Files
a1-stack_pytest_synthetic_g…/trainer_state.json
ModelHub XC abfc932b6b 初始化项目,由ModelHub XC社区提供模型
Model: DCAgent/a1-stack_pytest_synthetic_gpt5nano
Source: Original Platform
2026-05-13 11:25:29 +08:00

9651 lines
267 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 7.0,
"eval_steps": 500,
"global_step": 4368,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.008012820512820512,
"grad_norm": 18.729351054168752,
"learning_rate": 3.661327231121282e-07,
"loss": 0.6906,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7099630832672119,
"step": 5,
"valid_targets_mean": 5221.4,
"valid_targets_min": 629
},
{
"epoch": 0.016025641025641024,
"grad_norm": 17.547889588299615,
"learning_rate": 8.237986270022884e-07,
"loss": 0.6581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6654312610626221,
"step": 10,
"valid_targets_mean": 4460.4,
"valid_targets_min": 206
},
{
"epoch": 0.02403846153846154,
"grad_norm": 15.746202204380591,
"learning_rate": 1.2814645308924487e-06,
"loss": 0.6562,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6604777574539185,
"step": 15,
"valid_targets_mean": 5242.2,
"valid_targets_min": 1151
},
{
"epoch": 0.03205128205128205,
"grad_norm": 12.19240216279086,
"learning_rate": 1.7391304347826088e-06,
"loss": 0.6208,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6158360242843628,
"step": 20,
"valid_targets_mean": 5037.7,
"valid_targets_min": 595
},
{
"epoch": 0.04006410256410257,
"grad_norm": 6.686738986747432,
"learning_rate": 2.196796338672769e-06,
"loss": 0.5835,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5324317216873169,
"step": 25,
"valid_targets_mean": 4203.4,
"valid_targets_min": 291
},
{
"epoch": 0.04807692307692308,
"grad_norm": 4.140434130583081,
"learning_rate": 2.654462242562929e-06,
"loss": 0.5198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.48007187247276306,
"step": 30,
"valid_targets_mean": 4840.4,
"valid_targets_min": 973
},
{
"epoch": 0.05608974358974359,
"grad_norm": 2.9983399837391875,
"learning_rate": 3.1121281464530894e-06,
"loss": 0.4815,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.511635959148407,
"step": 35,
"valid_targets_mean": 3847.5,
"valid_targets_min": 600
},
{
"epoch": 0.0641025641025641,
"grad_norm": 1.5763363187190589,
"learning_rate": 3.56979405034325e-06,
"loss": 0.4481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4767676591873169,
"step": 40,
"valid_targets_mean": 5336.1,
"valid_targets_min": 1003
},
{
"epoch": 0.07211538461538461,
"grad_norm": 1.120301748537813,
"learning_rate": 4.0274599542334094e-06,
"loss": 0.4179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3749237060546875,
"step": 45,
"valid_targets_mean": 4185.1,
"valid_targets_min": 366
},
{
"epoch": 0.08012820512820513,
"grad_norm": 1.184336227704496,
"learning_rate": 4.48512585812357e-06,
"loss": 0.4285,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45978474617004395,
"step": 50,
"valid_targets_mean": 3175.8,
"valid_targets_min": 481
},
{
"epoch": 0.08814102564102565,
"grad_norm": 1.0750884426051721,
"learning_rate": 4.94279176201373e-06,
"loss": 0.4158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4851878583431244,
"step": 55,
"valid_targets_mean": 3279.4,
"valid_targets_min": 571
},
{
"epoch": 0.09615384615384616,
"grad_norm": 0.6543114948648834,
"learning_rate": 5.400457665903891e-06,
"loss": 0.3947,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3634079694747925,
"step": 60,
"valid_targets_mean": 4396.4,
"valid_targets_min": 314
},
{
"epoch": 0.10416666666666667,
"grad_norm": 0.7092849393150678,
"learning_rate": 5.858123569794051e-06,
"loss": 0.386,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38305991888046265,
"step": 65,
"valid_targets_mean": 4833.6,
"valid_targets_min": 577
},
{
"epoch": 0.11217948717948718,
"grad_norm": 0.7029561167078189,
"learning_rate": 6.31578947368421e-06,
"loss": 0.3719,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3460344076156616,
"step": 70,
"valid_targets_mean": 4298.4,
"valid_targets_min": 262
},
{
"epoch": 0.1201923076923077,
"grad_norm": 0.6651199886852767,
"learning_rate": 6.773455377574372e-06,
"loss": 0.3834,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3872331976890564,
"step": 75,
"valid_targets_mean": 3702.1,
"valid_targets_min": 323
},
{
"epoch": 0.1282051282051282,
"grad_norm": 0.5337310983788888,
"learning_rate": 7.231121281464531e-06,
"loss": 0.3602,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33362239599227905,
"step": 80,
"valid_targets_mean": 4769.6,
"valid_targets_min": 782
},
{
"epoch": 0.1362179487179487,
"grad_norm": 0.5291303347196413,
"learning_rate": 7.688787185354691e-06,
"loss": 0.3516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32650434970855713,
"step": 85,
"valid_targets_mean": 4916.9,
"valid_targets_min": 1924
},
{
"epoch": 0.14423076923076922,
"grad_norm": 0.5305118703534962,
"learning_rate": 8.146453089244852e-06,
"loss": 0.3338,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31834742426872253,
"step": 90,
"valid_targets_mean": 3782.9,
"valid_targets_min": 270
},
{
"epoch": 0.15224358974358973,
"grad_norm": 0.5781472009681354,
"learning_rate": 8.604118993135013e-06,
"loss": 0.3375,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3441351652145386,
"step": 95,
"valid_targets_mean": 3797.4,
"valid_targets_min": 270
},
{
"epoch": 0.16025641025641027,
"grad_norm": 0.5311825618685695,
"learning_rate": 9.061784897025172e-06,
"loss": 0.3222,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3036857545375824,
"step": 100,
"valid_targets_mean": 4236.0,
"valid_targets_min": 314
},
{
"epoch": 0.16826923076923078,
"grad_norm": 0.618750928746387,
"learning_rate": 9.519450800915333e-06,
"loss": 0.3295,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35067400336265564,
"step": 105,
"valid_targets_mean": 3614.1,
"valid_targets_min": 655
},
{
"epoch": 0.1762820512820513,
"grad_norm": 0.6218200582610693,
"learning_rate": 9.977116704805492e-06,
"loss": 0.3353,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33384716510772705,
"step": 110,
"valid_targets_mean": 3519.3,
"valid_targets_min": 420
},
{
"epoch": 0.1842948717948718,
"grad_norm": 0.5600251744508162,
"learning_rate": 1.0434782608695653e-05,
"loss": 0.3141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3137184977531433,
"step": 115,
"valid_targets_mean": 3834.7,
"valid_targets_min": 300
},
{
"epoch": 0.19230769230769232,
"grad_norm": 0.5633355235695061,
"learning_rate": 1.0892448512585814e-05,
"loss": 0.3276,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30305761098861694,
"step": 120,
"valid_targets_mean": 4371.9,
"valid_targets_min": 326
},
{
"epoch": 0.20032051282051283,
"grad_norm": 0.480113957158612,
"learning_rate": 1.1350114416475973e-05,
"loss": 0.3018,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31335359811782837,
"step": 125,
"valid_targets_mean": 5147.1,
"valid_targets_min": 515
},
{
"epoch": 0.20833333333333334,
"grad_norm": 0.5612560066947075,
"learning_rate": 1.1807780320366134e-05,
"loss": 0.2969,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3228274881839752,
"step": 130,
"valid_targets_mean": 5097.3,
"valid_targets_min": 807
},
{
"epoch": 0.21634615384615385,
"grad_norm": 0.5451540551560623,
"learning_rate": 1.2265446224256295e-05,
"loss": 0.2959,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2809686064720154,
"step": 135,
"valid_targets_mean": 4401.4,
"valid_targets_min": 1453
},
{
"epoch": 0.22435897435897437,
"grad_norm": 0.5264986458352455,
"learning_rate": 1.2723112128146454e-05,
"loss": 0.2956,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27167803049087524,
"step": 140,
"valid_targets_mean": 4431.8,
"valid_targets_min": 352
},
{
"epoch": 0.23237179487179488,
"grad_norm": 0.6848874138863399,
"learning_rate": 1.3180778032036615e-05,
"loss": 0.3094,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3045186996459961,
"step": 145,
"valid_targets_mean": 4910.6,
"valid_targets_min": 2777
},
{
"epoch": 0.2403846153846154,
"grad_norm": 0.5237843591461018,
"learning_rate": 1.3638443935926776e-05,
"loss": 0.3108,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29181188344955444,
"step": 150,
"valid_targets_mean": 5049.2,
"valid_targets_min": 271
},
{
"epoch": 0.2483974358974359,
"grad_norm": 0.5579434487220856,
"learning_rate": 1.4096109839816933e-05,
"loss": 0.3163,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29046428203582764,
"step": 155,
"valid_targets_mean": 4435.3,
"valid_targets_min": 1197
},
{
"epoch": 0.2564102564102564,
"grad_norm": 0.6032280986239427,
"learning_rate": 1.4553775743707096e-05,
"loss": 0.2983,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2903742492198944,
"step": 160,
"valid_targets_mean": 4507.2,
"valid_targets_min": 254
},
{
"epoch": 0.2644230769230769,
"grad_norm": 0.560762410271349,
"learning_rate": 1.5011441647597256e-05,
"loss": 0.3118,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32901275157928467,
"step": 165,
"valid_targets_mean": 4771.7,
"valid_targets_min": 641
},
{
"epoch": 0.2724358974358974,
"grad_norm": 0.5877937995218326,
"learning_rate": 1.5469107551487414e-05,
"loss": 0.2864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30902671813964844,
"step": 170,
"valid_targets_mean": 4174.0,
"valid_targets_min": 329
},
{
"epoch": 0.28044871794871795,
"grad_norm": 0.5318864709895957,
"learning_rate": 1.5926773455377575e-05,
"loss": 0.2789,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29658055305480957,
"step": 175,
"valid_targets_mean": 4351.1,
"valid_targets_min": 402
},
{
"epoch": 0.28846153846153844,
"grad_norm": 0.6071948506565317,
"learning_rate": 1.6384439359267736e-05,
"loss": 0.3129,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31895726919174194,
"step": 180,
"valid_targets_mean": 4309.3,
"valid_targets_min": 639
},
{
"epoch": 0.296474358974359,
"grad_norm": 0.5872471984259515,
"learning_rate": 1.6842105263157896e-05,
"loss": 0.2866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3174738585948944,
"step": 185,
"valid_targets_mean": 4299.6,
"valid_targets_min": 324
},
{
"epoch": 0.30448717948717946,
"grad_norm": 0.5559030415647039,
"learning_rate": 1.7299771167048057e-05,
"loss": 0.2988,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3075188100337982,
"step": 190,
"valid_targets_mean": 5057.0,
"valid_targets_min": 722
},
{
"epoch": 0.3125,
"grad_norm": 0.6272873152114179,
"learning_rate": 1.7757437070938218e-05,
"loss": 0.3034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3085539937019348,
"step": 195,
"valid_targets_mean": 4363.2,
"valid_targets_min": 1052
},
{
"epoch": 0.32051282051282054,
"grad_norm": 0.5888783998685431,
"learning_rate": 1.8215102974828376e-05,
"loss": 0.2775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2302914559841156,
"step": 200,
"valid_targets_mean": 4366.6,
"valid_targets_min": 1499
},
{
"epoch": 0.328525641025641,
"grad_norm": 0.5289587072572869,
"learning_rate": 1.8672768878718537e-05,
"loss": 0.2818,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25951194763183594,
"step": 205,
"valid_targets_mean": 4728.9,
"valid_targets_min": 241
},
{
"epoch": 0.33653846153846156,
"grad_norm": 0.5223715132881228,
"learning_rate": 1.9130434782608697e-05,
"loss": 0.2826,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3038901686668396,
"step": 210,
"valid_targets_mean": 5464.7,
"valid_targets_min": 2189
},
{
"epoch": 0.34455128205128205,
"grad_norm": 0.4901694809399787,
"learning_rate": 1.9588100686498858e-05,
"loss": 0.2854,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27634531259536743,
"step": 215,
"valid_targets_mean": 5030.6,
"valid_targets_min": 545
},
{
"epoch": 0.3525641025641026,
"grad_norm": 0.5453590954091112,
"learning_rate": 2.004576659038902e-05,
"loss": 0.2663,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25537845492362976,
"step": 220,
"valid_targets_mean": 4569.1,
"valid_targets_min": 392
},
{
"epoch": 0.3605769230769231,
"grad_norm": 0.5149470056447103,
"learning_rate": 2.050343249427918e-05,
"loss": 0.273,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29232102632522583,
"step": 225,
"valid_targets_mean": 5288.8,
"valid_targets_min": 971
},
{
"epoch": 0.3685897435897436,
"grad_norm": 0.5386671541393013,
"learning_rate": 2.0961098398169337e-05,
"loss": 0.263,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2573094069957733,
"step": 230,
"valid_targets_mean": 4266.0,
"valid_targets_min": 420
},
{
"epoch": 0.3766025641025641,
"grad_norm": 0.608341571305189,
"learning_rate": 2.14187643020595e-05,
"loss": 0.2702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.230138897895813,
"step": 235,
"valid_targets_mean": 5383.9,
"valid_targets_min": 554
},
{
"epoch": 0.38461538461538464,
"grad_norm": 0.5830602000597231,
"learning_rate": 2.187643020594966e-05,
"loss": 0.2884,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24141892790794373,
"step": 240,
"valid_targets_mean": 3893.4,
"valid_targets_min": 583
},
{
"epoch": 0.3926282051282051,
"grad_norm": 0.5624103345818707,
"learning_rate": 2.2334096109839817e-05,
"loss": 0.2713,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2784351706504822,
"step": 245,
"valid_targets_mean": 4039.4,
"valid_targets_min": 393
},
{
"epoch": 0.40064102564102566,
"grad_norm": 0.551812042341973,
"learning_rate": 2.279176201372998e-05,
"loss": 0.2731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2675885856151581,
"step": 250,
"valid_targets_mean": 4489.1,
"valid_targets_min": 315
},
{
"epoch": 0.40865384615384615,
"grad_norm": 0.4881981004835142,
"learning_rate": 2.3249427917620138e-05,
"loss": 0.269,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2349930703639984,
"step": 255,
"valid_targets_mean": 4811.7,
"valid_targets_min": 1026
},
{
"epoch": 0.4166666666666667,
"grad_norm": 0.540673404228557,
"learning_rate": 2.37070938215103e-05,
"loss": 0.2768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2768135368824005,
"step": 260,
"valid_targets_mean": 4714.4,
"valid_targets_min": 1014
},
{
"epoch": 0.42467948717948717,
"grad_norm": 0.7966716027921747,
"learning_rate": 2.4164759725400463e-05,
"loss": 0.2738,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.284927636384964,
"step": 265,
"valid_targets_mean": 3597.9,
"valid_targets_min": 228
},
{
"epoch": 0.4326923076923077,
"grad_norm": 0.6089687245644837,
"learning_rate": 2.462242562929062e-05,
"loss": 0.2673,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29529809951782227,
"step": 270,
"valid_targets_mean": 3442.1,
"valid_targets_min": 468
},
{
"epoch": 0.4407051282051282,
"grad_norm": 0.6222977093678537,
"learning_rate": 2.508009153318078e-05,
"loss": 0.2974,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3036815822124481,
"step": 275,
"valid_targets_mean": 3894.2,
"valid_targets_min": 304
},
{
"epoch": 0.44871794871794873,
"grad_norm": 0.5168382221700999,
"learning_rate": 2.5537757437070943e-05,
"loss": 0.2922,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.281344473361969,
"step": 280,
"valid_targets_mean": 4554.2,
"valid_targets_min": 287
},
{
"epoch": 0.4567307692307692,
"grad_norm": 0.5872781965575682,
"learning_rate": 2.59954233409611e-05,
"loss": 0.2773,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2553905248641968,
"step": 285,
"valid_targets_mean": 3695.9,
"valid_targets_min": 553
},
{
"epoch": 0.46474358974358976,
"grad_norm": 0.6324669780857965,
"learning_rate": 2.645308924485126e-05,
"loss": 0.2944,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2514147162437439,
"step": 290,
"valid_targets_mean": 4213.9,
"valid_targets_min": 260
},
{
"epoch": 0.47275641025641024,
"grad_norm": 0.5089251787451183,
"learning_rate": 2.6910755148741422e-05,
"loss": 0.2647,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25304847955703735,
"step": 295,
"valid_targets_mean": 4663.9,
"valid_targets_min": 329
},
{
"epoch": 0.4807692307692308,
"grad_norm": 0.5364973415599147,
"learning_rate": 2.7368421052631583e-05,
"loss": 0.2649,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24337220191955566,
"step": 300,
"valid_targets_mean": 4475.6,
"valid_targets_min": 322
},
{
"epoch": 0.48878205128205127,
"grad_norm": 0.6195475657680927,
"learning_rate": 2.782608695652174e-05,
"loss": 0.2783,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2563260793685913,
"step": 305,
"valid_targets_mean": 4387.9,
"valid_targets_min": 1790
},
{
"epoch": 0.4967948717948718,
"grad_norm": 0.5203206211291064,
"learning_rate": 2.8283752860411904e-05,
"loss": 0.2667,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2672708332538605,
"step": 310,
"valid_targets_mean": 4823.7,
"valid_targets_min": 683
},
{
"epoch": 0.5048076923076923,
"grad_norm": 0.5187640084196766,
"learning_rate": 2.8741418764302062e-05,
"loss": 0.2511,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23097531497478485,
"step": 315,
"valid_targets_mean": 4073.6,
"valid_targets_min": 440
},
{
"epoch": 0.5128205128205128,
"grad_norm": 0.5594458868665564,
"learning_rate": 2.9199084668192223e-05,
"loss": 0.2701,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27959808707237244,
"step": 320,
"valid_targets_mean": 5312.9,
"valid_targets_min": 1562
},
{
"epoch": 0.5208333333333334,
"grad_norm": 0.5572287731907528,
"learning_rate": 2.9656750572082384e-05,
"loss": 0.2576,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28152984380722046,
"step": 325,
"valid_targets_mean": 4349.9,
"valid_targets_min": 756
},
{
"epoch": 0.5288461538461539,
"grad_norm": 0.5025393252912373,
"learning_rate": 3.0114416475972544e-05,
"loss": 0.2637,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24482391774654388,
"step": 330,
"valid_targets_mean": 4584.7,
"valid_targets_min": 775
},
{
"epoch": 0.5368589743589743,
"grad_norm": 0.5316277354276495,
"learning_rate": 3.05720823798627e-05,
"loss": 0.2659,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23926711082458496,
"step": 335,
"valid_targets_mean": 4316.6,
"valid_targets_min": 1705
},
{
"epoch": 0.5448717948717948,
"grad_norm": 0.5926369540595648,
"learning_rate": 3.102974828375286e-05,
"loss": 0.2547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28076034784317017,
"step": 340,
"valid_targets_mean": 3431.1,
"valid_targets_min": 362
},
{
"epoch": 0.5528846153846154,
"grad_norm": 0.5060576505416645,
"learning_rate": 3.1487414187643024e-05,
"loss": 0.2724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2477727234363556,
"step": 345,
"valid_targets_mean": 4421.3,
"valid_targets_min": 2288
},
{
"epoch": 0.5608974358974359,
"grad_norm": 0.5175516881665588,
"learning_rate": 3.1945080091533184e-05,
"loss": 0.2597,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2503923177719116,
"step": 350,
"valid_targets_mean": 3955.1,
"valid_targets_min": 807
},
{
"epoch": 0.5689102564102564,
"grad_norm": 0.5946991554010684,
"learning_rate": 3.240274599542334e-05,
"loss": 0.2571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2551953196525574,
"step": 355,
"valid_targets_mean": 4417.1,
"valid_targets_min": 261
},
{
"epoch": 0.5769230769230769,
"grad_norm": 0.5927158003799573,
"learning_rate": 3.2860411899313506e-05,
"loss": 0.2578,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31151920557022095,
"step": 360,
"valid_targets_mean": 3531.7,
"valid_targets_min": 678
},
{
"epoch": 0.5849358974358975,
"grad_norm": 0.5891675778044099,
"learning_rate": 3.331807780320366e-05,
"loss": 0.2553,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26982879638671875,
"step": 365,
"valid_targets_mean": 3669.4,
"valid_targets_min": 464
},
{
"epoch": 0.592948717948718,
"grad_norm": 0.611509894731216,
"learning_rate": 3.377574370709382e-05,
"loss": 0.2728,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26848453283309937,
"step": 370,
"valid_targets_mean": 4375.4,
"valid_targets_min": 376
},
{
"epoch": 0.6009615384615384,
"grad_norm": 0.5196736341914934,
"learning_rate": 3.423340961098399e-05,
"loss": 0.2687,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.259830117225647,
"step": 375,
"valid_targets_mean": 4536.5,
"valid_targets_min": 278
},
{
"epoch": 0.6089743589743589,
"grad_norm": 0.5280807928384544,
"learning_rate": 3.469107551487414e-05,
"loss": 0.2625,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23455271124839783,
"step": 380,
"valid_targets_mean": 4795.4,
"valid_targets_min": 1419
},
{
"epoch": 0.6169871794871795,
"grad_norm": 0.5207215081373144,
"learning_rate": 3.5148741418764304e-05,
"loss": 0.2519,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25330376625061035,
"step": 385,
"valid_targets_mean": 4282.9,
"valid_targets_min": 324
},
{
"epoch": 0.625,
"grad_norm": 0.589894155996879,
"learning_rate": 3.5606407322654464e-05,
"loss": 0.2549,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2323845624923706,
"step": 390,
"valid_targets_mean": 4076.4,
"valid_targets_min": 703
},
{
"epoch": 0.6330128205128205,
"grad_norm": 0.5269124538117537,
"learning_rate": 3.6064073226544625e-05,
"loss": 0.2611,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27681398391723633,
"step": 395,
"valid_targets_mean": 4441.4,
"valid_targets_min": 1198
},
{
"epoch": 0.6410256410256411,
"grad_norm": 0.5128579930629643,
"learning_rate": 3.6521739130434786e-05,
"loss": 0.2591,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24692435562610626,
"step": 400,
"valid_targets_mean": 4770.9,
"valid_targets_min": 2265
},
{
"epoch": 0.6490384615384616,
"grad_norm": 0.5893572644841423,
"learning_rate": 3.697940503432495e-05,
"loss": 0.2429,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22618277370929718,
"step": 405,
"valid_targets_mean": 3721.7,
"valid_targets_min": 275
},
{
"epoch": 0.657051282051282,
"grad_norm": 0.5590239100077994,
"learning_rate": 3.743707093821511e-05,
"loss": 0.2635,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2413976639509201,
"step": 410,
"valid_targets_mean": 3935.8,
"valid_targets_min": 655
},
{
"epoch": 0.6650641025641025,
"grad_norm": 0.5150804243103135,
"learning_rate": 3.789473684210526e-05,
"loss": 0.2375,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27435582876205444,
"step": 415,
"valid_targets_mean": 4835.3,
"valid_targets_min": 828
},
{
"epoch": 0.6730769230769231,
"grad_norm": 0.5385482107772602,
"learning_rate": 3.835240274599543e-05,
"loss": 0.2507,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2416539043188095,
"step": 420,
"valid_targets_mean": 4454.2,
"valid_targets_min": 651
},
{
"epoch": 0.6810897435897436,
"grad_norm": 0.5117022146348998,
"learning_rate": 3.8810068649885584e-05,
"loss": 0.2552,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24814674258232117,
"step": 425,
"valid_targets_mean": 4624.2,
"valid_targets_min": 655
},
{
"epoch": 0.6891025641025641,
"grad_norm": 0.50963165573287,
"learning_rate": 3.9267734553775745e-05,
"loss": 0.2447,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22327658534049988,
"step": 430,
"valid_targets_mean": 4891.7,
"valid_targets_min": 1093
},
{
"epoch": 0.6971153846153846,
"grad_norm": 0.6814106307125438,
"learning_rate": 3.9725400457665905e-05,
"loss": 0.2697,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29641833901405334,
"step": 435,
"valid_targets_mean": 4282.2,
"valid_targets_min": 450
},
{
"epoch": 0.7051282051282052,
"grad_norm": 0.5610946816741627,
"learning_rate": 3.999997445219712e-05,
"loss": 0.2556,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27557656168937683,
"step": 440,
"valid_targets_mean": 4500.8,
"valid_targets_min": 610
},
{
"epoch": 0.7131410256410257,
"grad_norm": 0.4911274553640717,
"learning_rate": 3.999968704016428e-05,
"loss": 0.2465,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23766925930976868,
"step": 445,
"valid_targets_mean": 4024.7,
"valid_targets_min": 963
},
{
"epoch": 0.7211538461538461,
"grad_norm": 0.6304183675964307,
"learning_rate": 3.9999080285949514e-05,
"loss": 0.2422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2640378475189209,
"step": 450,
"valid_targets_mean": 5037.9,
"valid_targets_min": 915
},
{
"epoch": 0.7291666666666666,
"grad_norm": 0.5493336800995178,
"learning_rate": 3.999815419924108e-05,
"loss": 0.2681,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29132500290870667,
"step": 455,
"valid_targets_mean": 4322.1,
"valid_targets_min": 367
},
{
"epoch": 0.7371794871794872,
"grad_norm": 0.5950650375452828,
"learning_rate": 3.999690879482614e-05,
"loss": 0.2594,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25535181164741516,
"step": 460,
"valid_targets_mean": 4178.9,
"valid_targets_min": 1318
},
{
"epoch": 0.7451923076923077,
"grad_norm": 0.4995521707403712,
"learning_rate": 3.9995344092590506e-05,
"loss": 0.2616,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27951011061668396,
"step": 465,
"valid_targets_mean": 4545.2,
"valid_targets_min": 1084
},
{
"epoch": 0.7532051282051282,
"grad_norm": 0.6888046682474483,
"learning_rate": 3.999346011751835e-05,
"loss": 0.2433,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27734243869781494,
"step": 470,
"valid_targets_mean": 4337.5,
"valid_targets_min": 619
},
{
"epoch": 0.7612179487179487,
"grad_norm": 0.4672092070387373,
"learning_rate": 3.999125689969176e-05,
"loss": 0.2412,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24649207293987274,
"step": 475,
"valid_targets_mean": 4919.7,
"valid_targets_min": 467
},
{
"epoch": 0.7692307692307693,
"grad_norm": 0.5046724317006416,
"learning_rate": 3.9988734474290324e-05,
"loss": 0.2531,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23755775392055511,
"step": 480,
"valid_targets_mean": 4230.3,
"valid_targets_min": 714
},
{
"epoch": 0.7772435897435898,
"grad_norm": 0.5401478673844219,
"learning_rate": 3.9985892881590513e-05,
"loss": 0.2581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2535657286643982,
"step": 485,
"valid_targets_mean": 4308.9,
"valid_targets_min": 804
},
{
"epoch": 0.7852564102564102,
"grad_norm": 0.5114338731146565,
"learning_rate": 3.9982732166965054e-05,
"loss": 0.2544,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24153897166252136,
"step": 490,
"valid_targets_mean": 4224.6,
"valid_targets_min": 342
},
{
"epoch": 0.7932692307692307,
"grad_norm": 0.464321011079114,
"learning_rate": 3.997925238088221e-05,
"loss": 0.2497,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2033148854970932,
"step": 495,
"valid_targets_mean": 5041.6,
"valid_targets_min": 422
},
{
"epoch": 0.8012820512820513,
"grad_norm": 0.5487112664064877,
"learning_rate": 3.9975453578904975e-05,
"loss": 0.2502,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.245101660490036,
"step": 500,
"valid_targets_mean": 4924.8,
"valid_targets_min": 1336
},
{
"epoch": 0.8092948717948718,
"grad_norm": 0.727023852620155,
"learning_rate": 3.997133582169018e-05,
"loss": 0.2436,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2552146911621094,
"step": 505,
"valid_targets_mean": 4100.6,
"valid_targets_min": 550
},
{
"epoch": 0.8173076923076923,
"grad_norm": 0.6127494729153453,
"learning_rate": 3.996689917498754e-05,
"loss": 0.2508,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22979581356048584,
"step": 510,
"valid_targets_mean": 3454.3,
"valid_targets_min": 343
},
{
"epoch": 0.8253205128205128,
"grad_norm": 0.5715013359620111,
"learning_rate": 3.9962143709638585e-05,
"loss": 0.2351,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23442040383815765,
"step": 515,
"valid_targets_mean": 4235.6,
"valid_targets_min": 591
},
{
"epoch": 0.8333333333333334,
"grad_norm": 0.4712969356536701,
"learning_rate": 3.995706950157554e-05,
"loss": 0.252,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21304473280906677,
"step": 520,
"valid_targets_mean": 4792.1,
"valid_targets_min": 297
},
{
"epoch": 0.8413461538461539,
"grad_norm": 0.4915739565262899,
"learning_rate": 3.995167663182008e-05,
"loss": 0.2427,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24165457487106323,
"step": 525,
"valid_targets_mean": 4803.0,
"valid_targets_min": 1043
},
{
"epoch": 0.8493589743589743,
"grad_norm": 0.5001565793740625,
"learning_rate": 3.994596518648214e-05,
"loss": 0.2525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2570037543773651,
"step": 530,
"valid_targets_mean": 4697.3,
"valid_targets_min": 1050
},
{
"epoch": 0.8573717948717948,
"grad_norm": 0.5968934071913429,
"learning_rate": 3.993993525675838e-05,
"loss": 0.2275,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2219688594341278,
"step": 535,
"valid_targets_mean": 3579.8,
"valid_targets_min": 454
},
{
"epoch": 0.8653846153846154,
"grad_norm": 0.5526323789185988,
"learning_rate": 3.993358693893086e-05,
"loss": 0.2407,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22856299579143524,
"step": 540,
"valid_targets_mean": 4354.6,
"valid_targets_min": 384
},
{
"epoch": 0.8733974358974359,
"grad_norm": 0.6149255083510035,
"learning_rate": 3.9926920334365457e-05,
"loss": 0.2373,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25257566571235657,
"step": 545,
"valid_targets_mean": 3485.2,
"valid_targets_min": 395
},
{
"epoch": 0.8814102564102564,
"grad_norm": 0.5713833614585203,
"learning_rate": 3.991993554951023e-05,
"loss": 0.2549,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23296888172626495,
"step": 550,
"valid_targets_mean": 4345.3,
"valid_targets_min": 304
},
{
"epoch": 0.8894230769230769,
"grad_norm": 0.760910504834305,
"learning_rate": 3.991263269589376e-05,
"loss": 0.2469,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24157465994358063,
"step": 555,
"valid_targets_mean": 4510.9,
"valid_targets_min": 746
},
{
"epoch": 0.8974358974358975,
"grad_norm": 0.5356826828138093,
"learning_rate": 3.990501189012332e-05,
"loss": 0.2196,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2129981815814972,
"step": 560,
"valid_targets_mean": 4623.8,
"valid_targets_min": 286
},
{
"epoch": 0.905448717948718,
"grad_norm": 0.43050815793602054,
"learning_rate": 3.989707325388305e-05,
"loss": 0.2356,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2072744220495224,
"step": 565,
"valid_targets_mean": 4800.8,
"valid_targets_min": 1634
},
{
"epoch": 0.9134615384615384,
"grad_norm": 0.4934009867867431,
"learning_rate": 3.9888816913932016e-05,
"loss": 0.2445,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2358398288488388,
"step": 570,
"valid_targets_mean": 4539.9,
"valid_targets_min": 896
},
{
"epoch": 0.9214743589743589,
"grad_norm": 0.551013723763327,
"learning_rate": 3.988024300210215e-05,
"loss": 0.2349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23630398511886597,
"step": 575,
"valid_targets_mean": 4338.2,
"valid_targets_min": 327
},
{
"epoch": 0.9294871794871795,
"grad_norm": 0.4913475851614203,
"learning_rate": 3.987135165529618e-05,
"loss": 0.2341,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22553719580173492,
"step": 580,
"valid_targets_mean": 5399.9,
"valid_targets_min": 826
},
{
"epoch": 0.9375,
"grad_norm": 0.7451879820563563,
"learning_rate": 3.9862143015485446e-05,
"loss": 0.2524,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24659019708633423,
"step": 585,
"valid_targets_mean": 4197.9,
"valid_targets_min": 258
},
{
"epoch": 0.9455128205128205,
"grad_norm": 0.615645952040602,
"learning_rate": 3.985261722970759e-05,
"loss": 0.2504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2549634575843811,
"step": 590,
"valid_targets_mean": 3471.9,
"valid_targets_min": 642
},
{
"epoch": 0.9535256410256411,
"grad_norm": 0.5014073315079357,
"learning_rate": 3.984277445006426e-05,
"loss": 0.2412,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2386205494403839,
"step": 595,
"valid_targets_mean": 4235.6,
"valid_targets_min": 291
},
{
"epoch": 0.9615384615384616,
"grad_norm": 0.648859206902621,
"learning_rate": 3.9832614833718654e-05,
"loss": 0.2442,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2423931211233139,
"step": 600,
"valid_targets_mean": 3781.2,
"valid_targets_min": 529
},
{
"epoch": 0.969551282051282,
"grad_norm": 0.6337739882580341,
"learning_rate": 3.9822138542893005e-05,
"loss": 0.2442,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2040448784828186,
"step": 605,
"valid_targets_mean": 3677.9,
"valid_targets_min": 469
},
{
"epoch": 0.9775641025641025,
"grad_norm": 0.5987015608466942,
"learning_rate": 3.9811345744866014e-05,
"loss": 0.2482,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26201778650283813,
"step": 610,
"valid_targets_mean": 4079.4,
"valid_targets_min": 372
},
{
"epoch": 0.9855769230769231,
"grad_norm": 0.5754667756398731,
"learning_rate": 3.980023661197016e-05,
"loss": 0.2503,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2476460188627243,
"step": 615,
"valid_targets_mean": 3490.6,
"valid_targets_min": 321
},
{
"epoch": 0.9935897435897436,
"grad_norm": 0.46078820352529754,
"learning_rate": 3.978881132158896e-05,
"loss": 0.2439,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20165373384952545,
"step": 620,
"valid_targets_mean": 5304.0,
"valid_targets_min": 2754
},
{
"epoch": 1.001602564102564,
"grad_norm": 0.5183260222036216,
"learning_rate": 3.9777070056154124e-05,
"loss": 0.2372,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24027764797210693,
"step": 625,
"valid_targets_mean": 4369.7,
"valid_targets_min": 799
},
{
"epoch": 1.0096153846153846,
"grad_norm": 0.615519356411254,
"learning_rate": 3.976501300314264e-05,
"loss": 0.2268,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20788605511188507,
"step": 630,
"valid_targets_mean": 3541.9,
"valid_targets_min": 642
},
{
"epoch": 1.017628205128205,
"grad_norm": 0.5311097491711796,
"learning_rate": 3.9752640355073825e-05,
"loss": 0.2367,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24286183714866638,
"step": 635,
"valid_targets_mean": 4480.9,
"valid_targets_min": 726
},
{
"epoch": 1.0256410256410255,
"grad_norm": 0.523048437187609,
"learning_rate": 3.9739952309506175e-05,
"loss": 0.2175,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2169773131608963,
"step": 640,
"valid_targets_mean": 4380.6,
"valid_targets_min": 1969
},
{
"epoch": 1.0336538461538463,
"grad_norm": 0.5235971433465657,
"learning_rate": 3.972694906903427e-05,
"loss": 0.2333,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2523080110549927,
"step": 645,
"valid_targets_mean": 3814.9,
"valid_targets_min": 301
},
{
"epoch": 1.0416666666666667,
"grad_norm": 0.5744816977650197,
"learning_rate": 3.971363084128552e-05,
"loss": 0.2418,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2559906244277954,
"step": 650,
"valid_targets_mean": 4120.1,
"valid_targets_min": 287
},
{
"epoch": 1.0496794871794872,
"grad_norm": 0.5543283932610994,
"learning_rate": 3.969999783891685e-05,
"loss": 0.2457,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.255851149559021,
"step": 655,
"valid_targets_mean": 3687.6,
"valid_targets_min": 287
},
{
"epoch": 1.0576923076923077,
"grad_norm": 0.543824204802863,
"learning_rate": 3.96860502796113e-05,
"loss": 0.2178,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20594194531440735,
"step": 660,
"valid_targets_mean": 4272.8,
"valid_targets_min": 332
},
{
"epoch": 1.0657051282051282,
"grad_norm": 0.5429684185849859,
"learning_rate": 3.967178838607456e-05,
"loss": 0.2163,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2007255107164383,
"step": 665,
"valid_targets_mean": 3549.8,
"valid_targets_min": 364
},
{
"epoch": 1.0737179487179487,
"grad_norm": 0.5135653732491821,
"learning_rate": 3.965721238603139e-05,
"loss": 0.2278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2522657811641693,
"step": 670,
"valid_targets_mean": 4205.8,
"valid_targets_min": 263
},
{
"epoch": 1.0817307692307692,
"grad_norm": 0.5067143702824006,
"learning_rate": 3.964232251222203e-05,
"loss": 0.2258,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22779108583927155,
"step": 675,
"valid_targets_mean": 4169.6,
"valid_targets_min": 214
},
{
"epoch": 1.0897435897435896,
"grad_norm": 0.4736864391273317,
"learning_rate": 3.962711900239844e-05,
"loss": 0.2228,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20398220419883728,
"step": 680,
"valid_targets_mean": 4002.2,
"valid_targets_min": 1030
},
{
"epoch": 1.0977564102564104,
"grad_norm": 0.5843562547096047,
"learning_rate": 3.961160209932051e-05,
"loss": 0.2334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2181626260280609,
"step": 685,
"valid_targets_mean": 3384.9,
"valid_targets_min": 1357
},
{
"epoch": 1.1057692307692308,
"grad_norm": 0.5394495021787433,
"learning_rate": 3.95957720507522e-05,
"loss": 0.2312,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22555160522460938,
"step": 690,
"valid_targets_mean": 4161.2,
"valid_targets_min": 355
},
{
"epoch": 1.1137820512820513,
"grad_norm": 0.5767812746038268,
"learning_rate": 3.957962910945759e-05,
"loss": 0.2273,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2506183981895447,
"step": 695,
"valid_targets_mean": 4608.7,
"valid_targets_min": 321
},
{
"epoch": 1.1217948717948718,
"grad_norm": 0.499371761916278,
"learning_rate": 3.9563173533196805e-05,
"loss": 0.2275,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23579809069633484,
"step": 700,
"valid_targets_mean": 4448.9,
"valid_targets_min": 1902
},
{
"epoch": 1.1298076923076923,
"grad_norm": 0.5984748913222906,
"learning_rate": 3.954640558472195e-05,
"loss": 0.241,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23393550515174866,
"step": 705,
"valid_targets_mean": 5014.4,
"valid_targets_min": 2172
},
{
"epoch": 1.1378205128205128,
"grad_norm": 0.4646852043866937,
"learning_rate": 3.952932553177287e-05,
"loss": 0.2154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20852085947990417,
"step": 710,
"valid_targets_mean": 4491.9,
"valid_targets_min": 880
},
{
"epoch": 1.1458333333333333,
"grad_norm": 0.6696611112478394,
"learning_rate": 3.95119336470729e-05,
"loss": 0.2208,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21880467236042023,
"step": 715,
"valid_targets_mean": 4209.5,
"valid_targets_min": 920
},
{
"epoch": 1.1538461538461537,
"grad_norm": 0.5058932047031148,
"learning_rate": 3.949423020832451e-05,
"loss": 0.2192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2248716652393341,
"step": 720,
"valid_targets_mean": 3976.8,
"valid_targets_min": 317
},
{
"epoch": 1.1618589743589745,
"grad_norm": 0.5987009971967244,
"learning_rate": 3.947621549820485e-05,
"loss": 0.234,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23888376355171204,
"step": 725,
"valid_targets_mean": 4572.8,
"valid_targets_min": 1665
},
{
"epoch": 1.169871794871795,
"grad_norm": 0.435844427806574,
"learning_rate": 3.945788980436129e-05,
"loss": 0.236,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20528161525726318,
"step": 730,
"valid_targets_mean": 4759.8,
"valid_targets_min": 1807
},
{
"epoch": 1.1778846153846154,
"grad_norm": 0.5249046329450383,
"learning_rate": 3.943925341940673e-05,
"loss": 0.2367,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2288103699684143,
"step": 735,
"valid_targets_mean": 3814.1,
"valid_targets_min": 292
},
{
"epoch": 1.185897435897436,
"grad_norm": 0.5798938550572477,
"learning_rate": 3.942030664091503e-05,
"loss": 0.2313,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2773657441139221,
"step": 740,
"valid_targets_mean": 3972.2,
"valid_targets_min": 901
},
{
"epoch": 1.1939102564102564,
"grad_norm": 0.4912180000449908,
"learning_rate": 3.9401049771416214e-05,
"loss": 0.2249,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21388167142868042,
"step": 745,
"valid_targets_mean": 4266.1,
"valid_targets_min": 552
},
{
"epoch": 1.2019230769230769,
"grad_norm": 0.5472461605339294,
"learning_rate": 3.938148311839162e-05,
"loss": 0.2294,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24619370698928833,
"step": 750,
"valid_targets_mean": 3647.3,
"valid_targets_min": 263
},
{
"epoch": 1.2099358974358974,
"grad_norm": 0.6179919672963712,
"learning_rate": 3.9361606994269014e-05,
"loss": 0.2295,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22163723409175873,
"step": 755,
"valid_targets_mean": 3494.5,
"valid_targets_min": 304
},
{
"epoch": 1.217948717948718,
"grad_norm": 0.4820316573749669,
"learning_rate": 3.934142171641763e-05,
"loss": 0.2386,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23646295070648193,
"step": 760,
"valid_targets_mean": 5003.5,
"valid_targets_min": 300
},
{
"epoch": 1.2259615384615385,
"grad_norm": 0.6693435566711766,
"learning_rate": 3.9320927607143003e-05,
"loss": 0.2322,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23479235172271729,
"step": 765,
"valid_targets_mean": 4457.5,
"valid_targets_min": 364
},
{
"epoch": 1.233974358974359,
"grad_norm": 0.5586399655465525,
"learning_rate": 3.9300124993681976e-05,
"loss": 0.2332,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24570053815841675,
"step": 770,
"valid_targets_mean": 4665.1,
"valid_targets_min": 801
},
{
"epoch": 1.2419871794871795,
"grad_norm": 0.6193835746341648,
"learning_rate": 3.9279014208197317e-05,
"loss": 0.2341,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2803085148334503,
"step": 775,
"valid_targets_mean": 4447.7,
"valid_targets_min": 261
},
{
"epoch": 1.25,
"grad_norm": 0.5541353759085407,
"learning_rate": 3.925759558777252e-05,
"loss": 0.2184,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.212779238820076,
"step": 780,
"valid_targets_mean": 4576.8,
"valid_targets_min": 355
},
{
"epoch": 1.2580128205128205,
"grad_norm": 0.6445366005275464,
"learning_rate": 3.923586947440639e-05,
"loss": 0.2287,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27303412556648254,
"step": 785,
"valid_targets_mean": 3726.4,
"valid_targets_min": 313
},
{
"epoch": 1.266025641025641,
"grad_norm": 0.5351338916304204,
"learning_rate": 3.921383621500758e-05,
"loss": 0.2241,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21160487830638885,
"step": 790,
"valid_targets_mean": 3934.5,
"valid_targets_min": 605
},
{
"epoch": 1.2740384615384617,
"grad_norm": 0.5121690871356921,
"learning_rate": 3.919149616138906e-05,
"loss": 0.2373,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2367040514945984,
"step": 795,
"valid_targets_mean": 4527.7,
"valid_targets_min": 523
},
{
"epoch": 1.282051282051282,
"grad_norm": 0.5146096120322179,
"learning_rate": 3.916884967026246e-05,
"loss": 0.2082,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2128678858280182,
"step": 800,
"valid_targets_mean": 3931.8,
"valid_targets_min": 1003
},
{
"epoch": 1.2900641025641026,
"grad_norm": 0.5382862721846314,
"learning_rate": 3.914589710323245e-05,
"loss": 0.2325,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22114819288253784,
"step": 805,
"valid_targets_mean": 4836.5,
"valid_targets_min": 291
},
{
"epoch": 1.2980769230769231,
"grad_norm": 0.41690244719704006,
"learning_rate": 3.912263882679091e-05,
"loss": 0.2314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2032252997159958,
"step": 810,
"valid_targets_mean": 4809.9,
"valid_targets_min": 1751
},
{
"epoch": 1.3060897435897436,
"grad_norm": 0.5312757387265346,
"learning_rate": 3.9099075212311076e-05,
"loss": 0.2251,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24702531099319458,
"step": 815,
"valid_targets_mean": 4187.5,
"valid_targets_min": 583
},
{
"epoch": 1.314102564102564,
"grad_norm": 0.4967044642728975,
"learning_rate": 3.9075206636041646e-05,
"loss": 0.2132,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20572376251220703,
"step": 820,
"valid_targets_mean": 5312.8,
"valid_targets_min": 941
},
{
"epoch": 1.3221153846153846,
"grad_norm": 0.5253541673563855,
"learning_rate": 3.905103347910075e-05,
"loss": 0.218,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2204853892326355,
"step": 825,
"valid_targets_mean": 5501.3,
"valid_targets_min": 916
},
{
"epoch": 1.330128205128205,
"grad_norm": 0.45058459328927725,
"learning_rate": 3.902655612746985e-05,
"loss": 0.2265,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22448717057704926,
"step": 830,
"valid_targets_mean": 4959.6,
"valid_targets_min": 1714
},
{
"epoch": 1.3381410256410255,
"grad_norm": 0.4966056187290371,
"learning_rate": 3.900177497198761e-05,
"loss": 0.2327,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21923819184303284,
"step": 835,
"valid_targets_mean": 3980.8,
"valid_targets_min": 284
},
{
"epoch": 1.3461538461538463,
"grad_norm": 0.6217547453516624,
"learning_rate": 3.8976690408343635e-05,
"loss": 0.2231,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24218757450580597,
"step": 840,
"valid_targets_mean": 3487.1,
"valid_targets_min": 428
},
{
"epoch": 1.3541666666666667,
"grad_norm": 0.5355637761564441,
"learning_rate": 3.8951302837072165e-05,
"loss": 0.2168,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1972593516111374,
"step": 845,
"valid_targets_mean": 3523.4,
"valid_targets_min": 315
},
{
"epoch": 1.3621794871794872,
"grad_norm": 0.4905155050423892,
"learning_rate": 3.892561266354566e-05,
"loss": 0.2115,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21404795348644257,
"step": 850,
"valid_targets_mean": 4725.7,
"valid_targets_min": 1042
},
{
"epoch": 1.3701923076923077,
"grad_norm": 0.5675888302643592,
"learning_rate": 3.889962029796833e-05,
"loss": 0.2216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2576954662799835,
"step": 855,
"valid_targets_mean": 4191.7,
"valid_targets_min": 698
},
{
"epoch": 1.3782051282051282,
"grad_norm": 0.5802070948524329,
"learning_rate": 3.887332615536962e-05,
"loss": 0.2265,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2517366409301758,
"step": 860,
"valid_targets_mean": 3482.4,
"valid_targets_min": 244
},
{
"epoch": 1.3862179487179487,
"grad_norm": 0.501091816564502,
"learning_rate": 3.8846730655597535e-05,
"loss": 0.228,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23372066020965576,
"step": 865,
"valid_targets_mean": 3787.9,
"valid_targets_min": 339
},
{
"epoch": 1.3942307692307692,
"grad_norm": 0.5045732207468887,
"learning_rate": 3.881983422331198e-05,
"loss": 0.2329,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24290408194065094,
"step": 870,
"valid_targets_mean": 5002.9,
"valid_targets_min": 1983
},
{
"epoch": 1.4022435897435899,
"grad_norm": 0.4572209892415149,
"learning_rate": 3.879263728797792e-05,
"loss": 0.226,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23531322181224823,
"step": 875,
"valid_targets_mean": 4878.1,
"valid_targets_min": 759
},
{
"epoch": 1.4102564102564101,
"grad_norm": 0.5216848374755977,
"learning_rate": 3.876514028385861e-05,
"loss": 0.2492,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20825883746147156,
"step": 880,
"valid_targets_mean": 5603.9,
"valid_targets_min": 1833
},
{
"epoch": 1.4182692307692308,
"grad_norm": 1.1120321079733493,
"learning_rate": 3.873734365000857e-05,
"loss": 0.2277,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19830474257469177,
"step": 885,
"valid_targets_mean": 4210.7,
"valid_targets_min": 546
},
{
"epoch": 1.4262820512820513,
"grad_norm": 1.1474898006319683,
"learning_rate": 3.870924783026663e-05,
"loss": 0.2154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21796885132789612,
"step": 890,
"valid_targets_mean": 3727.3,
"valid_targets_min": 282
},
{
"epoch": 1.4342948717948718,
"grad_norm": 0.5425435044335261,
"learning_rate": 3.8680853273248826e-05,
"loss": 0.2398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2845466732978821,
"step": 895,
"valid_targets_mean": 3722.8,
"valid_targets_min": 400
},
{
"epoch": 1.4423076923076923,
"grad_norm": 0.5025723379803859,
"learning_rate": 3.865216043234126e-05,
"loss": 0.2142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2267698496580124,
"step": 900,
"valid_targets_mean": 4264.2,
"valid_targets_min": 989
},
{
"epoch": 1.4503205128205128,
"grad_norm": 0.48413187608652714,
"learning_rate": 3.862316976569281e-05,
"loss": 0.2176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22797155380249023,
"step": 905,
"valid_targets_mean": 4292.0,
"valid_targets_min": 262
},
{
"epoch": 1.4583333333333333,
"grad_norm": 0.5070571703128157,
"learning_rate": 3.859388173620785e-05,
"loss": 0.2176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1968487799167633,
"step": 910,
"valid_targets_mean": 4464.0,
"valid_targets_min": 336
},
{
"epoch": 1.4663461538461537,
"grad_norm": 0.46715568420258785,
"learning_rate": 3.8564296811538874e-05,
"loss": 0.2343,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22878612577915192,
"step": 915,
"valid_targets_mean": 4437.7,
"valid_targets_min": 580
},
{
"epoch": 1.4743589743589745,
"grad_norm": 0.48961696947712857,
"learning_rate": 3.853441546407898e-05,
"loss": 0.2159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2069419026374817,
"step": 920,
"valid_targets_mean": 3950.2,
"valid_targets_min": 310
},
{
"epoch": 1.482371794871795,
"grad_norm": 0.494501771770836,
"learning_rate": 3.850423817095438e-05,
"loss": 0.2211,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23218463361263275,
"step": 925,
"valid_targets_mean": 4225.6,
"valid_targets_min": 316
},
{
"epoch": 1.4903846153846154,
"grad_norm": 0.4405575665115888,
"learning_rate": 3.847376541401674e-05,
"loss": 0.2302,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21161818504333496,
"step": 930,
"valid_targets_mean": 4917.6,
"valid_targets_min": 375
},
{
"epoch": 1.498397435897436,
"grad_norm": 0.44717420584520595,
"learning_rate": 3.844299767983551e-05,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25120991468429565,
"step": 935,
"valid_targets_mean": 5459.8,
"valid_targets_min": 314
},
{
"epoch": 1.5064102564102564,
"grad_norm": 0.5649981869319821,
"learning_rate": 3.841193545969015e-05,
"loss": 0.2192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.223961740732193,
"step": 940,
"valid_targets_mean": 4889.3,
"valid_targets_min": 1141
},
{
"epoch": 1.5144230769230769,
"grad_norm": 0.5610756449476008,
"learning_rate": 3.8380579249562265e-05,
"loss": 0.2317,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2219979465007782,
"step": 945,
"valid_targets_mean": 3437.4,
"valid_targets_min": 614
},
{
"epoch": 1.5224358974358974,
"grad_norm": 0.5326466356734885,
"learning_rate": 3.8348929550127734e-05,
"loss": 0.2112,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2324865460395813,
"step": 950,
"valid_targets_mean": 4199.5,
"valid_targets_min": 300
},
{
"epoch": 1.530448717948718,
"grad_norm": 0.47579534122015166,
"learning_rate": 3.831698686674866e-05,
"loss": 0.2207,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22800025343894958,
"step": 955,
"valid_targets_mean": 5039.6,
"valid_targets_min": 640
},
{
"epoch": 1.5384615384615383,
"grad_norm": 0.5107178330292805,
"learning_rate": 3.828475170946534e-05,
"loss": 0.2165,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21851620078086853,
"step": 960,
"valid_targets_mean": 4972.2,
"valid_targets_min": 323
},
{
"epoch": 1.546474358974359,
"grad_norm": 0.5131620240873577,
"learning_rate": 3.8252224592988087e-05,
"loss": 0.2083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20986801385879517,
"step": 965,
"valid_targets_mean": 4395.8,
"valid_targets_min": 722
},
{
"epoch": 1.5544871794871795,
"grad_norm": 0.43516844997971743,
"learning_rate": 3.821940603668906e-05,
"loss": 0.2237,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2037906050682068,
"step": 970,
"valid_targets_mean": 5107.1,
"valid_targets_min": 1185
},
{
"epoch": 1.5625,
"grad_norm": 0.582189568389626,
"learning_rate": 3.8186296564593924e-05,
"loss": 0.2207,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20876115560531616,
"step": 975,
"valid_targets_mean": 4523.4,
"valid_targets_min": 256
},
{
"epoch": 1.5705128205128205,
"grad_norm": 0.49097961546973645,
"learning_rate": 3.815289670537351e-05,
"loss": 0.2432,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23335549235343933,
"step": 980,
"valid_targets_mean": 4919.2,
"valid_targets_min": 972
},
{
"epoch": 1.578525641025641,
"grad_norm": 0.5006639659606942,
"learning_rate": 3.811920699233535e-05,
"loss": 0.2215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.207101508975029,
"step": 985,
"valid_targets_mean": 4426.9,
"valid_targets_min": 533
},
{
"epoch": 1.5865384615384617,
"grad_norm": 0.644279729424333,
"learning_rate": 3.8085227963415186e-05,
"loss": 0.2229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22972708940505981,
"step": 990,
"valid_targets_mean": 4833.6,
"valid_targets_min": 2356
},
{
"epoch": 1.594551282051282,
"grad_norm": 0.5159507712065665,
"learning_rate": 3.805096016116838e-05,
"loss": 0.2307,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24587450921535492,
"step": 995,
"valid_targets_mean": 4441.9,
"valid_targets_min": 790
},
{
"epoch": 1.6025641025641026,
"grad_norm": 0.5149030759155213,
"learning_rate": 3.801640413276121e-05,
"loss": 0.2371,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23141878843307495,
"step": 1000,
"valid_targets_mean": 4767.9,
"valid_targets_min": 2032
},
{
"epoch": 1.6105769230769231,
"grad_norm": 0.46190768496009427,
"learning_rate": 3.7981560429962204e-05,
"loss": 0.2264,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2219385802745819,
"step": 1005,
"valid_targets_mean": 4610.0,
"valid_targets_min": 394
},
{
"epoch": 1.6185897435897436,
"grad_norm": 0.5907880679241649,
"learning_rate": 3.7946429609133274e-05,
"loss": 0.2252,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25632813572883606,
"step": 1010,
"valid_targets_mean": 3638.3,
"valid_targets_min": 270
},
{
"epoch": 1.626602564102564,
"grad_norm": 0.5288141242399365,
"learning_rate": 3.791101223122084e-05,
"loss": 0.2211,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23555824160575867,
"step": 1015,
"valid_targets_mean": 3608.1,
"valid_targets_min": 301
},
{
"epoch": 1.6346153846153846,
"grad_norm": 0.4855485033357978,
"learning_rate": 3.787530886174688e-05,
"loss": 0.2285,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23066557943820953,
"step": 1020,
"valid_targets_mean": 4533.7,
"valid_targets_min": 478
},
{
"epoch": 1.6426282051282053,
"grad_norm": 0.44129001593673417,
"learning_rate": 3.783932007079992e-05,
"loss": 0.2244,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21713760495185852,
"step": 1025,
"valid_targets_mean": 5150.0,
"valid_targets_min": 555
},
{
"epoch": 1.6506410256410255,
"grad_norm": 0.4869098130845637,
"learning_rate": 3.7803046433025905e-05,
"loss": 0.2154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21258345246315002,
"step": 1030,
"valid_targets_mean": 4247.4,
"valid_targets_min": 366
},
{
"epoch": 1.6586538461538463,
"grad_norm": 0.48150004210561737,
"learning_rate": 3.7766488527619024e-05,
"loss": 0.2307,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21068784594535828,
"step": 1035,
"valid_targets_mean": 4309.9,
"valid_targets_min": 592
},
{
"epoch": 1.6666666666666665,
"grad_norm": 0.46046811036917534,
"learning_rate": 3.772964693831247e-05,
"loss": 0.2278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23497143387794495,
"step": 1040,
"valid_targets_mean": 3955.5,
"valid_targets_min": 477
},
{
"epoch": 1.6746794871794872,
"grad_norm": 0.5054716091293595,
"learning_rate": 3.7692522253369136e-05,
"loss": 0.2327,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19847500324249268,
"step": 1045,
"valid_targets_mean": 4319.3,
"valid_targets_min": 315
},
{
"epoch": 1.6826923076923077,
"grad_norm": 0.46824714239053444,
"learning_rate": 3.7655115065572194e-05,
"loss": 0.2187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23209133744239807,
"step": 1050,
"valid_targets_mean": 4523.6,
"valid_targets_min": 750
},
{
"epoch": 1.6907051282051282,
"grad_norm": 0.5255497198453384,
"learning_rate": 3.7617425972215626e-05,
"loss": 0.216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20271123945713043,
"step": 1055,
"valid_targets_mean": 3911.7,
"valid_targets_min": 282
},
{
"epoch": 1.6987179487179487,
"grad_norm": 0.4449364726840892,
"learning_rate": 3.757945557509472e-05,
"loss": 0.2344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2020668238401413,
"step": 1060,
"valid_targets_mean": 4776.2,
"valid_targets_min": 290
},
{
"epoch": 1.7067307692307692,
"grad_norm": 0.45229917313087076,
"learning_rate": 3.7541204480496444e-05,
"loss": 0.223,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20702438056468964,
"step": 1065,
"valid_targets_mean": 4405.7,
"valid_targets_min": 1406
},
{
"epoch": 1.7147435897435899,
"grad_norm": 0.501236055742325,
"learning_rate": 3.7502673299189745e-05,
"loss": 0.2198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2156430035829544,
"step": 1070,
"valid_targets_mean": 3642.9,
"valid_targets_min": 277
},
{
"epoch": 1.7227564102564101,
"grad_norm": 0.49111402362665263,
"learning_rate": 3.746386264641583e-05,
"loss": 0.2304,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2629583775997162,
"step": 1075,
"valid_targets_mean": 4801.9,
"valid_targets_min": 1010
},
{
"epoch": 1.7307692307692308,
"grad_norm": 0.5386304799588276,
"learning_rate": 3.7424773141878324e-05,
"loss": 0.2217,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25513795018196106,
"step": 1080,
"valid_targets_mean": 3994.9,
"valid_targets_min": 423
},
{
"epoch": 1.7387820512820513,
"grad_norm": 0.6033801531402724,
"learning_rate": 3.738540540973338e-05,
"loss": 0.2256,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24040186405181885,
"step": 1085,
"valid_targets_mean": 5301.0,
"valid_targets_min": 1865
},
{
"epoch": 1.7467948717948718,
"grad_norm": 0.47797338225998187,
"learning_rate": 3.7345760078579695e-05,
"loss": 0.2079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21624310314655304,
"step": 1090,
"valid_targets_mean": 4393.9,
"valid_targets_min": 639
},
{
"epoch": 1.7548076923076923,
"grad_norm": 0.4962291315791365,
"learning_rate": 3.730583778144852e-05,
"loss": 0.2258,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22629985213279724,
"step": 1095,
"valid_targets_mean": 4448.3,
"valid_targets_min": 325
},
{
"epoch": 1.7628205128205128,
"grad_norm": 0.42542899702880366,
"learning_rate": 3.7265639155793494e-05,
"loss": 0.2169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21988126635551453,
"step": 1100,
"valid_targets_mean": 5359.6,
"valid_targets_min": 1318
},
{
"epoch": 1.7708333333333335,
"grad_norm": 0.5908032224965943,
"learning_rate": 3.7225164843480503e-05,
"loss": 0.2283,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23074805736541748,
"step": 1105,
"valid_targets_mean": 4068.8,
"valid_targets_min": 480
},
{
"epoch": 1.7788461538461537,
"grad_norm": 0.4467750332051743,
"learning_rate": 3.7184415490777426e-05,
"loss": 0.2249,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1936504989862442,
"step": 1110,
"valid_targets_mean": 3998.6,
"valid_targets_min": 365
},
{
"epoch": 1.7868589743589745,
"grad_norm": 0.5238360821909998,
"learning_rate": 3.714339174834379e-05,
"loss": 0.2256,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24224252998828888,
"step": 1115,
"valid_targets_mean": 3999.8,
"valid_targets_min": 319
},
{
"epoch": 1.7948717948717947,
"grad_norm": 0.41583700666530465,
"learning_rate": 3.710209427122044e-05,
"loss": 0.2301,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21709281206130981,
"step": 1120,
"valid_targets_mean": 5194.3,
"valid_targets_min": 722
},
{
"epoch": 1.8028846153846154,
"grad_norm": 0.9562822016716469,
"learning_rate": 3.7060523718819e-05,
"loss": 0.2143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2202528715133667,
"step": 1125,
"valid_targets_mean": 4092.9,
"valid_targets_min": 338
},
{
"epoch": 1.810897435897436,
"grad_norm": 0.5396342608547219,
"learning_rate": 3.701868075491139e-05,
"loss": 0.2274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24480941891670227,
"step": 1130,
"valid_targets_mean": 4232.1,
"valid_targets_min": 322
},
{
"epoch": 1.8189102564102564,
"grad_norm": 0.4755409098359112,
"learning_rate": 3.697656604761926e-05,
"loss": 0.2095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19729657471179962,
"step": 1135,
"valid_targets_mean": 4005.1,
"valid_targets_min": 1011
},
{
"epoch": 1.8269230769230769,
"grad_norm": 0.4354481953346282,
"learning_rate": 3.693418026940325e-05,
"loss": 0.2221,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2039467692375183,
"step": 1140,
"valid_targets_mean": 4886.9,
"valid_targets_min": 287
},
{
"epoch": 1.8349358974358974,
"grad_norm": 0.45733650099155515,
"learning_rate": 3.689152409705229e-05,
"loss": 0.2188,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20380863547325134,
"step": 1145,
"valid_targets_mean": 4495.7,
"valid_targets_min": 314
},
{
"epoch": 1.842948717948718,
"grad_norm": 0.5043570869567778,
"learning_rate": 3.6848598211672794e-05,
"loss": 0.216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25355690717697144,
"step": 1150,
"valid_targets_mean": 4753.0,
"valid_targets_min": 740
},
{
"epoch": 1.8509615384615383,
"grad_norm": 0.49031269098902835,
"learning_rate": 3.6805403298677797e-05,
"loss": 0.2213,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25376126170158386,
"step": 1155,
"valid_targets_mean": 4130.6,
"valid_targets_min": 565
},
{
"epoch": 1.858974358974359,
"grad_norm": 0.5561400017535065,
"learning_rate": 3.6761940047775966e-05,
"loss": 0.2175,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24656134843826294,
"step": 1160,
"valid_targets_mean": 4025.7,
"valid_targets_min": 448
},
{
"epoch": 1.8669871794871795,
"grad_norm": 0.550436453735836,
"learning_rate": 3.671820915296063e-05,
"loss": 0.2278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2330714762210846,
"step": 1165,
"valid_targets_mean": 3034.0,
"valid_targets_min": 329
},
{
"epoch": 1.875,
"grad_norm": 0.49546370898085806,
"learning_rate": 3.667421131249869e-05,
"loss": 0.2309,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24480220675468445,
"step": 1170,
"valid_targets_mean": 4666.2,
"valid_targets_min": 421
},
{
"epoch": 1.8830128205128205,
"grad_norm": 0.44700045416460626,
"learning_rate": 3.662994722891946e-05,
"loss": 0.2232,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2118150293827057,
"step": 1175,
"valid_targets_mean": 4497.4,
"valid_targets_min": 376
},
{
"epoch": 1.891025641025641,
"grad_norm": 0.4579009123490911,
"learning_rate": 3.658541760900344e-05,
"loss": 0.2237,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25270116329193115,
"step": 1180,
"valid_targets_mean": 4869.7,
"valid_targets_min": 1656
},
{
"epoch": 1.8990384615384617,
"grad_norm": 0.5239563528247507,
"learning_rate": 3.654062316377106e-05,
"loss": 0.2233,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20021352171897888,
"step": 1185,
"valid_targets_mean": 3814.4,
"valid_targets_min": 263
},
{
"epoch": 1.907051282051282,
"grad_norm": 0.5185947912859171,
"learning_rate": 3.649556460847131e-05,
"loss": 0.2231,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24012042582035065,
"step": 1190,
"valid_targets_mean": 4489.9,
"valid_targets_min": 2456
},
{
"epoch": 1.9150641025641026,
"grad_norm": 0.4980920242835685,
"learning_rate": 3.6450242662570314e-05,
"loss": 0.2299,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22117306292057037,
"step": 1195,
"valid_targets_mean": 4400.8,
"valid_targets_min": 1103
},
{
"epoch": 1.9230769230769231,
"grad_norm": 0.5357392080956894,
"learning_rate": 3.6404658049739854e-05,
"loss": 0.218,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21440961956977844,
"step": 1200,
"valid_targets_mean": 5076.9,
"valid_targets_min": 944
},
{
"epoch": 1.9310897435897436,
"grad_norm": 0.4886343852173284,
"learning_rate": 3.63588114978458e-05,
"loss": 0.2109,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2078753262758255,
"step": 1205,
"valid_targets_mean": 4187.8,
"valid_targets_min": 329
},
{
"epoch": 1.939102564102564,
"grad_norm": 0.6584715935039488,
"learning_rate": 3.6312703738936504e-05,
"loss": 0.2159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21551740169525146,
"step": 1210,
"valid_targets_mean": 5163.4,
"valid_targets_min": 961
},
{
"epoch": 1.9471153846153846,
"grad_norm": 0.5239612749129127,
"learning_rate": 3.626633550923111e-05,
"loss": 0.2188,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23797452449798584,
"step": 1215,
"valid_targets_mean": 4640.9,
"valid_targets_min": 375
},
{
"epoch": 1.9551282051282053,
"grad_norm": 0.4578287517382224,
"learning_rate": 3.621970754910778e-05,
"loss": 0.2156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20246237516403198,
"step": 1220,
"valid_targets_mean": 4311.1,
"valid_targets_min": 247
},
{
"epoch": 1.9631410256410255,
"grad_norm": 0.4362064569919698,
"learning_rate": 3.6172820603091885e-05,
"loss": 0.2159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19485518336296082,
"step": 1225,
"valid_targets_mean": 3943.1,
"valid_targets_min": 655
},
{
"epoch": 1.9711538461538463,
"grad_norm": 0.5477456689898786,
"learning_rate": 3.612567541984413e-05,
"loss": 0.2212,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22714458405971527,
"step": 1230,
"valid_targets_mean": 3563.1,
"valid_targets_min": 305
},
{
"epoch": 1.9791666666666665,
"grad_norm": 0.6337487772707655,
"learning_rate": 3.6078272752148574e-05,
"loss": 0.2224,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23142507672309875,
"step": 1235,
"valid_targets_mean": 3947.9,
"valid_targets_min": 700
},
{
"epoch": 1.9871794871794872,
"grad_norm": 0.5506228796626095,
"learning_rate": 3.6030613356900635e-05,
"loss": 0.2162,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19356246292591095,
"step": 1240,
"valid_targets_mean": 4835.5,
"valid_targets_min": 369
},
{
"epoch": 1.9951923076923077,
"grad_norm": 0.4810642061863289,
"learning_rate": 3.598269799509498e-05,
"loss": 0.2173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2160474956035614,
"step": 1245,
"valid_targets_mean": 4615.5,
"valid_targets_min": 909
},
{
"epoch": 2.003205128205128,
"grad_norm": 0.44492553185973127,
"learning_rate": 3.5934527431813385e-05,
"loss": 0.221,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1990196406841278,
"step": 1250,
"valid_targets_mean": 4164.4,
"valid_targets_min": 749
},
{
"epoch": 2.011217948717949,
"grad_norm": 0.5141387129358165,
"learning_rate": 3.5886102436212536e-05,
"loss": 0.1896,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21212483942508698,
"step": 1255,
"valid_targets_mean": 4521.2,
"valid_targets_min": 859
},
{
"epoch": 2.019230769230769,
"grad_norm": 0.5510786539932679,
"learning_rate": 3.583742378151171e-05,
"loss": 0.1873,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20346041023731232,
"step": 1260,
"valid_targets_mean": 4208.8,
"valid_targets_min": 309
},
{
"epoch": 2.02724358974359,
"grad_norm": 0.5128664909981213,
"learning_rate": 3.5788492244980464e-05,
"loss": 0.2018,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22063224017620087,
"step": 1265,
"valid_targets_mean": 4432.4,
"valid_targets_min": 400
},
{
"epoch": 2.03525641025641,
"grad_norm": 0.539847477724801,
"learning_rate": 3.573930860792621e-05,
"loss": 0.2005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2066233903169632,
"step": 1270,
"valid_targets_mean": 3667.3,
"valid_targets_min": 290
},
{
"epoch": 2.043269230769231,
"grad_norm": 0.5012268282180441,
"learning_rate": 3.568987365568173e-05,
"loss": 0.2037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2341419756412506,
"step": 1275,
"valid_targets_mean": 4575.6,
"valid_targets_min": 775
},
{
"epoch": 2.051282051282051,
"grad_norm": 0.5138807861065794,
"learning_rate": 3.564018817759266e-05,
"loss": 0.1898,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1783386915922165,
"step": 1280,
"valid_targets_mean": 3789.6,
"valid_targets_min": 349
},
{
"epoch": 2.059294871794872,
"grad_norm": 0.5192918589194834,
"learning_rate": 3.559025296700484e-05,
"loss": 0.2079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2053699791431427,
"step": 1285,
"valid_targets_mean": 3863.3,
"valid_targets_min": 790
},
{
"epoch": 2.0673076923076925,
"grad_norm": 0.567060835301027,
"learning_rate": 3.554006882125173e-05,
"loss": 0.2093,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2147003710269928,
"step": 1290,
"valid_targets_mean": 4759.8,
"valid_targets_min": 372
},
{
"epoch": 2.0753205128205128,
"grad_norm": 0.41784862516611243,
"learning_rate": 3.5489636541641586e-05,
"loss": 0.194,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18648558855056763,
"step": 1295,
"valid_targets_mean": 5099.5,
"valid_targets_min": 2640
},
{
"epoch": 2.0833333333333335,
"grad_norm": 0.5366646727369517,
"learning_rate": 3.543895693344472e-05,
"loss": 0.2141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20949724316596985,
"step": 1300,
"valid_targets_mean": 4668.2,
"valid_targets_min": 580
},
{
"epoch": 2.0913461538461537,
"grad_norm": 0.490644372801002,
"learning_rate": 3.538803080588063e-05,
"loss": 0.2091,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19683998823165894,
"step": 1305,
"valid_targets_mean": 4263.7,
"valid_targets_min": 710
},
{
"epoch": 2.0993589743589745,
"grad_norm": 0.5304325970617975,
"learning_rate": 3.5336858972105076e-05,
"loss": 0.1953,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1905965805053711,
"step": 1310,
"valid_targets_mean": 3748.9,
"valid_targets_min": 248
},
{
"epoch": 2.1073717948717947,
"grad_norm": 1.001857869004124,
"learning_rate": 3.528544224919708e-05,
"loss": 0.2055,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1772008240222931,
"step": 1315,
"valid_targets_mean": 5202.2,
"valid_targets_min": 326
},
{
"epoch": 2.1153846153846154,
"grad_norm": 0.6474174557505488,
"learning_rate": 3.5233781458145934e-05,
"loss": 0.1956,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19679662585258484,
"step": 1320,
"valid_targets_mean": 4880.0,
"valid_targets_min": 492
},
{
"epoch": 2.123397435897436,
"grad_norm": 0.4798308643141641,
"learning_rate": 3.5181877423838034e-05,
"loss": 0.2035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2163936197757721,
"step": 1325,
"valid_targets_mean": 4552.6,
"valid_targets_min": 1234
},
{
"epoch": 2.1314102564102564,
"grad_norm": 0.4468567518223701,
"learning_rate": 3.512973097504371e-05,
"loss": 0.2046,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19748911261558533,
"step": 1330,
"valid_targets_mean": 4543.7,
"valid_targets_min": 1775
},
{
"epoch": 2.139423076923077,
"grad_norm": 0.4605576275826261,
"learning_rate": 3.507734294440403e-05,
"loss": 0.1938,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19503527879714966,
"step": 1335,
"valid_targets_mean": 4819.1,
"valid_targets_min": 1328
},
{
"epoch": 2.1474358974358974,
"grad_norm": 0.5004694302157144,
"learning_rate": 3.50247141684175e-05,
"loss": 0.2003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18384021520614624,
"step": 1340,
"valid_targets_mean": 4215.6,
"valid_targets_min": 254
},
{
"epoch": 2.155448717948718,
"grad_norm": 0.484154975475951,
"learning_rate": 3.497184548742667e-05,
"loss": 0.1951,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18417516350746155,
"step": 1345,
"valid_targets_mean": 4054.4,
"valid_targets_min": 540
},
{
"epoch": 2.1634615384615383,
"grad_norm": 0.47253236152427297,
"learning_rate": 3.491873774560473e-05,
"loss": 0.1965,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19266608357429504,
"step": 1350,
"valid_targets_mean": 4708.6,
"valid_targets_min": 384
},
{
"epoch": 2.171474358974359,
"grad_norm": 0.5560736423220788,
"learning_rate": 3.486539179094208e-05,
"loss": 0.2274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22012293338775635,
"step": 1355,
"valid_targets_mean": 4776.8,
"valid_targets_min": 486
},
{
"epoch": 2.1794871794871793,
"grad_norm": 0.5656873903519819,
"learning_rate": 3.481180847523272e-05,
"loss": 0.2076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2384198009967804,
"step": 1360,
"valid_targets_mean": 4216.3,
"valid_targets_min": 331
},
{
"epoch": 2.1875,
"grad_norm": 0.46967006856533894,
"learning_rate": 3.4757988654060684e-05,
"loss": 0.1929,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18673837184906006,
"step": 1365,
"valid_targets_mean": 5035.9,
"valid_targets_min": 241
},
{
"epoch": 2.1955128205128207,
"grad_norm": 0.48951532813213844,
"learning_rate": 3.470393318678637e-05,
"loss": 0.1906,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21696701645851135,
"step": 1370,
"valid_targets_mean": 4745.4,
"valid_targets_min": 2575
},
{
"epoch": 2.203525641025641,
"grad_norm": 0.47640892100071497,
"learning_rate": 3.4649642936532836e-05,
"loss": 0.1963,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23376938700675964,
"step": 1375,
"valid_targets_mean": 4782.6,
"valid_targets_min": 871
},
{
"epoch": 2.2115384615384617,
"grad_norm": 0.4725006121778969,
"learning_rate": 3.4595118770171984e-05,
"loss": 0.2064,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2107929289340973,
"step": 1380,
"valid_targets_mean": 4737.5,
"valid_targets_min": 1012
},
{
"epoch": 2.219551282051282,
"grad_norm": 0.48220097297869996,
"learning_rate": 3.454036155831077e-05,
"loss": 0.1986,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20837053656578064,
"step": 1385,
"valid_targets_mean": 4410.6,
"valid_targets_min": 2651
},
{
"epoch": 2.2275641025641026,
"grad_norm": 0.5301789866248495,
"learning_rate": 3.4485372175277236e-05,
"loss": 0.203,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21922306716442108,
"step": 1390,
"valid_targets_mean": 3674.9,
"valid_targets_min": 985
},
{
"epoch": 2.235576923076923,
"grad_norm": 0.5083703396017198,
"learning_rate": 3.44301514991066e-05,
"loss": 0.1988,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19266681373119354,
"step": 1395,
"valid_targets_mean": 4855.9,
"valid_targets_min": 347
},
{
"epoch": 2.2435897435897436,
"grad_norm": 0.7519899916330522,
"learning_rate": 3.4374700411527225e-05,
"loss": 0.1979,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19227102398872375,
"step": 1400,
"valid_targets_mean": 5229.4,
"valid_targets_min": 367
},
{
"epoch": 2.251602564102564,
"grad_norm": 0.452648853680781,
"learning_rate": 3.431901979794653e-05,
"loss": 0.2039,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21651104092597961,
"step": 1405,
"valid_targets_mean": 4710.3,
"valid_targets_min": 796
},
{
"epoch": 2.2596153846153846,
"grad_norm": 0.47484015313546707,
"learning_rate": 3.426311054743685e-05,
"loss": 0.2042,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17482003569602966,
"step": 1410,
"valid_targets_mean": 4051.4,
"valid_targets_min": 350
},
{
"epoch": 2.2676282051282053,
"grad_norm": 0.5214453271367003,
"learning_rate": 3.420697355272127e-05,
"loss": 0.1943,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19362860918045044,
"step": 1415,
"valid_targets_mean": 5048.5,
"valid_targets_min": 1767
},
{
"epoch": 2.2756410256410255,
"grad_norm": 0.4518578178740669,
"learning_rate": 3.415060971015933e-05,
"loss": 0.1945,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.192762553691864,
"step": 1420,
"valid_targets_mean": 5874.1,
"valid_targets_min": 2366
},
{
"epoch": 2.2836538461538463,
"grad_norm": 0.4924190267193352,
"learning_rate": 3.4094019919732736e-05,
"loss": 0.2028,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2125452756881714,
"step": 1425,
"valid_targets_mean": 4768.4,
"valid_targets_min": 626
},
{
"epoch": 2.2916666666666665,
"grad_norm": 0.5471322360692432,
"learning_rate": 3.403720508503098e-05,
"loss": 0.2001,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23145927488803864,
"step": 1430,
"valid_targets_mean": 3752.8,
"valid_targets_min": 375
},
{
"epoch": 2.2996794871794872,
"grad_norm": 0.5877416566678065,
"learning_rate": 3.398016611323693e-05,
"loss": 0.2049,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20787060260772705,
"step": 1435,
"valid_targets_mean": 3692.1,
"valid_targets_min": 276
},
{
"epoch": 2.3076923076923075,
"grad_norm": 0.44006152633664714,
"learning_rate": 3.392290391511232e-05,
"loss": 0.2001,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19138392806053162,
"step": 1440,
"valid_targets_mean": 4439.2,
"valid_targets_min": 248
},
{
"epoch": 2.315705128205128,
"grad_norm": 0.4783400983594827,
"learning_rate": 3.386541940498322e-05,
"loss": 0.1954,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2313036024570465,
"step": 1445,
"valid_targets_mean": 4953.0,
"valid_targets_min": 279
},
{
"epoch": 2.323717948717949,
"grad_norm": 0.6300946546742549,
"learning_rate": 3.380771350072543e-05,
"loss": 0.197,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19289648532867432,
"step": 1450,
"valid_targets_mean": 4243.2,
"valid_targets_min": 2154
},
{
"epoch": 2.331730769230769,
"grad_norm": 0.5198671715818997,
"learning_rate": 3.374978712374986e-05,
"loss": 0.2146,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20773255825042725,
"step": 1455,
"valid_targets_mean": 3954.9,
"valid_targets_min": 581
},
{
"epoch": 2.33974358974359,
"grad_norm": 0.4814095866682044,
"learning_rate": 3.369164119898774e-05,
"loss": 0.2007,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2113143354654312,
"step": 1460,
"valid_targets_mean": 4341.0,
"valid_targets_min": 292
},
{
"epoch": 2.34775641025641,
"grad_norm": 0.41904367944641874,
"learning_rate": 3.363327665487593e-05,
"loss": 0.2065,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19306056201457977,
"step": 1465,
"valid_targets_mean": 4922.9,
"valid_targets_min": 1553
},
{
"epoch": 2.355769230769231,
"grad_norm": 0.6146893741741665,
"learning_rate": 3.357469442334206e-05,
"loss": 0.1989,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23453611135482788,
"step": 1470,
"valid_targets_mean": 4356.6,
"valid_targets_min": 407
},
{
"epoch": 2.363782051282051,
"grad_norm": 0.6257652187776819,
"learning_rate": 3.351589543978965e-05,
"loss": 0.2175,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1822822391986847,
"step": 1475,
"valid_targets_mean": 4928.3,
"valid_targets_min": 1870
},
{
"epoch": 2.371794871794872,
"grad_norm": 0.8474243335881568,
"learning_rate": 3.345688064308317e-05,
"loss": 0.2055,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18252535164356232,
"step": 1480,
"valid_targets_mean": 4439.1,
"valid_targets_min": 2665
},
{
"epoch": 2.3798076923076925,
"grad_norm": 0.5183229192698324,
"learning_rate": 3.339765097553307e-05,
"loss": 0.1961,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.217352032661438,
"step": 1485,
"valid_targets_mean": 4219.8,
"valid_targets_min": 766
},
{
"epoch": 2.3878205128205128,
"grad_norm": 0.4645772914411723,
"learning_rate": 3.33382073828807e-05,
"loss": 0.2093,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2134999930858612,
"step": 1490,
"valid_targets_mean": 5478.5,
"valid_targets_min": 1406
},
{
"epoch": 2.3958333333333335,
"grad_norm": 0.522344084450664,
"learning_rate": 3.327855081428326e-05,
"loss": 0.1938,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2056310921907425,
"step": 1495,
"valid_targets_mean": 4581.1,
"valid_targets_min": 690
},
{
"epoch": 2.4038461538461537,
"grad_norm": 0.5132076146193778,
"learning_rate": 3.3218682222298584e-05,
"loss": 0.2334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2013227939605713,
"step": 1500,
"valid_targets_mean": 4036.2,
"valid_targets_min": 1818
},
{
"epoch": 2.4118589743589745,
"grad_norm": 0.4533972377397368,
"learning_rate": 3.315860256286996e-05,
"loss": 0.1998,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1921844184398651,
"step": 1505,
"valid_targets_mean": 4405.9,
"valid_targets_min": 1653
},
{
"epoch": 2.4198717948717947,
"grad_norm": 0.4506805941862736,
"learning_rate": 3.3098312795310894e-05,
"loss": 0.2047,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2224891632795334,
"step": 1510,
"valid_targets_mean": 4855.4,
"valid_targets_min": 303
},
{
"epoch": 2.4278846153846154,
"grad_norm": 0.4843786932758061,
"learning_rate": 3.303781388228974e-05,
"loss": 0.2073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20204125344753265,
"step": 1515,
"valid_targets_mean": 4418.8,
"valid_targets_min": 371
},
{
"epoch": 2.435897435897436,
"grad_norm": 0.6362361847499759,
"learning_rate": 3.297710678981435e-05,
"loss": 0.2004,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18480180203914642,
"step": 1520,
"valid_targets_mean": 3908.8,
"valid_targets_min": 270
},
{
"epoch": 2.4439102564102564,
"grad_norm": 0.5563620720646466,
"learning_rate": 3.291619248721667e-05,
"loss": 0.1977,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18212661147117615,
"step": 1525,
"valid_targets_mean": 5110.9,
"valid_targets_min": 1862
},
{
"epoch": 2.451923076923077,
"grad_norm": 0.46504803690447954,
"learning_rate": 3.285507194713724e-05,
"loss": 0.2105,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21177895367145538,
"step": 1530,
"valid_targets_mean": 4464.4,
"valid_targets_min": 1003
},
{
"epoch": 2.4599358974358974,
"grad_norm": 0.46677861968790446,
"learning_rate": 3.279374614550966e-05,
"loss": 0.2061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22254544496536255,
"step": 1535,
"valid_targets_mean": 4727.5,
"valid_targets_min": 641
},
{
"epoch": 2.467948717948718,
"grad_norm": 0.6095887301271112,
"learning_rate": 3.2732216061545e-05,
"loss": 0.1958,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2057630717754364,
"step": 1540,
"valid_targets_mean": 3947.6,
"valid_targets_min": 369
},
{
"epoch": 2.4759615384615383,
"grad_norm": 0.4825034140657324,
"learning_rate": 3.2670482677716214e-05,
"loss": 0.1894,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18248946964740753,
"step": 1545,
"valid_targets_mean": 4655.3,
"valid_targets_min": 901
},
{
"epoch": 2.483974358974359,
"grad_norm": 0.6294546717480383,
"learning_rate": 3.2608546979742394e-05,
"loss": 0.1968,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21113872528076172,
"step": 1550,
"valid_targets_mean": 3027.1,
"valid_targets_min": 447
},
{
"epoch": 2.4919871794871793,
"grad_norm": 0.48163671469345276,
"learning_rate": 3.254640995657307e-05,
"loss": 0.2001,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2081471085548401,
"step": 1555,
"valid_targets_mean": 4833.1,
"valid_targets_min": 2316
},
{
"epoch": 2.5,
"grad_norm": 0.45452918565110073,
"learning_rate": 3.248407260037239e-05,
"loss": 0.2047,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20107614994049072,
"step": 1560,
"valid_targets_mean": 4901.2,
"valid_targets_min": 241
},
{
"epoch": 2.5080128205128203,
"grad_norm": 0.5107599829528497,
"learning_rate": 3.24215359065033e-05,
"loss": 0.1931,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1827908456325531,
"step": 1565,
"valid_targets_mean": 4124.0,
"valid_targets_min": 313
},
{
"epoch": 2.516025641025641,
"grad_norm": 0.4957715906581253,
"learning_rate": 3.235880087351164e-05,
"loss": 0.192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20882809162139893,
"step": 1570,
"valid_targets_mean": 4480.7,
"valid_targets_min": 400
},
{
"epoch": 2.5240384615384617,
"grad_norm": 0.6297348431855959,
"learning_rate": 3.2295868503110184e-05,
"loss": 0.2068,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21451301872730255,
"step": 1575,
"valid_targets_mean": 4339.9,
"valid_targets_min": 582
},
{
"epoch": 2.532051282051282,
"grad_norm": 0.5250285602419786,
"learning_rate": 3.22327398001627e-05,
"loss": 0.1939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1985049694776535,
"step": 1580,
"valid_targets_mean": 3999.4,
"valid_targets_min": 465
},
{
"epoch": 2.5400641025641026,
"grad_norm": 0.6197279110456299,
"learning_rate": 3.216941577266783e-05,
"loss": 0.2,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17681679129600525,
"step": 1585,
"valid_targets_mean": 3882.1,
"valid_targets_min": 291
},
{
"epoch": 2.5480769230769234,
"grad_norm": 0.5005218329405368,
"learning_rate": 3.210589743174308e-05,
"loss": 0.2051,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20417039096355438,
"step": 1590,
"valid_targets_mean": 4629.4,
"valid_targets_min": 579
},
{
"epoch": 2.5560897435897436,
"grad_norm": 0.41387673400073555,
"learning_rate": 3.204218579160857e-05,
"loss": 0.2003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1898219734430313,
"step": 1595,
"valid_targets_mean": 5402.8,
"valid_targets_min": 408
},
{
"epoch": 2.564102564102564,
"grad_norm": 0.4477860994730678,
"learning_rate": 3.197828186957094e-05,
"loss": 0.204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19756175577640533,
"step": 1600,
"valid_targets_mean": 4912.2,
"valid_targets_min": 945
},
{
"epoch": 2.5721153846153846,
"grad_norm": 0.5074741624158274,
"learning_rate": 3.191418668600705e-05,
"loss": 0.1906,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18758270144462585,
"step": 1605,
"valid_targets_mean": 3498.4,
"valid_targets_min": 480
},
{
"epoch": 2.5801282051282053,
"grad_norm": 0.5278803984751291,
"learning_rate": 3.184990126434771e-05,
"loss": 0.199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19920486211776733,
"step": 1610,
"valid_targets_mean": 4639.4,
"valid_targets_min": 2395
},
{
"epoch": 2.5881410256410255,
"grad_norm": 0.48250117538926984,
"learning_rate": 3.178542663106131e-05,
"loss": 0.1958,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1962428092956543,
"step": 1615,
"valid_targets_mean": 4020.0,
"valid_targets_min": 562
},
{
"epoch": 2.5961538461538463,
"grad_norm": 0.5105928181844743,
"learning_rate": 3.172076381563748e-05,
"loss": 0.1891,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21197693049907684,
"step": 1620,
"valid_targets_mean": 4043.7,
"valid_targets_min": 315
},
{
"epoch": 2.6041666666666665,
"grad_norm": 0.5978117090859352,
"learning_rate": 3.165591385057058e-05,
"loss": 0.2019,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2030148208141327,
"step": 1625,
"valid_targets_mean": 4966.9,
"valid_targets_min": 1656
},
{
"epoch": 2.6121794871794872,
"grad_norm": 0.5237714448931861,
"learning_rate": 3.1590877771343316e-05,
"loss": 0.215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24029159545898438,
"step": 1630,
"valid_targets_mean": 3557.9,
"valid_targets_min": 321
},
{
"epoch": 2.6201923076923075,
"grad_norm": 0.5455679124875518,
"learning_rate": 3.152565661641008e-05,
"loss": 0.2011,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19591960310935974,
"step": 1635,
"valid_targets_mean": 4300.9,
"valid_targets_min": 481
},
{
"epoch": 2.628205128205128,
"grad_norm": 0.4823671291679502,
"learning_rate": 3.1460251427180474e-05,
"loss": 0.1991,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20895536243915558,
"step": 1640,
"valid_targets_mean": 4238.1,
"valid_targets_min": 544
},
{
"epoch": 2.636217948717949,
"grad_norm": 0.4852485400713249,
"learning_rate": 3.139466324800263e-05,
"loss": 0.2034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.215309739112854,
"step": 1645,
"valid_targets_mean": 4590.0,
"valid_targets_min": 525
},
{
"epoch": 2.644230769230769,
"grad_norm": 0.4486123654518733,
"learning_rate": 3.132889312614655e-05,
"loss": 0.2001,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18489600718021393,
"step": 1650,
"valid_targets_mean": 5024.6,
"valid_targets_min": 759
},
{
"epoch": 2.65224358974359,
"grad_norm": 0.5156233789736453,
"learning_rate": 3.126294211178737e-05,
"loss": 0.1969,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2023034691810608,
"step": 1655,
"valid_targets_mean": 3512.4,
"valid_targets_min": 466
},
{
"epoch": 2.66025641025641,
"grad_norm": 0.4923304919454224,
"learning_rate": 3.1196811257988634e-05,
"loss": 0.1936,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19222262501716614,
"step": 1660,
"valid_targets_mean": 4250.3,
"valid_targets_min": 390
},
{
"epoch": 2.668269230769231,
"grad_norm": 0.4537269393201906,
"learning_rate": 3.1130501620685394e-05,
"loss": 0.1932,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17200437188148499,
"step": 1665,
"valid_targets_mean": 4017.3,
"valid_targets_min": 355
},
{
"epoch": 2.676282051282051,
"grad_norm": 0.5076260750156713,
"learning_rate": 3.106401425866745e-05,
"loss": 0.2057,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20404556393623352,
"step": 1670,
"valid_targets_mean": 4047.0,
"valid_targets_min": 295
},
{
"epoch": 2.684294871794872,
"grad_norm": 0.4472894855328836,
"learning_rate": 3.099735023356236e-05,
"loss": 0.1886,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1803874969482422,
"step": 1675,
"valid_targets_mean": 4669.4,
"valid_targets_min": 2161
},
{
"epoch": 2.6923076923076925,
"grad_norm": 0.9149817914733609,
"learning_rate": 3.0930510609818564e-05,
"loss": 0.2008,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19703227281570435,
"step": 1680,
"valid_targets_mean": 4526.1,
"valid_targets_min": 1447
},
{
"epoch": 2.7003205128205128,
"grad_norm": 0.47014351127108706,
"learning_rate": 3.086349645468831e-05,
"loss": 0.2038,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20951519906520844,
"step": 1685,
"valid_targets_mean": 4676.2,
"valid_targets_min": 312
},
{
"epoch": 2.7083333333333335,
"grad_norm": 0.4535281628543726,
"learning_rate": 3.079630883821067e-05,
"loss": 0.1848,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18936920166015625,
"step": 1690,
"valid_targets_mean": 4599.2,
"valid_targets_min": 584
},
{
"epoch": 2.7163461538461537,
"grad_norm": 0.5092708917911855,
"learning_rate": 3.0728948833194436e-05,
"loss": 0.1985,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21155445277690887,
"step": 1695,
"valid_targets_mean": 3958.4,
"valid_targets_min": 896
},
{
"epoch": 2.7243589743589745,
"grad_norm": 0.47364433042950727,
"learning_rate": 3.066141751520099e-05,
"loss": 0.193,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17155927419662476,
"step": 1700,
"valid_targets_mean": 4239.2,
"valid_targets_min": 527
},
{
"epoch": 2.7323717948717947,
"grad_norm": 0.44121738080356515,
"learning_rate": 3.059371596252712e-05,
"loss": 0.2005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1881178915500641,
"step": 1705,
"valid_targets_mean": 4979.3,
"valid_targets_min": 2321
},
{
"epoch": 2.7403846153846154,
"grad_norm": 0.47560643592872354,
"learning_rate": 3.0525845256187834e-05,
"loss": 0.208,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19123315811157227,
"step": 1710,
"valid_targets_mean": 4571.5,
"valid_targets_min": 1059
},
{
"epoch": 2.748397435897436,
"grad_norm": 0.4911772550975567,
"learning_rate": 3.0457806479899044e-05,
"loss": 0.2042,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1972346305847168,
"step": 1715,
"valid_targets_mean": 4060.8,
"valid_targets_min": 872
},
{
"epoch": 2.7564102564102564,
"grad_norm": 0.4778067655857801,
"learning_rate": 3.0389600720060318e-05,
"loss": 0.1974,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1982056349515915,
"step": 1720,
"valid_targets_mean": 4400.5,
"valid_targets_min": 740
},
{
"epoch": 2.7644230769230766,
"grad_norm": 0.482307688231173,
"learning_rate": 3.0321229065737522e-05,
"loss": 0.2232,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22422638535499573,
"step": 1725,
"valid_targets_mean": 4993.1,
"valid_targets_min": 2232
},
{
"epoch": 2.7724358974358974,
"grad_norm": 0.5306877581321934,
"learning_rate": 3.0252692608645384e-05,
"loss": 0.1982,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20194926857948303,
"step": 1730,
"valid_targets_mean": 4105.0,
"valid_targets_min": 1682
},
{
"epoch": 2.780448717948718,
"grad_norm": 0.7544920211605098,
"learning_rate": 3.0183992443130127e-05,
"loss": 0.2098,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18357272446155548,
"step": 1735,
"valid_targets_mean": 4308.6,
"valid_targets_min": 677
},
{
"epoch": 2.7884615384615383,
"grad_norm": 0.4658726829170583,
"learning_rate": 3.011512966615195e-05,
"loss": 0.1924,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2031126320362091,
"step": 1740,
"valid_targets_mean": 4896.4,
"valid_targets_min": 1347
},
{
"epoch": 2.796474358974359,
"grad_norm": 0.4238238277153858,
"learning_rate": 3.0046105377267523e-05,
"loss": 0.2007,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18918165564537048,
"step": 1745,
"valid_targets_mean": 4677.1,
"valid_targets_min": 328
},
{
"epoch": 2.8044871794871797,
"grad_norm": 0.5787489225777784,
"learning_rate": 2.9976920678612456e-05,
"loss": 0.2007,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23846130073070526,
"step": 1750,
"valid_targets_mean": 3442.1,
"valid_targets_min": 284
},
{
"epoch": 2.8125,
"grad_norm": 0.5637457344048898,
"learning_rate": 2.9907576674883664e-05,
"loss": 0.2043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19609586894512177,
"step": 1755,
"valid_targets_mean": 3967.2,
"valid_targets_min": 310
},
{
"epoch": 2.8205128205128203,
"grad_norm": 0.49314550364391757,
"learning_rate": 2.983807447332174e-05,
"loss": 0.201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20431850850582123,
"step": 1760,
"valid_targets_mean": 4017.9,
"valid_targets_min": 727
},
{
"epoch": 2.828525641025641,
"grad_norm": 0.5374500796292538,
"learning_rate": 2.9768415183693293e-05,
"loss": 0.1893,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17844527959823608,
"step": 1765,
"valid_targets_mean": 3471.4,
"valid_targets_min": 407
},
{
"epoch": 2.8365384615384617,
"grad_norm": 0.3870122789668203,
"learning_rate": 2.9698599918273197e-05,
"loss": 0.192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1775786131620407,
"step": 1770,
"valid_targets_mean": 5527.9,
"valid_targets_min": 788
},
{
"epoch": 2.844551282051282,
"grad_norm": 0.4903955718452372,
"learning_rate": 2.962862979182686e-05,
"loss": 0.2057,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2087697982788086,
"step": 1775,
"valid_targets_mean": 4066.7,
"valid_targets_min": 770
},
{
"epoch": 2.8525641025641026,
"grad_norm": 0.47987648157772644,
"learning_rate": 2.95585059215924e-05,
"loss": 0.1968,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18527767062187195,
"step": 1780,
"valid_targets_mean": 4840.9,
"valid_targets_min": 1309
},
{
"epoch": 2.8605769230769234,
"grad_norm": 0.481288453073141,
"learning_rate": 2.948822942726284e-05,
"loss": 0.1925,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18782569468021393,
"step": 1785,
"valid_targets_mean": 4274.6,
"valid_targets_min": 1453
},
{
"epoch": 2.8685897435897436,
"grad_norm": 0.6691171019406131,
"learning_rate": 2.941780143096817e-05,
"loss": 0.1926,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19146007299423218,
"step": 1790,
"valid_targets_mean": 4212.2,
"valid_targets_min": 282
},
{
"epoch": 2.876602564102564,
"grad_norm": 0.43782695397948074,
"learning_rate": 2.9347223057257505e-05,
"loss": 0.19,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.197881281375885,
"step": 1795,
"valid_targets_mean": 5566.2,
"valid_targets_min": 807
},
{
"epoch": 2.8846153846153846,
"grad_norm": 0.46771216138431343,
"learning_rate": 2.927649543308106e-05,
"loss": 0.1949,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1730491816997528,
"step": 1800,
"valid_targets_mean": 4313.1,
"valid_targets_min": 272
},
{
"epoch": 2.8926282051282053,
"grad_norm": 0.47730654869309064,
"learning_rate": 2.9205619687772212e-05,
"loss": 0.2031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2260577231645584,
"step": 1805,
"valid_targets_mean": 5125.2,
"valid_targets_min": 1026
},
{
"epoch": 2.9006410256410255,
"grad_norm": 0.46576437637786394,
"learning_rate": 2.9134596953029413e-05,
"loss": 0.1938,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19503527879714966,
"step": 1810,
"valid_targets_mean": 4975.1,
"valid_targets_min": 708
},
{
"epoch": 2.9086538461538463,
"grad_norm": 0.4568472142259264,
"learning_rate": 2.9063428362898168e-05,
"loss": 0.204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19240906834602356,
"step": 1815,
"valid_targets_mean": 4380.3,
"valid_targets_min": 330
},
{
"epoch": 2.9166666666666665,
"grad_norm": 0.48777948304189217,
"learning_rate": 2.8992115053752905e-05,
"loss": 0.2014,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18208041787147522,
"step": 1820,
"valid_targets_mean": 4136.9,
"valid_targets_min": 420
},
{
"epoch": 2.9246794871794872,
"grad_norm": 0.45630759408991867,
"learning_rate": 2.8920658164278816e-05,
"loss": 0.1992,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1998547464609146,
"step": 1825,
"valid_targets_mean": 4282.7,
"valid_targets_min": 1636
},
{
"epoch": 2.9326923076923075,
"grad_norm": 0.5026743912960341,
"learning_rate": 2.884905883545373e-05,
"loss": 0.1938,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2103625386953354,
"step": 1830,
"valid_targets_mean": 4018.1,
"valid_targets_min": 639
},
{
"epoch": 2.940705128205128,
"grad_norm": 0.4367135823349995,
"learning_rate": 2.877731821052981e-05,
"loss": 0.2029,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19900712370872498,
"step": 1835,
"valid_targets_mean": 4795.9,
"valid_targets_min": 542
},
{
"epoch": 2.948717948717949,
"grad_norm": 0.4974598396652497,
"learning_rate": 2.8705437435015375e-05,
"loss": 0.1948,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20310930907726288,
"step": 1840,
"valid_targets_mean": 4015.1,
"valid_targets_min": 651
},
{
"epoch": 2.956730769230769,
"grad_norm": 0.49114003004248646,
"learning_rate": 2.8633417656656566e-05,
"loss": 0.1957,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21181732416152954,
"step": 1845,
"valid_targets_mean": 3890.6,
"valid_targets_min": 263
},
{
"epoch": 2.96474358974359,
"grad_norm": 0.47027614491460185,
"learning_rate": 2.8561260025419036e-05,
"loss": 0.1918,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1937972456216812,
"step": 1850,
"valid_targets_mean": 4605.6,
"valid_targets_min": 504
},
{
"epoch": 2.97275641025641,
"grad_norm": 0.48681679964363767,
"learning_rate": 2.8488965693469583e-05,
"loss": 0.1967,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1900751292705536,
"step": 1855,
"valid_targets_mean": 4273.2,
"valid_targets_min": 2539
},
{
"epoch": 2.980769230769231,
"grad_norm": 0.5215812687238324,
"learning_rate": 2.8416535815157763e-05,
"loss": 0.2023,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20885762572288513,
"step": 1860,
"valid_targets_mean": 3585.7,
"valid_targets_min": 729
},
{
"epoch": 2.988782051282051,
"grad_norm": 0.4685130334546498,
"learning_rate": 2.8343971546997434e-05,
"loss": 0.193,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1861869990825653,
"step": 1865,
"valid_targets_mean": 4655.8,
"valid_targets_min": 1111
},
{
"epoch": 2.996794871794872,
"grad_norm": 0.4484362144716279,
"learning_rate": 2.827127404764831e-05,
"loss": 0.1932,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18164370954036713,
"step": 1870,
"valid_targets_mean": 4304.1,
"valid_targets_min": 315
},
{
"epoch": 3.0048076923076925,
"grad_norm": 0.432871791881375,
"learning_rate": 2.8198444477897467e-05,
"loss": 0.1802,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15751376748085022,
"step": 1875,
"valid_targets_mean": 4685.1,
"valid_targets_min": 985
},
{
"epoch": 3.0128205128205128,
"grad_norm": 0.4982687650306561,
"learning_rate": 2.8125484000640787e-05,
"loss": 0.1807,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1677863597869873,
"step": 1880,
"valid_targets_mean": 4489.8,
"valid_targets_min": 533
},
{
"epoch": 3.0208333333333335,
"grad_norm": 0.5809335653606537,
"learning_rate": 2.8052393780864394e-05,
"loss": 0.1898,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18472197651863098,
"step": 1885,
"valid_targets_mean": 4630.1,
"valid_targets_min": 2670
},
{
"epoch": 3.0288461538461537,
"grad_norm": 0.5208265863271694,
"learning_rate": 2.797917498562607e-05,
"loss": 0.1808,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15783464908599854,
"step": 1890,
"valid_targets_mean": 3420.9,
"valid_targets_min": 361
},
{
"epoch": 3.0368589743589745,
"grad_norm": 0.46181514502901877,
"learning_rate": 2.7905828784036596e-05,
"loss": 0.1718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1412910521030426,
"step": 1895,
"valid_targets_mean": 3576.5,
"valid_targets_min": 206
},
{
"epoch": 3.0448717948717947,
"grad_norm": 0.4815134522135764,
"learning_rate": 2.78323563472411e-05,
"loss": 0.1819,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15473149716854095,
"step": 1900,
"valid_targets_mean": 4453.2,
"valid_targets_min": 414
},
{
"epoch": 3.0528846153846154,
"grad_norm": 0.7240034799252051,
"learning_rate": 2.7758758848400354e-05,
"loss": 0.1833,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19129765033721924,
"step": 1905,
"valid_targets_mean": 4162.4,
"valid_targets_min": 565
},
{
"epoch": 3.0608974358974357,
"grad_norm": 0.4495139312276376,
"learning_rate": 2.7685037462672043e-05,
"loss": 0.1804,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18564610183238983,
"step": 1910,
"valid_targets_mean": 4587.6,
"valid_targets_min": 390
},
{
"epoch": 3.0689102564102564,
"grad_norm": 0.4688702604827096,
"learning_rate": 2.7611193367191993e-05,
"loss": 0.1743,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15809181332588196,
"step": 1915,
"valid_targets_mean": 5043.8,
"valid_targets_min": 629
},
{
"epoch": 3.076923076923077,
"grad_norm": 0.5381119371437599,
"learning_rate": 2.7537227741055378e-05,
"loss": 0.1811,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16532793641090393,
"step": 1920,
"valid_targets_mean": 4441.4,
"valid_targets_min": 293
},
{
"epoch": 3.0849358974358974,
"grad_norm": 0.4699984420397094,
"learning_rate": 2.746314176529791e-05,
"loss": 0.1832,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18074962496757507,
"step": 1925,
"valid_targets_mean": 4407.8,
"valid_targets_min": 551
},
{
"epoch": 3.092948717948718,
"grad_norm": 0.5163485151382384,
"learning_rate": 2.7388936622876957e-05,
"loss": 0.1792,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17316068708896637,
"step": 1930,
"valid_targets_mean": 3434.5,
"valid_targets_min": 403
},
{
"epoch": 3.1009615384615383,
"grad_norm": 0.569423090994062,
"learning_rate": 2.7314613498652663e-05,
"loss": 0.1851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1830349862575531,
"step": 1935,
"valid_targets_mean": 4831.6,
"valid_targets_min": 547
},
{
"epoch": 3.108974358974359,
"grad_norm": 0.4616476103781575,
"learning_rate": 2.7240173579369025e-05,
"loss": 0.1757,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17320474982261658,
"step": 1940,
"valid_targets_mean": 5128.4,
"valid_targets_min": 326
},
{
"epoch": 3.1169871794871793,
"grad_norm": 0.4462047818156271,
"learning_rate": 2.7165618053634962e-05,
"loss": 0.1864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16832458972930908,
"step": 1945,
"valid_targets_mean": 4711.2,
"valid_targets_min": 1873
},
{
"epoch": 3.125,
"grad_norm": 0.5837323818264617,
"learning_rate": 2.7090948111905304e-05,
"loss": 0.1894,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1754237413406372,
"step": 1950,
"valid_targets_mean": 4189.6,
"valid_targets_min": 337
},
{
"epoch": 3.1330128205128207,
"grad_norm": 0.49543553208737817,
"learning_rate": 2.701616494646183e-05,
"loss": 0.1868,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16249918937683105,
"step": 1955,
"valid_targets_mean": 4795.8,
"valid_targets_min": 1541
},
{
"epoch": 3.141025641025641,
"grad_norm": 0.4712994737284464,
"learning_rate": 2.6941269751394174e-05,
"loss": 0.1829,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1716231405735016,
"step": 1960,
"valid_targets_mean": 4722.0,
"valid_targets_min": 363
},
{
"epoch": 3.1490384615384617,
"grad_norm": 0.4967616547921004,
"learning_rate": 2.686626372258081e-05,
"loss": 0.182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19097980856895447,
"step": 1965,
"valid_targets_mean": 3979.8,
"valid_targets_min": 305
},
{
"epoch": 3.157051282051282,
"grad_norm": 0.48157592708107455,
"learning_rate": 2.6791148057669913e-05,
"loss": 0.1878,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17434343695640564,
"step": 1970,
"valid_targets_mean": 4615.7,
"valid_targets_min": 737
},
{
"epoch": 3.1650641025641026,
"grad_norm": 0.4575261738059591,
"learning_rate": 2.671592395606027e-05,
"loss": 0.1704,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1718798130750656,
"step": 1975,
"valid_targets_mean": 5020.6,
"valid_targets_min": 1658
},
{
"epoch": 3.173076923076923,
"grad_norm": 0.5297605829706495,
"learning_rate": 2.6640592618882114e-05,
"loss": 0.1891,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19246713817119598,
"step": 1980,
"valid_targets_mean": 3990.8,
"valid_targets_min": 298
},
{
"epoch": 3.1810897435897436,
"grad_norm": 0.5118387904089157,
"learning_rate": 2.656515524897795e-05,
"loss": 0.1746,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17611244320869446,
"step": 1985,
"valid_targets_mean": 3957.2,
"valid_targets_min": 859
},
{
"epoch": 3.189102564102564,
"grad_norm": 0.544896219833975,
"learning_rate": 2.6489613050883343e-05,
"loss": 0.1855,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1879330277442932,
"step": 1990,
"valid_targets_mean": 3862.2,
"valid_targets_min": 461
},
{
"epoch": 3.1971153846153846,
"grad_norm": 0.4991017840460863,
"learning_rate": 2.6413967230807677e-05,
"loss": 0.1809,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18983492255210876,
"step": 1995,
"valid_targets_mean": 3961.9,
"valid_targets_min": 613
},
{
"epoch": 3.2051282051282053,
"grad_norm": 0.5220268877267916,
"learning_rate": 2.6338218996614924e-05,
"loss": 0.1716,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16558729112148285,
"step": 2000,
"valid_targets_mean": 3282.6,
"valid_targets_min": 567
},
{
"epoch": 3.2131410256410255,
"grad_norm": 0.7444745196457476,
"learning_rate": 2.6262369557804325e-05,
"loss": 0.1725,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16176973283290863,
"step": 2005,
"valid_targets_mean": 5180.2,
"valid_targets_min": 1741
},
{
"epoch": 3.2211538461538463,
"grad_norm": 0.49912953557162126,
"learning_rate": 2.6186420125491094e-05,
"loss": 0.1793,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1665007770061493,
"step": 2010,
"valid_targets_mean": 4667.9,
"valid_targets_min": 770
},
{
"epoch": 3.2291666666666665,
"grad_norm": 0.5229896914021072,
"learning_rate": 2.6110371912387083e-05,
"loss": 0.1875,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22359222173690796,
"step": 2015,
"valid_targets_mean": 4303.2,
"valid_targets_min": 1288
},
{
"epoch": 3.2371794871794872,
"grad_norm": 0.49108056640788716,
"learning_rate": 2.6034226132781407e-05,
"loss": 0.1909,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2078736275434494,
"step": 2020,
"valid_targets_mean": 4791.1,
"valid_targets_min": 310
},
{
"epoch": 3.2451923076923075,
"grad_norm": 0.45678445870345497,
"learning_rate": 2.5957984002521066e-05,
"loss": 0.1724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1990813910961151,
"step": 2025,
"valid_targets_mean": 5513.9,
"valid_targets_min": 567
},
{
"epoch": 3.253205128205128,
"grad_norm": 0.5199553061087674,
"learning_rate": 2.588164673899151e-05,
"loss": 0.1675,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16644896566867828,
"step": 2030,
"valid_targets_mean": 3913.5,
"valid_targets_min": 901
},
{
"epoch": 3.261217948717949,
"grad_norm": 0.5788450224828576,
"learning_rate": 2.580521556109724e-05,
"loss": 0.1848,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16697050631046295,
"step": 2035,
"valid_targets_mean": 4462.6,
"valid_targets_min": 402
},
{
"epoch": 3.269230769230769,
"grad_norm": 0.4836978257195129,
"learning_rate": 2.57286916892423e-05,
"loss": 0.1815,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20908261835575104,
"step": 2040,
"valid_targets_mean": 5737.8,
"valid_targets_min": 3679
},
{
"epoch": 3.27724358974359,
"grad_norm": 0.527274680304656,
"learning_rate": 2.5652076345310822e-05,
"loss": 0.1786,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17558740079402924,
"step": 2045,
"valid_targets_mean": 4797.7,
"valid_targets_min": 277
},
{
"epoch": 3.28525641025641,
"grad_norm": 0.6026389291636117,
"learning_rate": 2.5575370752647507e-05,
"loss": 0.1847,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.216181218624115,
"step": 2050,
"valid_targets_mean": 3742.6,
"valid_targets_min": 244
},
{
"epoch": 3.293269230769231,
"grad_norm": 0.47211601487922494,
"learning_rate": 2.5498576136038077e-05,
"loss": 0.1892,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1865502893924713,
"step": 2055,
"valid_targets_mean": 4354.8,
"valid_targets_min": 352
},
{
"epoch": 3.301282051282051,
"grad_norm": 0.5975804347108851,
"learning_rate": 2.542169372168976e-05,
"loss": 0.1812,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17157131433486938,
"step": 2060,
"valid_targets_mean": 3102.9,
"valid_targets_min": 467
},
{
"epoch": 3.309294871794872,
"grad_norm": 0.545998134102454,
"learning_rate": 2.5344724737211646e-05,
"loss": 0.1799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19290784001350403,
"step": 2065,
"valid_targets_mean": 4107.3,
"valid_targets_min": 301
},
{
"epoch": 3.3173076923076925,
"grad_norm": 0.46834505408927857,
"learning_rate": 2.5267670411595152e-05,
"loss": 0.1813,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18823888897895813,
"step": 2070,
"valid_targets_mean": 5001.2,
"valid_targets_min": 1520
},
{
"epoch": 3.3253205128205128,
"grad_norm": 0.5400145376446488,
"learning_rate": 2.5190531975194345e-05,
"loss": 0.1901,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20545172691345215,
"step": 2075,
"valid_targets_mean": 4006.7,
"valid_targets_min": 364
},
{
"epoch": 3.3333333333333335,
"grad_norm": 0.5977656061760097,
"learning_rate": 2.5113310659706322e-05,
"loss": 0.1932,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2138136327266693,
"step": 2080,
"valid_targets_mean": 3805.7,
"valid_targets_min": 407
},
{
"epoch": 3.3413461538461537,
"grad_norm": 0.4915972267646394,
"learning_rate": 2.5036007698151553e-05,
"loss": 0.1748,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16997626423835754,
"step": 2085,
"valid_targets_mean": 4340.3,
"valid_targets_min": 314
},
{
"epoch": 3.3493589743589745,
"grad_norm": 0.46927707415170167,
"learning_rate": 2.4958624324854185e-05,
"loss": 0.1696,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15596602857112885,
"step": 2090,
"valid_targets_mean": 3905.6,
"valid_targets_min": 422
},
{
"epoch": 3.3573717948717947,
"grad_norm": 0.5518962576762444,
"learning_rate": 2.4881161775422303e-05,
"loss": 0.1707,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17724674940109253,
"step": 2095,
"valid_targets_mean": 4476.1,
"valid_targets_min": 536
},
{
"epoch": 3.3653846153846154,
"grad_norm": 0.6625888269246488,
"learning_rate": 2.480362128672824e-05,
"loss": 0.1812,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1620676964521408,
"step": 2100,
"valid_targets_mean": 4751.4,
"valid_targets_min": 562
},
{
"epoch": 3.373397435897436,
"grad_norm": 0.46650625422637937,
"learning_rate": 2.4726004096888817e-05,
"loss": 0.1853,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19178065657615662,
"step": 2105,
"valid_targets_mean": 5943.2,
"valid_targets_min": 1437
},
{
"epoch": 3.3814102564102564,
"grad_norm": 0.5041057009538759,
"learning_rate": 2.4648311445245558e-05,
"loss": 0.1734,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1935567557811737,
"step": 2110,
"valid_targets_mean": 4855.3,
"valid_targets_min": 606
},
{
"epoch": 3.389423076923077,
"grad_norm": 0.446579748191051,
"learning_rate": 2.457054457234493e-05,
"loss": 0.1832,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16537244617938995,
"step": 2115,
"valid_targets_mean": 4717.0,
"valid_targets_min": 804
},
{
"epoch": 3.3974358974358974,
"grad_norm": 0.6567244698169011,
"learning_rate": 2.4492704719918497e-05,
"loss": 0.1816,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18737126886844635,
"step": 2120,
"valid_targets_mean": 4250.7,
"valid_targets_min": 327
},
{
"epoch": 3.405448717948718,
"grad_norm": 0.4292920454857177,
"learning_rate": 2.4414793130863134e-05,
"loss": 0.1624,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16413454711437225,
"step": 2125,
"valid_targets_mean": 4819.0,
"valid_targets_min": 900
},
{
"epoch": 3.4134615384615383,
"grad_norm": 0.42995155247351313,
"learning_rate": 2.433681104922114e-05,
"loss": 0.1944,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1734238862991333,
"step": 2130,
"valid_targets_mean": 5333.6,
"valid_targets_min": 1029
},
{
"epoch": 3.421474358974359,
"grad_norm": 0.4692495357172585,
"learning_rate": 2.4258759720160412e-05,
"loss": 0.1701,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19220373034477234,
"step": 2135,
"valid_targets_mean": 4707.7,
"valid_targets_min": 400
},
{
"epoch": 3.4294871794871793,
"grad_norm": 0.48856621530653443,
"learning_rate": 2.4180640389954534e-05,
"loss": 0.1848,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19048014283180237,
"step": 2140,
"valid_targets_mean": 4252.0,
"valid_targets_min": 942
},
{
"epoch": 3.4375,
"grad_norm": 0.5807991608049384,
"learning_rate": 2.4102454305962892e-05,
"loss": 0.1833,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1901319921016693,
"step": 2145,
"valid_targets_mean": 4204.7,
"valid_targets_min": 316
},
{
"epoch": 3.4455128205128207,
"grad_norm": 0.5403028422540316,
"learning_rate": 2.402420271661076e-05,
"loss": 0.1843,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18668532371520996,
"step": 2150,
"valid_targets_mean": 4427.4,
"valid_targets_min": 315
},
{
"epoch": 3.453525641025641,
"grad_norm": 0.4830075430205197,
"learning_rate": 2.3945886871369338e-05,
"loss": 0.1866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17052114009857178,
"step": 2155,
"valid_targets_mean": 4845.3,
"valid_targets_min": 373
},
{
"epoch": 3.4615384615384617,
"grad_norm": 0.46511991263678365,
"learning_rate": 2.3867508020735865e-05,
"loss": 0.1648,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16577394306659698,
"step": 2160,
"valid_targets_mean": 5203.9,
"valid_targets_min": 1895
},
{
"epoch": 3.469551282051282,
"grad_norm": 0.4366391339477963,
"learning_rate": 2.3789067416213568e-05,
"loss": 0.1819,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17512130737304688,
"step": 2165,
"valid_targets_mean": 5543.1,
"valid_targets_min": 1790
},
{
"epoch": 3.4775641025641026,
"grad_norm": 0.5377721156152593,
"learning_rate": 2.3710566310291733e-05,
"loss": 0.1713,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1526574194431305,
"step": 2170,
"valid_targets_mean": 4432.9,
"valid_targets_min": 1174
},
{
"epoch": 3.485576923076923,
"grad_norm": 0.5053170339079525,
"learning_rate": 2.36320059564257e-05,
"loss": 0.1856,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16273434460163116,
"step": 2175,
"valid_targets_mean": 3832.2,
"valid_targets_min": 775
},
{
"epoch": 3.4935897435897436,
"grad_norm": 0.5384598378577018,
"learning_rate": 2.3553387609016833e-05,
"loss": 0.1727,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18388324975967407,
"step": 2180,
"valid_targets_mean": 4589.6,
"valid_targets_min": 655
},
{
"epoch": 3.501602564102564,
"grad_norm": 0.4708209980462298,
"learning_rate": 2.347471252339252e-05,
"loss": 0.1863,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16591985523700714,
"step": 2185,
"valid_targets_mean": 4598.1,
"valid_targets_min": 2195
},
{
"epoch": 3.5096153846153846,
"grad_norm": 0.7426834042504328,
"learning_rate": 2.339598195578608e-05,
"loss": 0.1896,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22950224578380585,
"step": 2190,
"valid_targets_mean": 4666.9,
"valid_targets_min": 1636
},
{
"epoch": 3.5176282051282053,
"grad_norm": 0.48512702777522804,
"learning_rate": 2.3317197163316757e-05,
"loss": 0.1826,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21147963404655457,
"step": 2195,
"valid_targets_mean": 4816.8,
"valid_targets_min": 1018
},
{
"epoch": 3.5256410256410255,
"grad_norm": 0.5617877956310096,
"learning_rate": 2.3238359403969608e-05,
"loss": 0.1958,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21842700242996216,
"step": 2200,
"valid_targets_mean": 4281.9,
"valid_targets_min": 1328
},
{
"epoch": 3.5336538461538463,
"grad_norm": 0.5076850539925842,
"learning_rate": 2.315946993657543e-05,
"loss": 0.1726,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17994599044322968,
"step": 2205,
"valid_targets_mean": 3793.8,
"valid_targets_min": 279
},
{
"epoch": 3.5416666666666665,
"grad_norm": 0.466810006203134,
"learning_rate": 2.3080530020790673e-05,
"loss": 0.1752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18654845654964447,
"step": 2210,
"valid_targets_mean": 4783.2,
"valid_targets_min": 291
},
{
"epoch": 3.5496794871794872,
"grad_norm": 0.45446292558753937,
"learning_rate": 2.300154091707731e-05,
"loss": 0.1758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16973653435707092,
"step": 2215,
"valid_targets_mean": 4705.4,
"valid_targets_min": 1973
},
{
"epoch": 3.5576923076923075,
"grad_norm": 0.4949737313911632,
"learning_rate": 2.2922503886682706e-05,
"loss": 0.1771,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17169412970542908,
"step": 2220,
"valid_targets_mean": 4416.4,
"valid_targets_min": 318
},
{
"epoch": 3.565705128205128,
"grad_norm": 0.5267191493066402,
"learning_rate": 2.28434201916195e-05,
"loss": 0.1758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19544604420661926,
"step": 2225,
"valid_targets_mean": 4748.6,
"valid_targets_min": 804
},
{
"epoch": 3.573717948717949,
"grad_norm": 0.4823975173668488,
"learning_rate": 2.2764291094645446e-05,
"loss": 0.1725,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1826593279838562,
"step": 2230,
"valid_targets_mean": 4681.9,
"valid_targets_min": 672
},
{
"epoch": 3.581730769230769,
"grad_norm": 0.5072413918849578,
"learning_rate": 2.2685117859243223e-05,
"loss": 0.1795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17262235283851624,
"step": 2235,
"valid_targets_mean": 4077.1,
"valid_targets_min": 1175
},
{
"epoch": 3.58974358974359,
"grad_norm": 0.5702808405967386,
"learning_rate": 2.2605901749600312e-05,
"loss": 0.1771,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1898045539855957,
"step": 2240,
"valid_targets_mean": 3809.8,
"valid_targets_min": 1011
},
{
"epoch": 3.59775641025641,
"grad_norm": 0.49312205734475056,
"learning_rate": 2.2526644030588764e-05,
"loss": 0.1759,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17452372610569,
"step": 2245,
"valid_targets_mean": 3606.4,
"valid_targets_min": 714
},
{
"epoch": 3.605769230769231,
"grad_norm": 0.477566947797979,
"learning_rate": 2.2447345967745036e-05,
"loss": 0.1707,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18020571768283844,
"step": 2250,
"valid_targets_mean": 4380.3,
"valid_targets_min": 447
},
{
"epoch": 3.613782051282051,
"grad_norm": 0.49750178086286223,
"learning_rate": 2.2368008827249756e-05,
"loss": 0.1765,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1591065526008606,
"step": 2255,
"valid_targets_mean": 3981.6,
"valid_targets_min": 719
},
{
"epoch": 3.621794871794872,
"grad_norm": 0.4742607510220615,
"learning_rate": 2.228863387590752e-05,
"loss": 0.1727,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17601871490478516,
"step": 2260,
"valid_targets_mean": 4223.6,
"valid_targets_min": 639
},
{
"epoch": 3.6298076923076925,
"grad_norm": 0.6789824725869934,
"learning_rate": 2.2209222381126687e-05,
"loss": 0.1858,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20665745437145233,
"step": 2265,
"valid_targets_mean": 5236.2,
"valid_targets_min": 1059
},
{
"epoch": 3.6378205128205128,
"grad_norm": 0.45121336650640365,
"learning_rate": 2.212977561089908e-05,
"loss": 0.1775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1601269394159317,
"step": 2270,
"valid_targets_mean": 4738.1,
"valid_targets_min": 953
},
{
"epoch": 3.6458333333333335,
"grad_norm": 0.6270472198328433,
"learning_rate": 2.20502948337798e-05,
"loss": 0.174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1846921741962433,
"step": 2275,
"valid_targets_mean": 4177.4,
"valid_targets_min": 263
},
{
"epoch": 3.6538461538461537,
"grad_norm": 0.4821813201480846,
"learning_rate": 2.1970781318866953e-05,
"loss": 0.1692,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15647317469120026,
"step": 2280,
"valid_targets_mean": 5159.3,
"valid_targets_min": 468
},
{
"epoch": 3.6618589743589745,
"grad_norm": 0.536104982395625,
"learning_rate": 2.1891236335781363e-05,
"loss": 0.1999,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15523283183574677,
"step": 2285,
"valid_targets_mean": 4150.9,
"valid_targets_min": 420
},
{
"epoch": 3.6698717948717947,
"grad_norm": 0.48454892470305433,
"learning_rate": 2.1811661154646332e-05,
"loss": 0.1895,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1858227252960205,
"step": 2290,
"valid_targets_mean": 4461.4,
"valid_targets_min": 552
},
{
"epoch": 3.6778846153846154,
"grad_norm": 0.6521908432001564,
"learning_rate": 2.173205704606735e-05,
"loss": 0.181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1667114794254303,
"step": 2295,
"valid_targets_mean": 3821.4,
"valid_targets_min": 325
},
{
"epoch": 3.685897435897436,
"grad_norm": 0.5212736900902799,
"learning_rate": 2.1652425281111785e-05,
"loss": 0.1856,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.192377969622612,
"step": 2300,
"valid_targets_mean": 3940.4,
"valid_targets_min": 270
},
{
"epoch": 3.6939102564102564,
"grad_norm": 0.464091179210996,
"learning_rate": 2.1572767131288607e-05,
"loss": 0.1844,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17901402711868286,
"step": 2305,
"valid_targets_mean": 4782.1,
"valid_targets_min": 359
},
{
"epoch": 3.7019230769230766,
"grad_norm": 0.48937900923316263,
"learning_rate": 2.1493083868528095e-05,
"loss": 0.1824,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19445034861564636,
"step": 2310,
"valid_targets_mean": 4362.8,
"valid_targets_min": 242
},
{
"epoch": 3.7099358974358974,
"grad_norm": 0.5484852155939033,
"learning_rate": 2.141337676516151e-05,
"loss": 0.182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17409178614616394,
"step": 2315,
"valid_targets_mean": 3832.2,
"valid_targets_min": 369
},
{
"epoch": 3.717948717948718,
"grad_norm": 0.4914131321920882,
"learning_rate": 2.1333647093900772e-05,
"loss": 0.1893,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18725253641605377,
"step": 2320,
"valid_targets_mean": 4433.8,
"valid_targets_min": 329
},
{
"epoch": 3.7259615384615383,
"grad_norm": 0.49751968748487585,
"learning_rate": 2.1253896127818175e-05,
"loss": 0.173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16333390772342682,
"step": 2325,
"valid_targets_mean": 4464.2,
"valid_targets_min": 308
},
{
"epoch": 3.733974358974359,
"grad_norm": 0.4719919215752472,
"learning_rate": 2.1174125140326013e-05,
"loss": 0.1829,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20246420800685883,
"step": 2330,
"valid_targets_mean": 5023.8,
"valid_targets_min": 260
},
{
"epoch": 3.7419871794871797,
"grad_norm": 0.4213132842531635,
"learning_rate": 2.1094335405156277e-05,
"loss": 0.1661,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14584484696388245,
"step": 2335,
"valid_targets_mean": 4837.5,
"valid_targets_min": 377
},
{
"epoch": 3.75,
"grad_norm": 0.48964346201552494,
"learning_rate": 2.1014528196340316e-05,
"loss": 0.1675,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15814045071601868,
"step": 2340,
"valid_targets_mean": 4158.0,
"valid_targets_min": 369
},
{
"epoch": 3.7580128205128203,
"grad_norm": 0.5682661229174749,
"learning_rate": 2.093470478818847e-05,
"loss": 0.1838,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19296470284461975,
"step": 2345,
"valid_targets_mean": 4751.6,
"valid_targets_min": 257
},
{
"epoch": 3.766025641025641,
"grad_norm": 0.5126486537228652,
"learning_rate": 2.0854866455269756e-05,
"loss": 0.1867,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19226203858852386,
"step": 2350,
"valid_targets_mean": 4039.4,
"valid_targets_min": 278
},
{
"epoch": 3.7740384615384617,
"grad_norm": 0.5107515928876354,
"learning_rate": 2.0775014472391496e-05,
"loss": 0.1702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15904046595096588,
"step": 2355,
"valid_targets_mean": 3409.4,
"valid_targets_min": 342
},
{
"epoch": 3.782051282051282,
"grad_norm": 0.6170717641446949,
"learning_rate": 2.0695150114578958e-05,
"loss": 0.1803,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1688968539237976,
"step": 2360,
"valid_targets_mean": 4900.8,
"valid_targets_min": 676
},
{
"epoch": 3.7900641025641026,
"grad_norm": 0.5510100171818819,
"learning_rate": 2.061527465705502e-05,
"loss": 0.1859,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18843808770179749,
"step": 2365,
"valid_targets_mean": 3456.8,
"valid_targets_min": 299
},
{
"epoch": 3.7980769230769234,
"grad_norm": 0.5167212186335401,
"learning_rate": 2.0535389375219773e-05,
"loss": 0.1787,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19306384027004242,
"step": 2370,
"valid_targets_mean": 4672.0,
"valid_targets_min": 447
},
{
"epoch": 3.8060897435897436,
"grad_norm": 0.5182182761001438,
"learning_rate": 2.045549554463019e-05,
"loss": 0.2006,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20985566079616547,
"step": 2375,
"valid_targets_mean": 4507.6,
"valid_targets_min": 1376
},
{
"epoch": 3.814102564102564,
"grad_norm": 0.4472978698135491,
"learning_rate": 2.0375594440979744e-05,
"loss": 0.1843,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17899033427238464,
"step": 2380,
"valid_targets_mean": 4987.6,
"valid_targets_min": 442
},
{
"epoch": 3.8221153846153846,
"grad_norm": 1.0280283195638396,
"learning_rate": 2.0295687340078037e-05,
"loss": 0.1857,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19626322388648987,
"step": 2385,
"valid_targets_mean": 4670.8,
"valid_targets_min": 705
},
{
"epoch": 3.8301282051282053,
"grad_norm": 0.4859354043955413,
"learning_rate": 2.0215775517830437e-05,
"loss": 0.1791,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1650046408176422,
"step": 2390,
"valid_targets_mean": 4451.2,
"valid_targets_min": 339
},
{
"epoch": 3.8381410256410255,
"grad_norm": 0.4541989606998763,
"learning_rate": 2.013586025021769e-05,
"loss": 0.1734,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1853248029947281,
"step": 2395,
"valid_targets_mean": 5229.6,
"valid_targets_min": 2083
},
{
"epoch": 3.8461538461538463,
"grad_norm": 0.4783189276714589,
"learning_rate": 2.0055942813275564e-05,
"loss": 0.1822,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20412176847457886,
"step": 2400,
"valid_targets_mean": 4903.4,
"valid_targets_min": 302
},
{
"epoch": 3.8541666666666665,
"grad_norm": 0.519475629785975,
"learning_rate": 1.9976024483074456e-05,
"loss": 0.1779,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16651460528373718,
"step": 2405,
"valid_targets_mean": 3584.3,
"valid_targets_min": 400
},
{
"epoch": 3.8621794871794872,
"grad_norm": 0.5257911360055391,
"learning_rate": 1.9896106535699025e-05,
"loss": 0.1819,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20910142362117767,
"step": 2410,
"valid_targets_mean": 4189.2,
"valid_targets_min": 939
},
{
"epoch": 3.8701923076923075,
"grad_norm": 0.49577646060120845,
"learning_rate": 1.9816190247227834e-05,
"loss": 0.1869,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19056084752082825,
"step": 2415,
"valid_targets_mean": 4303.1,
"valid_targets_min": 595
},
{
"epoch": 3.878205128205128,
"grad_norm": 0.48158937567078663,
"learning_rate": 1.9736276893712954e-05,
"loss": 0.1899,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18846407532691956,
"step": 2420,
"valid_targets_mean": 4915.5,
"valid_targets_min": 315
},
{
"epoch": 3.886217948717949,
"grad_norm": 0.45923546909130153,
"learning_rate": 1.9656367751159565e-05,
"loss": 0.1748,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16031868755817413,
"step": 2425,
"valid_targets_mean": 4236.9,
"valid_targets_min": 521
},
{
"epoch": 3.894230769230769,
"grad_norm": 0.5335889864225674,
"learning_rate": 1.957646409550565e-05,
"loss": 0.1903,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18086227774620056,
"step": 2430,
"valid_targets_mean": 4418.6,
"valid_targets_min": 1054
},
{
"epoch": 3.90224358974359,
"grad_norm": 0.44724006922714,
"learning_rate": 1.9496567202601545e-05,
"loss": 0.1771,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16453775763511658,
"step": 2435,
"valid_targets_mean": 4851.1,
"valid_targets_min": 948
},
{
"epoch": 3.91025641025641,
"grad_norm": 0.5730851103153727,
"learning_rate": 1.9416678348189627e-05,
"loss": 0.177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1829206496477127,
"step": 2440,
"valid_targets_mean": 3174.9,
"valid_targets_min": 308
},
{
"epoch": 3.918269230769231,
"grad_norm": 0.44487083041464287,
"learning_rate": 1.9336798807883907e-05,
"loss": 0.1856,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18125095963478088,
"step": 2445,
"valid_targets_mean": 4873.8,
"valid_targets_min": 342
},
{
"epoch": 3.926282051282051,
"grad_norm": 0.45290515642223383,
"learning_rate": 1.9256929857149686e-05,
"loss": 0.1867,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18533127009868622,
"step": 2450,
"valid_targets_mean": 5464.2,
"valid_targets_min": 3248
},
{
"epoch": 3.934294871794872,
"grad_norm": 0.46895364145179663,
"learning_rate": 1.9177072771283167e-05,
"loss": 0.1814,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17094582319259644,
"step": 2455,
"valid_targets_mean": 4383.3,
"valid_targets_min": 914
},
{
"epoch": 3.9423076923076925,
"grad_norm": 0.567104525557713,
"learning_rate": 1.9097228825391087e-05,
"loss": 0.185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19724735617637634,
"step": 2460,
"valid_targets_mean": 3731.2,
"valid_targets_min": 307
},
{
"epoch": 3.9503205128205128,
"grad_norm": 0.5005409537469364,
"learning_rate": 1.9017399294370413e-05,
"loss": 0.1949,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20221903920173645,
"step": 2465,
"valid_targets_mean": 4243.8,
"valid_targets_min": 859
},
{
"epoch": 3.9583333333333335,
"grad_norm": 0.4802288259424973,
"learning_rate": 1.893758545288791e-05,
"loss": 0.1737,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1752682328224182,
"step": 2470,
"valid_targets_mean": 4303.4,
"valid_targets_min": 355
},
{
"epoch": 3.9663461538461537,
"grad_norm": 0.49411499890561694,
"learning_rate": 1.8857788575359847e-05,
"loss": 0.1702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16602462530136108,
"step": 2475,
"valid_targets_mean": 3958.0,
"valid_targets_min": 704
},
{
"epoch": 3.9743589743589745,
"grad_norm": 0.5444640086912708,
"learning_rate": 1.87780099359316e-05,
"loss": 0.1907,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1942613124847412,
"step": 2480,
"valid_targets_mean": 4002.6,
"valid_targets_min": 247
},
{
"epoch": 3.9823717948717947,
"grad_norm": 0.4716709515190074,
"learning_rate": 1.869825080845734e-05,
"loss": 0.1769,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17910870909690857,
"step": 2485,
"valid_targets_mean": 4894.6,
"valid_targets_min": 372
},
{
"epoch": 3.9903846153846154,
"grad_norm": 0.5078299463470789,
"learning_rate": 1.8618512466479686e-05,
"loss": 0.1743,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17169637978076935,
"step": 2490,
"valid_targets_mean": 3555.8,
"valid_targets_min": 740
},
{
"epoch": 3.998397435897436,
"grad_norm": 0.4533322628939216,
"learning_rate": 1.8538796183209373e-05,
"loss": 0.1694,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16294938325881958,
"step": 2495,
"valid_targets_mean": 4419.5,
"valid_targets_min": 753
},
{
"epoch": 4.006410256410256,
"grad_norm": 0.4616051446634912,
"learning_rate": 1.845910323150491e-05,
"loss": 0.1665,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15906602144241333,
"step": 2500,
"valid_targets_mean": 4619.8,
"valid_targets_min": 971
},
{
"epoch": 4.014423076923077,
"grad_norm": 0.4710253956026651,
"learning_rate": 1.8379434883852255e-05,
"loss": 0.1628,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15674439072608948,
"step": 2505,
"valid_targets_mean": 4605.7,
"valid_targets_min": 421
},
{
"epoch": 4.022435897435898,
"grad_norm": 0.4475134646768596,
"learning_rate": 1.8299792412344524e-05,
"loss": 0.1569,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16481932997703552,
"step": 2510,
"valid_targets_mean": 5058.6,
"valid_targets_min": 1403
},
{
"epoch": 4.030448717948718,
"grad_norm": 0.5061051434972497,
"learning_rate": 1.8220177088661635e-05,
"loss": 0.1646,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16660577058792114,
"step": 2515,
"valid_targets_mean": 4046.9,
"valid_targets_min": 315
},
{
"epoch": 4.038461538461538,
"grad_norm": 0.5246503130237942,
"learning_rate": 1.814059018405004e-05,
"loss": 0.1647,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16199514269828796,
"step": 2520,
"valid_targets_mean": 4784.2,
"valid_targets_min": 279
},
{
"epoch": 4.046474358974359,
"grad_norm": 0.5356191839028323,
"learning_rate": 1.806103296930243e-05,
"loss": 0.17,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16863659024238586,
"step": 2525,
"valid_targets_mean": 4192.4,
"valid_targets_min": 502
},
{
"epoch": 4.05448717948718,
"grad_norm": 0.532429349635364,
"learning_rate": 1.7981506714737392e-05,
"loss": 0.1634,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1873507797718048,
"step": 2530,
"valid_targets_mean": 4682.2,
"valid_targets_min": 826
},
{
"epoch": 4.0625,
"grad_norm": 0.5140341570871965,
"learning_rate": 1.7902012690179188e-05,
"loss": 0.1541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13633522391319275,
"step": 2535,
"valid_targets_mean": 3657.9,
"valid_targets_min": 564
},
{
"epoch": 4.07051282051282,
"grad_norm": 0.5946396546810248,
"learning_rate": 1.7822552164937437e-05,
"loss": 0.1638,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1841537058353424,
"step": 2540,
"valid_targets_mean": 3993.6,
"valid_targets_min": 313
},
{
"epoch": 4.078525641025641,
"grad_norm": 0.6116212018925045,
"learning_rate": 1.7743126407786873e-05,
"loss": 0.1629,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18144303560256958,
"step": 2545,
"valid_targets_mean": 3568.0,
"valid_targets_min": 325
},
{
"epoch": 4.086538461538462,
"grad_norm": 0.4871191766999381,
"learning_rate": 1.766373668694707e-05,
"loss": 0.1561,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13382868468761444,
"step": 2550,
"valid_targets_mean": 4527.5,
"valid_targets_min": 1309
},
{
"epoch": 4.094551282051282,
"grad_norm": 0.5624675358720111,
"learning_rate": 1.7584384270062195e-05,
"loss": 0.1507,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14146824181079865,
"step": 2555,
"valid_targets_mean": 4255.1,
"valid_targets_min": 275
},
{
"epoch": 4.102564102564102,
"grad_norm": 0.6610949977381224,
"learning_rate": 1.7505070424180772e-05,
"loss": 0.1641,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18343916535377502,
"step": 2560,
"valid_targets_mean": 5185.1,
"valid_targets_min": 291
},
{
"epoch": 4.110576923076923,
"grad_norm": 0.5053616332013421,
"learning_rate": 1.7425796415735454e-05,
"loss": 0.1695,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16558396816253662,
"step": 2565,
"valid_targets_mean": 4672.1,
"valid_targets_min": 338
},
{
"epoch": 4.118589743589744,
"grad_norm": 0.6376023459688379,
"learning_rate": 1.7346563510522783e-05,
"loss": 0.1665,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19944004714488983,
"step": 2570,
"valid_targets_mean": 3546.6,
"valid_targets_min": 481
},
{
"epoch": 4.126602564102564,
"grad_norm": 0.48817072155559843,
"learning_rate": 1.7267372973682998e-05,
"loss": 0.1648,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15739847719669342,
"step": 2575,
"valid_targets_mean": 4519.3,
"valid_targets_min": 331
},
{
"epoch": 4.134615384615385,
"grad_norm": 0.5137497648831978,
"learning_rate": 1.7188226069679834e-05,
"loss": 0.1686,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1627301275730133,
"step": 2580,
"valid_targets_mean": 4443.8,
"valid_targets_min": 1847
},
{
"epoch": 4.142628205128205,
"grad_norm": 0.48896271459330987,
"learning_rate": 1.7109124062280307e-05,
"loss": 0.1603,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16580283641815186,
"step": 2585,
"valid_targets_mean": 5335.2,
"valid_targets_min": 3074
},
{
"epoch": 4.1506410256410255,
"grad_norm": 0.49943715572069164,
"learning_rate": 1.7030068214534567e-05,
"loss": 0.1683,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17405760288238525,
"step": 2590,
"valid_targets_mean": 4867.4,
"valid_targets_min": 1511
},
{
"epoch": 4.158653846153846,
"grad_norm": 0.5073969495513736,
"learning_rate": 1.695105978875572e-05,
"loss": 0.1613,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1704675704240799,
"step": 2595,
"valid_targets_mean": 4079.5,
"valid_targets_min": 818
},
{
"epoch": 4.166666666666667,
"grad_norm": 0.5380632952001784,
"learning_rate": 1.687210004649965e-05,
"loss": 0.1673,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17288878560066223,
"step": 2600,
"valid_targets_mean": 4840.8,
"valid_targets_min": 1050
},
{
"epoch": 4.174679487179487,
"grad_norm": 0.5188135731308313,
"learning_rate": 1.679319024854491e-05,
"loss": 0.1677,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14789575338363647,
"step": 2605,
"valid_targets_mean": 3799.6,
"valid_targets_min": 682
},
{
"epoch": 4.1826923076923075,
"grad_norm": 0.4845157899020588,
"learning_rate": 1.6714331654872564e-05,
"loss": 0.1563,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14494550228118896,
"step": 2610,
"valid_targets_mean": 4274.5,
"valid_targets_min": 615
},
{
"epoch": 4.190705128205128,
"grad_norm": 0.5015702063651578,
"learning_rate": 1.663552552464609e-05,
"loss": 0.1686,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20149420201778412,
"step": 2615,
"valid_targets_mean": 4455.5,
"valid_targets_min": 369
},
{
"epoch": 4.198717948717949,
"grad_norm": 0.529580289652153,
"learning_rate": 1.6556773116191257e-05,
"loss": 0.17,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15997329354286194,
"step": 2620,
"valid_targets_mean": 4087.1,
"valid_targets_min": 520
},
{
"epoch": 4.206730769230769,
"grad_norm": 0.4987637425946955,
"learning_rate": 1.647807568697603e-05,
"loss": 0.1589,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16673317551612854,
"step": 2625,
"valid_targets_mean": 4702.3,
"valid_targets_min": 1560
},
{
"epoch": 4.214743589743589,
"grad_norm": 0.5234398671686435,
"learning_rate": 1.6399434493590524e-05,
"loss": 0.1675,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17398235201835632,
"step": 2630,
"valid_targets_mean": 3758.5,
"valid_targets_min": 318
},
{
"epoch": 4.222756410256411,
"grad_norm": 0.4631849315487262,
"learning_rate": 1.6320850791726884e-05,
"loss": 0.1594,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16157153248786926,
"step": 2635,
"valid_targets_mean": 5235.2,
"valid_targets_min": 275
},
{
"epoch": 4.230769230769231,
"grad_norm": 0.5672934802083003,
"learning_rate": 1.6242325836159304e-05,
"loss": 0.1683,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1700442135334015,
"step": 2640,
"valid_targets_mean": 4639.9,
"valid_targets_min": 533
},
{
"epoch": 4.238782051282051,
"grad_norm": 0.4605935211501636,
"learning_rate": 1.6163860880723923e-05,
"loss": 0.1649,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.152215838432312,
"step": 2645,
"valid_targets_mean": 5314.2,
"valid_targets_min": 1148
},
{
"epoch": 4.246794871794872,
"grad_norm": 0.5304100909181085,
"learning_rate": 1.6085457178298866e-05,
"loss": 0.1585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15569767355918884,
"step": 2650,
"valid_targets_mean": 3955.1,
"valid_targets_min": 330
},
{
"epoch": 4.2548076923076925,
"grad_norm": 0.4925124164425331,
"learning_rate": 1.6007115980784182e-05,
"loss": 0.1603,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16387639939785004,
"step": 2655,
"valid_targets_mean": 4616.6,
"valid_targets_min": 351
},
{
"epoch": 4.262820512820513,
"grad_norm": 0.5422773357903093,
"learning_rate": 1.592883853908188e-05,
"loss": 0.1599,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15925085544586182,
"step": 2660,
"valid_targets_mean": 4170.6,
"valid_targets_min": 578
},
{
"epoch": 4.270833333333333,
"grad_norm": 0.5175884973410019,
"learning_rate": 1.585062610307599e-05,
"loss": 0.1627,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16652101278305054,
"step": 2665,
"valid_targets_mean": 4090.0,
"valid_targets_min": 345
},
{
"epoch": 4.278846153846154,
"grad_norm": 0.514815039494445,
"learning_rate": 1.5772479921612543e-05,
"loss": 0.1577,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1476632058620453,
"step": 2670,
"valid_targets_mean": 4571.9,
"valid_targets_min": 1460
},
{
"epoch": 4.2868589743589745,
"grad_norm": 0.4729061874382982,
"learning_rate": 1.5694401242479677e-05,
"loss": 0.1556,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17110750079154968,
"step": 2675,
"valid_targets_mean": 4617.8,
"valid_targets_min": 302
},
{
"epoch": 4.294871794871795,
"grad_norm": 0.47983614643304373,
"learning_rate": 1.5616391312387683e-05,
"loss": 0.1617,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17012767493724823,
"step": 2680,
"valid_targets_mean": 4972.3,
"valid_targets_min": 1081
},
{
"epoch": 4.302884615384615,
"grad_norm": 0.5900398045525906,
"learning_rate": 1.5538451376949106e-05,
"loss": 0.1791,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19347208738327026,
"step": 2685,
"valid_targets_mean": 3360.1,
"valid_targets_min": 600
},
{
"epoch": 4.310897435897436,
"grad_norm": 0.4988276519939084,
"learning_rate": 1.5460582680658888e-05,
"loss": 0.1583,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15606293082237244,
"step": 2690,
"valid_targets_mean": 4457.4,
"valid_targets_min": 312
},
{
"epoch": 4.318910256410256,
"grad_norm": 0.4801685572727678,
"learning_rate": 1.5382786466874446e-05,
"loss": 0.1648,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14603829383850098,
"step": 2695,
"valid_targets_mean": 4662.8,
"valid_targets_min": 621
},
{
"epoch": 4.326923076923077,
"grad_norm": 0.5092745697416226,
"learning_rate": 1.5305063977795856e-05,
"loss": 0.168,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15981602668762207,
"step": 2700,
"valid_targets_mean": 4408.5,
"valid_targets_min": 337
},
{
"epoch": 4.334935897435898,
"grad_norm": 0.4719408239108926,
"learning_rate": 1.5227416454445995e-05,
"loss": 0.1702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18617725372314453,
"step": 2705,
"valid_targets_mean": 4869.6,
"valid_targets_min": 2399
},
{
"epoch": 4.342948717948718,
"grad_norm": 0.4953069472625189,
"learning_rate": 1.5149845136650748e-05,
"loss": 0.1558,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14147809147834778,
"step": 2710,
"valid_targets_mean": 4625.7,
"valid_targets_min": 292
},
{
"epoch": 4.350961538461538,
"grad_norm": 0.5316899843184331,
"learning_rate": 1.5072351263019177e-05,
"loss": 0.1586,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16226962208747864,
"step": 2715,
"valid_targets_mean": 4195.4,
"valid_targets_min": 401
},
{
"epoch": 4.358974358974359,
"grad_norm": 0.5064292174513952,
"learning_rate": 1.4994936070923784e-05,
"loss": 0.1692,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2049139440059662,
"step": 2720,
"valid_targets_mean": 5046.2,
"valid_targets_min": 1461
},
{
"epoch": 4.36698717948718,
"grad_norm": 0.5268832681196608,
"learning_rate": 1.4917600796480745e-05,
"loss": 0.1636,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17457829415798187,
"step": 2725,
"valid_targets_mean": 4273.8,
"valid_targets_min": 1504
},
{
"epoch": 4.375,
"grad_norm": 0.5316672633509928,
"learning_rate": 1.4840346674530122e-05,
"loss": 0.1697,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17298303544521332,
"step": 2730,
"valid_targets_mean": 3933.4,
"valid_targets_min": 271
},
{
"epoch": 4.38301282051282,
"grad_norm": 0.539251520668442,
"learning_rate": 1.4763174938616232e-05,
"loss": 0.156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15699277818202972,
"step": 2735,
"valid_targets_mean": 4498.7,
"valid_targets_min": 1500
},
{
"epoch": 4.391025641025641,
"grad_norm": 0.47337427219021855,
"learning_rate": 1.4686086820967865e-05,
"loss": 0.1621,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15189063549041748,
"step": 2740,
"valid_targets_mean": 4710.5,
"valid_targets_min": 355
},
{
"epoch": 4.399038461538462,
"grad_norm": 0.577115269190156,
"learning_rate": 1.460908355247868e-05,
"loss": 0.156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15899528563022614,
"step": 2745,
"valid_targets_mean": 4288.1,
"valid_targets_min": 1867
},
{
"epoch": 4.407051282051282,
"grad_norm": 0.5629175225975437,
"learning_rate": 1.4532166362687507e-05,
"loss": 0.1723,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19520245492458344,
"step": 2750,
"valid_targets_mean": 4262.2,
"valid_targets_min": 328
},
{
"epoch": 4.415064102564102,
"grad_norm": 0.47052408378056876,
"learning_rate": 1.445533647975871e-05,
"loss": 0.1548,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14956176280975342,
"step": 2755,
"valid_targets_mean": 4230.2,
"valid_targets_min": 254
},
{
"epoch": 4.423076923076923,
"grad_norm": 0.5292543787006989,
"learning_rate": 1.437859513046263e-05,
"loss": 0.1687,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1721234917640686,
"step": 2760,
"valid_targets_mean": 4353.6,
"valid_targets_min": 660
},
{
"epoch": 4.431089743589744,
"grad_norm": 0.4808265395153408,
"learning_rate": 1.4301943540155914e-05,
"loss": 0.1581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16927722096443176,
"step": 2765,
"valid_targets_mean": 5056.1,
"valid_targets_min": 1437
},
{
"epoch": 4.439102564102564,
"grad_norm": 0.5726650998016874,
"learning_rate": 1.4225382932762033e-05,
"loss": 0.1645,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15824058651924133,
"step": 2770,
"valid_targets_mean": 3288.9,
"valid_targets_min": 349
},
{
"epoch": 4.447115384615385,
"grad_norm": 0.44969969631353335,
"learning_rate": 1.4148914530751681e-05,
"loss": 0.1663,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15940533578395844,
"step": 2775,
"valid_targets_mean": 5797.3,
"valid_targets_min": 3434
},
{
"epoch": 4.455128205128205,
"grad_norm": 0.49741975836768354,
"learning_rate": 1.4072539555123292e-05,
"loss": 0.1631,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16189683973789215,
"step": 2780,
"valid_targets_mean": 4238.9,
"valid_targets_min": 291
},
{
"epoch": 4.4631410256410255,
"grad_norm": 0.5494976681695789,
"learning_rate": 1.3996259225383514e-05,
"loss": 0.1584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14954495429992676,
"step": 2785,
"valid_targets_mean": 4269.6,
"valid_targets_min": 759
},
{
"epoch": 4.471153846153846,
"grad_norm": 0.48614623992317596,
"learning_rate": 1.3920074759527737e-05,
"loss": 0.1626,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15910828113555908,
"step": 2790,
"valid_targets_mean": 4590.2,
"valid_targets_min": 316
},
{
"epoch": 4.479166666666667,
"grad_norm": 0.4555536743685927,
"learning_rate": 1.3843987374020689e-05,
"loss": 0.1758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1782645285129547,
"step": 2795,
"valid_targets_mean": 5682.2,
"valid_targets_min": 2418
},
{
"epoch": 4.487179487179487,
"grad_norm": 0.4939829320446489,
"learning_rate": 1.376799828377696e-05,
"loss": 0.1526,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14899791777133942,
"step": 2800,
"valid_targets_mean": 4733.5,
"valid_targets_min": 976
},
{
"epoch": 4.4951923076923075,
"grad_norm": 0.5156690114128816,
"learning_rate": 1.3692108702141642e-05,
"loss": 0.1581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17382262647151947,
"step": 2805,
"valid_targets_mean": 5299.5,
"valid_targets_min": 2987
},
{
"epoch": 4.503205128205128,
"grad_norm": 0.594337874393995,
"learning_rate": 1.361631984087091e-05,
"loss": 0.1579,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15712302923202515,
"step": 2810,
"valid_targets_mean": 3587.2,
"valid_targets_min": 366
},
{
"epoch": 4.511217948717949,
"grad_norm": 0.5033350436107449,
"learning_rate": 1.354063291011273e-05,
"loss": 0.1602,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16081839799880981,
"step": 2815,
"valid_targets_mean": 4126.3,
"valid_targets_min": 416
},
{
"epoch": 4.519230769230769,
"grad_norm": 0.6020936145530169,
"learning_rate": 1.3465049118387486e-05,
"loss": 0.174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18222561478614807,
"step": 2820,
"valid_targets_mean": 4309.9,
"valid_targets_min": 572
},
{
"epoch": 4.527243589743589,
"grad_norm": 0.5183615754616125,
"learning_rate": 1.3389569672568707e-05,
"loss": 0.1721,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14816981554031372,
"step": 2825,
"valid_targets_mean": 5247.9,
"valid_targets_min": 1944
},
{
"epoch": 4.535256410256411,
"grad_norm": 0.5887404633737963,
"learning_rate": 1.331419577786381e-05,
"loss": 0.1808,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18965160846710205,
"step": 2830,
"valid_targets_mean": 4065.2,
"valid_targets_min": 282
},
{
"epoch": 4.543269230769231,
"grad_norm": 0.5586143524883943,
"learning_rate": 1.3238928637794816e-05,
"loss": 0.1631,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13714498281478882,
"step": 2835,
"valid_targets_mean": 4856.2,
"valid_targets_min": 604
},
{
"epoch": 4.551282051282051,
"grad_norm": 0.5501511440044995,
"learning_rate": 1.3163769454179183e-05,
"loss": 0.1587,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1634409874677658,
"step": 2840,
"valid_targets_mean": 3328.1,
"valid_targets_min": 386
},
{
"epoch": 4.559294871794872,
"grad_norm": 0.5887621490305616,
"learning_rate": 1.3088719427110552e-05,
"loss": 0.1808,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22348913550376892,
"step": 2845,
"valid_targets_mean": 4610.0,
"valid_targets_min": 362
},
{
"epoch": 4.5673076923076925,
"grad_norm": 0.49930142538677474,
"learning_rate": 1.3013779754939666e-05,
"loss": 0.1592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15172190964221954,
"step": 2850,
"valid_targets_mean": 4027.2,
"valid_targets_min": 450
},
{
"epoch": 4.575320512820513,
"grad_norm": 0.48743984423580505,
"learning_rate": 1.2938951634255164e-05,
"loss": 0.1692,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16398422420024872,
"step": 2855,
"valid_targets_mean": 4630.4,
"valid_targets_min": 787
},
{
"epoch": 4.583333333333333,
"grad_norm": 0.5674408533279952,
"learning_rate": 1.2864236259864495e-05,
"loss": 0.1746,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18284529447555542,
"step": 2860,
"valid_targets_mean": 3369.7,
"valid_targets_min": 326
},
{
"epoch": 4.591346153846154,
"grad_norm": 0.55651818091062,
"learning_rate": 1.2789634824774887e-05,
"loss": 0.1683,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15348376333713531,
"step": 2865,
"valid_targets_mean": 3940.6,
"valid_targets_min": 241
},
{
"epoch": 4.5993589743589745,
"grad_norm": 0.5825742654396743,
"learning_rate": 1.2715148520174206e-05,
"loss": 0.1718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17244982719421387,
"step": 2870,
"valid_targets_mean": 3691.9,
"valid_targets_min": 321
},
{
"epoch": 4.607371794871795,
"grad_norm": 0.5135084268953117,
"learning_rate": 1.2640778535412036e-05,
"loss": 0.1688,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16129691898822784,
"step": 2875,
"valid_targets_mean": 4357.0,
"valid_targets_min": 939
},
{
"epoch": 4.615384615384615,
"grad_norm": 0.5523073586760865,
"learning_rate": 1.2566526057980608e-05,
"loss": 0.163,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15679003298282623,
"step": 2880,
"valid_targets_mean": 4061.8,
"valid_targets_min": 850
},
{
"epoch": 4.623397435897436,
"grad_norm": 0.5250521879110595,
"learning_rate": 1.2492392273495879e-05,
"loss": 0.1646,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16140484809875488,
"step": 2885,
"valid_targets_mean": 3998.6,
"valid_targets_min": 600
},
{
"epoch": 4.631410256410256,
"grad_norm": 0.5296491323750893,
"learning_rate": 1.2418378365678612e-05,
"loss": 0.1517,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15377673506736755,
"step": 2890,
"valid_targets_mean": 4099.4,
"valid_targets_min": 214
},
{
"epoch": 4.639423076923077,
"grad_norm": 0.5751739853610986,
"learning_rate": 1.234448551633542e-05,
"loss": 0.1726,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18292954564094543,
"step": 2895,
"valid_targets_mean": 4731.6,
"valid_targets_min": 801
},
{
"epoch": 4.647435897435898,
"grad_norm": 0.4741401252221975,
"learning_rate": 1.2270714905339969e-05,
"loss": 0.1576,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1418485939502716,
"step": 2900,
"valid_targets_mean": 4557.6,
"valid_targets_min": 347
},
{
"epoch": 4.655448717948718,
"grad_norm": 0.4526800307047848,
"learning_rate": 1.2197067710614075e-05,
"loss": 0.1609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14657928049564362,
"step": 2905,
"valid_targets_mean": 4691.9,
"valid_targets_min": 314
},
{
"epoch": 4.663461538461538,
"grad_norm": 0.5871345652949138,
"learning_rate": 1.2123545108108943e-05,
"loss": 0.1598,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14322693645954132,
"step": 2910,
"valid_targets_mean": 4301.7,
"valid_targets_min": 858
},
{
"epoch": 4.671474358974359,
"grad_norm": 0.5369766340649303,
"learning_rate": 1.2050148271786348e-05,
"loss": 0.1647,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1703229695558548,
"step": 2915,
"valid_targets_mean": 4125.8,
"valid_targets_min": 431
},
{
"epoch": 4.67948717948718,
"grad_norm": 0.4635354546791655,
"learning_rate": 1.1976878373599928e-05,
"loss": 0.167,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17637383937835693,
"step": 2920,
"valid_targets_mean": 5448.9,
"valid_targets_min": 336
},
{
"epoch": 4.6875,
"grad_norm": 0.4350740367186194,
"learning_rate": 1.1903736583476441e-05,
"loss": 0.1775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14900848269462585,
"step": 2925,
"valid_targets_mean": 4991.1,
"valid_targets_min": 1751
},
{
"epoch": 4.69551282051282,
"grad_norm": 0.4894468743125967,
"learning_rate": 1.1830724069297106e-05,
"loss": 0.17,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1520778238773346,
"step": 2930,
"valid_targets_mean": 4512.1,
"valid_targets_min": 1011
},
{
"epoch": 4.703525641025641,
"grad_norm": 0.5464655254035435,
"learning_rate": 1.1757841996878957e-05,
"loss": 0.1611,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16754302382469177,
"step": 2935,
"valid_targets_mean": 4386.4,
"valid_targets_min": 277
},
{
"epoch": 4.711538461538462,
"grad_norm": 0.7045566101812637,
"learning_rate": 1.1685091529956187e-05,
"loss": 0.156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15815463662147522,
"step": 2940,
"valid_targets_mean": 4289.8,
"valid_targets_min": 447
},
{
"epoch": 4.719551282051282,
"grad_norm": 0.4722129291117032,
"learning_rate": 1.161247383016163e-05,
"loss": 0.164,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1460554301738739,
"step": 2945,
"valid_targets_mean": 4285.4,
"valid_targets_min": 406
},
{
"epoch": 4.727564102564102,
"grad_norm": 0.4949417798189962,
"learning_rate": 1.1539990057008166e-05,
"loss": 0.1662,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14685088396072388,
"step": 2950,
"valid_targets_mean": 4511.9,
"valid_targets_min": 850
},
{
"epoch": 4.735576923076923,
"grad_norm": 0.45146352160046904,
"learning_rate": 1.1467641367870198e-05,
"loss": 0.1678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15919725596904755,
"step": 2955,
"valid_targets_mean": 5230.2,
"valid_targets_min": 2388
},
{
"epoch": 4.743589743589744,
"grad_norm": 0.493056612872455,
"learning_rate": 1.1395428917965239e-05,
"loss": 0.1621,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1494591236114502,
"step": 2960,
"valid_targets_mean": 4480.7,
"valid_targets_min": 359
},
{
"epoch": 4.751602564102564,
"grad_norm": 0.5600120299632549,
"learning_rate": 1.1323353860335385e-05,
"loss": 0.1672,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1780986487865448,
"step": 2965,
"valid_targets_mean": 3990.8,
"valid_targets_min": 531
},
{
"epoch": 4.759615384615385,
"grad_norm": 0.43441709493515895,
"learning_rate": 1.1251417345828962e-05,
"loss": 0.1525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13580113649368286,
"step": 2970,
"valid_targets_mean": 4967.2,
"valid_targets_min": 286
},
{
"epoch": 4.767628205128205,
"grad_norm": 0.5731718333915521,
"learning_rate": 1.1179620523082107e-05,
"loss": 0.163,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1791650652885437,
"step": 2975,
"valid_targets_mean": 3753.9,
"valid_targets_min": 323
},
{
"epoch": 4.7756410256410255,
"grad_norm": 0.5073366871473234,
"learning_rate": 1.110796453850047e-05,
"loss": 0.1624,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15835461020469666,
"step": 2980,
"valid_targets_mean": 3953.3,
"valid_targets_min": 367
},
{
"epoch": 4.783653846153846,
"grad_norm": 0.5448235725326789,
"learning_rate": 1.1036450536240877e-05,
"loss": 0.1621,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1594708263874054,
"step": 2985,
"valid_targets_mean": 4184.6,
"valid_targets_min": 329
},
{
"epoch": 4.791666666666667,
"grad_norm": 0.5112990620786744,
"learning_rate": 1.0965079658193068e-05,
"loss": 0.162,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15943202376365662,
"step": 2990,
"valid_targets_mean": 4657.1,
"valid_targets_min": 721
},
{
"epoch": 4.799679487179487,
"grad_norm": 0.5452656536159444,
"learning_rate": 1.0893853043961475e-05,
"loss": 0.1631,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17918306589126587,
"step": 2995,
"valid_targets_mean": 4030.8,
"valid_targets_min": 308
},
{
"epoch": 4.8076923076923075,
"grad_norm": 0.5065220928275539,
"learning_rate": 1.0822771830847011e-05,
"loss": 0.1722,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15812425315380096,
"step": 3000,
"valid_targets_mean": 4248.4,
"valid_targets_min": 764
},
{
"epoch": 4.815705128205128,
"grad_norm": 0.512162623492872,
"learning_rate": 1.0751837153828926e-05,
"loss": 0.1683,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15615516901016235,
"step": 3005,
"valid_targets_mean": 3870.6,
"valid_targets_min": 257
},
{
"epoch": 4.823717948717949,
"grad_norm": 0.5381812456187758,
"learning_rate": 1.0681050145546666e-05,
"loss": 0.1627,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16726335883140564,
"step": 3010,
"valid_targets_mean": 4017.4,
"valid_targets_min": 726
},
{
"epoch": 4.831730769230769,
"grad_norm": 0.6445496146480065,
"learning_rate": 1.0610411936281801e-05,
"loss": 0.1581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1484619677066803,
"step": 3015,
"valid_targets_mean": 3612.6,
"valid_targets_min": 383
},
{
"epoch": 4.839743589743589,
"grad_norm": 0.5702030900204609,
"learning_rate": 1.0539923653939978e-05,
"loss": 0.1666,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18328678607940674,
"step": 3020,
"valid_targets_mean": 4482.2,
"valid_targets_min": 1817
},
{
"epoch": 4.847756410256411,
"grad_norm": 0.5008784624956089,
"learning_rate": 1.0469586424032903e-05,
"loss": 0.165,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15659628808498383,
"step": 3025,
"valid_targets_mean": 4508.0,
"valid_targets_min": 375
},
{
"epoch": 4.855769230769231,
"grad_norm": 0.5530899557457086,
"learning_rate": 1.0399401369660369e-05,
"loss": 0.1723,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15809404850006104,
"step": 3030,
"valid_targets_mean": 4167.8,
"valid_targets_min": 324
},
{
"epoch": 4.863782051282051,
"grad_norm": 0.516645985654954,
"learning_rate": 1.0329369611492334e-05,
"loss": 0.16,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18036146461963654,
"step": 3035,
"valid_targets_mean": 4836.8,
"valid_targets_min": 299
},
{
"epoch": 4.871794871794872,
"grad_norm": 0.5112506058910664,
"learning_rate": 1.0259492267751022e-05,
"loss": 0.1801,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17415016889572144,
"step": 3040,
"valid_targets_mean": 4220.0,
"valid_targets_min": 628
},
{
"epoch": 4.8798076923076925,
"grad_norm": 0.48779955052469653,
"learning_rate": 1.0189770454193052e-05,
"loss": 0.1609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17102153599262238,
"step": 3045,
"valid_targets_mean": 4791.1,
"valid_targets_min": 700
},
{
"epoch": 4.887820512820513,
"grad_norm": 0.5316391723419246,
"learning_rate": 1.0120205284091673e-05,
"loss": 0.1731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16392341256141663,
"step": 3050,
"valid_targets_mean": 3791.6,
"valid_targets_min": 297
},
{
"epoch": 4.895833333333333,
"grad_norm": 0.4978850535617155,
"learning_rate": 1.0050797868218907e-05,
"loss": 0.1725,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15801748633384705,
"step": 3055,
"valid_targets_mean": 4318.2,
"valid_targets_min": 345
},
{
"epoch": 4.903846153846154,
"grad_norm": 0.5519780914058718,
"learning_rate": 9.981549314827876e-06,
"loss": 0.1702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18745127320289612,
"step": 3060,
"valid_targets_mean": 4001.9,
"valid_targets_min": 566
},
{
"epoch": 4.9118589743589745,
"grad_norm": 0.5125865980897772,
"learning_rate": 9.912460729635097e-06,
"loss": 0.1662,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1603458821773529,
"step": 3065,
"valid_targets_mean": 4251.4,
"valid_targets_min": 248
},
{
"epoch": 4.919871794871795,
"grad_norm": 0.6477506853730828,
"learning_rate": 9.843533215802796e-06,
"loss": 0.1573,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17034149169921875,
"step": 3070,
"valid_targets_mean": 3570.2,
"valid_targets_min": 313
},
{
"epoch": 4.927884615384615,
"grad_norm": 0.609986938655745,
"learning_rate": 9.774767873921357e-06,
"loss": 0.1678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1743059754371643,
"step": 3075,
"valid_targets_mean": 3627.1,
"valid_targets_min": 307
},
{
"epoch": 4.935897435897436,
"grad_norm": 0.48465311874450473,
"learning_rate": 9.706165801991651e-06,
"loss": 0.1585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14823710918426514,
"step": 3080,
"valid_targets_mean": 4433.3,
"valid_targets_min": 544
},
{
"epoch": 4.943910256410256,
"grad_norm": 0.48714354138964694,
"learning_rate": 9.637728095407593e-06,
"loss": 0.1543,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16156205534934998,
"step": 3085,
"valid_targets_mean": 4500.5,
"valid_targets_min": 775
},
{
"epoch": 4.951923076923077,
"grad_norm": 0.49670826195331386,
"learning_rate": 9.56945584693861e-06,
"loss": 0.1756,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16287025809288025,
"step": 3090,
"valid_targets_mean": 4310.7,
"valid_targets_min": 349
},
{
"epoch": 4.959935897435898,
"grad_norm": 0.5652295836822062,
"learning_rate": 9.501350146712193e-06,
"loss": 0.1762,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18979312479496002,
"step": 3095,
"valid_targets_mean": 3869.1,
"valid_targets_min": 364
},
{
"epoch": 4.967948717948718,
"grad_norm": 0.5025036450764108,
"learning_rate": 9.433412082196527e-06,
"loss": 0.1587,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1535206437110901,
"step": 3100,
"valid_targets_mean": 5336.6,
"valid_targets_min": 2466
},
{
"epoch": 4.975961538461538,
"grad_norm": 0.6281243792714412,
"learning_rate": 9.365642738183044e-06,
"loss": 0.1693,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16985513269901276,
"step": 3105,
"valid_targets_mean": 3242.9,
"valid_targets_min": 409
},
{
"epoch": 4.983974358974359,
"grad_norm": 0.5085204053134876,
"learning_rate": 9.298043196769217e-06,
"loss": 0.1632,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18689662218093872,
"step": 3110,
"valid_targets_mean": 4230.7,
"valid_targets_min": 742
},
{
"epoch": 4.99198717948718,
"grad_norm": 0.5426966023363969,
"learning_rate": 9.230614537341167e-06,
"loss": 0.171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1662767231464386,
"step": 3115,
"valid_targets_mean": 3954.7,
"valid_targets_min": 726
},
{
"epoch": 5.0,
"grad_norm": 0.48507717102566583,
"learning_rate": 9.163357836556498e-06,
"loss": 0.1729,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17505590617656708,
"step": 3120,
"valid_targets_mean": 4747.4,
"valid_targets_min": 828
},
{
"epoch": 5.00801282051282,
"grad_norm": 0.541614590181428,
"learning_rate": 9.096274168327122e-06,
"loss": 0.1553,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17244088649749756,
"step": 3125,
"valid_targets_mean": 4650.6,
"valid_targets_min": 1198
},
{
"epoch": 5.016025641025641,
"grad_norm": 0.490165923718776,
"learning_rate": 9.029364603802017e-06,
"loss": 0.1465,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1379947066307068,
"step": 3130,
"valid_targets_mean": 4759.4,
"valid_targets_min": 369
},
{
"epoch": 5.024038461538462,
"grad_norm": 0.4945283605481377,
"learning_rate": 8.962630211350248e-06,
"loss": 0.1618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16132843494415283,
"step": 3135,
"valid_targets_mean": 5059.5,
"valid_targets_min": 2727
},
{
"epoch": 5.032051282051282,
"grad_norm": 0.5774557427950784,
"learning_rate": 8.89607205654378e-06,
"loss": 0.1562,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15898734331130981,
"step": 3140,
"valid_targets_mean": 5315.1,
"valid_targets_min": 2030
},
{
"epoch": 5.040064102564102,
"grad_norm": 0.496524355084867,
"learning_rate": 8.829691202140591e-06,
"loss": 0.1507,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1663457453250885,
"step": 3145,
"valid_targets_mean": 5069.9,
"valid_targets_min": 1902
},
{
"epoch": 5.048076923076923,
"grad_norm": 0.4769123283394398,
"learning_rate": 8.763488708067604e-06,
"loss": 0.1592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12692353129386902,
"step": 3150,
"valid_targets_mean": 4791.1,
"valid_targets_min": 584
},
{
"epoch": 5.056089743589744,
"grad_norm": 0.5648032914676866,
"learning_rate": 8.69746563140379e-06,
"loss": 0.1473,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1447281837463379,
"step": 3155,
"valid_targets_mean": 4017.8,
"valid_targets_min": 468
},
{
"epoch": 5.064102564102564,
"grad_norm": 0.4835731191010093,
"learning_rate": 8.631623026363331e-06,
"loss": 0.1568,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13092973828315735,
"step": 3160,
"valid_targets_mean": 4022.8,
"valid_targets_min": 804
},
{
"epoch": 5.072115384615385,
"grad_norm": 0.5281358905201415,
"learning_rate": 8.56596194427873e-06,
"loss": 0.1698,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14482533931732178,
"step": 3165,
"valid_targets_mean": 4089.5,
"valid_targets_min": 481
},
{
"epoch": 5.080128205128205,
"grad_norm": 0.5826067069371261,
"learning_rate": 8.500483433584054e-06,
"loss": 0.1604,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1394755095243454,
"step": 3170,
"valid_targets_mean": 3875.4,
"valid_targets_min": 278
},
{
"epoch": 5.0881410256410255,
"grad_norm": 0.5022855307411574,
"learning_rate": 8.435188539798187e-06,
"loss": 0.1523,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14776669442653656,
"step": 3175,
"valid_targets_mean": 4571.3,
"valid_targets_min": 1668
},
{
"epoch": 5.096153846153846,
"grad_norm": 0.5400450765734484,
"learning_rate": 8.370078305508136e-06,
"loss": 0.1501,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13629215955734253,
"step": 3180,
"valid_targets_mean": 3904.1,
"valid_targets_min": 365
},
{
"epoch": 5.104166666666667,
"grad_norm": 0.5924876493518854,
"learning_rate": 8.305153770352384e-06,
"loss": 0.1514,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15243536233901978,
"step": 3185,
"valid_targets_mean": 4082.4,
"valid_targets_min": 2034
},
{
"epoch": 5.112179487179487,
"grad_norm": 0.5920292524598976,
"learning_rate": 8.240415971004285e-06,
"loss": 0.1481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13546709716320038,
"step": 3190,
"valid_targets_mean": 4605.2,
"valid_targets_min": 895
},
{
"epoch": 5.1201923076923075,
"grad_norm": 0.5311433425596007,
"learning_rate": 8.175865941155525e-06,
"loss": 0.1464,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1275891810655594,
"step": 3195,
"valid_targets_mean": 4334.2,
"valid_targets_min": 1431
},
{
"epoch": 5.128205128205128,
"grad_norm": 0.5474838651429441,
"learning_rate": 8.111504711499598e-06,
"loss": 0.1595,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1427532434463501,
"step": 3200,
"valid_targets_mean": 4057.6,
"valid_targets_min": 727
},
{
"epoch": 5.136217948717949,
"grad_norm": 0.5282654860128789,
"learning_rate": 8.04733330971536e-06,
"loss": 0.1499,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14707237482070923,
"step": 3205,
"valid_targets_mean": 4233.0,
"valid_targets_min": 688
},
{
"epoch": 5.144230769230769,
"grad_norm": 0.7915017071999582,
"learning_rate": 7.983352760450618e-06,
"loss": 0.1536,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1558932065963745,
"step": 3210,
"valid_targets_mean": 4670.9,
"valid_targets_min": 257
},
{
"epoch": 5.152243589743589,
"grad_norm": 1.4919243745130015,
"learning_rate": 7.919564085305768e-06,
"loss": 0.1617,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1521928608417511,
"step": 3215,
"valid_targets_mean": 4375.9,
"valid_targets_min": 638
},
{
"epoch": 5.160256410256411,
"grad_norm": 0.523390692498547,
"learning_rate": 7.855968302817487e-06,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15986919403076172,
"step": 3220,
"valid_targets_mean": 4230.8,
"valid_targets_min": 244
},
{
"epoch": 5.168269230769231,
"grad_norm": 0.5099071327956128,
"learning_rate": 7.792566428442456e-06,
"loss": 0.1405,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1378299444913864,
"step": 3225,
"valid_targets_mean": 4137.8,
"valid_targets_min": 1500
},
{
"epoch": 5.176282051282051,
"grad_norm": 0.4836650713912454,
"learning_rate": 7.729359474541168e-06,
"loss": 0.1591,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13931715488433838,
"step": 3230,
"valid_targets_mean": 4412.4,
"valid_targets_min": 963
},
{
"epoch": 5.184294871794872,
"grad_norm": 0.5641916486797564,
"learning_rate": 7.666348450361737e-06,
"loss": 0.1503,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16095387935638428,
"step": 3235,
"valid_targets_mean": 4122.6,
"valid_targets_min": 799
},
{
"epoch": 5.1923076923076925,
"grad_norm": 0.6174454057400754,
"learning_rate": 7.60353436202381e-06,
"loss": 0.1516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15927666425704956,
"step": 3240,
"valid_targets_mean": 3910.2,
"valid_targets_min": 260
},
{
"epoch": 5.200320512820513,
"grad_norm": 0.5406036266792001,
"learning_rate": 7.540918212502479e-06,
"loss": 0.1662,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1532440483570099,
"step": 3245,
"valid_targets_mean": 4377.1,
"valid_targets_min": 704
},
{
"epoch": 5.208333333333333,
"grad_norm": 0.5010408207165512,
"learning_rate": 7.478501001612281e-06,
"loss": 0.1585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16027632355690002,
"step": 3250,
"valid_targets_mean": 5156.2,
"valid_targets_min": 731
},
{
"epoch": 5.216346153846154,
"grad_norm": 0.4806860831503525,
"learning_rate": 7.416283725991229e-06,
"loss": 0.1627,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14156237244606018,
"step": 3255,
"valid_targets_mean": 4912.4,
"valid_targets_min": 1574
},
{
"epoch": 5.2243589743589745,
"grad_norm": 0.5086086398888948,
"learning_rate": 7.354267379084896e-06,
"loss": 0.1564,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15885037183761597,
"step": 3260,
"valid_targets_mean": 4458.6,
"valid_targets_min": 359
},
{
"epoch": 5.232371794871795,
"grad_norm": 0.5316155743604632,
"learning_rate": 7.292452951130548e-06,
"loss": 0.15,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1323135793209076,
"step": 3265,
"valid_targets_mean": 4457.7,
"valid_targets_min": 1685
},
{
"epoch": 5.240384615384615,
"grad_norm": 0.5492125064022614,
"learning_rate": 7.230841429141347e-06,
"loss": 0.1602,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20732444524765015,
"step": 3270,
"valid_targets_mean": 4965.0,
"valid_targets_min": 853
},
{
"epoch": 5.248397435897436,
"grad_norm": 0.5193973157192391,
"learning_rate": 7.169433796890595e-06,
"loss": 0.1617,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15308740735054016,
"step": 3275,
"valid_targets_mean": 4178.1,
"valid_targets_min": 641
},
{
"epoch": 5.256410256410256,
"grad_norm": 0.5209009706836608,
"learning_rate": 7.108231034895976e-06,
"loss": 0.1521,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16298431158065796,
"step": 3280,
"valid_targets_mean": 4440.5,
"valid_targets_min": 343
},
{
"epoch": 5.264423076923077,
"grad_norm": 0.514990254693756,
"learning_rate": 7.047234120403972e-06,
"loss": 0.1495,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1427186131477356,
"step": 3285,
"valid_targets_mean": 4187.7,
"valid_targets_min": 859
},
{
"epoch": 5.272435897435898,
"grad_norm": 0.5123977736090055,
"learning_rate": 6.986444027374211e-06,
"loss": 0.1465,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14697766304016113,
"step": 3290,
"valid_targets_mean": 4412.2,
"valid_targets_min": 594
},
{
"epoch": 5.280448717948718,
"grad_norm": 0.5370919233708519,
"learning_rate": 6.925861726463919e-06,
"loss": 0.1462,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12968498468399048,
"step": 3295,
"valid_targets_mean": 4780.1,
"valid_targets_min": 343
},
{
"epoch": 5.288461538461538,
"grad_norm": 0.5893789549229775,
"learning_rate": 6.865488185012464e-06,
"loss": 0.1502,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.147515669465065,
"step": 3300,
"valid_targets_mean": 4007.0,
"valid_targets_min": 328
},
{
"epoch": 5.296474358974359,
"grad_norm": 0.557032974245759,
"learning_rate": 6.805324367025825e-06,
"loss": 0.1524,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14404591917991638,
"step": 3305,
"valid_targets_mean": 3388.2,
"valid_targets_min": 317
},
{
"epoch": 5.30448717948718,
"grad_norm": 0.5245596795012754,
"learning_rate": 6.745371233161309e-06,
"loss": 0.1583,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17218780517578125,
"step": 3310,
"valid_targets_mean": 4910.2,
"valid_targets_min": 1329
},
{
"epoch": 5.3125,
"grad_norm": 0.4572021511074751,
"learning_rate": 6.685629740712103e-06,
"loss": 0.1467,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15132874250411987,
"step": 3315,
"valid_targets_mean": 5872.9,
"valid_targets_min": 3284
},
{
"epoch": 5.32051282051282,
"grad_norm": 0.5046465518598903,
"learning_rate": 6.6261008435920605e-06,
"loss": 0.1483,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14855624735355377,
"step": 3320,
"valid_targets_mean": 4805.1,
"valid_targets_min": 357
},
{
"epoch": 5.328525641025641,
"grad_norm": 0.5800484814127054,
"learning_rate": 6.566785492320471e-06,
"loss": 0.1574,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16996949911117554,
"step": 3325,
"valid_targets_mean": 4481.9,
"valid_targets_min": 324
},
{
"epoch": 5.336538461538462,
"grad_norm": 0.4914190952839447,
"learning_rate": 6.507684634006815e-06,
"loss": 0.1546,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1526101678609848,
"step": 3330,
"valid_targets_mean": 4738.5,
"valid_targets_min": 504
},
{
"epoch": 5.344551282051282,
"grad_norm": 0.5520175051676571,
"learning_rate": 6.448799212335734e-06,
"loss": 0.1475,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13613277673721313,
"step": 3335,
"valid_targets_mean": 5095.2,
"valid_targets_min": 2477
},
{
"epoch": 5.352564102564102,
"grad_norm": 0.4986142452528356,
"learning_rate": 6.390130167551869e-06,
"loss": 0.144,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13920287787914276,
"step": 3340,
"valid_targets_mean": 4504.7,
"valid_targets_min": 386
},
{
"epoch": 5.360576923076923,
"grad_norm": 0.5956829941579933,
"learning_rate": 6.331678436444939e-06,
"loss": 0.1496,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15984797477722168,
"step": 3345,
"valid_targets_mean": 4252.8,
"valid_targets_min": 407
},
{
"epoch": 5.368589743589744,
"grad_norm": 0.5302709581166493,
"learning_rate": 6.273444952334713e-06,
"loss": 0.1537,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1777845025062561,
"step": 3350,
"valid_targets_mean": 4262.6,
"valid_targets_min": 971
},
{
"epoch": 5.376602564102564,
"grad_norm": 0.5293650615014065,
"learning_rate": 6.2154306450561175e-06,
"loss": 0.1448,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14727038145065308,
"step": 3355,
"valid_targets_mean": 4646.3,
"valid_targets_min": 580
},
{
"epoch": 5.384615384615385,
"grad_norm": 0.549526964949152,
"learning_rate": 6.157636440944445e-06,
"loss": 0.1527,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17116007208824158,
"step": 3360,
"valid_targets_mean": 4763.5,
"valid_targets_min": 282
},
{
"epoch": 5.392628205128205,
"grad_norm": 0.5801151632305452,
"learning_rate": 6.100063262820474e-06,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15987835824489594,
"step": 3365,
"valid_targets_mean": 4164.9,
"valid_targets_min": 617
},
{
"epoch": 5.4006410256410255,
"grad_norm": 0.6209518500735427,
"learning_rate": 6.0427120299758236e-06,
"loss": 0.1519,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17719699442386627,
"step": 3370,
"valid_targets_mean": 3831.4,
"valid_targets_min": 355
},
{
"epoch": 5.408653846153846,
"grad_norm": 0.5085229893370715,
"learning_rate": 5.985583658158212e-06,
"loss": 0.1392,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1472761332988739,
"step": 3375,
"valid_targets_mean": 4365.6,
"valid_targets_min": 1037
},
{
"epoch": 5.416666666666667,
"grad_norm": 0.7120850511044665,
"learning_rate": 5.928679059556852e-06,
"loss": 0.1601,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1559578776359558,
"step": 3380,
"valid_targets_mean": 3396.2,
"valid_targets_min": 298
},
{
"epoch": 5.424679487179487,
"grad_norm": 0.571318505016176,
"learning_rate": 5.871999142787908e-06,
"loss": 0.1518,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15362244844436646,
"step": 3385,
"valid_targets_mean": 4239.9,
"valid_targets_min": 737
},
{
"epoch": 5.4326923076923075,
"grad_norm": 0.6141401475589294,
"learning_rate": 5.815544812879936e-06,
"loss": 0.1482,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1585809737443924,
"step": 3390,
"valid_targets_mean": 3491.8,
"valid_targets_min": 754
},
{
"epoch": 5.440705128205128,
"grad_norm": 0.581667982340018,
"learning_rate": 5.759316971259503e-06,
"loss": 0.1481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15154506266117096,
"step": 3395,
"valid_targets_mean": 3723.0,
"valid_targets_min": 550
},
{
"epoch": 5.448717948717949,
"grad_norm": 0.5041207130915869,
"learning_rate": 5.703316515736734e-06,
"loss": 0.1425,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1316491812467575,
"step": 3400,
"valid_targets_mean": 4177.4,
"valid_targets_min": 317
},
{
"epoch": 5.456730769230769,
"grad_norm": 0.5342499521472172,
"learning_rate": 5.647544340491007e-06,
"loss": 0.1461,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1405012309551239,
"step": 3405,
"valid_targets_mean": 4178.2,
"valid_targets_min": 857
},
{
"epoch": 5.464743589743589,
"grad_norm": 0.498509557024922,
"learning_rate": 5.592001336056659e-06,
"loss": 0.1498,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16636775434017181,
"step": 3410,
"valid_targets_mean": 4985.4,
"valid_targets_min": 827
},
{
"epoch": 5.472756410256411,
"grad_norm": 0.5028998726460713,
"learning_rate": 5.536688389308782e-06,
"loss": 0.1569,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15679937601089478,
"step": 3415,
"valid_targets_mean": 5159.8,
"valid_targets_min": 2222
},
{
"epoch": 5.480769230769231,
"grad_norm": 0.5993362430915836,
"learning_rate": 5.4816063834490496e-06,
"loss": 0.1477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14045289158821106,
"step": 3420,
"valid_targets_mean": 4467.2,
"valid_targets_min": 717
},
{
"epoch": 5.488782051282051,
"grad_norm": 0.5684427211796031,
"learning_rate": 5.426756197991625e-06,
"loss": 0.1526,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13909168541431427,
"step": 3425,
"valid_targets_mean": 5007.2,
"valid_targets_min": 271
},
{
"epoch": 5.496794871794872,
"grad_norm": 0.5479529660747777,
"learning_rate": 5.372138708749104e-06,
"loss": 0.1532,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1453046202659607,
"step": 3430,
"valid_targets_mean": 3937.9,
"valid_targets_min": 762
},
{
"epoch": 5.5048076923076925,
"grad_norm": 0.5378471898494906,
"learning_rate": 5.3177547878185436e-06,
"loss": 0.1582,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14055317640304565,
"step": 3435,
"valid_targets_mean": 4476.1,
"valid_targets_min": 292
},
{
"epoch": 5.512820512820513,
"grad_norm": 0.5582338792476373,
"learning_rate": 5.263605303567532e-06,
"loss": 0.1371,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14210942387580872,
"step": 3440,
"valid_targets_mean": 4133.8,
"valid_targets_min": 282
},
{
"epoch": 5.520833333333333,
"grad_norm": 0.6703529081897249,
"learning_rate": 5.20969112062032e-06,
"loss": 0.1576,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1555454134941101,
"step": 3445,
"valid_targets_mean": 3554.4,
"valid_targets_min": 278
},
{
"epoch": 5.528846153846154,
"grad_norm": 0.5194988424541571,
"learning_rate": 5.156013099844017e-06,
"loss": 0.1477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15153059363365173,
"step": 3450,
"valid_targets_mean": 4626.6,
"valid_targets_min": 1208
},
{
"epoch": 5.5368589743589745,
"grad_norm": 0.47911289768246845,
"learning_rate": 5.1025720983348544e-06,
"loss": 0.1507,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15803536772727966,
"step": 3455,
"valid_targets_mean": 5187.6,
"valid_targets_min": 1174
},
{
"epoch": 5.544871794871795,
"grad_norm": 0.5403400880109693,
"learning_rate": 5.049368969404484e-06,
"loss": 0.15,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15584543347358704,
"step": 3460,
"valid_targets_mean": 4910.1,
"valid_targets_min": 1069
},
{
"epoch": 5.552884615384615,
"grad_norm": 0.564398776018862,
"learning_rate": 4.99640456256636e-06,
"loss": 0.1493,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16660058498382568,
"step": 3465,
"valid_targets_mean": 4017.1,
"valid_targets_min": 261
},
{
"epoch": 5.560897435897436,
"grad_norm": 0.578778128504491,
"learning_rate": 4.9436797235221814e-06,
"loss": 0.1513,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13579995930194855,
"step": 3470,
"valid_targets_mean": 3500.8,
"valid_targets_min": 339
},
{
"epoch": 5.568910256410256,
"grad_norm": 0.49961375549558895,
"learning_rate": 4.891195294148376e-06,
"loss": 0.1542,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15672512352466583,
"step": 3475,
"valid_targets_mean": 4762.4,
"valid_targets_min": 782
},
{
"epoch": 5.576923076923077,
"grad_norm": 0.5634541995131147,
"learning_rate": 4.838952112482671e-06,
"loss": 0.1442,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14387747645378113,
"step": 3480,
"valid_targets_mean": 3813.2,
"valid_targets_min": 287
},
{
"epoch": 5.584935897435898,
"grad_norm": 0.49090730478236594,
"learning_rate": 4.786951012710699e-06,
"loss": 0.1637,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1561790108680725,
"step": 3485,
"valid_targets_mean": 4932.4,
"valid_targets_min": 353
},
{
"epoch": 5.592948717948718,
"grad_norm": 0.563494138678453,
"learning_rate": 4.735192825152686e-06,
"loss": 0.1587,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16302964091300964,
"step": 3490,
"valid_targets_mean": 4066.9,
"valid_targets_min": 328
},
{
"epoch": 5.600961538461538,
"grad_norm": 0.5529911390773293,
"learning_rate": 4.683678376250189e-06,
"loss": 0.1488,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15432584285736084,
"step": 3495,
"valid_targets_mean": 3943.9,
"valid_targets_min": 221
},
{
"epoch": 5.608974358974359,
"grad_norm": 0.4733764042906081,
"learning_rate": 4.6324084885529086e-06,
"loss": 0.1637,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1666216254234314,
"step": 3500,
"valid_targets_mean": 5284.3,
"valid_targets_min": 1895
},
{
"epoch": 5.61698717948718,
"grad_norm": 0.5031850788326319,
"learning_rate": 4.581383980705538e-06,
"loss": 0.1548,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16201062500476837,
"step": 3505,
"valid_targets_mean": 5158.6,
"valid_targets_min": 1210
},
{
"epoch": 5.625,
"grad_norm": 0.48243575002706995,
"learning_rate": 4.530605667434727e-06,
"loss": 0.1524,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1448429375886917,
"step": 3510,
"valid_targets_mean": 5058.6,
"valid_targets_min": 1376
},
{
"epoch": 5.63301282051282,
"grad_norm": 0.6117296477169918,
"learning_rate": 4.480074359536013e-06,
"loss": 0.1589,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15564385056495667,
"step": 3515,
"valid_targets_mean": 4181.0,
"valid_targets_min": 548
},
{
"epoch": 5.641025641025641,
"grad_norm": 0.6085977315520913,
"learning_rate": 4.429790863860934e-06,
"loss": 0.1503,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1697201430797577,
"step": 3520,
"valid_targets_mean": 4754.1,
"valid_targets_min": 613
},
{
"epoch": 5.649038461538462,
"grad_norm": 0.5115852970565925,
"learning_rate": 4.3797559833041146e-06,
"loss": 0.1716,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1357446312904358,
"step": 3525,
"valid_targets_mean": 4338.3,
"valid_targets_min": 828
},
{
"epoch": 5.657051282051282,
"grad_norm": 0.5561427762694048,
"learning_rate": 4.329970516790447e-06,
"loss": 0.1482,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15750139951705933,
"step": 3530,
"valid_targets_mean": 3765.8,
"valid_targets_min": 1009
},
{
"epoch": 5.665064102564102,
"grad_norm": 0.5642997757040915,
"learning_rate": 4.280435259262363e-06,
"loss": 0.1599,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13392791152000427,
"step": 3535,
"valid_targets_mean": 3417.7,
"valid_targets_min": 294
},
{
"epoch": 5.673076923076923,
"grad_norm": 0.6082540644447687,
"learning_rate": 4.231151001667077e-06,
"loss": 0.1501,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14874929189682007,
"step": 3540,
"valid_targets_mean": 4694.6,
"valid_targets_min": 2471
},
{
"epoch": 5.681089743589744,
"grad_norm": 0.6548954718367228,
"learning_rate": 4.182118530944044e-06,
"loss": 0.1561,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16230438649654388,
"step": 3545,
"valid_targets_mean": 3323.4,
"valid_targets_min": 329
},
{
"epoch": 5.689102564102564,
"grad_norm": 0.5430618343101518,
"learning_rate": 4.133338630012307e-06,
"loss": 0.159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15367120504379272,
"step": 3550,
"valid_targets_mean": 3942.6,
"valid_targets_min": 314
},
{
"epoch": 5.697115384615385,
"grad_norm": 0.5304908674268932,
"learning_rate": 4.0848120777580554e-06,
"loss": 0.1575,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15589433908462524,
"step": 3555,
"valid_targets_mean": 4362.6,
"valid_targets_min": 959
},
{
"epoch": 5.705128205128205,
"grad_norm": 0.44297758952057176,
"learning_rate": 4.036539649022182e-06,
"loss": 0.1585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14082220196723938,
"step": 3560,
"valid_targets_mean": 5507.7,
"valid_targets_min": 3086
},
{
"epoch": 5.7131410256410255,
"grad_norm": 2.605302944239685,
"learning_rate": 3.988522114587865e-06,
"loss": 0.1436,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14218877255916595,
"step": 3565,
"valid_targets_mean": 3798.8,
"valid_targets_min": 1200
},
{
"epoch": 5.721153846153846,
"grad_norm": 0.5413953038685451,
"learning_rate": 3.940760241168331e-06,
"loss": 0.1421,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14577752351760864,
"step": 3570,
"valid_targets_mean": 4153.8,
"valid_targets_min": 598
},
{
"epoch": 5.729166666666667,
"grad_norm": 0.4983998941473086,
"learning_rate": 3.893254791394541e-06,
"loss": 0.1468,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1316983699798584,
"step": 3575,
"valid_targets_mean": 4661.9,
"valid_targets_min": 900
},
{
"epoch": 5.737179487179487,
"grad_norm": 0.5273843179411134,
"learning_rate": 3.846006523803074e-06,
"loss": 0.1401,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1403467357158661,
"step": 3580,
"valid_targets_mean": 4550.4,
"valid_targets_min": 369
},
{
"epoch": 5.7451923076923075,
"grad_norm": 0.5442771085390932,
"learning_rate": 3.799016192823981e-06,
"loss": 0.156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16002416610717773,
"step": 3585,
"valid_targets_mean": 4726.9,
"valid_targets_min": 330
},
{
"epoch": 5.753205128205128,
"grad_norm": 0.6434299034962991,
"learning_rate": 3.7522845487687276e-06,
"loss": 0.1428,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14445558190345764,
"step": 3590,
"valid_targets_mean": 3594.9,
"valid_targets_min": 367
},
{
"epoch": 5.761217948717949,
"grad_norm": 0.5580586982907163,
"learning_rate": 3.7058123378182664e-06,
"loss": 0.1565,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15552376210689545,
"step": 3595,
"valid_targets_mean": 3636.8,
"valid_targets_min": 448
},
{
"epoch": 5.769230769230769,
"grad_norm": 0.5173425863593317,
"learning_rate": 3.6596003020110636e-06,
"loss": 0.156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1508871614933014,
"step": 3600,
"valid_targets_mean": 4426.9,
"valid_targets_min": 840
},
{
"epoch": 5.777243589743589,
"grad_norm": 0.5244756610226968,
"learning_rate": 3.613649179231287e-06,
"loss": 0.1601,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18764357268810272,
"step": 3605,
"valid_targets_mean": 5464.5,
"valid_targets_min": 1508
},
{
"epoch": 5.785256410256411,
"grad_norm": 0.5531139491697628,
"learning_rate": 3.5679597031970017e-06,
"loss": 0.1462,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.139120951294899,
"step": 3610,
"valid_targets_mean": 3985.1,
"valid_targets_min": 554
},
{
"epoch": 5.793269230769231,
"grad_norm": 0.4818465181569853,
"learning_rate": 3.5225326034484764e-06,
"loss": 0.1475,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13170459866523743,
"step": 3615,
"valid_targets_mean": 4955.9,
"valid_targets_min": 774
},
{
"epoch": 5.801282051282051,
"grad_norm": 0.5282655948022116,
"learning_rate": 3.4773686053365197e-06,
"loss": 0.1571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13457638025283813,
"step": 3620,
"valid_targets_mean": 5036.7,
"valid_targets_min": 366
},
{
"epoch": 5.809294871794872,
"grad_norm": 0.5515934639916573,
"learning_rate": 3.4324684300109003e-06,
"loss": 0.1451,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14944568276405334,
"step": 3625,
"valid_targets_mean": 3912.7,
"valid_targets_min": 476
},
{
"epoch": 5.8173076923076925,
"grad_norm": 0.5320429061168414,
"learning_rate": 3.387832794408832e-06,
"loss": 0.1449,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1446266621351242,
"step": 3630,
"valid_targets_mean": 4598.1,
"valid_targets_min": 468
},
{
"epoch": 5.825320512820513,
"grad_norm": 0.5461834831411754,
"learning_rate": 3.3434624112435342e-06,
"loss": 0.1432,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14564934372901917,
"step": 3635,
"valid_targets_mean": 4962.0,
"valid_targets_min": 329
},
{
"epoch": 5.833333333333333,
"grad_norm": 0.5126166462482566,
"learning_rate": 3.2993579889928397e-06,
"loss": 0.1486,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13514655828475952,
"step": 3640,
"valid_targets_mean": 4625.7,
"valid_targets_min": 698
},
{
"epoch": 5.841346153846154,
"grad_norm": 0.5429971801064682,
"learning_rate": 3.25552023188789e-06,
"loss": 0.1657,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18387652933597565,
"step": 3645,
"valid_targets_mean": 4552.1,
"valid_targets_min": 875
},
{
"epoch": 5.8493589743589745,
"grad_norm": 0.5198227665075318,
"learning_rate": 3.211949839901889e-06,
"loss": 0.1684,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18851862847805023,
"step": 3650,
"valid_targets_mean": 4796.6,
"valid_targets_min": 2009
},
{
"epoch": 5.857371794871795,
"grad_norm": 0.4727083304631456,
"learning_rate": 3.168647508738927e-06,
"loss": 0.1527,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13420718908309937,
"step": 3655,
"valid_targets_mean": 4606.7,
"valid_targets_min": 440
},
{
"epoch": 5.865384615384615,
"grad_norm": 0.851609931655878,
"learning_rate": 3.125613929822866e-06,
"loss": 0.1566,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16550441086292267,
"step": 3660,
"valid_targets_mean": 3886.1,
"valid_targets_min": 301
},
{
"epoch": 5.873397435897436,
"grad_norm": 0.5728801995776978,
"learning_rate": 3.0828497902863106e-06,
"loss": 0.1501,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15078899264335632,
"step": 3665,
"valid_targets_mean": 3929.6,
"valid_targets_min": 265
},
{
"epoch": 5.881410256410256,
"grad_norm": 0.5101270696396807,
"learning_rate": 3.0403557729596267e-06,
"loss": 0.156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16278427839279175,
"step": 3670,
"valid_targets_mean": 4321.5,
"valid_targets_min": 565
},
{
"epoch": 5.889423076923077,
"grad_norm": 0.5730655916243329,
"learning_rate": 2.998132556360038e-06,
"loss": 0.1569,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1519847810268402,
"step": 3675,
"valid_targets_mean": 4000.5,
"valid_targets_min": 963
},
{
"epoch": 5.897435897435898,
"grad_norm": 0.5108431077755453,
"learning_rate": 2.9561808146808068e-06,
"loss": 0.1545,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14279256761074066,
"step": 3680,
"valid_targets_mean": 4451.8,
"valid_targets_min": 1587
},
{
"epoch": 5.905448717948718,
"grad_norm": 0.5395993486339674,
"learning_rate": 2.9145012177804476e-06,
"loss": 0.1604,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1897404044866562,
"step": 3685,
"valid_targets_mean": 4698.1,
"valid_targets_min": 1394
},
{
"epoch": 5.913461538461538,
"grad_norm": 0.45559505670781114,
"learning_rate": 2.8730944311720454e-06,
"loss": 0.1516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1436762660741806,
"step": 3690,
"valid_targets_mean": 5500.9,
"valid_targets_min": 2335
},
{
"epoch": 5.921474358974359,
"grad_norm": 0.46835082427644875,
"learning_rate": 2.8319611160126226e-06,
"loss": 0.153,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13440626859664917,
"step": 3695,
"valid_targets_mean": 4717.1,
"valid_targets_min": 431
},
{
"epoch": 5.92948717948718,
"grad_norm": 0.5260596052221173,
"learning_rate": 2.791101929092592e-06,
"loss": 0.1462,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11741216480731964,
"step": 3700,
"valid_targets_mean": 3852.9,
"valid_targets_min": 376
},
{
"epoch": 5.9375,
"grad_norm": 0.5526501708660039,
"learning_rate": 2.750517522825251e-06,
"loss": 0.1579,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17630136013031006,
"step": 3705,
"valid_targets_mean": 4239.2,
"valid_targets_min": 322
},
{
"epoch": 5.94551282051282,
"grad_norm": 0.6435269055745118,
"learning_rate": 2.710208545236397e-06,
"loss": 0.1534,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14900657534599304,
"step": 3710,
"valid_targets_mean": 3492.5,
"valid_targets_min": 284
},
{
"epoch": 5.953525641025641,
"grad_norm": 0.5299716566704048,
"learning_rate": 2.670175639953929e-06,
"loss": 0.1539,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14511975646018982,
"step": 3715,
"valid_targets_mean": 4305.9,
"valid_targets_min": 1833
},
{
"epoch": 5.961538461538462,
"grad_norm": 0.5345896991387908,
"learning_rate": 2.6304194461976207e-06,
"loss": 0.1563,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18142619729042053,
"step": 3720,
"valid_targets_mean": 4770.4,
"valid_targets_min": 504
},
{
"epoch": 5.969551282051282,
"grad_norm": 0.5336780870911274,
"learning_rate": 2.5909405987688896e-06,
"loss": 0.1413,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1686202585697174,
"step": 3725,
"valid_targets_mean": 4325.5,
"valid_targets_min": 369
},
{
"epoch": 5.977564102564102,
"grad_norm": 0.5869722476313707,
"learning_rate": 2.5517397280406565e-06,
"loss": 0.1598,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16414906084537506,
"step": 3730,
"valid_targets_mean": 3551.8,
"valid_targets_min": 349
},
{
"epoch": 5.985576923076923,
"grad_norm": 0.5483917312386541,
"learning_rate": 2.512817459947312e-06,
"loss": 0.1555,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16464398801326752,
"step": 3735,
"valid_targets_mean": 4415.8,
"valid_targets_min": 248
},
{
"epoch": 5.993589743589744,
"grad_norm": 0.5263572937304016,
"learning_rate": 2.4741744159746618e-06,
"loss": 0.1525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17008954286575317,
"step": 3740,
"valid_targets_mean": 4621.6,
"valid_targets_min": 361
},
{
"epoch": 6.001602564102564,
"grad_norm": 0.47762361284306515,
"learning_rate": 2.435811213150079e-06,
"loss": 0.1523,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15137696266174316,
"step": 3745,
"valid_targets_mean": 4790.8,
"valid_targets_min": 730
},
{
"epoch": 6.009615384615385,
"grad_norm": 0.5190852836567575,
"learning_rate": 2.3977284640325805e-06,
"loss": 0.1445,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14654141664505005,
"step": 3750,
"valid_targets_mean": 4255.8,
"valid_targets_min": 756
},
{
"epoch": 6.017628205128205,
"grad_norm": 0.6379150408372559,
"learning_rate": 2.359926776703092e-06,
"loss": 0.1415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15680421888828278,
"step": 3755,
"valid_targets_mean": 3829.4,
"valid_targets_min": 352
},
{
"epoch": 6.0256410256410255,
"grad_norm": 0.531872526304696,
"learning_rate": 2.3224067547547357e-06,
"loss": 0.1418,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14205306768417358,
"step": 3760,
"valid_targets_mean": 4010.1,
"valid_targets_min": 304
},
{
"epoch": 6.033653846153846,
"grad_norm": 0.5111476706637856,
"learning_rate": 2.2851689972831536e-06,
"loss": 0.1394,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12912288308143616,
"step": 3765,
"valid_targets_mean": 4093.1,
"valid_targets_min": 263
},
{
"epoch": 6.041666666666667,
"grad_norm": 0.561717144522184,
"learning_rate": 2.248214098877002e-06,
"loss": 0.1481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15707916021347046,
"step": 3770,
"valid_targets_mean": 4488.9,
"valid_targets_min": 475
},
{
"epoch": 6.049679487179487,
"grad_norm": 0.5416285337842635,
"learning_rate": 2.2115426496083958e-06,
"loss": 0.147,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14775054156780243,
"step": 3775,
"valid_targets_mean": 4594.3,
"valid_targets_min": 1199
},
{
"epoch": 6.0576923076923075,
"grad_norm": 0.6638036250954644,
"learning_rate": 2.175155235023536e-06,
"loss": 0.1481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16961002349853516,
"step": 3780,
"valid_targets_mean": 2815.1,
"valid_targets_min": 309
},
{
"epoch": 6.065705128205129,
"grad_norm": 0.5181575462611893,
"learning_rate": 2.1390524361333355e-06,
"loss": 0.15,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14543813467025757,
"step": 3785,
"valid_targets_mean": 4435.9,
"valid_targets_min": 850
},
{
"epoch": 6.073717948717949,
"grad_norm": 0.4868506111028336,
"learning_rate": 2.1032348294041305e-06,
"loss": 0.1508,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13385263085365295,
"step": 3790,
"valid_targets_mean": 4814.3,
"valid_targets_min": 597
},
{
"epoch": 6.081730769230769,
"grad_norm": 0.526299230617338,
"learning_rate": 2.067702986748521e-06,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1468874216079712,
"step": 3795,
"valid_targets_mean": 4299.1,
"valid_targets_min": 999
},
{
"epoch": 6.089743589743589,
"grad_norm": 0.5000911718103765,
"learning_rate": 2.0324574755161764e-06,
"loss": 0.1413,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14721500873565674,
"step": 3800,
"valid_targets_mean": 4727.7,
"valid_targets_min": 738
},
{
"epoch": 6.097756410256411,
"grad_norm": 0.6204025423572928,
"learning_rate": 1.9974988584848385e-06,
"loss": 0.1548,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13772644102573395,
"step": 3805,
"valid_targets_mean": 3074.3,
"valid_targets_min": 276
},
{
"epoch": 6.105769230769231,
"grad_norm": 0.5619309333152206,
"learning_rate": 1.96282769385129e-06,
"loss": 0.1584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17289316654205322,
"step": 3810,
"valid_targets_mean": 3989.2,
"valid_targets_min": 609
},
{
"epoch": 6.113782051282051,
"grad_norm": 0.5166149929094362,
"learning_rate": 1.9284445352224625e-06,
"loss": 0.1417,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14835289120674133,
"step": 3815,
"valid_targets_mean": 5252.2,
"valid_targets_min": 1031
},
{
"epoch": 6.121794871794871,
"grad_norm": 0.5325132133546706,
"learning_rate": 1.894349931606596e-06,
"loss": 0.1476,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1437477320432663,
"step": 3820,
"valid_targets_mean": 4024.3,
"valid_targets_min": 316
},
{
"epoch": 6.1298076923076925,
"grad_norm": 0.578833412719083,
"learning_rate": 1.8605444274044493e-06,
"loss": 0.1517,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14157095551490784,
"step": 3825,
"valid_targets_mean": 3536.2,
"valid_targets_min": 504
},
{
"epoch": 6.137820512820513,
"grad_norm": 0.5800455571158836,
"learning_rate": 1.827028562400659e-06,
"loss": 0.1425,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13655886054039001,
"step": 3830,
"valid_targets_mean": 4052.8,
"valid_targets_min": 428
},
{
"epoch": 6.145833333333333,
"grad_norm": 0.5440250130354706,
"learning_rate": 1.793802871755066e-06,
"loss": 0.1434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14248904585838318,
"step": 3835,
"valid_targets_mean": 4093.3,
"valid_targets_min": 392
},
{
"epoch": 6.153846153846154,
"grad_norm": 0.5185132694276166,
"learning_rate": 1.760867885994202e-06,
"loss": 0.1441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1571086049079895,
"step": 3840,
"valid_targets_mean": 4376.5,
"valid_targets_min": 738
},
{
"epoch": 6.1618589743589745,
"grad_norm": 0.49356945325327367,
"learning_rate": 1.7282241310028047e-06,
"loss": 0.1469,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15342417359352112,
"step": 3845,
"valid_targets_mean": 5123.1,
"valid_targets_min": 390
},
{
"epoch": 6.169871794871795,
"grad_norm": 0.7010774342976863,
"learning_rate": 1.6958721280154232e-06,
"loss": 0.1505,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1650376170873642,
"step": 3850,
"valid_targets_mean": 2733.2,
"valid_targets_min": 343
},
{
"epoch": 6.177884615384615,
"grad_norm": 0.5311203670433022,
"learning_rate": 1.6638123936081085e-06,
"loss": 0.1412,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12942928075790405,
"step": 3855,
"valid_targets_mean": 4368.7,
"valid_targets_min": 972
},
{
"epoch": 6.185897435897436,
"grad_norm": 0.5513999117684938,
"learning_rate": 1.6320454396901463e-06,
"loss": 0.1541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1369420886039734,
"step": 3860,
"valid_targets_mean": 3812.2,
"valid_targets_min": 787
},
{
"epoch": 6.193910256410256,
"grad_norm": 0.6453448983571556,
"learning_rate": 1.6005717734958914e-06,
"loss": 0.1468,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14638645946979523,
"step": 3865,
"valid_targets_mean": 4997.4,
"valid_targets_min": 294
},
{
"epoch": 6.201923076923077,
"grad_norm": 0.6009865962850838,
"learning_rate": 1.569391897576671e-06,
"loss": 0.1537,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14904412627220154,
"step": 3870,
"valid_targets_mean": 3487.3,
"valid_targets_min": 818
},
{
"epoch": 6.209935897435898,
"grad_norm": 0.5716136038698605,
"learning_rate": 1.5385063097927533e-06,
"loss": 0.1566,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14329415559768677,
"step": 3875,
"valid_targets_mean": 3903.8,
"valid_targets_min": 636
},
{
"epoch": 6.217948717948718,
"grad_norm": 0.5548927611136607,
"learning_rate": 1.5079155033054104e-06,
"loss": 0.1529,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13876593112945557,
"step": 3880,
"valid_targets_mean": 4137.1,
"valid_targets_min": 328
},
{
"epoch": 6.225961538461538,
"grad_norm": 1.0123565969239923,
"learning_rate": 1.4776199665690239e-06,
"loss": 0.1509,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16170409321784973,
"step": 3885,
"valid_targets_mean": 4509.0,
"valid_targets_min": 359
},
{
"epoch": 6.233974358974359,
"grad_norm": 0.5007254747887637,
"learning_rate": 1.4476201833233084e-06,
"loss": 0.1397,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13085854053497314,
"step": 3890,
"valid_targets_mean": 4747.0,
"valid_targets_min": 304
},
{
"epoch": 6.24198717948718,
"grad_norm": 0.5753859156747111,
"learning_rate": 1.4179166325855676e-06,
"loss": 0.147,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14013399183750153,
"step": 3895,
"valid_targets_mean": 3678.2,
"valid_targets_min": 719
},
{
"epoch": 6.25,
"grad_norm": 0.5053064988268128,
"learning_rate": 1.3885097886430599e-06,
"loss": 0.1541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14340434968471527,
"step": 3900,
"valid_targets_mean": 4563.3,
"valid_targets_min": 714
},
{
"epoch": 6.25801282051282,
"grad_norm": 0.6021561958634815,
"learning_rate": 1.35940012104542e-06,
"loss": 0.1469,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13250254094600677,
"step": 3905,
"valid_targets_mean": 3129.2,
"valid_targets_min": 321
},
{
"epoch": 6.266025641025641,
"grad_norm": 0.5883374347242302,
"learning_rate": 1.3305880945971583e-06,
"loss": 0.1462,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15309929847717285,
"step": 3910,
"valid_targets_mean": 3602.2,
"valid_targets_min": 272
},
{
"epoch": 6.274038461538462,
"grad_norm": 0.5456172725218881,
"learning_rate": 1.3020741693502403e-06,
"loss": 0.1424,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13385576009750366,
"step": 3915,
"valid_targets_mean": 3685.2,
"valid_targets_min": 345
},
{
"epoch": 6.282051282051282,
"grad_norm": 0.5367518745325782,
"learning_rate": 1.27385880059675e-06,
"loss": 0.1584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17513702809810638,
"step": 3920,
"valid_targets_mean": 4339.4,
"valid_targets_min": 249
},
{
"epoch": 6.290064102564102,
"grad_norm": 0.5011171324116793,
"learning_rate": 1.245942438861607e-06,
"loss": 0.1359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12527501583099365,
"step": 3925,
"valid_targets_mean": 5390.5,
"valid_targets_min": 327
},
{
"epoch": 6.298076923076923,
"grad_norm": 0.4966399917498914,
"learning_rate": 1.2183255298953788e-06,
"loss": 0.148,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1518949568271637,
"step": 3930,
"valid_targets_mean": 5114.8,
"valid_targets_min": 345
},
{
"epoch": 6.306089743589744,
"grad_norm": 0.5197828590593763,
"learning_rate": 1.1910085146671645e-06,
"loss": 0.1505,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15573599934577942,
"step": 3935,
"valid_targets_mean": 4435.1,
"valid_targets_min": 305
},
{
"epoch": 6.314102564102564,
"grad_norm": 0.6117197793928126,
"learning_rate": 1.1639918293575492e-06,
"loss": 0.1474,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1521931290626526,
"step": 3940,
"valid_targets_mean": 3810.1,
"valid_targets_min": 285
},
{
"epoch": 6.322115384615385,
"grad_norm": 0.5837468298978788,
"learning_rate": 1.1372759053516536e-06,
"loss": 0.1392,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13840635120868683,
"step": 3945,
"valid_targets_mean": 4503.8,
"valid_targets_min": 705
},
{
"epoch": 6.330128205128205,
"grad_norm": 0.49506682443475447,
"learning_rate": 1.1108611692322157e-06,
"loss": 0.1452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15087629854679108,
"step": 3950,
"valid_targets_mean": 4704.2,
"valid_targets_min": 331
},
{
"epoch": 6.3381410256410255,
"grad_norm": 0.5334501616182925,
"learning_rate": 1.0847480427728142e-06,
"loss": 0.1431,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1342017948627472,
"step": 3955,
"valid_targets_mean": 4061.0,
"valid_targets_min": 314
},
{
"epoch": 6.346153846153846,
"grad_norm": 0.5296906078090796,
"learning_rate": 1.0589369429311125e-06,
"loss": 0.1408,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15812858939170837,
"step": 3960,
"valid_targets_mean": 4588.1,
"valid_targets_min": 486
},
{
"epoch": 6.354166666666667,
"grad_norm": 0.5297442507517544,
"learning_rate": 1.0334282818422037e-06,
"loss": 0.1457,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13695786893367767,
"step": 3965,
"valid_targets_mean": 4689.1,
"valid_targets_min": 515
},
{
"epoch": 6.362179487179487,
"grad_norm": 0.5230304600132254,
"learning_rate": 1.008222466812041e-06,
"loss": 0.1541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16781078279018402,
"step": 3970,
"valid_targets_mean": 4798.9,
"valid_targets_min": 288
},
{
"epoch": 6.3701923076923075,
"grad_norm": 0.5782207964280378,
"learning_rate": 9.83319900310915e-07,
"loss": 0.1415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1245032325387001,
"step": 3975,
"valid_targets_mean": 3486.5,
"valid_targets_min": 486
},
{
"epoch": 6.378205128205128,
"grad_norm": 0.606191653182962,
"learning_rate": 9.587209799670495e-07,
"loss": 0.1423,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1584639549255371,
"step": 3980,
"valid_targets_mean": 4010.9,
"valid_targets_min": 971
},
{
"epoch": 6.386217948717949,
"grad_norm": 0.6505585302385207,
"learning_rate": 9.344260985602327e-07,
"loss": 0.147,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1441328525543213,
"step": 3985,
"valid_targets_mean": 4138.5,
"valid_targets_min": 326
},
{
"epoch": 6.394230769230769,
"grad_norm": 0.6178226975889122,
"learning_rate": 9.104356440155526e-07,
"loss": 0.1471,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1467789262533188,
"step": 3990,
"valid_targets_mean": 3504.4,
"valid_targets_min": 249
},
{
"epoch": 6.402243589743589,
"grad_norm": 0.5144408005336827,
"learning_rate": 8.867499993972162e-07,
"loss": 0.1353,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1299908459186554,
"step": 3995,
"valid_targets_mean": 4057.5,
"valid_targets_min": 645
},
{
"epoch": 6.410256410256411,
"grad_norm": 0.5332751953538373,
"learning_rate": 8.633695429024058e-07,
"loss": 0.1425,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14507465064525604,
"step": 4000,
"valid_targets_mean": 4870.1,
"valid_targets_min": 1254
},
{
"epoch": 6.418269230769231,
"grad_norm": 0.5075025825439389,
"learning_rate": 8.402946478552732e-07,
"loss": 0.1446,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14383724331855774,
"step": 4005,
"valid_targets_mean": 5179.3,
"valid_targets_min": 1751
},
{
"epoch": 6.426282051282051,
"grad_norm": 0.546020569963982,
"learning_rate": 8.175256827009392e-07,
"loss": 0.149,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1376463621854782,
"step": 4010,
"valid_targets_mean": 4594.8,
"valid_targets_min": 2366
},
{
"epoch": 6.434294871794872,
"grad_norm": 0.4965358151990727,
"learning_rate": 7.95063010999646e-07,
"loss": 0.1462,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14377667009830475,
"step": 4015,
"valid_targets_mean": 4992.8,
"valid_targets_min": 291
},
{
"epoch": 6.4423076923076925,
"grad_norm": 0.7087733569746311,
"learning_rate": 7.729069914209409e-07,
"loss": 0.1467,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15065652132034302,
"step": 4020,
"valid_targets_mean": 4950.9,
"valid_targets_min": 1005
},
{
"epoch": 6.450320512820513,
"grad_norm": 0.6033203685286874,
"learning_rate": 7.510579777379345e-07,
"loss": 0.1532,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16875869035720825,
"step": 4025,
"valid_targets_mean": 3811.9,
"valid_targets_min": 290
},
{
"epoch": 6.458333333333333,
"grad_norm": 0.4980909889651939,
"learning_rate": 7.295163188216792e-07,
"loss": 0.1424,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14184656739234924,
"step": 4030,
"valid_targets_mean": 4936.7,
"valid_targets_min": 2948
},
{
"epoch": 6.466346153846154,
"grad_norm": 0.5159055312431465,
"learning_rate": 7.08282358635568e-07,
"loss": 0.1514,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16906698048114777,
"step": 4035,
"valid_targets_mean": 5111.2,
"valid_targets_min": 328
},
{
"epoch": 6.4743589743589745,
"grad_norm": 0.542993534747814,
"learning_rate": 6.87356436229869e-07,
"loss": 0.1517,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1458635777235031,
"step": 4040,
"valid_targets_mean": 4206.2,
"valid_targets_min": 361
},
{
"epoch": 6.482371794871795,
"grad_norm": 0.5033162912489548,
"learning_rate": 6.667388857362977e-07,
"loss": 0.1502,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14949876070022583,
"step": 4045,
"valid_targets_mean": 4878.2,
"valid_targets_min": 363
},
{
"epoch": 6.490384615384615,
"grad_norm": 0.536431728991935,
"learning_rate": 6.464300363626797e-07,
"loss": 0.1417,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14794233441352844,
"step": 4050,
"valid_targets_mean": 4281.0,
"valid_targets_min": 315
},
{
"epoch": 6.498397435897436,
"grad_norm": 0.49206665425573526,
"learning_rate": 6.264302123877053e-07,
"loss": 0.1461,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13767631351947784,
"step": 4055,
"valid_targets_mean": 4549.7,
"valid_targets_min": 1725
},
{
"epoch": 6.506410256410256,
"grad_norm": 0.49235139585399057,
"learning_rate": 6.067397331557412e-07,
"loss": 0.1422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12583093345165253,
"step": 4060,
"valid_targets_mean": 4645.6,
"valid_targets_min": 450
},
{
"epoch": 6.514423076923077,
"grad_norm": 0.5398770589588451,
"learning_rate": 5.873589130717405e-07,
"loss": 0.1499,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14772993326187134,
"step": 4065,
"valid_targets_mean": 4355.2,
"valid_targets_min": 1113
},
{
"epoch": 6.522435897435898,
"grad_norm": 0.5248422830965018,
"learning_rate": 5.682880615962116e-07,
"loss": 0.1436,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13549655675888062,
"step": 4070,
"valid_targets_mean": 4379.6,
"valid_targets_min": 305
},
{
"epoch": 6.530448717948718,
"grad_norm": 0.5191480840331407,
"learning_rate": 5.495274832402841e-07,
"loss": 0.1433,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1365729123353958,
"step": 4075,
"valid_targets_mean": 4191.1,
"valid_targets_min": 299
},
{
"epoch": 6.538461538461538,
"grad_norm": 0.5661765501858168,
"learning_rate": 5.310774775608529e-07,
"loss": 0.1408,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1464567482471466,
"step": 4080,
"valid_targets_mean": 4098.4,
"valid_targets_min": 1278
},
{
"epoch": 6.546474358974359,
"grad_norm": 0.5181494542656516,
"learning_rate": 5.129383391557751e-07,
"loss": 0.1381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12176834791898727,
"step": 4085,
"valid_targets_mean": 4236.1,
"valid_targets_min": 312
},
{
"epoch": 6.55448717948718,
"grad_norm": 0.6372296579862414,
"learning_rate": 4.951103576591876e-07,
"loss": 0.1547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15168729424476624,
"step": 4090,
"valid_targets_mean": 3568.4,
"valid_targets_min": 342
},
{
"epoch": 6.5625,
"grad_norm": 0.5814754485987583,
"learning_rate": 4.7759381773687e-07,
"loss": 0.1501,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14342455565929413,
"step": 4095,
"valid_targets_mean": 5122.5,
"valid_targets_min": 349
},
{
"epoch": 6.57051282051282,
"grad_norm": 0.5411496888165933,
"learning_rate": 4.6038899908170234e-07,
"loss": 0.1516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1414349377155304,
"step": 4100,
"valid_targets_mean": 4128.8,
"valid_targets_min": 307
},
{
"epoch": 6.578525641025641,
"grad_norm": 0.5538171157501636,
"learning_rate": 4.4349617640920164e-07,
"loss": 0.1448,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14773766696453094,
"step": 4105,
"valid_targets_mean": 4527.3,
"valid_targets_min": 278
},
{
"epoch": 6.586538461538462,
"grad_norm": 0.8331185304777967,
"learning_rate": 4.2691561945312764e-07,
"loss": 0.1367,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1279963254928589,
"step": 4110,
"valid_targets_mean": 4435.6,
"valid_targets_min": 314
},
{
"epoch": 6.594551282051282,
"grad_norm": 0.5256963290934564,
"learning_rate": 4.106475929611886e-07,
"loss": 0.1449,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14436779916286469,
"step": 4115,
"valid_targets_mean": 4174.6,
"valid_targets_min": 1166
},
{
"epoch": 6.602564102564102,
"grad_norm": 0.6068393297118765,
"learning_rate": 3.9469235669080007e-07,
"loss": 0.1585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1584828644990921,
"step": 4120,
"valid_targets_mean": 4141.7,
"valid_targets_min": 895
},
{
"epoch": 6.610576923076923,
"grad_norm": 0.6704731153506479,
"learning_rate": 3.7905016540495053e-07,
"loss": 0.1416,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15452314913272858,
"step": 4125,
"valid_targets_mean": 3474.4,
"valid_targets_min": 254
},
{
"epoch": 6.618589743589744,
"grad_norm": 0.5533839901432267,
"learning_rate": 3.63721268868118e-07,
"loss": 0.1512,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1592683494091034,
"step": 4130,
"valid_targets_mean": 5466.2,
"valid_targets_min": 1865
},
{
"epoch": 6.626602564102564,
"grad_norm": 0.5590315085900683,
"learning_rate": 3.487059118422997e-07,
"loss": 0.1451,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15265530347824097,
"step": 4135,
"valid_targets_mean": 3547.8,
"valid_targets_min": 333
},
{
"epoch": 6.634615384615385,
"grad_norm": 0.5170149615127346,
"learning_rate": 3.3400433408308895e-07,
"loss": 0.1444,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14670473337173462,
"step": 4140,
"valid_targets_mean": 4572.9,
"valid_targets_min": 264
},
{
"epoch": 6.642628205128205,
"grad_norm": 0.5456088831608593,
"learning_rate": 3.196167703358577e-07,
"loss": 0.1376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12917637825012207,
"step": 4145,
"valid_targets_mean": 4298.1,
"valid_targets_min": 1364
},
{
"epoch": 6.6506410256410255,
"grad_norm": 0.613685220864772,
"learning_rate": 3.0554345033199985e-07,
"loss": 0.1449,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14656884968280792,
"step": 4150,
"valid_targets_mean": 3745.9,
"valid_targets_min": 330
},
{
"epoch": 6.658653846153846,
"grad_norm": 0.5522760605557776,
"learning_rate": 2.917845987852652e-07,
"loss": 0.1423,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1271848827600479,
"step": 4155,
"valid_targets_mean": 3934.1,
"valid_targets_min": 276
},
{
"epoch": 6.666666666666667,
"grad_norm": 0.6132961950364434,
"learning_rate": 2.783404353881758e-07,
"loss": 0.1467,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17348499596118927,
"step": 4160,
"valid_targets_mean": 3738.6,
"valid_targets_min": 282
},
{
"epoch": 6.674679487179487,
"grad_norm": 0.5560899625247082,
"learning_rate": 2.652111748085151e-07,
"loss": 0.1638,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16242700815200806,
"step": 4165,
"valid_targets_mean": 4589.6,
"valid_targets_min": 511
},
{
"epoch": 6.6826923076923075,
"grad_norm": 0.517619665822479,
"learning_rate": 2.523970266859044e-07,
"loss": 0.1472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16085945069789886,
"step": 4170,
"valid_targets_mean": 4618.5,
"valid_targets_min": 710
},
{
"epoch": 6.690705128205128,
"grad_norm": 0.5172675196014065,
"learning_rate": 2.398981956284363e-07,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1684911698102951,
"step": 4175,
"valid_targets_mean": 4658.2,
"valid_targets_min": 905
},
{
"epoch": 6.698717948717949,
"grad_norm": 0.5234117919054867,
"learning_rate": 2.2771488120944207e-07,
"loss": 0.1425,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14477473497390747,
"step": 4180,
"valid_targets_mean": 4637.7,
"valid_targets_min": 1779
},
{
"epoch": 6.706730769230769,
"grad_norm": 0.5890137860841297,
"learning_rate": 2.1584727796427174e-07,
"loss": 0.1461,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15862077474594116,
"step": 4185,
"valid_targets_mean": 3699.8,
"valid_targets_min": 272
},
{
"epoch": 6.714743589743589,
"grad_norm": 0.5589382681617442,
"learning_rate": 2.0429557538720556e-07,
"loss": 0.1502,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17357629537582397,
"step": 4190,
"valid_targets_mean": 5303.4,
"valid_targets_min": 301
},
{
"epoch": 6.722756410256411,
"grad_norm": 0.5349887371115216,
"learning_rate": 1.930599579284298e-07,
"loss": 0.155,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1659969538450241,
"step": 4195,
"valid_targets_mean": 4496.1,
"valid_targets_min": 377
},
{
"epoch": 6.730769230769231,
"grad_norm": 0.617253840479543,
"learning_rate": 1.8214060499107679e-07,
"loss": 0.1489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15575742721557617,
"step": 4200,
"valid_targets_mean": 3445.6,
"valid_targets_min": 760
},
{
"epoch": 6.738782051282051,
"grad_norm": 0.5386620896534594,
"learning_rate": 1.7153769092837614e-07,
"loss": 0.1472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14661836624145508,
"step": 4205,
"valid_targets_mean": 4364.1,
"valid_targets_min": 595
},
{
"epoch": 6.746794871794872,
"grad_norm": 0.7436036826677052,
"learning_rate": 1.6125138504086146e-07,
"loss": 0.1662,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14223778247833252,
"step": 4210,
"valid_targets_mean": 4072.8,
"valid_targets_min": 494
},
{
"epoch": 6.7548076923076925,
"grad_norm": 0.5576846452057956,
"learning_rate": 1.5128185157367247e-07,
"loss": 0.1373,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15080441534519196,
"step": 4215,
"valid_targets_mean": 4338.8,
"valid_targets_min": 625
},
{
"epoch": 6.762820512820513,
"grad_norm": 0.5111944334838397,
"learning_rate": 1.4162924971393044e-07,
"loss": 0.1466,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14644336700439453,
"step": 4220,
"valid_targets_mean": 5156.4,
"valid_targets_min": 228
},
{
"epoch": 6.770833333333333,
"grad_norm": 0.5238746873593738,
"learning_rate": 1.322937335881891e-07,
"loss": 0.1421,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.134772390127182,
"step": 4225,
"valid_targets_mean": 4097.3,
"valid_targets_min": 791
},
{
"epoch": 6.778846153846154,
"grad_norm": 0.5359397004694126,
"learning_rate": 1.2327545225999215e-07,
"loss": 0.1361,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13360458612442017,
"step": 4230,
"valid_targets_mean": 3873.0,
"valid_targets_min": 592
},
{
"epoch": 6.7868589743589745,
"grad_norm": 0.5966939669148095,
"learning_rate": 1.145745497274664e-07,
"loss": 0.1435,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1785121113061905,
"step": 4235,
"valid_targets_mean": 4373.3,
"valid_targets_min": 367
},
{
"epoch": 6.794871794871795,
"grad_norm": 0.537670114900445,
"learning_rate": 1.061911649210523e-07,
"loss": 0.1431,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14339633285999298,
"step": 4240,
"valid_targets_mean": 5094.1,
"valid_targets_min": 726
},
{
"epoch": 6.802884615384615,
"grad_norm": 0.5601268825053515,
"learning_rate": 9.812543170126365e-08,
"loss": 0.1434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1433734893798828,
"step": 4245,
"valid_targets_mean": 4290.6,
"valid_targets_min": 784
},
{
"epoch": 6.810897435897436,
"grad_norm": 0.5477258052862168,
"learning_rate": 9.03774788565559e-08,
"loss": 0.1435,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14027948677539825,
"step": 4250,
"valid_targets_mean": 3909.4,
"valid_targets_min": 480
},
{
"epoch": 6.818910256410256,
"grad_norm": 0.489311097609643,
"learning_rate": 8.294743010127448e-08,
"loss": 0.1525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15101996064186096,
"step": 4255,
"valid_targets_mean": 5336.7,
"valid_targets_min": 2103
},
{
"epoch": 6.826923076923077,
"grad_norm": 0.5868738981698945,
"learning_rate": 7.583540407367418e-08,
"loss": 0.1576,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1847933530807495,
"step": 4260,
"valid_targets_mean": 4343.7,
"valid_targets_min": 1212
},
{
"epoch": 6.834935897435898,
"grad_norm": 0.5855857213187571,
"learning_rate": 6.904151433402728e-08,
"loss": 0.1453,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15804800391197205,
"step": 4265,
"valid_targets_mean": 4060.9,
"valid_targets_min": 260
},
{
"epoch": 6.842948717948718,
"grad_norm": 0.5341073827330711,
"learning_rate": 6.256586936281172e-08,
"loss": 0.1516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13401219248771667,
"step": 4270,
"valid_targets_mean": 4276.9,
"valid_targets_min": 1521
},
{
"epoch": 6.850961538461538,
"grad_norm": 0.5596872071474632,
"learning_rate": 5.6408572558972475e-08,
"loss": 0.1447,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14991730451583862,
"step": 4275,
"valid_targets_mean": 4012.2,
"valid_targets_min": 313
},
{
"epoch": 6.858974358974359,
"grad_norm": 0.5127650079320776,
"learning_rate": 5.0569722238280605e-08,
"loss": 0.1517,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15968577563762665,
"step": 4280,
"valid_targets_mean": 4922.4,
"valid_targets_min": 1029
},
{
"epoch": 6.86698717948718,
"grad_norm": 0.556044222304068,
"learning_rate": 4.504941163175236e-08,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14913472533226013,
"step": 4285,
"valid_targets_mean": 4234.5,
"valid_targets_min": 315
},
{
"epoch": 6.875,
"grad_norm": 0.5173955328620299,
"learning_rate": 3.984772888417032e-08,
"loss": 0.1484,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14511007070541382,
"step": 4290,
"valid_targets_mean": 4831.1,
"valid_targets_min": 640
},
{
"epoch": 6.88301282051282,
"grad_norm": 0.5135904529444902,
"learning_rate": 3.4964757052671216e-08,
"loss": 0.1423,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1329515427350998,
"step": 4295,
"valid_targets_mean": 4548.6,
"valid_targets_min": 325
},
{
"epoch": 6.891025641025641,
"grad_norm": 0.5273812631403141,
"learning_rate": 3.0400574105415856e-08,
"loss": 0.1445,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1466929018497467,
"step": 4300,
"valid_targets_mean": 4443.2,
"valid_targets_min": 1286
},
{
"epoch": 6.899038461538462,
"grad_norm": 0.5112152128099553,
"learning_rate": 2.615525292035459e-08,
"loss": 0.1529,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14834466576576233,
"step": 4305,
"valid_targets_mean": 4525.5,
"valid_targets_min": 748
},
{
"epoch": 6.907051282051282,
"grad_norm": 0.4787950069823875,
"learning_rate": 2.222886128405266e-08,
"loss": 0.1482,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1258067488670349,
"step": 4310,
"valid_targets_mean": 5091.2,
"valid_targets_min": 630
},
{
"epoch": 6.915064102564102,
"grad_norm": 0.4902158430906645,
"learning_rate": 1.8621461890617752e-08,
"loss": 0.1512,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13186883926391602,
"step": 4315,
"valid_targets_mean": 5605.1,
"valid_targets_min": 1534
},
{
"epoch": 6.923076923076923,
"grad_norm": 0.5753741635709525,
"learning_rate": 1.5333112340687463e-08,
"loss": 0.1452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11813374608755112,
"step": 4320,
"valid_targets_mean": 3545.7,
"valid_targets_min": 215
},
{
"epoch": 6.931089743589744,
"grad_norm": 0.753356471718685,
"learning_rate": 1.2363865140518905e-08,
"loss": 0.1461,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14750586450099945,
"step": 4325,
"valid_targets_mean": 4844.5,
"valid_targets_min": 1011
},
{
"epoch": 6.939102564102564,
"grad_norm": 0.51962392181577,
"learning_rate": 9.713767701151621e-09,
"loss": 0.1456,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14431287348270416,
"step": 4330,
"valid_targets_mean": 4353.6,
"valid_targets_min": 693
},
{
"epoch": 6.947115384615385,
"grad_norm": 0.546166440798925,
"learning_rate": 7.382862337641516e-09,
"loss": 0.1567,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16867202520370483,
"step": 4335,
"valid_targets_mean": 4406.1,
"valid_targets_min": 370
},
{
"epoch": 6.955128205128205,
"grad_norm": 0.5208849270537348,
"learning_rate": 5.371186268390283e-09,
"loss": 0.1543,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16302883625030518,
"step": 4340,
"valid_targets_mean": 4673.9,
"valid_targets_min": 525
},
{
"epoch": 6.9631410256410255,
"grad_norm": 0.5897064804269239,
"learning_rate": 3.678771614550325e-09,
"loss": 0.1531,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15750685334205627,
"step": 4345,
"valid_targets_mean": 3988.9,
"valid_targets_min": 278
},
{
"epoch": 6.971153846153846,
"grad_norm": 0.5357145537981339,
"learning_rate": 2.3056453995162763e-09,
"loss": 0.1434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15159642696380615,
"step": 4350,
"valid_targets_mean": 4613.0,
"valid_targets_min": 698
},
{
"epoch": 6.979166666666667,
"grad_norm": 0.583373224453449,
"learning_rate": 1.2518295484875708e-09,
"loss": 0.1388,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14054948091506958,
"step": 4355,
"valid_targets_mean": 4758.1,
"valid_targets_min": 549
},
{
"epoch": 6.987179487179487,
"grad_norm": 0.5456294360171976,
"learning_rate": 5.173408881198328e-10,
"loss": 0.1517,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14700648188591003,
"step": 4360,
"valid_targets_mean": 5204.9,
"valid_targets_min": 1503
},
{
"epoch": 6.9951923076923075,
"grad_norm": 0.4782723455270486,
"learning_rate": 1.0219114625398263e-10,
"loss": 0.1475,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13123342394828796,
"step": 4365,
"valid_targets_mean": 4757.3,
"valid_targets_min": 244
},
{
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1500842571258545,
"step": 4368,
"total_flos": 1538335469010944.0,
"train_loss": 0.1953298668530616,
"train_runtime": 23876.8249,
"train_samples_per_second": 2.926,
"train_steps_per_second": 0.183,
"valid_targets_mean": 4269.2,
"valid_targets_min": 580
}
],
"logging_steps": 5,
"max_steps": 4368,
"num_input_tokens_seen": 0,
"num_train_epochs": 7,
"save_steps": 1500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 1538335469010944.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}