Files
Qwen3-8B_exp_tas_temp_0.25_…/trainer_state.json
ModelHub XC 79b3720dba 初始化项目,由ModelHub XC社区提供模型
Model: laion/Qwen3-8B_exp_tas_temp_0.25_traces_save-strategy_steps
Source: Original Platform
2026-09-28 07:53:18 +08:00

5361 lines
149 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 8.0,
"eval_steps": 500,
"global_step": 2416,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.016556291390728478,
"grad_norm": 1.90683513530528,
"learning_rate": 3.0769230769230774e-05,
"loss": 0.6488,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5742146968841553,
"step": 5,
"valid_targets_mean": 3701.2,
"valid_targets_min": 762
},
{
"epoch": 0.033112582781456956,
"grad_norm": 0.7651158048605883,
"learning_rate": 6.923076923076924e-05,
"loss": 0.5233,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4385795593261719,
"step": 10,
"valid_targets_mean": 3411.1,
"valid_targets_min": 1350
},
{
"epoch": 0.04966887417218543,
"grad_norm": 0.6226221872511525,
"learning_rate": 9.999995727007303e-05,
"loss": 0.4642,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40634483098983765,
"step": 15,
"valid_targets_mean": 3474.5,
"valid_targets_min": 290
},
{
"epoch": 0.06622516556291391,
"grad_norm": 0.5429012364756655,
"learning_rate": 9.999846173029751e-05,
"loss": 0.4167,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.382730096578598,
"step": 20,
"valid_targets_mean": 3201.2,
"valid_targets_min": 503
},
{
"epoch": 0.08278145695364239,
"grad_norm": 0.5015777202545035,
"learning_rate": 9.999482976720664e-05,
"loss": 0.3938,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3730844259262085,
"step": 25,
"valid_targets_mean": 3639.4,
"valid_targets_min": 980
},
{
"epoch": 0.09933774834437085,
"grad_norm": 0.521689084445223,
"learning_rate": 9.99890615359934e-05,
"loss": 0.3851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4082399606704712,
"step": 30,
"valid_targets_mean": 3125.9,
"valid_targets_min": 304
},
{
"epoch": 0.11589403973509933,
"grad_norm": 0.46242768252784705,
"learning_rate": 9.998115728313304e-05,
"loss": 0.3636,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36418813467025757,
"step": 35,
"valid_targets_mean": 3603.2,
"valid_targets_min": 389
},
{
"epoch": 0.13245033112582782,
"grad_norm": 0.46062437918525106,
"learning_rate": 9.997111734637259e-05,
"loss": 0.3636,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3387767970561981,
"step": 40,
"valid_targets_mean": 3010.2,
"valid_targets_min": 914
},
{
"epoch": 0.1490066225165563,
"grad_norm": 0.5474805806560548,
"learning_rate": 9.99589421547163e-05,
"loss": 0.3648,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3573765754699707,
"step": 45,
"valid_targets_mean": 2723.1,
"valid_targets_min": 348
},
{
"epoch": 0.16556291390728478,
"grad_norm": 0.46112806799631295,
"learning_rate": 9.994463222840746e-05,
"loss": 0.3642,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3329419493675232,
"step": 50,
"valid_targets_mean": 3779.4,
"valid_targets_min": 1089
},
{
"epoch": 0.18211920529801323,
"grad_norm": 0.5889981153153853,
"learning_rate": 9.992818817890609e-05,
"loss": 0.3489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34022992849349976,
"step": 55,
"valid_targets_mean": 3695.0,
"valid_targets_min": 683
},
{
"epoch": 0.1986754966887417,
"grad_norm": 0.485890501225745,
"learning_rate": 9.990961070886282e-05,
"loss": 0.3323,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3380316197872162,
"step": 60,
"valid_targets_mean": 3077.4,
"valid_targets_min": 484
},
{
"epoch": 0.2152317880794702,
"grad_norm": 0.46871870615048855,
"learning_rate": 9.988890061208888e-05,
"loss": 0.3246,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34749019145965576,
"step": 65,
"valid_targets_mean": 2666.1,
"valid_targets_min": 942
},
{
"epoch": 0.23178807947019867,
"grad_norm": 0.49981407253174,
"learning_rate": 9.986605877352216e-05,
"loss": 0.3281,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32183465361595154,
"step": 70,
"valid_targets_mean": 2898.4,
"valid_targets_min": 948
},
{
"epoch": 0.24834437086092714,
"grad_norm": 0.465970055557152,
"learning_rate": 9.984108616918944e-05,
"loss": 0.3472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3638273775577545,
"step": 75,
"valid_targets_mean": 3322.7,
"valid_targets_min": 829
},
{
"epoch": 0.26490066225165565,
"grad_norm": 0.42923109727006264,
"learning_rate": 9.981398386616458e-05,
"loss": 0.3028,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3329468369483948,
"step": 80,
"valid_targets_mean": 4030.4,
"valid_targets_min": 1243
},
{
"epoch": 0.2814569536423841,
"grad_norm": 0.4908534436328601,
"learning_rate": 9.97847530225231e-05,
"loss": 0.3206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34494367241859436,
"step": 85,
"valid_targets_mean": 2818.8,
"valid_targets_min": 397
},
{
"epoch": 0.2980132450331126,
"grad_norm": 0.4682338476431765,
"learning_rate": 9.975339488729252e-05,
"loss": 0.3303,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36933451890945435,
"step": 90,
"valid_targets_mean": 3035.2,
"valid_targets_min": 769
},
{
"epoch": 0.31456953642384106,
"grad_norm": 0.5703627164016276,
"learning_rate": 9.97199108003991e-05,
"loss": 0.3398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34210559725761414,
"step": 95,
"valid_targets_mean": 2703.5,
"valid_targets_min": 659
},
{
"epoch": 0.33112582781456956,
"grad_norm": 0.5694870237263695,
"learning_rate": 9.968430219261054e-05,
"loss": 0.3416,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3571714758872986,
"step": 100,
"valid_targets_mean": 2942.2,
"valid_targets_min": 414
},
{
"epoch": 0.347682119205298,
"grad_norm": 0.5447155856810172,
"learning_rate": 9.964657058547482e-05,
"loss": 0.3229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33061397075653076,
"step": 105,
"valid_targets_mean": 3210.2,
"valid_targets_min": 811
},
{
"epoch": 0.36423841059602646,
"grad_norm": 0.49822723862480545,
"learning_rate": 9.960671759125528e-05,
"loss": 0.3151,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3498014807701111,
"step": 110,
"valid_targets_mean": 2995.3,
"valid_targets_min": 960
},
{
"epoch": 0.38079470198675497,
"grad_norm": 0.4240177497240811,
"learning_rate": 9.956474491286163e-05,
"loss": 0.3078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3021423816680908,
"step": 115,
"valid_targets_mean": 3478.9,
"valid_targets_min": 807
},
{
"epoch": 0.3973509933774834,
"grad_norm": 0.4308477993032867,
"learning_rate": 9.95206543437772e-05,
"loss": 0.3225,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3106687664985657,
"step": 120,
"valid_targets_mean": 3050.9,
"valid_targets_min": 343
},
{
"epoch": 0.4139072847682119,
"grad_norm": 0.4521634245429716,
"learning_rate": 9.947444776798235e-05,
"loss": 0.3254,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31832602620124817,
"step": 125,
"valid_targets_mean": 2670.3,
"valid_targets_min": 682
},
{
"epoch": 0.4304635761589404,
"grad_norm": 0.42557498087453505,
"learning_rate": 9.942612715987396e-05,
"loss": 0.3261,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30098408460617065,
"step": 130,
"valid_targets_mean": 2869.0,
"valid_targets_min": 491
},
{
"epoch": 0.4470198675496689,
"grad_norm": 0.40476946449698925,
"learning_rate": 9.9375694584181e-05,
"loss": 0.3092,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2903498113155365,
"step": 135,
"valid_targets_mean": 3383.0,
"valid_targets_min": 639
},
{
"epoch": 0.46357615894039733,
"grad_norm": 0.3849739662867801,
"learning_rate": 9.93231521958764e-05,
"loss": 0.3004,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2814164459705353,
"step": 140,
"valid_targets_mean": 3503.0,
"valid_targets_min": 748
},
{
"epoch": 0.48013245033112584,
"grad_norm": 0.43542799451802044,
"learning_rate": 9.92685022400849e-05,
"loss": 0.3188,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30971819162368774,
"step": 145,
"valid_targets_mean": 3022.8,
"valid_targets_min": 716
},
{
"epoch": 0.4966887417218543,
"grad_norm": 0.4303859316551023,
"learning_rate": 9.921174705198713e-05,
"loss": 0.3333,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3176947236061096,
"step": 150,
"valid_targets_mean": 2879.0,
"valid_targets_min": 927
},
{
"epoch": 0.5132450331125827,
"grad_norm": 0.4048521162688901,
"learning_rate": 9.915288905671986e-05,
"loss": 0.3132,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31546807289123535,
"step": 155,
"valid_targets_mean": 2984.2,
"valid_targets_min": 697
},
{
"epoch": 0.5298013245033113,
"grad_norm": 0.41318383113790297,
"learning_rate": 9.90919307692723e-05,
"loss": 0.3122,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.307787150144577,
"step": 160,
"valid_targets_mean": 3121.4,
"valid_targets_min": 691
},
{
"epoch": 0.5463576158940397,
"grad_norm": 0.42344683174956227,
"learning_rate": 9.902887479437874e-05,
"loss": 0.3043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31755536794662476,
"step": 165,
"valid_targets_mean": 3221.2,
"valid_targets_min": 383
},
{
"epoch": 0.5629139072847682,
"grad_norm": 0.44999121888304794,
"learning_rate": 9.896372382640717e-05,
"loss": 0.3054,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29850849509239197,
"step": 170,
"valid_targets_mean": 3293.4,
"valid_targets_min": 708
},
{
"epoch": 0.5794701986754967,
"grad_norm": 0.4174723020518441,
"learning_rate": 9.889648064924417e-05,
"loss": 0.2959,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29357242584228516,
"step": 175,
"valid_targets_mean": 2921.2,
"valid_targets_min": 337
},
{
"epoch": 0.5960264900662252,
"grad_norm": 0.3934992601560117,
"learning_rate": 9.882714813617597e-05,
"loss": 0.2917,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.314992755651474,
"step": 180,
"valid_targets_mean": 2771.6,
"valid_targets_min": 742
},
{
"epoch": 0.6125827814569537,
"grad_norm": 0.4319366520390652,
"learning_rate": 9.875572924976568e-05,
"loss": 0.3032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30686649680137634,
"step": 185,
"valid_targets_mean": 3044.5,
"valid_targets_min": 996
},
{
"epoch": 0.6291390728476821,
"grad_norm": 0.39663117081681576,
"learning_rate": 9.868222704172663e-05,
"loss": 0.2968,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3096313774585724,
"step": 190,
"valid_targets_mean": 3284.1,
"valid_targets_min": 704
},
{
"epoch": 0.6456953642384106,
"grad_norm": 0.40930203334490584,
"learning_rate": 9.86066446527921e-05,
"loss": 0.3108,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3373670279979706,
"step": 195,
"valid_targets_mean": 2972.9,
"valid_targets_min": 388
},
{
"epoch": 0.6622516556291391,
"grad_norm": 0.41458528813092926,
"learning_rate": 9.852898531258101e-05,
"loss": 0.2974,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29588669538497925,
"step": 200,
"valid_targets_mean": 2968.2,
"valid_targets_min": 417
},
{
"epoch": 0.6788079470198676,
"grad_norm": 0.4127726707868181,
"learning_rate": 9.844925233945993e-05,
"loss": 0.2982,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29582294821739197,
"step": 205,
"valid_targets_mean": 3085.2,
"valid_targets_min": 669
},
{
"epoch": 0.695364238410596,
"grad_norm": 0.38150171227295054,
"learning_rate": 9.836744914040135e-05,
"loss": 0.3035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2959309220314026,
"step": 210,
"valid_targets_mean": 3154.6,
"valid_targets_min": 1046
},
{
"epoch": 0.7119205298013245,
"grad_norm": 0.3907574242481699,
"learning_rate": 9.828357921083803e-05,
"loss": 0.2916,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2875044345855713,
"step": 215,
"valid_targets_mean": 2839.2,
"valid_targets_min": 441
},
{
"epoch": 0.7284768211920529,
"grad_norm": 0.47734635556487637,
"learning_rate": 9.819764613451369e-05,
"loss": 0.3061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3034651577472687,
"step": 220,
"valid_targets_mean": 2299.7,
"valid_targets_min": 456
},
{
"epoch": 0.7450331125827815,
"grad_norm": 0.5493306341853688,
"learning_rate": 9.810965358332986e-05,
"loss": 0.3261,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4191819429397583,
"step": 225,
"valid_targets_mean": 2425.1,
"valid_targets_min": 881
},
{
"epoch": 0.7615894039735099,
"grad_norm": 0.4231239429385647,
"learning_rate": 9.801960531718896e-05,
"loss": 0.3043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30527618527412415,
"step": 230,
"valid_targets_mean": 3004.8,
"valid_targets_min": 1166
},
{
"epoch": 0.7781456953642384,
"grad_norm": 0.40273349353223326,
"learning_rate": 9.792750518383369e-05,
"loss": 0.2881,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.291836678981781,
"step": 235,
"valid_targets_mean": 3289.2,
"valid_targets_min": 755
},
{
"epoch": 0.7947019867549668,
"grad_norm": 0.34979086947645943,
"learning_rate": 9.783335711868258e-05,
"loss": 0.2879,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26166123151779175,
"step": 240,
"valid_targets_mean": 3777.2,
"valid_targets_min": 443
},
{
"epoch": 0.8112582781456954,
"grad_norm": 0.3814946213086308,
"learning_rate": 9.773716514466178e-05,
"loss": 0.2908,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2830836772918701,
"step": 245,
"valid_targets_mean": 3064.7,
"valid_targets_min": 787
},
{
"epoch": 0.8278145695364238,
"grad_norm": 0.4291462186338508,
"learning_rate": 9.76389333720333e-05,
"loss": 0.2867,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3208215534687042,
"step": 250,
"valid_targets_mean": 2699.3,
"valid_targets_min": 960
},
{
"epoch": 0.8443708609271523,
"grad_norm": 0.38599331383215324,
"learning_rate": 9.753866599821926e-05,
"loss": 0.2821,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29871106147766113,
"step": 255,
"valid_targets_mean": 3240.2,
"valid_targets_min": 517
},
{
"epoch": 0.8609271523178808,
"grad_norm": 0.4041941643222489,
"learning_rate": 9.743636730762258e-05,
"loss": 0.2895,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2891217768192291,
"step": 260,
"valid_targets_mean": 3215.0,
"valid_targets_min": 460
},
{
"epoch": 0.8774834437086093,
"grad_norm": 0.40792448674007137,
"learning_rate": 9.733204167144388e-05,
"loss": 0.289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30060410499572754,
"step": 265,
"valid_targets_mean": 2794.2,
"valid_targets_min": 348
},
{
"epoch": 0.8940397350993378,
"grad_norm": 0.42348183649389975,
"learning_rate": 9.722569354749474e-05,
"loss": 0.2795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3023919463157654,
"step": 270,
"valid_targets_mean": 2564.1,
"valid_targets_min": 364
},
{
"epoch": 0.9105960264900662,
"grad_norm": 0.37266274235459473,
"learning_rate": 9.71173274800072e-05,
"loss": 0.2737,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2739112973213196,
"step": 275,
"valid_targets_mean": 2655.0,
"valid_targets_min": 682
},
{
"epoch": 0.9271523178807947,
"grad_norm": 0.41833539615129917,
"learning_rate": 9.700694809943956e-05,
"loss": 0.289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2915274500846863,
"step": 280,
"valid_targets_mean": 2742.8,
"valid_targets_min": 322
},
{
"epoch": 0.9437086092715232,
"grad_norm": 0.39705257637999314,
"learning_rate": 9.689456012227862e-05,
"loss": 0.2864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2892838716506958,
"step": 285,
"valid_targets_mean": 2693.2,
"valid_targets_min": 524
},
{
"epoch": 0.9602649006622517,
"grad_norm": 0.3402610597897325,
"learning_rate": 9.678016835083797e-05,
"loss": 0.2901,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27023911476135254,
"step": 290,
"valid_targets_mean": 3718.4,
"valid_targets_min": 400
},
{
"epoch": 0.9768211920529801,
"grad_norm": 0.3736572088860731,
"learning_rate": 9.666377767305297e-05,
"loss": 0.2749,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2962803244590759,
"step": 295,
"valid_targets_mean": 3088.6,
"valid_targets_min": 379
},
{
"epoch": 0.9933774834437086,
"grad_norm": 0.3660796505630214,
"learning_rate": 9.654539306227177e-05,
"loss": 0.2851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28717029094696045,
"step": 300,
"valid_targets_mean": 3238.6,
"valid_targets_min": 662
},
{
"epoch": 1.009933774834437,
"grad_norm": 0.3389126925709842,
"learning_rate": 9.642501957704287e-05,
"loss": 0.2312,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1948443055152893,
"step": 305,
"valid_targets_mean": 3947.9,
"valid_targets_min": 460
},
{
"epoch": 1.0264900662251655,
"grad_norm": 0.43389381563954094,
"learning_rate": 9.630266236089888e-05,
"loss": 0.2131,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2011556476354599,
"step": 310,
"valid_targets_mean": 2415.0,
"valid_targets_min": 417
},
{
"epoch": 1.0430463576158941,
"grad_norm": 0.4779658551075087,
"learning_rate": 9.617832664213687e-05,
"loss": 0.2083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2396453469991684,
"step": 315,
"valid_targets_mean": 2805.7,
"valid_targets_min": 539
},
{
"epoch": 1.0596026490066226,
"grad_norm": 0.3700054585822443,
"learning_rate": 9.605201773359485e-05,
"loss": 0.1941,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20710405707359314,
"step": 320,
"valid_targets_mean": 3428.3,
"valid_targets_min": 624
},
{
"epoch": 1.076158940397351,
"grad_norm": 0.3638894195356153,
"learning_rate": 9.59237410324248e-05,
"loss": 0.1921,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18649369478225708,
"step": 325,
"valid_targets_mean": 3425.1,
"valid_targets_min": 810
},
{
"epoch": 1.0927152317880795,
"grad_norm": 0.4112539243976021,
"learning_rate": 9.579350201986209e-05,
"loss": 0.216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23401358723640442,
"step": 330,
"valid_targets_mean": 2886.5,
"valid_targets_min": 382
},
{
"epoch": 1.109271523178808,
"grad_norm": 0.3912355263814285,
"learning_rate": 9.566130626099117e-05,
"loss": 0.201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19619351625442505,
"step": 335,
"valid_targets_mean": 2849.8,
"valid_targets_min": 923
},
{
"epoch": 1.1258278145695364,
"grad_norm": 0.40578552480984786,
"learning_rate": 9.552715940450785e-05,
"loss": 0.2009,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21668368577957153,
"step": 340,
"valid_targets_mean": 2849.9,
"valid_targets_min": 738
},
{
"epoch": 1.1423841059602649,
"grad_norm": 0.35947855551809216,
"learning_rate": 9.53910671824779e-05,
"loss": 0.1972,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18576085567474365,
"step": 345,
"valid_targets_mean": 3117.2,
"valid_targets_min": 669
},
{
"epoch": 1.1589403973509933,
"grad_norm": 0.39321523941764297,
"learning_rate": 9.525303541009218e-05,
"loss": 0.211,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21488025784492493,
"step": 350,
"valid_targets_mean": 3144.1,
"valid_targets_min": 668
},
{
"epoch": 1.1754966887417218,
"grad_norm": 0.3954111253599595,
"learning_rate": 9.511306998541809e-05,
"loss": 0.2049,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20759233832359314,
"step": 355,
"valid_targets_mean": 3138.0,
"valid_targets_min": 337
},
{
"epoch": 1.1920529801324504,
"grad_norm": 0.36777226391572226,
"learning_rate": 9.497117688914753e-05,
"loss": 0.2121,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20897117257118225,
"step": 360,
"valid_targets_mean": 3266.3,
"valid_targets_min": 387
},
{
"epoch": 1.2086092715231789,
"grad_norm": 0.37073365877461406,
"learning_rate": 9.482736218434143e-05,
"loss": 0.204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22039398550987244,
"step": 365,
"valid_targets_mean": 3285.4,
"valid_targets_min": 612
},
{
"epoch": 1.2251655629139073,
"grad_norm": 0.4196541012562941,
"learning_rate": 9.468163201617062e-05,
"loss": 0.2139,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22531884908676147,
"step": 370,
"valid_targets_mean": 2320.7,
"valid_targets_min": 297
},
{
"epoch": 1.2417218543046358,
"grad_norm": 0.43178877348718864,
"learning_rate": 9.453399261165326e-05,
"loss": 0.2103,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24533167481422424,
"step": 375,
"valid_targets_mean": 2644.3,
"valid_targets_min": 363
},
{
"epoch": 1.2582781456953642,
"grad_norm": 0.4135336905960593,
"learning_rate": 9.438445027938872e-05,
"loss": 0.2083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22220832109451294,
"step": 380,
"valid_targets_mean": 2890.8,
"valid_targets_min": 1041
},
{
"epoch": 1.2748344370860927,
"grad_norm": 0.36590976016057225,
"learning_rate": 9.423301140928815e-05,
"loss": 0.2135,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22077831625938416,
"step": 385,
"valid_targets_mean": 3366.4,
"valid_targets_min": 760
},
{
"epoch": 1.2913907284768211,
"grad_norm": 0.40764100499602923,
"learning_rate": 9.407968247230127e-05,
"loss": 0.2169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24488572776317596,
"step": 390,
"valid_targets_mean": 2984.4,
"valid_targets_min": 691
},
{
"epoch": 1.3079470198675498,
"grad_norm": 0.3821438803487412,
"learning_rate": 9.392447002013996e-05,
"loss": 0.2071,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2059929072856903,
"step": 395,
"valid_targets_mean": 2712.9,
"valid_targets_min": 515
},
{
"epoch": 1.3245033112582782,
"grad_norm": 0.3899862302231617,
"learning_rate": 9.376738068499828e-05,
"loss": 0.2127,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2190270870923996,
"step": 400,
"valid_targets_mean": 2747.4,
"valid_targets_min": 286
},
{
"epoch": 1.3410596026490067,
"grad_norm": 0.3325867378595477,
"learning_rate": 9.360842117926912e-05,
"loss": 0.209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20354317128658295,
"step": 405,
"valid_targets_mean": 3225.6,
"valid_targets_min": 845
},
{
"epoch": 1.3576158940397351,
"grad_norm": 0.3916211490204987,
"learning_rate": 9.344759829525733e-05,
"loss": 0.2186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2249426394701004,
"step": 410,
"valid_targets_mean": 2838.7,
"valid_targets_min": 705
},
{
"epoch": 1.3741721854304636,
"grad_norm": 0.42291313422988946,
"learning_rate": 9.32849189048895e-05,
"loss": 0.2181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2273331880569458,
"step": 415,
"valid_targets_mean": 2553.2,
"valid_targets_min": 675
},
{
"epoch": 1.390728476821192,
"grad_norm": 0.33151027581689263,
"learning_rate": 9.312038995942036e-05,
"loss": 0.2082,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19271031022071838,
"step": 420,
"valid_targets_mean": 3639.4,
"valid_targets_min": 1068
},
{
"epoch": 1.4072847682119205,
"grad_norm": 0.37349219015627133,
"learning_rate": 9.295401848913568e-05,
"loss": 0.2104,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20354311168193817,
"step": 425,
"valid_targets_mean": 2989.4,
"valid_targets_min": 351
},
{
"epoch": 1.423841059602649,
"grad_norm": 0.3786604220131624,
"learning_rate": 9.278581160305195e-05,
"loss": 0.214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23767903447151184,
"step": 430,
"valid_targets_mean": 2826.8,
"valid_targets_min": 659
},
{
"epoch": 1.4403973509933774,
"grad_norm": 0.34311764570841485,
"learning_rate": 9.261577648861255e-05,
"loss": 0.2104,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20151343941688538,
"step": 435,
"valid_targets_mean": 3389.7,
"valid_targets_min": 257
},
{
"epoch": 1.4569536423841059,
"grad_norm": 0.35759471917482405,
"learning_rate": 9.244392041138068e-05,
"loss": 0.2119,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20437508821487427,
"step": 440,
"valid_targets_mean": 3178.9,
"valid_targets_min": 341
},
{
"epoch": 1.4735099337748343,
"grad_norm": 0.37973399919655415,
"learning_rate": 9.227025071472884e-05,
"loss": 0.2174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21793994307518005,
"step": 445,
"valid_targets_mean": 2845.3,
"valid_targets_min": 511
},
{
"epoch": 1.490066225165563,
"grad_norm": 0.3199299678974692,
"learning_rate": 9.209477481952513e-05,
"loss": 0.2088,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1964888721704483,
"step": 450,
"valid_targets_mean": 3549.1,
"valid_targets_min": 1175
},
{
"epoch": 1.5066225165562914,
"grad_norm": 0.3360596112867529,
"learning_rate": 9.191750022381614e-05,
"loss": 0.2097,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2100779265165329,
"step": 455,
"valid_targets_mean": 3224.5,
"valid_targets_min": 732
},
{
"epoch": 1.5231788079470199,
"grad_norm": 0.41319883720928263,
"learning_rate": 9.173843450250644e-05,
"loss": 0.2287,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23697538673877716,
"step": 460,
"valid_targets_mean": 2307.2,
"valid_targets_min": 436
},
{
"epoch": 1.5397350993377483,
"grad_norm": 0.3674808668270343,
"learning_rate": 9.155758530703512e-05,
"loss": 0.2102,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2140740156173706,
"step": 465,
"valid_targets_mean": 2787.8,
"valid_targets_min": 389
},
{
"epoch": 1.5562913907284768,
"grad_norm": 0.34510698134019835,
"learning_rate": 9.137496036504867e-05,
"loss": 0.2044,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20614945888519287,
"step": 470,
"valid_targets_mean": 3011.7,
"valid_targets_min": 348
},
{
"epoch": 1.5728476821192054,
"grad_norm": 0.33486924083957836,
"learning_rate": 9.119056748007083e-05,
"loss": 0.2134,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21356387436389923,
"step": 475,
"valid_targets_mean": 3417.7,
"valid_targets_min": 343
},
{
"epoch": 1.589403973509934,
"grad_norm": 0.33794193739460154,
"learning_rate": 9.10044145311692e-05,
"loss": 0.2084,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22062477469444275,
"step": 480,
"valid_targets_mean": 2954.5,
"valid_targets_min": 1084
},
{
"epoch": 1.6059602649006623,
"grad_norm": 0.4350940200625618,
"learning_rate": 9.081650947261848e-05,
"loss": 0.2192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23036962747573853,
"step": 485,
"valid_targets_mean": 2063.4,
"valid_targets_min": 370
},
{
"epoch": 1.6225165562913908,
"grad_norm": 0.3613490352121763,
"learning_rate": 9.062686033356064e-05,
"loss": 0.2246,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23221859335899353,
"step": 490,
"valid_targets_mean": 3110.2,
"valid_targets_min": 422
},
{
"epoch": 1.6390728476821192,
"grad_norm": 0.3659110415801921,
"learning_rate": 9.043547521766192e-05,
"loss": 0.2031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22216351330280304,
"step": 495,
"valid_targets_mean": 3123.1,
"valid_targets_min": 697
},
{
"epoch": 1.6556291390728477,
"grad_norm": 0.29058141650832486,
"learning_rate": 9.024236230276629e-05,
"loss": 0.2095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18544334173202515,
"step": 500,
"valid_targets_mean": 3743.5,
"valid_targets_min": 1147
},
{
"epoch": 1.6721854304635762,
"grad_norm": 0.32947959713235797,
"learning_rate": 9.004752984054636e-05,
"loss": 0.2066,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19370371103286743,
"step": 505,
"valid_targets_mean": 3204.0,
"valid_targets_min": 616
},
{
"epoch": 1.6887417218543046,
"grad_norm": 0.32243645481261785,
"learning_rate": 8.98509861561505e-05,
"loss": 0.2089,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19874104857444763,
"step": 510,
"valid_targets_mean": 3478.5,
"valid_targets_min": 284
},
{
"epoch": 1.705298013245033,
"grad_norm": 0.3205590404946569,
"learning_rate": 8.965273964784735e-05,
"loss": 0.2161,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2074926495552063,
"step": 515,
"valid_targets_mean": 3786.1,
"valid_targets_min": 889
},
{
"epoch": 1.7218543046357615,
"grad_norm": 0.32380087461319645,
"learning_rate": 8.945279878666673e-05,
"loss": 0.2092,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2074010670185089,
"step": 520,
"valid_targets_mean": 3235.7,
"valid_targets_min": 400
},
{
"epoch": 1.73841059602649,
"grad_norm": 0.31773031899011506,
"learning_rate": 8.925117211603784e-05,
"loss": 0.2079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1998700499534607,
"step": 525,
"valid_targets_mean": 3762.6,
"valid_targets_min": 979
},
{
"epoch": 1.7549668874172184,
"grad_norm": 0.3458480669482276,
"learning_rate": 8.904786825142414e-05,
"loss": 0.2076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20790527760982513,
"step": 530,
"valid_targets_mean": 3102.6,
"valid_targets_min": 1066
},
{
"epoch": 1.771523178807947,
"grad_norm": 0.30366598541080403,
"learning_rate": 8.884289587995524e-05,
"loss": 0.2055,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19465413689613342,
"step": 535,
"valid_targets_mean": 3343.0,
"valid_targets_min": 432
},
{
"epoch": 1.7880794701986755,
"grad_norm": 0.33604048544483084,
"learning_rate": 8.863626376005562e-05,
"loss": 0.216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22073376178741455,
"step": 540,
"valid_targets_mean": 3205.7,
"valid_targets_min": 720
},
{
"epoch": 1.804635761589404,
"grad_norm": 0.37007779611361397,
"learning_rate": 8.842798072107054e-05,
"loss": 0.2204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2214677929878235,
"step": 545,
"valid_targets_mean": 2804.7,
"valid_targets_min": 676
},
{
"epoch": 1.8211920529801324,
"grad_norm": 0.38478783228066465,
"learning_rate": 8.82180556628886e-05,
"loss": 0.2105,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22936037182807922,
"step": 550,
"valid_targets_mean": 2943.2,
"valid_targets_min": 639
},
{
"epoch": 1.8377483443708609,
"grad_norm": 0.34543010326577117,
"learning_rate": 8.800649755556157e-05,
"loss": 0.2138,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2245236337184906,
"step": 555,
"valid_targets_mean": 3434.0,
"valid_targets_min": 595
},
{
"epoch": 1.8543046357615895,
"grad_norm": 0.2932756071886955,
"learning_rate": 8.779331543892097e-05,
"loss": 0.2,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1866540163755417,
"step": 560,
"valid_targets_mean": 3893.2,
"valid_targets_min": 848
},
{
"epoch": 1.870860927152318,
"grad_norm": 0.32659918090044693,
"learning_rate": 8.757851842219198e-05,
"loss": 0.2052,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1978355050086975,
"step": 565,
"valid_targets_mean": 3045.8,
"valid_targets_min": 494
},
{
"epoch": 1.8874172185430464,
"grad_norm": 0.35574882988452505,
"learning_rate": 8.736211568360405e-05,
"loss": 0.2157,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2048686295747757,
"step": 570,
"valid_targets_mean": 2578.5,
"valid_targets_min": 367
},
{
"epoch": 1.903973509933775,
"grad_norm": 0.30787265425987176,
"learning_rate": 8.714411646999878e-05,
"loss": 0.2049,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2000887393951416,
"step": 575,
"valid_targets_mean": 3594.4,
"valid_targets_min": 510
},
{
"epoch": 1.9205298013245033,
"grad_norm": 0.32779303458954434,
"learning_rate": 8.692453009643482e-05,
"loss": 0.207,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18430213630199432,
"step": 580,
"valid_targets_mean": 3034.0,
"valid_targets_min": 295
},
{
"epoch": 1.9370860927152318,
"grad_norm": 0.3557072523105563,
"learning_rate": 8.670336594578981e-05,
"loss": 0.2015,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20450301468372345,
"step": 585,
"valid_targets_mean": 3029.6,
"valid_targets_min": 451
},
{
"epoch": 1.9536423841059603,
"grad_norm": 0.3665345270635697,
"learning_rate": 8.648063346835942e-05,
"loss": 0.2035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2252351939678192,
"step": 590,
"valid_targets_mean": 2917.3,
"valid_targets_min": 690
},
{
"epoch": 1.9701986754966887,
"grad_norm": 0.3421886637999956,
"learning_rate": 8.625634218145365e-05,
"loss": 0.1975,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20249176025390625,
"step": 595,
"valid_targets_mean": 3078.8,
"valid_targets_min": 336
},
{
"epoch": 1.9867549668874172,
"grad_norm": 0.33199110988226277,
"learning_rate": 8.603050166899e-05,
"loss": 0.2226,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21750187873840332,
"step": 600,
"valid_targets_mean": 3147.6,
"valid_targets_min": 918
},
{
"epoch": 2.0033112582781456,
"grad_norm": 0.4035685357179341,
"learning_rate": 8.580312158108413e-05,
"loss": 0.195,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14211350679397583,
"step": 605,
"valid_targets_mean": 2804.1,
"valid_targets_min": 614
},
{
"epoch": 2.019867549668874,
"grad_norm": 0.8295625743759771,
"learning_rate": 8.557421163363735e-05,
"loss": 0.1267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13873261213302612,
"step": 610,
"valid_targets_mean": 3058.0,
"valid_targets_min": 503
},
{
"epoch": 2.0364238410596025,
"grad_norm": 0.31084477092205376,
"learning_rate": 8.534378160792159e-05,
"loss": 0.1194,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11524184048175812,
"step": 615,
"valid_targets_mean": 3675.6,
"valid_targets_min": 397
},
{
"epoch": 2.052980132450331,
"grad_norm": 0.3630068106894064,
"learning_rate": 8.511184135016134e-05,
"loss": 0.1251,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12994585931301117,
"step": 620,
"valid_targets_mean": 3232.6,
"valid_targets_min": 933
},
{
"epoch": 2.0695364238410594,
"grad_norm": 0.3642377532132808,
"learning_rate": 8.487840077111302e-05,
"loss": 0.1241,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12537671625614166,
"step": 625,
"valid_targets_mean": 3249.1,
"valid_targets_min": 809
},
{
"epoch": 2.0860927152317883,
"grad_norm": 0.3358941305355801,
"learning_rate": 8.464346984564138e-05,
"loss": 0.1288,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1243174821138382,
"step": 630,
"valid_targets_mean": 3312.7,
"valid_targets_min": 422
},
{
"epoch": 2.1026490066225167,
"grad_norm": 0.35151837218149495,
"learning_rate": 8.440705861229344e-05,
"loss": 0.1317,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12233859300613403,
"step": 635,
"valid_targets_mean": 3643.2,
"valid_targets_min": 1188
},
{
"epoch": 2.119205298013245,
"grad_norm": 0.3565052347840762,
"learning_rate": 8.416917717286939e-05,
"loss": 0.1197,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12404929101467133,
"step": 640,
"valid_targets_mean": 3201.8,
"valid_targets_min": 798
},
{
"epoch": 2.1357615894039736,
"grad_norm": 0.32279967266373655,
"learning_rate": 8.392983569199102e-05,
"loss": 0.1294,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11819366365671158,
"step": 645,
"valid_targets_mean": 3314.7,
"valid_targets_min": 452
},
{
"epoch": 2.152317880794702,
"grad_norm": 0.41776582766355863,
"learning_rate": 8.368904439666738e-05,
"loss": 0.13,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1355164498090744,
"step": 650,
"valid_targets_mean": 2200.0,
"valid_targets_min": 304
},
{
"epoch": 2.1688741721854305,
"grad_norm": 0.4057220710157651,
"learning_rate": 8.344681357585773e-05,
"loss": 0.1288,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13748055696487427,
"step": 655,
"valid_targets_mean": 2722.0,
"valid_targets_min": 308
},
{
"epoch": 2.185430463576159,
"grad_norm": 0.3537858088211586,
"learning_rate": 8.3203153580032e-05,
"loss": 0.1278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12586785852909088,
"step": 660,
"valid_targets_mean": 2863.7,
"valid_targets_min": 254
},
{
"epoch": 2.2019867549668874,
"grad_norm": 0.39201784897237507,
"learning_rate": 8.295807482072842e-05,
"loss": 0.1246,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13615946471691132,
"step": 665,
"valid_targets_mean": 2952.1,
"valid_targets_min": 849
},
{
"epoch": 2.218543046357616,
"grad_norm": 0.31295795477583843,
"learning_rate": 8.271158777010868e-05,
"loss": 0.1281,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12042751163244247,
"step": 670,
"valid_targets_mean": 3436.0,
"valid_targets_min": 541
},
{
"epoch": 2.2350993377483444,
"grad_norm": 0.33578417162244156,
"learning_rate": 8.246370296051045e-05,
"loss": 0.1273,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11698208749294281,
"step": 675,
"valid_targets_mean": 3144.2,
"valid_targets_min": 379
},
{
"epoch": 2.251655629139073,
"grad_norm": 0.3402187670378416,
"learning_rate": 8.221443098399732e-05,
"loss": 0.1217,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11877850443124771,
"step": 680,
"valid_targets_mean": 2992.1,
"valid_targets_min": 758
},
{
"epoch": 2.2682119205298013,
"grad_norm": 0.3602241680417054,
"learning_rate": 8.196378249190626e-05,
"loss": 0.1336,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13987311720848083,
"step": 685,
"valid_targets_mean": 3183.0,
"valid_targets_min": 678
},
{
"epoch": 2.2847682119205297,
"grad_norm": 0.34828354873641876,
"learning_rate": 8.17117681943924e-05,
"loss": 0.1349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14888405799865723,
"step": 690,
"valid_targets_mean": 3266.5,
"valid_targets_min": 511
},
{
"epoch": 2.301324503311258,
"grad_norm": 0.3702564459028072,
"learning_rate": 8.145839885997146e-05,
"loss": 0.128,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13586509227752686,
"step": 695,
"valid_targets_mean": 2921.9,
"valid_targets_min": 585
},
{
"epoch": 2.3178807947019866,
"grad_norm": 0.31817008615640685,
"learning_rate": 8.120368531505961e-05,
"loss": 0.128,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12238563597202301,
"step": 700,
"valid_targets_mean": 3795.2,
"valid_targets_min": 810
},
{
"epoch": 2.334437086092715,
"grad_norm": 0.31644194321275865,
"learning_rate": 8.094763844351078e-05,
"loss": 0.1267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12229268252849579,
"step": 705,
"valid_targets_mean": 3440.1,
"valid_targets_min": 705
},
{
"epoch": 2.3509933774834435,
"grad_norm": 0.31935712150450846,
"learning_rate": 8.069026918615172e-05,
"loss": 0.1323,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13009533286094666,
"step": 710,
"valid_targets_mean": 3700.7,
"valid_targets_min": 455
},
{
"epoch": 2.3675496688741724,
"grad_norm": 0.3223577073595666,
"learning_rate": 8.04315885403144e-05,
"loss": 0.1235,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11702422797679901,
"step": 715,
"valid_targets_mean": 3570.3,
"valid_targets_min": 257
},
{
"epoch": 2.384105960264901,
"grad_norm": 0.37461445017048595,
"learning_rate": 8.017160755936614e-05,
"loss": 0.1282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14748725295066833,
"step": 720,
"valid_targets_mean": 3237.3,
"valid_targets_min": 288
},
{
"epoch": 2.4006622516556293,
"grad_norm": 0.333468121399865,
"learning_rate": 7.991033735223729e-05,
"loss": 0.1341,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13057513535022736,
"step": 725,
"valid_targets_mean": 3064.9,
"valid_targets_min": 316
},
{
"epoch": 2.4172185430463577,
"grad_norm": 0.38403609747016776,
"learning_rate": 7.964778908294657e-05,
"loss": 0.1339,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13628852367401123,
"step": 730,
"valid_targets_mean": 2438.7,
"valid_targets_min": 748
},
{
"epoch": 2.433774834437086,
"grad_norm": 0.35988887050000995,
"learning_rate": 7.938397397012402e-05,
"loss": 0.1355,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1316654086112976,
"step": 735,
"valid_targets_mean": 3053.1,
"valid_targets_min": 703
},
{
"epoch": 2.4503311258278146,
"grad_norm": 0.35485379782062176,
"learning_rate": 7.911890328653156e-05,
"loss": 0.133,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1298218071460724,
"step": 740,
"valid_targets_mean": 3239.6,
"valid_targets_min": 805
},
{
"epoch": 2.466887417218543,
"grad_norm": 0.30427836115700374,
"learning_rate": 7.885258835858148e-05,
"loss": 0.1321,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12050586193799973,
"step": 745,
"valid_targets_mean": 3601.6,
"valid_targets_min": 1207
},
{
"epoch": 2.4834437086092715,
"grad_norm": 0.3444059855869074,
"learning_rate": 7.858504056585227e-05,
"loss": 0.1285,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12549395859241486,
"step": 750,
"valid_targets_mean": 3273.3,
"valid_targets_min": 852
},
{
"epoch": 2.5,
"grad_norm": 0.3677265207302808,
"learning_rate": 7.83162713406025e-05,
"loss": 0.1352,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1380949467420578,
"step": 755,
"valid_targets_mean": 2687.6,
"valid_targets_min": 417
},
{
"epoch": 2.5165562913907285,
"grad_norm": 0.3986486315910722,
"learning_rate": 7.804629216728227e-05,
"loss": 0.1422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1460767537355423,
"step": 760,
"valid_targets_mean": 3007.6,
"valid_targets_min": 755
},
{
"epoch": 2.533112582781457,
"grad_norm": 0.3613678844297741,
"learning_rate": 7.777511458204252e-05,
"loss": 0.1333,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13656137883663177,
"step": 765,
"valid_targets_mean": 2719.0,
"valid_targets_min": 563
},
{
"epoch": 2.5496688741721854,
"grad_norm": 0.3717037676122537,
"learning_rate": 7.750275017224207e-05,
"loss": 0.1306,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13513994216918945,
"step": 770,
"valid_targets_mean": 2590.1,
"valid_targets_min": 489
},
{
"epoch": 2.566225165562914,
"grad_norm": 0.30605938884997713,
"learning_rate": 7.722921057595244e-05,
"loss": 0.1351,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1248236745595932,
"step": 775,
"valid_targets_mean": 3916.7,
"valid_targets_min": 385
},
{
"epoch": 2.5827814569536423,
"grad_norm": 0.28594294280425714,
"learning_rate": 7.695450748146069e-05,
"loss": 0.1324,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11908982694149017,
"step": 780,
"valid_targets_mean": 3497.2,
"valid_targets_min": 346
},
{
"epoch": 2.5993377483443707,
"grad_norm": 0.33275588358626473,
"learning_rate": 7.66786526267698e-05,
"loss": 0.1314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12390835583209991,
"step": 785,
"valid_targets_mean": 3211.5,
"valid_targets_min": 430
},
{
"epoch": 2.6158940397350996,
"grad_norm": 0.3383366193385283,
"learning_rate": 7.640165779909735e-05,
"loss": 0.1359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13580447435379028,
"step": 790,
"valid_targets_mean": 3250.2,
"valid_targets_min": 639
},
{
"epoch": 2.6324503311258276,
"grad_norm": 0.3107558295311167,
"learning_rate": 7.612353483437158e-05,
"loss": 0.1367,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13051705062389374,
"step": 795,
"valid_targets_mean": 3639.9,
"valid_targets_min": 795
},
{
"epoch": 2.6490066225165565,
"grad_norm": 0.36184191442055263,
"learning_rate": 7.584429561672586e-05,
"loss": 0.1326,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1366756111383438,
"step": 800,
"valid_targets_mean": 3003.5,
"valid_targets_min": 476
},
{
"epoch": 2.6655629139072845,
"grad_norm": 0.3413224205451187,
"learning_rate": 7.556395207799077e-05,
"loss": 0.1307,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13229820132255554,
"step": 805,
"valid_targets_mean": 3005.9,
"valid_targets_min": 637
},
{
"epoch": 2.6821192052980134,
"grad_norm": 0.28247014882550064,
"learning_rate": 7.528251619718431e-05,
"loss": 0.1317,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11838458478450775,
"step": 810,
"valid_targets_mean": 3923.9,
"valid_targets_min": 530
},
{
"epoch": 2.698675496688742,
"grad_norm": 0.36971304320381926,
"learning_rate": 7.500000000000001e-05,
"loss": 0.1313,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13400927186012268,
"step": 815,
"valid_targets_mean": 3018.0,
"valid_targets_min": 290
},
{
"epoch": 2.7152317880794703,
"grad_norm": 0.32652007842308994,
"learning_rate": 7.471641555829307e-05,
"loss": 0.1289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12828126549720764,
"step": 820,
"valid_targets_mean": 3543.0,
"valid_targets_min": 1118
},
{
"epoch": 2.7317880794701987,
"grad_norm": 0.29930593803859495,
"learning_rate": 7.443177498956453e-05,
"loss": 0.1349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12653376162052155,
"step": 825,
"valid_targets_mean": 3559.8,
"valid_targets_min": 325
},
{
"epoch": 2.748344370860927,
"grad_norm": 0.30332685268784965,
"learning_rate": 7.414609045644356e-05,
"loss": 0.1302,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12665782868862152,
"step": 830,
"valid_targets_mean": 3858.2,
"valid_targets_min": 669
},
{
"epoch": 2.7649006622516556,
"grad_norm": 0.2847169886980276,
"learning_rate": 7.385937416616767e-05,
"loss": 0.1282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11664550006389618,
"step": 835,
"valid_targets_mean": 3870.9,
"valid_targets_min": 681
},
{
"epoch": 2.781456953642384,
"grad_norm": 0.3643915403596576,
"learning_rate": 7.357163837006111e-05,
"loss": 0.1394,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14505338668823242,
"step": 840,
"valid_targets_mean": 2784.2,
"valid_targets_min": 953
},
{
"epoch": 2.7980132450331126,
"grad_norm": 0.3451967187612331,
"learning_rate": 7.32828953630114e-05,
"loss": 0.1337,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1330152451992035,
"step": 845,
"valid_targets_mean": 3206.1,
"valid_targets_min": 632
},
{
"epoch": 2.814569536423841,
"grad_norm": 0.2753986431142239,
"learning_rate": 7.2993157482944e-05,
"loss": 0.1323,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12577594816684723,
"step": 850,
"valid_targets_mean": 4482.1,
"valid_targets_min": 1162
},
{
"epoch": 2.8311258278145695,
"grad_norm": 0.3639461654352118,
"learning_rate": 7.2702437110295e-05,
"loss": 0.1301,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14124904572963715,
"step": 855,
"valid_targets_mean": 2888.8,
"valid_targets_min": 1064
},
{
"epoch": 2.847682119205298,
"grad_norm": 0.3360649001314586,
"learning_rate": 7.241074666748227e-05,
"loss": 0.1349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13045209646224976,
"step": 860,
"valid_targets_mean": 3387.1,
"valid_targets_min": 455
},
{
"epoch": 2.8642384105960264,
"grad_norm": 0.31720005711656,
"learning_rate": 7.211809861837451e-05,
"loss": 0.1284,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11310358345508575,
"step": 865,
"valid_targets_mean": 2787.1,
"valid_targets_min": 341
},
{
"epoch": 2.880794701986755,
"grad_norm": 0.31969092965104995,
"learning_rate": 7.182450546775874e-05,
"loss": 0.129,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14063483476638794,
"step": 870,
"valid_targets_mean": 3512.4,
"valid_targets_min": 577
},
{
"epoch": 2.8973509933774837,
"grad_norm": 0.31847021760829747,
"learning_rate": 7.152997976080595e-05,
"loss": 0.1331,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1267443597316742,
"step": 875,
"valid_targets_mean": 3578.7,
"valid_targets_min": 669
},
{
"epoch": 2.9139072847682117,
"grad_norm": 0.36679827710956553,
"learning_rate": 7.123453408253508e-05,
"loss": 0.1373,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1326959729194641,
"step": 880,
"valid_targets_mean": 2838.6,
"valid_targets_min": 598
},
{
"epoch": 2.9304635761589406,
"grad_norm": 0.32409518547104016,
"learning_rate": 7.093818105727522e-05,
"loss": 0.1301,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1208677589893341,
"step": 885,
"valid_targets_mean": 3129.2,
"valid_targets_min": 811
},
{
"epoch": 2.9470198675496686,
"grad_norm": 0.3336307356393745,
"learning_rate": 7.064093334812623e-05,
"loss": 0.1333,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13538572192192078,
"step": 890,
"valid_targets_mean": 3566.7,
"valid_targets_min": 773
},
{
"epoch": 2.9635761589403975,
"grad_norm": 0.3228439365679719,
"learning_rate": 7.034280365641759e-05,
"loss": 0.1281,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1289815604686737,
"step": 895,
"valid_targets_mean": 3346.9,
"valid_targets_min": 890
},
{
"epoch": 2.980132450331126,
"grad_norm": 0.32150791807579643,
"learning_rate": 7.00438047211657e-05,
"loss": 0.1334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1305575668811798,
"step": 900,
"valid_targets_mean": 3525.2,
"valid_targets_min": 1008
},
{
"epoch": 2.9966887417218544,
"grad_norm": 0.376747524173855,
"learning_rate": 6.974394931852956e-05,
"loss": 0.1371,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1445886343717575,
"step": 905,
"valid_targets_mean": 2936.7,
"valid_targets_min": 316
},
{
"epoch": 3.013245033112583,
"grad_norm": 0.4015710050531782,
"learning_rate": 6.944325026126484e-05,
"loss": 0.0797,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07109177857637405,
"step": 910,
"valid_targets_mean": 2848.7,
"valid_targets_min": 991
},
{
"epoch": 3.0298013245033113,
"grad_norm": 0.3290408043520342,
"learning_rate": 6.914172039817634e-05,
"loss": 0.0712,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07119590044021606,
"step": 915,
"valid_targets_mean": 3237.0,
"valid_targets_min": 362
},
{
"epoch": 3.0463576158940397,
"grad_norm": 0.29292122484351646,
"learning_rate": 6.883937261356909e-05,
"loss": 0.0671,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05982881411910057,
"step": 920,
"valid_targets_mean": 2963.9,
"valid_targets_min": 254
},
{
"epoch": 3.062913907284768,
"grad_norm": 0.3691527549235315,
"learning_rate": 6.853621982669765e-05,
"loss": 0.0689,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07329269498586655,
"step": 925,
"valid_targets_mean": 3019.8,
"valid_targets_min": 1032
},
{
"epoch": 3.0794701986754967,
"grad_norm": 0.308885957961717,
"learning_rate": 6.82322749912142e-05,
"loss": 0.0673,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06925823539495468,
"step": 930,
"valid_targets_mean": 3703.9,
"valid_targets_min": 720
},
{
"epoch": 3.096026490066225,
"grad_norm": 0.3770486735440148,
"learning_rate": 6.792755109461499e-05,
"loss": 0.0687,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07204819470643997,
"step": 935,
"valid_targets_mean": 2704.8,
"valid_targets_min": 689
},
{
"epoch": 3.1125827814569536,
"grad_norm": 0.3484668358061707,
"learning_rate": 6.762206115768529e-05,
"loss": 0.0752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08134443312883377,
"step": 940,
"valid_targets_mean": 3387.0,
"valid_targets_min": 656
},
{
"epoch": 3.129139072847682,
"grad_norm": 0.33306284819882687,
"learning_rate": 6.731581823394324e-05,
"loss": 0.0698,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.070025235414505,
"step": 945,
"valid_targets_mean": 3198.5,
"valid_targets_min": 436
},
{
"epoch": 3.1456953642384105,
"grad_norm": 0.3140376381190174,
"learning_rate": 6.700883540908184e-05,
"loss": 0.0731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06381855905056,
"step": 950,
"valid_targets_mean": 3394.9,
"valid_targets_min": 900
},
{
"epoch": 3.162251655629139,
"grad_norm": 0.3097637254784413,
"learning_rate": 6.670112580040996e-05,
"loss": 0.0721,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0707109123468399,
"step": 955,
"valid_targets_mean": 3627.6,
"valid_targets_min": 598
},
{
"epoch": 3.1788079470198674,
"grad_norm": 0.30806814298493973,
"learning_rate": 6.639270255629176e-05,
"loss": 0.071,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06404514610767365,
"step": 960,
"valid_targets_mean": 3344.4,
"valid_targets_min": 585
},
{
"epoch": 3.195364238410596,
"grad_norm": 0.3387776387762296,
"learning_rate": 6.608357885558485e-05,
"loss": 0.0696,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07401485741138458,
"step": 965,
"valid_targets_mean": 3363.1,
"valid_targets_min": 717
},
{
"epoch": 3.2119205298013247,
"grad_norm": 0.30970754443570575,
"learning_rate": 6.577376790707729e-05,
"loss": 0.0735,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06643778085708618,
"step": 970,
"valid_targets_mean": 3226.3,
"valid_targets_min": 332
},
{
"epoch": 3.228476821192053,
"grad_norm": 0.33424610516875725,
"learning_rate": 6.5463282948923e-05,
"loss": 0.0713,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06977947056293488,
"step": 975,
"valid_targets_mean": 2801.7,
"valid_targets_min": 895
},
{
"epoch": 3.2450331125827816,
"grad_norm": 0.3106924553989161,
"learning_rate": 6.51521372480762e-05,
"loss": 0.0737,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07378740608692169,
"step": 980,
"valid_targets_mean": 3694.0,
"valid_targets_min": 791
},
{
"epoch": 3.26158940397351,
"grad_norm": 0.3612557078304956,
"learning_rate": 6.484034409972457e-05,
"loss": 0.0731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07424025982618332,
"step": 985,
"valid_targets_mean": 2580.8,
"valid_targets_min": 501
},
{
"epoch": 3.2781456953642385,
"grad_norm": 0.31997339393997704,
"learning_rate": 6.452791682672097e-05,
"loss": 0.0738,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06445827335119247,
"step": 990,
"valid_targets_mean": 3426.0,
"valid_targets_min": 844
},
{
"epoch": 3.294701986754967,
"grad_norm": 0.34998633130027557,
"learning_rate": 6.421486877901437e-05,
"loss": 0.078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07630739361047745,
"step": 995,
"valid_targets_mean": 3084.0,
"valid_targets_min": 577
},
{
"epoch": 3.3112582781456954,
"grad_norm": 0.32037831037767844,
"learning_rate": 6.390121333307921e-05,
"loss": 0.0703,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0728875920176506,
"step": 1000,
"valid_targets_mean": 3137.2,
"valid_targets_min": 895
},
{
"epoch": 3.327814569536424,
"grad_norm": 0.31228161182414027,
"learning_rate": 6.358696389134402e-05,
"loss": 0.075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06733287870883942,
"step": 1005,
"valid_targets_mean": 3217.8,
"valid_targets_min": 325
},
{
"epoch": 3.3443708609271523,
"grad_norm": 0.36518588719864375,
"learning_rate": 6.327213388161859e-05,
"loss": 0.079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07602456212043762,
"step": 1010,
"valid_targets_mean": 2730.5,
"valid_targets_min": 794
},
{
"epoch": 3.3609271523178808,
"grad_norm": 0.2905425966478273,
"learning_rate": 6.295673675652024e-05,
"loss": 0.0766,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07344558835029602,
"step": 1015,
"valid_targets_mean": 3587.8,
"valid_targets_min": 811
},
{
"epoch": 3.377483443708609,
"grad_norm": 0.3181322828637574,
"learning_rate": 6.264078599289903e-05,
"loss": 0.0773,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07851684093475342,
"step": 1020,
"valid_targets_mean": 3359.2,
"valid_targets_min": 754
},
{
"epoch": 3.3940397350993377,
"grad_norm": 0.38499597702794097,
"learning_rate": 6.232429509126189e-05,
"loss": 0.0784,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08062493801116943,
"step": 1025,
"valid_targets_mean": 2458.4,
"valid_targets_min": 300
},
{
"epoch": 3.410596026490066,
"grad_norm": 0.3364693576315686,
"learning_rate": 6.200727757519565e-05,
"loss": 0.0731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08122321963310242,
"step": 1030,
"valid_targets_mean": 3520.3,
"valid_targets_min": 814
},
{
"epoch": 3.4271523178807946,
"grad_norm": 0.31839705885816855,
"learning_rate": 6.168974699078936e-05,
"loss": 0.0733,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07207996398210526,
"step": 1035,
"valid_targets_mean": 3285.1,
"valid_targets_min": 388
},
{
"epoch": 3.443708609271523,
"grad_norm": 0.3617165965061275,
"learning_rate": 6.137171690605533e-05,
"loss": 0.0752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07620041817426682,
"step": 1040,
"valid_targets_mean": 2670.6,
"valid_targets_min": 430
},
{
"epoch": 3.460264900662252,
"grad_norm": 0.33864904521389044,
"learning_rate": 6.105320091034936e-05,
"loss": 0.0758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07385677099227905,
"step": 1045,
"valid_targets_mean": 3056.7,
"valid_targets_min": 316
},
{
"epoch": 3.47682119205298,
"grad_norm": 0.33071889128562487,
"learning_rate": 6.0734212613790206e-05,
"loss": 0.0753,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07516039162874222,
"step": 1050,
"valid_targets_mean": 3002.4,
"valid_targets_min": 848
},
{
"epoch": 3.493377483443709,
"grad_norm": 0.3666520598517765,
"learning_rate": 6.041476564667785e-05,
"loss": 0.0745,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08016708493232727,
"step": 1055,
"valid_targets_mean": 2683.3,
"valid_targets_min": 476
},
{
"epoch": 3.5099337748344372,
"grad_norm": 0.3299274873874484,
"learning_rate": 6.009487365891123e-05,
"loss": 0.0768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07682723551988602,
"step": 1060,
"valid_targets_mean": 3033.1,
"valid_targets_min": 681
},
{
"epoch": 3.5264900662251657,
"grad_norm": 0.30672017302586696,
"learning_rate": 5.977455031940491e-05,
"loss": 0.0718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07092280685901642,
"step": 1065,
"valid_targets_mean": 3331.8,
"valid_targets_min": 327
},
{
"epoch": 3.543046357615894,
"grad_norm": 0.2829467497663885,
"learning_rate": 5.9453809315504975e-05,
"loss": 0.0732,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06973779201507568,
"step": 1070,
"valid_targets_mean": 4080.5,
"valid_targets_min": 1063
},
{
"epoch": 3.5596026490066226,
"grad_norm": 0.2998251440174961,
"learning_rate": 5.913266435240429e-05,
"loss": 0.0751,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0719188004732132,
"step": 1075,
"valid_targets_mean": 3714.2,
"valid_targets_min": 1184
},
{
"epoch": 3.576158940397351,
"grad_norm": 0.311775415344261,
"learning_rate": 5.881112915255672e-05,
"loss": 0.0774,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07658076286315918,
"step": 1080,
"valid_targets_mean": 3538.7,
"valid_targets_min": 830
},
{
"epoch": 3.5927152317880795,
"grad_norm": 0.3169375832902381,
"learning_rate": 5.848921745509094e-05,
"loss": 0.0735,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07392723858356476,
"step": 1085,
"valid_targets_mean": 3601.6,
"valid_targets_min": 946
},
{
"epoch": 3.609271523178808,
"grad_norm": 0.336927142842697,
"learning_rate": 5.8166943015223226e-05,
"loss": 0.0728,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07585244625806808,
"step": 1090,
"valid_targets_mean": 3366.2,
"valid_targets_min": 367
},
{
"epoch": 3.6258278145695364,
"grad_norm": 0.3548225489642839,
"learning_rate": 5.784431960366981e-05,
"loss": 0.0797,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08079524338245392,
"step": 1095,
"valid_targets_mean": 3059.3,
"valid_targets_min": 1245
},
{
"epoch": 3.642384105960265,
"grad_norm": 0.31619920886400144,
"learning_rate": 5.7521361006058394e-05,
"loss": 0.0742,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07311496138572693,
"step": 1100,
"valid_targets_mean": 3403.0,
"valid_targets_min": 878
},
{
"epoch": 3.6589403973509933,
"grad_norm": 0.3610815294100483,
"learning_rate": 5.719808102233907e-05,
"loss": 0.0759,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07423543930053711,
"step": 1105,
"valid_targets_mean": 2798.7,
"valid_targets_min": 829
},
{
"epoch": 3.6754966887417218,
"grad_norm": 0.32273266216313284,
"learning_rate": 5.687449346619471e-05,
"loss": 0.073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07126402109861374,
"step": 1110,
"valid_targets_mean": 3248.9,
"valid_targets_min": 414
},
{
"epoch": 3.69205298013245,
"grad_norm": 0.3520674586724713,
"learning_rate": 5.65506121644507e-05,
"loss": 0.0717,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07669555395841599,
"step": 1115,
"valid_targets_mean": 3141.8,
"valid_targets_min": 511
},
{
"epoch": 3.7086092715231787,
"grad_norm": 0.31248502452072835,
"learning_rate": 5.622645095648411e-05,
"loss": 0.0757,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06829120218753815,
"step": 1120,
"valid_targets_mean": 3081.1,
"valid_targets_min": 257
},
{
"epoch": 3.725165562913907,
"grad_norm": 0.34994651254868825,
"learning_rate": 5.590202369363233e-05,
"loss": 0.0774,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08172380924224854,
"step": 1125,
"valid_targets_mean": 3339.2,
"valid_targets_min": 813
},
{
"epoch": 3.741721854304636,
"grad_norm": 0.32204782136596366,
"learning_rate": 5.557734423860123e-05,
"loss": 0.0737,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07117553055286407,
"step": 1130,
"valid_targets_mean": 3050.5,
"valid_targets_min": 486
},
{
"epoch": 3.758278145695364,
"grad_norm": 0.33760167441590533,
"learning_rate": 5.5252426464872773e-05,
"loss": 0.0748,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07491283863782883,
"step": 1135,
"valid_targets_mean": 3249.1,
"valid_targets_min": 598
},
{
"epoch": 3.774834437086093,
"grad_norm": 0.30529221520921,
"learning_rate": 5.4927284256112244e-05,
"loss": 0.0731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07238440215587616,
"step": 1140,
"valid_targets_mean": 3514.3,
"valid_targets_min": 649
},
{
"epoch": 3.7913907284768213,
"grad_norm": 0.3448828844921914,
"learning_rate": 5.4601931505575e-05,
"loss": 0.0748,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07113992422819138,
"step": 1145,
"valid_targets_mean": 3024.5,
"valid_targets_min": 1354
},
{
"epoch": 3.80794701986755,
"grad_norm": 0.36021028175064357,
"learning_rate": 5.427638211551278e-05,
"loss": 0.0741,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0711931586265564,
"step": 1150,
"valid_targets_mean": 3122.4,
"valid_targets_min": 450
},
{
"epoch": 3.8245033112582782,
"grad_norm": 0.30920120954263597,
"learning_rate": 5.39506499965797e-05,
"loss": 0.0737,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07597890496253967,
"step": 1155,
"valid_targets_mean": 3739.3,
"valid_targets_min": 669
},
{
"epoch": 3.8410596026490067,
"grad_norm": 0.2838397306045775,
"learning_rate": 5.362474906723781e-05,
"loss": 0.075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07047129422426224,
"step": 1160,
"valid_targets_mean": 4206.1,
"valid_targets_min": 875
},
{
"epoch": 3.857615894039735,
"grad_norm": 0.31355250168525867,
"learning_rate": 5.329869325316246e-05,
"loss": 0.0721,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0692002922296524,
"step": 1165,
"valid_targets_mean": 3252.1,
"valid_targets_min": 382
},
{
"epoch": 3.8741721854304636,
"grad_norm": 0.3178345488543423,
"learning_rate": 5.297249648664712e-05,
"loss": 0.073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07440881431102753,
"step": 1170,
"valid_targets_mean": 3362.3,
"valid_targets_min": 674
},
{
"epoch": 3.890728476821192,
"grad_norm": 0.33347003831907135,
"learning_rate": 5.2646172706008156e-05,
"loss": 0.0724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07261810451745987,
"step": 1175,
"valid_targets_mean": 2850.4,
"valid_targets_min": 516
},
{
"epoch": 3.9072847682119205,
"grad_norm": 0.30528083077165835,
"learning_rate": 5.231973585498924e-05,
"loss": 0.0739,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07194554805755615,
"step": 1180,
"valid_targets_mean": 3237.2,
"valid_targets_min": 592
},
{
"epoch": 3.923841059602649,
"grad_norm": 0.30601906819816244,
"learning_rate": 5.199319988216547e-05,
"loss": 0.0725,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0687810480594635,
"step": 1185,
"valid_targets_mean": 3128.5,
"valid_targets_min": 731
},
{
"epoch": 3.9403973509933774,
"grad_norm": 0.32887230299137504,
"learning_rate": 5.166657874034745e-05,
"loss": 0.0734,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07794922590255737,
"step": 1190,
"valid_targets_mean": 3184.7,
"valid_targets_min": 683
},
{
"epoch": 3.956953642384106,
"grad_norm": 0.3513693229868733,
"learning_rate": 5.133988638598497e-05,
"loss": 0.0737,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07598121464252472,
"step": 1195,
"valid_targets_mean": 2701.5,
"valid_targets_min": 749
},
{
"epoch": 3.9735099337748343,
"grad_norm": 0.321162310475212,
"learning_rate": 5.101313677857078e-05,
"loss": 0.0718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06860232353210449,
"step": 1200,
"valid_targets_mean": 3029.6,
"valid_targets_min": 708
},
{
"epoch": 3.9900662251655628,
"grad_norm": 0.3250245639860087,
"learning_rate": 5.068634388004404e-05,
"loss": 0.0719,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07109785079956055,
"step": 1205,
"valid_targets_mean": 3261.3,
"valid_targets_min": 612
},
{
"epoch": 4.006622516556291,
"grad_norm": 0.2533770649612179,
"learning_rate": 5.0359521654193654e-05,
"loss": 0.0539,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03473834693431854,
"step": 1210,
"valid_targets_mean": 3162.8,
"valid_targets_min": 705
},
{
"epoch": 4.02317880794702,
"grad_norm": 0.3188909597866215,
"learning_rate": 5.003268406606176e-05,
"loss": 0.0346,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03440415859222412,
"step": 1215,
"valid_targets_mean": 3360.6,
"valid_targets_min": 714
},
{
"epoch": 4.039735099337748,
"grad_norm": 0.2726442325163795,
"learning_rate": 4.9705845081346894e-05,
"loss": 0.034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0340653732419014,
"step": 1220,
"valid_targets_mean": 3739.6,
"valid_targets_min": 1043
},
{
"epoch": 4.056291390728477,
"grad_norm": 0.31745511288964384,
"learning_rate": 4.9379018665807245e-05,
"loss": 0.038,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04656948894262314,
"step": 1225,
"valid_targets_mean": 3245.9,
"valid_targets_min": 839
},
{
"epoch": 4.072847682119205,
"grad_norm": 0.2991000198836208,
"learning_rate": 4.9052218784663947e-05,
"loss": 0.0369,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03539630025625229,
"step": 1230,
"valid_targets_mean": 3093.5,
"valid_targets_min": 413
},
{
"epoch": 4.089403973509934,
"grad_norm": 0.25948368003040384,
"learning_rate": 4.8725459402004345e-05,
"loss": 0.035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.02952433004975319,
"step": 1235,
"valid_targets_mean": 3793.8,
"valid_targets_min": 735
},
{
"epoch": 4.105960264900662,
"grad_norm": 0.34359458492349904,
"learning_rate": 4.839875448018524e-05,
"loss": 0.0378,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.041338708251714706,
"step": 1240,
"valid_targets_mean": 2485.8,
"valid_targets_min": 430
},
{
"epoch": 4.122516556291391,
"grad_norm": 0.2908865347610644,
"learning_rate": 4.807211797923639e-05,
"loss": 0.0396,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.036823902279138565,
"step": 1245,
"valid_targets_mean": 3467.5,
"valid_targets_min": 798
},
{
"epoch": 4.139072847682119,
"grad_norm": 0.2798763256222066,
"learning_rate": 4.7745563856263856e-05,
"loss": 0.0375,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.034374311566352844,
"step": 1250,
"valid_targets_mean": 3414.0,
"valid_targets_min": 587
},
{
"epoch": 4.155629139072848,
"grad_norm": 0.27868108487255805,
"learning_rate": 4.741910606485385e-05,
"loss": 0.0382,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.038354240357875824,
"step": 1255,
"valid_targets_mean": 3173.6,
"valid_targets_min": 665
},
{
"epoch": 4.172185430463577,
"grad_norm": 0.32533822087411113,
"learning_rate": 4.709275855447621e-05,
"loss": 0.0376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03936341404914856,
"step": 1260,
"valid_targets_mean": 2791.5,
"valid_targets_min": 849
},
{
"epoch": 4.188741721854305,
"grad_norm": 0.2822116754105137,
"learning_rate": 4.676653526988858e-05,
"loss": 0.0389,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.033856384456157684,
"step": 1265,
"valid_targets_mean": 2821.0,
"valid_targets_min": 397
},
{
"epoch": 4.2052980132450335,
"grad_norm": 0.3267014135141966,
"learning_rate": 4.644045015054045e-05,
"loss": 0.0379,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.040697213262319565,
"step": 1270,
"valid_targets_mean": 2822.8,
"valid_targets_min": 693
},
{
"epoch": 4.2218543046357615,
"grad_norm": 0.2860739980937301,
"learning_rate": 4.61145171299775e-05,
"loss": 0.0398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03861942142248154,
"step": 1275,
"valid_targets_mean": 3329.6,
"valid_targets_min": 426
},
{
"epoch": 4.23841059602649,
"grad_norm": 0.2877709521571235,
"learning_rate": 4.578875013524631e-05,
"loss": 0.0388,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03869316354393959,
"step": 1280,
"valid_targets_mean": 3138.3,
"valid_targets_min": 337
},
{
"epoch": 4.254966887417218,
"grad_norm": 0.3060549292112492,
"learning_rate": 4.546316308629915e-05,
"loss": 0.04,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0416206419467926,
"step": 1285,
"valid_targets_mean": 3077.0,
"valid_targets_min": 422
},
{
"epoch": 4.271523178807947,
"grad_norm": 0.29934154912049604,
"learning_rate": 4.513776989539934e-05,
"loss": 0.0372,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03901611641049385,
"step": 1290,
"valid_targets_mean": 3431.0,
"valid_targets_min": 1016
},
{
"epoch": 4.288079470198675,
"grad_norm": 0.30153816528954963,
"learning_rate": 4.481258446652662e-05,
"loss": 0.0401,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03662698715925217,
"step": 1295,
"valid_targets_mean": 2781.1,
"valid_targets_min": 489
},
{
"epoch": 4.304635761589404,
"grad_norm": 0.2902613469378036,
"learning_rate": 4.448762069478309e-05,
"loss": 0.0406,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.042478203773498535,
"step": 1300,
"valid_targets_mean": 3234.8,
"valid_targets_min": 461
},
{
"epoch": 4.321192052980132,
"grad_norm": 0.30546029481723797,
"learning_rate": 4.416289246579954e-05,
"loss": 0.0403,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04381430894136429,
"step": 1305,
"valid_targets_mean": 3041.8,
"valid_targets_min": 840
},
{
"epoch": 4.337748344370861,
"grad_norm": 0.3149357538714551,
"learning_rate": 4.383841365514208e-05,
"loss": 0.04,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04181423783302307,
"step": 1310,
"valid_targets_mean": 2837.9,
"valid_targets_min": 325
},
{
"epoch": 4.354304635761589,
"grad_norm": 0.3354678050399959,
"learning_rate": 4.3514198127719143e-05,
"loss": 0.0379,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04507702961564064,
"step": 1315,
"valid_targets_mean": 2825.4,
"valid_targets_min": 401
},
{
"epoch": 4.370860927152318,
"grad_norm": 0.29063622571729403,
"learning_rate": 4.31902597371892e-05,
"loss": 0.04,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03782668709754944,
"step": 1320,
"valid_targets_mean": 3161.4,
"valid_targets_min": 624
},
{
"epoch": 4.387417218543046,
"grad_norm": 0.2760851354461555,
"learning_rate": 4.2866612325368726e-05,
"loss": 0.0403,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.035794857889413834,
"step": 1325,
"valid_targets_mean": 3501.7,
"valid_targets_min": 295
},
{
"epoch": 4.403973509933775,
"grad_norm": 0.29707936753598824,
"learning_rate": 4.25432697216407e-05,
"loss": 0.043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04659942537546158,
"step": 1330,
"valid_targets_mean": 3504.5,
"valid_targets_min": 996
},
{
"epoch": 4.420529801324503,
"grad_norm": 0.2964424869739796,
"learning_rate": 4.2220245742363726e-05,
"loss": 0.0382,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03739917278289795,
"step": 1335,
"valid_targets_mean": 3310.0,
"valid_targets_min": 373
},
{
"epoch": 4.437086092715232,
"grad_norm": 0.2834180767058586,
"learning_rate": 4.1897554190281694e-05,
"loss": 0.0397,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.038497645407915115,
"step": 1340,
"valid_targets_mean": 3069.7,
"valid_targets_min": 314
},
{
"epoch": 4.45364238410596,
"grad_norm": 0.34378559520925767,
"learning_rate": 4.157520885393392e-05,
"loss": 0.0394,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.041668981313705444,
"step": 1345,
"valid_targets_mean": 2610.1,
"valid_targets_min": 451
},
{
"epoch": 4.470198675496689,
"grad_norm": 0.3045695069414927,
"learning_rate": 4.125322350706598e-05,
"loss": 0.0394,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04001860320568085,
"step": 1350,
"valid_targets_mean": 2893.6,
"valid_targets_min": 464
},
{
"epoch": 4.486754966887418,
"grad_norm": 0.2858627188968983,
"learning_rate": 4.09316119080412e-05,
"loss": 0.0379,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.040242284536361694,
"step": 1355,
"valid_targets_mean": 3633.8,
"valid_targets_min": 1144
},
{
"epoch": 4.503311258278146,
"grad_norm": 0.30965194360042086,
"learning_rate": 4.0610387799252785e-05,
"loss": 0.0389,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04035414755344391,
"step": 1360,
"valid_targets_mean": 2557.2,
"valid_targets_min": 270
},
{
"epoch": 4.5198675496688745,
"grad_norm": 0.29416379069351983,
"learning_rate": 4.0289564906536486e-05,
"loss": 0.0381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03775478154420853,
"step": 1365,
"valid_targets_mean": 3066.1,
"valid_targets_min": 981
},
{
"epoch": 4.5364238410596025,
"grad_norm": 0.3218461006752068,
"learning_rate": 3.9969156938584215e-05,
"loss": 0.0407,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04487745091319084,
"step": 1370,
"valid_targets_mean": 2454.5,
"valid_targets_min": 389
},
{
"epoch": 4.552980132450331,
"grad_norm": 0.3193960141162944,
"learning_rate": 3.964917758635828e-05,
"loss": 0.0393,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04038995876908302,
"step": 1375,
"valid_targets_mean": 2990.7,
"valid_targets_min": 970
},
{
"epoch": 4.569536423841059,
"grad_norm": 0.3027889611241044,
"learning_rate": 3.932964052250626e-05,
"loss": 0.0392,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0377410426735878,
"step": 1380,
"valid_targets_mean": 2974.6,
"valid_targets_min": 673
},
{
"epoch": 4.586092715231788,
"grad_norm": 0.2891951857349599,
"learning_rate": 3.9010559400776915e-05,
"loss": 0.0371,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03795380890369415,
"step": 1385,
"valid_targets_mean": 2941.1,
"valid_targets_min": 683
},
{
"epoch": 4.602649006622516,
"grad_norm": 0.3108065930208171,
"learning_rate": 3.869194785543662e-05,
"loss": 0.0388,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.039802391082048416,
"step": 1390,
"valid_targets_mean": 3064.7,
"valid_targets_min": 626
},
{
"epoch": 4.619205298013245,
"grad_norm": 0.29665474107833184,
"learning_rate": 3.837381950068695e-05,
"loss": 0.0388,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04021240025758743,
"step": 1395,
"valid_targets_mean": 3332.1,
"valid_targets_min": 800
},
{
"epoch": 4.635761589403973,
"grad_norm": 0.2877332706410405,
"learning_rate": 3.805618793008279e-05,
"loss": 0.0371,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.036268629133701324,
"step": 1400,
"valid_targets_mean": 3082.1,
"valid_targets_min": 417
},
{
"epoch": 4.652317880794702,
"grad_norm": 0.3095652551769783,
"learning_rate": 3.7739066715951535e-05,
"loss": 0.0378,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03903259336948395,
"step": 1405,
"valid_targets_mean": 3157.4,
"valid_targets_min": 598
},
{
"epoch": 4.66887417218543,
"grad_norm": 0.289136626618137,
"learning_rate": 3.742246940881321e-05,
"loss": 0.0381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03652704134583473,
"step": 1410,
"valid_targets_mean": 3652.9,
"valid_targets_min": 370
},
{
"epoch": 4.685430463576159,
"grad_norm": 0.29141869790970154,
"learning_rate": 3.7106409536801397e-05,
"loss": 0.0392,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04092412814497948,
"step": 1415,
"valid_targets_mean": 2997.0,
"valid_targets_min": 657
},
{
"epoch": 4.701986754966887,
"grad_norm": 0.28988699702479676,
"learning_rate": 3.679090060508516e-05,
"loss": 0.0384,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03866320475935936,
"step": 1420,
"valid_targets_mean": 3081.1,
"valid_targets_min": 254
},
{
"epoch": 4.718543046357616,
"grad_norm": 0.3018786382186051,
"learning_rate": 3.6475956095292014e-05,
"loss": 0.0396,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03999405354261398,
"step": 1425,
"valid_targets_mean": 3246.0,
"valid_targets_min": 400
},
{
"epoch": 4.735099337748345,
"grad_norm": 0.28281841868896745,
"learning_rate": 3.616158946493188e-05,
"loss": 0.0394,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.038476429879665375,
"step": 1430,
"valid_targets_mean": 3319.0,
"valid_targets_min": 1001
},
{
"epoch": 4.751655629139073,
"grad_norm": 0.2783485708553036,
"learning_rate": 3.5847814146822005e-05,
"loss": 0.0357,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03993133455514908,
"step": 1435,
"valid_targets_mean": 3534.0,
"valid_targets_min": 741
},
{
"epoch": 4.768211920529802,
"grad_norm": 0.30766454854220066,
"learning_rate": 3.5534643548512944e-05,
"loss": 0.0388,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03968173265457153,
"step": 1440,
"valid_targets_mean": 2908.2,
"valid_targets_min": 972
},
{
"epoch": 4.78476821192053,
"grad_norm": 0.23848795045982416,
"learning_rate": 3.52220910517158e-05,
"loss": 0.0374,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.032834168523550034,
"step": 1445,
"valid_targets_mean": 4154.5,
"valid_targets_min": 366
},
{
"epoch": 4.801324503311259,
"grad_norm": 0.31666877207307315,
"learning_rate": 3.491017001173026e-05,
"loss": 0.0382,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.039865173399448395,
"step": 1450,
"valid_targets_mean": 2633.1,
"valid_targets_min": 322
},
{
"epoch": 4.817880794701987,
"grad_norm": 0.2856103512654398,
"learning_rate": 3.459889375687402e-05,
"loss": 0.0388,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0413297638297081,
"step": 1455,
"valid_targets_mean": 3688.2,
"valid_targets_min": 1237
},
{
"epoch": 4.8344370860927155,
"grad_norm": 0.2755977131103906,
"learning_rate": 3.428827558791323e-05,
"loss": 0.0364,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03599392622709274,
"step": 1460,
"valid_targets_mean": 3727.7,
"valid_targets_min": 484
},
{
"epoch": 4.8509933774834435,
"grad_norm": 0.33607680738922974,
"learning_rate": 3.397832877749425e-05,
"loss": 0.0386,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.040633976459503174,
"step": 1465,
"valid_targets_mean": 2460.4,
"valid_targets_min": 886
},
{
"epoch": 4.867549668874172,
"grad_norm": 0.29556721957743776,
"learning_rate": 3.3669066569576334e-05,
"loss": 0.0398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.038377679884433746,
"step": 1470,
"valid_targets_mean": 3235.2,
"valid_targets_min": 306
},
{
"epoch": 4.8841059602649,
"grad_norm": 0.2718708461860853,
"learning_rate": 3.336050217886588e-05,
"loss": 0.037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.033986881375312805,
"step": 1475,
"valid_targets_mean": 3247.7,
"valid_targets_min": 337
},
{
"epoch": 4.900662251655629,
"grad_norm": 0.29046988783095157,
"learning_rate": 3.3052648790251715e-05,
"loss": 0.0379,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.039075400680303574,
"step": 1480,
"valid_targets_mean": 3426.2,
"valid_targets_min": 863
},
{
"epoch": 4.917218543046357,
"grad_norm": 0.3162724666592099,
"learning_rate": 3.2745519558241664e-05,
"loss": 0.038,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0377536416053772,
"step": 1485,
"valid_targets_mean": 2567.1,
"valid_targets_min": 417
},
{
"epoch": 4.933774834437086,
"grad_norm": 0.3186894653624437,
"learning_rate": 3.243912760640054e-05,
"loss": 0.0403,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04340728744864464,
"step": 1490,
"valid_targets_mean": 3049.3,
"valid_targets_min": 514
},
{
"epoch": 4.950331125827814,
"grad_norm": 0.25287546174236397,
"learning_rate": 3.213348602678928e-05,
"loss": 0.0351,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03173156827688217,
"step": 1495,
"valid_targets_mean": 4157.4,
"valid_targets_min": 1014
},
{
"epoch": 4.966887417218543,
"grad_norm": 0.3106237713164603,
"learning_rate": 3.1828607879405674e-05,
"loss": 0.0376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.03871217370033264,
"step": 1500,
"valid_targets_mean": 2624.1,
"valid_targets_min": 336
},
{
"epoch": 4.983443708609272,
"grad_norm": 0.2760613262824866,
"learning_rate": 3.1524506191626116e-05,
"loss": 0.0362,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0380619652569294,
"step": 1505,
"valid_targets_mean": 3306.6,
"valid_targets_min": 813
},
{
"epoch": 5.0,
"grad_norm": 0.24703833134333728,
"learning_rate": 3.122119395764911e-05,
"loss": 0.0362,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0335039347410202,
"step": 1510,
"valid_targets_mean": 3556.3,
"valid_targets_min": 288
},
{
"epoch": 5.016556291390729,
"grad_norm": 0.22730118261563992,
"learning_rate": 3.0918684137939975e-05,
"loss": 0.0181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.019685471430420876,
"step": 1515,
"valid_targets_mean": 3281.3,
"valid_targets_min": 666
},
{
"epoch": 5.033112582781457,
"grad_norm": 0.24352277353000806,
"learning_rate": 3.061698965867701e-05,
"loss": 0.0169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.015633970499038696,
"step": 1520,
"valid_targets_mean": 3149.2,
"valid_targets_min": 624
},
{
"epoch": 5.049668874172186,
"grad_norm": 0.2914984372902733,
"learning_rate": 3.0316123411199228e-05,
"loss": 0.0187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.02213025465607643,
"step": 1525,
"valid_targets_mean": 2653.5,
"valid_targets_min": 682
},
{
"epoch": 5.066225165562914,
"grad_norm": 0.22916458288710354,
"learning_rate": 3.0016098251455442e-05,
"loss": 0.0181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017979608848690987,
"step": 1530,
"valid_targets_mean": 3157.4,
"valid_targets_min": 689
},
{
"epoch": 5.082781456953643,
"grad_norm": 0.22184729474777812,
"learning_rate": 2.971692699945502e-05,
"loss": 0.0185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0169589314609766,
"step": 1535,
"valid_targets_mean": 2940.4,
"valid_targets_min": 1030
},
{
"epoch": 5.099337748344371,
"grad_norm": 0.258694518076171,
"learning_rate": 2.9418622438720017e-05,
"loss": 0.0185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.019805902615189552,
"step": 1540,
"valid_targets_mean": 2852.8,
"valid_targets_min": 680
},
{
"epoch": 5.1158940397351,
"grad_norm": 0.2364864679821581,
"learning_rate": 2.9121197315738917e-05,
"loss": 0.0179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018344629555940628,
"step": 1545,
"valid_targets_mean": 3181.3,
"valid_targets_min": 452
},
{
"epoch": 5.132450331125828,
"grad_norm": 0.2366564946920007,
"learning_rate": 2.8824664339422114e-05,
"loss": 0.018,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.016553200781345367,
"step": 1550,
"valid_targets_mean": 3262.7,
"valid_targets_min": 423
},
{
"epoch": 5.1490066225165565,
"grad_norm": 0.2252293381022784,
"learning_rate": 2.8529036180558665e-05,
"loss": 0.0186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0168086439371109,
"step": 1555,
"valid_targets_mean": 3429.9,
"valid_targets_min": 454
},
{
"epoch": 5.1655629139072845,
"grad_norm": 0.22113286785928268,
"learning_rate": 2.8234325471275058e-05,
"loss": 0.0194,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01686927117407322,
"step": 1560,
"valid_targets_mean": 3416.8,
"valid_targets_min": 287
},
{
"epoch": 5.182119205298013,
"grad_norm": 0.21212610340392146,
"learning_rate": 2.7940544804495344e-05,
"loss": 0.0192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018345486372709274,
"step": 1565,
"valid_targets_mean": 3362.6,
"valid_targets_min": 432
},
{
"epoch": 5.198675496688741,
"grad_norm": 0.2175722329615146,
"learning_rate": 2.7647706733403033e-05,
"loss": 0.0187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.02010449767112732,
"step": 1570,
"valid_targets_mean": 4109.9,
"valid_targets_min": 1048
},
{
"epoch": 5.21523178807947,
"grad_norm": 0.23306063110602354,
"learning_rate": 2.7355823770904736e-05,
"loss": 0.0175,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017950814217329025,
"step": 1575,
"valid_targets_mean": 3748.8,
"valid_targets_min": 322
},
{
"epoch": 5.231788079470198,
"grad_norm": 0.23672388097998792,
"learning_rate": 2.7064908389095468e-05,
"loss": 0.0182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018415657803416252,
"step": 1580,
"valid_targets_mean": 3489.2,
"valid_targets_min": 735
},
{
"epoch": 5.248344370860927,
"grad_norm": 0.23742057100355848,
"learning_rate": 2.6774973018725815e-05,
"loss": 0.018,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01655556634068489,
"step": 1585,
"valid_targets_mean": 2962.5,
"valid_targets_min": 531
},
{
"epoch": 5.264900662251655,
"grad_norm": 0.2549485071123772,
"learning_rate": 2.6486030048670597e-05,
"loss": 0.0195,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018138859421014786,
"step": 1590,
"valid_targets_mean": 2896.8,
"valid_targets_min": 389
},
{
"epoch": 5.281456953642384,
"grad_norm": 0.2353608786308097,
"learning_rate": 2.6198091825399605e-05,
"loss": 0.0183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018565738573670387,
"step": 1595,
"valid_targets_mean": 3267.8,
"valid_targets_min": 791
},
{
"epoch": 5.298013245033113,
"grad_norm": 0.21746537321208256,
"learning_rate": 2.5911170652450067e-05,
"loss": 0.0187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017724383622407913,
"step": 1600,
"valid_targets_mean": 3414.1,
"valid_targets_min": 1092
},
{
"epoch": 5.314569536423841,
"grad_norm": 0.23948446337932847,
"learning_rate": 2.562527878990081e-05,
"loss": 0.0199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018817899748682976,
"step": 1605,
"valid_targets_mean": 3105.8,
"valid_targets_min": 257
},
{
"epoch": 5.33112582781457,
"grad_norm": 0.22886880328166953,
"learning_rate": 2.534042845384851e-05,
"loss": 0.0197,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.019166797399520874,
"step": 1610,
"valid_targets_mean": 3858.0,
"valid_targets_min": 806
},
{
"epoch": 5.347682119205298,
"grad_norm": 0.22910938446198675,
"learning_rate": 2.5056631815885585e-05,
"loss": 0.0182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017277346923947334,
"step": 1615,
"valid_targets_mean": 2916.8,
"valid_targets_min": 420
},
{
"epoch": 5.364238410596027,
"grad_norm": 0.215045884428276,
"learning_rate": 2.4773901002580235e-05,
"loss": 0.0178,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01621703989803791,
"step": 1620,
"valid_targets_mean": 3402.4,
"valid_targets_min": 348
},
{
"epoch": 5.380794701986755,
"grad_norm": 0.26029562494835984,
"learning_rate": 2.4492248094958147e-05,
"loss": 0.0188,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.020647404715418816,
"step": 1625,
"valid_targets_mean": 2778.4,
"valid_targets_min": 559
},
{
"epoch": 5.397350993377484,
"grad_norm": 0.28643082761962496,
"learning_rate": 2.421168512798634e-05,
"loss": 0.0195,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.020385634154081345,
"step": 1630,
"valid_targets_mean": 2362.4,
"valid_targets_min": 762
},
{
"epoch": 5.413907284768212,
"grad_norm": 0.1997655418670906,
"learning_rate": 2.3932224090058935e-05,
"loss": 0.0184,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01593804359436035,
"step": 1635,
"valid_targets_mean": 3627.2,
"valid_targets_min": 723
},
{
"epoch": 5.430463576158941,
"grad_norm": 0.2584843082826377,
"learning_rate": 2.365387692248488e-05,
"loss": 0.0187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.019819604232907295,
"step": 1640,
"valid_targets_mean": 3059.1,
"valid_targets_min": 532
},
{
"epoch": 5.447019867549669,
"grad_norm": 0.23352818903841596,
"learning_rate": 2.3376655518977642e-05,
"loss": 0.0184,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018485667183995247,
"step": 1645,
"valid_targets_mean": 3387.3,
"valid_targets_min": 494
},
{
"epoch": 5.4635761589403975,
"grad_norm": 0.2363955583281061,
"learning_rate": 2.310057172514703e-05,
"loss": 0.0192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017806580290198326,
"step": 1650,
"valid_targets_mean": 2877.9,
"valid_targets_min": 348
},
{
"epoch": 5.4801324503311255,
"grad_norm": 0.24086123457076253,
"learning_rate": 2.2825637337993093e-05,
"loss": 0.0176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017760343849658966,
"step": 1655,
"valid_targets_mean": 3551.2,
"valid_targets_min": 823
},
{
"epoch": 5.496688741721854,
"grad_norm": 0.2514965806315985,
"learning_rate": 2.255186410540197e-05,
"loss": 0.0198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.019075598567724228,
"step": 1660,
"valid_targets_mean": 2926.9,
"valid_targets_min": 769
},
{
"epoch": 5.513245033112582,
"grad_norm": 0.23477018024739887,
"learning_rate": 2.2279263725643874e-05,
"loss": 0.0177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.019219625741243362,
"step": 1665,
"valid_targets_mean": 3019.4,
"valid_targets_min": 981
},
{
"epoch": 5.529801324503311,
"grad_norm": 0.24151585501562628,
"learning_rate": 2.200784784687334e-05,
"loss": 0.0183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018403450027108192,
"step": 1670,
"valid_targets_mean": 2868.7,
"valid_targets_min": 348
},
{
"epoch": 5.54635761589404,
"grad_norm": 0.217747489117389,
"learning_rate": 2.173762806663139e-05,
"loss": 0.0185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.016213811933994293,
"step": 1675,
"valid_targets_mean": 3463.8,
"valid_targets_min": 912
},
{
"epoch": 5.562913907284768,
"grad_norm": 0.2512929589670852,
"learning_rate": 2.146861593135e-05,
"loss": 0.0188,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017997808754444122,
"step": 1680,
"valid_targets_mean": 2735.8,
"valid_targets_min": 577
},
{
"epoch": 5.579470198675496,
"grad_norm": 0.2511463527628348,
"learning_rate": 2.1200822935858804e-05,
"loss": 0.0189,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017626876011490822,
"step": 1685,
"valid_targets_mean": 2815.3,
"valid_targets_min": 849
},
{
"epoch": 5.596026490066225,
"grad_norm": 0.2316035151107184,
"learning_rate": 2.093426052289384e-05,
"loss": 0.0178,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01970108412206173,
"step": 1690,
"valid_targets_mean": 2911.6,
"valid_targets_min": 444
},
{
"epoch": 5.612582781456954,
"grad_norm": 0.22216530929238976,
"learning_rate": 2.066894008260859e-05,
"loss": 0.019,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01790759339928627,
"step": 1695,
"valid_targets_mean": 3455.8,
"valid_targets_min": 602
},
{
"epoch": 5.629139072847682,
"grad_norm": 0.22871616417797874,
"learning_rate": 2.040487295208732e-05,
"loss": 0.0174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017838913947343826,
"step": 1700,
"valid_targets_mean": 3140.1,
"valid_targets_min": 684
},
{
"epoch": 5.645695364238411,
"grad_norm": 0.21448146414148817,
"learning_rate": 2.0142070414860704e-05,
"loss": 0.0176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017433665692806244,
"step": 1705,
"valid_targets_mean": 3913.5,
"valid_targets_min": 1149
},
{
"epoch": 5.662251655629139,
"grad_norm": 0.22850635912200845,
"learning_rate": 1.9880543700423533e-05,
"loss": 0.0178,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01749793067574501,
"step": 1710,
"valid_targets_mean": 3097.8,
"valid_targets_min": 792
},
{
"epoch": 5.678807947019868,
"grad_norm": 0.22682977769005455,
"learning_rate": 1.962030398375506e-05,
"loss": 0.0173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.016696149483323097,
"step": 1715,
"valid_targets_mean": 3026.1,
"valid_targets_min": 346
},
{
"epoch": 5.695364238410596,
"grad_norm": 0.23859152675019352,
"learning_rate": 1.9361362384841325e-05,
"loss": 0.0173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01651284471154213,
"step": 1720,
"valid_targets_mean": 3026.2,
"valid_targets_min": 1106
},
{
"epoch": 5.711920529801325,
"grad_norm": 0.23879084915470672,
"learning_rate": 1.9103729968200147e-05,
"loss": 0.0168,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01774653047323227,
"step": 1725,
"valid_targets_mean": 2829.7,
"valid_targets_min": 543
},
{
"epoch": 5.728476821192053,
"grad_norm": 0.19141628322677554,
"learning_rate": 1.884741774240823e-05,
"loss": 0.0177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.015631206333637238,
"step": 1730,
"valid_targets_mean": 3749.4,
"valid_targets_min": 618
},
{
"epoch": 5.745033112582782,
"grad_norm": 0.21457866617694085,
"learning_rate": 1.8592436659630784e-05,
"loss": 0.0185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.016092777252197266,
"step": 1735,
"valid_targets_mean": 2978.6,
"valid_targets_min": 355
},
{
"epoch": 5.76158940397351,
"grad_norm": 0.23151354938535157,
"learning_rate": 1.8338797615153697e-05,
"loss": 0.0162,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.016310635954141617,
"step": 1740,
"valid_targets_mean": 3089.6,
"valid_targets_min": 611
},
{
"epoch": 5.7781456953642385,
"grad_norm": 0.238729370589449,
"learning_rate": 1.8086511446917714e-05,
"loss": 0.0177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018275607377290726,
"step": 1745,
"valid_targets_mean": 2858.8,
"valid_targets_min": 316
},
{
"epoch": 5.7947019867549665,
"grad_norm": 0.2599873172719797,
"learning_rate": 1.783558893505554e-05,
"loss": 0.0183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.021223030984401703,
"step": 1750,
"valid_targets_mean": 2747.5,
"valid_targets_min": 439
},
{
"epoch": 5.811258278145695,
"grad_norm": 0.22273793371432132,
"learning_rate": 1.7586040801431113e-05,
"loss": 0.0181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.019926520064473152,
"step": 1755,
"valid_targets_mean": 3707.9,
"valid_targets_min": 270
},
{
"epoch": 5.827814569536423,
"grad_norm": 0.22093047468014035,
"learning_rate": 1.7337877709181526e-05,
"loss": 0.0169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.016475237905979156,
"step": 1760,
"valid_targets_mean": 3472.9,
"valid_targets_min": 430
},
{
"epoch": 5.844370860927152,
"grad_norm": 0.2957621995179929,
"learning_rate": 1.7091110262261357e-05,
"loss": 0.0186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.020524853840470314,
"step": 1765,
"valid_targets_mean": 1976.7,
"valid_targets_min": 254
},
{
"epoch": 5.860927152317881,
"grad_norm": 0.22145363280009814,
"learning_rate": 1.684574900498951e-05,
"loss": 0.0169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.016672097146511078,
"step": 1770,
"valid_targets_mean": 3131.5,
"valid_targets_min": 486
},
{
"epoch": 5.877483443708609,
"grad_norm": 0.23266223923508544,
"learning_rate": 1.660180442159879e-05,
"loss": 0.017,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017092088237404823,
"step": 1775,
"valid_targets_mean": 3181.2,
"valid_targets_min": 1343
},
{
"epoch": 5.894039735099338,
"grad_norm": 0.22311044609403474,
"learning_rate": 1.635928693578777e-05,
"loss": 0.0173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017362073063850403,
"step": 1780,
"valid_targets_mean": 3538.1,
"valid_targets_min": 1045
},
{
"epoch": 5.910596026490066,
"grad_norm": 0.24068492138740707,
"learning_rate": 1.611820691027548e-05,
"loss": 0.0169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.017458714544773102,
"step": 1785,
"valid_targets_mean": 2655.7,
"valid_targets_min": 341
},
{
"epoch": 5.927152317880795,
"grad_norm": 0.19647494257878567,
"learning_rate": 1.5878574646358607e-05,
"loss": 0.0152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01394198089838028,
"step": 1790,
"valid_targets_mean": 3739.9,
"valid_targets_min": 281
},
{
"epoch": 5.943708609271523,
"grad_norm": 0.24224754470083248,
"learning_rate": 1.5640400383471294e-05,
"loss": 0.0169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.016955027356743813,
"step": 1795,
"valid_targets_mean": 2705.8,
"valid_targets_min": 413
},
{
"epoch": 5.960264900662252,
"grad_norm": 0.20955855105752547,
"learning_rate": 1.5403694298747602e-05,
"loss": 0.0162,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.015804769471287727,
"step": 1800,
"valid_targets_mean": 3339.2,
"valid_targets_min": 756
},
{
"epoch": 5.97682119205298,
"grad_norm": 0.22573055304516038,
"learning_rate": 1.5168466506586654e-05,
"loss": 0.0178,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018291154876351357,
"step": 1805,
"valid_targets_mean": 3340.3,
"valid_targets_min": 704
},
{
"epoch": 5.993377483443709,
"grad_norm": 0.2341222928786568,
"learning_rate": 1.49347270582205e-05,
"loss": 0.0172,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.018087171018123627,
"step": 1810,
"valid_targets_mean": 3556.8,
"valid_targets_min": 388
},
{
"epoch": 6.009933774834437,
"grad_norm": 0.13229679791376994,
"learning_rate": 1.4702485941284534e-05,
"loss": 0.0107,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00863178726285696,
"step": 1815,
"valid_targets_mean": 3255.4,
"valid_targets_min": 572
},
{
"epoch": 6.026490066225166,
"grad_norm": 0.15539123475586689,
"learning_rate": 1.4471753079390815e-05,
"loss": 0.0078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006660331506282091,
"step": 1820,
"valid_targets_mean": 3072.4,
"valid_targets_min": 364
},
{
"epoch": 6.043046357615894,
"grad_norm": 0.15832598103201723,
"learning_rate": 1.4242538331703969e-05,
"loss": 0.0076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007331097032874823,
"step": 1825,
"valid_targets_mean": 3698.4,
"valid_targets_min": 489
},
{
"epoch": 6.059602649006623,
"grad_norm": 0.17744650731127176,
"learning_rate": 1.401485149251996e-05,
"loss": 0.0079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007538849022239447,
"step": 1830,
"valid_targets_mean": 3227.4,
"valid_targets_min": 867
},
{
"epoch": 6.076158940397351,
"grad_norm": 0.1726740301592844,
"learning_rate": 1.3788702290847517e-05,
"loss": 0.0082,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008650300092995167,
"step": 1835,
"valid_targets_mean": 3441.4,
"valid_targets_min": 861
},
{
"epoch": 6.0927152317880795,
"grad_norm": 0.16979066431789921,
"learning_rate": 1.3564100389992496e-05,
"loss": 0.0079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008822856470942497,
"step": 1840,
"valid_targets_mean": 3106.0,
"valid_targets_min": 719
},
{
"epoch": 6.109271523178808,
"grad_norm": 0.15499865855490422,
"learning_rate": 1.3341055387144923e-05,
"loss": 0.0078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007064931560307741,
"step": 1845,
"valid_targets_mean": 2872.3,
"valid_targets_min": 446
},
{
"epoch": 6.125827814569536,
"grad_norm": 0.1703416545422938,
"learning_rate": 1.3119576812968892e-05,
"loss": 0.0077,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007549528032541275,
"step": 1850,
"valid_targets_mean": 2797.3,
"valid_targets_min": 674
},
{
"epoch": 6.142384105960265,
"grad_norm": 0.1697437858396481,
"learning_rate": 1.289967413119535e-05,
"loss": 0.0082,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007537458557635546,
"step": 1855,
"valid_targets_mean": 3088.2,
"valid_targets_min": 692
},
{
"epoch": 6.158940397350993,
"grad_norm": 0.15740217689907615,
"learning_rate": 1.2681356738217692e-05,
"loss": 0.0075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00699217151850462,
"step": 1860,
"valid_targets_mean": 2920.7,
"valid_targets_min": 651
},
{
"epoch": 6.175496688741722,
"grad_norm": 0.17589779228732438,
"learning_rate": 1.2464633962690303e-05,
"loss": 0.0079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0072303046472370625,
"step": 1865,
"valid_targets_mean": 2953.4,
"valid_targets_min": 304
},
{
"epoch": 6.19205298013245,
"grad_norm": 0.18210559491489264,
"learning_rate": 1.2249515065129918e-05,
"loss": 0.0078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008227629587054253,
"step": 1870,
"valid_targets_mean": 3361.9,
"valid_targets_min": 607
},
{
"epoch": 6.208609271523179,
"grad_norm": 0.1698046233572755,
"learning_rate": 1.2036009237519869e-05,
"loss": 0.0075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0076033310033380985,
"step": 1875,
"valid_targets_mean": 3130.8,
"valid_targets_min": 703
},
{
"epoch": 6.225165562913907,
"grad_norm": 0.17529068128382608,
"learning_rate": 1.1824125602917412e-05,
"loss": 0.0074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008096292614936829,
"step": 1880,
"valid_targets_mean": 2944.0,
"valid_targets_min": 981
},
{
"epoch": 6.241721854304636,
"grad_norm": 0.19205932233382772,
"learning_rate": 1.161387321506383e-05,
"loss": 0.0083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.009108491241931915,
"step": 1885,
"valid_targets_mean": 2701.5,
"valid_targets_min": 1111
},
{
"epoch": 6.258278145695364,
"grad_norm": 0.13824188190877879,
"learning_rate": 1.1405261057997563e-05,
"loss": 0.0072,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006597296334803104,
"step": 1890,
"valid_targets_mean": 3353.2,
"valid_targets_min": 382
},
{
"epoch": 6.274834437086093,
"grad_norm": 0.17419244467055003,
"learning_rate": 1.1198298045670402e-05,
"loss": 0.0079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008129781112074852,
"step": 1895,
"valid_targets_mean": 2993.9,
"valid_targets_min": 700
},
{
"epoch": 6.291390728476821,
"grad_norm": 0.16667513994242347,
"learning_rate": 1.0992993021566528e-05,
"loss": 0.0082,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007786220870912075,
"step": 1900,
"valid_targets_mean": 2825.1,
"valid_targets_min": 707
},
{
"epoch": 6.30794701986755,
"grad_norm": 0.15185724867583905,
"learning_rate": 1.0789354758324621e-05,
"loss": 0.0076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006867009215056896,
"step": 1905,
"valid_targets_mean": 3570.2,
"valid_targets_min": 835
},
{
"epoch": 6.324503311258278,
"grad_norm": 0.17028287276176793,
"learning_rate": 1.0587391957363053e-05,
"loss": 0.008,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008131728507578373,
"step": 1910,
"valid_targets_mean": 3156.1,
"valid_targets_min": 950
},
{
"epoch": 6.341059602649007,
"grad_norm": 0.1704347813207607,
"learning_rate": 1.0387113248508062e-05,
"loss": 0.008,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008300486952066422,
"step": 1915,
"valid_targets_mean": 2814.2,
"valid_targets_min": 405
},
{
"epoch": 6.357615894039735,
"grad_norm": 0.17922969456288812,
"learning_rate": 1.0188527189625013e-05,
"loss": 0.0087,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008780170232057571,
"step": 1920,
"valid_targets_mean": 3106.3,
"valid_targets_min": 1088
},
{
"epoch": 6.374172185430464,
"grad_norm": 0.17009754517034342,
"learning_rate": 9.991642266252672e-06,
"loss": 0.0078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00821577850729227,
"step": 1925,
"valid_targets_mean": 2887.9,
"valid_targets_min": 419
},
{
"epoch": 6.390728476821192,
"grad_norm": 0.16323074864551892,
"learning_rate": 9.796466891240702e-06,
"loss": 0.0078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0070761325769126415,
"step": 1930,
"valid_targets_mean": 2889.1,
"valid_targets_min": 559
},
{
"epoch": 6.4072847682119205,
"grad_norm": 0.15644221550249918,
"learning_rate": 9.603009404390096e-06,
"loss": 0.008,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008021621964871883,
"step": 1935,
"valid_targets_mean": 2841.0,
"valid_targets_min": 598
},
{
"epoch": 6.423841059602649,
"grad_norm": 0.16598367190829177,
"learning_rate": 9.41127807209688e-06,
"loss": 0.0075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008119653910398483,
"step": 1940,
"valid_targets_mean": 2805.6,
"valid_targets_min": 322
},
{
"epoch": 6.440397350993377,
"grad_norm": 0.18393145097087016,
"learning_rate": 9.221281086998878e-06,
"loss": 0.008,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008811881765723228,
"step": 1945,
"valid_targets_mean": 2820.2,
"valid_targets_min": 889
},
{
"epoch": 6.456953642384106,
"grad_norm": 0.13861147326112663,
"learning_rate": 9.033026567625652e-06,
"loss": 0.0072,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006629860028624535,
"step": 1950,
"valid_targets_mean": 3483.8,
"valid_targets_min": 1206
},
{
"epoch": 6.473509933774834,
"grad_norm": 0.14577823804947748,
"learning_rate": 8.846522558051561e-06,
"loss": 0.0083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006631593219935894,
"step": 1955,
"valid_targets_mean": 3314.4,
"valid_targets_min": 708
},
{
"epoch": 6.490066225165563,
"grad_norm": 0.20041476288284338,
"learning_rate": 8.66177702755206e-06,
"loss": 0.0084,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008733078837394714,
"step": 1960,
"valid_targets_mean": 2331.3,
"valid_targets_min": 290
},
{
"epoch": 6.506622516556291,
"grad_norm": 0.16541272094090137,
"learning_rate": 8.478797870263205e-06,
"loss": 0.008,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.009385555982589722,
"step": 1965,
"valid_targets_mean": 2827.2,
"valid_targets_min": 278
},
{
"epoch": 6.52317880794702,
"grad_norm": 0.15698878725456727,
"learning_rate": 8.29759290484428e-06,
"loss": 0.0077,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00760327372699976,
"step": 1970,
"valid_targets_mean": 3569.4,
"valid_targets_min": 373
},
{
"epoch": 6.539735099337748,
"grad_norm": 0.17204858194232356,
"learning_rate": 8.118169874143783e-06,
"loss": 0.0078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008203571662306786,
"step": 1975,
"valid_targets_mean": 3637.4,
"valid_targets_min": 1020
},
{
"epoch": 6.556291390728477,
"grad_norm": 0.15428596015755208,
"learning_rate": 7.94053644486848e-06,
"loss": 0.0072,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008218035101890564,
"step": 1980,
"valid_targets_mean": 3484.9,
"valid_targets_min": 534
},
{
"epoch": 6.572847682119205,
"grad_norm": 0.1616016188453889,
"learning_rate": 7.764700207255903e-06,
"loss": 0.0079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00741251464933157,
"step": 1985,
"valid_targets_mean": 3015.1,
"valid_targets_min": 295
},
{
"epoch": 6.589403973509934,
"grad_norm": 0.14681513537537527,
"learning_rate": 7.590668674749946e-06,
"loss": 0.0078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007753623183816671,
"step": 1990,
"valid_targets_mean": 3305.1,
"valid_targets_min": 884
},
{
"epoch": 6.605960264900662,
"grad_norm": 0.1712467224481615,
"learning_rate": 7.418449283679868e-06,
"loss": 0.0076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00784437358379364,
"step": 1995,
"valid_targets_mean": 2848.3,
"valid_targets_min": 343
},
{
"epoch": 6.622516556291391,
"grad_norm": 0.18351826309232944,
"learning_rate": 7.2480493929425295e-06,
"loss": 0.008,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008334236219525337,
"step": 2000,
"valid_targets_mean": 2850.3,
"valid_targets_min": 587
},
{
"epoch": 6.639072847682119,
"grad_norm": 0.15546663258853366,
"learning_rate": 7.079476283687925e-06,
"loss": 0.0076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006955362390726805,
"step": 2005,
"valid_targets_mean": 2787.5,
"valid_targets_min": 499
},
{
"epoch": 6.655629139072848,
"grad_norm": 0.162553849391359,
"learning_rate": 6.91273715900807e-06,
"loss": 0.0074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007508080918341875,
"step": 2010,
"valid_targets_mean": 2807.0,
"valid_targets_min": 373
},
{
"epoch": 6.672185430463577,
"grad_norm": 0.153580733052077,
"learning_rate": 6.7478391436292195e-06,
"loss": 0.0074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007931060157716274,
"step": 2015,
"valid_targets_mean": 3213.3,
"valid_targets_min": 1152
},
{
"epoch": 6.688741721854305,
"grad_norm": 0.17060332282647817,
"learning_rate": 6.584789283607457e-06,
"loss": 0.0076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007987936958670616,
"step": 2020,
"valid_targets_mean": 3311.6,
"valid_targets_min": 615
},
{
"epoch": 6.7052980132450335,
"grad_norm": 0.19120715814803146,
"learning_rate": 6.423594546027589e-06,
"loss": 0.0074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008832000195980072,
"step": 2025,
"valid_targets_mean": 2429.2,
"valid_targets_min": 284
},
{
"epoch": 6.7218543046357615,
"grad_norm": 0.19343662927403713,
"learning_rate": 6.264261818705419e-06,
"loss": 0.0073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008886952884495258,
"step": 2030,
"valid_targets_mean": 2019.9,
"valid_targets_min": 415
},
{
"epoch": 6.73841059602649,
"grad_norm": 0.18857879508841674,
"learning_rate": 6.1067979098934856e-06,
"loss": 0.0079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00828949548304081,
"step": 2035,
"valid_targets_mean": 3015.3,
"valid_targets_min": 649
},
{
"epoch": 6.754966887417218,
"grad_norm": 0.1447938749916527,
"learning_rate": 5.951209547990111e-06,
"loss": 0.0073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007432145066559315,
"step": 2040,
"valid_targets_mean": 3884.9,
"valid_targets_min": 388
},
{
"epoch": 6.771523178807947,
"grad_norm": 0.14172834465758732,
"learning_rate": 5.797503381251895e-06,
"loss": 0.0077,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006457366980612278,
"step": 2045,
"valid_targets_mean": 3344.8,
"valid_targets_min": 390
},
{
"epoch": 6.788079470198675,
"grad_norm": 0.14749223026339864,
"learning_rate": 5.6456859775096795e-06,
"loss": 0.007,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007586037740111351,
"step": 2050,
"valid_targets_mean": 3450.8,
"valid_targets_min": 619
},
{
"epoch": 6.804635761589404,
"grad_norm": 0.16530330834885876,
"learning_rate": 5.495763823887878e-06,
"loss": 0.0072,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00706671504303813,
"step": 2055,
"valid_targets_mean": 3170.5,
"valid_targets_min": 632
},
{
"epoch": 6.821192052980132,
"grad_norm": 0.1300971919954987,
"learning_rate": 5.347743326527255e-06,
"loss": 0.0071,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006268251687288284,
"step": 2060,
"valid_targets_mean": 3392.6,
"valid_targets_min": 918
},
{
"epoch": 6.837748344370861,
"grad_norm": 0.18960162476599696,
"learning_rate": 5.201630810311214e-06,
"loss": 0.0074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007977061904966831,
"step": 2065,
"valid_targets_mean": 2266.9,
"valid_targets_min": 388
},
{
"epoch": 6.854304635761589,
"grad_norm": 0.17487639851676667,
"learning_rate": 5.057432518595584e-06,
"loss": 0.0077,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00837306771427393,
"step": 2070,
"valid_targets_mean": 2775.6,
"valid_targets_min": 383
},
{
"epoch": 6.870860927152318,
"grad_norm": 0.16074739777183703,
"learning_rate": 4.9151546129417804e-06,
"loss": 0.0072,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0074305785819888115,
"step": 2075,
"valid_targets_mean": 3030.6,
"valid_targets_min": 1008
},
{
"epoch": 6.887417218543046,
"grad_norm": 0.14508618603276802,
"learning_rate": 4.774803172853526e-06,
"loss": 0.0069,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006585285998880863,
"step": 2080,
"valid_targets_mean": 3171.8,
"valid_targets_min": 355
},
{
"epoch": 6.903973509933775,
"grad_norm": 0.14413225453441192,
"learning_rate": 4.6363841955171015e-06,
"loss": 0.0068,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.006275333929806948,
"step": 2085,
"valid_targets_mean": 3006.4,
"valid_targets_min": 503
},
{
"epoch": 6.920529801324504,
"grad_norm": 0.15618331040692643,
"learning_rate": 4.4999035955450965e-06,
"loss": 0.007,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008417507633566856,
"step": 2090,
"valid_targets_mean": 3133.2,
"valid_targets_min": 577
},
{
"epoch": 6.937086092715232,
"grad_norm": 0.1720181738017788,
"learning_rate": 4.365367204723636e-06,
"loss": 0.0067,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007837580516934395,
"step": 2095,
"valid_targets_mean": 2914.9,
"valid_targets_min": 852
},
{
"epoch": 6.95364238410596,
"grad_norm": 0.1783704715121628,
"learning_rate": 4.2327807717632175e-06,
"loss": 0.0071,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.008706760592758656,
"step": 2100,
"valid_targets_mean": 2637.3,
"valid_targets_min": 464
},
{
"epoch": 6.970198675496689,
"grad_norm": 0.1654582247696629,
"learning_rate": 4.102149962053098e-06,
"loss": 0.0074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00754096545279026,
"step": 2105,
"valid_targets_mean": 2690.4,
"valid_targets_min": 743
},
{
"epoch": 6.986754966887418,
"grad_norm": 0.15223737269399398,
"learning_rate": 3.973480357419135e-06,
"loss": 0.0066,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.007146907038986683,
"step": 2110,
"valid_targets_mean": 3258.3,
"valid_targets_min": 953
},
{
"epoch": 7.003311258278146,
"grad_norm": 0.08128187572886887,
"learning_rate": 3.846777455885348e-06,
"loss": 0.0058,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003688330762088299,
"step": 2115,
"valid_targets_mean": 3406.6,
"valid_targets_min": 782
},
{
"epoch": 7.0198675496688745,
"grad_norm": 0.09009236582405541,
"learning_rate": 3.72204667143895e-06,
"loss": 0.0036,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0035016811452805996,
"step": 2120,
"valid_targets_mean": 2913.0,
"valid_targets_min": 556
},
{
"epoch": 7.0364238410596025,
"grad_norm": 0.0851022230759573,
"learning_rate": 3.599293333799042e-06,
"loss": 0.0031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003222863655537367,
"step": 2125,
"valid_targets_mean": 3323.1,
"valid_targets_min": 417
},
{
"epoch": 7.052980132450331,
"grad_norm": 0.11953730654430537,
"learning_rate": 3.4785226881888565e-06,
"loss": 0.0035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.004369686357676983,
"step": 2130,
"valid_targets_mean": 2389.0,
"valid_targets_min": 300
},
{
"epoch": 7.069536423841059,
"grad_norm": 0.0865913074699349,
"learning_rate": 3.3597398951116023e-06,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0032925319392234087,
"step": 2135,
"valid_targets_mean": 3149.0,
"valid_targets_min": 896
},
{
"epoch": 7.086092715231788,
"grad_norm": 0.10422960589397072,
"learning_rate": 3.2429500301300207e-06,
"loss": 0.0035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00322504760697484,
"step": 2140,
"valid_targets_mean": 2631.8,
"valid_targets_min": 780
},
{
"epoch": 7.102649006622516,
"grad_norm": 0.09907891778313167,
"learning_rate": 3.128158083649435e-06,
"loss": 0.0036,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003162794280797243,
"step": 2145,
"valid_targets_mean": 3677.4,
"valid_targets_min": 769
},
{
"epoch": 7.119205298013245,
"grad_norm": 0.0878263506832752,
"learning_rate": 3.0153689607045845e-06,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0030086166225373745,
"step": 2150,
"valid_targets_mean": 3633.9,
"valid_targets_min": 716
},
{
"epoch": 7.135761589403973,
"grad_norm": 0.09115363415822987,
"learning_rate": 2.904587480749965e-06,
"loss": 0.0035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0030223748181015253,
"step": 2155,
"valid_targets_mean": 3305.4,
"valid_targets_min": 577
},
{
"epoch": 7.152317880794702,
"grad_norm": 0.10049634929373007,
"learning_rate": 2.795818377453968e-06,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003773024072870612,
"step": 2160,
"valid_targets_mean": 2846.4,
"valid_targets_min": 419
},
{
"epoch": 7.16887417218543,
"grad_norm": 0.09295466192159862,
"learning_rate": 2.6890662984965232e-06,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0032892152667045593,
"step": 2165,
"valid_targets_mean": 3238.7,
"valid_targets_min": 254
},
{
"epoch": 7.185430463576159,
"grad_norm": 0.10700211443995879,
"learning_rate": 2.5843358053705855e-06,
"loss": 0.0038,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003560242708772421,
"step": 2170,
"valid_targets_mean": 2918.7,
"valid_targets_min": 486
},
{
"epoch": 7.201986754966887,
"grad_norm": 0.13131171797026508,
"learning_rate": 2.481631373187171e-06,
"loss": 0.0037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00407462939620018,
"step": 2175,
"valid_targets_mean": 2111.6,
"valid_targets_min": 499
},
{
"epoch": 7.218543046357616,
"grad_norm": 0.10308336571764598,
"learning_rate": 2.3809573904841843e-06,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0035824584774672985,
"step": 2180,
"valid_targets_mean": 3713.8,
"valid_targets_min": 920
},
{
"epoch": 7.235099337748345,
"grad_norm": 0.08657031530805787,
"learning_rate": 2.282318159038838e-06,
"loss": 0.0031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.002839879598468542,
"step": 2185,
"valid_targets_mean": 3494.3,
"valid_targets_min": 768
},
{
"epoch": 7.251655629139073,
"grad_norm": 0.09344464755413119,
"learning_rate": 2.1857178936838695e-06,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0030079518910497427,
"step": 2190,
"valid_targets_mean": 3014.6,
"valid_targets_min": 304
},
{
"epoch": 7.268211920529802,
"grad_norm": 0.10484727324190744,
"learning_rate": 2.091160722127472e-06,
"loss": 0.0037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00350088719278574,
"step": 2195,
"valid_targets_mean": 3021.6,
"valid_targets_min": 366
},
{
"epoch": 7.28476821192053,
"grad_norm": 0.09818715383735736,
"learning_rate": 1.9986506847768595e-06,
"loss": 0.0035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0031390986405313015,
"step": 2200,
"valid_targets_mean": 2969.8,
"valid_targets_min": 321
},
{
"epoch": 7.301324503311259,
"grad_norm": 0.10139056224803883,
"learning_rate": 1.908191734565651e-06,
"loss": 0.0037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003458566265180707,
"step": 2205,
"valid_targets_mean": 2844.0,
"valid_targets_min": 364
},
{
"epoch": 7.317880794701987,
"grad_norm": 0.09585169261666918,
"learning_rate": 1.8197877367849947e-06,
"loss": 0.003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0032619377598166466,
"step": 2210,
"valid_targets_mean": 2803.5,
"valid_targets_min": 626
},
{
"epoch": 7.3344370860927155,
"grad_norm": 0.07902836181860146,
"learning_rate": 1.7334424689183282e-06,
"loss": 0.0035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.002633770927786827,
"step": 2215,
"valid_targets_mean": 3056.4,
"valid_targets_min": 760
},
{
"epoch": 7.3509933774834435,
"grad_norm": 0.10759390801373159,
"learning_rate": 1.6491596204800342e-06,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0036994030233472586,
"step": 2220,
"valid_targets_mean": 2629.9,
"valid_targets_min": 325
},
{
"epoch": 7.367549668874172,
"grad_norm": 0.10089653202938675,
"learning_rate": 1.5669427928577452e-06,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0032046681735664606,
"step": 2225,
"valid_targets_mean": 2745.7,
"valid_targets_min": 346
},
{
"epoch": 7.3841059602649,
"grad_norm": 0.07487394480939905,
"learning_rate": 1.486795499158522e-06,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00264303432777524,
"step": 2230,
"valid_targets_mean": 3534.2,
"valid_targets_min": 800
},
{
"epoch": 7.400662251655629,
"grad_norm": 0.09883115831122584,
"learning_rate": 1.408721164058646e-06,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0040969038382172585,
"step": 2235,
"valid_targets_mean": 3726.2,
"valid_targets_min": 1257
},
{
"epoch": 7.417218543046357,
"grad_norm": 0.1042504124419704,
"learning_rate": 1.3327231236573478e-06,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.004593914374709129,
"step": 2240,
"valid_targets_mean": 3748.6,
"valid_targets_min": 611
},
{
"epoch": 7.433774834437086,
"grad_norm": 0.0892579190482292,
"learning_rate": 1.258804625334259e-06,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.004084902815520763,
"step": 2245,
"valid_targets_mean": 3592.8,
"valid_targets_min": 577
},
{
"epoch": 7.450331125827814,
"grad_norm": 0.09204143659719179,
"learning_rate": 1.1869688276106016e-06,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0031076439190655947,
"step": 2250,
"valid_targets_mean": 3252.0,
"valid_targets_min": 595
},
{
"epoch": 7.466887417218543,
"grad_norm": 0.09192476694452797,
"learning_rate": 1.1172188000142802e-06,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0029827728867530823,
"step": 2255,
"valid_targets_mean": 3090.6,
"valid_targets_min": 759
},
{
"epoch": 7.483443708609272,
"grad_norm": 0.11844295275347291,
"learning_rate": 1.0495575229486908e-06,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003052099607884884,
"step": 2260,
"valid_targets_mean": 2804.4,
"valid_targets_min": 669
},
{
"epoch": 7.5,
"grad_norm": 0.09338515630700445,
"learning_rate": 9.839878875653852e-07,
"loss": 0.003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.002875866135582328,
"step": 2265,
"valid_targets_mean": 3126.4,
"valid_targets_min": 781
},
{
"epoch": 7.516556291390728,
"grad_norm": 0.09957575628188654,
"learning_rate": 9.205126956405075e-07,
"loss": 0.0031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0031447166111320257,
"step": 2270,
"valid_targets_mean": 3081.4,
"valid_targets_min": 572
},
{
"epoch": 7.533112582781457,
"grad_norm": 0.09838701800490543,
"learning_rate": 8.591346594551019e-07,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0033545675687491894,
"step": 2275,
"valid_targets_mean": 3185.1,
"valid_targets_min": 288
},
{
"epoch": 7.549668874172186,
"grad_norm": 0.08434326160507456,
"learning_rate": 7.998564016792109e-07,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.002952792914584279,
"step": 2280,
"valid_targets_mean": 3549.2,
"valid_targets_min": 316
},
{
"epoch": 7.566225165562914,
"grad_norm": 0.09210381252686883,
"learning_rate": 7.426804552598088e-07,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.00304861506447196,
"step": 2285,
"valid_targets_mean": 3414.8,
"valid_targets_min": 1159
},
{
"epoch": 7.582781456953643,
"grad_norm": 0.09875567223953083,
"learning_rate": 6.876092633125442e-07,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0034329225309193134,
"step": 2290,
"valid_targets_mean": 3022.8,
"valid_targets_min": 1288
},
{
"epoch": 7.599337748344371,
"grad_norm": 0.08062460264401002,
"learning_rate": 6.346451790173957e-07,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0025075674057006836,
"step": 2295,
"valid_targets_mean": 3623.2,
"valid_targets_min": 804
},
{
"epoch": 7.6158940397351,
"grad_norm": 0.11274456189356268,
"learning_rate": 5.837904655180748e-07,
"loss": 0.0034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0035490221343934536,
"step": 2300,
"valid_targets_mean": 2890.7,
"valid_targets_min": 1088
},
{
"epoch": 7.632450331125828,
"grad_norm": 0.08519509405552562,
"learning_rate": 5.350472958253416e-07,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0033322363160550594,
"step": 2305,
"valid_targets_mean": 3351.2,
"valid_targets_min": 586
},
{
"epoch": 7.6490066225165565,
"grad_norm": 0.08988934531200792,
"learning_rate": 4.884177527241685e-07,
"loss": 0.003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003177615813910961,
"step": 2310,
"valid_targets_mean": 3015.3,
"valid_targets_min": 681
},
{
"epoch": 7.6655629139072845,
"grad_norm": 0.09090553633828964,
"learning_rate": 4.4390382868471635e-07,
"loss": 0.0031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0028612101450562477,
"step": 2315,
"valid_targets_mean": 3220.5,
"valid_targets_min": 935
},
{
"epoch": 7.682119205298013,
"grad_norm": 0.0954091623079664,
"learning_rate": 4.0150742577720337e-07,
"loss": 0.0031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0036504012532532215,
"step": 2320,
"valid_targets_mean": 3196.7,
"valid_targets_min": 337
},
{
"epoch": 7.698675496688741,
"grad_norm": 0.10115796151070462,
"learning_rate": 3.6123035559063044e-07,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003108232282102108,
"step": 2325,
"valid_targets_mean": 3396.1,
"valid_targets_min": 708
},
{
"epoch": 7.71523178807947,
"grad_norm": 0.1253424543081462,
"learning_rate": 3.230743391553881e-07,
"loss": 0.0035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003925042692571878,
"step": 2330,
"valid_targets_mean": 2870.8,
"valid_targets_min": 658
},
{
"epoch": 7.731788079470198,
"grad_norm": 0.08160777605246404,
"learning_rate": 2.870410068697038e-07,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0029902588576078415,
"step": 2335,
"valid_targets_mean": 3878.4,
"valid_targets_min": 980
},
{
"epoch": 7.748344370860927,
"grad_norm": 0.0842341192661009,
"learning_rate": 2.5313189842996466e-07,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0028664730489253998,
"step": 2340,
"valid_targets_mean": 3451.0,
"valid_targets_min": 786
},
{
"epoch": 7.764900662251655,
"grad_norm": 0.09047877313490682,
"learning_rate": 2.2134846276494202e-07,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0031933397985994816,
"step": 2345,
"valid_targets_mean": 2653.2,
"valid_targets_min": 619
},
{
"epoch": 7.781456953642384,
"grad_norm": 0.10761805615611171,
"learning_rate": 1.9169205797388568e-07,
"loss": 0.0037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0038718143478035927,
"step": 2350,
"valid_targets_mean": 2614.0,
"valid_targets_min": 774
},
{
"epoch": 7.798013245033113,
"grad_norm": 0.09833482709882478,
"learning_rate": 1.6416395126847005e-07,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0033078936394304037,
"step": 2355,
"valid_targets_mean": 2854.1,
"valid_targets_min": 385
},
{
"epoch": 7.814569536423841,
"grad_norm": 0.09233759067686517,
"learning_rate": 1.3876531891867105e-07,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003295790869742632,
"step": 2360,
"valid_targets_mean": 3389.0,
"valid_targets_min": 697
},
{
"epoch": 7.83112582781457,
"grad_norm": 0.10654260534095188,
"learning_rate": 1.1549724620247282e-07,
"loss": 0.0031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0033516387920826674,
"step": 2365,
"valid_targets_mean": 2905.3,
"valid_targets_min": 373
},
{
"epoch": 7.847682119205298,
"grad_norm": 0.08260614976749385,
"learning_rate": 9.436072735953261e-08,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.002684560138732195,
"step": 2370,
"valid_targets_mean": 3532.7,
"valid_targets_min": 682
},
{
"epoch": 7.864238410596027,
"grad_norm": 0.1150149750499018,
"learning_rate": 7.535666554866483e-08,
"loss": 0.0035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0036380125675350428,
"step": 2375,
"valid_targets_mean": 2499.6,
"valid_targets_min": 379
},
{
"epoch": 7.880794701986755,
"grad_norm": 0.07875445250051372,
"learning_rate": 5.848587280927187e-08,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0027318689972162247,
"step": 2380,
"valid_targets_mean": 3901.9,
"valid_targets_min": 516
},
{
"epoch": 7.897350993377484,
"grad_norm": 0.10759875899535781,
"learning_rate": 4.3749070026627426e-08,
"loss": 0.0035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0036383490078151226,
"step": 2385,
"valid_targets_mean": 3146.1,
"valid_targets_min": 911
},
{
"epoch": 7.913907284768212,
"grad_norm": 0.07262667473159329,
"learning_rate": 3.1146886901090025e-08,
"loss": 0.0031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0026026968844234943,
"step": 2390,
"valid_targets_mean": 3588.0,
"valid_targets_min": 1220
},
{
"epoch": 7.930463576158941,
"grad_norm": 0.09815063878272579,
"learning_rate": 2.0679861921174548e-08,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0035322587937116623,
"step": 2395,
"valid_targets_mean": 3375.3,
"valid_targets_min": 362
},
{
"epoch": 7.947019867549669,
"grad_norm": 0.12163070968554349,
"learning_rate": 1.2348442340576195e-08,
"loss": 0.0035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.004631766118109226,
"step": 2400,
"valid_targets_mean": 2323.2,
"valid_targets_min": 633
},
{
"epoch": 7.9635761589403975,
"grad_norm": 0.0798070406036093,
"learning_rate": 6.152984159024655e-09,
"loss": 0.0032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.002762910909950733,
"step": 2405,
"valid_targets_mean": 3653.4,
"valid_targets_min": 602
},
{
"epoch": 7.9801324503311255,
"grad_norm": 0.12592119011618041,
"learning_rate": 2.0937521070851783e-09,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.003920475486665964,
"step": 2410,
"valid_targets_mean": 2392.9,
"valid_targets_min": 809
},
{
"epoch": 7.996688741721854,
"grad_norm": 0.08386812464212423,
"learning_rate": 1.7091963485649232e-10,
"loss": 0.0033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0034379262942820787,
"step": 2415,
"valid_targets_mean": 3860.8,
"valid_targets_min": 1169
},
{
"epoch": 8.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.002475286368280649,
"step": 2416,
"total_flos": 1139460138729472.0,
"train_loss": 0.08529083973322209,
"train_runtime": 37249.2574,
"train_samples_per_second": 2.073,
"train_steps_per_second": 0.065,
"valid_targets_mean": 2886.6,
"valid_targets_min": 328
}
],
"logging_steps": 5,
"max_steps": 2416,
"num_input_tokens_seen": 0,
"num_train_epochs": 8,
"save_steps": 100,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 1139460138729472.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}