Files
exp-syh-tezos-stackoverflow…/trainer_state.json
ModelHub XC 329f9fd66c 初始化项目,由ModelHub XC社区提供模型
Model: laion/exp-syh-tezos-stackoverflow-mixed_glm_4_7_traces_jupiter_cleaned
Source: Original Platform
2026-08-22 05:39:18 +08:00

9178 lines
255 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 7.0,
"eval_steps": 500,
"global_step": 4151,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.008431703204047217,
"grad_norm": 11.501664361122982,
"learning_rate": 3.846153846153847e-07,
"loss": 0.5979,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36830034852027893,
"step": 5,
"valid_targets_mean": 7085.2,
"valid_targets_min": 269
},
{
"epoch": 0.016863406408094434,
"grad_norm": 9.798136578054267,
"learning_rate": 8.653846153846154e-07,
"loss": 0.5376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2571937143802643,
"step": 10,
"valid_targets_mean": 8085.4,
"valid_targets_min": 812
},
{
"epoch": 0.025295109612141653,
"grad_norm": 8.045896861375654,
"learning_rate": 1.3461538461538462e-06,
"loss": 0.5633,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3087533414363861,
"step": 15,
"valid_targets_mean": 10508.8,
"valid_targets_min": 2586
},
{
"epoch": 0.03372681281618887,
"grad_norm": 7.371631685286674,
"learning_rate": 1.826923076923077e-06,
"loss": 0.5413,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22022399306297302,
"step": 20,
"valid_targets_mean": 7668.2,
"valid_targets_min": 557
},
{
"epoch": 0.04215851602023609,
"grad_norm": 4.81283816877758,
"learning_rate": 2.307692307692308e-06,
"loss": 0.5043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22741477191448212,
"step": 25,
"valid_targets_mean": 6946.5,
"valid_targets_min": 3160
},
{
"epoch": 0.050590219224283306,
"grad_norm": 3.1051127569800414,
"learning_rate": 2.7884615384615386e-06,
"loss": 0.4539,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2195306420326233,
"step": 30,
"valid_targets_mean": 5693.1,
"valid_targets_min": 383
},
{
"epoch": 0.05902192242833052,
"grad_norm": 1.6765063848632142,
"learning_rate": 3.2692307692307696e-06,
"loss": 0.44,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14829085767269135,
"step": 35,
"valid_targets_mean": 5373.4,
"valid_targets_min": 399
},
{
"epoch": 0.06745362563237774,
"grad_norm": 0.9832461239275655,
"learning_rate": 3.7500000000000005e-06,
"loss": 0.4021,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19490928947925568,
"step": 40,
"valid_targets_mean": 7610.6,
"valid_targets_min": 1151
},
{
"epoch": 0.07588532883642496,
"grad_norm": 0.7804969166595506,
"learning_rate": 4.230769230769231e-06,
"loss": 0.3699,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18852092325687408,
"step": 45,
"valid_targets_mean": 7590.9,
"valid_targets_min": 863
},
{
"epoch": 0.08431703204047218,
"grad_norm": 0.5870246584538654,
"learning_rate": 4.711538461538462e-06,
"loss": 0.3904,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10879742354154587,
"step": 50,
"valid_targets_mean": 6546.2,
"valid_targets_min": 378
},
{
"epoch": 0.09274873524451939,
"grad_norm": 0.501164831410169,
"learning_rate": 5.192307692307693e-06,
"loss": 0.3702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20236611366271973,
"step": 55,
"valid_targets_mean": 9931.6,
"valid_targets_min": 686
},
{
"epoch": 0.10118043844856661,
"grad_norm": 0.5431051882856589,
"learning_rate": 5.6730769230769235e-06,
"loss": 0.4056,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16180011630058289,
"step": 60,
"valid_targets_mean": 5995.6,
"valid_targets_min": 1060
},
{
"epoch": 0.10961214165261383,
"grad_norm": 0.48199191017963605,
"learning_rate": 6.153846153846155e-06,
"loss": 0.4106,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17160961031913757,
"step": 65,
"valid_targets_mean": 6642.9,
"valid_targets_min": 1480
},
{
"epoch": 0.11804384485666104,
"grad_norm": 0.5618924814018379,
"learning_rate": 6.6346153846153846e-06,
"loss": 0.3876,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1253470927476883,
"step": 70,
"valid_targets_mean": 3935.2,
"valid_targets_min": 288
},
{
"epoch": 0.12647554806070826,
"grad_norm": 0.4544912456760567,
"learning_rate": 7.115384615384616e-06,
"loss": 0.3666,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14284873008728027,
"step": 75,
"valid_targets_mean": 7864.8,
"valid_targets_min": 847
},
{
"epoch": 0.13490725126475547,
"grad_norm": 0.4726839176042114,
"learning_rate": 7.5961538461538465e-06,
"loss": 0.3675,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19099323451519012,
"step": 80,
"valid_targets_mean": 8334.2,
"valid_targets_min": 417
},
{
"epoch": 0.1433389544688027,
"grad_norm": 0.43962745167958156,
"learning_rate": 8.076923076923077e-06,
"loss": 0.3527,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2015790194272995,
"step": 85,
"valid_targets_mean": 9258.0,
"valid_targets_min": 426
},
{
"epoch": 0.15177065767284992,
"grad_norm": 0.4035809358036227,
"learning_rate": 8.557692307692308e-06,
"loss": 0.3635,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21266590058803558,
"step": 90,
"valid_targets_mean": 13562.1,
"valid_targets_min": 529
},
{
"epoch": 0.16020236087689713,
"grad_norm": 0.445011113431071,
"learning_rate": 9.03846153846154e-06,
"loss": 0.3567,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28299641609191895,
"step": 95,
"valid_targets_mean": 10170.6,
"valid_targets_min": 432
},
{
"epoch": 0.16863406408094436,
"grad_norm": 0.49617697007002176,
"learning_rate": 9.51923076923077e-06,
"loss": 0.3434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1918463110923767,
"step": 100,
"valid_targets_mean": 8421.4,
"valid_targets_min": 660
},
{
"epoch": 0.17706576728499157,
"grad_norm": 0.4246194873244484,
"learning_rate": 1e-05,
"loss": 0.3422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17130586504936218,
"step": 105,
"valid_targets_mean": 7335.1,
"valid_targets_min": 906
},
{
"epoch": 0.18549747048903878,
"grad_norm": 0.45355672981374173,
"learning_rate": 1.0480769230769232e-05,
"loss": 0.334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14784985780715942,
"step": 110,
"valid_targets_mean": 9566.6,
"valid_targets_min": 1034
},
{
"epoch": 0.19392917369308602,
"grad_norm": 0.4861981444096039,
"learning_rate": 1.0961538461538464e-05,
"loss": 0.3314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1463363766670227,
"step": 115,
"valid_targets_mean": 9278.0,
"valid_targets_min": 943
},
{
"epoch": 0.20236087689713322,
"grad_norm": 0.3513696103102496,
"learning_rate": 1.1442307692307693e-05,
"loss": 0.3302,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12183891236782074,
"step": 120,
"valid_targets_mean": 7784.9,
"valid_targets_min": 363
},
{
"epoch": 0.21079258010118043,
"grad_norm": 0.41302762583163183,
"learning_rate": 1.1923076923076925e-05,
"loss": 0.3238,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15790146589279175,
"step": 125,
"valid_targets_mean": 9032.8,
"valid_targets_min": 343
},
{
"epoch": 0.21922428330522767,
"grad_norm": 0.36833098746671605,
"learning_rate": 1.2403846153846156e-05,
"loss": 0.3184,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1425168365240097,
"step": 130,
"valid_targets_mean": 9530.8,
"valid_targets_min": 927
},
{
"epoch": 0.22765598650927488,
"grad_norm": 0.35632750309477546,
"learning_rate": 1.2884615384615386e-05,
"loss": 0.3287,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1511588990688324,
"step": 135,
"valid_targets_mean": 9320.0,
"valid_targets_min": 525
},
{
"epoch": 0.23608768971332209,
"grad_norm": 0.3839760500603974,
"learning_rate": 1.3365384615384615e-05,
"loss": 0.3401,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14773255586624146,
"step": 140,
"valid_targets_mean": 7942.8,
"valid_targets_min": 551
},
{
"epoch": 0.24451939291736932,
"grad_norm": 0.4027919633372496,
"learning_rate": 1.3846153846153847e-05,
"loss": 0.3314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11750765144824982,
"step": 145,
"valid_targets_mean": 5743.8,
"valid_targets_min": 386
},
{
"epoch": 0.25295109612141653,
"grad_norm": 0.39509014637894146,
"learning_rate": 1.4326923076923078e-05,
"loss": 0.3344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21389496326446533,
"step": 150,
"valid_targets_mean": 11631.9,
"valid_targets_min": 814
},
{
"epoch": 0.26138279932546377,
"grad_norm": 0.42934642768276293,
"learning_rate": 1.480769230769231e-05,
"loss": 0.3317,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17534306645393372,
"step": 155,
"valid_targets_mean": 6293.1,
"valid_targets_min": 653
},
{
"epoch": 0.26981450252951095,
"grad_norm": 0.5099995962262183,
"learning_rate": 1.528846153846154e-05,
"loss": 0.352,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18245422840118408,
"step": 160,
"valid_targets_mean": 5486.6,
"valid_targets_min": 281
},
{
"epoch": 0.2782462057335582,
"grad_norm": 0.34814582034242464,
"learning_rate": 1.576923076923077e-05,
"loss": 0.3368,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12868812680244446,
"step": 165,
"valid_targets_mean": 12304.4,
"valid_targets_min": 4453
},
{
"epoch": 0.2866779089376054,
"grad_norm": 0.4013671644627058,
"learning_rate": 1.6250000000000002e-05,
"loss": 0.3247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16478392481803894,
"step": 170,
"valid_targets_mean": 10902.6,
"valid_targets_min": 2086
},
{
"epoch": 0.2951096121416526,
"grad_norm": 0.36982305265389137,
"learning_rate": 1.673076923076923e-05,
"loss": 0.3223,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13514260947704315,
"step": 175,
"valid_targets_mean": 9129.5,
"valid_targets_min": 905
},
{
"epoch": 0.30354131534569984,
"grad_norm": 0.43492479353639024,
"learning_rate": 1.7211538461538465e-05,
"loss": 0.3298,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15397945046424866,
"step": 180,
"valid_targets_mean": 7150.8,
"valid_targets_min": 2491
},
{
"epoch": 0.31197301854974707,
"grad_norm": 0.4064615035314157,
"learning_rate": 1.7692307692307694e-05,
"loss": 0.3118,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18773217499256134,
"step": 185,
"valid_targets_mean": 9813.6,
"valid_targets_min": 1698
},
{
"epoch": 0.32040472175379425,
"grad_norm": 0.5033090972495295,
"learning_rate": 1.8173076923076924e-05,
"loss": 0.3449,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1756390631198883,
"step": 190,
"valid_targets_mean": 11611.0,
"valid_targets_min": 693
},
{
"epoch": 0.3288364249578415,
"grad_norm": 0.41263139500583995,
"learning_rate": 1.8653846153846157e-05,
"loss": 0.3169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16576656699180603,
"step": 195,
"valid_targets_mean": 8001.4,
"valid_targets_min": 377
},
{
"epoch": 0.3372681281618887,
"grad_norm": 0.420993712284801,
"learning_rate": 1.9134615384615387e-05,
"loss": 0.3171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1707465499639511,
"step": 200,
"valid_targets_mean": 10204.0,
"valid_targets_min": 1077
},
{
"epoch": 0.3456998313659359,
"grad_norm": 0.4367613634632769,
"learning_rate": 1.9615384615384617e-05,
"loss": 0.309,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16906008124351501,
"step": 205,
"valid_targets_mean": 7979.0,
"valid_targets_min": 1875
},
{
"epoch": 0.35413153456998314,
"grad_norm": 0.3248744949429436,
"learning_rate": 2.0096153846153846e-05,
"loss": 0.2988,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.113541379570961,
"step": 210,
"valid_targets_mean": 9570.4,
"valid_targets_min": 773
},
{
"epoch": 0.3625632377740304,
"grad_norm": 0.4311412288707716,
"learning_rate": 2.0576923076923076e-05,
"loss": 0.3063,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1082523912191391,
"step": 215,
"valid_targets_mean": 7513.6,
"valid_targets_min": 377
},
{
"epoch": 0.37099494097807756,
"grad_norm": 0.6146608530881765,
"learning_rate": 2.105769230769231e-05,
"loss": 0.3381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22556598484516144,
"step": 220,
"valid_targets_mean": 12560.1,
"valid_targets_min": 1625
},
{
"epoch": 0.3794266441821248,
"grad_norm": 0.42361126511900266,
"learning_rate": 2.153846153846154e-05,
"loss": 0.3208,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17275100946426392,
"step": 225,
"valid_targets_mean": 7761.5,
"valid_targets_min": 4376
},
{
"epoch": 0.38785834738617203,
"grad_norm": 0.3851240502238606,
"learning_rate": 2.2019230769230772e-05,
"loss": 0.2997,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12829145789146423,
"step": 230,
"valid_targets_mean": 11647.0,
"valid_targets_min": 697
},
{
"epoch": 0.3962900505902192,
"grad_norm": 0.422677449061301,
"learning_rate": 2.25e-05,
"loss": 0.3474,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17590142786502838,
"step": 235,
"valid_targets_mean": 8475.9,
"valid_targets_min": 1812
},
{
"epoch": 0.40472175379426645,
"grad_norm": 0.383519360660951,
"learning_rate": 2.298076923076923e-05,
"loss": 0.2915,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13052451610565186,
"step": 240,
"valid_targets_mean": 8538.6,
"valid_targets_min": 381
},
{
"epoch": 0.4131534569983137,
"grad_norm": 0.45528412455691736,
"learning_rate": 2.3461538461538464e-05,
"loss": 0.3093,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13710317015647888,
"step": 245,
"valid_targets_mean": 4980.2,
"valid_targets_min": 783
},
{
"epoch": 0.42158516020236086,
"grad_norm": 0.4498638169869249,
"learning_rate": 2.3942307692307694e-05,
"loss": 0.3206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15775921940803528,
"step": 250,
"valid_targets_mean": 9238.1,
"valid_targets_min": 502
},
{
"epoch": 0.4300168634064081,
"grad_norm": 0.3466983971141771,
"learning_rate": 2.4423076923076927e-05,
"loss": 0.3113,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14348097145557404,
"step": 255,
"valid_targets_mean": 10720.5,
"valid_targets_min": 1295
},
{
"epoch": 0.43844856661045534,
"grad_norm": 0.48376064213768327,
"learning_rate": 2.4903846153846157e-05,
"loss": 0.288,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1243462860584259,
"step": 260,
"valid_targets_mean": 6862.2,
"valid_targets_min": 3385
},
{
"epoch": 0.4468802698145025,
"grad_norm": 0.5664338352052242,
"learning_rate": 2.5384615384615386e-05,
"loss": 0.2971,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14861498773097992,
"step": 265,
"valid_targets_mean": 4057.9,
"valid_targets_min": 698
},
{
"epoch": 0.45531197301854975,
"grad_norm": 0.46505223769019355,
"learning_rate": 2.586538461538462e-05,
"loss": 0.3319,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1883731633424759,
"step": 270,
"valid_targets_mean": 9529.6,
"valid_targets_min": 4630
},
{
"epoch": 0.463743676222597,
"grad_norm": 0.45717493502019835,
"learning_rate": 2.634615384615385e-05,
"loss": 0.3314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1497742384672165,
"step": 275,
"valid_targets_mean": 6601.8,
"valid_targets_min": 620
},
{
"epoch": 0.47217537942664417,
"grad_norm": 0.4141179235701937,
"learning_rate": 2.682692307692308e-05,
"loss": 0.2737,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11978630721569061,
"step": 280,
"valid_targets_mean": 7221.5,
"valid_targets_min": 1058
},
{
"epoch": 0.4806070826306914,
"grad_norm": 0.41937811057772423,
"learning_rate": 2.7307692307692312e-05,
"loss": 0.3209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1403650939464569,
"step": 285,
"valid_targets_mean": 9229.1,
"valid_targets_min": 1553
},
{
"epoch": 0.48903878583473864,
"grad_norm": 0.3886910109512036,
"learning_rate": 2.7788461538461538e-05,
"loss": 0.2951,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1693798154592514,
"step": 290,
"valid_targets_mean": 10075.0,
"valid_targets_min": 311
},
{
"epoch": 0.4974704890387858,
"grad_norm": 0.37334853416933333,
"learning_rate": 2.8269230769230775e-05,
"loss": 0.3026,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1481993943452835,
"step": 295,
"valid_targets_mean": 8932.2,
"valid_targets_min": 370
},
{
"epoch": 0.5059021922428331,
"grad_norm": 0.4967423402394348,
"learning_rate": 2.875e-05,
"loss": 0.3222,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1374732404947281,
"step": 300,
"valid_targets_mean": 6898.2,
"valid_targets_min": 358
},
{
"epoch": 0.5143338954468802,
"grad_norm": 0.9725423049042601,
"learning_rate": 2.923076923076923e-05,
"loss": 0.3005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18834879994392395,
"step": 305,
"valid_targets_mean": 4743.8,
"valid_targets_min": 363
},
{
"epoch": 0.5227655986509275,
"grad_norm": 0.47242935241554923,
"learning_rate": 2.9711538461538464e-05,
"loss": 0.2975,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16733747720718384,
"step": 310,
"valid_targets_mean": 9543.1,
"valid_targets_min": 1586
},
{
"epoch": 0.5311973018549747,
"grad_norm": 0.42713571875902523,
"learning_rate": 3.0192307692307693e-05,
"loss": 0.294,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14805594086647034,
"step": 315,
"valid_targets_mean": 10864.9,
"valid_targets_min": 1301
},
{
"epoch": 0.5396290050590219,
"grad_norm": 0.39812719316988654,
"learning_rate": 3.0673076923076926e-05,
"loss": 0.3154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1822110116481781,
"step": 320,
"valid_targets_mean": 11715.2,
"valid_targets_min": 6752
},
{
"epoch": 0.5480607082630692,
"grad_norm": 0.34414563816337385,
"learning_rate": 3.115384615384616e-05,
"loss": 0.2842,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14371591806411743,
"step": 325,
"valid_targets_mean": 11877.0,
"valid_targets_min": 3622
},
{
"epoch": 0.5564924114671164,
"grad_norm": 0.39758762182996543,
"learning_rate": 3.1634615384615386e-05,
"loss": 0.2779,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11213129758834839,
"step": 330,
"valid_targets_mean": 7999.9,
"valid_targets_min": 386
},
{
"epoch": 0.5649241146711635,
"grad_norm": 0.4933599455984154,
"learning_rate": 3.211538461538462e-05,
"loss": 0.2799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11450517177581787,
"step": 335,
"valid_targets_mean": 9544.8,
"valid_targets_min": 339
},
{
"epoch": 0.5733558178752108,
"grad_norm": 0.400174996142026,
"learning_rate": 3.2596153846153845e-05,
"loss": 0.2947,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1516113579273224,
"step": 340,
"valid_targets_mean": 8616.9,
"valid_targets_min": 1736
},
{
"epoch": 0.581787521079258,
"grad_norm": 0.7922767325843253,
"learning_rate": 3.307692307692308e-05,
"loss": 0.3104,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13221657276153564,
"step": 345,
"valid_targets_mean": 7844.9,
"valid_targets_min": 630
},
{
"epoch": 0.5902192242833052,
"grad_norm": 0.44237832083725176,
"learning_rate": 3.355769230769231e-05,
"loss": 0.2946,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24423298239707947,
"step": 350,
"valid_targets_mean": 10679.9,
"valid_targets_min": 480
},
{
"epoch": 0.5986509274873525,
"grad_norm": 0.40483887095614407,
"learning_rate": 3.403846153846154e-05,
"loss": 0.261,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1261899769306183,
"step": 355,
"valid_targets_mean": 8614.1,
"valid_targets_min": 408
},
{
"epoch": 0.6070826306913997,
"grad_norm": 0.3922488584283575,
"learning_rate": 3.451923076923077e-05,
"loss": 0.2931,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15898962318897247,
"step": 360,
"valid_targets_mean": 8970.9,
"valid_targets_min": 3601
},
{
"epoch": 0.6155143338954469,
"grad_norm": 0.5799076928477381,
"learning_rate": 3.5000000000000004e-05,
"loss": 0.3169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1344633251428604,
"step": 365,
"valid_targets_mean": 5862.9,
"valid_targets_min": 326
},
{
"epoch": 0.6239460370994941,
"grad_norm": 0.3455705419094975,
"learning_rate": 3.548076923076923e-05,
"loss": 0.2712,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13601480424404144,
"step": 370,
"valid_targets_mean": 9698.6,
"valid_targets_min": 1987
},
{
"epoch": 0.6323777403035413,
"grad_norm": 0.4656835602219374,
"learning_rate": 3.596153846153846e-05,
"loss": 0.2762,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08869156241416931,
"step": 375,
"valid_targets_mean": 4697.8,
"valid_targets_min": 281
},
{
"epoch": 0.6408094435075885,
"grad_norm": 0.370038842023246,
"learning_rate": 3.6442307692307696e-05,
"loss": 0.2823,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14282071590423584,
"step": 380,
"valid_targets_mean": 13545.9,
"valid_targets_min": 4232
},
{
"epoch": 0.6492411467116358,
"grad_norm": 0.3471590069265448,
"learning_rate": 3.692307692307693e-05,
"loss": 0.2422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.114535853266716,
"step": 385,
"valid_targets_mean": 12262.2,
"valid_targets_min": 670
},
{
"epoch": 0.657672849915683,
"grad_norm": 0.33624284252036996,
"learning_rate": 3.7403846153846156e-05,
"loss": 0.2822,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1334913820028305,
"step": 390,
"valid_targets_mean": 15694.5,
"valid_targets_min": 7354
},
{
"epoch": 0.6661045531197302,
"grad_norm": 0.6275522268915146,
"learning_rate": 3.788461538461539e-05,
"loss": 0.3027,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2870558500289917,
"step": 395,
"valid_targets_mean": 12156.4,
"valid_targets_min": 3198
},
{
"epoch": 0.6745362563237775,
"grad_norm": 0.6853044078395961,
"learning_rate": 3.836538461538462e-05,
"loss": 0.3176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07187429070472717,
"step": 400,
"valid_targets_mean": 3046.5,
"valid_targets_min": 337
},
{
"epoch": 0.6829679595278246,
"grad_norm": 0.5158045814934572,
"learning_rate": 3.884615384615385e-05,
"loss": 0.2829,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20460447669029236,
"step": 405,
"valid_targets_mean": 7685.5,
"valid_targets_min": 341
},
{
"epoch": 0.6913996627318718,
"grad_norm": 0.3472704184795827,
"learning_rate": 3.932692307692308e-05,
"loss": 0.2825,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1535644829273224,
"step": 410,
"valid_targets_mean": 14589.9,
"valid_targets_min": 4397
},
{
"epoch": 0.6998313659359191,
"grad_norm": 0.44609082219875507,
"learning_rate": 3.9807692307692314e-05,
"loss": 0.3052,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1818748116493225,
"step": 415,
"valid_targets_mean": 10170.5,
"valid_targets_min": 5058
},
{
"epoch": 0.7082630691399663,
"grad_norm": 0.4056039188987142,
"learning_rate": 3.9999936326193684e-05,
"loss": 0.3077,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1235802099108696,
"step": 420,
"valid_targets_mean": 6829.8,
"valid_targets_min": 425
},
{
"epoch": 0.7166947723440135,
"grad_norm": 0.495248406558465,
"learning_rate": 3.9999547209956636e-05,
"loss": 0.3034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21021106839179993,
"step": 425,
"valid_targets_mean": 8639.9,
"valid_targets_min": 1592
},
{
"epoch": 0.7251264755480608,
"grad_norm": 0.33932203874046385,
"learning_rate": 3.99988043586934e-05,
"loss": 0.2769,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13366064429283142,
"step": 430,
"valid_targets_mean": 11280.4,
"valid_targets_min": 3475
},
{
"epoch": 0.7335581787521079,
"grad_norm": 0.32343033178896646,
"learning_rate": 3.9997707785542906e-05,
"loss": 0.3012,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13561661541461945,
"step": 435,
"valid_targets_mean": 14800.1,
"valid_targets_min": 5913
},
{
"epoch": 0.7419898819561551,
"grad_norm": 0.3581601753024619,
"learning_rate": 3.99962575099004e-05,
"loss": 0.2997,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1740327924489975,
"step": 440,
"valid_targets_mean": 13191.9,
"valid_targets_min": 366
},
{
"epoch": 0.7504215851602024,
"grad_norm": 0.5345261119909862,
"learning_rate": 3.999445355741713e-05,
"loss": 0.2934,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10350354015827179,
"step": 445,
"valid_targets_mean": 6364.1,
"valid_targets_min": 408
},
{
"epoch": 0.7588532883642496,
"grad_norm": 0.49017912866920044,
"learning_rate": 3.999229595999988e-05,
"loss": 0.3025,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14110887050628662,
"step": 450,
"valid_targets_mean": 6424.6,
"valid_targets_min": 294
},
{
"epoch": 0.7672849915682968,
"grad_norm": 0.5241374141499904,
"learning_rate": 3.9989784755810396e-05,
"loss": 0.2837,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12570369243621826,
"step": 455,
"valid_targets_mean": 5407.5,
"valid_targets_min": 1112
},
{
"epoch": 0.7757166947723441,
"grad_norm": 0.3619396161375343,
"learning_rate": 3.998691998926474e-05,
"loss": 0.2787,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11693060398101807,
"step": 460,
"valid_targets_mean": 7457.5,
"valid_targets_min": 836
},
{
"epoch": 0.7841483979763912,
"grad_norm": 0.4058596669434105,
"learning_rate": 3.998370171103247e-05,
"loss": 0.3263,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18975314497947693,
"step": 465,
"valid_targets_mean": 10665.6,
"valid_targets_min": 293
},
{
"epoch": 0.7925801011804384,
"grad_norm": 0.5036259810181237,
"learning_rate": 3.998012997803575e-05,
"loss": 0.2933,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09080538153648376,
"step": 470,
"valid_targets_mean": 6615.0,
"valid_targets_min": 461
},
{
"epoch": 0.8010118043844857,
"grad_norm": 0.3721888632659276,
"learning_rate": 3.9976204853448364e-05,
"loss": 0.2883,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12514978647232056,
"step": 475,
"valid_targets_mean": 10071.4,
"valid_targets_min": 400
},
{
"epoch": 0.8094435075885329,
"grad_norm": 0.4864349693874674,
"learning_rate": 3.99719264066946e-05,
"loss": 0.2979,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14365720748901367,
"step": 480,
"valid_targets_mean": 9047.9,
"valid_targets_min": 301
},
{
"epoch": 0.8178752107925801,
"grad_norm": 0.42768639289045784,
"learning_rate": 3.996729471344799e-05,
"loss": 0.2897,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1308303326368332,
"step": 485,
"valid_targets_mean": 6141.0,
"valid_targets_min": 358
},
{
"epoch": 0.8263069139966274,
"grad_norm": 0.4468150898842549,
"learning_rate": 3.9962309855629986e-05,
"loss": 0.3124,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1209115982055664,
"step": 490,
"valid_targets_mean": 6994.2,
"valid_targets_min": 1649
},
{
"epoch": 0.8347386172006745,
"grad_norm": 0.48270199920599527,
"learning_rate": 3.9956971921408536e-05,
"loss": 0.2751,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18245559930801392,
"step": 495,
"valid_targets_mean": 8594.0,
"valid_targets_min": 916
},
{
"epoch": 0.8431703204047217,
"grad_norm": 0.48374297943081757,
"learning_rate": 3.9951281005196486e-05,
"loss": 0.2812,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10876193642616272,
"step": 500,
"valid_targets_mean": 8478.4,
"valid_targets_min": 322
},
{
"epoch": 0.851602023608769,
"grad_norm": 0.4768340838074662,
"learning_rate": 3.994523720764994e-05,
"loss": 0.2939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11042595654726028,
"step": 505,
"valid_targets_mean": 6606.9,
"valid_targets_min": 480
},
{
"epoch": 0.8600337268128162,
"grad_norm": 0.4211313114119396,
"learning_rate": 3.9938840635666454e-05,
"loss": 0.2876,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16031618416309357,
"step": 510,
"valid_targets_mean": 7723.9,
"valid_targets_min": 417
},
{
"epoch": 0.8684654300168634,
"grad_norm": 0.36260680145366225,
"learning_rate": 3.9932091402383177e-05,
"loss": 0.2809,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13239243626594543,
"step": 515,
"valid_targets_mean": 8837.8,
"valid_targets_min": 380
},
{
"epoch": 0.8768971332209107,
"grad_norm": 0.39921032663295486,
"learning_rate": 3.992498962717482e-05,
"loss": 0.2902,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11992882937192917,
"step": 520,
"valid_targets_mean": 7356.0,
"valid_targets_min": 298
},
{
"epoch": 0.8853288364249579,
"grad_norm": 0.4441914351777187,
"learning_rate": 3.991753543565154e-05,
"loss": 0.2791,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13742601871490479,
"step": 525,
"valid_targets_mean": 8246.5,
"valid_targets_min": 1017
},
{
"epoch": 0.893760539629005,
"grad_norm": 0.38881209631352615,
"learning_rate": 3.9909728959656784e-05,
"loss": 0.2665,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1489628255367279,
"step": 530,
"valid_targets_mean": 10495.8,
"valid_targets_min": 717
},
{
"epoch": 0.9021922428330523,
"grad_norm": 0.6556634958333262,
"learning_rate": 3.990157033726486e-05,
"loss": 0.2773,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13200673460960388,
"step": 535,
"valid_targets_mean": 5240.0,
"valid_targets_min": 357
},
{
"epoch": 0.9106239460370995,
"grad_norm": 0.3972308918515509,
"learning_rate": 3.9893059712778564e-05,
"loss": 0.2717,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11547436565160751,
"step": 540,
"valid_targets_mean": 6159.5,
"valid_targets_min": 526
},
{
"epoch": 0.9190556492411467,
"grad_norm": 0.43637253018320094,
"learning_rate": 3.988419723672659e-05,
"loss": 0.3064,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17712107300758362,
"step": 545,
"valid_targets_mean": 7416.1,
"valid_targets_min": 702
},
{
"epoch": 0.927487352445194,
"grad_norm": 0.48708780993587675,
"learning_rate": 3.9874983065860925e-05,
"loss": 0.2886,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1344929039478302,
"step": 550,
"valid_targets_mean": 6310.8,
"valid_targets_min": 560
},
{
"epoch": 0.9359190556492412,
"grad_norm": 0.4158960883689957,
"learning_rate": 3.9865417363153985e-05,
"loss": 0.2706,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15958233177661896,
"step": 555,
"valid_targets_mean": 10025.2,
"valid_targets_min": 606
},
{
"epoch": 0.9443507588532883,
"grad_norm": 0.448891659644444,
"learning_rate": 3.985550029779581e-05,
"loss": 0.2791,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16829639673233032,
"step": 560,
"valid_targets_mean": 8129.9,
"valid_targets_min": 954
},
{
"epoch": 0.9527824620573356,
"grad_norm": 0.37824873620047234,
"learning_rate": 3.984523204519104e-05,
"loss": 0.2705,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12859921157360077,
"step": 565,
"valid_targets_mean": 11816.6,
"valid_targets_min": 809
},
{
"epoch": 0.9612141652613828,
"grad_norm": 0.3252086645984893,
"learning_rate": 3.983461278695582e-05,
"loss": 0.295,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12258483469486237,
"step": 570,
"valid_targets_mean": 11992.9,
"valid_targets_min": 1618
},
{
"epoch": 0.96964586846543,
"grad_norm": 0.3364280967862715,
"learning_rate": 3.9823642710914574e-05,
"loss": 0.2674,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1522078812122345,
"step": 575,
"valid_targets_mean": 10680.9,
"valid_targets_min": 470
},
{
"epoch": 0.9780775716694773,
"grad_norm": 0.41088830829568085,
"learning_rate": 3.981232201109672e-05,
"loss": 0.2661,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14503081142902374,
"step": 580,
"valid_targets_mean": 9152.5,
"valid_targets_min": 997
},
{
"epoch": 0.9865092748735245,
"grad_norm": 0.4495718088116054,
"learning_rate": 3.980065088773317e-05,
"loss": 0.2663,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12942713499069214,
"step": 585,
"valid_targets_mean": 9048.1,
"valid_targets_min": 963
},
{
"epoch": 0.9949409780775716,
"grad_norm": 0.31625013276386793,
"learning_rate": 3.978862954725286e-05,
"loss": 0.2775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13362139463424683,
"step": 590,
"valid_targets_mean": 12946.0,
"valid_targets_min": 1879
},
{
"epoch": 1.0033726812816188,
"grad_norm": 0.4086794978706987,
"learning_rate": 3.977625820227908e-05,
"loss": 0.2817,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14433297514915466,
"step": 595,
"valid_targets_mean": 7765.5,
"valid_targets_min": 3078
},
{
"epoch": 1.0118043844856661,
"grad_norm": 0.531377272532303,
"learning_rate": 3.976353707162566e-05,
"loss": 0.2477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10834267735481262,
"step": 600,
"valid_targets_mean": 9053.9,
"valid_targets_min": 442
},
{
"epoch": 1.0202360876897134,
"grad_norm": 0.5469161869078016,
"learning_rate": 3.9750466380293194e-05,
"loss": 0.2552,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14577242732048035,
"step": 605,
"valid_targets_mean": 5780.5,
"valid_targets_min": 846
},
{
"epoch": 1.0286677908937605,
"grad_norm": 0.4406295188135976,
"learning_rate": 3.973704635946496e-05,
"loss": 0.2739,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1358659565448761,
"step": 610,
"valid_targets_mean": 8506.8,
"valid_targets_min": 326
},
{
"epoch": 1.0370994940978078,
"grad_norm": 0.40830057795355124,
"learning_rate": 3.972327724650292e-05,
"loss": 0.2573,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12361744046211243,
"step": 615,
"valid_targets_mean": 7929.0,
"valid_targets_min": 998
},
{
"epoch": 1.045531197301855,
"grad_norm": 0.46518023435730416,
"learning_rate": 3.970915928494345e-05,
"loss": 0.2678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12349513173103333,
"step": 620,
"valid_targets_mean": 8301.1,
"valid_targets_min": 336
},
{
"epoch": 1.0539629005059021,
"grad_norm": 0.7385293753723227,
"learning_rate": 3.969469272449309e-05,
"loss": 0.2779,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1458127349615097,
"step": 625,
"valid_targets_mean": 11355.5,
"valid_targets_min": 7497
},
{
"epoch": 1.0623946037099494,
"grad_norm": 0.46536339139049315,
"learning_rate": 3.9679877821024085e-05,
"loss": 0.276,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12034212797880173,
"step": 630,
"valid_targets_mean": 6638.2,
"valid_targets_min": 293
},
{
"epoch": 1.0708263069139967,
"grad_norm": 0.44673112244724983,
"learning_rate": 3.9664714836569886e-05,
"loss": 0.2733,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1443609595298767,
"step": 635,
"valid_targets_mean": 7901.8,
"valid_targets_min": 706
},
{
"epoch": 1.0792580101180438,
"grad_norm": 0.4149634259908432,
"learning_rate": 3.964920403932049e-05,
"loss": 0.2728,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11244809627532959,
"step": 640,
"valid_targets_mean": 8597.0,
"valid_targets_min": 1115
},
{
"epoch": 1.087689713322091,
"grad_norm": 0.35941321517624936,
"learning_rate": 3.963334570361773e-05,
"loss": 0.2709,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16259606182575226,
"step": 645,
"valid_targets_mean": 11493.0,
"valid_targets_min": 783
},
{
"epoch": 1.0961214165261384,
"grad_norm": 0.4588759882130647,
"learning_rate": 3.9617140109950386e-05,
"loss": 0.2823,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12466984987258911,
"step": 650,
"valid_targets_mean": 4532.2,
"valid_targets_min": 925
},
{
"epoch": 1.1045531197301854,
"grad_norm": 0.38759449134947127,
"learning_rate": 3.9600587544949264e-05,
"loss": 0.2621,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10265229642391205,
"step": 655,
"valid_targets_mean": 5710.9,
"valid_targets_min": 333
},
{
"epoch": 1.1129848229342327,
"grad_norm": 0.4052464587985377,
"learning_rate": 3.958368830138209e-05,
"loss": 0.2581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10479821264743805,
"step": 660,
"valid_targets_mean": 6957.2,
"valid_targets_min": 634
},
{
"epoch": 1.12141652613828,
"grad_norm": 0.3906662278632303,
"learning_rate": 3.9566442678148336e-05,
"loss": 0.2883,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10750792920589447,
"step": 665,
"valid_targets_mean": 5905.9,
"valid_targets_min": 732
},
{
"epoch": 1.129848229342327,
"grad_norm": 0.45255978199163177,
"learning_rate": 3.9548850980273974e-05,
"loss": 0.2673,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16962644457817078,
"step": 670,
"valid_targets_mean": 7284.1,
"valid_targets_min": 796
},
{
"epoch": 1.1382799325463744,
"grad_norm": 0.3504412436456499,
"learning_rate": 3.9530913518906024e-05,
"loss": 0.2555,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14275503158569336,
"step": 675,
"valid_targets_mean": 11249.0,
"valid_targets_min": 2427
},
{
"epoch": 1.1467116357504217,
"grad_norm": 0.4001267364533494,
"learning_rate": 3.9512630611307094e-05,
"loss": 0.2674,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13394565880298615,
"step": 680,
"valid_targets_mean": 9590.2,
"valid_targets_min": 530
},
{
"epoch": 1.1551433389544687,
"grad_norm": 0.3892461759546036,
"learning_rate": 3.949400258084974e-05,
"loss": 0.2666,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11819396913051605,
"step": 685,
"valid_targets_mean": 7665.6,
"valid_targets_min": 804
},
{
"epoch": 1.163575042158516,
"grad_norm": 0.48480294672272733,
"learning_rate": 3.9475029757010757e-05,
"loss": 0.2641,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18022677302360535,
"step": 690,
"valid_targets_mean": 9717.2,
"valid_targets_min": 566
},
{
"epoch": 1.1720067453625633,
"grad_norm": 0.4667586921560271,
"learning_rate": 3.9455712475365375e-05,
"loss": 0.2581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10755648463964462,
"step": 695,
"valid_targets_mean": 8896.0,
"valid_targets_min": 345
},
{
"epoch": 1.1804384485666104,
"grad_norm": 0.39210941883931105,
"learning_rate": 3.943605107758128e-05,
"loss": 0.2559,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11234095692634583,
"step": 700,
"valid_targets_mean": 6575.5,
"valid_targets_min": 259
},
{
"epoch": 1.1888701517706577,
"grad_norm": 0.4956198197840669,
"learning_rate": 3.941604591141259e-05,
"loss": 0.2891,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15952013432979584,
"step": 705,
"valid_targets_mean": 7830.8,
"valid_targets_min": 588
},
{
"epoch": 1.197301854974705,
"grad_norm": 0.39855501357685447,
"learning_rate": 3.9395697330693716e-05,
"loss": 0.255,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11256325244903564,
"step": 710,
"valid_targets_mean": 7493.6,
"valid_targets_min": 358
},
{
"epoch": 1.205733558178752,
"grad_norm": 0.42976074090126715,
"learning_rate": 3.9375005695333096e-05,
"loss": 0.2764,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10012060403823853,
"step": 715,
"valid_targets_mean": 4766.8,
"valid_targets_min": 903
},
{
"epoch": 1.2141652613827993,
"grad_norm": 0.35320193580943593,
"learning_rate": 3.935397137130683e-05,
"loss": 0.2552,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09375806152820587,
"step": 720,
"valid_targets_mean": 8264.0,
"valid_targets_min": 767
},
{
"epoch": 1.2225969645868466,
"grad_norm": 0.3552798626518043,
"learning_rate": 3.933259473065219e-05,
"loss": 0.2584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07506534457206726,
"step": 725,
"valid_targets_mean": 4044.2,
"valid_targets_min": 406
},
{
"epoch": 1.2310286677908937,
"grad_norm": 0.3279744763698071,
"learning_rate": 3.9310876151461076e-05,
"loss": 0.2674,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11524611711502075,
"step": 730,
"valid_targets_mean": 11077.2,
"valid_targets_min": 3133
},
{
"epoch": 1.239460370994941,
"grad_norm": 0.4197193685482744,
"learning_rate": 3.9288816017873276e-05,
"loss": 0.2502,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14202597737312317,
"step": 735,
"valid_targets_mean": 8796.4,
"valid_targets_min": 2948
},
{
"epoch": 1.2478920741989883,
"grad_norm": 0.5244827723249347,
"learning_rate": 3.926641472006973e-05,
"loss": 0.2705,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13600367307662964,
"step": 740,
"valid_targets_mean": 11799.1,
"valid_targets_min": 2080
},
{
"epoch": 1.2563237774030354,
"grad_norm": 0.4452162802267489,
"learning_rate": 3.9243672654265584e-05,
"loss": 0.2671,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14343160390853882,
"step": 745,
"valid_targets_mean": 5635.2,
"valid_targets_min": 1453
},
{
"epoch": 1.2647554806070826,
"grad_norm": 0.43088680911544774,
"learning_rate": 3.922059022270322e-05,
"loss": 0.2652,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13595256209373474,
"step": 750,
"valid_targets_mean": 5123.9,
"valid_targets_min": 455
},
{
"epoch": 1.27318718381113,
"grad_norm": 0.45710506860573047,
"learning_rate": 3.919716783364509e-05,
"loss": 0.2918,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1817789077758789,
"step": 755,
"valid_targets_mean": 8357.0,
"valid_targets_min": 415
},
{
"epoch": 1.281618887015177,
"grad_norm": 0.3382425264276293,
"learning_rate": 3.917340590136656e-05,
"loss": 0.2562,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13356652855873108,
"step": 760,
"valid_targets_mean": 12070.2,
"valid_targets_min": 809
},
{
"epoch": 1.2900505902192243,
"grad_norm": 0.3645508103646137,
"learning_rate": 3.914930484614853e-05,
"loss": 0.2523,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13618983328342438,
"step": 765,
"valid_targets_mean": 9179.8,
"valid_targets_min": 404
},
{
"epoch": 1.2984822934232714,
"grad_norm": 0.44506810019376114,
"learning_rate": 3.912486509427e-05,
"loss": 0.2812,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1779347062110901,
"step": 770,
"valid_targets_mean": 8573.1,
"valid_targets_min": 968
},
{
"epoch": 1.3069139966273187,
"grad_norm": 0.34517379735534626,
"learning_rate": 3.91000870780006e-05,
"loss": 0.262,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10514889657497406,
"step": 775,
"valid_targets_mean": 9975.4,
"valid_targets_min": 382
},
{
"epoch": 1.315345699831366,
"grad_norm": 0.4672966212039155,
"learning_rate": 3.9074971235592845e-05,
"loss": 0.2477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14455226063728333,
"step": 780,
"valid_targets_mean": 10837.4,
"valid_targets_min": 2223
},
{
"epoch": 1.3237774030354132,
"grad_norm": 0.33589833915681067,
"learning_rate": 3.904951801127445e-05,
"loss": 0.2335,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1113758236169815,
"step": 785,
"valid_targets_mean": 10369.4,
"valid_targets_min": 1960
},
{
"epoch": 1.3322091062394603,
"grad_norm": 0.41505796496484565,
"learning_rate": 3.902372785524047e-05,
"loss": 0.2478,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11151224374771118,
"step": 790,
"valid_targets_mean": 9147.4,
"valid_targets_min": 436
},
{
"epoch": 1.3406408094435076,
"grad_norm": 0.35362193015989385,
"learning_rate": 3.899760122364529e-05,
"loss": 0.2559,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1082097664475441,
"step": 795,
"valid_targets_mean": 9177.0,
"valid_targets_min": 694
},
{
"epoch": 1.3490725126475547,
"grad_norm": 0.399434241991854,
"learning_rate": 3.8971138578594614e-05,
"loss": 0.2679,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10963332653045654,
"step": 800,
"valid_targets_mean": 6511.8,
"valid_targets_min": 549
},
{
"epoch": 1.357504215851602,
"grad_norm": 0.41789077878526437,
"learning_rate": 3.894434038813727e-05,
"loss": 0.2394,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10442526638507843,
"step": 805,
"valid_targets_mean": 6595.8,
"valid_targets_min": 365
},
{
"epoch": 1.3659359190556493,
"grad_norm": 0.32750530474164086,
"learning_rate": 3.891720712625691e-05,
"loss": 0.2443,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12113337963819504,
"step": 810,
"valid_targets_mean": 10112.2,
"valid_targets_min": 520
},
{
"epoch": 1.3743676222596966,
"grad_norm": 0.4400244122230931,
"learning_rate": 3.8889739272863656e-05,
"loss": 0.2579,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10030785202980042,
"step": 815,
"valid_targets_mean": 5051.5,
"valid_targets_min": 222
},
{
"epoch": 1.3827993254637436,
"grad_norm": 0.39769323582525967,
"learning_rate": 3.8861937313785594e-05,
"loss": 0.2659,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14163155853748322,
"step": 820,
"valid_targets_mean": 10737.9,
"valid_targets_min": 344
},
{
"epoch": 1.391231028667791,
"grad_norm": 0.3761948323757739,
"learning_rate": 3.883380174076019e-05,
"loss": 0.2695,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1856262981891632,
"step": 825,
"valid_targets_mean": 12689.9,
"valid_targets_min": 888
},
{
"epoch": 1.399662731871838,
"grad_norm": 0.36493042129332787,
"learning_rate": 3.880533305142561e-05,
"loss": 0.2581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13061369955539703,
"step": 830,
"valid_targets_mean": 9754.2,
"valid_targets_min": 351
},
{
"epoch": 1.4080944350758853,
"grad_norm": 0.4617319031043397,
"learning_rate": 3.8776531749311876e-05,
"loss": 0.2474,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10686546564102173,
"step": 835,
"valid_targets_mean": 6909.0,
"valid_targets_min": 376
},
{
"epoch": 1.4165261382799326,
"grad_norm": 0.3455386084444083,
"learning_rate": 3.874739834383198e-05,
"loss": 0.2653,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11929252743721008,
"step": 840,
"valid_targets_mean": 8891.5,
"valid_targets_min": 792
},
{
"epoch": 1.4249578414839799,
"grad_norm": 0.4155551633924909,
"learning_rate": 3.871793335027289e-05,
"loss": 0.2436,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15894144773483276,
"step": 845,
"valid_targets_mean": 9947.6,
"valid_targets_min": 5068
},
{
"epoch": 1.433389544688027,
"grad_norm": 0.5682007014740678,
"learning_rate": 3.868813728978642e-05,
"loss": 0.2788,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12174543738365173,
"step": 850,
"valid_targets_mean": 6866.1,
"valid_targets_min": 624
},
{
"epoch": 1.4418212478920742,
"grad_norm": 0.38686445291667304,
"learning_rate": 3.8658010689380006e-05,
"loss": 0.2795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1785919964313507,
"step": 855,
"valid_targets_mean": 12040.2,
"valid_targets_min": 5105
},
{
"epoch": 1.4502529510961213,
"grad_norm": 0.38596085771693356,
"learning_rate": 3.862755408190741e-05,
"loss": 0.2765,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1205671951174736,
"step": 860,
"valid_targets_mean": 9189.2,
"valid_targets_min": 604
},
{
"epoch": 1.4586846543001686,
"grad_norm": 0.4030940289603823,
"learning_rate": 3.859676800605927e-05,
"loss": 0.265,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14316019415855408,
"step": 865,
"valid_targets_mean": 11254.5,
"valid_targets_min": 1863
},
{
"epoch": 1.4671163575042159,
"grad_norm": 0.5253535388278335,
"learning_rate": 3.856565300635357e-05,
"loss": 0.2651,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15311139822006226,
"step": 870,
"valid_targets_mean": 6861.6,
"valid_targets_min": 2461
},
{
"epoch": 1.4755480607082632,
"grad_norm": 0.454012982470006,
"learning_rate": 3.853420963312604e-05,
"loss": 0.2651,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12477818131446838,
"step": 875,
"valid_targets_mean": 6861.6,
"valid_targets_min": 979
},
{
"epoch": 1.4839797639123102,
"grad_norm": 0.4858203152713948,
"learning_rate": 3.850243844252039e-05,
"loss": 0.2753,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11629155278205872,
"step": 880,
"valid_targets_mean": 6302.1,
"valid_targets_min": 518
},
{
"epoch": 1.4924114671163575,
"grad_norm": 0.38202978922147174,
"learning_rate": 3.84703399964785e-05,
"loss": 0.2503,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10060980916023254,
"step": 885,
"valid_targets_mean": 7047.6,
"valid_targets_min": 1226
},
{
"epoch": 1.5008431703204046,
"grad_norm": 0.3944619245767814,
"learning_rate": 3.843791486273044e-05,
"loss": 0.249,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13788242638111115,
"step": 890,
"valid_targets_mean": 9494.1,
"valid_targets_min": 2762
},
{
"epoch": 1.5092748735244519,
"grad_norm": 0.5175959705697554,
"learning_rate": 3.840516361478448e-05,
"loss": 0.2689,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1204976812005043,
"step": 895,
"valid_targets_mean": 5397.2,
"valid_targets_min": 351
},
{
"epoch": 1.5177065767284992,
"grad_norm": 0.36406206830176197,
"learning_rate": 3.83720868319169e-05,
"loss": 0.2619,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08684238791465759,
"step": 900,
"valid_targets_mean": 4748.6,
"valid_targets_min": 325
},
{
"epoch": 1.5261382799325465,
"grad_norm": 0.4141581318321356,
"learning_rate": 3.8338685099161786e-05,
"loss": 0.2595,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12677925825119019,
"step": 905,
"valid_targets_mean": 9134.6,
"valid_targets_min": 337
},
{
"epoch": 1.5345699831365935,
"grad_norm": 0.37424254246833477,
"learning_rate": 3.830495900730066e-05,
"loss": 0.2661,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1666191816329956,
"step": 910,
"valid_targets_mean": 11004.0,
"valid_targets_min": 938
},
{
"epoch": 1.5430016863406408,
"grad_norm": 0.43848418303412,
"learning_rate": 3.827090915285202e-05,
"loss": 0.2465,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10807475447654724,
"step": 915,
"valid_targets_mean": 6321.6,
"valid_targets_min": 321
},
{
"epoch": 1.551433389544688,
"grad_norm": 0.42584527254316873,
"learning_rate": 3.823653613806081e-05,
"loss": 0.2516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1449190080165863,
"step": 920,
"valid_targets_mean": 7579.2,
"valid_targets_min": 668
},
{
"epoch": 1.5598650927487352,
"grad_norm": 0.617051389343634,
"learning_rate": 3.820184057088775e-05,
"loss": 0.2591,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11032834649085999,
"step": 925,
"valid_targets_mean": 8036.0,
"valid_targets_min": 396
},
{
"epoch": 1.5682967959527825,
"grad_norm": 0.3918057261501475,
"learning_rate": 3.8166823064998616e-05,
"loss": 0.2434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1440655142068863,
"step": 930,
"valid_targets_mean": 8829.9,
"valid_targets_min": 311
},
{
"epoch": 1.5767284991568298,
"grad_norm": 0.43036041868101893,
"learning_rate": 3.813148423975334e-05,
"loss": 0.2775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12346546351909637,
"step": 935,
"valid_targets_mean": 7158.2,
"valid_targets_min": 944
},
{
"epoch": 1.5851602023608768,
"grad_norm": 0.3617832859971793,
"learning_rate": 3.8095824720195084e-05,
"loss": 0.2544,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10149797797203064,
"step": 940,
"valid_targets_mean": 6945.8,
"valid_targets_min": 402
},
{
"epoch": 1.5935919055649241,
"grad_norm": 0.4624336807030225,
"learning_rate": 3.8059845137039185e-05,
"loss": 0.2581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13834244012832642,
"step": 945,
"valid_targets_mean": 7080.4,
"valid_targets_min": 308
},
{
"epoch": 1.6020236087689712,
"grad_norm": 0.5232035750373686,
"learning_rate": 3.8023546126661985e-05,
"loss": 0.2536,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13808363676071167,
"step": 950,
"valid_targets_mean": 7353.9,
"valid_targets_min": 337
},
{
"epoch": 1.6104553119730185,
"grad_norm": 0.4502188033315801,
"learning_rate": 3.79869283310896e-05,
"loss": 0.2524,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1706855744123459,
"step": 955,
"valid_targets_mean": 7502.1,
"valid_targets_min": 291
},
{
"epoch": 1.6188870151770658,
"grad_norm": 0.37935208777575025,
"learning_rate": 3.794999239798652e-05,
"loss": 0.2739,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10850772261619568,
"step": 960,
"valid_targets_mean": 9034.1,
"valid_targets_min": 825
},
{
"epoch": 1.627318718381113,
"grad_norm": 0.353470675725493,
"learning_rate": 3.7912738980644207e-05,
"loss": 0.2676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14227962493896484,
"step": 965,
"valid_targets_mean": 10924.2,
"valid_targets_min": 930
},
{
"epoch": 1.6357504215851602,
"grad_norm": 0.550912070286999,
"learning_rate": 3.787516873796952e-05,
"loss": 0.264,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1475759595632553,
"step": 970,
"valid_targets_mean": 10371.0,
"valid_targets_min": 3893
},
{
"epoch": 1.6441821247892074,
"grad_norm": 0.40196305394182497,
"learning_rate": 3.783728233447301e-05,
"loss": 0.2589,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11004417389631271,
"step": 975,
"valid_targets_mean": 7482.1,
"valid_targets_min": 456
},
{
"epoch": 1.6526138279932545,
"grad_norm": 0.43619391128793084,
"learning_rate": 3.7799080440257274e-05,
"loss": 0.2585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1114400252699852,
"step": 980,
"valid_targets_mean": 6542.5,
"valid_targets_min": 827
},
{
"epoch": 1.6610455311973018,
"grad_norm": 0.37358403531640966,
"learning_rate": 3.7760563731004996e-05,
"loss": 0.2596,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12266801297664642,
"step": 985,
"valid_targets_mean": 9331.2,
"valid_targets_min": 1551
},
{
"epoch": 1.669477234401349,
"grad_norm": 0.38113570878869735,
"learning_rate": 3.7721732887967064e-05,
"loss": 0.2513,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1315096616744995,
"step": 990,
"valid_targets_mean": 10204.2,
"valid_targets_min": 1214
},
{
"epoch": 1.6779089376053964,
"grad_norm": 0.3874697957784334,
"learning_rate": 3.7682588597950486e-05,
"loss": 0.2607,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07295676320791245,
"step": 995,
"valid_targets_mean": 5080.4,
"valid_targets_min": 738
},
{
"epoch": 1.6863406408094435,
"grad_norm": 0.4048040646367245,
"learning_rate": 3.7643131553306265e-05,
"loss": 0.271,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15146978199481964,
"step": 1000,
"valid_targets_mean": 8687.5,
"valid_targets_min": 506
},
{
"epoch": 1.6947723440134908,
"grad_norm": 0.3088102071154041,
"learning_rate": 3.760336245191713e-05,
"loss": 0.2573,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1238112673163414,
"step": 1005,
"valid_targets_mean": 13495.8,
"valid_targets_min": 978
},
{
"epoch": 1.7032040472175378,
"grad_norm": 0.45321234529448673,
"learning_rate": 3.756328199718521e-05,
"loss": 0.2676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14644251763820648,
"step": 1010,
"valid_targets_mean": 7968.9,
"valid_targets_min": 636
},
{
"epoch": 1.7116357504215851,
"grad_norm": 0.4116938034812059,
"learning_rate": 3.7522890898019606e-05,
"loss": 0.275,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16921404004096985,
"step": 1015,
"valid_targets_mean": 9373.9,
"valid_targets_min": 831
},
{
"epoch": 1.7200674536256324,
"grad_norm": 0.4257742881827407,
"learning_rate": 3.74821898688238e-05,
"loss": 0.2605,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13965681195259094,
"step": 1020,
"valid_targets_mean": 7445.8,
"valid_targets_min": 386
},
{
"epoch": 1.7284991568296797,
"grad_norm": 0.5296707452699168,
"learning_rate": 3.744117962948308e-05,
"loss": 0.2598,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15512895584106445,
"step": 1025,
"valid_targets_mean": 9550.2,
"valid_targets_min": 347
},
{
"epoch": 1.7369308600337268,
"grad_norm": 0.41924028368143645,
"learning_rate": 3.7399860905351774e-05,
"loss": 0.2583,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12704220414161682,
"step": 1030,
"valid_targets_mean": 9305.8,
"valid_targets_min": 414
},
{
"epoch": 1.745362563237774,
"grad_norm": 0.3439601346938688,
"learning_rate": 3.735823442724041e-05,
"loss": 0.2397,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12454552948474884,
"step": 1035,
"valid_targets_mean": 11324.1,
"valid_targets_min": 963
},
{
"epoch": 1.7537942664418211,
"grad_norm": 0.40841599404438367,
"learning_rate": 3.731630093140284e-05,
"loss": 0.2589,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14869846403598785,
"step": 1040,
"valid_targets_mean": 11499.2,
"valid_targets_min": 752
},
{
"epoch": 1.7622259696458684,
"grad_norm": 0.28808993643642733,
"learning_rate": 3.727406115952316e-05,
"loss": 0.2561,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10977204889059067,
"step": 1045,
"valid_targets_mean": 10032.9,
"valid_targets_min": 297
},
{
"epoch": 1.7706576728499157,
"grad_norm": 0.4254650249214162,
"learning_rate": 3.723151585870263e-05,
"loss": 0.2664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06899794936180115,
"step": 1050,
"valid_targets_mean": 5476.8,
"valid_targets_min": 979
},
{
"epoch": 1.779089376053963,
"grad_norm": 0.43170439104254554,
"learning_rate": 3.718866578144643e-05,
"loss": 0.2638,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12306337058544159,
"step": 1055,
"valid_targets_mean": 7639.1,
"valid_targets_min": 954
},
{
"epoch": 1.78752107925801,
"grad_norm": 0.3931205703458597,
"learning_rate": 3.7145511685650384e-05,
"loss": 0.2643,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13113872706890106,
"step": 1060,
"valid_targets_mean": 8158.8,
"valid_targets_min": 905
},
{
"epoch": 1.7959527824620574,
"grad_norm": 0.4442848310904727,
"learning_rate": 3.710205433458754e-05,
"loss": 0.2595,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08252847194671631,
"step": 1065,
"valid_targets_mean": 6959.9,
"valid_targets_min": 313
},
{
"epoch": 1.8043844856661044,
"grad_norm": 0.3653847035020948,
"learning_rate": 3.7058294496894654e-05,
"loss": 0.2586,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13877543807029724,
"step": 1070,
"valid_targets_mean": 10058.1,
"valid_targets_min": 1016
},
{
"epoch": 1.8128161888701517,
"grad_norm": 0.4060756156547807,
"learning_rate": 3.701423294655862e-05,
"loss": 0.2508,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16855351626873016,
"step": 1075,
"valid_targets_mean": 8304.2,
"valid_targets_min": 1043
},
{
"epoch": 1.821247892074199,
"grad_norm": 0.3623407879266793,
"learning_rate": 3.696987046290275e-05,
"loss": 0.2643,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11014987528324127,
"step": 1080,
"valid_targets_mean": 8889.1,
"valid_targets_min": 502
},
{
"epoch": 1.8296795952782463,
"grad_norm": 0.6702584879528396,
"learning_rate": 3.692520783057304e-05,
"loss": 0.2501,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08813249319791794,
"step": 1085,
"valid_targets_mean": 5269.0,
"valid_targets_min": 325
},
{
"epoch": 1.8381112984822934,
"grad_norm": 0.4300736478674192,
"learning_rate": 3.688024583952424e-05,
"loss": 0.2594,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09172368794679642,
"step": 1090,
"valid_targets_mean": 5124.2,
"valid_targets_min": 358
},
{
"epoch": 1.8465430016863407,
"grad_norm": 0.3937632121196448,
"learning_rate": 3.683498528500591e-05,
"loss": 0.2642,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1894649863243103,
"step": 1095,
"valid_targets_mean": 10807.6,
"valid_targets_min": 416
},
{
"epoch": 1.8549747048903877,
"grad_norm": 0.4993468960085093,
"learning_rate": 3.6789426967548336e-05,
"loss": 0.2623,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09895464777946472,
"step": 1100,
"valid_targets_mean": 5041.5,
"valid_targets_min": 991
},
{
"epoch": 1.863406408094435,
"grad_norm": 0.4043071364973763,
"learning_rate": 3.6743571692948404e-05,
"loss": 0.2743,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13687078654766083,
"step": 1105,
"valid_targets_mean": 9585.1,
"valid_targets_min": 671
},
{
"epoch": 1.8718381112984823,
"grad_norm": 0.4279505676538887,
"learning_rate": 3.669742027225531e-05,
"loss": 0.2423,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1596437245607376,
"step": 1110,
"valid_targets_mean": 11203.6,
"valid_targets_min": 1427
},
{
"epoch": 1.8802698145025296,
"grad_norm": 0.42828704109613036,
"learning_rate": 3.665097352175623e-05,
"loss": 0.2566,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16299262642860413,
"step": 1115,
"valid_targets_mean": 9912.1,
"valid_targets_min": 444
},
{
"epoch": 1.8887015177065767,
"grad_norm": 0.379019714561486,
"learning_rate": 3.660423226296189e-05,
"loss": 0.2453,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13470494747161865,
"step": 1120,
"valid_targets_mean": 10075.8,
"valid_targets_min": 5917
},
{
"epoch": 1.897133220910624,
"grad_norm": 0.38820825028867945,
"learning_rate": 3.6557197322592025e-05,
"loss": 0.2543,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11981119215488434,
"step": 1125,
"valid_targets_mean": 6938.1,
"valid_targets_min": 480
},
{
"epoch": 1.905564924114671,
"grad_norm": 0.43216509186117696,
"learning_rate": 3.6509869532560774e-05,
"loss": 0.2589,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10838529467582703,
"step": 1130,
"valid_targets_mean": 4512.1,
"valid_targets_min": 409
},
{
"epoch": 1.9139966273187183,
"grad_norm": 0.3489952111326072,
"learning_rate": 3.646224972996193e-05,
"loss": 0.2259,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13144941627979279,
"step": 1135,
"valid_targets_mean": 11339.9,
"valid_targets_min": 298
},
{
"epoch": 1.9224283305227656,
"grad_norm": 0.4887893957379976,
"learning_rate": 3.641433875705416e-05,
"loss": 0.2475,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08402438461780548,
"step": 1140,
"valid_targets_mean": 4971.4,
"valid_targets_min": 346
},
{
"epoch": 1.930860033726813,
"grad_norm": 0.3530167611870837,
"learning_rate": 3.6366137461246125e-05,
"loss": 0.2567,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12010815739631653,
"step": 1145,
"valid_targets_mean": 7790.6,
"valid_targets_min": 3622
},
{
"epoch": 1.93929173693086,
"grad_norm": 0.34737101474998555,
"learning_rate": 3.6317646695081456e-05,
"loss": 0.2622,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15191681683063507,
"step": 1150,
"valid_targets_mean": 16865.1,
"valid_targets_min": 7486
},
{
"epoch": 1.9477234401349073,
"grad_norm": 0.39244966265849895,
"learning_rate": 3.626886731622369e-05,
"loss": 0.2551,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10733728110790253,
"step": 1155,
"valid_targets_mean": 6567.4,
"valid_targets_min": 469
},
{
"epoch": 1.9561551433389543,
"grad_norm": 0.42843991272304527,
"learning_rate": 3.62198001874411e-05,
"loss": 0.2703,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09612288326025009,
"step": 1160,
"valid_targets_mean": 5330.1,
"valid_targets_min": 425
},
{
"epoch": 1.9645868465430016,
"grad_norm": 0.3636856534965485,
"learning_rate": 3.617044617659144e-05,
"loss": 0.2393,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12395919114351273,
"step": 1165,
"valid_targets_mean": 11198.0,
"valid_targets_min": 6435
},
{
"epoch": 1.973018549747049,
"grad_norm": 0.34470711338480514,
"learning_rate": 3.612080615660659e-05,
"loss": 0.2377,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07975724339485168,
"step": 1170,
"valid_targets_mean": 7043.8,
"valid_targets_min": 337
},
{
"epoch": 1.9814502529510962,
"grad_norm": 0.47677922288791613,
"learning_rate": 3.607088100547712e-05,
"loss": 0.2655,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12454728037118912,
"step": 1175,
"valid_targets_mean": 7636.0,
"valid_targets_min": 293
},
{
"epoch": 1.9898819561551433,
"grad_norm": 0.47000117167387817,
"learning_rate": 3.602067160623674e-05,
"loss": 0.2525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1344081461429596,
"step": 1180,
"valid_targets_mean": 7248.4,
"valid_targets_min": 362
},
{
"epoch": 1.9983136593591906,
"grad_norm": 0.4878808720162569,
"learning_rate": 3.597017884694673e-05,
"loss": 0.2659,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11134347319602966,
"step": 1185,
"valid_targets_mean": 5186.8,
"valid_targets_min": 383
},
{
"epoch": 2.0067453625632377,
"grad_norm": 0.42778116752181633,
"learning_rate": 3.591940362068017e-05,
"loss": 0.2224,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09581497311592102,
"step": 1190,
"valid_targets_mean": 6296.5,
"valid_targets_min": 336
},
{
"epoch": 2.015177065767285,
"grad_norm": 0.5332014550946883,
"learning_rate": 3.58683468255062e-05,
"loss": 0.2367,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12219070643186569,
"step": 1195,
"valid_targets_mean": 9144.5,
"valid_targets_min": 4232
},
{
"epoch": 2.0236087689713322,
"grad_norm": 0.40362443515749874,
"learning_rate": 3.5817009364474094e-05,
"loss": 0.2347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08650688081979752,
"step": 1200,
"valid_targets_mean": 5156.5,
"valid_targets_min": 356
},
{
"epoch": 2.0320404721753795,
"grad_norm": 0.4038075778302946,
"learning_rate": 3.576539214559733e-05,
"loss": 0.2435,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15470555424690247,
"step": 1205,
"valid_targets_mean": 12063.8,
"valid_targets_min": 3480
},
{
"epoch": 2.040472175379427,
"grad_norm": 0.3857470735302077,
"learning_rate": 3.571349608183749e-05,
"loss": 0.2356,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07464306056499481,
"step": 1210,
"valid_targets_mean": 6025.9,
"valid_targets_min": 324
},
{
"epoch": 2.0489038785834737,
"grad_norm": 0.43294227018948844,
"learning_rate": 3.566132209108811e-05,
"loss": 0.2553,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07677531987428665,
"step": 1215,
"valid_targets_mean": 5413.2,
"valid_targets_min": 549
},
{
"epoch": 2.057335581787521,
"grad_norm": 0.5196418448866507,
"learning_rate": 3.560887109615849e-05,
"loss": 0.2263,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09884986281394958,
"step": 1220,
"valid_targets_mean": 4875.6,
"valid_targets_min": 1061
},
{
"epoch": 2.0657672849915683,
"grad_norm": 0.39204656552573625,
"learning_rate": 3.5556144024757354e-05,
"loss": 0.2348,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09521162509918213,
"step": 1225,
"valid_targets_mean": 5792.0,
"valid_targets_min": 357
},
{
"epoch": 2.0741989881956155,
"grad_norm": 0.42196011022231245,
"learning_rate": 3.550314180947639e-05,
"loss": 0.2436,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11704714596271515,
"step": 1230,
"valid_targets_mean": 6740.1,
"valid_targets_min": 307
},
{
"epoch": 2.082630691399663,
"grad_norm": 0.4513823157681111,
"learning_rate": 3.544986538777383e-05,
"loss": 0.2451,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21808815002441406,
"step": 1235,
"valid_targets_mean": 11486.5,
"valid_targets_min": 3727
},
{
"epoch": 2.09106239460371,
"grad_norm": 0.4074008564872491,
"learning_rate": 3.539631570195783e-05,
"loss": 0.2151,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08867260068655014,
"step": 1240,
"valid_targets_mean": 7512.5,
"valid_targets_min": 351
},
{
"epoch": 2.099494097807757,
"grad_norm": 0.4701622279643828,
"learning_rate": 3.534249369916981e-05,
"loss": 0.2469,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14414334297180176,
"step": 1245,
"valid_targets_mean": 8328.4,
"valid_targets_min": 1454
},
{
"epoch": 2.1079258010118043,
"grad_norm": 0.4747155712417444,
"learning_rate": 3.5288400331367704e-05,
"loss": 0.23,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15197674930095673,
"step": 1250,
"valid_targets_mean": 10064.8,
"valid_targets_min": 2997
},
{
"epoch": 2.1163575042158516,
"grad_norm": 0.4422628226214092,
"learning_rate": 3.52340365553091e-05,
"loss": 0.2389,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13354700803756714,
"step": 1255,
"valid_targets_mean": 7158.2,
"valid_targets_min": 331
},
{
"epoch": 2.124789207419899,
"grad_norm": 0.31313363098013364,
"learning_rate": 3.517940333253438e-05,
"loss": 0.2268,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09569864720106125,
"step": 1260,
"valid_targets_mean": 14181.9,
"valid_targets_min": 5435
},
{
"epoch": 2.133220910623946,
"grad_norm": 0.4711973267556688,
"learning_rate": 3.512450162934962e-05,
"loss": 0.2509,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09335250407457352,
"step": 1265,
"valid_targets_mean": 5510.4,
"valid_targets_min": 404
},
{
"epoch": 2.1416526138279934,
"grad_norm": 0.4192438355410711,
"learning_rate": 3.5069332416809586e-05,
"loss": 0.2488,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1286272406578064,
"step": 1270,
"valid_targets_mean": 7201.1,
"valid_targets_min": 409
},
{
"epoch": 2.1500843170320403,
"grad_norm": 0.36121572104819333,
"learning_rate": 3.501389667070051e-05,
"loss": 0.2324,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10193075984716415,
"step": 1275,
"valid_targets_mean": 8809.8,
"valid_targets_min": 340
},
{
"epoch": 2.1585160202360876,
"grad_norm": 0.4970124962922364,
"learning_rate": 3.4958195371522854e-05,
"loss": 0.2362,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1110304594039917,
"step": 1280,
"valid_targets_mean": 4530.2,
"valid_targets_min": 1078
},
{
"epoch": 2.166947723440135,
"grad_norm": 0.39166210712072147,
"learning_rate": 3.490222950447393e-05,
"loss": 0.2313,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13261935114860535,
"step": 1285,
"valid_targets_mean": 11243.1,
"valid_targets_min": 917
},
{
"epoch": 2.175379426644182,
"grad_norm": 0.36414968376545864,
"learning_rate": 3.484600005943054e-05,
"loss": 0.241,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12285901606082916,
"step": 1290,
"valid_targets_mean": 10942.2,
"valid_targets_min": 2530
},
{
"epoch": 2.1838111298482294,
"grad_norm": 0.3682650459030714,
"learning_rate": 3.478950803093139e-05,
"loss": 0.2302,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11753889918327332,
"step": 1295,
"valid_targets_mean": 10945.0,
"valid_targets_min": 1214
},
{
"epoch": 2.1922428330522767,
"grad_norm": 0.42791583013870965,
"learning_rate": 3.473275441815958e-05,
"loss": 0.2482,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16097962856292725,
"step": 1300,
"valid_targets_mean": 8892.6,
"valid_targets_min": 269
},
{
"epoch": 2.2006745362563236,
"grad_norm": 0.47967583289541527,
"learning_rate": 3.467574022492486e-05,
"loss": 0.2368,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08637973666191101,
"step": 1305,
"valid_targets_mean": 6649.8,
"valid_targets_min": 738
},
{
"epoch": 2.209106239460371,
"grad_norm": 0.5397656034343836,
"learning_rate": 3.4618466459645926e-05,
"loss": 0.2363,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13329580426216125,
"step": 1310,
"valid_targets_mean": 6628.6,
"valid_targets_min": 1041
},
{
"epoch": 2.217537942664418,
"grad_norm": 0.3132329309997544,
"learning_rate": 3.4560934135332544e-05,
"loss": 0.214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0959177017211914,
"step": 1315,
"valid_targets_mean": 10313.0,
"valid_targets_min": 420
},
{
"epoch": 2.2259696458684655,
"grad_norm": 0.49934744334683756,
"learning_rate": 3.4503144269567665e-05,
"loss": 0.2363,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11804713308811188,
"step": 1320,
"valid_targets_mean": 6283.4,
"valid_targets_min": 456
},
{
"epoch": 2.2344013490725128,
"grad_norm": 0.576617072939063,
"learning_rate": 3.444509788448942e-05,
"loss": 0.2366,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10578224807977676,
"step": 1325,
"valid_targets_mean": 9095.5,
"valid_targets_min": 293
},
{
"epoch": 2.24283305227656,
"grad_norm": 0.26810749955326857,
"learning_rate": 3.438679600677303e-05,
"loss": 0.2118,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08233433961868286,
"step": 1330,
"valid_targets_mean": 14900.1,
"valid_targets_min": 2911
},
{
"epoch": 2.251264755480607,
"grad_norm": 0.3560222465235831,
"learning_rate": 3.432823966761264e-05,
"loss": 0.2402,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11099645495414734,
"step": 1335,
"valid_targets_mean": 9408.1,
"valid_targets_min": 289
},
{
"epoch": 2.259696458684654,
"grad_norm": 0.34412164666015127,
"learning_rate": 3.4269429902703124e-05,
"loss": 0.2234,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11150509864091873,
"step": 1340,
"valid_targets_mean": 6347.5,
"valid_targets_min": 1959
},
{
"epoch": 2.2681281618887015,
"grad_norm": 0.4576081261157346,
"learning_rate": 3.42103677522217e-05,
"loss": 0.2461,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14542993903160095,
"step": 1345,
"valid_targets_mean": 7249.5,
"valid_targets_min": 509
},
{
"epoch": 2.2765598650927488,
"grad_norm": 0.3766355304696296,
"learning_rate": 3.41510542608096e-05,
"loss": 0.2398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.154470294713974,
"step": 1350,
"valid_targets_mean": 10789.6,
"valid_targets_min": 1353
},
{
"epoch": 2.284991568296796,
"grad_norm": 0.5209190432275469,
"learning_rate": 3.409149047755355e-05,
"loss": 0.2641,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19939953088760376,
"step": 1355,
"valid_targets_mean": 8834.2,
"valid_targets_min": 397
},
{
"epoch": 2.2934232715008434,
"grad_norm": 0.40313897602674903,
"learning_rate": 3.4031677455967206e-05,
"loss": 0.2441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0792250782251358,
"step": 1360,
"valid_targets_mean": 6457.6,
"valid_targets_min": 408
},
{
"epoch": 2.30185497470489,
"grad_norm": 0.3302250871799064,
"learning_rate": 3.397161625397256e-05,
"loss": 0.2247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09746218472719193,
"step": 1365,
"valid_targets_mean": 12090.6,
"valid_targets_min": 530
},
{
"epoch": 2.3102866779089375,
"grad_norm": 0.44149606462678914,
"learning_rate": 3.3911307933881215e-05,
"loss": 0.2437,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14116719365119934,
"step": 1370,
"valid_targets_mean": 8646.6,
"valid_targets_min": 2155
},
{
"epoch": 2.318718381112985,
"grad_norm": 0.3858720627714008,
"learning_rate": 3.3850753562375575e-05,
"loss": 0.2443,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13359633088111877,
"step": 1375,
"valid_targets_mean": 11386.2,
"valid_targets_min": 3601
},
{
"epoch": 2.327150084317032,
"grad_norm": 0.43482563677582836,
"learning_rate": 3.378995421048999e-05,
"loss": 0.2151,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11568767577409744,
"step": 1380,
"valid_targets_mean": 10007.5,
"valid_targets_min": 698
},
{
"epoch": 2.3355817875210794,
"grad_norm": 0.35878295012614225,
"learning_rate": 3.372891095359184e-05,
"loss": 0.2215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1057644709944725,
"step": 1385,
"valid_targets_mean": 9573.9,
"valid_targets_min": 673
},
{
"epoch": 2.3440134907251267,
"grad_norm": 0.40257070755382157,
"learning_rate": 3.366762487136245e-05,
"loss": 0.2428,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10868300497531891,
"step": 1390,
"valid_targets_mean": 6542.4,
"valid_targets_min": 396
},
{
"epoch": 2.3524451939291735,
"grad_norm": 0.36944669846673517,
"learning_rate": 3.3606097047778074e-05,
"loss": 0.2504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11331410706043243,
"step": 1395,
"valid_targets_mean": 8214.6,
"valid_targets_min": 1077
},
{
"epoch": 2.360876897133221,
"grad_norm": 0.4178255236672399,
"learning_rate": 3.354432857109065e-05,
"loss": 0.2286,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1203223168849945,
"step": 1400,
"valid_targets_mean": 7750.2,
"valid_targets_min": 460
},
{
"epoch": 2.369308600337268,
"grad_norm": 0.389255395480486,
"learning_rate": 3.348232053380861e-05,
"loss": 0.2359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09197911620140076,
"step": 1405,
"valid_targets_mean": 5984.4,
"valid_targets_min": 1145
},
{
"epoch": 2.3777403035413154,
"grad_norm": 0.3631213477709467,
"learning_rate": 3.342007403267751e-05,
"loss": 0.2424,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09749580174684525,
"step": 1410,
"valid_targets_mean": 6949.8,
"valid_targets_min": 605
},
{
"epoch": 2.3861720067453627,
"grad_norm": 0.3860771156929493,
"learning_rate": 3.3357590168660665e-05,
"loss": 0.2348,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08600738644599915,
"step": 1415,
"valid_targets_mean": 9047.9,
"valid_targets_min": 702
},
{
"epoch": 2.39460370994941,
"grad_norm": 0.3896198303910411,
"learning_rate": 3.329487004691966e-05,
"loss": 0.247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10782908648252487,
"step": 1420,
"valid_targets_mean": 6262.6,
"valid_targets_min": 770
},
{
"epoch": 2.403035413153457,
"grad_norm": 0.47055782024931386,
"learning_rate": 3.32319147767948e-05,
"loss": 0.2355,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0995263159275055,
"step": 1425,
"valid_targets_mean": 5901.6,
"valid_targets_min": 768
},
{
"epoch": 2.411467116357504,
"grad_norm": 0.43639696470712896,
"learning_rate": 3.31687254717855e-05,
"loss": 0.2299,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09814895689487457,
"step": 1430,
"valid_targets_mean": 7444.5,
"valid_targets_min": 2119
},
{
"epoch": 2.4198988195615514,
"grad_norm": 0.41479745042846444,
"learning_rate": 3.310530324953059e-05,
"loss": 0.2272,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16535836458206177,
"step": 1435,
"valid_targets_mean": 10602.2,
"valid_targets_min": 1337
},
{
"epoch": 2.4283305227655987,
"grad_norm": 0.3985924118370244,
"learning_rate": 3.3041649231788546e-05,
"loss": 0.2542,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1435907483100891,
"step": 1440,
"valid_targets_mean": 7868.2,
"valid_targets_min": 1961
},
{
"epoch": 2.436762225969646,
"grad_norm": 0.36547182330517286,
"learning_rate": 3.297776454441762e-05,
"loss": 0.2346,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12050890922546387,
"step": 1445,
"valid_targets_mean": 9597.6,
"valid_targets_min": 1751
},
{
"epoch": 2.4451939291736933,
"grad_norm": 0.46395487518125,
"learning_rate": 3.291365031735599e-05,
"loss": 0.2323,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14683182537555695,
"step": 1450,
"valid_targets_mean": 10892.5,
"valid_targets_min": 4657
},
{
"epoch": 2.45362563237774,
"grad_norm": 0.3835780732716753,
"learning_rate": 3.2849307684601704e-05,
"loss": 0.2418,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11021892726421356,
"step": 1455,
"valid_targets_mean": 8311.6,
"valid_targets_min": 454
},
{
"epoch": 2.4620573355817874,
"grad_norm": 0.45722431411277564,
"learning_rate": 3.278473778419267e-05,
"loss": 0.2511,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11257103830575943,
"step": 1460,
"valid_targets_mean": 6888.8,
"valid_targets_min": 381
},
{
"epoch": 2.4704890387858347,
"grad_norm": 0.4092066613934668,
"learning_rate": 3.2719941758186534e-05,
"loss": 0.232,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08980157971382141,
"step": 1465,
"valid_targets_mean": 6509.9,
"valid_targets_min": 359
},
{
"epoch": 2.478920741989882,
"grad_norm": 0.3490162796424915,
"learning_rate": 3.265492075264042e-05,
"loss": 0.2288,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1272692084312439,
"step": 1470,
"valid_targets_mean": 8615.0,
"valid_targets_min": 1216
},
{
"epoch": 2.4873524451939293,
"grad_norm": 0.49817268688983485,
"learning_rate": 3.2589675917590726e-05,
"loss": 0.2412,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08574163168668747,
"step": 1475,
"valid_targets_mean": 5245.6,
"valid_targets_min": 1085
},
{
"epoch": 2.4957841483979766,
"grad_norm": 0.3709735875200464,
"learning_rate": 3.252420840703275e-05,
"loss": 0.2422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11666274070739746,
"step": 1480,
"valid_targets_mean": 7359.1,
"valid_targets_min": 498
},
{
"epoch": 2.504215851602024,
"grad_norm": 0.5080160545393337,
"learning_rate": 3.2458519378900293e-05,
"loss": 0.2381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20056697726249695,
"step": 1485,
"valid_targets_mean": 6719.1,
"valid_targets_min": 630
},
{
"epoch": 2.5126475548060707,
"grad_norm": 0.4589959212269124,
"learning_rate": 3.239260999504514e-05,
"loss": 0.237,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12830400466918945,
"step": 1490,
"valid_targets_mean": 7291.8,
"valid_targets_min": 379
},
{
"epoch": 2.521079258010118,
"grad_norm": 0.45754240505403293,
"learning_rate": 3.232648142121659e-05,
"loss": 0.2291,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14555633068084717,
"step": 1495,
"valid_targets_mean": 7025.8,
"valid_targets_min": 716
},
{
"epoch": 2.5295109612141653,
"grad_norm": 0.37137687255044743,
"learning_rate": 3.226013482704075e-05,
"loss": 0.2333,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16324038803577423,
"step": 1500,
"valid_targets_mean": 13254.4,
"valid_targets_min": 3178
},
{
"epoch": 2.5379426644182126,
"grad_norm": 0.3784626334505863,
"learning_rate": 3.21935713859999e-05,
"loss": 0.2595,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12599334120750427,
"step": 1505,
"valid_targets_mean": 8803.9,
"valid_targets_min": 634
},
{
"epoch": 2.54637436762226,
"grad_norm": 0.34564233813183176,
"learning_rate": 3.2126792275411716e-05,
"loss": 0.2438,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0991472527384758,
"step": 1510,
"valid_targets_mean": 9658.2,
"valid_targets_min": 2993
},
{
"epoch": 2.5548060708263067,
"grad_norm": 0.35819464047300553,
"learning_rate": 3.205979867640847e-05,
"loss": 0.2356,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12249651551246643,
"step": 1515,
"valid_targets_mean": 9737.4,
"valid_targets_min": 1078
},
{
"epoch": 2.563237774030354,
"grad_norm": 0.4002051818031895,
"learning_rate": 3.1992591773916106e-05,
"loss": 0.2457,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11378714442253113,
"step": 1520,
"valid_targets_mean": 7550.8,
"valid_targets_min": 399
},
{
"epoch": 2.5716694772344013,
"grad_norm": 0.34169333294865023,
"learning_rate": 3.192517275663331e-05,
"loss": 0.2213,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12551595270633698,
"step": 1525,
"valid_targets_mean": 11425.9,
"valid_targets_min": 1141
},
{
"epoch": 2.5801011804384486,
"grad_norm": 0.3689171012091987,
"learning_rate": 3.1857542817010474e-05,
"loss": 0.2446,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1461847424507141,
"step": 1530,
"valid_targets_mean": 10272.0,
"valid_targets_min": 692
},
{
"epoch": 2.588532883642496,
"grad_norm": 0.48712689067743287,
"learning_rate": 3.17897031512286e-05,
"loss": 0.2482,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15296855568885803,
"step": 1535,
"valid_targets_mean": 5793.5,
"valid_targets_min": 949
},
{
"epoch": 2.5969645868465427,
"grad_norm": 0.3785871954335419,
"learning_rate": 3.172165495917814e-05,
"loss": 0.2261,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10480280220508575,
"step": 1540,
"valid_targets_mean": 9115.9,
"valid_targets_min": 732
},
{
"epoch": 2.6053962900505905,
"grad_norm": 0.46409576423589155,
"learning_rate": 3.16533994444378e-05,
"loss": 0.2344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16393154859542847,
"step": 1545,
"valid_targets_mean": 10088.2,
"valid_targets_min": 429
},
{
"epoch": 2.6138279932546373,
"grad_norm": 0.40228366174412017,
"learning_rate": 3.158493781425324e-05,
"loss": 0.2292,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14488348364830017,
"step": 1550,
"valid_targets_mean": 11732.1,
"valid_targets_min": 363
},
{
"epoch": 2.6222596964586846,
"grad_norm": 0.515622162568922,
"learning_rate": 3.151627127951567e-05,
"loss": 0.2374,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08075758814811707,
"step": 1555,
"valid_targets_mean": 3552.0,
"valid_targets_min": 543
},
{
"epoch": 2.630691399662732,
"grad_norm": 0.49578885346640933,
"learning_rate": 3.144740105474054e-05,
"loss": 0.2269,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14041131734848022,
"step": 1560,
"valid_targets_mean": 9398.6,
"valid_targets_min": 4073
},
{
"epoch": 2.639123102866779,
"grad_norm": 0.3105458669025041,
"learning_rate": 3.1378328358045944e-05,
"loss": 0.2102,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12387597560882568,
"step": 1565,
"valid_targets_mean": 13636.8,
"valid_targets_min": 2747
},
{
"epoch": 2.6475548060708265,
"grad_norm": 0.38420478785731715,
"learning_rate": 3.1309054411131174e-05,
"loss": 0.2266,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11250080913305283,
"step": 1570,
"valid_targets_mean": 9039.5,
"valid_targets_min": 4688
},
{
"epoch": 2.6559865092748733,
"grad_norm": 0.37827614298397416,
"learning_rate": 3.123958043925501e-05,
"loss": 0.235,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08225288987159729,
"step": 1575,
"valid_targets_mean": 10769.5,
"valid_targets_min": 415
},
{
"epoch": 2.6644182124789206,
"grad_norm": 0.3625970410285234,
"learning_rate": 3.116990767121415e-05,
"loss": 0.2376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09313942492008209,
"step": 1580,
"valid_targets_mean": 7659.0,
"valid_targets_min": 694
},
{
"epoch": 2.672849915682968,
"grad_norm": 0.40383363277266426,
"learning_rate": 3.110003733932142e-05,
"loss": 0.241,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09080016613006592,
"step": 1585,
"valid_targets_mean": 5186.0,
"valid_targets_min": 324
},
{
"epoch": 2.681281618887015,
"grad_norm": 0.4425937927341585,
"learning_rate": 3.102997067938397e-05,
"loss": 0.2383,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09839118272066116,
"step": 1590,
"valid_targets_mean": 6000.0,
"valid_targets_min": 600
},
{
"epoch": 2.6897133220910625,
"grad_norm": 0.3933564319605015,
"learning_rate": 3.095970893068145e-05,
"loss": 0.2229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09367253631353378,
"step": 1595,
"valid_targets_mean": 8437.2,
"valid_targets_min": 323
},
{
"epoch": 2.6981450252951094,
"grad_norm": 0.3562783975818187,
"learning_rate": 3.088925333594408e-05,
"loss": 0.2175,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14288729429244995,
"step": 1600,
"valid_targets_mean": 11796.9,
"valid_targets_min": 5290
},
{
"epoch": 2.706576728499157,
"grad_norm": 0.380644774240712,
"learning_rate": 3.081860514133065e-05,
"loss": 0.2377,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12001267820596695,
"step": 1605,
"valid_targets_mean": 8599.2,
"valid_targets_min": 358
},
{
"epoch": 2.715008431703204,
"grad_norm": 0.33406246843937587,
"learning_rate": 3.0747765596406484e-05,
"loss": 0.2515,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10506954789161682,
"step": 1610,
"valid_targets_mean": 11808.6,
"valid_targets_min": 672
},
{
"epoch": 2.7234401349072512,
"grad_norm": 0.36218443952934953,
"learning_rate": 3.0676735954121384e-05,
"loss": 0.2295,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13790374994277954,
"step": 1615,
"valid_targets_mean": 11083.4,
"valid_targets_min": 311
},
{
"epoch": 2.7318718381112985,
"grad_norm": 0.4080144440097816,
"learning_rate": 3.06055174707874e-05,
"loss": 0.2467,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09257596731185913,
"step": 1620,
"valid_targets_mean": 5696.1,
"valid_targets_min": 854
},
{
"epoch": 2.740303541315346,
"grad_norm": 0.4511164300531131,
"learning_rate": 3.053411140605665e-05,
"loss": 0.2381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10837613046169281,
"step": 1625,
"valid_targets_mean": 5226.2,
"valid_targets_min": 803
},
{
"epoch": 2.748735244519393,
"grad_norm": 0.3471411790500541,
"learning_rate": 3.0462519022899038e-05,
"loss": 0.2326,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11439405381679535,
"step": 1630,
"valid_targets_mean": 9148.2,
"valid_targets_min": 312
},
{
"epoch": 2.75716694772344,
"grad_norm": 0.3546524249512062,
"learning_rate": 3.0390741587579914e-05,
"loss": 0.2253,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09082471579313278,
"step": 1635,
"valid_targets_mean": 6696.4,
"valid_targets_min": 357
},
{
"epoch": 2.7655986509274872,
"grad_norm": 0.35716266789636414,
"learning_rate": 3.031878036963766e-05,
"loss": 0.245,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11040914058685303,
"step": 1640,
"valid_targets_mean": 8842.2,
"valid_targets_min": 266
},
{
"epoch": 2.7740303541315345,
"grad_norm": 0.40585102257411554,
"learning_rate": 3.0246636641861257e-05,
"loss": 0.2345,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10097295790910721,
"step": 1645,
"valid_targets_mean": 5520.5,
"valid_targets_min": 738
},
{
"epoch": 2.782462057335582,
"grad_norm": 0.3240449373758223,
"learning_rate": 3.017431168026778e-05,
"loss": 0.2171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10409976541996002,
"step": 1650,
"valid_targets_mean": 10074.4,
"valid_targets_min": 4948
},
{
"epoch": 2.790893760539629,
"grad_norm": 0.39552508563073313,
"learning_rate": 3.010180676407978e-05,
"loss": 0.2406,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07744117081165314,
"step": 1655,
"valid_targets_mean": 4831.1,
"valid_targets_min": 600
},
{
"epoch": 2.799325463743676,
"grad_norm": 0.4017436521300533,
"learning_rate": 3.002912317570272e-05,
"loss": 0.2295,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11097027361392975,
"step": 1660,
"valid_targets_mean": 7394.6,
"valid_targets_min": 402
},
{
"epoch": 2.8077571669477237,
"grad_norm": 0.31899895526449445,
"learning_rate": 2.9956262200702254e-05,
"loss": 0.2107,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10917623341083527,
"step": 1665,
"valid_targets_mean": 10247.6,
"valid_targets_min": 3143
},
{
"epoch": 2.8161888701517706,
"grad_norm": 0.37915179092747875,
"learning_rate": 2.9883225127781514e-05,
"loss": 0.2354,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1212180033326149,
"step": 1670,
"valid_targets_mean": 9372.8,
"valid_targets_min": 5565
},
{
"epoch": 2.824620573355818,
"grad_norm": 0.36252558888365816,
"learning_rate": 2.981001324875827e-05,
"loss": 0.2445,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10705649852752686,
"step": 1675,
"valid_targets_mean": 8583.4,
"valid_targets_min": 2156
},
{
"epoch": 2.833052276559865,
"grad_norm": 0.4078423841107869,
"learning_rate": 2.9736627858542153e-05,
"loss": 0.2435,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11869801580905914,
"step": 1680,
"valid_targets_mean": 7660.8,
"valid_targets_min": 560
},
{
"epoch": 2.8414839797639124,
"grad_norm": 0.4149489858814276,
"learning_rate": 2.966307025511169e-05,
"loss": 0.2449,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12520082294940948,
"step": 1685,
"valid_targets_mean": 8460.2,
"valid_targets_min": 473
},
{
"epoch": 2.8499156829679597,
"grad_norm": 0.35764247192273574,
"learning_rate": 2.9589341739491368e-05,
"loss": 0.2309,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10460789501667023,
"step": 1690,
"valid_targets_mean": 8293.0,
"valid_targets_min": 958
},
{
"epoch": 2.8583473861720066,
"grad_norm": 0.44300351622692524,
"learning_rate": 2.9515443615728635e-05,
"loss": 0.2363,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15430548787117004,
"step": 1695,
"valid_targets_mean": 6620.4,
"valid_targets_min": 290
},
{
"epoch": 2.866779089376054,
"grad_norm": 0.46008998718975563,
"learning_rate": 2.9441377190870827e-05,
"loss": 0.2299,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1321791112422943,
"step": 1700,
"valid_targets_mean": 10277.5,
"valid_targets_min": 412
},
{
"epoch": 2.875210792580101,
"grad_norm": 0.4190175096430771,
"learning_rate": 2.9367143774942027e-05,
"loss": 0.2134,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08706319332122803,
"step": 1705,
"valid_targets_mean": 6096.9,
"valid_targets_min": 334
},
{
"epoch": 2.8836424957841484,
"grad_norm": 0.3548914508562864,
"learning_rate": 2.9292744680919925e-05,
"loss": 0.2436,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08381716907024384,
"step": 1710,
"valid_targets_mean": 5408.6,
"valid_targets_min": 343
},
{
"epoch": 2.8920741989881957,
"grad_norm": 0.3782542490473102,
"learning_rate": 2.9218181224712602e-05,
"loss": 0.2262,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1749817430973053,
"step": 1715,
"valid_targets_mean": 11363.5,
"valid_targets_min": 648
},
{
"epoch": 2.9005059021922426,
"grad_norm": 0.35337679220379553,
"learning_rate": 2.914345472513521e-05,
"loss": 0.2305,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13214825093746185,
"step": 1720,
"valid_targets_mean": 10012.0,
"valid_targets_min": 349
},
{
"epoch": 2.9089376053962903,
"grad_norm": 0.358317608921764,
"learning_rate": 2.906856650388669e-05,
"loss": 0.2377,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10796335339546204,
"step": 1725,
"valid_targets_mean": 9324.6,
"valid_targets_min": 909
},
{
"epoch": 2.917369308600337,
"grad_norm": 0.46725671873033586,
"learning_rate": 2.8993517885526362e-05,
"loss": 0.238,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10520040988922119,
"step": 1730,
"valid_targets_mean": 9691.5,
"valid_targets_min": 298
},
{
"epoch": 2.9258010118043845,
"grad_norm": 0.4321854689439293,
"learning_rate": 2.891831019745055e-05,
"loss": 0.2142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12195204198360443,
"step": 1735,
"valid_targets_mean": 8222.9,
"valid_targets_min": 309
},
{
"epoch": 2.9342327150084317,
"grad_norm": 0.48478128335919357,
"learning_rate": 2.8842944769869023e-05,
"loss": 0.2353,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15228696167469025,
"step": 1740,
"valid_targets_mean": 11364.2,
"valid_targets_min": 1490
},
{
"epoch": 2.942664418212479,
"grad_norm": 0.3417932677621535,
"learning_rate": 2.876742293578155e-05,
"loss": 0.2283,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17082005739212036,
"step": 1745,
"valid_targets_mean": 13373.2,
"valid_targets_min": 930
},
{
"epoch": 2.9510961214165263,
"grad_norm": 0.3643115303642522,
"learning_rate": 2.8691746030954264e-05,
"loss": 0.2139,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07021670788526535,
"step": 1750,
"valid_targets_mean": 5016.6,
"valid_targets_min": 421
},
{
"epoch": 2.959527824620573,
"grad_norm": 0.32649340453242226,
"learning_rate": 2.861591539389606e-05,
"loss": 0.2337,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08613084256649017,
"step": 1755,
"valid_targets_mean": 11152.2,
"valid_targets_min": 1446
},
{
"epoch": 2.9679595278246205,
"grad_norm": 0.35696233922270276,
"learning_rate": 2.8539932365834924e-05,
"loss": 0.2379,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13745978474617004,
"step": 1760,
"valid_targets_mean": 8880.1,
"valid_targets_min": 1189
},
{
"epoch": 2.9763912310286678,
"grad_norm": 0.39482880286048994,
"learning_rate": 2.8463798290694208e-05,
"loss": 0.2342,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11653143167495728,
"step": 1765,
"valid_targets_mean": 6762.2,
"valid_targets_min": 434
},
{
"epoch": 2.984822934232715,
"grad_norm": 0.3890314447078271,
"learning_rate": 2.8387514515068847e-05,
"loss": 0.2398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14811716973781586,
"step": 1770,
"valid_targets_mean": 10809.4,
"valid_targets_min": 842
},
{
"epoch": 2.9932546374367623,
"grad_norm": 0.7005390348207277,
"learning_rate": 2.8311082388201554e-05,
"loss": 0.2389,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11304610967636108,
"step": 1775,
"valid_targets_mean": 8430.0,
"valid_targets_min": 912
},
{
"epoch": 3.0016863406408096,
"grad_norm": 0.3834505967793565,
"learning_rate": 2.8234503261958956e-05,
"loss": 0.2323,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11922605335712433,
"step": 1780,
"valid_targets_mean": 10550.1,
"valid_targets_min": 310
},
{
"epoch": 3.0101180438448565,
"grad_norm": 0.39043143151890214,
"learning_rate": 2.8157778490807688e-05,
"loss": 0.216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1152990460395813,
"step": 1785,
"valid_targets_mean": 8333.8,
"valid_targets_min": 460
},
{
"epoch": 3.0185497470489038,
"grad_norm": 0.41272642688965594,
"learning_rate": 2.80809094317904e-05,
"loss": 0.2252,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11227001249790192,
"step": 1790,
"valid_targets_mean": 8792.1,
"valid_targets_min": 257
},
{
"epoch": 3.026981450252951,
"grad_norm": 0.44702227130757705,
"learning_rate": 2.8003897444501816e-05,
"loss": 0.2288,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10921060293912888,
"step": 1795,
"valid_targets_mean": 9690.8,
"valid_targets_min": 3681
},
{
"epoch": 3.0354131534569984,
"grad_norm": 0.4599681956775396,
"learning_rate": 2.7926743891064646e-05,
"loss": 0.2303,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10466659069061279,
"step": 1800,
"valid_targets_mean": 7824.0,
"valid_targets_min": 753
},
{
"epoch": 3.0438448566610457,
"grad_norm": 0.4130391285616059,
"learning_rate": 2.784945013610548e-05,
"loss": 0.2228,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08191810548305511,
"step": 1805,
"valid_targets_mean": 6233.0,
"valid_targets_min": 289
},
{
"epoch": 3.052276559865093,
"grad_norm": 0.464574473121947,
"learning_rate": 2.77720175467307e-05,
"loss": 0.2171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09094714373350143,
"step": 1810,
"valid_targets_mean": 4645.2,
"valid_targets_min": 1091
},
{
"epoch": 3.06070826306914,
"grad_norm": 0.42611726310533476,
"learning_rate": 2.769444749250227e-05,
"loss": 0.2214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11890680342912674,
"step": 1815,
"valid_targets_mean": 8863.0,
"valid_targets_min": 1156
},
{
"epoch": 3.069139966273187,
"grad_norm": 0.42458130331013944,
"learning_rate": 2.7616741345413502e-05,
"loss": 0.21,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13504037261009216,
"step": 1820,
"valid_targets_mean": 9772.2,
"valid_targets_min": 1784
},
{
"epoch": 3.0775716694772344,
"grad_norm": 0.39481238852614353,
"learning_rate": 2.753890047986481e-05,
"loss": 0.2128,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09317506104707718,
"step": 1825,
"valid_targets_mean": 6470.6,
"valid_targets_min": 560
},
{
"epoch": 3.0860033726812817,
"grad_norm": 0.38934697459591255,
"learning_rate": 2.7460926272639388e-05,
"loss": 0.1978,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13523752987384796,
"step": 1830,
"valid_targets_mean": 10915.8,
"valid_targets_min": 2280
},
{
"epoch": 3.094435075885329,
"grad_norm": 0.45980474399479077,
"learning_rate": 2.738282010287889e-05,
"loss": 0.201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11868693679571152,
"step": 1835,
"valid_targets_mean": 6256.2,
"valid_targets_min": 316
},
{
"epoch": 3.1028667790893762,
"grad_norm": 0.35014682371982186,
"learning_rate": 2.730458335205897e-05,
"loss": 0.2162,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09542815387248993,
"step": 1840,
"valid_targets_mean": 10432.6,
"valid_targets_min": 4378
},
{
"epoch": 3.111298482293423,
"grad_norm": 0.4217951971598071,
"learning_rate": 2.722621740396493e-05,
"loss": 0.219,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12616297602653503,
"step": 1845,
"valid_targets_mean": 10411.6,
"valid_targets_min": 6334
},
{
"epoch": 3.1197301854974704,
"grad_norm": 0.35047555372322786,
"learning_rate": 2.7147723644667175e-05,
"loss": 0.2109,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14423929154872894,
"step": 1850,
"valid_targets_mean": 11663.2,
"valid_targets_min": 5313
},
{
"epoch": 3.1281618887015177,
"grad_norm": 0.3201725997737384,
"learning_rate": 2.7069103462496733e-05,
"loss": 0.2333,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08930090069770813,
"step": 1855,
"valid_targets_mean": 11771.2,
"valid_targets_min": 362
},
{
"epoch": 3.136593591905565,
"grad_norm": 0.35859105414215114,
"learning_rate": 2.6990358248020703e-05,
"loss": 0.2109,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13119398057460785,
"step": 1860,
"valid_targets_mean": 12700.8,
"valid_targets_min": 5451
},
{
"epoch": 3.1450252951096123,
"grad_norm": 0.5230614226456466,
"learning_rate": 2.691148939401764e-05,
"loss": 0.2173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12492654472589493,
"step": 1865,
"valid_targets_mean": 6815.6,
"valid_targets_min": 727
},
{
"epoch": 3.1534569983136596,
"grad_norm": 0.5159759527757263,
"learning_rate": 2.6832498295452934e-05,
"loss": 0.2232,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19276994466781616,
"step": 1870,
"valid_targets_mean": 7993.9,
"valid_targets_min": 301
},
{
"epoch": 3.1618887015177064,
"grad_norm": 0.44394571755263484,
"learning_rate": 2.675338634945413e-05,
"loss": 0.2159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15400463342666626,
"step": 1875,
"valid_targets_mean": 9766.9,
"valid_targets_min": 3305
},
{
"epoch": 3.1703204047217537,
"grad_norm": 0.43074593883986284,
"learning_rate": 2.6674154955286215e-05,
"loss": 0.2124,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09096457809209824,
"step": 1880,
"valid_targets_mean": 7118.2,
"valid_targets_min": 478
},
{
"epoch": 3.178752107925801,
"grad_norm": 0.5395361693890871,
"learning_rate": 2.6594805514326888e-05,
"loss": 0.1988,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14631852507591248,
"step": 1885,
"valid_targets_mean": 4927.8,
"valid_targets_min": 371
},
{
"epoch": 3.1871838111298483,
"grad_norm": 0.38288869216989413,
"learning_rate": 2.651533943004175e-05,
"loss": 0.2139,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10795603692531586,
"step": 1890,
"valid_targets_mean": 10605.5,
"valid_targets_min": 3259
},
{
"epoch": 3.1956155143338956,
"grad_norm": 0.40070690403513437,
"learning_rate": 2.6435758107959486e-05,
"loss": 0.2118,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11431564390659332,
"step": 1895,
"valid_targets_mean": 8460.5,
"valid_targets_min": 434
},
{
"epoch": 3.204047217537943,
"grad_norm": 0.4296533136029398,
"learning_rate": 2.635606295564703e-05,
"loss": 0.2133,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08163078129291534,
"step": 1900,
"valid_targets_mean": 6234.2,
"valid_targets_min": 865
},
{
"epoch": 3.2124789207419897,
"grad_norm": 0.3550161324555439,
"learning_rate": 2.627625538268461e-05,
"loss": 0.1972,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09761563688516617,
"step": 1905,
"valid_targets_mean": 10543.2,
"valid_targets_min": 869
},
{
"epoch": 3.220910623946037,
"grad_norm": 0.39310789631787746,
"learning_rate": 2.61963368006409e-05,
"loss": 0.2211,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10359895974397659,
"step": 1910,
"valid_targets_mean": 8548.1,
"valid_targets_min": 366
},
{
"epoch": 3.2293423271500843,
"grad_norm": 0.3918928116724224,
"learning_rate": 2.6116308623047974e-05,
"loss": 0.2194,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10392262786626816,
"step": 1915,
"valid_targets_mean": 10654.4,
"valid_targets_min": 2030
},
{
"epoch": 3.2377740303541316,
"grad_norm": 0.3556648033168638,
"learning_rate": 2.6036172265376368e-05,
"loss": 0.222,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10652333498001099,
"step": 1920,
"valid_targets_mean": 10933.1,
"valid_targets_min": 2324
},
{
"epoch": 3.246205733558179,
"grad_norm": 0.431931819186922,
"learning_rate": 2.595592914500998e-05,
"loss": 0.2109,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13332435488700867,
"step": 1925,
"valid_targets_mean": 8789.0,
"valid_targets_min": 916
},
{
"epoch": 3.254637436762226,
"grad_norm": 0.6531524405760268,
"learning_rate": 2.587558068122107e-05,
"loss": 0.2159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09719350934028625,
"step": 1930,
"valid_targets_mean": 2899.8,
"valid_targets_min": 336
},
{
"epoch": 3.263069139966273,
"grad_norm": 0.5196176805447351,
"learning_rate": 2.5795128295145103e-05,
"loss": 0.2222,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09347108006477356,
"step": 1935,
"valid_targets_mean": 5663.1,
"valid_targets_min": 314
},
{
"epoch": 3.2715008431703203,
"grad_norm": 0.4317365231348586,
"learning_rate": 2.571457340975564e-05,
"loss": 0.2126,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13472044467926025,
"step": 1940,
"valid_targets_mean": 10182.0,
"valid_targets_min": 954
},
{
"epoch": 3.2799325463743676,
"grad_norm": 0.45257765029764635,
"learning_rate": 2.5633917449839163e-05,
"loss": 0.2074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1266043335199356,
"step": 1945,
"valid_targets_mean": 7762.2,
"valid_targets_min": 1203
},
{
"epoch": 3.288364249578415,
"grad_norm": 0.42216246501665816,
"learning_rate": 2.555316184196988e-05,
"loss": 0.2165,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11517626792192459,
"step": 1950,
"valid_targets_mean": 7952.8,
"valid_targets_min": 3343
},
{
"epoch": 3.296795952782462,
"grad_norm": 0.3651456628081294,
"learning_rate": 2.5472308014484465e-05,
"loss": 0.2293,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11096100509166718,
"step": 1955,
"valid_targets_mean": 11392.0,
"valid_targets_min": 4951
},
{
"epoch": 3.305227655986509,
"grad_norm": 0.4266727408287084,
"learning_rate": 2.5391357397456846e-05,
"loss": 0.2121,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14057181775569916,
"step": 1960,
"valid_targets_mean": 9008.0,
"valid_targets_min": 739
},
{
"epoch": 3.3136593591905563,
"grad_norm": 0.45259911849439555,
"learning_rate": 2.5310311422672863e-05,
"loss": 0.2207,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16516533493995667,
"step": 1965,
"valid_targets_mean": 10782.2,
"valid_targets_min": 989
},
{
"epoch": 3.3220910623946036,
"grad_norm": 0.4316771533489028,
"learning_rate": 2.5229171523604983e-05,
"loss": 0.2298,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12440809607505798,
"step": 1970,
"valid_targets_mean": 7694.5,
"valid_targets_min": 1649
},
{
"epoch": 3.330522765598651,
"grad_norm": 0.35622576569106335,
"learning_rate": 2.514793913538689e-05,
"loss": 0.235,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0877189189195633,
"step": 1975,
"valid_targets_mean": 7153.5,
"valid_targets_min": 656
},
{
"epoch": 3.338954468802698,
"grad_norm": 0.39472509058026534,
"learning_rate": 2.5066615694788183e-05,
"loss": 0.2224,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09721790254116058,
"step": 1980,
"valid_targets_mean": 7481.8,
"valid_targets_min": 530
},
{
"epoch": 3.3473861720067455,
"grad_norm": 0.3934149171722683,
"learning_rate": 2.498520264018888e-05,
"loss": 0.2152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07264173030853271,
"step": 1985,
"valid_targets_mean": 4662.0,
"valid_targets_min": 406
},
{
"epoch": 3.3558178752107928,
"grad_norm": 0.36649393469746905,
"learning_rate": 2.490370141155404e-05,
"loss": 0.2225,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10559599101543427,
"step": 1990,
"valid_targets_mean": 10137.5,
"valid_targets_min": 1433
},
{
"epoch": 3.3642495784148396,
"grad_norm": 0.44287707050000535,
"learning_rate": 2.4822113450408255e-05,
"loss": 0.2074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1425827592611313,
"step": 1995,
"valid_targets_mean": 9634.0,
"valid_targets_min": 762
},
{
"epoch": 3.372681281618887,
"grad_norm": 0.3807138984959313,
"learning_rate": 2.474044019981018e-05,
"loss": 0.2333,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12348232418298721,
"step": 2000,
"valid_targets_mean": 10872.6,
"valid_targets_min": 906
},
{
"epoch": 3.381112984822934,
"grad_norm": 0.36949096130526543,
"learning_rate": 2.4658683104326983e-05,
"loss": 0.2094,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11024417728185654,
"step": 2005,
"valid_targets_mean": 10341.0,
"valid_targets_min": 672
},
{
"epoch": 3.3895446880269815,
"grad_norm": 0.44271860175348,
"learning_rate": 2.4576843610008828e-05,
"loss": 0.2156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08421653509140015,
"step": 2010,
"valid_targets_mean": 4437.8,
"valid_targets_min": 371
},
{
"epoch": 3.397976391231029,
"grad_norm": 0.40221918634684056,
"learning_rate": 2.4494923164363255e-05,
"loss": 0.2179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08152516186237335,
"step": 2015,
"valid_targets_mean": 7382.4,
"valid_targets_min": 2938
},
{
"epoch": 3.4064080944350756,
"grad_norm": 0.40584720968267507,
"learning_rate": 2.4412923216329637e-05,
"loss": 0.2296,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14482375979423523,
"step": 2020,
"valid_targets_mean": 10600.0,
"valid_targets_min": 799
},
{
"epoch": 3.414839797639123,
"grad_norm": 0.41202575318784723,
"learning_rate": 2.4330845216253475e-05,
"loss": 0.201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09250207245349884,
"step": 2025,
"valid_targets_mean": 8144.6,
"valid_targets_min": 337
},
{
"epoch": 3.4232715008431702,
"grad_norm": 0.44449927132476624,
"learning_rate": 2.4248690615860823e-05,
"loss": 0.2051,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08777996897697449,
"step": 2030,
"valid_targets_mean": 6045.5,
"valid_targets_min": 843
},
{
"epoch": 3.4317032040472175,
"grad_norm": 0.3199903870739605,
"learning_rate": 2.4166460868232563e-05,
"loss": 0.2123,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08690337091684341,
"step": 2035,
"valid_targets_mean": 10358.9,
"valid_targets_min": 2839
},
{
"epoch": 3.440134907251265,
"grad_norm": 0.34941837945268367,
"learning_rate": 2.408415742777872e-05,
"loss": 0.2015,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09891988337039948,
"step": 2040,
"valid_targets_mean": 8918.2,
"valid_targets_min": 1022
},
{
"epoch": 3.448566610455312,
"grad_norm": 0.40055263834328714,
"learning_rate": 2.4001781750212727e-05,
"loss": 0.2155,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11856140196323395,
"step": 2045,
"valid_targets_mean": 9873.0,
"valid_targets_min": 341
},
{
"epoch": 3.4569983136593594,
"grad_norm": 0.3874825338894945,
"learning_rate": 2.3919335292525697e-05,
"loss": 0.2042,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11520876735448837,
"step": 2050,
"valid_targets_mean": 9808.8,
"valid_targets_min": 1034
},
{
"epoch": 3.4654300168634062,
"grad_norm": 0.43259033453784296,
"learning_rate": 2.383681951296064e-05,
"loss": 0.2,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07498389482498169,
"step": 2055,
"valid_targets_mean": 8465.6,
"valid_targets_min": 568
},
{
"epoch": 3.4738617200674535,
"grad_norm": 0.4593606180276576,
"learning_rate": 2.375423587098667e-05,
"loss": 0.2338,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14210501313209534,
"step": 2060,
"valid_targets_mean": 7498.2,
"valid_targets_min": 3047
},
{
"epoch": 3.482293423271501,
"grad_norm": 0.4647440496904315,
"learning_rate": 2.3671585827273184e-05,
"loss": 0.2262,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09839943051338196,
"step": 2065,
"valid_targets_mean": 6991.1,
"valid_targets_min": 435
},
{
"epoch": 3.490725126475548,
"grad_norm": 0.3748096268915676,
"learning_rate": 2.358887084366406e-05,
"loss": 0.2201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1162882074713707,
"step": 2070,
"valid_targets_mean": 9071.9,
"valid_targets_min": 2223
},
{
"epoch": 3.4991568296795954,
"grad_norm": 0.42442032907589594,
"learning_rate": 2.3506092383151752e-05,
"loss": 0.2419,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09524928033351898,
"step": 2075,
"valid_targets_mean": 6554.6,
"valid_targets_min": 931
},
{
"epoch": 3.5075885328836423,
"grad_norm": 0.427573653914135,
"learning_rate": 2.3423251909851454e-05,
"loss": 0.2094,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09386489540338516,
"step": 2080,
"valid_targets_mean": 6484.1,
"valid_targets_min": 880
},
{
"epoch": 3.51602023608769,
"grad_norm": 0.3221188681202846,
"learning_rate": 2.3340350888975204e-05,
"loss": 0.214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09149964153766632,
"step": 2085,
"valid_targets_mean": 11289.5,
"valid_targets_min": 850
},
{
"epoch": 3.524451939291737,
"grad_norm": 0.4119202095864077,
"learning_rate": 2.3257390786805908e-05,
"loss": 0.2104,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11636510491371155,
"step": 2090,
"valid_targets_mean": 11048.6,
"valid_targets_min": 408
},
{
"epoch": 3.532883642495784,
"grad_norm": 0.3787703131353681,
"learning_rate": 2.3174373070671505e-05,
"loss": 0.2097,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12298138439655304,
"step": 2095,
"valid_targets_mean": 8265.9,
"valid_targets_min": 1869
},
{
"epoch": 3.5413153456998314,
"grad_norm": 0.46114011196844584,
"learning_rate": 2.3091299208918932e-05,
"loss": 0.215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1401064395904541,
"step": 2100,
"valid_targets_mean": 7136.9,
"valid_targets_min": 300
},
{
"epoch": 3.5497470489038787,
"grad_norm": 0.3737211981688352,
"learning_rate": 2.3008170670888176e-05,
"loss": 0.2186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10425862669944763,
"step": 2105,
"valid_targets_mean": 9846.5,
"valid_targets_min": 322
},
{
"epoch": 3.558178752107926,
"grad_norm": 0.43185564888539874,
"learning_rate": 2.2924988926886313e-05,
"loss": 0.228,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09306024014949799,
"step": 2110,
"valid_targets_mean": 6222.8,
"valid_targets_min": 266
},
{
"epoch": 3.566610455311973,
"grad_norm": 0.36449632153893213,
"learning_rate": 2.2841755448161472e-05,
"loss": 0.2158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10744164884090424,
"step": 2115,
"valid_targets_mean": 11188.5,
"valid_targets_min": 814
},
{
"epoch": 3.57504215851602,
"grad_norm": 0.46458157253483046,
"learning_rate": 2.275847170687682e-05,
"loss": 0.2378,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08195170760154724,
"step": 2120,
"valid_targets_mean": 5738.5,
"valid_targets_min": 1068
},
{
"epoch": 3.5834738617200674,
"grad_norm": 0.42863361267874706,
"learning_rate": 2.2675139176084522e-05,
"loss": 0.2182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11137376725673676,
"step": 2125,
"valid_targets_mean": 6784.4,
"valid_targets_min": 803
},
{
"epoch": 3.5919055649241147,
"grad_norm": 0.6266401362041654,
"learning_rate": 2.2591759329699695e-05,
"loss": 0.2164,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09482762962579727,
"step": 2130,
"valid_targets_mean": 7434.1,
"valid_targets_min": 394
},
{
"epoch": 3.600337268128162,
"grad_norm": 0.3873895422493245,
"learning_rate": 2.250833364247434e-05,
"loss": 0.2122,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12493032962083817,
"step": 2135,
"valid_targets_mean": 10374.8,
"valid_targets_min": 456
},
{
"epoch": 3.608768971332209,
"grad_norm": 0.3752587257213234,
"learning_rate": 2.2424863589971243e-05,
"loss": 0.2096,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10249622166156769,
"step": 2140,
"valid_targets_mean": 10302.1,
"valid_targets_min": 2159
},
{
"epoch": 3.6172006745362566,
"grad_norm": 0.3512104493867019,
"learning_rate": 2.2341350648537883e-05,
"loss": 0.2119,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09728104621171951,
"step": 2145,
"valid_targets_mean": 8586.5,
"valid_targets_min": 431
},
{
"epoch": 3.6256323777403034,
"grad_norm": 0.36388904172308867,
"learning_rate": 2.2257796295280332e-05,
"loss": 0.2047,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07765563577413559,
"step": 2150,
"valid_targets_mean": 9299.4,
"valid_targets_min": 925
},
{
"epoch": 3.6340640809443507,
"grad_norm": 0.43717459593598945,
"learning_rate": 2.2174202008037117e-05,
"loss": 0.2299,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07985242456197739,
"step": 2155,
"valid_targets_mean": 5606.0,
"valid_targets_min": 744
},
{
"epoch": 3.642495784148398,
"grad_norm": 0.4344362556259223,
"learning_rate": 2.209056926535307e-05,
"loss": 0.1891,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09671647846698761,
"step": 2160,
"valid_targets_mean": 6012.9,
"valid_targets_min": 294
},
{
"epoch": 3.6509274873524453,
"grad_norm": 0.3500143523965317,
"learning_rate": 2.2006899546453218e-05,
"loss": 0.21,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08989816904067993,
"step": 2165,
"valid_targets_mean": 9664.4,
"valid_targets_min": 308
},
{
"epoch": 3.6593591905564926,
"grad_norm": 0.3309035370226495,
"learning_rate": 2.192319433121656e-05,
"loss": 0.2194,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11954857409000397,
"step": 2170,
"valid_targets_mean": 14107.2,
"valid_targets_min": 6334
},
{
"epoch": 3.6677908937605395,
"grad_norm": 0.4520375990396482,
"learning_rate": 2.1839455100149944e-05,
"loss": 0.2292,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16944167017936707,
"step": 2175,
"valid_targets_mean": 9309.6,
"valid_targets_min": 350
},
{
"epoch": 3.6762225969645868,
"grad_norm": 0.4102783079070375,
"learning_rate": 2.1755683334361857e-05,
"loss": 0.2044,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11651168763637543,
"step": 2180,
"valid_targets_mean": 11568.4,
"valid_targets_min": 569
},
{
"epoch": 3.684654300168634,
"grad_norm": 0.3910787864596934,
"learning_rate": 2.1671880515536243e-05,
"loss": 0.2267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09499253332614899,
"step": 2185,
"valid_targets_mean": 8530.5,
"valid_targets_min": 346
},
{
"epoch": 3.6930860033726813,
"grad_norm": 0.3810676736237859,
"learning_rate": 2.1588048125906245e-05,
"loss": 0.2334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09765785932540894,
"step": 2190,
"valid_targets_mean": 7940.8,
"valid_targets_min": 477
},
{
"epoch": 3.7015177065767286,
"grad_norm": 0.35352026319569063,
"learning_rate": 2.1504187648228088e-05,
"loss": 0.2142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08722476661205292,
"step": 2195,
"valid_targets_mean": 10774.8,
"valid_targets_min": 2198
},
{
"epoch": 3.7099494097807755,
"grad_norm": 0.4240684692415063,
"learning_rate": 2.1420300565754758e-05,
"loss": 0.2096,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08883196115493774,
"step": 2200,
"valid_targets_mean": 5211.9,
"valid_targets_min": 323
},
{
"epoch": 3.718381112984823,
"grad_norm": 0.33866007881775945,
"learning_rate": 2.1336388362209806e-05,
"loss": 0.2152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10407601296901703,
"step": 2205,
"valid_targets_mean": 10577.0,
"valid_targets_min": 1102
},
{
"epoch": 3.72681281618887,
"grad_norm": 0.3398218680883839,
"learning_rate": 2.1252452521761103e-05,
"loss": 0.2064,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10325972735881805,
"step": 2210,
"valid_targets_mean": 12926.5,
"valid_targets_min": 7614
},
{
"epoch": 3.7352445193929174,
"grad_norm": 0.5785595595756501,
"learning_rate": 2.116849452899461e-05,
"loss": 0.218,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08407862484455109,
"step": 2215,
"valid_targets_mean": 8438.0,
"valid_targets_min": 729
},
{
"epoch": 3.7436762225969646,
"grad_norm": 0.4500293969918961,
"learning_rate": 2.108451586888808e-05,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1524084061384201,
"step": 2220,
"valid_targets_mean": 10042.6,
"valid_targets_min": 426
},
{
"epoch": 3.752107925801012,
"grad_norm": 0.39548811886153096,
"learning_rate": 2.1000518026784814e-05,
"loss": 0.2144,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08035552501678467,
"step": 2225,
"valid_targets_mean": 5907.1,
"valid_targets_min": 502
},
{
"epoch": 3.7605396290050592,
"grad_norm": 0.41828430433984387,
"learning_rate": 2.0916502488367384e-05,
"loss": 0.2197,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10711421072483063,
"step": 2230,
"valid_targets_mean": 11928.2,
"valid_targets_min": 473
},
{
"epoch": 3.768971332209106,
"grad_norm": 0.38193978853199584,
"learning_rate": 2.0832470739631384e-05,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15970773994922638,
"step": 2235,
"valid_targets_mean": 11664.6,
"valid_targets_min": 7940
},
{
"epoch": 3.7774030354131534,
"grad_norm": 0.3770102699396087,
"learning_rate": 2.0748424266859094e-05,
"loss": 0.2263,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12455850839614868,
"step": 2240,
"valid_targets_mean": 7823.0,
"valid_targets_min": 334
},
{
"epoch": 3.7858347386172007,
"grad_norm": 0.39691019355457885,
"learning_rate": 2.0664364556593227e-05,
"loss": 0.204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1314055621623993,
"step": 2245,
"valid_targets_mean": 9411.6,
"valid_targets_min": 345
},
{
"epoch": 3.794266441821248,
"grad_norm": 0.3494313839044746,
"learning_rate": 2.0580293095610644e-05,
"loss": 0.2219,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14222031831741333,
"step": 2250,
"valid_targets_mean": 11202.4,
"valid_targets_min": 768
},
{
"epoch": 3.8026981450252952,
"grad_norm": 0.36732969021535383,
"learning_rate": 2.049621137089602e-05,
"loss": 0.2037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11348112672567368,
"step": 2255,
"valid_targets_mean": 10394.2,
"valid_targets_min": 472
},
{
"epoch": 3.811129848229342,
"grad_norm": 0.3474125445060669,
"learning_rate": 2.0412120869615593e-05,
"loss": 0.2086,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08484344184398651,
"step": 2260,
"valid_targets_mean": 8623.4,
"valid_targets_min": 1713
},
{
"epoch": 3.8195615514333894,
"grad_norm": 0.3294024911189811,
"learning_rate": 2.03280230790908e-05,
"loss": 0.2034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0919605940580368,
"step": 2265,
"valid_targets_mean": 10850.1,
"valid_targets_min": 606
},
{
"epoch": 3.8279932546374367,
"grad_norm": 0.40121199221610404,
"learning_rate": 2.0243919486772047e-05,
"loss": 0.2106,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07793623954057693,
"step": 2270,
"valid_targets_mean": 10134.5,
"valid_targets_min": 4012
},
{
"epoch": 3.836424957841484,
"grad_norm": 0.5822867662253917,
"learning_rate": 2.015981158021231e-05,
"loss": 0.2082,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15234428644180298,
"step": 2275,
"valid_targets_mean": 7661.0,
"valid_targets_min": 297
},
{
"epoch": 3.8448566610455313,
"grad_norm": 0.41743278046327703,
"learning_rate": 2.0075700847040908e-05,
"loss": 0.2236,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11013621091842651,
"step": 2280,
"valid_targets_mean": 6445.6,
"valid_targets_min": 792
},
{
"epoch": 3.8532883642495785,
"grad_norm": 0.39611232961352805,
"learning_rate": 1.999158877493714e-05,
"loss": 0.2232,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08640918135643005,
"step": 2285,
"valid_targets_mean": 5883.1,
"valid_targets_min": 2225
},
{
"epoch": 3.861720067453626,
"grad_norm": 0.35831965778793473,
"learning_rate": 1.990747685160398e-05,
"loss": 0.2331,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08304017037153244,
"step": 2290,
"valid_targets_mean": 8774.2,
"valid_targets_min": 1976
},
{
"epoch": 3.8701517706576727,
"grad_norm": 0.35020268643093844,
"learning_rate": 1.9823366564741787e-05,
"loss": 0.2236,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10700672119855881,
"step": 2295,
"valid_targets_mean": 10642.8,
"valid_targets_min": 778
},
{
"epoch": 3.87858347386172,
"grad_norm": 0.3939271969588139,
"learning_rate": 1.9739259402021966e-05,
"loss": 0.2139,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14130660891532898,
"step": 2300,
"valid_targets_mean": 11587.1,
"valid_targets_min": 3489
},
{
"epoch": 3.8870151770657673,
"grad_norm": 0.3944801101942699,
"learning_rate": 1.9655156851060664e-05,
"loss": 0.2266,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11037345230579376,
"step": 2305,
"valid_targets_mean": 7375.0,
"valid_targets_min": 984
},
{
"epoch": 3.8954468802698146,
"grad_norm": 0.3816963239744615,
"learning_rate": 1.9571060399392474e-05,
"loss": 0.2016,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09082368016242981,
"step": 2310,
"valid_targets_mean": 6941.4,
"valid_targets_min": 3473
},
{
"epoch": 3.903878583473862,
"grad_norm": 0.5395551969001765,
"learning_rate": 1.9486971534444073e-05,
"loss": 0.2322,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12743373215198517,
"step": 2315,
"valid_targets_mean": 4947.2,
"valid_targets_min": 325
},
{
"epoch": 3.9123102866779087,
"grad_norm": 0.3586563291682806,
"learning_rate": 1.9402891743507993e-05,
"loss": 0.2126,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09804892539978027,
"step": 2320,
"valid_targets_mean": 9274.0,
"valid_targets_min": 498
},
{
"epoch": 3.920741989881956,
"grad_norm": 0.3911219655771001,
"learning_rate": 1.9318822513716245e-05,
"loss": 0.2043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11526736617088318,
"step": 2325,
"valid_targets_mean": 11002.8,
"valid_targets_min": 3500
},
{
"epoch": 3.9291736930860033,
"grad_norm": 0.560180652510681,
"learning_rate": 1.9234765332014064e-05,
"loss": 0.2124,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07197198271751404,
"step": 2330,
"valid_targets_mean": 5063.4,
"valid_targets_min": 327
},
{
"epoch": 3.9376053962900506,
"grad_norm": 0.3502407076014703,
"learning_rate": 1.9150721685133564e-05,
"loss": 0.2393,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09542004764080048,
"step": 2335,
"valid_targets_mean": 10259.4,
"valid_targets_min": 1911
},
{
"epoch": 3.946037099494098,
"grad_norm": 0.3959763666580073,
"learning_rate": 1.9066693059567493e-05,
"loss": 0.2201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1020418256521225,
"step": 2340,
"valid_targets_mean": 9561.0,
"valid_targets_min": 391
},
{
"epoch": 3.954468802698145,
"grad_norm": 0.42848301011020556,
"learning_rate": 1.8982680941542894e-05,
"loss": 0.2132,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10610627382993698,
"step": 2345,
"valid_targets_mean": 6985.9,
"valid_targets_min": 1291
},
{
"epoch": 3.9629005059021924,
"grad_norm": 0.4200973209913442,
"learning_rate": 1.889868681699485e-05,
"loss": 0.2121,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09834475070238113,
"step": 2350,
"valid_targets_mean": 8992.5,
"valid_targets_min": 345
},
{
"epoch": 3.9713322091062393,
"grad_norm": 0.47575113036815586,
"learning_rate": 1.8814712171540175e-05,
"loss": 0.2154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10570941865444183,
"step": 2355,
"valid_targets_mean": 6888.5,
"valid_targets_min": 673
},
{
"epoch": 3.9797639123102866,
"grad_norm": 0.4122881449208728,
"learning_rate": 1.873075849045118e-05,
"loss": 0.2168,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08295533806085587,
"step": 2360,
"valid_targets_mean": 4970.6,
"valid_targets_min": 297
},
{
"epoch": 3.988195615514334,
"grad_norm": 0.6649514184330403,
"learning_rate": 1.8646827258629353e-05,
"loss": 0.2082,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08931984007358551,
"step": 2365,
"valid_targets_mean": 5128.4,
"valid_targets_min": 461
},
{
"epoch": 3.996627318718381,
"grad_norm": 0.3801700622216366,
"learning_rate": 1.8562919960579137e-05,
"loss": 0.2075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07104817032814026,
"step": 2370,
"valid_targets_mean": 5829.2,
"valid_targets_min": 574
},
{
"epoch": 4.0050590219224285,
"grad_norm": 0.3355472333489663,
"learning_rate": 1.847903808038164e-05,
"loss": 0.191,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08120846003293991,
"step": 2375,
"valid_targets_mean": 8994.0,
"valid_targets_min": 345
},
{
"epoch": 4.013490725126475,
"grad_norm": 0.45839060937238785,
"learning_rate": 1.839518310166839e-05,
"loss": 0.2006,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08377303183078766,
"step": 2380,
"valid_targets_mean": 7782.8,
"valid_targets_min": 297
},
{
"epoch": 4.021922428330523,
"grad_norm": 0.4128781782024508,
"learning_rate": 1.8311356507595138e-05,
"loss": 0.2042,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08560852706432343,
"step": 2385,
"valid_targets_mean": 7383.2,
"valid_targets_min": 962
},
{
"epoch": 4.03035413153457,
"grad_norm": 0.3635085508263796,
"learning_rate": 1.8227559780815567e-05,
"loss": 0.2062,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1110738068819046,
"step": 2390,
"valid_targets_mean": 9381.0,
"valid_targets_min": 269
},
{
"epoch": 4.038785834738618,
"grad_norm": 0.3633681472451052,
"learning_rate": 1.8143794403455096e-05,
"loss": 0.1912,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08923216164112091,
"step": 2395,
"valid_targets_mean": 10084.1,
"valid_targets_min": 4064
},
{
"epoch": 4.0472175379426645,
"grad_norm": 0.39724329101887473,
"learning_rate": 1.806006185708467e-05,
"loss": 0.2033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1681937277317047,
"step": 2400,
"valid_targets_mean": 11645.0,
"valid_targets_min": 1038
},
{
"epoch": 4.055649241146711,
"grad_norm": 0.40479971853807256,
"learning_rate": 1.7976363622694535e-05,
"loss": 0.2001,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11657126247882843,
"step": 2405,
"valid_targets_mean": 9778.9,
"valid_targets_min": 1021
},
{
"epoch": 4.064080944350759,
"grad_norm": 0.49323342103577894,
"learning_rate": 1.7892701180668077e-05,
"loss": 0.2022,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10341711342334747,
"step": 2410,
"valid_targets_mean": 6735.1,
"valid_targets_min": 371
},
{
"epoch": 4.072512647554806,
"grad_norm": 0.44659979962312035,
"learning_rate": 1.7809076010755587e-05,
"loss": 0.1869,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08914995193481445,
"step": 2415,
"valid_targets_mean": 7525.8,
"valid_targets_min": 488
},
{
"epoch": 4.080944350758854,
"grad_norm": 0.432511642405327,
"learning_rate": 1.772548959204814e-05,
"loss": 0.2038,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09358938038349152,
"step": 2420,
"valid_targets_mean": 7997.4,
"valid_targets_min": 889
},
{
"epoch": 4.0893760539629005,
"grad_norm": 0.4330476744442892,
"learning_rate": 1.764194340295141e-05,
"loss": 0.2073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1349794566631317,
"step": 2425,
"valid_targets_mean": 9290.2,
"valid_targets_min": 1005
},
{
"epoch": 4.097807757166947,
"grad_norm": 0.44334024456337917,
"learning_rate": 1.7558438921159516e-05,
"loss": 0.2024,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10061223059892654,
"step": 2430,
"valid_targets_mean": 8054.2,
"valid_targets_min": 511
},
{
"epoch": 4.106239460370995,
"grad_norm": 0.3236404272211875,
"learning_rate": 1.7474977623628897e-05,
"loss": 0.1966,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08139478415250778,
"step": 2435,
"valid_targets_mean": 11942.1,
"valid_targets_min": 4544
},
{
"epoch": 4.114671163575042,
"grad_norm": 0.3872791318022624,
"learning_rate": 1.7391560986552195e-05,
"loss": 0.1988,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05941583216190338,
"step": 2440,
"valid_targets_mean": 5195.1,
"valid_targets_min": 354
},
{
"epoch": 4.12310286677909,
"grad_norm": 0.4064364968552249,
"learning_rate": 1.7308190485332126e-05,
"loss": 0.2053,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07335732132196426,
"step": 2445,
"valid_targets_mean": 5803.5,
"valid_targets_min": 442
},
{
"epoch": 4.1315345699831365,
"grad_norm": 0.36627317070240883,
"learning_rate": 1.7224867594555382e-05,
"loss": 0.2025,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08833503723144531,
"step": 2450,
"valid_targets_mean": 11778.0,
"valid_targets_min": 256
},
{
"epoch": 4.139966273187184,
"grad_norm": 0.4926982503988765,
"learning_rate": 1.714159378796659e-05,
"loss": 0.1864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11034771800041199,
"step": 2455,
"valid_targets_mean": 5916.9,
"valid_targets_min": 792
},
{
"epoch": 4.148397976391231,
"grad_norm": 0.3795234617582888,
"learning_rate": 1.7058370538442196e-05,
"loss": 0.1947,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09642472118139267,
"step": 2460,
"valid_targets_mean": 10773.2,
"valid_targets_min": 1713
},
{
"epoch": 4.156829679595278,
"grad_norm": 0.451985096314112,
"learning_rate": 1.6975199317964446e-05,
"loss": 0.1933,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.059830036014318466,
"step": 2465,
"valid_targets_mean": 5588.0,
"valid_targets_min": 357
},
{
"epoch": 4.165261382799326,
"grad_norm": 0.427634421374401,
"learning_rate": 1.6892081597595333e-05,
"loss": 0.1966,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1389504373073578,
"step": 2470,
"valid_targets_mean": 9490.8,
"valid_targets_min": 3792
},
{
"epoch": 4.1736930860033725,
"grad_norm": 0.44642151562325016,
"learning_rate": 1.6809018847450606e-05,
"loss": 0.2036,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0874238833785057,
"step": 2475,
"valid_targets_mean": 6834.1,
"valid_targets_min": 345
},
{
"epoch": 4.18212478920742,
"grad_norm": 0.3965509646274704,
"learning_rate": 1.67260125366737e-05,
"loss": 0.2086,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10369640588760376,
"step": 2480,
"valid_targets_mean": 8980.4,
"valid_targets_min": 1139
},
{
"epoch": 4.190556492411467,
"grad_norm": 0.4705984235793704,
"learning_rate": 1.664306413340985e-05,
"loss": 0.2032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09745734930038452,
"step": 2485,
"valid_targets_mean": 8568.1,
"valid_targets_min": 369
},
{
"epoch": 4.198988195615514,
"grad_norm": 0.393288899402507,
"learning_rate": 1.6560175104780042e-05,
"loss": 0.2019,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08372732996940613,
"step": 2490,
"valid_targets_mean": 6360.6,
"valid_targets_min": 473
},
{
"epoch": 4.207419898819562,
"grad_norm": 0.5825525211951375,
"learning_rate": 1.647734691685511e-05,
"loss": 0.2022,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09355568885803223,
"step": 2495,
"valid_targets_mean": 9860.6,
"valid_targets_min": 2816
},
{
"epoch": 4.2158516020236085,
"grad_norm": 0.38066039681641317,
"learning_rate": 1.6394581034629775e-05,
"loss": 0.2002,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04843580350279808,
"step": 2500,
"valid_targets_mean": 5227.1,
"valid_targets_min": 322
},
{
"epoch": 4.224283305227656,
"grad_norm": 0.36855119466490527,
"learning_rate": 1.631187892199676e-05,
"loss": 0.1852,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1059439405798912,
"step": 2505,
"valid_targets_mean": 9693.5,
"valid_targets_min": 1035
},
{
"epoch": 4.232715008431703,
"grad_norm": 0.4667528179939355,
"learning_rate": 1.622924204172089e-05,
"loss": 0.209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0987122505903244,
"step": 2510,
"valid_targets_mean": 7231.9,
"valid_targets_min": 316
},
{
"epoch": 4.24114671163575,
"grad_norm": 0.523577052181026,
"learning_rate": 1.6146671855413183e-05,
"loss": 0.2118,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11328664422035217,
"step": 2515,
"valid_targets_mean": 8359.0,
"valid_targets_min": 329
},
{
"epoch": 4.249578414839798,
"grad_norm": 0.40019452511137854,
"learning_rate": 1.6064169823505062e-05,
"loss": 0.2053,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09944941848516464,
"step": 2520,
"valid_targets_mean": 9803.6,
"valid_targets_min": 351
},
{
"epoch": 4.2580101180438445,
"grad_norm": 0.35118266100535617,
"learning_rate": 1.5981737405222494e-05,
"loss": 0.1993,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13614186644554138,
"step": 2525,
"valid_targets_mean": 11232.9,
"valid_targets_min": 770
},
{
"epoch": 4.266441821247892,
"grad_norm": 0.4101784506333653,
"learning_rate": 1.589937605856015e-05,
"loss": 0.1974,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10815351456403732,
"step": 2530,
"valid_targets_mean": 7728.8,
"valid_targets_min": 710
},
{
"epoch": 4.274873524451939,
"grad_norm": 0.40182601555910885,
"learning_rate": 1.5817087240255687e-05,
"loss": 0.2006,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07623255997896194,
"step": 2535,
"valid_targets_mean": 8092.9,
"valid_targets_min": 472
},
{
"epoch": 4.283305227655987,
"grad_norm": 0.3718743658930698,
"learning_rate": 1.5734872405763907e-05,
"loss": 0.197,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07010039687156677,
"step": 2540,
"valid_targets_mean": 8235.4,
"valid_targets_min": 336
},
{
"epoch": 4.291736930860034,
"grad_norm": 0.5875499334687437,
"learning_rate": 1.5652733009231074e-05,
"loss": 0.2041,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08835016191005707,
"step": 2545,
"valid_targets_mean": 4037.8,
"valid_targets_min": 443
},
{
"epoch": 4.300168634064081,
"grad_norm": 0.520042273475445,
"learning_rate": 1.5570670503469148e-05,
"loss": 0.2051,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1299225240945816,
"step": 2550,
"valid_targets_mean": 8634.2,
"valid_targets_min": 313
},
{
"epoch": 4.308600337268128,
"grad_norm": 0.42947045696460956,
"learning_rate": 1.5488686339930122e-05,
"loss": 0.2104,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10042064636945724,
"step": 2555,
"valid_targets_mean": 7048.2,
"valid_targets_min": 337
},
{
"epoch": 4.317032040472175,
"grad_norm": 0.33691756632877196,
"learning_rate": 1.540678196868033e-05,
"loss": 0.2002,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09054474532604218,
"step": 2560,
"valid_targets_mean": 9284.5,
"valid_targets_min": 396
},
{
"epoch": 4.325463743676223,
"grad_norm": 0.4766172427929607,
"learning_rate": 1.532495883837483e-05,
"loss": 0.2141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13609734177589417,
"step": 2565,
"valid_targets_mean": 8044.9,
"valid_targets_min": 771
},
{
"epoch": 4.33389544688027,
"grad_norm": 0.4392320616069225,
"learning_rate": 1.524321839623173e-05,
"loss": 0.2146,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11402477324008942,
"step": 2570,
"valid_targets_mean": 8781.5,
"valid_targets_min": 560
},
{
"epoch": 4.3423271500843175,
"grad_norm": 0.4737094772885574,
"learning_rate": 1.5161562088006649e-05,
"loss": 0.1944,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10010764002799988,
"step": 2575,
"valid_targets_mean": 6023.6,
"valid_targets_min": 630
},
{
"epoch": 4.350758853288364,
"grad_norm": 0.42870232650888085,
"learning_rate": 1.5079991357967086e-05,
"loss": 0.1973,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08779013156890869,
"step": 2580,
"valid_targets_mean": 6640.1,
"valid_targets_min": 354
},
{
"epoch": 4.359190556492411,
"grad_norm": 0.41426782391649497,
"learning_rate": 1.4998507648866922e-05,
"loss": 0.1967,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11851321160793304,
"step": 2585,
"valid_targets_mean": 8884.9,
"valid_targets_min": 952
},
{
"epoch": 4.367622259696459,
"grad_norm": 0.42714856144281727,
"learning_rate": 1.491711240192089e-05,
"loss": 0.2068,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10884803533554077,
"step": 2590,
"valid_targets_mean": 9772.4,
"valid_targets_min": 1973
},
{
"epoch": 4.376053962900506,
"grad_norm": 0.4444292297126014,
"learning_rate": 1.4835807056779068e-05,
"loss": 0.2161,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15461134910583496,
"step": 2595,
"valid_targets_mean": 7702.2,
"valid_targets_min": 3710
},
{
"epoch": 4.3844856661045535,
"grad_norm": 0.4155763024767392,
"learning_rate": 1.4754593051501433e-05,
"loss": 0.2206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1142769306898117,
"step": 2600,
"valid_targets_mean": 7896.8,
"valid_targets_min": 332
},
{
"epoch": 4.3929173693086,
"grad_norm": 0.41711066357272963,
"learning_rate": 1.4673471822532428e-05,
"loss": 0.2058,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10899996757507324,
"step": 2605,
"valid_targets_mean": 8563.4,
"valid_targets_min": 761
},
{
"epoch": 4.401349072512647,
"grad_norm": 0.4262288399273641,
"learning_rate": 1.4592444804675536e-05,
"loss": 0.2118,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10033875703811646,
"step": 2610,
"valid_targets_mean": 6677.1,
"valid_targets_min": 344
},
{
"epoch": 4.409780775716695,
"grad_norm": 0.4500335335045232,
"learning_rate": 1.4511513431067905e-05,
"loss": 0.2071,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14068397879600525,
"step": 2615,
"valid_targets_mean": 10767.9,
"valid_targets_min": 1880
},
{
"epoch": 4.418212478920742,
"grad_norm": 0.4072267854252546,
"learning_rate": 1.4430679133155033e-05,
"loss": 0.2001,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09835230559110641,
"step": 2620,
"valid_targets_mean": 7725.8,
"valid_targets_min": 1067
},
{
"epoch": 4.4266441821247895,
"grad_norm": 0.39866867565970393,
"learning_rate": 1.4349943340665405e-05,
"loss": 0.2056,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07181677967309952,
"step": 2625,
"valid_targets_mean": 5884.4,
"valid_targets_min": 717
},
{
"epoch": 4.435075885328836,
"grad_norm": 0.39393915405620455,
"learning_rate": 1.4269307481585237e-05,
"loss": 0.2071,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10145594924688339,
"step": 2630,
"valid_targets_mean": 8737.6,
"valid_targets_min": 869
},
{
"epoch": 4.443507588532883,
"grad_norm": 0.39192907071866456,
"learning_rate": 1.4188772982133205e-05,
"loss": 0.2079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09700833261013031,
"step": 2635,
"valid_targets_mean": 7039.6,
"valid_targets_min": 810
},
{
"epoch": 4.451939291736931,
"grad_norm": 0.5661472203570502,
"learning_rate": 1.4108341266735221e-05,
"loss": 0.2062,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12680977582931519,
"step": 2640,
"valid_targets_mean": 7524.9,
"valid_targets_min": 543
},
{
"epoch": 4.460370994940978,
"grad_norm": 0.40650561558878656,
"learning_rate": 1.4028013757999243e-05,
"loss": 0.1973,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09107547998428345,
"step": 2645,
"valid_targets_mean": 6485.2,
"valid_targets_min": 435
},
{
"epoch": 4.4688026981450255,
"grad_norm": 0.4419881481373133,
"learning_rate": 1.39477918766901e-05,
"loss": 0.2079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0989055335521698,
"step": 2650,
"valid_targets_mean": 6257.4,
"valid_targets_min": 1130
},
{
"epoch": 4.477234401349072,
"grad_norm": 0.39945995166584847,
"learning_rate": 1.386767704170438e-05,
"loss": 0.2081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07877545058727264,
"step": 2655,
"valid_targets_mean": 6927.5,
"valid_targets_min": 1879
},
{
"epoch": 4.48566610455312,
"grad_norm": 0.46875286541908295,
"learning_rate": 1.3787670670045328e-05,
"loss": 0.1991,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1038471907377243,
"step": 2660,
"valid_targets_mean": 6607.0,
"valid_targets_min": 891
},
{
"epoch": 4.494097807757167,
"grad_norm": 0.43934702862292974,
"learning_rate": 1.3707774176797775e-05,
"loss": 0.196,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1389627456665039,
"step": 2665,
"valid_targets_mean": 8963.9,
"valid_targets_min": 272
},
{
"epoch": 4.502529510961214,
"grad_norm": 0.4032375621649299,
"learning_rate": 1.362798897510313e-05,
"loss": 0.1858,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.054929155856370926,
"step": 2670,
"valid_targets_mean": 4053.8,
"valid_targets_min": 329
},
{
"epoch": 4.5109612141652615,
"grad_norm": 0.38037923209294977,
"learning_rate": 1.354831647613436e-05,
"loss": 0.1977,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09493321925401688,
"step": 2675,
"valid_targets_mean": 8310.4,
"valid_targets_min": 1151
},
{
"epoch": 4.519392917369308,
"grad_norm": 0.4216375834882622,
"learning_rate": 1.3468758089071034e-05,
"loss": 0.2019,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13784663379192352,
"step": 2680,
"valid_targets_mean": 10557.8,
"valid_targets_min": 452
},
{
"epoch": 4.527824620573356,
"grad_norm": 0.34237173543113175,
"learning_rate": 1.3389315221074422e-05,
"loss": 0.1924,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11419700086116791,
"step": 2685,
"valid_targets_mean": 11369.1,
"valid_targets_min": 2828
},
{
"epoch": 4.536256323777403,
"grad_norm": 0.37709479567988785,
"learning_rate": 1.3309989277262587e-05,
"loss": 0.1972,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09897217154502869,
"step": 2690,
"valid_targets_mean": 7574.6,
"valid_targets_min": 3911
},
{
"epoch": 4.544688026981451,
"grad_norm": 0.5075465641452338,
"learning_rate": 1.3230781660685526e-05,
"loss": 0.2024,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11697357147932053,
"step": 2695,
"valid_targets_mean": 7323.1,
"valid_targets_min": 387
},
{
"epoch": 4.5531197301854975,
"grad_norm": 0.8991864830826296,
"learning_rate": 1.3151693772300385e-05,
"loss": 0.1985,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10646720230579376,
"step": 2700,
"valid_targets_mean": 7788.4,
"valid_targets_min": 826
},
{
"epoch": 4.561551433389544,
"grad_norm": 0.41367794684475806,
"learning_rate": 1.307272701094663e-05,
"loss": 0.2175,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08782824128866196,
"step": 2705,
"valid_targets_mean": 7461.2,
"valid_targets_min": 459
},
{
"epoch": 4.569983136593592,
"grad_norm": 0.532327140632588,
"learning_rate": 1.2993882773321374e-05,
"loss": 0.2038,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10838079452514648,
"step": 2710,
"valid_targets_mean": 11126.9,
"valid_targets_min": 371
},
{
"epoch": 4.578414839797639,
"grad_norm": 0.41037121957802364,
"learning_rate": 1.2915162453954593e-05,
"loss": 0.2022,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08384490013122559,
"step": 2715,
"valid_targets_mean": 6725.2,
"valid_targets_min": 625
},
{
"epoch": 4.586846543001687,
"grad_norm": 0.4546951582916679,
"learning_rate": 1.2836567445184523e-05,
"loss": 0.2003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12524966895580292,
"step": 2720,
"valid_targets_mean": 8212.5,
"valid_targets_min": 2757
},
{
"epoch": 4.5952782462057336,
"grad_norm": 0.3645815399365509,
"learning_rate": 1.275809913713301e-05,
"loss": 0.2143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09839334338903427,
"step": 2725,
"valid_targets_mean": 9590.9,
"valid_targets_min": 591
},
{
"epoch": 4.60370994940978,
"grad_norm": 0.4014827324178174,
"learning_rate": 1.2679758917680932e-05,
"loss": 0.2054,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12857840955257416,
"step": 2730,
"valid_targets_mean": 11657.8,
"valid_targets_min": 994
},
{
"epoch": 4.612141652613828,
"grad_norm": 0.48512111837186883,
"learning_rate": 1.2601548172443627e-05,
"loss": 0.2063,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09481889009475708,
"step": 2735,
"valid_targets_mean": 7878.0,
"valid_targets_min": 762
},
{
"epoch": 4.620573355817875,
"grad_norm": 0.33154465551256584,
"learning_rate": 1.2523468284746424e-05,
"loss": 0.2076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07845214009284973,
"step": 2740,
"valid_targets_mean": 9652.8,
"valid_targets_min": 404
},
{
"epoch": 4.629005059021923,
"grad_norm": 0.3557786955215219,
"learning_rate": 1.2445520635600148e-05,
"loss": 0.1867,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05851577967405319,
"step": 2745,
"valid_targets_mean": 8467.0,
"valid_targets_min": 312
},
{
"epoch": 4.63743676222597,
"grad_norm": 0.5995658237130046,
"learning_rate": 1.2367706603676677e-05,
"loss": 0.2045,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08317027986049652,
"step": 2750,
"valid_targets_mean": 5703.2,
"valid_targets_min": 954
},
{
"epoch": 4.645868465430016,
"grad_norm": 0.6404453456393473,
"learning_rate": 1.2290027565284618e-05,
"loss": 0.2085,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12514594197273254,
"step": 2755,
"valid_targets_mean": 8836.5,
"valid_targets_min": 660
},
{
"epoch": 4.654300168634064,
"grad_norm": 0.43738036584192136,
"learning_rate": 1.2212484894344908e-05,
"loss": 0.2156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11289623379707336,
"step": 2760,
"valid_targets_mean": 8789.9,
"valid_targets_min": 1148
},
{
"epoch": 4.662731871838111,
"grad_norm": 0.41008154555326876,
"learning_rate": 1.2135079962366523e-05,
"loss": 0.2183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0784025639295578,
"step": 2765,
"valid_targets_mean": 6381.6,
"valid_targets_min": 976
},
{
"epoch": 4.671163575042159,
"grad_norm": 0.4096150862236262,
"learning_rate": 1.2057814138422255e-05,
"loss": 0.2003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10011357069015503,
"step": 2770,
"valid_targets_mean": 9003.6,
"valid_targets_min": 1138
},
{
"epoch": 4.679595278246206,
"grad_norm": 0.40135740186587276,
"learning_rate": 1.1980688789124453e-05,
"loss": 0.201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10500173270702362,
"step": 2775,
"valid_targets_mean": 11512.5,
"valid_targets_min": 1631
},
{
"epoch": 4.688026981450253,
"grad_norm": 0.3804216131852594,
"learning_rate": 1.1903705278600859e-05,
"loss": 0.1905,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11023178696632385,
"step": 2780,
"valid_targets_mean": 9925.9,
"valid_targets_min": 791
},
{
"epoch": 4.6964586846543,
"grad_norm": 0.34675811611679774,
"learning_rate": 1.182686496847052e-05,
"loss": 0.196,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10560382157564163,
"step": 2785,
"valid_targets_mean": 11970.8,
"valid_targets_min": 350
},
{
"epoch": 4.704890387858347,
"grad_norm": 0.3853485431136969,
"learning_rate": 1.1750169217819654e-05,
"loss": 0.1998,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07867911458015442,
"step": 2790,
"valid_targets_mean": 7721.4,
"valid_targets_min": 341
},
{
"epoch": 4.713322091062395,
"grad_norm": 0.32500929872245643,
"learning_rate": 1.1673619383177644e-05,
"loss": 0.1932,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09103326499462128,
"step": 2795,
"valid_targets_mean": 12047.6,
"valid_targets_min": 2446
},
{
"epoch": 4.721753794266442,
"grad_norm": 0.4724301295183875,
"learning_rate": 1.1597216818493031e-05,
"loss": 0.2124,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07927206158638,
"step": 2800,
"valid_targets_mean": 6524.8,
"valid_targets_min": 326
},
{
"epoch": 4.730185497470489,
"grad_norm": 0.39383856506332326,
"learning_rate": 1.1520962875109567e-05,
"loss": 0.1993,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11373279988765717,
"step": 2805,
"valid_targets_mean": 9380.9,
"valid_targets_min": 938
},
{
"epoch": 4.738617200674536,
"grad_norm": 0.41066224485195185,
"learning_rate": 1.1444858901742319e-05,
"loss": 0.2022,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15671461820602417,
"step": 2810,
"valid_targets_mean": 13779.6,
"valid_targets_min": 3160
},
{
"epoch": 4.747048903878584,
"grad_norm": 0.497762838269459,
"learning_rate": 1.1368906244453814e-05,
"loss": 0.2054,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11875620484352112,
"step": 2815,
"valid_targets_mean": 7145.9,
"valid_targets_min": 847
},
{
"epoch": 4.755480607082631,
"grad_norm": 0.40933169187723045,
"learning_rate": 1.1293106246630226e-05,
"loss": 0.2195,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10948336124420166,
"step": 2820,
"valid_targets_mean": 6835.0,
"valid_targets_min": 362
},
{
"epoch": 4.763912310286678,
"grad_norm": 0.41367940810587267,
"learning_rate": 1.121746024895761e-05,
"loss": 0.2062,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11677373945713043,
"step": 2825,
"valid_targets_mean": 8926.5,
"valid_targets_min": 4433
},
{
"epoch": 4.772344013490725,
"grad_norm": 0.5448841225109203,
"learning_rate": 1.1141969589398218e-05,
"loss": 0.2066,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07498321682214737,
"step": 2830,
"valid_targets_mean": 3636.8,
"valid_targets_min": 420
},
{
"epoch": 4.780775716694772,
"grad_norm": 0.4506471214535243,
"learning_rate": 1.1066635603166796e-05,
"loss": 0.2218,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14314770698547363,
"step": 2835,
"valid_targets_mean": 9107.4,
"valid_targets_min": 698
},
{
"epoch": 4.78920741989882,
"grad_norm": 0.4280881513831746,
"learning_rate": 1.099145962270699e-05,
"loss": 0.1966,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09459936618804932,
"step": 2840,
"valid_targets_mean": 8531.1,
"valid_targets_min": 2839
},
{
"epoch": 4.797639123102867,
"grad_norm": 0.5142528187116223,
"learning_rate": 1.0916442977667772e-05,
"loss": 0.2179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19257615506649017,
"step": 2845,
"valid_targets_mean": 7679.0,
"valid_targets_min": 1527
},
{
"epoch": 4.806070826306914,
"grad_norm": 0.3251283967457832,
"learning_rate": 1.0841586994879924e-05,
"loss": 0.2026,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08114565908908844,
"step": 2850,
"valid_targets_mean": 9102.5,
"valid_targets_min": 325
},
{
"epoch": 4.814502529510961,
"grad_norm": 0.4668273292821014,
"learning_rate": 1.0766892998332574e-05,
"loss": 0.2049,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10245705395936966,
"step": 2855,
"valid_targets_mean": 7755.4,
"valid_targets_min": 357
},
{
"epoch": 4.822934232715008,
"grad_norm": 0.4761124561273808,
"learning_rate": 1.0692362309149772e-05,
"loss": 0.1944,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07718995958566666,
"step": 2860,
"valid_targets_mean": 8755.8,
"valid_targets_min": 455
},
{
"epoch": 4.831365935919056,
"grad_norm": 0.419533605358227,
"learning_rate": 1.0617996245567121e-05,
"loss": 0.2126,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1411946415901184,
"step": 2865,
"valid_targets_mean": 10821.2,
"valid_targets_min": 301
},
{
"epoch": 4.839797639123103,
"grad_norm": 0.594193980678808,
"learning_rate": 1.0543796122908487e-05,
"loss": 0.202,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09805161505937576,
"step": 2870,
"valid_targets_mean": 6501.5,
"valid_targets_min": 1235
},
{
"epoch": 4.84822934232715,
"grad_norm": 0.37494130321292113,
"learning_rate": 1.0469763253562696e-05,
"loss": 0.2094,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.061359234154224396,
"step": 2875,
"valid_targets_mean": 5916.5,
"valid_targets_min": 415
},
{
"epoch": 4.856661045531197,
"grad_norm": 0.4274712102068083,
"learning_rate": 1.0395898946960324e-05,
"loss": 0.2007,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09225556254386902,
"step": 2880,
"valid_targets_mean": 7796.1,
"valid_targets_min": 2728
},
{
"epoch": 4.865092748735244,
"grad_norm": 0.41125192126831395,
"learning_rate": 1.0322204509550576e-05,
"loss": 0.2023,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16514438390731812,
"step": 2885,
"valid_targets_mean": 9705.9,
"valid_targets_min": 1005
},
{
"epoch": 4.873524451939292,
"grad_norm": 0.416620395072027,
"learning_rate": 1.024868124477816e-05,
"loss": 0.1967,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0866679698228836,
"step": 2890,
"valid_targets_mean": 8153.6,
"valid_targets_min": 441
},
{
"epoch": 4.881956155143339,
"grad_norm": 0.4512025980656363,
"learning_rate": 1.017533045306021e-05,
"loss": 0.2045,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10361349582672119,
"step": 2895,
"valid_targets_mean": 7783.6,
"valid_targets_min": 325
},
{
"epoch": 4.8903878583473865,
"grad_norm": 0.36056932448054463,
"learning_rate": 1.0102153431763311e-05,
"loss": 0.1937,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08864520490169525,
"step": 2900,
"valid_targets_mean": 12234.0,
"valid_targets_min": 3138
},
{
"epoch": 4.898819561551433,
"grad_norm": 0.439306907536818,
"learning_rate": 1.0029151475180531e-05,
"loss": 0.1963,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07438446581363678,
"step": 2905,
"valid_targets_mean": 8522.6,
"valid_targets_min": 345
},
{
"epoch": 4.90725126475548,
"grad_norm": 0.4467566606110265,
"learning_rate": 9.956325874508578e-06,
"loss": 0.2154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0826200619339943,
"step": 2910,
"valid_targets_mean": 8011.5,
"valid_targets_min": 441
},
{
"epoch": 4.915682967959528,
"grad_norm": 0.5096872378402495,
"learning_rate": 9.883677917824882e-06,
"loss": 0.1989,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14158344268798828,
"step": 2915,
"valid_targets_mean": 12193.5,
"valid_targets_min": 399
},
{
"epoch": 4.924114671163575,
"grad_norm": 0.3705569692680681,
"learning_rate": 9.811208890064876e-06,
"loss": 0.1876,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09020857512950897,
"step": 2920,
"valid_targets_mean": 8963.5,
"valid_targets_min": 441
},
{
"epoch": 4.9325463743676226,
"grad_norm": 0.46485952981284817,
"learning_rate": 9.738920072999246e-06,
"loss": 0.1941,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1549614667892456,
"step": 2925,
"valid_targets_mean": 10742.4,
"valid_targets_min": 4729
},
{
"epoch": 4.940978077571669,
"grad_norm": 0.3456950754437106,
"learning_rate": 9.66681274521129e-06,
"loss": 0.1901,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09097681939601898,
"step": 2930,
"valid_targets_mean": 8989.6,
"valid_targets_min": 1111
},
{
"epoch": 4.949409780775717,
"grad_norm": 0.4324881365953452,
"learning_rate": 9.59488818207424e-06,
"loss": 0.2041,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07879682630300522,
"step": 2935,
"valid_targets_mean": 4875.8,
"valid_targets_min": 446
},
{
"epoch": 4.957841483979764,
"grad_norm": 0.40295636457408185,
"learning_rate": 9.523147655728761e-06,
"loss": 0.1904,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12001125514507294,
"step": 2940,
"valid_targets_mean": 11100.8,
"valid_targets_min": 425
},
{
"epoch": 4.966273187183811,
"grad_norm": 0.5594541849945658,
"learning_rate": 9.451592435060429e-06,
"loss": 0.1844,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09189236164093018,
"step": 2945,
"valid_targets_mean": 10513.8,
"valid_targets_min": 494
},
{
"epoch": 4.974704890387859,
"grad_norm": 0.5093667369057596,
"learning_rate": 9.380223785677287e-06,
"loss": 0.1912,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.106390580534935,
"step": 2950,
"valid_targets_mean": 8018.8,
"valid_targets_min": 342
},
{
"epoch": 4.983136593591905,
"grad_norm": 0.42681493508273965,
"learning_rate": 9.30904296988747e-06,
"loss": 0.2048,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12143994122743607,
"step": 2955,
"valid_targets_mean": 7984.0,
"valid_targets_min": 605
},
{
"epoch": 4.991568296795953,
"grad_norm": 0.36506433253296383,
"learning_rate": 9.238051246676856e-06,
"loss": 0.1918,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07516197115182877,
"step": 2960,
"valid_targets_mean": 6477.4,
"valid_targets_min": 463
},
{
"epoch": 5.0,
"grad_norm": 0.4022073667048992,
"learning_rate": 9.167249871686842e-06,
"loss": 0.2097,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11054430902004242,
"step": 2965,
"valid_targets_mean": 8887.5,
"valid_targets_min": 334
},
{
"epoch": 5.008431703204047,
"grad_norm": 0.542954679647001,
"learning_rate": 9.096640097192084e-06,
"loss": 0.1921,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10823874175548553,
"step": 2970,
"valid_targets_mean": 4924.9,
"valid_targets_min": 334
},
{
"epoch": 5.016863406408095,
"grad_norm": 0.33362047960000063,
"learning_rate": 9.026223172078385e-06,
"loss": 0.1975,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08372687548398972,
"step": 2975,
"valid_targets_mean": 13982.6,
"valid_targets_min": 392
},
{
"epoch": 5.025295109612141,
"grad_norm": 0.4402994446799318,
"learning_rate": 8.956000341820563e-06,
"loss": 0.1871,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10522448271512985,
"step": 2980,
"valid_targets_mean": 10494.6,
"valid_targets_min": 905
},
{
"epoch": 5.033726812816189,
"grad_norm": 0.4224129002624222,
"learning_rate": 8.885972848460492e-06,
"loss": 0.1832,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10956309735774994,
"step": 2985,
"valid_targets_mean": 11723.6,
"valid_targets_min": 1368
},
{
"epoch": 5.042158516020236,
"grad_norm": 0.40787386286532296,
"learning_rate": 8.816141930585067e-06,
"loss": 0.1901,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08430559933185577,
"step": 2990,
"valid_targets_mean": 6385.1,
"valid_targets_min": 478
},
{
"epoch": 5.050590219224283,
"grad_norm": 0.4000267866531839,
"learning_rate": 8.746508823304327e-06,
"loss": 0.193,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.088462695479393,
"step": 2995,
"valid_targets_mean": 8791.6,
"valid_targets_min": 293
},
{
"epoch": 5.059021922428331,
"grad_norm": 0.38493511198891794,
"learning_rate": 8.677074758229602e-06,
"loss": 0.1995,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13261154294013977,
"step": 3000,
"valid_targets_mean": 12564.2,
"valid_targets_min": 1788
},
{
"epoch": 5.0674536256323774,
"grad_norm": 0.40422334788246445,
"learning_rate": 8.607840963451748e-06,
"loss": 0.1889,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09068082273006439,
"step": 3005,
"valid_targets_mean": 10961.4,
"valid_targets_min": 771
},
{
"epoch": 5.075885328836425,
"grad_norm": 0.4246270455404128,
"learning_rate": 8.538808663519402e-06,
"loss": 0.1869,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09276852756738663,
"step": 3010,
"valid_targets_mean": 7172.2,
"valid_targets_min": 818
},
{
"epoch": 5.084317032040472,
"grad_norm": 0.48203104879676767,
"learning_rate": 8.46997907941731e-06,
"loss": 0.2098,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1207263246178627,
"step": 3015,
"valid_targets_mean": 8630.8,
"valid_targets_min": 587
},
{
"epoch": 5.09274873524452,
"grad_norm": 0.3550425581234542,
"learning_rate": 8.401353428544772e-06,
"loss": 0.1794,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05964341759681702,
"step": 3020,
"valid_targets_mean": 6539.1,
"valid_targets_min": 373
},
{
"epoch": 5.101180438448567,
"grad_norm": 0.43530667490714414,
"learning_rate": 8.332932924694102e-06,
"loss": 0.2039,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11769139766693115,
"step": 3025,
"valid_targets_mean": 8755.1,
"valid_targets_min": 1286
},
{
"epoch": 5.1096121416526135,
"grad_norm": 0.47113960625764684,
"learning_rate": 8.264718778029125e-06,
"loss": 0.1863,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09204412251710892,
"step": 3030,
"valid_targets_mean": 6050.2,
"valid_targets_min": 475
},
{
"epoch": 5.118043844856661,
"grad_norm": 0.4962792386315349,
"learning_rate": 8.196712195063807e-06,
"loss": 0.2076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08696837723255157,
"step": 3035,
"valid_targets_mean": 4815.0,
"valid_targets_min": 226
},
{
"epoch": 5.126475548060708,
"grad_norm": 0.42737540999525847,
"learning_rate": 8.128914378640891e-06,
"loss": 0.1909,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11643768101930618,
"step": 3040,
"valid_targets_mean": 9874.6,
"valid_targets_min": 3577
},
{
"epoch": 5.134907251264756,
"grad_norm": 0.5059683529475366,
"learning_rate": 8.061326527910665e-06,
"loss": 0.1893,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09628903865814209,
"step": 3045,
"valid_targets_mean": 6064.0,
"valid_targets_min": 349
},
{
"epoch": 5.143338954468803,
"grad_norm": 0.4358546731249333,
"learning_rate": 7.993949838309676e-06,
"loss": 0.2027,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08835709095001221,
"step": 3050,
"valid_targets_mean": 8530.4,
"valid_targets_min": 918
},
{
"epoch": 5.15177065767285,
"grad_norm": 0.4768347098429266,
"learning_rate": 7.926785501539669e-06,
"loss": 0.2004,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1063929945230484,
"step": 3055,
"valid_targets_mean": 8721.1,
"valid_targets_min": 925
},
{
"epoch": 5.160202360876897,
"grad_norm": 0.38275172753616354,
"learning_rate": 7.859834705546451e-06,
"loss": 0.1958,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08256031572818756,
"step": 3060,
"valid_targets_mean": 9673.1,
"valid_targets_min": 3673
},
{
"epoch": 5.168634064080944,
"grad_norm": 0.47628184130280693,
"learning_rate": 7.793098634498923e-06,
"loss": 0.1884,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09466478228569031,
"step": 3065,
"valid_targets_mean": 6457.4,
"valid_targets_min": 281
},
{
"epoch": 5.177065767284992,
"grad_norm": 0.4128531546997656,
"learning_rate": 7.726578468768094e-06,
"loss": 0.1864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06623177230358124,
"step": 3070,
"valid_targets_mean": 4651.8,
"valid_targets_min": 1954
},
{
"epoch": 5.185497470489039,
"grad_norm": 0.4914160764988135,
"learning_rate": 7.660275384906227e-06,
"loss": 0.1893,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.101621612906456,
"step": 3075,
"valid_targets_mean": 8757.8,
"valid_targets_min": 365
},
{
"epoch": 5.193929173693086,
"grad_norm": 0.44961650974591305,
"learning_rate": 7.594190555626027e-06,
"loss": 0.1871,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09812674671411514,
"step": 3080,
"valid_targets_mean": 9011.4,
"valid_targets_min": 370
},
{
"epoch": 5.202360876897133,
"grad_norm": 0.3534849224113734,
"learning_rate": 7.528325149779898e-06,
"loss": 0.1903,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0962594598531723,
"step": 3085,
"valid_targets_mean": 12906.1,
"valid_targets_min": 7617
},
{
"epoch": 5.21079258010118,
"grad_norm": 0.4204713177217243,
"learning_rate": 7.46268033233927e-06,
"loss": 0.1736,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.055231936275959015,
"step": 3090,
"valid_targets_mean": 5581.2,
"valid_targets_min": 246
},
{
"epoch": 5.219224283305228,
"grad_norm": 0.39636053007368066,
"learning_rate": 7.397257264373985e-06,
"loss": 0.1874,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07444168627262115,
"step": 3095,
"valid_targets_mean": 6763.2,
"valid_targets_min": 614
},
{
"epoch": 5.227655986509275,
"grad_norm": 0.3837355423718371,
"learning_rate": 7.332057103031771e-06,
"loss": 0.1945,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12513524293899536,
"step": 3100,
"valid_targets_mean": 10223.1,
"valid_targets_min": 3093
},
{
"epoch": 5.236087689713322,
"grad_norm": 0.4185680144029909,
"learning_rate": 7.267081001517784e-06,
"loss": 0.186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09677357971668243,
"step": 3105,
"valid_targets_mean": 8877.6,
"valid_targets_min": 1491
},
{
"epoch": 5.244519392917369,
"grad_norm": 0.37878564248168,
"learning_rate": 7.202330109074196e-06,
"loss": 0.1722,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04259276017546654,
"step": 3110,
"valid_targets_mean": 5835.1,
"valid_targets_min": 375
},
{
"epoch": 5.252951096121416,
"grad_norm": 0.35220160892272456,
"learning_rate": 7.137805570959842e-06,
"loss": 0.1823,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08681666105985641,
"step": 3115,
"valid_targets_mean": 9533.8,
"valid_targets_min": 644
},
{
"epoch": 5.261382799325464,
"grad_norm": 0.44824635374002586,
"learning_rate": 7.07350852843004e-06,
"loss": 0.2014,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08400139212608337,
"step": 3120,
"valid_targets_mean": 6737.8,
"valid_targets_min": 967
},
{
"epoch": 5.269814502529511,
"grad_norm": 0.4272901630988954,
"learning_rate": 7.009440118716331e-06,
"loss": 0.1909,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1326947808265686,
"step": 3125,
"valid_targets_mean": 9065.4,
"valid_targets_min": 843
},
{
"epoch": 5.278246205733558,
"grad_norm": 0.46597594817586707,
"learning_rate": 6.945601475006401e-06,
"loss": 0.1827,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12690648436546326,
"step": 3130,
"valid_targets_mean": 11650.1,
"valid_targets_min": 353
},
{
"epoch": 5.286677908937605,
"grad_norm": 0.4015307152717381,
"learning_rate": 6.881993726424025e-06,
"loss": 0.1869,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10109364986419678,
"step": 3135,
"valid_targets_mean": 10740.2,
"valid_targets_min": 2767
},
{
"epoch": 5.295109612141653,
"grad_norm": 0.42727062870231347,
"learning_rate": 6.8186179980090985e-06,
"loss": 0.1884,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07408735156059265,
"step": 3140,
"valid_targets_mean": 6522.6,
"valid_targets_min": 341
},
{
"epoch": 5.3035413153457,
"grad_norm": 0.43780808596515997,
"learning_rate": 6.7554754106977674e-06,
"loss": 0.1929,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1048104465007782,
"step": 3145,
"valid_targets_mean": 8526.0,
"valid_targets_min": 414
},
{
"epoch": 5.311973018549747,
"grad_norm": 0.4659780925074606,
"learning_rate": 6.692567081302534e-06,
"loss": 0.1931,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11064967513084412,
"step": 3150,
"valid_targets_mean": 10817.0,
"valid_targets_min": 4731
},
{
"epoch": 5.320404721753794,
"grad_norm": 0.46196684917388225,
"learning_rate": 6.629894122492566e-06,
"loss": 0.1931,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0807664543390274,
"step": 3155,
"valid_targets_mean": 6031.0,
"valid_targets_min": 378
},
{
"epoch": 5.328836424957841,
"grad_norm": 0.42465530757557785,
"learning_rate": 6.5674576427740065e-06,
"loss": 0.1797,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12117722630500793,
"step": 3160,
"valid_targets_mean": 10268.4,
"valid_targets_min": 311
},
{
"epoch": 5.337268128161889,
"grad_norm": 0.5422597299842946,
"learning_rate": 6.505258746470344e-06,
"loss": 0.1924,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12142455577850342,
"step": 3165,
"valid_targets_mean": 7019.5,
"valid_targets_min": 371
},
{
"epoch": 5.345699831365936,
"grad_norm": 0.39262012640464106,
"learning_rate": 6.443298533702891e-06,
"loss": 0.197,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08644361793994904,
"step": 3170,
"valid_targets_mean": 10032.6,
"valid_targets_min": 892
},
{
"epoch": 5.354131534569984,
"grad_norm": 0.3489263600265207,
"learning_rate": 6.381578100371335e-06,
"loss": 0.2061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08918486535549164,
"step": 3175,
"valid_targets_mean": 13995.2,
"valid_targets_min": 5190
},
{
"epoch": 5.36256323777403,
"grad_norm": 0.3401478743098609,
"learning_rate": 6.320098538134347e-06,
"loss": 0.1971,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07504022121429443,
"step": 3180,
"valid_targets_mean": 10153.9,
"valid_targets_min": 1353
},
{
"epoch": 5.370994940978077,
"grad_norm": 0.3833692675819442,
"learning_rate": 6.25886093439027e-06,
"loss": 0.1896,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11480212211608887,
"step": 3185,
"valid_targets_mean": 9889.5,
"valid_targets_min": 353
},
{
"epoch": 5.379426644182125,
"grad_norm": 0.35002475320513415,
"learning_rate": 6.197866372257894e-06,
"loss": 0.1838,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08806844055652618,
"step": 3190,
"valid_targets_mean": 10578.5,
"valid_targets_min": 1437
},
{
"epoch": 5.387858347386172,
"grad_norm": 0.5632716464911012,
"learning_rate": 6.137115930557289e-06,
"loss": 0.1844,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10343332588672638,
"step": 3195,
"valid_targets_mean": 12329.5,
"valid_targets_min": 656
},
{
"epoch": 5.39629005059022,
"grad_norm": 0.37300531609136794,
"learning_rate": 6.076610683790749e-06,
"loss": 0.182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1235748678445816,
"step": 3200,
"valid_targets_mean": 11331.5,
"valid_targets_min": 5204
},
{
"epoch": 5.4047217537942664,
"grad_norm": 0.5294802349153991,
"learning_rate": 6.0163517021237465e-06,
"loss": 0.198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09362947940826416,
"step": 3205,
"valid_targets_mean": 7213.9,
"valid_targets_min": 341
},
{
"epoch": 5.413153456998313,
"grad_norm": 0.430780074949685,
"learning_rate": 5.956340051366041e-06,
"loss": 0.2011,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10629715770483017,
"step": 3210,
"valid_targets_mean": 10477.6,
"valid_targets_min": 4219
},
{
"epoch": 5.421585160202361,
"grad_norm": 0.43243513236232256,
"learning_rate": 5.896576792952786e-06,
"loss": 0.1973,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08778179436922073,
"step": 3215,
"valid_targets_mean": 6120.5,
"valid_targets_min": 1019
},
{
"epoch": 5.430016863406408,
"grad_norm": 0.5042367272239958,
"learning_rate": 5.837062983925821e-06,
"loss": 0.1959,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1380157768726349,
"step": 3220,
"valid_targets_mean": 7510.8,
"valid_targets_min": 803
},
{
"epoch": 5.438448566610456,
"grad_norm": 0.40245700101978,
"learning_rate": 5.77779967691491e-06,
"loss": 0.1867,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07026177644729614,
"step": 3225,
"valid_targets_mean": 5899.8,
"valid_targets_min": 335
},
{
"epoch": 5.4468802698145025,
"grad_norm": 0.40833943734168165,
"learning_rate": 5.718787920119158e-06,
"loss": 0.2074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08088932931423187,
"step": 3230,
"valid_targets_mean": 8356.6,
"valid_targets_min": 318
},
{
"epoch": 5.455311973018549,
"grad_norm": 0.42269376913150103,
"learning_rate": 5.660028757288454e-06,
"loss": 0.2056,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0773058831691742,
"step": 3235,
"valid_targets_mean": 6245.8,
"valid_targets_min": 1973
},
{
"epoch": 5.463743676222597,
"grad_norm": 0.3706934825696817,
"learning_rate": 5.601523227705039e-06,
"loss": 0.1963,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09028086066246033,
"step": 3240,
"valid_targets_mean": 8493.2,
"valid_targets_min": 1034
},
{
"epoch": 5.472175379426644,
"grad_norm": 0.3791710396477716,
"learning_rate": 5.543272366165093e-06,
"loss": 0.1869,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07896457612514496,
"step": 3245,
"valid_targets_mean": 8438.2,
"valid_targets_min": 773
},
{
"epoch": 5.480607082630692,
"grad_norm": 0.4253325417718827,
"learning_rate": 5.48527720296043e-06,
"loss": 0.1978,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09559831768274307,
"step": 3250,
"valid_targets_mean": 9056.0,
"valid_targets_min": 979
},
{
"epoch": 5.4890387858347385,
"grad_norm": 0.47577391439289596,
"learning_rate": 5.427538763860303e-06,
"loss": 0.1939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13109445571899414,
"step": 3255,
"valid_targets_mean": 7825.8,
"valid_targets_min": 1016
},
{
"epoch": 5.497470489038786,
"grad_norm": 0.41933904607731515,
"learning_rate": 5.370058070093256e-06,
"loss": 0.1909,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08193913102149963,
"step": 3260,
"valid_targets_mean": 6718.5,
"valid_targets_min": 660
},
{
"epoch": 5.505902192242833,
"grad_norm": 0.3768333664461666,
"learning_rate": 5.3128361383290365e-06,
"loss": 0.1794,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08452702313661575,
"step": 3265,
"valid_targets_mean": 9978.6,
"valid_targets_min": 348
},
{
"epoch": 5.51433389544688,
"grad_norm": 0.34975600663021716,
"learning_rate": 5.255873980660635e-06,
"loss": 0.1961,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06400668621063232,
"step": 3270,
"valid_targets_mean": 10418.1,
"valid_targets_min": 875
},
{
"epoch": 5.522765598650928,
"grad_norm": 0.41374417726609014,
"learning_rate": 5.199172604586376e-06,
"loss": 0.1936,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10672137141227722,
"step": 3275,
"valid_targets_mean": 9393.5,
"valid_targets_min": 579
},
{
"epoch": 5.5311973018549745,
"grad_norm": 0.40412250533224975,
"learning_rate": 5.142733012992101e-06,
"loss": 0.1924,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09195329248905182,
"step": 3280,
"valid_targets_mean": 7222.5,
"valid_targets_min": 2840
},
{
"epoch": 5.539629005059022,
"grad_norm": 0.4052025025562296,
"learning_rate": 5.0865562041334345e-06,
"loss": 0.1834,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.093361034989357,
"step": 3285,
"valid_targets_mean": 8084.6,
"valid_targets_min": 472
},
{
"epoch": 5.548060708263069,
"grad_norm": 0.5182940378275435,
"learning_rate": 5.030643171618114e-06,
"loss": 0.1899,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07985076308250427,
"step": 3290,
"valid_targets_mean": 6370.2,
"valid_targets_min": 768
},
{
"epoch": 5.556492411467117,
"grad_norm": 0.4435721182119751,
"learning_rate": 4.974994904388428e-06,
"loss": 0.1841,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08876698464155197,
"step": 3295,
"valid_targets_mean": 8310.0,
"valid_targets_min": 384
},
{
"epoch": 5.564924114671164,
"grad_norm": 0.4421625692167533,
"learning_rate": 4.919612386703734e-06,
"loss": 0.1838,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10462569445371628,
"step": 3300,
"valid_targets_mean": 10635.9,
"valid_targets_min": 303
},
{
"epoch": 5.5733558178752105,
"grad_norm": 0.517167431587648,
"learning_rate": 4.864496598123018e-06,
"loss": 0.1927,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09126631915569305,
"step": 3305,
"valid_targets_mean": 6775.0,
"valid_targets_min": 2960
},
{
"epoch": 5.581787521079258,
"grad_norm": 0.48675891081245587,
"learning_rate": 4.809648513487597e-06,
"loss": 0.2053,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18233448266983032,
"step": 3310,
"valid_targets_mean": 8862.5,
"valid_targets_min": 1021
},
{
"epoch": 5.590219224283305,
"grad_norm": 0.3954989936732149,
"learning_rate": 4.755069102903864e-06,
"loss": 0.1957,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.042476288974285126,
"step": 3315,
"valid_targets_mean": 4860.1,
"valid_targets_min": 337
},
{
"epoch": 5.598650927487353,
"grad_norm": 0.49198169538331343,
"learning_rate": 4.700759331726141e-06,
"loss": 0.1817,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11045876890420914,
"step": 3320,
"valid_targets_mean": 8451.1,
"valid_targets_min": 1049
},
{
"epoch": 5.6070826306914,
"grad_norm": 0.5546956669073048,
"learning_rate": 4.646720160539588e-06,
"loss": 0.1871,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10236077010631561,
"step": 3325,
"valid_targets_mean": 7463.2,
"valid_targets_min": 420
},
{
"epoch": 5.6155143338954465,
"grad_norm": 0.41030137569219405,
"learning_rate": 4.592952545143219e-06,
"loss": 0.1987,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12685737013816833,
"step": 3330,
"valid_targets_mean": 10448.1,
"valid_targets_min": 1195
},
{
"epoch": 5.623946037099494,
"grad_norm": 0.4111210424791622,
"learning_rate": 4.539457436533021e-06,
"loss": 0.1876,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09167680144309998,
"step": 3335,
"valid_targets_mean": 7712.4,
"valid_targets_min": 1464
},
{
"epoch": 5.632377740303541,
"grad_norm": 0.40593337171096666,
"learning_rate": 4.4862357808850886e-06,
"loss": 0.2033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1007004827260971,
"step": 3340,
"valid_targets_mean": 9487.2,
"valid_targets_min": 353
},
{
"epoch": 5.640809443507589,
"grad_norm": 0.4216170786376396,
"learning_rate": 4.433288519538927e-06,
"loss": 0.1992,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07875614613294601,
"step": 3345,
"valid_targets_mean": 7743.2,
"valid_targets_min": 377
},
{
"epoch": 5.649241146711636,
"grad_norm": 0.5245222210778443,
"learning_rate": 4.380616588980766e-06,
"loss": 0.2132,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08431693911552429,
"step": 3350,
"valid_targets_mean": 4591.6,
"valid_targets_min": 360
},
{
"epoch": 5.6576728499156825,
"grad_norm": 0.3937933693810972,
"learning_rate": 4.328220920827051e-06,
"loss": 0.1949,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09154506027698517,
"step": 3355,
"valid_targets_mean": 8693.4,
"valid_targets_min": 300
},
{
"epoch": 5.66610455311973,
"grad_norm": 0.3927120766275851,
"learning_rate": 4.276102441807905e-06,
"loss": 0.1949,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09336018562316895,
"step": 3360,
"valid_targets_mean": 7453.0,
"valid_targets_min": 357
},
{
"epoch": 5.674536256323777,
"grad_norm": 0.4308075282932863,
"learning_rate": 4.224262073750782e-06,
"loss": 0.1902,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07926037907600403,
"step": 3365,
"valid_targets_mean": 7407.9,
"valid_targets_min": 406
},
{
"epoch": 5.682967959527825,
"grad_norm": 0.45235140682189356,
"learning_rate": 4.1727007335641255e-06,
"loss": 0.1987,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06334839761257172,
"step": 3370,
"valid_targets_mean": 4346.6,
"valid_targets_min": 626
},
{
"epoch": 5.691399662731872,
"grad_norm": 0.5091975020185051,
"learning_rate": 4.121419333221204e-06,
"loss": 0.1994,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09873318672180176,
"step": 3375,
"valid_targets_mean": 6263.6,
"valid_targets_min": 633
},
{
"epoch": 5.699831365935919,
"grad_norm": 0.3982928916991725,
"learning_rate": 4.07041877974391e-06,
"loss": 0.2018,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08585284650325775,
"step": 3380,
"valid_targets_mean": 6567.2,
"valid_targets_min": 855
},
{
"epoch": 5.708263069139966,
"grad_norm": 0.4320229561019046,
"learning_rate": 4.019699975186775e-06,
"loss": 0.1975,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06424611806869507,
"step": 3385,
"valid_targets_mean": 5432.8,
"valid_targets_min": 321
},
{
"epoch": 5.716694772344013,
"grad_norm": 0.4455282618518715,
"learning_rate": 3.9692638166209856e-06,
"loss": 0.1941,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07248486578464508,
"step": 3390,
"valid_targets_mean": 5735.0,
"valid_targets_min": 342
},
{
"epoch": 5.725126475548061,
"grad_norm": 0.43391690908209707,
"learning_rate": 3.9191111961185455e-06,
"loss": 0.184,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12507924437522888,
"step": 3395,
"valid_targets_mean": 11789.4,
"valid_targets_min": 4882
},
{
"epoch": 5.733558178752108,
"grad_norm": 0.35637435071519863,
"learning_rate": 3.869243000736453e-06,
"loss": 0.1919,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08329631388187408,
"step": 3400,
"valid_targets_mean": 10558.8,
"valid_targets_min": 988
},
{
"epoch": 5.7419898819561555,
"grad_norm": 0.3808225282694838,
"learning_rate": 3.819660112501053e-06,
"loss": 0.1898,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08843482285737991,
"step": 3405,
"valid_targets_mean": 8562.9,
"valid_targets_min": 1553
},
{
"epoch": 5.750421585160202,
"grad_norm": 0.45994942554587803,
"learning_rate": 3.7703634083924077e-06,
"loss": 0.1895,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1378982812166214,
"step": 3410,
"valid_targets_mean": 8156.8,
"valid_targets_min": 2758
},
{
"epoch": 5.75885328836425,
"grad_norm": 0.5351419944888521,
"learning_rate": 3.7213537603288076e-06,
"loss": 0.2002,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06130992993712425,
"step": 3415,
"valid_targets_mean": 5763.0,
"valid_targets_min": 778
},
{
"epoch": 5.767284991568297,
"grad_norm": 0.39137074642187103,
"learning_rate": 3.672632035151329e-06,
"loss": 0.1885,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13084037601947784,
"step": 3420,
"valid_targets_mean": 10961.6,
"valid_targets_min": 622
},
{
"epoch": 5.775716694772344,
"grad_norm": 0.42432261805845334,
"learning_rate": 3.624199094608525e-06,
"loss": 0.1774,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10581144690513611,
"step": 3425,
"valid_targets_mean": 10391.4,
"valid_targets_min": 367
},
{
"epoch": 5.7841483979763915,
"grad_norm": 0.507886487836701,
"learning_rate": 3.576055795341151e-06,
"loss": 0.206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1146765947341919,
"step": 3430,
"valid_targets_mean": 6854.1,
"valid_targets_min": 365
},
{
"epoch": 5.792580101180438,
"grad_norm": 0.4923692971326073,
"learning_rate": 3.528202988867064e-06,
"loss": 0.2062,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10118594765663147,
"step": 3435,
"valid_targets_mean": 6734.0,
"valid_targets_min": 2606
},
{
"epoch": 5.801011804384486,
"grad_norm": 0.3927658690898022,
"learning_rate": 3.480641521566104e-06,
"loss": 0.2103,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09115704149007797,
"step": 3440,
"valid_targets_mean": 7548.8,
"valid_targets_min": 383
},
{
"epoch": 5.809443507588533,
"grad_norm": 0.3786521741084374,
"learning_rate": 3.43337223466516e-06,
"loss": 0.1917,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08883464336395264,
"step": 3445,
"valid_targets_mean": 10119.4,
"valid_targets_min": 664
},
{
"epoch": 5.81787521079258,
"grad_norm": 0.4791078054909208,
"learning_rate": 3.3863959642232835e-06,
"loss": 0.1933,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0843275710940361,
"step": 3450,
"valid_targets_mean": 9600.6,
"valid_targets_min": 880
},
{
"epoch": 5.8263069139966275,
"grad_norm": 0.44212239728889885,
"learning_rate": 3.339713541116896e-06,
"loss": 0.1998,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0958564430475235,
"step": 3455,
"valid_targets_mean": 6854.5,
"valid_targets_min": 619
},
{
"epoch": 5.834738617200674,
"grad_norm": 0.38638112047451684,
"learning_rate": 3.2933257910250992e-06,
"loss": 0.1804,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10496068000793457,
"step": 3460,
"valid_targets_mean": 10717.6,
"valid_targets_min": 1578
},
{
"epoch": 5.843170320404722,
"grad_norm": 0.42030119984356595,
"learning_rate": 3.2472335344150706e-06,
"loss": 0.1821,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09116186201572418,
"step": 3465,
"valid_targets_mean": 10487.4,
"valid_targets_min": 3413
},
{
"epoch": 5.851602023608769,
"grad_norm": 0.43361249227791726,
"learning_rate": 3.201437586527536e-06,
"loss": 0.1941,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10730093717575073,
"step": 3470,
"valid_targets_mean": 8992.0,
"valid_targets_min": 2075
},
{
"epoch": 5.860033726812816,
"grad_norm": 0.47565732973739727,
"learning_rate": 3.155938757362391e-06,
"loss": 0.1992,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06313558667898178,
"step": 3475,
"valid_targets_mean": 4916.1,
"valid_targets_min": 313
},
{
"epoch": 5.8684654300168635,
"grad_norm": 0.35497310348718236,
"learning_rate": 3.110737851664327e-06,
"loss": 0.1836,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10307668149471283,
"step": 3480,
"valid_targets_mean": 13386.8,
"valid_targets_min": 6356
},
{
"epoch": 5.87689713322091,
"grad_norm": 0.4378804415607339,
"learning_rate": 3.0658356689086077e-06,
"loss": 0.1863,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10347874462604523,
"step": 3485,
"valid_targets_mean": 7905.1,
"valid_targets_min": 983
},
{
"epoch": 5.885328836424958,
"grad_norm": 0.4592228896506163,
"learning_rate": 3.0212330032869654e-06,
"loss": 0.1851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08304422348737717,
"step": 3490,
"valid_targets_mean": 5670.1,
"valid_targets_min": 415
},
{
"epoch": 5.893760539629005,
"grad_norm": 0.5410260915083691,
"learning_rate": 2.9769306436935143e-06,
"loss": 0.2036,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13733431696891785,
"step": 3495,
"valid_targets_mean": 5478.2,
"valid_targets_min": 912
},
{
"epoch": 5.902192242833053,
"grad_norm": 0.3536098383617728,
"learning_rate": 2.932929373710809e-06,
"loss": 0.1864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11266787350177765,
"step": 3500,
"valid_targets_mean": 16139.5,
"valid_targets_min": 6959
},
{
"epoch": 5.9106239460370995,
"grad_norm": 0.48244894934533417,
"learning_rate": 2.8892299715959883e-06,
"loss": 0.1796,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07873134315013885,
"step": 3505,
"valid_targets_mean": 4343.0,
"valid_targets_min": 415
},
{
"epoch": 5.919055649241146,
"grad_norm": 0.43211360706070095,
"learning_rate": 2.8458332102670084e-06,
"loss": 0.2004,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10615306347608566,
"step": 3510,
"valid_targets_mean": 8088.8,
"valid_targets_min": 747
},
{
"epoch": 5.927487352445194,
"grad_norm": 0.35823778826768493,
"learning_rate": 2.80273985728897e-06,
"loss": 0.1815,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11246617883443832,
"step": 3515,
"valid_targets_mean": 13532.5,
"valid_targets_min": 5383
},
{
"epoch": 5.935919055649241,
"grad_norm": 0.38932413905042784,
"learning_rate": 2.7599506748605497e-06,
"loss": 0.1911,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06973576545715332,
"step": 3520,
"valid_targets_mean": 12441.8,
"valid_targets_min": 369
},
{
"epoch": 5.944350758853289,
"grad_norm": 0.436906691946414,
"learning_rate": 2.717466419800501e-06,
"loss": 0.2035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11446712911128998,
"step": 3525,
"valid_targets_mean": 6150.2,
"valid_targets_min": 432
},
{
"epoch": 5.9527824620573355,
"grad_norm": 0.5871472870386228,
"learning_rate": 2.675287843534302e-06,
"loss": 0.1843,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1034948080778122,
"step": 3530,
"valid_targets_mean": 10024.1,
"valid_targets_min": 281
},
{
"epoch": 5.961214165261383,
"grad_norm": 0.3693077593622342,
"learning_rate": 2.633415692080825e-06,
"loss": 0.1912,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10228648781776428,
"step": 3535,
"valid_targets_mean": 12193.2,
"valid_targets_min": 345
},
{
"epoch": 5.96964586846543,
"grad_norm": 0.4980807003649807,
"learning_rate": 2.5918507060391675e-06,
"loss": 0.1811,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10854673385620117,
"step": 3540,
"valid_targets_mean": 10953.9,
"valid_targets_min": 307
},
{
"epoch": 5.978077571669477,
"grad_norm": 0.41095231338907384,
"learning_rate": 2.5505936205755457e-06,
"loss": 0.1909,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1156267374753952,
"step": 3545,
"valid_targets_mean": 10766.9,
"valid_targets_min": 1665
},
{
"epoch": 5.986509274873525,
"grad_norm": 0.6549168564838099,
"learning_rate": 2.509645165410286e-06,
"loss": 0.2029,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10136865079402924,
"step": 3550,
"valid_targets_mean": 7702.6,
"valid_targets_min": 434
},
{
"epoch": 5.9949409780775715,
"grad_norm": 0.4166122990295687,
"learning_rate": 2.469006064804933e-06,
"loss": 0.2013,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15833088755607605,
"step": 3555,
"valid_targets_mean": 11934.6,
"valid_targets_min": 4787
},
{
"epoch": 6.003372681281619,
"grad_norm": 0.3476955340898102,
"learning_rate": 2.428677037549423e-06,
"loss": 0.2001,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07825460284948349,
"step": 3560,
"valid_targets_mean": 11951.6,
"valid_targets_min": 752
},
{
"epoch": 6.011804384485666,
"grad_norm": 0.39509700057369984,
"learning_rate": 2.3886587969493767e-06,
"loss": 0.1965,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08286819607019424,
"step": 3565,
"valid_targets_mean": 8729.2,
"valid_targets_min": 598
},
{
"epoch": 6.020236087689713,
"grad_norm": 0.38858873890934376,
"learning_rate": 2.348952050813498e-06,
"loss": 0.1819,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10276305675506592,
"step": 3570,
"valid_targets_mean": 9669.2,
"valid_targets_min": 3729
},
{
"epoch": 6.028667790893761,
"grad_norm": 0.3879309134922529,
"learning_rate": 2.3095575014410223e-06,
"loss": 0.1911,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05445010960102081,
"step": 3575,
"valid_targets_mean": 8770.6,
"valid_targets_min": 321
},
{
"epoch": 6.0370994940978076,
"grad_norm": 0.47955820839979224,
"learning_rate": 2.2704758456093323e-06,
"loss": 0.1799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13328570127487183,
"step": 3580,
"valid_targets_mean": 9508.8,
"valid_targets_min": 854
},
{
"epoch": 6.045531197301855,
"grad_norm": 0.3965160429865374,
"learning_rate": 2.231707774561587e-06,
"loss": 0.1941,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08635030686855316,
"step": 3585,
"valid_targets_mean": 6504.8,
"valid_targets_min": 1365
},
{
"epoch": 6.053962900505902,
"grad_norm": 0.4829259765929293,
"learning_rate": 2.1932539739945556e-06,
"loss": 0.1785,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10407707095146179,
"step": 3590,
"valid_targets_mean": 6806.8,
"valid_targets_min": 1469
},
{
"epoch": 6.062394603709949,
"grad_norm": 0.6239373558597963,
"learning_rate": 2.155115124046443e-06,
"loss": 0.1868,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10681603103876114,
"step": 3595,
"valid_targets_mean": 3905.0,
"valid_targets_min": 783
},
{
"epoch": 6.070826306913997,
"grad_norm": 0.7112210538102081,
"learning_rate": 2.1172918992848703e-06,
"loss": 0.1967,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09597411751747131,
"step": 3600,
"valid_targets_mean": 9401.6,
"valid_targets_min": 1165
},
{
"epoch": 6.079258010118044,
"grad_norm": 0.4437129173467961,
"learning_rate": 2.0797849686949512e-06,
"loss": 0.2039,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1332988440990448,
"step": 3605,
"valid_targets_mean": 11532.4,
"valid_targets_min": 382
},
{
"epoch": 6.087689713322091,
"grad_norm": 0.5989071853334794,
"learning_rate": 2.0425949956674707e-06,
"loss": 0.2022,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09734189510345459,
"step": 3610,
"valid_targets_mean": 4232.8,
"valid_targets_min": 402
},
{
"epoch": 6.096121416526138,
"grad_norm": 0.425722035068979,
"learning_rate": 2.005722637987111e-06,
"loss": 0.1792,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0664922371506691,
"step": 3615,
"valid_targets_mean": 6101.6,
"valid_targets_min": 563
},
{
"epoch": 6.104553119730186,
"grad_norm": 0.388321424149101,
"learning_rate": 1.9691685478208656e-06,
"loss": 0.172,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07667747139930725,
"step": 3620,
"valid_targets_mean": 9815.0,
"valid_targets_min": 2836
},
{
"epoch": 6.112984822934233,
"grad_norm": 0.5089004999538798,
"learning_rate": 1.9329333717064646e-06,
"loss": 0.1982,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08871015906333923,
"step": 3625,
"valid_targets_mean": 5355.0,
"valid_targets_min": 467
},
{
"epoch": 6.12141652613828,
"grad_norm": 0.41186027692617166,
"learning_rate": 1.897017750540977e-06,
"loss": 0.214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0746651291847229,
"step": 3630,
"valid_targets_mean": 5351.1,
"valid_targets_min": 316
},
{
"epoch": 6.129848229342327,
"grad_norm": 0.4033931496113947,
"learning_rate": 1.8614223195694391e-06,
"loss": 0.1969,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12525403499603271,
"step": 3635,
"valid_targets_mean": 10918.5,
"valid_targets_min": 1034
},
{
"epoch": 6.138279932546374,
"grad_norm": 0.4180432966870388,
"learning_rate": 1.8261477083736424e-06,
"loss": 0.2014,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09297751635313034,
"step": 3640,
"valid_targets_mean": 7989.0,
"valid_targets_min": 1193
},
{
"epoch": 6.146711635750422,
"grad_norm": 0.4120584918028444,
"learning_rate": 1.7911945408609831e-06,
"loss": 0.1858,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05716902017593384,
"step": 3645,
"valid_targets_mean": 6517.0,
"valid_targets_min": 3647
},
{
"epoch": 6.155143338954469,
"grad_norm": 0.6654652736553199,
"learning_rate": 1.7565634352534423e-06,
"loss": 0.1879,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14306357502937317,
"step": 3650,
"valid_targets_mean": 11061.5,
"valid_targets_min": 3933
},
{
"epoch": 6.163575042158516,
"grad_norm": 0.49845993321749793,
"learning_rate": 1.7222550040766383e-06,
"loss": 0.1933,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07705598324537277,
"step": 3655,
"valid_targets_mean": 7277.2,
"valid_targets_min": 371
},
{
"epoch": 6.172006745362563,
"grad_norm": 0.3673149535093275,
"learning_rate": 1.6882698541489983e-06,
"loss": 0.173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10980324447154999,
"step": 3660,
"valid_targets_mean": 11853.2,
"valid_targets_min": 855
},
{
"epoch": 6.18043844856661,
"grad_norm": 0.525351664206511,
"learning_rate": 1.654608586571025e-06,
"loss": 0.1935,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1117447018623352,
"step": 3665,
"valid_targets_mean": 6436.1,
"valid_targets_min": 396
},
{
"epoch": 6.188870151770658,
"grad_norm": 0.42828561542168314,
"learning_rate": 1.6212717967146697e-06,
"loss": 0.1922,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08651744574308395,
"step": 3670,
"valid_targets_mean": 7135.6,
"valid_targets_min": 2446
},
{
"epoch": 6.197301854974705,
"grad_norm": 0.3499028629016545,
"learning_rate": 1.5882600742127907e-06,
"loss": 0.1905,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1005062460899353,
"step": 3675,
"valid_targets_mean": 11420.9,
"valid_targets_min": 3060
},
{
"epoch": 6.2057335581787525,
"grad_norm": 0.414669111143042,
"learning_rate": 1.5555740029487342e-06,
"loss": 0.191,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1186700239777565,
"step": 3680,
"valid_targets_mean": 10277.9,
"valid_targets_min": 3175
},
{
"epoch": 6.214165261382799,
"grad_norm": 0.42984613987771814,
"learning_rate": 1.5232141610460027e-06,
"loss": 0.1833,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09360063821077347,
"step": 3685,
"valid_targets_mean": 11024.8,
"valid_targets_min": 397
},
{
"epoch": 6.222596964586846,
"grad_norm": 0.4032180299798233,
"learning_rate": 1.4911811208580274e-06,
"loss": 0.1943,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09980033338069916,
"step": 3690,
"valid_targets_mean": 11117.1,
"valid_targets_min": 2765
},
{
"epoch": 6.231028667790894,
"grad_norm": 0.39477195854702574,
"learning_rate": 1.4594754489580542e-06,
"loss": 0.1815,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09655362367630005,
"step": 3695,
"valid_targets_mean": 10964.9,
"valid_targets_min": 4454
},
{
"epoch": 6.239460370994941,
"grad_norm": 0.597528128491468,
"learning_rate": 1.4280977061291057e-06,
"loss": 0.1964,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09477049112319946,
"step": 3700,
"valid_targets_mean": 3841.9,
"valid_targets_min": 366
},
{
"epoch": 6.2478920741989885,
"grad_norm": 0.40751123350107793,
"learning_rate": 1.3970484473540903e-06,
"loss": 0.1945,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11050431430339813,
"step": 3705,
"valid_targets_mean": 8705.6,
"valid_targets_min": 278
},
{
"epoch": 6.256323777403035,
"grad_norm": 0.4561285621281205,
"learning_rate": 1.366328221805957e-06,
"loss": 0.1664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06762471795082092,
"step": 3710,
"valid_targets_mean": 6710.5,
"valid_targets_min": 377
},
{
"epoch": 6.264755480607082,
"grad_norm": 0.4161372484695771,
"learning_rate": 1.3359375728379908e-06,
"loss": 0.1994,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08619184792041779,
"step": 3715,
"valid_targets_mean": 7908.9,
"valid_targets_min": 1353
},
{
"epoch": 6.27318718381113,
"grad_norm": 0.3631362510222921,
"learning_rate": 1.3058770379742125e-06,
"loss": 0.1742,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09839427471160889,
"step": 3720,
"valid_targets_mean": 12389.8,
"valid_targets_min": 2618
},
{
"epoch": 6.281618887015177,
"grad_norm": 0.6626543574275667,
"learning_rate": 1.2761471488998688e-06,
"loss": 0.2011,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1290425956249237,
"step": 3725,
"valid_targets_mean": 8339.4,
"valid_targets_min": 354
},
{
"epoch": 6.2900505902192245,
"grad_norm": 0.4205118346707732,
"learning_rate": 1.2467484314520163e-06,
"loss": 0.1768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09689425677061081,
"step": 3730,
"valid_targets_mean": 11722.4,
"valid_targets_min": 809
},
{
"epoch": 6.298482293423271,
"grad_norm": 0.33424990841600827,
"learning_rate": 1.2176814056102315e-06,
"loss": 0.1828,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07166987657546997,
"step": 3735,
"valid_targets_mean": 9408.5,
"valid_targets_min": 3333
},
{
"epoch": 6.306913996627319,
"grad_norm": 0.41163755574802646,
"learning_rate": 1.1889465854874072e-06,
"loss": 0.1879,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0651257187128067,
"step": 3740,
"valid_targets_mean": 6899.1,
"valid_targets_min": 431
},
{
"epoch": 6.315345699831366,
"grad_norm": 0.4029882943819463,
"learning_rate": 1.1605444793206755e-06,
"loss": 0.1905,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11918431520462036,
"step": 3745,
"valid_targets_mean": 12357.1,
"valid_targets_min": 5465
},
{
"epoch": 6.323777403035413,
"grad_norm": 0.4862994210000376,
"learning_rate": 1.1324755894623917e-06,
"loss": 0.198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10200175642967224,
"step": 3750,
"valid_targets_mean": 7314.1,
"valid_targets_min": 935
},
{
"epoch": 6.3322091062394605,
"grad_norm": 0.4173844093535785,
"learning_rate": 1.1047404123712703e-06,
"loss": 0.179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09026925265789032,
"step": 3755,
"valid_targets_mean": 7606.4,
"valid_targets_min": 283
},
{
"epoch": 6.340640809443507,
"grad_norm": 0.377749525655152,
"learning_rate": 1.0773394386035952e-06,
"loss": 0.1854,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14625495672225952,
"step": 3760,
"valid_targets_mean": 13962.8,
"valid_targets_min": 4814
},
{
"epoch": 6.349072512647555,
"grad_norm": 0.46413305788873693,
"learning_rate": 1.0502731528045506e-06,
"loss": 0.1861,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1087455302476883,
"step": 3765,
"valid_targets_mean": 7880.9,
"valid_targets_min": 367
},
{
"epoch": 6.357504215851602,
"grad_norm": 0.5316678889087798,
"learning_rate": 1.0235420336996405e-06,
"loss": 0.1987,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.054866641759872437,
"step": 3770,
"valid_targets_mean": 4618.1,
"valid_targets_min": 297
},
{
"epoch": 6.36593591905565,
"grad_norm": 0.3858366746472398,
"learning_rate": 9.971465540862236e-07,
"loss": 0.18,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10158047080039978,
"step": 3775,
"valid_targets_mean": 11926.2,
"valid_targets_min": 1077
},
{
"epoch": 6.3743676222596966,
"grad_norm": 0.4323590839158096,
"learning_rate": 9.710871808251543e-07,
"loss": 0.1881,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09787815809249878,
"step": 3780,
"valid_targets_mean": 9206.2,
"valid_targets_min": 942
},
{
"epoch": 6.382799325463743,
"grad_norm": 0.43843492824609853,
"learning_rate": 9.453643748325203e-07,
"loss": 0.1861,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0957077145576477,
"step": 3785,
"valid_targets_mean": 7901.9,
"valid_targets_min": 463
},
{
"epoch": 6.391231028667791,
"grad_norm": 0.475875860513937,
"learning_rate": 9.199785910714931e-07,
"loss": 0.1775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08440780639648438,
"step": 3790,
"valid_targets_mean": 5652.4,
"valid_targets_min": 550
},
{
"epoch": 6.399662731871838,
"grad_norm": 0.4545314041282514,
"learning_rate": 8.949302785442837e-07,
"loss": 0.1809,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08174890279769897,
"step": 3795,
"valid_targets_mean": 6672.9,
"valid_targets_min": 1952
},
{
"epoch": 6.408094435075886,
"grad_norm": 0.3875408365201888,
"learning_rate": 8.702198802841933e-07,
"loss": 0.2026,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.052681341767311096,
"step": 3800,
"valid_targets_mean": 5883.5,
"valid_targets_min": 312
},
{
"epoch": 6.416526138279933,
"grad_norm": 0.4239419221970245,
"learning_rate": 8.458478333477905e-07,
"loss": 0.1805,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09456834197044373,
"step": 3805,
"valid_targets_mean": 7683.1,
"valid_targets_min": 322
},
{
"epoch": 6.424957841483979,
"grad_norm": 0.38844660665865827,
"learning_rate": 8.218145688071688e-07,
"loss": 0.1964,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07516494393348694,
"step": 3810,
"valid_targets_mean": 6942.9,
"valid_targets_min": 382
},
{
"epoch": 6.433389544688027,
"grad_norm": 0.4294453036657181,
"learning_rate": 7.981205117423152e-07,
"loss": 0.1857,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09382139891386032,
"step": 3815,
"valid_targets_mean": 8716.8,
"valid_targets_min": 1586
},
{
"epoch": 6.441821247892074,
"grad_norm": 0.42069204347590333,
"learning_rate": 7.747660812336221e-07,
"loss": 0.1819,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07698909193277359,
"step": 3820,
"valid_targets_mean": 6424.9,
"valid_targets_min": 694
},
{
"epoch": 6.450252951096122,
"grad_norm": 0.4599498885386992,
"learning_rate": 7.517516903544431e-07,
"loss": 0.1892,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12698175013065338,
"step": 3825,
"valid_targets_mean": 9654.4,
"valid_targets_min": 837
},
{
"epoch": 6.458684654300169,
"grad_norm": 0.7712462092850062,
"learning_rate": 7.290777461637999e-07,
"loss": 0.2004,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1415184736251831,
"step": 3830,
"valid_targets_mean": 10770.4,
"valid_targets_min": 498
},
{
"epoch": 6.467116357504215,
"grad_norm": 0.542667149001183,
"learning_rate": 7.067446496991847e-07,
"loss": 0.2092,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.128068745136261,
"step": 3835,
"valid_targets_mean": 6307.2,
"valid_targets_min": 774
},
{
"epoch": 6.475548060708263,
"grad_norm": 0.47926135197936093,
"learning_rate": 6.847527959694611e-07,
"loss": 0.1812,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10486174374818802,
"step": 3840,
"valid_targets_mean": 8118.8,
"valid_targets_min": 606
},
{
"epoch": 6.48397976391231,
"grad_norm": 0.5036598357941342,
"learning_rate": 6.631025739478913e-07,
"loss": 0.1915,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07612927258014679,
"step": 3845,
"valid_targets_mean": 7741.8,
"valid_targets_min": 668
},
{
"epoch": 6.492411467116358,
"grad_norm": 0.4357150404565448,
"learning_rate": 6.417943665652271e-07,
"loss": 0.1811,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08846313506364822,
"step": 3850,
"valid_targets_mean": 9150.2,
"valid_targets_min": 3277
},
{
"epoch": 6.500843170320405,
"grad_norm": 0.6377900967894415,
"learning_rate": 6.208285507029699e-07,
"loss": 0.1925,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10835045576095581,
"step": 3855,
"valid_targets_mean": 4396.9,
"valid_targets_min": 756
},
{
"epoch": 6.509274873524452,
"grad_norm": 0.39725170664049797,
"learning_rate": 6.002054971866878e-07,
"loss": 0.1879,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08391515910625458,
"step": 3860,
"valid_targets_mean": 7921.2,
"valid_targets_min": 455
},
{
"epoch": 6.517706576728499,
"grad_norm": 0.3870252470281087,
"learning_rate": 5.799255707794582e-07,
"loss": 0.1785,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10050440579652786,
"step": 3865,
"valid_targets_mean": 10204.1,
"valid_targets_min": 2202
},
{
"epoch": 6.526138279932546,
"grad_norm": 0.3520994940405098,
"learning_rate": 5.599891301754112e-07,
"loss": 0.1805,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09250039607286453,
"step": 3870,
"valid_targets_mean": 10682.0,
"valid_targets_min": 2141
},
{
"epoch": 6.534569983136594,
"grad_norm": 0.357653116068236,
"learning_rate": 5.403965279934031e-07,
"loss": 0.1752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06915470957756042,
"step": 3875,
"valid_targets_mean": 9516.2,
"valid_targets_min": 345
},
{
"epoch": 6.543001686340641,
"grad_norm": 0.37657294225932925,
"learning_rate": 5.211481107707594e-07,
"loss": 0.1806,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06967167556285858,
"step": 3880,
"valid_targets_mean": 7734.0,
"valid_targets_min": 459
},
{
"epoch": 6.551433389544688,
"grad_norm": 0.489758332221818,
"learning_rate": 5.02244218957153e-07,
"loss": 0.1807,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06643298268318176,
"step": 3885,
"valid_targets_mean": 6383.5,
"valid_targets_min": 354
},
{
"epoch": 6.559865092748735,
"grad_norm": 0.4303130737500959,
"learning_rate": 4.836851869085912e-07,
"loss": 0.2005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.083314910531044,
"step": 3890,
"valid_targets_mean": 7267.8,
"valid_targets_min": 847
},
{
"epoch": 6.568296795952783,
"grad_norm": 0.4034285591836649,
"learning_rate": 4.654713428814872e-07,
"loss": 0.1939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09345631301403046,
"step": 3895,
"valid_targets_mean": 11566.1,
"valid_targets_min": 675
},
{
"epoch": 6.57672849915683,
"grad_norm": 0.3803969636361645,
"learning_rate": 4.476030090268668e-07,
"loss": 0.1718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09347693622112274,
"step": 3900,
"valid_targets_mean": 10816.2,
"valid_targets_min": 920
},
{
"epoch": 6.585160202360877,
"grad_norm": 0.363474674513869,
"learning_rate": 4.3008050138465986e-07,
"loss": 0.1907,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0996813178062439,
"step": 3905,
"valid_targets_mean": 10656.5,
"valid_targets_min": 3872
},
{
"epoch": 6.593591905564924,
"grad_norm": 0.6224908161326272,
"learning_rate": 4.1290412987812223e-07,
"loss": 0.1829,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10408461093902588,
"step": 3910,
"valid_targets_mean": 5273.6,
"valid_targets_min": 381
},
{
"epoch": 6.602023608768971,
"grad_norm": 0.3820118876077167,
"learning_rate": 3.960741983083338e-07,
"loss": 0.178,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08072896301746368,
"step": 3915,
"valid_targets_mean": 10781.8,
"valid_targets_min": 976
},
{
"epoch": 6.610455311973018,
"grad_norm": 0.4083201485109933,
"learning_rate": 3.79591004348856e-07,
"loss": 0.1908,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07945813983678818,
"step": 3920,
"valid_targets_mean": 6424.4,
"valid_targets_min": 364
},
{
"epoch": 6.618887015177066,
"grad_norm": 0.4553759784656866,
"learning_rate": 3.6345483954044023e-07,
"loss": 0.1908,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06289969384670258,
"step": 3925,
"valid_targets_mean": 4279.8,
"valid_targets_min": 533
},
{
"epoch": 6.627318718381113,
"grad_norm": 0.32911050934059227,
"learning_rate": 3.476659892858791e-07,
"loss": 0.1843,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08094698190689087,
"step": 3930,
"valid_targets_mean": 12525.2,
"valid_targets_min": 744
},
{
"epoch": 6.63575042158516,
"grad_norm": 0.509363253976109,
"learning_rate": 3.322247328449635e-07,
"loss": 0.2061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12460733950138092,
"step": 3935,
"valid_targets_mean": 8913.8,
"valid_targets_min": 514
},
{
"epoch": 6.644182124789207,
"grad_norm": 0.3084520468384684,
"learning_rate": 3.1713134332954196e-07,
"loss": 0.1879,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05299391224980354,
"step": 3940,
"valid_targets_mean": 9411.2,
"valid_targets_min": 340
},
{
"epoch": 6.652613827993255,
"grad_norm": 0.34913707909326797,
"learning_rate": 3.0238608769868507e-07,
"loss": 0.1866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09478899836540222,
"step": 3945,
"valid_targets_mean": 12532.2,
"valid_targets_min": 1973
},
{
"epoch": 6.661045531197302,
"grad_norm": 0.3521666734928031,
"learning_rate": 2.879892267539619e-07,
"loss": 0.1981,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.106800377368927,
"step": 3950,
"valid_targets_mean": 11360.5,
"valid_targets_min": 3358
},
{
"epoch": 6.669477234401349,
"grad_norm": 0.5266740955307603,
"learning_rate": 2.739410151348398e-07,
"loss": 0.1995,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1133960708975792,
"step": 3955,
"valid_targets_mean": 6220.2,
"valid_targets_min": 304
},
{
"epoch": 6.677908937605396,
"grad_norm": 0.7229878003746588,
"learning_rate": 2.6024170131417006e-07,
"loss": 0.1821,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0908622071146965,
"step": 3960,
"valid_targets_mean": 9095.1,
"valid_targets_min": 2182
},
{
"epoch": 6.686340640809443,
"grad_norm": 0.3821560852666492,
"learning_rate": 2.468915275937933e-07,
"loss": 0.1862,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10074663162231445,
"step": 3965,
"valid_targets_mean": 13656.5,
"valid_targets_min": 1262
},
{
"epoch": 6.694772344013491,
"grad_norm": 0.4278114552962226,
"learning_rate": 2.338907301002591e-07,
"loss": 0.1949,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08412501215934753,
"step": 3970,
"valid_targets_mean": 7315.1,
"valid_targets_min": 496
},
{
"epoch": 6.703204047217538,
"grad_norm": 0.4122975342322843,
"learning_rate": 2.212395387806421e-07,
"loss": 0.171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.049946848303079605,
"step": 3975,
"valid_targets_mean": 4569.8,
"valid_targets_min": 336
},
{
"epoch": 6.7116357504215856,
"grad_norm": 0.41133716427599926,
"learning_rate": 2.0893817739848776e-07,
"loss": 0.1796,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11801201105117798,
"step": 3980,
"valid_targets_mean": 11205.4,
"valid_targets_min": 4372
},
{
"epoch": 6.720067453625632,
"grad_norm": 0.4058901354047069,
"learning_rate": 1.9698686352984e-07,
"loss": 0.1842,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07888654619455338,
"step": 3985,
"valid_targets_mean": 9879.1,
"valid_targets_min": 2324
},
{
"epoch": 6.728499156829679,
"grad_norm": 0.5447890367194643,
"learning_rate": 1.8538580855939737e-07,
"loss": 0.1968,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12148575484752655,
"step": 3990,
"valid_targets_mean": 8304.2,
"valid_targets_min": 689
},
{
"epoch": 6.736930860033727,
"grad_norm": 0.4615151417402277,
"learning_rate": 1.74135217676783e-07,
"loss": 0.2141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13751620054244995,
"step": 3995,
"valid_targets_mean": 10580.4,
"valid_targets_min": 1291
},
{
"epoch": 6.745362563237774,
"grad_norm": 0.49283827619760107,
"learning_rate": 1.6323528987290282e-07,
"loss": 0.1959,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10771192610263824,
"step": 4000,
"valid_targets_mean": 7280.6,
"valid_targets_min": 350
},
{
"epoch": 6.753794266441822,
"grad_norm": 0.3533313885283157,
"learning_rate": 1.5268621793643745e-07,
"loss": 0.187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07847689837217331,
"step": 4005,
"valid_targets_mean": 8697.6,
"valid_targets_min": 325
},
{
"epoch": 6.762225969645868,
"grad_norm": 0.4735294541383007,
"learning_rate": 1.4248818845041813e-07,
"loss": 0.1782,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1132201999425888,
"step": 4010,
"valid_targets_mean": 7712.0,
"valid_targets_min": 354
},
{
"epoch": 6.770657672849916,
"grad_norm": 0.489860771207464,
"learning_rate": 1.326413817889427e-07,
"loss": 0.1968,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12657737731933594,
"step": 4015,
"valid_targets_mean": 7175.0,
"valid_targets_min": 954
},
{
"epoch": 6.779089376053963,
"grad_norm": 0.4454708210625732,
"learning_rate": 1.2314597211397383e-07,
"loss": 0.1876,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08421581238508224,
"step": 4020,
"valid_targets_mean": 7775.4,
"valid_targets_min": 670
},
{
"epoch": 6.78752107925801,
"grad_norm": 0.5953361581533753,
"learning_rate": 1.1400212737226357e-07,
"loss": 0.1973,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15385916829109192,
"step": 4025,
"valid_targets_mean": 9723.5,
"valid_targets_min": 1545
},
{
"epoch": 6.795952782462058,
"grad_norm": 0.47875771190702116,
"learning_rate": 1.0521000929238023e-07,
"loss": 0.1861,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10643937438726425,
"step": 4030,
"valid_targets_mean": 7134.4,
"valid_targets_min": 2042
},
{
"epoch": 6.804384485666104,
"grad_norm": 0.4601576179120863,
"learning_rate": 9.676977338185067e-08,
"loss": 0.1779,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07497909665107727,
"step": 4035,
"valid_targets_mean": 4190.4,
"valid_targets_min": 364
},
{
"epoch": 6.812816188870151,
"grad_norm": 0.4494519508138053,
"learning_rate": 8.868156892440915e-08,
"loss": 0.1849,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06891386210918427,
"step": 4040,
"valid_targets_mean": 6610.9,
"valid_targets_min": 885
},
{
"epoch": 6.821247892074199,
"grad_norm": 0.467701242366306,
"learning_rate": 8.094553897735725e-08,
"loss": 0.1953,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13082212209701538,
"step": 4045,
"valid_targets_mean": 9018.0,
"valid_targets_min": 285
},
{
"epoch": 6.829679595278246,
"grad_norm": 0.44396369664602076,
"learning_rate": 7.356182036902581e-08,
"loss": 0.1845,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09440740942955017,
"step": 4050,
"valid_targets_mean": 6712.1,
"valid_targets_min": 414
},
{
"epoch": 6.838111298482294,
"grad_norm": 0.47889821218002804,
"learning_rate": 6.653054369637035e-08,
"loss": 0.1943,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11628532409667969,
"step": 4055,
"valid_targets_mean": 7981.0,
"valid_targets_min": 823
},
{
"epoch": 6.8465430016863404,
"grad_norm": 0.48363734233322697,
"learning_rate": 5.985183332264388e-08,
"loss": 0.1907,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0821453109383583,
"step": 4060,
"valid_targets_mean": 7526.6,
"valid_targets_min": 865
},
{
"epoch": 6.854974704890388,
"grad_norm": 0.4302032727475301,
"learning_rate": 5.352580737521429e-08,
"loss": 0.1991,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12737418711185455,
"step": 4065,
"valid_targets_mean": 9708.8,
"valid_targets_min": 814
},
{
"epoch": 6.863406408094435,
"grad_norm": 0.39344642623473,
"learning_rate": 4.755257774346156e-08,
"loss": 0.1845,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10401811450719833,
"step": 4070,
"valid_targets_mean": 14280.1,
"valid_targets_min": 3238
},
{
"epoch": 6.871838111298482,
"grad_norm": 0.4464654101453913,
"learning_rate": 4.1932250076803794e-08,
"loss": 0.1928,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09252262115478516,
"step": 4075,
"valid_targets_mean": 10928.8,
"valid_targets_min": 958
},
{
"epoch": 6.88026981450253,
"grad_norm": 0.43320851215498196,
"learning_rate": 3.66649237828276e-08,
"loss": 0.1785,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1057780385017395,
"step": 4080,
"valid_targets_mean": 8528.8,
"valid_targets_min": 2957
},
{
"epoch": 6.8887015177065765,
"grad_norm": 0.4076199555405414,
"learning_rate": 3.1750692025536156e-08,
"loss": 0.1812,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0982946902513504,
"step": 4085,
"valid_targets_mean": 7892.0,
"valid_targets_min": 1461
},
{
"epoch": 6.897133220910624,
"grad_norm": 0.48872884480219175,
"learning_rate": 2.718964172368832e-08,
"loss": 0.1828,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07660526782274246,
"step": 4090,
"valid_targets_mean": 4845.9,
"valid_targets_min": 943
},
{
"epoch": 6.905564924114671,
"grad_norm": 0.5572251344793053,
"learning_rate": 2.2981853549275402e-08,
"loss": 0.1887,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11709411442279816,
"step": 4095,
"valid_targets_mean": 11627.6,
"valid_targets_min": 5151
},
{
"epoch": 6.913996627318719,
"grad_norm": 0.4949295277300872,
"learning_rate": 1.9127401926084532e-08,
"loss": 0.1895,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08785691112279892,
"step": 4100,
"valid_targets_mean": 6636.1,
"valid_targets_min": 1505
},
{
"epoch": 6.922428330522766,
"grad_norm": 0.4204847614941541,
"learning_rate": 1.5626355028388608e-08,
"loss": 0.183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08516311645507812,
"step": 4105,
"valid_targets_mean": 8957.0,
"valid_targets_min": 726
},
{
"epoch": 6.9308600337268125,
"grad_norm": 0.3992935865792739,
"learning_rate": 1.2478774779729474e-08,
"loss": 0.2106,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09479379653930664,
"step": 4110,
"valid_targets_mean": 7951.5,
"valid_targets_min": 294
},
{
"epoch": 6.93929173693086,
"grad_norm": 0.4333349486494216,
"learning_rate": 9.684716851843245e-09,
"loss": 0.1748,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08893856406211853,
"step": 4115,
"valid_targets_mean": 7130.9,
"valid_targets_min": 841
},
{
"epoch": 6.947723440134907,
"grad_norm": 0.352325908959814,
"learning_rate": 7.2442306636522075e-09,
"loss": 0.1716,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05718475952744484,
"step": 4120,
"valid_targets_mean": 6363.8,
"valid_targets_min": 987
},
{
"epoch": 6.956155143338955,
"grad_norm": 0.43718165158532285,
"learning_rate": 5.157359380409954e-09,
"loss": 0.1886,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11359106749296188,
"step": 4125,
"valid_targets_mean": 10956.0,
"valid_targets_min": 383
},
{
"epoch": 6.964586846543002,
"grad_norm": 0.48690295899601443,
"learning_rate": 3.4241399129197883e-09,
"loss": 0.1723,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07518358528614044,
"step": 4130,
"valid_targets_mean": 5246.6,
"valid_targets_min": 778
},
{
"epoch": 6.973018549747049,
"grad_norm": 0.431758458042417,
"learning_rate": 2.044602916901894e-09,
"loss": 0.1879,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09651060402393341,
"step": 4135,
"valid_targets_mean": 7997.0,
"valid_targets_min": 3606
},
{
"epoch": 6.981450252951096,
"grad_norm": 0.35135568021055524,
"learning_rate": 1.0187727924337864e-09,
"loss": 0.1903,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10192978382110596,
"step": 4140,
"valid_targets_mean": 9242.6,
"valid_targets_min": 5021
},
{
"epoch": 6.989881956155143,
"grad_norm": 0.5710269543348359,
"learning_rate": 3.4666768353064375e-10,
"loss": 0.1864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09490747004747391,
"step": 4145,
"valid_targets_mean": 4301.8,
"valid_targets_min": 665
},
{
"epoch": 6.998313659359191,
"grad_norm": 0.40148854512095367,
"learning_rate": 2.8299477818904965e-11,
"loss": 0.1877,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09348092973232269,
"step": 4150,
"valid_targets_mean": 9353.0,
"valid_targets_min": 375
},
{
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08177298307418823,
"step": 4151,
"total_flos": 2.1664892224861635e+18,
"train_loss": 0.23105641345867506,
"train_runtime": 62878.6857,
"train_samples_per_second": 1.056,
"train_steps_per_second": 0.066,
"valid_targets_mean": 5219.9,
"valid_targets_min": 359
}
],
"logging_steps": 5,
"max_steps": 4151,
"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": 2.1664892224861635e+18,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}