941,
      "step": 745,
      "valid_targets_mean": 2937.9,
      "valid_targets_min": 1579
    },
    {
      "epoch": 2.3968,
      "grad_norm": 0.49031185792544935,
      "learning_rate": 2.494613001108431e-05,
      "loss": 0.4928,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25457414984703064,
      "step": 750,
      "valid_targets_mean": 3221.4,
      "valid_targets_min": 1247
    },
    {
      "epoch": 2.4128,
      "grad_norm": 0.5242476263518132,
      "learning_rate": 2.4729633068003466e-05,
      "loss": 0.4789,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24022498726844788,
      "step": 755,
      "valid_targets_mean": 2896.2,
      "valid_targets_min": 1320
    },
    {
      "epoch": 2.4288,
      "grad_norm": 0.4866222176129818,
      "learning_rate": 2.4512547474596624e-05,
      "loss": 0.5092,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2546798288822174,
      "step": 760,
      "valid_targets_mean": 3466.6,
      "valid_targets_min": 1631
    },
    {
      "epoch": 2.4448,
      "grad_norm": 0.5137377843083281,
      "learning_rate": 2.429490024934566e-05,
      "loss": 0.4808,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2947126030921936,
      "step": 765,
      "valid_targets_mean": 2776.8,
      "valid_targets_min": 947
    },
    {
      "epoch": 2.4608,
      "grad_norm": 0.4869533055861981,
      "learning_rate": 2.4076718480633178e-05,
      "loss": 0.4933,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20823511481285095,
      "step": 770,
      "valid_targets_mean": 2971.9,
      "valid_targets_min": 1278
    },
    {
      "epoch": 2.4768,
      "grad_norm": 0.45539327422923814,
      "learning_rate": 2.3858029323371067e-05,
      "loss": 0.5015,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20980717241764069,
      "step": 775,
      "valid_targets_mean": 2973.5,
      "valid_targets_min": 1445
    },
    {
      "epoch": 2.4928,
      "grad_norm": 0.5287411648228943,
      "learning_rate": 2.363885999562084e-05,
      "loss": 0.4865,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23313599824905396,
      "step": 780,
      "valid_targets_mean": 3004.2,
      "valid_targets_min": 969
    },
    {
      "epoch": 2.5088,
      "grad_norm": 0.4543483221665911,
      "learning_rate": 2.3419237775206026e-05,
      "loss": 0.5005,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.28633490204811096,
      "step": 785,
      "valid_targets_mean": 3227.5,
      "valid_targets_min": 985
    },
    {
      "epoch": 2.5248,
      "grad_norm": 0.4651724232555382,
      "learning_rate": 2.3199189996317205e-05,
      "loss": 0.487,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24572448432445526,
      "step": 790,
      "valid_targets_mean": 2907.9,
      "valid_targets_min": 1056
    },
    {
      "epoch": 2.5408,
      "grad_norm": 0.49088060050512494,
      "learning_rate": 2.297874404610998e-05,
      "loss": 0.4848,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24726298451423645,
      "step": 795,
      "valid_targets_mean": 2909.0,
      "valid_targets_min": 1397
    },
    {
      "epoch": 2.5568,
      "grad_norm": 0.4412421616860558,
      "learning_rate": 2.2757927361296376e-05,
      "loss": 0.5024,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2554225027561188,
      "step": 800,
      "valid_targets_mean": 3597.6,
      "valid_targets_min": 1393
    },
    {
      "epoch": 2.5728,
      "grad_norm": 0.4511584722794911,
      "learning_rate": 2.2536767424730052e-05,
      "loss": 0.4783,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25293684005737305,
      "step": 805,
      "valid_targets_mean": 3084.6,
      "valid_targets_min": 1588
    },
    {
      "epoch": 2.5888,
      "grad_norm": 0.504484886615669,
      "learning_rate": 2.2315291761985803e-05,
      "loss": 0.4975,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23480279743671417,
      "step": 810,
      "valid_targets_mean": 3016.9,
      "valid_targets_min": 1040
    },
    {
      "epoch": 2.6048,
      "grad_norm": 0.49526946333824134,
      "learning_rate": 2.2093527937933716e-05,
      "loss": 0.4726,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2735830545425415,
      "step": 815,
      "valid_targets_mean": 3600.4,
      "valid_targets_min": 1429
    },
    {
      "epoch": 2.6208,
      "grad_norm": 0.49005069499970527,
      "learning_rate": 2.1871503553308447e-05,
      "loss": 0.5052,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24496214091777802,
      "step": 820,
      "valid_targets_mean": 2981.4,
      "valid_targets_min": 1634
    },
    {
      "epoch": 2.6368,
      "grad_norm": 0.4596594761364903,
      "learning_rate": 2.164924624127403e-05,
      "loss": 0.5024,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25404804944992065,
      "step": 825,
      "valid_targets_mean": 3347.8,
      "valid_targets_min": 1674
    },
    {
      "epoch": 2.6528,
      "grad_norm": 0.46699363461367677,
      "learning_rate": 2.1426783663984648e-05,
      "loss": 0.4922,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24780279397964478,
      "step": 830,
      "valid_targets_mean": 3289.5,
      "valid_targets_min": 1097
    },
    {
      "epoch": 2.6688,
      "grad_norm": 0.4530913646330165,
      "learning_rate": 2.1204143509141818e-05,
      "loss": 0.4895,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23123887181282043,
      "step": 835,
      "valid_targets_mean": 3102.1,
      "valid_targets_min": 1058
    },
    {
      "epoch": 2.6848,
      "grad_norm": 0.48488083478979,
      "learning_rate": 2.0981353486548363e-05,
      "loss": 0.5005,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.29003238677978516,
      "step": 840,
      "valid_targets_mean": 3341.6,
      "valid_targets_min": 2051
    },
    {
      "epoch": 2.7008,
      "grad_norm": 0.50927260798858,
      "learning_rate": 2.075844132465964e-05,
      "loss": 0.4862,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.262538343667984,
      "step": 845,
      "valid_targets_mean": 3244.6,
      "valid_targets_min": 1111
    },
    {
      "epoch": 2.7168,
      "grad_norm": 0.483799594883633,
      "learning_rate": 2.0535434767132495e-05,
      "loss": 0.4741,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24056734144687653,
      "step": 850,
      "valid_targets_mean": 2681.6,
      "valid_targets_min": 1340
    },
    {
      "epoch": 2.7328,
      "grad_norm": 0.4614773306975341,
      "learning_rate": 2.0312361569372215e-05,
      "loss": 0.4723,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24297058582305908,
      "step": 855,
      "valid_targets_mean": 3447.4,
      "valid_targets_min": 2402
    },
    {
      "epoch": 2.7488,
      "grad_norm": 0.47344331595599315,
      "learning_rate": 2.0089249495078186e-05,
      "loss": 0.4977,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2512054443359375,
      "step": 860,
      "valid_targets_mean": 2892.4,
      "valid_targets_min": 1023
    },
    {
      "epoch": 2.7648,
      "grad_norm": 0.4312360053420849,
      "learning_rate": 1.9866126312788333e-05,
      "loss": 0.4805,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2499227225780487,
      "step": 865,
      "valid_targets_mean": 3342.8,
      "valid_targets_min": 667
    },
    {
      "epoch": 2.7808,
      "grad_norm": 0.4804917083313915,
      "learning_rate": 1.964301979242308e-05,
      "loss": 0.488,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23908847570419312,
      "step": 870,
      "valid_targets_mean": 2947.2,
      "valid_targets_min": 1752
    },
    {
      "epoch": 2.7968,
      "grad_norm": 0.8762633926530743,
      "learning_rate": 1.9419957701829138e-05,
      "loss": 0.485,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2324860394001007,
      "step": 875,
      "valid_targets_mean": 3017.0,
      "valid_targets_min": 965
    },
    {
      "epoch": 2.8128,
      "grad_norm": 0.4556434340662145,
      "learning_rate": 1.9196967803323464e-05,
      "loss": 0.4974,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23595423996448517,
      "step": 880,
      "valid_targets_mean": 3213.3,
      "valid_targets_min": 884
    },
    {
      "epoch": 2.8288,
      "grad_norm": 0.4393572955542803,
      "learning_rate": 1.8974077850237983e-05,
      "loss": 0.4754,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2087182253599167,
      "step": 885,
      "valid_targets_mean": 2638.4,
      "valid_targets_min": 955
    },
    {
      "epoch": 2.8448,
      "grad_norm": 0.46271408791167346,
      "learning_rate": 1.875131558346542e-05,
      "loss": 0.485,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23028402030467987,
      "step": 890,
      "valid_targets_mean": 3273.6,
      "valid_targets_min": 1544
    },
    {
      "epoch": 2.8608000000000002,
      "grad_norm": 0.4513553889559316,
      "learning_rate": 1.8528708728006654e-05,
      "loss": 0.4693,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21357670426368713,
      "step": 895,
      "valid_targets_mean": 2945.9,
      "valid_targets_min": 985
    },
    {
      "epoch": 2.8768000000000002,
      "grad_norm": 0.4397598056676539,
      "learning_rate": 1.8306284989520055e-05,
      "loss": 0.4913,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2715904116630554,
      "step": 900,
      "valid_targets_mean": 3446.7,
      "valid_targets_min": 1012
    },
    {
      "epoch": 2.8928000000000003,
      "grad_norm": 0.4842128353083839,
      "learning_rate": 1.8084072050873265e-05,
      "loss": 0.4821,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22894605994224548,
      "step": 905,
      "valid_targets_mean": 2792.0,
      "valid_targets_min": 1913
    },
    {
      "epoch": 2.9088000000000003,
      "grad_norm": 0.46238383721407217,
      "learning_rate": 1.786209756869775e-05,
      "loss": 0.4701,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2544752359390259,
      "step": 910,
      "valid_targets_mean": 3167.1,
      "valid_targets_min": 2189
    },
    {
      "epoch": 2.9248,
      "grad_norm": 0.47372661840165786,
      "learning_rate": 1.764038916994669e-05,
      "loss": 0.4818,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2640783488750458,
      "step": 915,
      "valid_targets_mean": 3044.8,
      "valid_targets_min": 1011
    },
    {
      "epoch": 2.9408,
      "grad_norm": 0.6128265221061115,
      "learning_rate": 1.741897444845649e-05,
      "loss": 0.4837,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.27274250984191895,
      "step": 920,
      "valid_targets_mean": 3361.2,
      "valid_targets_min": 1932
    },
    {
      "epoch": 2.9568,
      "grad_norm": 0.4542712376697493,
      "learning_rate": 1.7197880961512498e-05,
      "loss": 0.4857,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25496718287467957,
      "step": 925,
      "valid_targets_mean": 2903.7,
      "valid_targets_min": 749
    },
    {
      "epoch": 2.9728,
      "grad_norm": 0.4615603875036676,
      "learning_rate": 1.6977136226419187e-05,
      "loss": 0.4896,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2545563578605652,
      "step": 930,
      "valid_targets_mean": 3328.4,
      "valid_targets_min": 1912
    },
    {
      "epoch": 2.9888,
      "grad_norm": 0.4329963191374008,
      "learning_rate": 1.6756767717075354e-05,
      "loss": 0.4827,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2186550498008728,
      "step": 935,
      "valid_targets_mean": 3280.2,
      "valid_targets_min": 1515
    },
    {
      "epoch": 3.0032,
      "grad_norm": 0.4275403372883427,
      "learning_rate": 1.6536802860554723e-05,
      "loss": 0.4587,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2611745595932007,
      "step": 940,
      "valid_targets_mean": 3092.6,
      "valid_targets_min": 1401
    },
    {
      "epoch": 3.0192,
      "grad_norm": 0.450471108392735,
      "learning_rate": 1.631726903369238e-05,
      "loss": 0.4665,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22990155220031738,
      "step": 945,
      "valid_targets_mean": 2913.1,
      "valid_targets_min": 881
    },
    {
      "epoch": 3.0352,
      "grad_norm": 0.5160073223042104,
      "learning_rate": 1.609819355967744e-05,
      "loss": 0.4673,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2828558087348938,
      "step": 950,
      "valid_targets_mean": 3116.0,
      "valid_targets_min": 1473
    },
    {
      "epoch": 3.0512,
      "grad_norm": 0.49174113573385936,
      "learning_rate": 1.587960370465239e-05,
      "loss": 0.469,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2625318467617035,
      "step": 955,
      "valid_targets_mean": 2854.3,
      "valid_targets_min": 1621
    },
    {
      "epoch": 3.0672,
      "grad_norm": 0.45141979426038364,
      "learning_rate": 1.5661526674319582e-05,
      "loss": 0.46,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2091599404811859,
      "step": 960,
      "valid_targets_mean": 3043.4,
      "valid_targets_min": 1375
    },
    {
      "epoch": 3.0832,
      "grad_norm": 0.43282987334993606,
      "learning_rate": 1.544398961055516e-05,
      "loss": 0.4575,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22479215264320374,
      "step": 965,
      "valid_targets_mean": 3451.6,
      "valid_targets_min": 1395
    },
    {
      "epoch": 3.0992,
      "grad_norm": 0.4509947466200453,
      "learning_rate": 1.5227019588031035e-05,
      "loss": 0.4436,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.19393005967140198,
      "step": 970,
      "valid_targets_mean": 3079.7,
      "valid_targets_min": 1488
    },
    {
      "epoch": 3.1152,
      "grad_norm": 0.5160194179383659,
      "learning_rate": 1.501064361084511e-05,
      "loss": 0.4778,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2354956418275833,
      "step": 975,
      "valid_targets_mean": 2606.4,
      "valid_targets_min": 640
    },
    {
      "epoch": 3.1312,
      "grad_norm": 0.49691820011535437,
      "learning_rate": 1.47948886091604e-05,
      "loss": 0.4802,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24869810044765472,
      "step": 980,
      "valid_targets_mean": 2802.9,
      "valid_targets_min": 1420
    },
    {
      "epoch": 3.1471999999999998,
      "grad_norm": 0.5229876254819794,
      "learning_rate": 1.4579781435853289e-05,
      "loss": 0.4973,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2780328392982483,
      "step": 985,
      "valid_targets_mean": 3310.3,
      "valid_targets_min": 1977
    },
    {
      "epoch": 3.1632,
      "grad_norm": 0.4757032753580724,
      "learning_rate": 1.4365348863171406e-05,
      "loss": 0.4749,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23341715335845947,
      "step": 990,
      "valid_targets_mean": 2850.2,
      "valid_targets_min": 1371
    },
    {
      "epoch": 3.1792,
      "grad_norm": 0.49397855207266117,
      "learning_rate": 1.4151617579401551e-05,
      "loss": 0.4664,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2610412538051605,
      "step": 995,
      "valid_targets_mean": 3235.2,
      "valid_targets_min": 1271
    },
    {
      "epoch": 3.1952,
      "grad_norm": 0.5634169766980279,
      "learning_rate": 1.3938614185548094e-05,
      "loss": 0.4764,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24418114125728607,
      "step": 1000,
      "valid_targets_mean": 2622.4,
      "valid_targets_min": 1061
    },
    {
      "epoch": 3.2112,
      "grad_norm": 0.43284546857851314,
      "learning_rate": 1.3726365192022173e-05,
      "loss": 0.4714,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20849572122097015,
      "step": 1005,
      "valid_targets_mean": 2964.6,
      "valid_targets_min": 1005
    },
    {
      "epoch": 3.2272,
      "grad_norm": 0.4425874996893573,
      "learning_rate": 1.3514897015342257e-05,
      "loss": 0.4613,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21642732620239258,
      "step": 1010,
      "valid_targets_mean": 2798.9,
      "valid_targets_min": 1047
    },
    {
      "epoch": 3.2432,
      "grad_norm": 0.520474503449209,
      "learning_rate": 1.3304235974846295e-05,
      "loss": 0.4632,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.258159875869751,
      "step": 1015,
      "valid_targets_mean": 2868.6,
      "valid_targets_min": 1428
    },
    {
      "epoch": 3.2592,
      "grad_norm": 0.5144656747497975,
      "learning_rate": 1.3094408289416052e-05,
      "loss": 0.4683,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2635853886604309,
      "step": 1020,
      "valid_targets_mean": 3025.1,
      "valid_targets_min": 1064
    },
    {
      "epoch": 3.2752,
      "grad_norm": 0.45505799007822667,
      "learning_rate": 1.2885440074213877e-05,
      "loss": 0.4732,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22156450152397156,
      "step": 1025,
      "valid_targets_mean": 3054.5,
      "valid_targets_min": 918
    },
    {
      "epoch": 3.2912,
      "grad_norm": 0.49329116427210845,
      "learning_rate": 1.267735733743242e-05,
      "loss": 0.4789,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2545117437839508,
      "step": 1030,
      "valid_targets_mean": 2909.3,
      "valid_targets_min": 1099
    },
    {
      "epoch": 3.3072,
      "grad_norm": 0.46073569463876607,
      "learning_rate": 1.2470185977057643e-05,
      "loss": 0.4565,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23767399787902832,
      "step": 1035,
      "valid_targets_mean": 3542.4,
      "valid_targets_min": 1528
    },
    {
      "epoch": 3.3232,
      "grad_norm": 1.1380968918785033,
      "learning_rate": 1.2263951777645588e-05,
      "loss": 0.4662,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22722697257995605,
      "step": 1040,
      "valid_targets_mean": 2743.5,
      "valid_targets_min": 1074
    },
    {
      "epoch": 3.3392,
      "grad_norm": 0.5152322146587375,
      "learning_rate": 1.2058680407113176e-05,
      "loss": 0.4919,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26898497343063354,
      "step": 1045,
      "valid_targets_mean": 3478.0,
      "valid_targets_min": 2344
    },
    {
      "epoch": 3.3552,
      "grad_norm": 0.4689141874042101,
      "learning_rate": 1.1854397413543626e-05,
      "loss": 0.4404,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.19132088124752045,
      "step": 1050,
      "valid_targets_mean": 2550.5,
      "valid_targets_min": 641
    },
    {
      "epoch": 3.3712,
      "grad_norm": 0.4569028553905679,
      "learning_rate": 1.1651128222006713e-05,
      "loss": 0.4689,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20203140377998352,
      "step": 1055,
      "valid_targets_mean": 3417.3,
      "valid_targets_min": 1001
    },
    {
      "epoch": 3.3872,
      "grad_norm": 0.4386665491317407,
      "learning_rate": 1.1448898131394364e-05,
      "loss": 0.4599,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2558937072753906,
      "step": 1060,
      "valid_targets_mean": 3618.4,
      "valid_targets_min": 1132
    },
    {
      "epoch": 3.4032,
      "grad_norm": 0.4857080547943075,
      "learning_rate": 1.124773231127196e-05,
      "loss": 0.4735,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22912126779556274,
      "step": 1065,
      "valid_targets_mean": 2826.2,
      "valid_targets_min": 1249
    },
    {
      "epoch": 3.4192,
      "grad_norm": 0.7075154349707183,
      "learning_rate": 1.1047655798745752e-05,
      "loss": 0.4616,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23784030973911285,
      "step": 1070,
      "valid_targets_mean": 3468.4,
      "valid_targets_min": 1545
    },
    {
      "epoch": 3.4352,
      "grad_norm": 0.4741688779214942,
      "learning_rate": 1.084869349534671e-05,
      "loss": 0.464,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2626760005950928,
      "step": 1075,
      "valid_targets_mean": 3524.6,
      "valid_targets_min": 1213
    },
    {
      "epoch": 3.4512,
      "grad_norm": 0.4874858260927396,
      "learning_rate": 1.0650870163931275e-05,
      "loss": 0.4724,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20847836136817932,
      "step": 1080,
      "valid_targets_mean": 2429.3,
      "valid_targets_min": 1342
    },
    {
      "epoch": 3.4672,
      "grad_norm": 0.4876869164388468,
      "learning_rate": 1.0454210425599426e-05,
      "loss": 0.4674,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24478471279144287,
      "step": 1085,
      "valid_targets_mean": 2773.8,
      "valid_targets_min": 938
    },
    {
      "epoch": 3.4832,
      "grad_norm": 0.4752134862059384,
      "learning_rate": 1.0258738756630255e-05,
      "loss": 0.4595,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23344089090824127,
      "step": 1090,
      "valid_targets_mean": 2999.8,
      "valid_targets_min": 1357
    },
    {
      "epoch": 3.4992,
      "grad_norm": 0.44670646778927636,
      "learning_rate": 1.0064479485435737e-05,
      "loss": 0.4726,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.19428589940071106,
      "step": 1095,
      "valid_targets_mean": 2776.6,
      "valid_targets_min": 1594
    },
    {
      "epoch": 3.5152,
      "grad_norm": 0.4732489224718485,
      "learning_rate": 9.871456789532736e-06,
      "loss": 0.4733,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22874388098716736,
      "step": 1100,
      "valid_targets_mean": 2985.3,
      "valid_targets_min": 1850
    },
    {
      "epoch": 3.5312,
      "grad_norm": 0.4507617362751026,
      "learning_rate": 9.679694692533909e-06,
      "loss": 0.4557,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24319292604923248,
      "step": 1105,
      "valid_targets_mean": 3500.3,
      "valid_targets_min": 1069
    },
    {
      "epoch": 3.5472,
      "grad_norm": 0.47845324805606443,
      "learning_rate": 9.489217061157744e-06,
      "loss": 0.4747,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22654272615909576,
      "step": 1110,
      "valid_targets_mean": 3058.6,
      "valid_targets_min": 830
    },
    {
      "epoch": 3.5632,
      "grad_norm": 0.44930562581242833,
      "learning_rate": 9.30004760225806e-06,
      "loss": 0.4635,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21433737874031067,
      "step": 1115,
      "valid_targets_mean": 2892.0,
      "valid_targets_min": 1652
    },
    {
      "epoch": 3.5792,
      "grad_norm": 0.46051845797382324,
      "learning_rate": 9.112209859873479e-06,
      "loss": 0.4767,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23646137118339539,
      "step": 1120,
      "valid_targets_mean": 3047.9,
      "valid_targets_min": 1204
    },
    {
      "epoch": 3.5952,
      "grad_norm": 0.518040425573914,
      "learning_rate": 8.925727212297154e-06,
      "loss": 0.471,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24463368952274323,
      "step": 1125,
      "valid_targets_mean": 2889.5,
      "valid_targets_min": 1435
    },
    {
      "epoch": 3.6112,
      "grad_norm": 0.5282336939379751,
      "learning_rate": 8.74062286916705e-06,
      "loss": 0.4557,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23317721486091614,
      "step": 1130,
      "valid_targets_mean": 2804.1,
      "valid_targets_min": 957
    },
    {
      "epoch": 3.6272,
      "grad_norm": 0.42473984542421317,
      "learning_rate": 8.55691986857733e-06,
      "loss": 0.457,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23406881093978882,
      "step": 1135,
      "valid_targets_mean": 3556.3,
      "valid_targets_min": 1793
    },
    {
      "epoch": 3.6432,
      "grad_norm": 0.4763111625783207,
      "learning_rate": 8.374641074210979e-06,
      "loss": 0.475,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2574063539505005,
      "step": 1140,
      "valid_targets_mean": 2914.0,
      "valid_targets_min": 1130
    },
    {
      "epoch": 3.6592000000000002,
      "grad_norm": 0.4627805208631961,
      "learning_rate": 8.193809172494249e-06,
      "loss": 0.4642,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21742022037506104,
      "step": 1145,
      "valid_targets_mean": 3212.3,
      "valid_targets_min": 1433
    },
    {
      "epoch": 3.6752000000000002,
      "grad_norm": 0.4583153261416385,
      "learning_rate": 8.014446669773061e-06,
      "loss": 0.4541,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22402873635292053,
      "step": 1150,
      "valid_targets_mean": 3207.1,
      "valid_targets_min": 2009
    },
    {
      "epoch": 3.6912000000000003,
      "grad_norm": 0.5684605679181882,
      "learning_rate": 7.83657588951187e-06,
      "loss": 0.4922,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22889772057533264,
      "step": 1155,
      "valid_targets_mean": 2417.2,
      "valid_targets_min": 1240
    },
    {
      "epoch": 3.7072000000000003,
      "grad_norm": 0.48549138988607504,
      "learning_rate": 7.66021896951529e-06,
      "loss": 0.4582,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22643694281578064,
      "step": 1160,
      "valid_targets_mean": 2842.7,
      "valid_targets_min": 1629
    },
    {
      "epoch": 3.7232,
      "grad_norm": 0.48999301596541023,
      "learning_rate": 7.485397859172841e-06,
      "loss": 0.4819,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24150823056697845,
      "step": 1165,
      "valid_targets_mean": 2834.5,
      "valid_targets_min": 1642
    },
    {
      "epoch": 3.7392,
      "grad_norm": 0.5234114662032995,
      "learning_rate": 7.312134316727093e-06,
      "loss": 0.4681,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23145684599876404,
      "step": 1170,
      "valid_targets_mean": 2914.3,
      "valid_targets_min": 1406
    },
    {
      "epoch": 3.7552,
      "grad_norm": 0.48976591667376895,
      "learning_rate": 7.140449906565656e-06,
      "loss": 0.4681,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2297079712152481,
      "step": 1175,
      "valid_targets_mean": 3307.9,
      "valid_targets_min": 1882
    },
    {
      "epoch": 3.7712,
      "grad_norm": 0.4975298653538736,
      "learning_rate": 6.970365996537285e-06,
      "loss": 0.4611,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22569899260997772,
      "step": 1180,
      "valid_targets_mean": 2788.9,
      "valid_targets_min": 752
    },
    {
      "epoch": 3.7872,
      "grad_norm": 0.48902216227735895,
      "learning_rate": 6.801903755292403e-06,
      "loss": 0.4712,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.241786390542984,
      "step": 1185,
      "valid_targets_mean": 3153.2,
      "valid_targets_min": 2123
    },
    {
      "epoch": 3.8032,
      "grad_norm": 0.5098309755198677,
      "learning_rate": 6.635084149648481e-06,
      "loss": 0.4777,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2331969290971756,
      "step": 1190,
      "valid_targets_mean": 3060.3,
      "valid_targets_min": 1051
    },
    {
      "epoch": 3.8192,
      "grad_norm": 0.5079088842619889,
      "learning_rate": 6.469927941980483e-06,
      "loss": 0.4689,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1932951956987381,
      "step": 1195,
      "valid_targets_mean": 2451.8,
      "valid_targets_min": 1247
    },
    {
      "epoch": 3.8352,
      "grad_norm": 0.51113694581853,
      "learning_rate": 6.30645568763681e-06,
      "loss": 0.463,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24860982596874237,
      "step": 1200,
      "valid_targets_mean": 3285.6,
      "valid_targets_min": 1472
    },
    {
      "epoch": 3.8512,
      "grad_norm": 0.5202062021434569,
      "learning_rate": 6.144687732380963e-06,
      "loss": 0.4662,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23511703312397003,
      "step": 1205,
      "valid_targets_mean": 2735.3,
      "valid_targets_min": 1513
    },
    {
      "epoch": 3.8672,
      "grad_norm": 0.46415649801108116,
      "learning_rate": 5.9846442098592895e-06,
      "loss": 0.4604,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24313569068908691,
      "step": 1210,
      "valid_targets_mean": 2886.9,
      "valid_targets_min": 726
    },
    {
      "epoch": 3.8832,
      "grad_norm": 0.48843310938787426,
      "learning_rate": 5.826345039095178e-06,
      "loss": 0.4821,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2138245403766632,
      "step": 1215,
      "valid_targets_mean": 2952.6,
      "valid_targets_min": 719
    },
    {
      "epoch": 3.8992,
      "grad_norm": 0.45187325642880627,
      "learning_rate": 5.669809922009937e-06,
      "loss": 0.4893,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21185560524463654,
      "step": 1220,
      "valid_targets_mean": 3259.9,
      "valid_targets_min": 1767
    },
    {
      "epoch": 3.9152,
      "grad_norm": 0.4512362910904971,
      "learning_rate": 5.515058340970665e-06,
      "loss": 0.464,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23606804013252258,
      "step": 1225,
      "valid_targets_mean": 3170.3,
      "valid_targets_min": 693
    },
    {
      "epoch": 3.9312,
      "grad_norm": 0.563692302304579,
      "learning_rate": 5.362109556365496e-06,
      "loss": 0.4762,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23333798348903656,
      "step": 1230,
      "valid_targets_mean": 3311.6,
      "valid_targets_min": 1618
    },
    {
      "epoch": 3.9472,
      "grad_norm": 0.4768555432720377,
      "learning_rate": 5.2109826042064445e-06,
      "loss": 0.4876,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22933447360992432,
      "step": 1235,
      "valid_targets_mean": 3085.6,
      "valid_targets_min": 1800
    },
    {
      "epoch": 3.9632,
      "grad_norm": 0.5103664581840371,
      "learning_rate": 5.0616962937601945e-06,
      "loss": 0.4806,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23675984144210815,
      "step": 1240,
      "valid_targets_mean": 3059.2,
      "valid_targets_min": 1210
    },
    {
      "epoch": 3.9792,
      "grad_norm": 0.46382472726277013,
      "learning_rate": 4.914269205207076e-06,
      "loss": 0.4661,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2383194863796234,
      "step": 1245,
      "valid_targets_mean": 3139.2,
      "valid_targets_min": 846
    },
    {
      "epoch": 3.9952,
      "grad_norm": 0.4268050053278703,
      "learning_rate": 4.76871968732858e-06,
      "loss": 0.4592,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23828884959220886,
      "step": 1250,
      "valid_targets_mean": 3040.5,
      "valid_targets_min": 1095
    },
    {
      "epoch": 4.0096,
      "grad_norm": 0.49732816108587136,
      "learning_rate": 4.625065855223689e-06,
      "loss": 0.4567,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2624007761478424,
      "step": 1255,
      "valid_targets_mean": 2886.8,
      "valid_targets_min": 750
    },
    {
      "epoch": 4.0256,
      "grad_norm": 0.4831609813667501,
      "learning_rate": 4.483325588054259e-06,
      "loss": 0.4591,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.27929121255874634,
      "step": 1260,
      "valid_targets_mean": 3247.8,
      "valid_targets_min": 881
    },
    {
      "epoch": 4.0416,
      "grad_norm": 0.4399957174180194,
      "learning_rate": 4.343516526819755e-06,
      "loss": 0.455,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21088378131389618,
      "step": 1265,
      "valid_targets_mean": 3649.4,
      "valid_targets_min": 1935
    },
    {
      "epoch": 4.0576,
      "grad_norm": 0.4922430221954871,
      "learning_rate": 4.205656072161681e-06,
      "loss": 0.4581,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25056588649749756,
      "step": 1270,
      "valid_targets_mean": 3408.4,
      "valid_targets_min": 1894
    },
    {
      "epoch": 4.0736,
      "grad_norm": 0.46350490712051406,
      "learning_rate": 4.069761382197901e-06,
      "loss": 0.4558,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2043137550354004,
      "step": 1275,
      "valid_targets_mean": 2936.8,
      "valid_targets_min": 1347
    },
    {
      "epoch": 4.0896,
      "grad_norm": 0.4994153609615677,
      "learning_rate": 3.935849370387104e-06,
      "loss": 0.446,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22454729676246643,
      "step": 1280,
      "valid_targets_mean": 2772.5,
      "valid_targets_min": 1484
    },
    {
      "epoch": 4.1056,
      "grad_norm": 0.5499827117625989,
      "learning_rate": 3.803936703423783e-06,
      "loss": 0.4846,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24745602905750275,
      "step": 1285,
      "valid_targets_mean": 3050.7,
      "valid_targets_min": 1528
    },
    {
      "epoch": 4.1216,
      "grad_norm": 0.4406929832672752,
      "learning_rate": 3.6740397991638864e-06,
      "loss": 0.4527,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22310206294059753,
      "step": 1290,
      "valid_targets_mean": 3021.6,
      "valid_targets_min": 1125
    },
    {
      "epoch": 4.1376,
      "grad_norm": 0.44705321465815584,
      "learning_rate": 3.5461748245814633e-06,
      "loss": 0.4589,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22481977939605713,
      "step": 1295,
      "valid_targets_mean": 3618.6,
      "valid_targets_min": 1692
    },
    {
      "epoch": 4.1536,
      "grad_norm": 0.49516970608135413,
      "learning_rate": 3.420357693756502e-06,
      "loss": 0.4585,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26202359795570374,
      "step": 1300,
      "valid_targets_mean": 3241.0,
      "valid_targets_min": 1015
    },
    {
      "epoch": 4.1696,
      "grad_norm": 0.5360703828757977,
      "learning_rate": 3.2966040658942666e-06,
      "loss": 0.4664,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.266828715801239,
      "step": 1305,
      "valid_targets_mean": 2974.2,
      "valid_targets_min": 828
    },
    {
      "epoch": 4.1856,
      "grad_norm": 0.4950544833096414,
      "learning_rate": 3.174929343376374e-06,
      "loss": 0.4451,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22401739656925201,
      "step": 1310,
      "valid_targets_mean": 2744.4,
      "valid_targets_min": 846
    },
    {
      "epoch": 4.2016,
      "grad_norm": 0.4311500601540355,
      "learning_rate": 3.055348669843794e-06,
      "loss": 0.4495,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2312304973602295,
      "step": 1315,
      "valid_targets_mean": 3403.9,
      "valid_targets_min": 1499
    },
    {
      "epoch": 4.2176,
      "grad_norm": 0.5154587342573326,
      "learning_rate": 2.937876928312062e-06,
      "loss": 0.4868,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3122970461845398,
      "step": 1320,
      "valid_targets_mean": 3197.4,
      "valid_targets_min": 1823
    },
    {
      "epoch": 4.2336,
      "grad_norm": 0.4559328431144319,
      "learning_rate": 2.8225287393189547e-06,
      "loss": 0.4747,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20903021097183228,
      "step": 1325,
      "valid_targets_mean": 2931.4,
      "valid_targets_min": 1677
    },
    {
      "epoch": 4.2496,
      "grad_norm": 0.4806364348701646,
      "learning_rate": 2.709318459104815e-06,
      "loss": 0.4605,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23968324065208435,
      "step": 1330,
      "valid_targets_mean": 2946.6,
      "valid_targets_min": 1566
    },
    {
      "epoch": 4.2656,
      "grad_norm": 0.45289745050543984,
      "learning_rate": 2.5982601778257733e-06,
      "loss": 0.4515,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22231417894363403,
      "step": 1335,
      "valid_targets_mean": 3004.3,
      "valid_targets_min": 1691
    },
    {
      "epoch": 4.2816,
      "grad_norm": 0.47798104194681423,
      "learning_rate": 2.4893677178000797e-06,
      "loss": 0.4687,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22800111770629883,
      "step": 1340,
      "valid_targets_mean": 2962.9,
      "valid_targets_min": 1148
    },
    {
      "epoch": 4.2976,
      "grad_norm": 0.4487440581613835,
      "learning_rate": 2.3826546317877795e-06,
      "loss": 0.467,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2532956600189209,
      "step": 1345,
      "valid_targets_mean": 3242.4,
      "valid_targets_min": 1630
    },
    {
      "epoch": 4.3136,
      "grad_norm": 0.49067799892929675,
      "learning_rate": 2.278134201303952e-06,
      "loss": 0.4539,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.19229218363761902,
      "step": 1350,
      "valid_targets_mean": 2673.9,
      "valid_targets_min": 767
    },
    {
      "epoch": 4.3296,
      "grad_norm": 0.4932440061265886,
      "learning_rate": 2.1758194349656624e-06,
      "loss": 0.4551,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24860474467277527,
      "step": 1355,
      "valid_targets_mean": 3220.6,
      "valid_targets_min": 1626
    },
    {
      "epoch": 4.3456,
      "grad_norm": 0.4731361095577356,
      "learning_rate": 2.075723066872939e-06,
      "loss": 0.4662,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21686381101608276,
      "step": 1360,
      "valid_targets_mean": 3198.9,
      "valid_targets_min": 1827
    },
    {
      "epoch": 4.3616,
      "grad_norm": 0.4276096023507029,
      "learning_rate": 1.977857555023854e-06,
      "loss": 0.4351,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.25182968378067017,
      "step": 1365,
      "valid_targets_mean": 3999.2,
      "valid_targets_min": 1878
    },
    {
      "epoch": 4.3776,
      "grad_norm": 0.48625622975368676,
      "learning_rate": 1.8822350797640543e-06,
      "loss": 0.443,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24654650688171387,
      "step": 1370,
      "valid_targets_mean": 2706.6,
      "valid_targets_min": 1803
    },
    {
      "epoch": 4.3936,
      "grad_norm": 0.45154196191194956,
      "learning_rate": 1.788867542270729e-06,
      "loss": 0.4621,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23573841154575348,
      "step": 1375,
      "valid_targets_mean": 2747.8,
      "valid_targets_min": 1506
    },
    {
      "epoch": 4.4096,
      "grad_norm": 0.47622906189431036,
      "learning_rate": 1.6977665630714345e-06,
      "loss": 0.4714,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2745051980018616,
      "step": 1380,
      "valid_targets_mean": 3200.2,
      "valid_targets_min": 1554
    },
    {
      "epoch": 4.4256,
      "grad_norm": 0.4747768382814133,
      "learning_rate": 1.6089434805977799e-06,
      "loss": 0.4421,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20941603183746338,
      "step": 1385,
      "valid_targets_mean": 2683.9,
      "valid_targets_min": 1099
    },
    {
      "epoch": 4.4416,
      "grad_norm": 0.4898721125591949,
      "learning_rate": 1.5224093497742654e-06,
      "loss": 0.4746,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2430362105369568,
      "step": 1390,
      "valid_targets_mean": 3143.4,
      "valid_targets_min": 1850
    },
    {
      "epoch": 4.4576,
      "grad_norm": 0.45924092288013746,
      "learning_rate": 1.4381749406423695e-06,
      "loss": 0.4566,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22313876450061798,
      "step": 1395,
      "valid_targets_mean": 2977.5,
      "valid_targets_min": 1612
    },
    {
      "epoch": 4.4736,
      "grad_norm": 0.45355426629210616,
      "learning_rate": 1.3562507370201062e-06,
      "loss": 0.4624,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21281218528747559,
      "step": 1400,
      "valid_targets_mean": 2939.4,
      "valid_targets_min": 1601
    },
    {
      "epoch": 4.4896,
      "grad_norm": 0.5066555022569265,
      "learning_rate": 1.2766469351972345e-06,
      "loss": 0.4741,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21767839789390564,
      "step": 1405,
      "valid_targets_mean": 2710.8,
      "valid_targets_min": 1358
    },
    {
      "epoch": 4.5056,
      "grad_norm": 0.43358289004641076,
      "learning_rate": 1.1993734426661985e-06,
      "loss": 0.4472,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22621162235736847,
      "step": 1410,
      "valid_targets_mean": 3720.8,
      "valid_targets_min": 1983
    },
    {
      "epoch": 4.5216,
      "grad_norm": 0.5044761882250215,
      "learning_rate": 1.1244398768890496e-06,
      "loss": 0.4828,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23491895198822021,
      "step": 1415,
      "valid_targets_mean": 2813.6,
      "valid_targets_min": 1320
    },
    {
      "epoch": 4.5376,
      "grad_norm": 0.5330550467460243,
      "learning_rate": 1.0518555641004613e-06,
      "loss": 0.46,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2191581130027771,
      "step": 1420,
      "valid_targets_mean": 2327.1,
      "valid_targets_min": 1040
    },
    {
      "epoch": 4.5536,
      "grad_norm": 0.7320905137758208,
      "learning_rate": 9.816295381469954e-07,
      "loss": 0.4597,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2201734483242035,
      "step": 1425,
      "valid_targets_mean": 2678.6,
      "valid_targets_min": 957
    },
    {
      "epoch": 4.5696,
      "grad_norm": 0.4551873603504526,
      "learning_rate": 9.137705393627239e-07,
      "loss": 0.445,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.19194208085536957,
      "step": 1430,
      "valid_targets_mean": 3091.6,
      "valid_targets_min": 713
    },
    {
      "epoch": 4.5856,
      "grad_norm": 0.4698529743121287,
      "learning_rate": 8.482870134814214e-07,
      "loss": 0.4441,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2153075933456421,
      "step": 1435,
      "valid_targets_mean": 2814.6,
      "valid_targets_min": 1242
    },
    {
      "epoch": 4.6016,
      "grad_norm": 0.46872541762060055,
      "learning_rate": 7.851871105854125e-07,
      "loss": 0.4715,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24948900938034058,
      "step": 1440,
      "valid_targets_mean": 2967.8,
      "valid_targets_min": 1511
    },
    {
      "epoch": 4.6176,
      "grad_norm": 0.4489060249996938,
      "learning_rate": 7.244786840912033e-07,
      "loss": 0.4428,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.1973823457956314,
      "step": 1445,
      "valid_targets_mean": 2968.2,
      "valid_targets_min": 1703
    },
    {
      "epoch": 4.6336,
      "grad_norm": 0.45728313875611964,
      "learning_rate": 6.661692897720517e-07,
      "loss": 0.4658,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24835816025733948,
      "step": 1450,
      "valid_targets_mean": 3327.2,
      "valid_targets_min": 1195
    },
    {
      "epoch": 4.6495999999999995,
      "grad_norm": 0.4747389713592,
      "learning_rate": 6.10266184817565e-07,
      "loss": 0.4831,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24206383526325226,
      "step": 1455,
      "valid_targets_mean": 3087.4,
      "valid_targets_min": 1463
    },
    {
      "epoch": 4.6655999999999995,
      "grad_norm": 0.4508085344360603,
      "learning_rate": 5.567763269304927e-07,
      "loss": 0.4598,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2053154706954956,
      "step": 1460,
      "valid_targets_mean": 2956.6,
      "valid_targets_min": 1527
    },
    {
      "epoch": 4.6815999999999995,
      "grad_norm": 0.530546828942126,
      "learning_rate": 5.057063734607392e-07,
      "loss": 0.4725,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2210843563079834,
      "step": 1465,
      "valid_targets_mean": 2872.1,
      "valid_targets_min": 1062
    },
    {
      "epoch": 4.6975999999999996,
      "grad_norm": 0.44783159799278005,
      "learning_rate": 4.570626805768119e-07,
      "loss": 0.4597,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20867741107940674,
      "step": 1470,
      "valid_targets_mean": 2984.9,
      "valid_targets_min": 1101
    },
    {
      "epoch": 4.7136,
      "grad_norm": 0.4833208583687667,
      "learning_rate": 4.1085130247472625e-07,
      "loss": 0.4601,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21128685772418976,
      "step": 1475,
      "valid_targets_mean": 2744.1,
      "valid_targets_min": 1689
    },
    {
      "epoch": 4.7296,
      "grad_norm": 0.4382426941125842,
      "learning_rate": 3.670779906244981e-07,
      "loss": 0.4423,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22669294476509094,
      "step": 1480,
      "valid_targets_mean": 3158.8,
      "valid_targets_min": 1065
    },
    {
      "epoch": 4.7456,
      "grad_norm": 0.46954837995443893,
      "learning_rate": 3.2574819305432713e-07,
      "loss": 0.4629,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23599302768707275,
      "step": 1485,
      "valid_targets_mean": 3009.8,
      "valid_targets_min": 1451
    },
    {
      "epoch": 4.7616,
      "grad_norm": 0.4705667047472302,
      "learning_rate": 2.8686705367250824e-07,
      "loss": 0.4486,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.19904813170433044,
      "step": 1490,
      "valid_targets_mean": 2777.3,
      "valid_targets_min": 1237
    },
    {
      "epoch": 4.7776,
      "grad_norm": 0.4405020382668668,
      "learning_rate": 2.504394116272502e-07,
      "loss": 0.4318,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21686434745788574,
      "step": 1495,
      "valid_targets_mean": 3203.4,
      "valid_targets_min": 1366
    },
    {
      "epoch": 4.7936,
      "grad_norm": 0.47490195586595313,
      "learning_rate": 2.1646980070437973e-07,
      "loss": 0.4653,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.232308030128479,
      "step": 1500,
      "valid_targets_mean": 2665.0,
      "valid_targets_min": 1263
    },
    {
      "epoch": 4.8096,
      "grad_norm": 0.45026205218493226,
      "learning_rate": 1.8496244876306858e-07,
      "loss": 0.4433,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.21020200848579407,
      "step": 1505,
      "valid_targets_mean": 3043.4,
      "valid_targets_min": 1670
    },
    {
      "epoch": 4.8256,
      "grad_norm": 0.49102319598317373,
      "learning_rate": 1.559212772096319e-07,
      "loss": 0.4596,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.26330631971359253,
      "step": 1510,
      "valid_targets_mean": 3498.3,
      "valid_targets_min": 1323
    },
    {
      "epoch": 4.8416,
      "grad_norm": 0.4885125920804189,
      "learning_rate": 1.2934990050947228e-07,
      "loss": 0.4785,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2583045959472656,
      "step": 1515,
      "valid_targets_mean": 3124.8,
      "valid_targets_min": 2100
    },
    {
      "epoch": 4.8576,
      "grad_norm": 0.4276707026867548,
      "learning_rate": 1.0525162573723269e-07,
      "loss": 0.4768,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22992771863937378,
      "step": 1520,
      "valid_targets_mean": 3659.1,
      "valid_targets_min": 1498
    },
    {
      "epoch": 4.8736,
      "grad_norm": 0.4795330410563074,
      "learning_rate": 8.362945216517704e-08,
      "loss": 0.4758,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2545624375343323,
      "step": 1525,
      "valid_targets_mean": 3432.0,
      "valid_targets_min": 1324
    },
    {
      "epoch": 4.8896,
      "grad_norm": 0.5315564601564388,
      "learning_rate": 6.448607088991532e-08,
      "loss": 0.4684,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.2378055900335312,
      "step": 1530,
      "valid_targets_mean": 2588.7,
      "valid_targets_min": 1200
    },
    {
      "epoch": 4.9056,
      "grad_norm": 0.5164470981252668,
      "learning_rate": 4.782386449746934e-08,
      "loss": 0.4545,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.22151297330856323,
      "step": 1535,
      "valid_targets_mean": 2466.6,
      "valid_targets_min": 1493
    },
    {
      "epoch": 4.9216,
      "grad_norm": 0.4620064992407382,
      "learning_rate": 3.3644906766734374e-08,
      "loss": 0.4573,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24147440493106842,
      "step": 1540,
      "valid_targets_mean": 3112.8,
      "valid_targets_min": 1163
    },
    {
      "epoch": 4.9376,
      "grad_norm": 0.4824052579676531,
      "learning_rate": 2.1950962411367848e-08,
      "loss": 0.4525,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24404051899909973,
      "step": 1545,
      "valid_targets_mean": 3064.6,
      "valid_targets_min": 2201
    },
    {
      "epoch": 4.9536,
      "grad_norm": 0.4604378529000815,
      "learning_rate": 1.2743486860165022e-08,
      "loss": 0.4586,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.24580053985118866,
      "step": 1550,
      "valid_targets_mean": 3306.9,
      "valid_targets_min": 1074
    },
    {
      "epoch": 4.9696,
      "grad_norm": 0.4753199018242153,
      "learning_rate": 6.023626075915001e-09,
      "loss": 0.448,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.20023775100708008,
      "step": 1555,
      "valid_targets_mean": 2623.0,
      "valid_targets_min": 1790
    },
    {
      "epoch": 4.9856,
      "grad_norm": 0.4994766980539709,
      "learning_rate": 1.7922164127659457e-09,
      "loss": 0.4676,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.23927080631256104,
      "step": 1560,
      "valid_targets_mean": 3136.8,
      "valid_targets_min": 1188
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.6369392275743723,
      "learning_rate": 4.978451213499824e-11,
      "loss": 0.4567,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.399383544921875,
      "step": 1565,
      "valid_targets_mean": 3486.9,
      "valid_targets_min": 1658
    },
    {
      "epoch": 5.0,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.399383544921875,
      "step": 1565,
      "total_flos": 6.872336138627973e+17,
      "train_loss": 0.5079075557355301,
      "train_runtime": 11661.1311,
      "train_samples_per_second": 4.283,
      "train_steps_per_second": 0.134,
      "valid_targets_mean": 3486.9,
      "valid_targets_min": 1658
    }
  ],
  "logging_steps": 5,
  "max_steps": 1565,
  "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": 6.872336138627973e+17,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}