Files
g1_min_episodes_e1_gpt_long…/trainer_state.json
ModelHub XC c9cfe5c0fd 初始化项目,由ModelHub XC社区提供模型
Model: DCAgent/g1_min_episodes_e1_gpt_long_tacc
Source: Original Platform
2026-05-03 12:08:43 +08:00

9222 lines
256 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 7.0,
"eval_steps": 500,
"global_step": 4172,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.008389261744966443,
"grad_norm": 11.429916174694034,
"learning_rate": 3.827751196172249e-07,
"loss": 0.8029,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7515180706977844,
"step": 5,
"valid_targets_mean": 8496.2,
"valid_targets_min": 3717
},
{
"epoch": 0.016778523489932886,
"grad_norm": 11.304896612592945,
"learning_rate": 8.612440191387561e-07,
"loss": 0.7977,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.8521906137466431,
"step": 10,
"valid_targets_mean": 8275.7,
"valid_targets_min": 5484
},
{
"epoch": 0.025167785234899327,
"grad_norm": 10.737776300507816,
"learning_rate": 1.3397129186602873e-06,
"loss": 0.7715,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6780750751495361,
"step": 15,
"valid_targets_mean": 7025.2,
"valid_targets_min": 1165
},
{
"epoch": 0.03355704697986577,
"grad_norm": 7.680186888965264,
"learning_rate": 1.8181818181818183e-06,
"loss": 0.7771,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7101215124130249,
"step": 20,
"valid_targets_mean": 6943.1,
"valid_targets_min": 2898
},
{
"epoch": 0.04194630872483222,
"grad_norm": 5.533977671773871,
"learning_rate": 2.2966507177033497e-06,
"loss": 0.7319,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7111684083938599,
"step": 25,
"valid_targets_mean": 8003.8,
"valid_targets_min": 3472
},
{
"epoch": 0.050335570469798654,
"grad_norm": 3.2759545986806122,
"learning_rate": 2.7751196172248807e-06,
"loss": 0.6565,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6436762809753418,
"step": 30,
"valid_targets_mean": 7904.6,
"valid_targets_min": 3065
},
{
"epoch": 0.0587248322147651,
"grad_norm": 2.1640611214803243,
"learning_rate": 3.2535885167464117e-06,
"loss": 0.6337,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6023802757263184,
"step": 35,
"valid_targets_mean": 7352.1,
"valid_targets_min": 3473
},
{
"epoch": 0.06711409395973154,
"grad_norm": 1.385676572466764,
"learning_rate": 3.732057416267943e-06,
"loss": 0.6477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5922027826309204,
"step": 40,
"valid_targets_mean": 8099.6,
"valid_targets_min": 3148
},
{
"epoch": 0.07550335570469799,
"grad_norm": 1.169631616709564,
"learning_rate": 4.210526315789474e-06,
"loss": 0.5833,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5523771047592163,
"step": 45,
"valid_targets_mean": 7346.4,
"valid_targets_min": 4070
},
{
"epoch": 0.08389261744966443,
"grad_norm": 0.995287650972169,
"learning_rate": 4.6889952153110055e-06,
"loss": 0.5862,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5546296238899231,
"step": 50,
"valid_targets_mean": 7365.1,
"valid_targets_min": 3180
},
{
"epoch": 0.09228187919463088,
"grad_norm": 0.9232100535542159,
"learning_rate": 5.167464114832537e-06,
"loss": 0.5651,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5748593211174011,
"step": 55,
"valid_targets_mean": 7270.7,
"valid_targets_min": 3439
},
{
"epoch": 0.10067114093959731,
"grad_norm": 0.8804743342181288,
"learning_rate": 5.645933014354067e-06,
"loss": 0.5593,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.558745265007019,
"step": 60,
"valid_targets_mean": 6689.4,
"valid_targets_min": 2870
},
{
"epoch": 0.10906040268456375,
"grad_norm": 0.6669968293281203,
"learning_rate": 6.124401913875598e-06,
"loss": 0.5485,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5656333565711975,
"step": 65,
"valid_targets_mean": 7673.3,
"valid_targets_min": 3397
},
{
"epoch": 0.1174496644295302,
"grad_norm": 0.5650591764206403,
"learning_rate": 6.6028708133971295e-06,
"loss": 0.5291,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4837729334831238,
"step": 70,
"valid_targets_mean": 7722.0,
"valid_targets_min": 4597
},
{
"epoch": 0.12583892617449666,
"grad_norm": 0.5626711579351399,
"learning_rate": 7.081339712918661e-06,
"loss": 0.504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5010769367218018,
"step": 75,
"valid_targets_mean": 7659.4,
"valid_targets_min": 2432
},
{
"epoch": 0.1342281879194631,
"grad_norm": 0.558815995790594,
"learning_rate": 7.5598086124401915e-06,
"loss": 0.483,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5433139801025391,
"step": 80,
"valid_targets_mean": 7931.2,
"valid_targets_min": 3733
},
{
"epoch": 0.14261744966442952,
"grad_norm": 0.5323244299099542,
"learning_rate": 8.038277511961722e-06,
"loss": 0.4863,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4403553009033203,
"step": 85,
"valid_targets_mean": 7204.1,
"valid_targets_min": 3650
},
{
"epoch": 0.15100671140939598,
"grad_norm": 0.5234182981110849,
"learning_rate": 8.516746411483254e-06,
"loss": 0.4688,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.47406071424484253,
"step": 90,
"valid_targets_mean": 6685.4,
"valid_targets_min": 2822
},
{
"epoch": 0.1593959731543624,
"grad_norm": 0.49999255866557624,
"learning_rate": 8.995215311004785e-06,
"loss": 0.472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45585063099861145,
"step": 95,
"valid_targets_mean": 7153.2,
"valid_targets_min": 2650
},
{
"epoch": 0.16778523489932887,
"grad_norm": 0.5107533377375633,
"learning_rate": 9.473684210526315e-06,
"loss": 0.4309,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41738754510879517,
"step": 100,
"valid_targets_mean": 6273.4,
"valid_targets_min": 2485
},
{
"epoch": 0.1761744966442953,
"grad_norm": 0.5692737338534906,
"learning_rate": 9.952153110047848e-06,
"loss": 0.4675,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.49137014150619507,
"step": 105,
"valid_targets_mean": 6593.8,
"valid_targets_min": 3230
},
{
"epoch": 0.18456375838926176,
"grad_norm": 0.4545919532663208,
"learning_rate": 1.0430622009569378e-05,
"loss": 0.4452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44480836391448975,
"step": 110,
"valid_targets_mean": 7739.9,
"valid_targets_min": 2932
},
{
"epoch": 0.1929530201342282,
"grad_norm": 0.577714358727239,
"learning_rate": 1.0909090909090909e-05,
"loss": 0.4479,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.46256864070892334,
"step": 115,
"valid_targets_mean": 8366.8,
"valid_targets_min": 2430
},
{
"epoch": 0.20134228187919462,
"grad_norm": 0.5355160779727121,
"learning_rate": 1.1387559808612441e-05,
"loss": 0.4206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4319803714752197,
"step": 120,
"valid_targets_mean": 7350.1,
"valid_targets_min": 4405
},
{
"epoch": 0.20973154362416108,
"grad_norm": 0.8404470547291075,
"learning_rate": 1.1866028708133972e-05,
"loss": 0.4221,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5280655026435852,
"step": 125,
"valid_targets_mean": 5855.0,
"valid_targets_min": 2012
},
{
"epoch": 0.2181208053691275,
"grad_norm": 0.884879506452225,
"learning_rate": 1.2344497607655504e-05,
"loss": 0.5864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6223379969596863,
"step": 130,
"valid_targets_mean": 4784.8,
"valid_targets_min": 2094
},
{
"epoch": 0.22651006711409397,
"grad_norm": 0.8886013989047753,
"learning_rate": 1.2822966507177035e-05,
"loss": 0.5988,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6249760389328003,
"step": 135,
"valid_targets_mean": 3623.2,
"valid_targets_min": 1395
},
{
"epoch": 0.2348993288590604,
"grad_norm": 0.7498606562311375,
"learning_rate": 1.3301435406698567e-05,
"loss": 0.57,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.6269269585609436,
"step": 140,
"valid_targets_mean": 4250.2,
"valid_targets_min": 2286
},
{
"epoch": 0.24328859060402686,
"grad_norm": 0.6857742614223888,
"learning_rate": 1.3779904306220097e-05,
"loss": 0.5547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5776923894882202,
"step": 145,
"valid_targets_mean": 5208.2,
"valid_targets_min": 2893
},
{
"epoch": 0.2516778523489933,
"grad_norm": 0.6468252829304791,
"learning_rate": 1.4258373205741626e-05,
"loss": 0.5314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.48802947998046875,
"step": 150,
"valid_targets_mean": 4094.2,
"valid_targets_min": 2428
},
{
"epoch": 0.2600671140939597,
"grad_norm": 0.7024802120973274,
"learning_rate": 1.4736842105263159e-05,
"loss": 0.5609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5435975790023804,
"step": 155,
"valid_targets_mean": 3976.8,
"valid_targets_min": 2196
},
{
"epoch": 0.2684563758389262,
"grad_norm": 0.6244080831093797,
"learning_rate": 1.5215311004784689e-05,
"loss": 0.5432,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5432044267654419,
"step": 160,
"valid_targets_mean": 4607.0,
"valid_targets_min": 2387
},
{
"epoch": 0.27684563758389263,
"grad_norm": 0.7859205405156159,
"learning_rate": 1.569377990430622e-05,
"loss": 0.5299,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5337243676185608,
"step": 165,
"valid_targets_mean": 3891.6,
"valid_targets_min": 2396
},
{
"epoch": 0.28523489932885904,
"grad_norm": 0.6249553291441904,
"learning_rate": 1.6172248803827754e-05,
"loss": 0.5199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.48773542046546936,
"step": 170,
"valid_targets_mean": 4769.2,
"valid_targets_min": 2532
},
{
"epoch": 0.2936241610738255,
"grad_norm": 0.6425081204598898,
"learning_rate": 1.6650717703349284e-05,
"loss": 0.5152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4805818796157837,
"step": 175,
"valid_targets_mean": 4180.4,
"valid_targets_min": 2412
},
{
"epoch": 0.30201342281879195,
"grad_norm": 0.644697648163711,
"learning_rate": 1.7129186602870815e-05,
"loss": 0.5358,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5350947380065918,
"step": 180,
"valid_targets_mean": 4524.7,
"valid_targets_min": 1851
},
{
"epoch": 0.3104026845637584,
"grad_norm": 0.6397675681350429,
"learning_rate": 1.7607655502392345e-05,
"loss": 0.5147,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4973926246166229,
"step": 185,
"valid_targets_mean": 4274.9,
"valid_targets_min": 2163
},
{
"epoch": 0.3187919463087248,
"grad_norm": 0.7096262995210569,
"learning_rate": 1.8086124401913876e-05,
"loss": 0.5015,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5184868574142456,
"step": 190,
"valid_targets_mean": 3913.9,
"valid_targets_min": 2006
},
{
"epoch": 0.3271812080536913,
"grad_norm": 0.8716944536624738,
"learning_rate": 1.8564593301435407e-05,
"loss": 0.4985,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.48212093114852905,
"step": 195,
"valid_targets_mean": 5175.9,
"valid_targets_min": 2342
},
{
"epoch": 0.33557046979865773,
"grad_norm": 0.7263035771601677,
"learning_rate": 1.904306220095694e-05,
"loss": 0.4964,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5013434886932373,
"step": 200,
"valid_targets_mean": 4147.4,
"valid_targets_min": 2188
},
{
"epoch": 0.34395973154362414,
"grad_norm": 0.5109831835045102,
"learning_rate": 1.952153110047847e-05,
"loss": 0.4847,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4357479214668274,
"step": 205,
"valid_targets_mean": 6421.8,
"valid_targets_min": 2865
},
{
"epoch": 0.3523489932885906,
"grad_norm": 0.7306050630698964,
"learning_rate": 2e-05,
"loss": 0.4729,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.516329288482666,
"step": 210,
"valid_targets_mean": 6181.7,
"valid_targets_min": 2653
},
{
"epoch": 0.36073825503355705,
"grad_norm": 0.5894334301414923,
"learning_rate": 2.0478468899521532e-05,
"loss": 0.4411,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4444805979728699,
"step": 215,
"valid_targets_mean": 4921.6,
"valid_targets_min": 1130
},
{
"epoch": 0.3691275167785235,
"grad_norm": 0.6038404012268255,
"learning_rate": 2.0956937799043063e-05,
"loss": 0.4686,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4492901563644409,
"step": 220,
"valid_targets_mean": 4941.5,
"valid_targets_min": 2248
},
{
"epoch": 0.3775167785234899,
"grad_norm": 0.5368532521629573,
"learning_rate": 2.1435406698564597e-05,
"loss": 0.4244,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4112545847892761,
"step": 225,
"valid_targets_mean": 5334.0,
"valid_targets_min": 2202
},
{
"epoch": 0.3859060402684564,
"grad_norm": 0.5398889974747488,
"learning_rate": 2.1913875598086127e-05,
"loss": 0.4441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4427616000175476,
"step": 230,
"valid_targets_mean": 5717.2,
"valid_targets_min": 1693
},
{
"epoch": 0.39429530201342283,
"grad_norm": 0.6504692937019585,
"learning_rate": 2.2392344497607658e-05,
"loss": 0.4527,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.49989527463912964,
"step": 235,
"valid_targets_mean": 5333.2,
"valid_targets_min": 2446
},
{
"epoch": 0.40268456375838924,
"grad_norm": 0.6058996240121997,
"learning_rate": 2.287081339712919e-05,
"loss": 0.452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4401355981826782,
"step": 240,
"valid_targets_mean": 4057.7,
"valid_targets_min": 2037
},
{
"epoch": 0.4110738255033557,
"grad_norm": 0.5436440925207906,
"learning_rate": 2.3349282296650722e-05,
"loss": 0.4386,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40055838227272034,
"step": 245,
"valid_targets_mean": 5323.5,
"valid_targets_min": 2568
},
{
"epoch": 0.41946308724832215,
"grad_norm": 0.5244406484607159,
"learning_rate": 2.3827751196172253e-05,
"loss": 0.4459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4190705120563507,
"step": 250,
"valid_targets_mean": 6049.9,
"valid_targets_min": 2894
},
{
"epoch": 0.4278523489932886,
"grad_norm": 0.5870528969037385,
"learning_rate": 2.430622009569378e-05,
"loss": 0.4539,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4495871663093567,
"step": 255,
"valid_targets_mean": 4810.2,
"valid_targets_min": 2273
},
{
"epoch": 0.436241610738255,
"grad_norm": 0.5479113458584991,
"learning_rate": 2.478468899521531e-05,
"loss": 0.4352,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.436789870262146,
"step": 260,
"valid_targets_mean": 5604.8,
"valid_targets_min": 1967
},
{
"epoch": 0.4446308724832215,
"grad_norm": 0.6440537028665174,
"learning_rate": 2.526315789473684e-05,
"loss": 0.4466,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4923563599586487,
"step": 265,
"valid_targets_mean": 4861.6,
"valid_targets_min": 2047
},
{
"epoch": 0.45302013422818793,
"grad_norm": 0.551766371057539,
"learning_rate": 2.5741626794258375e-05,
"loss": 0.4307,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4165174961090088,
"step": 270,
"valid_targets_mean": 5708.2,
"valid_targets_min": 2660
},
{
"epoch": 0.46140939597315433,
"grad_norm": 0.5649909158521907,
"learning_rate": 2.6220095693779906e-05,
"loss": 0.4277,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.423203706741333,
"step": 275,
"valid_targets_mean": 5425.9,
"valid_targets_min": 1784
},
{
"epoch": 0.4697986577181208,
"grad_norm": 0.6980086453467145,
"learning_rate": 2.6698564593301436e-05,
"loss": 0.4453,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4403393566608429,
"step": 280,
"valid_targets_mean": 4510.3,
"valid_targets_min": 2328
},
{
"epoch": 0.47818791946308725,
"grad_norm": 0.6489874685006204,
"learning_rate": 2.7177033492822967e-05,
"loss": 0.4289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4224075376987457,
"step": 285,
"valid_targets_mean": 5093.3,
"valid_targets_min": 2668
},
{
"epoch": 0.4865771812080537,
"grad_norm": 0.66483255908447,
"learning_rate": 2.76555023923445e-05,
"loss": 0.4415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4568319320678711,
"step": 290,
"valid_targets_mean": 5909.3,
"valid_targets_min": 2398
},
{
"epoch": 0.4949664429530201,
"grad_norm": 0.547172056963068,
"learning_rate": 2.813397129186603e-05,
"loss": 0.434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4222297668457031,
"step": 295,
"valid_targets_mean": 5225.2,
"valid_targets_min": 2706
},
{
"epoch": 0.5033557046979866,
"grad_norm": 0.5955693953705371,
"learning_rate": 2.8612440191387562e-05,
"loss": 0.4385,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42588669061660767,
"step": 300,
"valid_targets_mean": 4588.1,
"valid_targets_min": 2204
},
{
"epoch": 0.511744966442953,
"grad_norm": 0.5710584591581995,
"learning_rate": 2.9090909090909093e-05,
"loss": 0.4189,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41106200218200684,
"step": 305,
"valid_targets_mean": 5412.2,
"valid_targets_min": 2473
},
{
"epoch": 0.5201342281879194,
"grad_norm": 0.5717962187504413,
"learning_rate": 2.9569377990430627e-05,
"loss": 0.4133,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3966302275657654,
"step": 310,
"valid_targets_mean": 5886.4,
"valid_targets_min": 2693
},
{
"epoch": 0.5285234899328859,
"grad_norm": 0.5896096749709734,
"learning_rate": 3.0047846889952157e-05,
"loss": 0.4418,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4589608907699585,
"step": 315,
"valid_targets_mean": 5643.7,
"valid_targets_min": 2087
},
{
"epoch": 0.5369127516778524,
"grad_norm": 0.6053308422150226,
"learning_rate": 3.052631578947369e-05,
"loss": 0.4131,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.414186030626297,
"step": 320,
"valid_targets_mean": 5636.5,
"valid_targets_min": 2342
},
{
"epoch": 0.5453020134228188,
"grad_norm": 0.5913310924717964,
"learning_rate": 3.100478468899522e-05,
"loss": 0.4001,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3703305721282959,
"step": 325,
"valid_targets_mean": 5548.0,
"valid_targets_min": 2174
},
{
"epoch": 0.5536912751677853,
"grad_norm": 0.5561479952898906,
"learning_rate": 3.148325358851675e-05,
"loss": 0.417,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43574556708335876,
"step": 330,
"valid_targets_mean": 5774.5,
"valid_targets_min": 2188
},
{
"epoch": 0.5620805369127517,
"grad_norm": 0.7888172523730118,
"learning_rate": 3.1961722488038276e-05,
"loss": 0.4233,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4162256121635437,
"step": 335,
"valid_targets_mean": 5900.6,
"valid_targets_min": 2320
},
{
"epoch": 0.5704697986577181,
"grad_norm": 0.5741463457887236,
"learning_rate": 3.244019138755981e-05,
"loss": 0.4334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4329298138618469,
"step": 340,
"valid_targets_mean": 5101.9,
"valid_targets_min": 1893
},
{
"epoch": 0.5788590604026845,
"grad_norm": 0.5607409905024663,
"learning_rate": 3.2918660287081344e-05,
"loss": 0.4226,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3906022906303406,
"step": 345,
"valid_targets_mean": 5646.4,
"valid_targets_min": 1766
},
{
"epoch": 0.587248322147651,
"grad_norm": 0.5549767545802637,
"learning_rate": 3.3397129186602875e-05,
"loss": 0.4149,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4468402862548828,
"step": 350,
"valid_targets_mean": 6108.6,
"valid_targets_min": 1998
},
{
"epoch": 0.5956375838926175,
"grad_norm": 0.5678095067095734,
"learning_rate": 3.3875598086124405e-05,
"loss": 0.4233,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4491395950317383,
"step": 355,
"valid_targets_mean": 5481.2,
"valid_targets_min": 2120
},
{
"epoch": 0.6040268456375839,
"grad_norm": 0.5552326561046017,
"learning_rate": 3.4354066985645936e-05,
"loss": 0.4194,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.412387490272522,
"step": 360,
"valid_targets_mean": 5890.2,
"valid_targets_min": 3253
},
{
"epoch": 0.6124161073825504,
"grad_norm": 0.627469143674013,
"learning_rate": 3.4832535885167466e-05,
"loss": 0.4397,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4570452570915222,
"step": 365,
"valid_targets_mean": 4708.5,
"valid_targets_min": 1846
},
{
"epoch": 0.6208053691275168,
"grad_norm": 0.6615735846017682,
"learning_rate": 3.5311004784689e-05,
"loss": 0.4108,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4102242588996887,
"step": 370,
"valid_targets_mean": 5491.2,
"valid_targets_min": 1920
},
{
"epoch": 0.6291946308724832,
"grad_norm": 0.6296619900954828,
"learning_rate": 3.578947368421053e-05,
"loss": 0.4226,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4719841480255127,
"step": 375,
"valid_targets_mean": 4956.1,
"valid_targets_min": 2345
},
{
"epoch": 0.6375838926174496,
"grad_norm": 0.5554628372527716,
"learning_rate": 3.626794258373206e-05,
"loss": 0.3858,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36292698979377747,
"step": 380,
"valid_targets_mean": 5131.4,
"valid_targets_min": 1719
},
{
"epoch": 0.6459731543624161,
"grad_norm": 0.6044788537839234,
"learning_rate": 3.6746411483253595e-05,
"loss": 0.4182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4428744316101074,
"step": 385,
"valid_targets_mean": 4511.2,
"valid_targets_min": 2213
},
{
"epoch": 0.6543624161073825,
"grad_norm": 0.580910814357443,
"learning_rate": 3.7224880382775126e-05,
"loss": 0.4073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43189340829849243,
"step": 390,
"valid_targets_mean": 5346.6,
"valid_targets_min": 2552
},
{
"epoch": 0.662751677852349,
"grad_norm": 0.5483656010470783,
"learning_rate": 3.7703349282296657e-05,
"loss": 0.4081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3961476683616638,
"step": 395,
"valid_targets_mean": 5379.6,
"valid_targets_min": 2402
},
{
"epoch": 0.6711409395973155,
"grad_norm": 0.6038431643889872,
"learning_rate": 3.818181818181819e-05,
"loss": 0.3978,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4378660023212433,
"step": 400,
"valid_targets_mean": 6561.0,
"valid_targets_min": 2063
},
{
"epoch": 0.6795302013422819,
"grad_norm": 0.5530116276983784,
"learning_rate": 3.866028708133972e-05,
"loss": 0.4066,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38150209188461304,
"step": 405,
"valid_targets_mean": 5516.9,
"valid_targets_min": 2094
},
{
"epoch": 0.6879194630872483,
"grad_norm": 0.4865194676585773,
"learning_rate": 3.913875598086125e-05,
"loss": 0.3972,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3626737892627716,
"step": 410,
"valid_targets_mean": 6917.1,
"valid_targets_min": 2649
},
{
"epoch": 0.6963087248322147,
"grad_norm": 0.5912590224147495,
"learning_rate": 3.961722488038278e-05,
"loss": 0.403,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38366463780403137,
"step": 415,
"valid_targets_mean": 5088.7,
"valid_targets_min": 2515
},
{
"epoch": 0.7046979865771812,
"grad_norm": 0.7908250017137203,
"learning_rate": 3.999999299656369e-05,
"loss": 0.42,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.46073949337005615,
"step": 420,
"valid_targets_mean": 3442.6,
"valid_targets_min": 2112
},
{
"epoch": 0.7130872483221476,
"grad_norm": 0.5870245117493919,
"learning_rate": 3.9999747876807854e-05,
"loss": 0.4234,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43893224000930786,
"step": 425,
"valid_targets_mean": 5594.4,
"valid_targets_min": 2832
},
{
"epoch": 0.7214765100671141,
"grad_norm": 0.5569426589223803,
"learning_rate": 3.999915259014132e-05,
"loss": 0.4104,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39234989881515503,
"step": 430,
"valid_targets_mean": 5105.4,
"valid_targets_min": 2265
},
{
"epoch": 0.7298657718120806,
"grad_norm": 0.552003952423719,
"learning_rate": 3.999820714698672e-05,
"loss": 0.4093,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38001513481140137,
"step": 435,
"valid_targets_mean": 5383.6,
"valid_targets_min": 2041
},
{
"epoch": 0.738255033557047,
"grad_norm": 0.5499205814382019,
"learning_rate": 3.9996911563897385e-05,
"loss": 0.4158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38236403465270996,
"step": 440,
"valid_targets_mean": 6037.8,
"valid_targets_min": 2956
},
{
"epoch": 0.7466442953020134,
"grad_norm": 0.648282130093116,
"learning_rate": 3.9995265863557134e-05,
"loss": 0.361,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34265291690826416,
"step": 445,
"valid_targets_mean": 6712.6,
"valid_targets_min": 3122
},
{
"epoch": 0.7550335570469798,
"grad_norm": 0.5341218249928086,
"learning_rate": 3.9993270074779816e-05,
"loss": 0.3056,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2955394983291626,
"step": 450,
"valid_targets_mean": 5881.0,
"valid_targets_min": 1614
},
{
"epoch": 0.7634228187919463,
"grad_norm": 0.5006261532403555,
"learning_rate": 3.9990924232508825e-05,
"loss": 0.3051,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26478704810142517,
"step": 455,
"valid_targets_mean": 5759.3,
"valid_targets_min": 2511
},
{
"epoch": 0.7718120805369127,
"grad_norm": 0.4690860350726601,
"learning_rate": 3.99882283778165e-05,
"loss": 0.3007,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3271682858467102,
"step": 460,
"valid_targets_mean": 6722.5,
"valid_targets_min": 3416
},
{
"epoch": 0.7802013422818792,
"grad_norm": 0.4574459537198095,
"learning_rate": 3.99851825579034e-05,
"loss": 0.2881,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2718518376350403,
"step": 465,
"valid_targets_mean": 6207.9,
"valid_targets_min": 3333
},
{
"epoch": 0.7885906040268457,
"grad_norm": 0.4738737124309521,
"learning_rate": 3.998178682609746e-05,
"loss": 0.3043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32907524704933167,
"step": 470,
"valid_targets_mean": 6317.9,
"valid_targets_min": 3749
},
{
"epoch": 0.7969798657718121,
"grad_norm": 0.6429859269520458,
"learning_rate": 3.997804124185307e-05,
"loss": 0.2814,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2754654884338379,
"step": 475,
"valid_targets_mean": 5869.5,
"valid_targets_min": 3804
},
{
"epoch": 0.8053691275167785,
"grad_norm": 0.4421765489204953,
"learning_rate": 3.997394587075004e-05,
"loss": 0.3087,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30648085474967957,
"step": 480,
"valid_targets_mean": 6835.0,
"valid_targets_min": 3131
},
{
"epoch": 0.8137583892617449,
"grad_norm": 0.47260237115341625,
"learning_rate": 3.996950078449245e-05,
"loss": 0.2891,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28972387313842773,
"step": 485,
"valid_targets_mean": 6305.6,
"valid_targets_min": 3498
},
{
"epoch": 0.8221476510067114,
"grad_norm": 0.5415368111551728,
"learning_rate": 3.9964706060907385e-05,
"loss": 0.2837,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2475288212299347,
"step": 490,
"valid_targets_mean": 7068.8,
"valid_targets_min": 4193
},
{
"epoch": 0.8305369127516778,
"grad_norm": 0.5625730961753714,
"learning_rate": 3.9959561783943585e-05,
"loss": 0.321,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3490103483200073,
"step": 495,
"valid_targets_mean": 5792.0,
"valid_targets_min": 1274
},
{
"epoch": 0.8389261744966443,
"grad_norm": 0.4875004918737747,
"learning_rate": 3.9954068043669966e-05,
"loss": 0.2806,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36193621158599854,
"step": 500,
"valid_targets_mean": 7094.4,
"valid_targets_min": 4609
},
{
"epoch": 0.8473154362416108,
"grad_norm": 0.4101522124633319,
"learning_rate": 3.994822493627403e-05,
"loss": 0.2682,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2873650789260864,
"step": 505,
"valid_targets_mean": 7215.2,
"valid_targets_min": 3897
},
{
"epoch": 0.8557046979865772,
"grad_norm": 0.4781257402151834,
"learning_rate": 3.9942032564060225e-05,
"loss": 0.2774,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28978726267814636,
"step": 510,
"valid_targets_mean": 6558.3,
"valid_targets_min": 3351
},
{
"epoch": 0.8640939597315436,
"grad_norm": 0.42936514162823514,
"learning_rate": 3.99354910354481e-05,
"loss": 0.2997,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31615179777145386,
"step": 515,
"valid_targets_mean": 6824.3,
"valid_targets_min": 2436
},
{
"epoch": 0.87248322147651,
"grad_norm": 0.41376060988986474,
"learning_rate": 3.9928600464970445e-05,
"loss": 0.2611,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24269354343414307,
"step": 520,
"valid_targets_mean": 6103.6,
"valid_targets_min": 3898
},
{
"epoch": 0.8808724832214765,
"grad_norm": 0.47130177205334933,
"learning_rate": 3.9921360973271266e-05,
"loss": 0.2813,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2576438784599304,
"step": 525,
"valid_targets_mean": 6659.6,
"valid_targets_min": 3256
},
{
"epoch": 0.889261744966443,
"grad_norm": 0.5220966461848282,
"learning_rate": 3.9913772687103695e-05,
"loss": 0.2704,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24460357427597046,
"step": 530,
"valid_targets_mean": 5699.6,
"valid_targets_min": 3248
},
{
"epoch": 0.8976510067114094,
"grad_norm": 0.42277794317550865,
"learning_rate": 3.990583573932774e-05,
"loss": 0.2852,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2582475244998932,
"step": 535,
"valid_targets_mean": 7132.4,
"valid_targets_min": 3729
},
{
"epoch": 0.9060402684563759,
"grad_norm": 0.45911239233133105,
"learning_rate": 3.9897550268907955e-05,
"loss": 0.2821,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26020509004592896,
"step": 540,
"valid_targets_mean": 5948.7,
"valid_targets_min": 2287
},
{
"epoch": 0.9144295302013423,
"grad_norm": 0.45306664341179165,
"learning_rate": 3.988891642091108e-05,
"loss": 0.2745,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2483942210674286,
"step": 545,
"valid_targets_mean": 5960.7,
"valid_targets_min": 1667
},
{
"epoch": 0.9228187919463087,
"grad_norm": 0.45005593832383756,
"learning_rate": 3.9879934346503395e-05,
"loss": 0.2789,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27685582637786865,
"step": 550,
"valid_targets_mean": 6103.1,
"valid_targets_min": 4121
},
{
"epoch": 0.9312080536912751,
"grad_norm": 0.4630172640729001,
"learning_rate": 3.987060420294815e-05,
"loss": 0.2718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2772757112979889,
"step": 555,
"valid_targets_mean": 7280.8,
"valid_targets_min": 4421
},
{
"epoch": 0.9395973154362416,
"grad_norm": 0.5905547100848773,
"learning_rate": 3.9860926153602785e-05,
"loss": 0.2818,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2750881314277649,
"step": 560,
"valid_targets_mean": 5570.1,
"valid_targets_min": 3241
},
{
"epoch": 0.947986577181208,
"grad_norm": 0.5418578863071151,
"learning_rate": 3.985090036791606e-05,
"loss": 0.2712,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24656403064727783,
"step": 565,
"valid_targets_mean": 5923.2,
"valid_targets_min": 1037
},
{
"epoch": 0.9563758389261745,
"grad_norm": 0.4794976278033037,
"learning_rate": 3.9840527021425115e-05,
"loss": 0.2608,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26753291487693787,
"step": 570,
"valid_targets_mean": 6194.7,
"valid_targets_min": 3559
},
{
"epoch": 0.964765100671141,
"grad_norm": 0.48506846188834934,
"learning_rate": 3.9829806295752374e-05,
"loss": 0.2756,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2682664394378662,
"step": 575,
"valid_targets_mean": 5676.1,
"valid_targets_min": 3539
},
{
"epoch": 0.9731543624161074,
"grad_norm": 0.4493346429692844,
"learning_rate": 3.981873837860236e-05,
"loss": 0.2946,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2867574989795685,
"step": 580,
"valid_targets_mean": 6225.4,
"valid_targets_min": 3347
},
{
"epoch": 0.9815436241610739,
"grad_norm": 0.44332825538625303,
"learning_rate": 3.980732346375846e-05,
"loss": 0.283,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25628867745399475,
"step": 585,
"valid_targets_mean": 6355.5,
"valid_targets_min": 3652
},
{
"epoch": 0.9899328859060402,
"grad_norm": 0.4398555371126775,
"learning_rate": 3.979556175107943e-05,
"loss": 0.2934,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27157896757125854,
"step": 590,
"valid_targets_mean": 6742.0,
"valid_targets_min": 4016
},
{
"epoch": 0.9983221476510067,
"grad_norm": 0.44388257975561,
"learning_rate": 3.978345344649603e-05,
"loss": 0.267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27930933237075806,
"step": 595,
"valid_targets_mean": 7195.1,
"valid_targets_min": 3331
},
{
"epoch": 1.0067114093959733,
"grad_norm": 0.4918875160068219,
"learning_rate": 3.97709987620073e-05,
"loss": 0.3834,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4197412431240082,
"step": 600,
"valid_targets_mean": 7649.6,
"valid_targets_min": 3784
},
{
"epoch": 1.0151006711409396,
"grad_norm": 0.5310303672634428,
"learning_rate": 3.9758197915676904e-05,
"loss": 0.3687,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3886641561985016,
"step": 605,
"valid_targets_mean": 8470.1,
"valid_targets_min": 2859
},
{
"epoch": 1.023489932885906,
"grad_norm": 0.45174431853802494,
"learning_rate": 3.974505113162931e-05,
"loss": 0.3857,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3814656734466553,
"step": 610,
"valid_targets_mean": 7579.4,
"valid_targets_min": 2671
},
{
"epoch": 1.0318791946308725,
"grad_norm": 0.45128010552811726,
"learning_rate": 3.973155864004587e-05,
"loss": 0.3768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3784426748752594,
"step": 615,
"valid_targets_mean": 7706.4,
"valid_targets_min": 2385
},
{
"epoch": 1.0402684563758389,
"grad_norm": 0.5026432056256217,
"learning_rate": 3.971772067716074e-05,
"loss": 0.3696,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3807569146156311,
"step": 620,
"valid_targets_mean": 6949.2,
"valid_targets_min": 2196
},
{
"epoch": 1.0486577181208054,
"grad_norm": 0.4451124033509271,
"learning_rate": 3.970353748525684e-05,
"loss": 0.362,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3572586178779602,
"step": 625,
"valid_targets_mean": 7776.6,
"valid_targets_min": 3797
},
{
"epoch": 1.0570469798657718,
"grad_norm": 0.45463886405258913,
"learning_rate": 3.96890093126615e-05,
"loss": 0.3634,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3626330494880676,
"step": 630,
"valid_targets_mean": 7734.9,
"valid_targets_min": 3678
},
{
"epoch": 1.0654362416107384,
"grad_norm": 0.5349550966760981,
"learning_rate": 3.967413641374222e-05,
"loss": 0.384,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37619084119796753,
"step": 635,
"valid_targets_mean": 6914.5,
"valid_targets_min": 2870
},
{
"epoch": 1.0738255033557047,
"grad_norm": 0.4731864320121969,
"learning_rate": 3.9658919048902105e-05,
"loss": 0.3524,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3684079647064209,
"step": 640,
"valid_targets_mean": 7338.3,
"valid_targets_min": 1373
},
{
"epoch": 1.082214765100671,
"grad_norm": 0.5129409519188545,
"learning_rate": 3.9643357484575416e-05,
"loss": 0.3629,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40892213582992554,
"step": 645,
"valid_targets_mean": 7915.4,
"valid_targets_min": 2332
},
{
"epoch": 1.0906040268456376,
"grad_norm": 0.481788755603101,
"learning_rate": 3.962745199322282e-05,
"loss": 0.3538,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33942824602127075,
"step": 650,
"valid_targets_mean": 7457.8,
"valid_targets_min": 2888
},
{
"epoch": 1.098993288590604,
"grad_norm": 0.473979688800083,
"learning_rate": 3.961120285332668e-05,
"loss": 0.3632,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3505626320838928,
"step": 655,
"valid_targets_mean": 6991.1,
"valid_targets_min": 2394
},
{
"epoch": 1.1073825503355705,
"grad_norm": 0.45358904944497436,
"learning_rate": 3.959461034938613e-05,
"loss": 0.3618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3826666474342346,
"step": 660,
"valid_targets_mean": 7398.9,
"valid_targets_min": 3358
},
{
"epoch": 1.1157718120805369,
"grad_norm": 0.5846108369658202,
"learning_rate": 3.957767477191214e-05,
"loss": 0.3726,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36829644441604614,
"step": 665,
"valid_targets_mean": 6854.7,
"valid_targets_min": 1726
},
{
"epoch": 1.1241610738255035,
"grad_norm": 0.47276083457917234,
"learning_rate": 3.9560396417422353e-05,
"loss": 0.3439,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34974294900894165,
"step": 670,
"valid_targets_mean": 6910.1,
"valid_targets_min": 3314
},
{
"epoch": 1.1325503355704698,
"grad_norm": 0.5301998627075942,
"learning_rate": 3.954277558843602e-05,
"loss": 0.3329,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3520258069038391,
"step": 675,
"valid_targets_mean": 6882.6,
"valid_targets_min": 3063
},
{
"epoch": 1.1409395973154361,
"grad_norm": 0.4919053681795801,
"learning_rate": 3.952481259346858e-05,
"loss": 0.3648,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3271327614784241,
"step": 680,
"valid_targets_mean": 7284.4,
"valid_targets_min": 1161
},
{
"epoch": 1.1493288590604027,
"grad_norm": 0.46496702920155447,
"learning_rate": 3.950650774702631e-05,
"loss": 0.3317,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.375588595867157,
"step": 685,
"valid_targets_mean": 6990.2,
"valid_targets_min": 1791
},
{
"epoch": 1.157718120805369,
"grad_norm": 0.47632054224978215,
"learning_rate": 3.9487861369600846e-05,
"loss": 0.3524,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36229509115219116,
"step": 690,
"valid_targets_mean": 7225.6,
"valid_targets_min": 3609
},
{
"epoch": 1.1661073825503356,
"grad_norm": 0.4703395549281827,
"learning_rate": 3.946887378766351e-05,
"loss": 0.3248,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.332710325717926,
"step": 695,
"valid_targets_mean": 7060.9,
"valid_targets_min": 2549
},
{
"epoch": 1.174496644295302,
"grad_norm": 0.4899534692866401,
"learning_rate": 3.944954533365964e-05,
"loss": 0.343,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3061160445213318,
"step": 700,
"valid_targets_mean": 6822.6,
"valid_targets_min": 2505
},
{
"epoch": 1.1828859060402686,
"grad_norm": 0.4286459332210617,
"learning_rate": 3.9429876346002755e-05,
"loss": 0.348,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3403472304344177,
"step": 705,
"valid_targets_mean": 8112.8,
"valid_targets_min": 1985
},
{
"epoch": 1.191275167785235,
"grad_norm": 0.4406751869329674,
"learning_rate": 3.940986716906863e-05,
"loss": 0.3427,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3227345049381256,
"step": 710,
"valid_targets_mean": 7395.3,
"valid_targets_min": 3697
},
{
"epoch": 1.1996644295302012,
"grad_norm": 0.4733460815311179,
"learning_rate": 3.938951815318926e-05,
"loss": 0.3323,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3289448022842407,
"step": 715,
"valid_targets_mean": 6519.2,
"valid_targets_min": 1969
},
{
"epoch": 1.2080536912751678,
"grad_norm": 0.5411475724037658,
"learning_rate": 3.9368829654646734e-05,
"loss": 0.3161,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2890268862247467,
"step": 720,
"valid_targets_mean": 8276.1,
"valid_targets_min": 6762
},
{
"epoch": 1.2164429530201342,
"grad_norm": 0.6378899815113951,
"learning_rate": 3.9347802035667014e-05,
"loss": 0.4495,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4517311453819275,
"step": 725,
"valid_targets_mean": 4914.0,
"valid_targets_min": 2880
},
{
"epoch": 1.2248322147651007,
"grad_norm": 0.763443367566175,
"learning_rate": 3.932643566441356e-05,
"loss": 0.4763,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4668845534324646,
"step": 730,
"valid_targets_mean": 4679.4,
"valid_targets_min": 1761
},
{
"epoch": 1.233221476510067,
"grad_norm": 0.6697593205294177,
"learning_rate": 3.930473091498088e-05,
"loss": 0.4574,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44335824251174927,
"step": 735,
"valid_targets_mean": 4310.0,
"valid_targets_min": 2294
},
{
"epoch": 1.2416107382550337,
"grad_norm": 0.6466850171189275,
"learning_rate": 3.9282688167388025e-05,
"loss": 0.4633,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.410359263420105,
"step": 740,
"valid_targets_mean": 4812.1,
"valid_targets_min": 2737
},
{
"epoch": 1.25,
"grad_norm": 0.6342579807355135,
"learning_rate": 3.926030780757192e-05,
"loss": 0.454,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4025290310382843,
"step": 745,
"valid_targets_mean": 4642.6,
"valid_targets_min": 2195
},
{
"epoch": 1.2583892617449663,
"grad_norm": 0.6920352891192058,
"learning_rate": 3.923759022738055e-05,
"loss": 0.4562,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4718177914619446,
"step": 750,
"valid_targets_mean": 4054.5,
"valid_targets_min": 2282
},
{
"epoch": 1.266778523489933,
"grad_norm": 0.5670671818223533,
"learning_rate": 3.921453582456618e-05,
"loss": 0.452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4622154235839844,
"step": 755,
"valid_targets_mean": 5353.2,
"valid_targets_min": 1488
},
{
"epoch": 1.2751677852348993,
"grad_norm": 0.6162714789949248,
"learning_rate": 3.9191145002778364e-05,
"loss": 0.4483,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44496333599090576,
"step": 760,
"valid_targets_mean": 5255.4,
"valid_targets_min": 2195
},
{
"epoch": 1.2835570469798658,
"grad_norm": 0.6656888218443957,
"learning_rate": 3.9167418171556836e-05,
"loss": 0.4448,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.42560693621635437,
"step": 765,
"valid_targets_mean": 4070.5,
"valid_targets_min": 2035
},
{
"epoch": 1.2919463087248322,
"grad_norm": 0.6788263203506376,
"learning_rate": 3.9143355746324415e-05,
"loss": 0.4371,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4444505572319031,
"step": 770,
"valid_targets_mean": 4327.9,
"valid_targets_min": 2313
},
{
"epoch": 1.3003355704697985,
"grad_norm": 0.6042180471896506,
"learning_rate": 3.911895814837964e-05,
"loss": 0.4444,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4616854190826416,
"step": 775,
"valid_targets_mean": 4667.8,
"valid_targets_min": 1749
},
{
"epoch": 1.308724832214765,
"grad_norm": 0.8379564960108428,
"learning_rate": 3.90942258048895e-05,
"loss": 0.4448,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43143102526664734,
"step": 780,
"valid_targets_mean": 4554.0,
"valid_targets_min": 1891
},
{
"epoch": 1.3171140939597317,
"grad_norm": 0.5987165486002523,
"learning_rate": 3.906915914888184e-05,
"loss": 0.4245,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40693628787994385,
"step": 785,
"valid_targets_mean": 4446.8,
"valid_targets_min": 2098
},
{
"epoch": 1.325503355704698,
"grad_norm": 0.6116395444800412,
"learning_rate": 3.904375861923789e-05,
"loss": 0.4223,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36410653591156006,
"step": 790,
"valid_targets_mean": 4459.8,
"valid_targets_min": 2331
},
{
"epoch": 1.3338926174496644,
"grad_norm": 0.5515440713674635,
"learning_rate": 3.901802466068449e-05,
"loss": 0.3979,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4054684042930603,
"step": 795,
"valid_targets_mean": 6275.2,
"valid_targets_min": 2533
},
{
"epoch": 1.342281879194631,
"grad_norm": 0.560316520971745,
"learning_rate": 3.8991957723786365e-05,
"loss": 0.4032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3985890746116638,
"step": 800,
"valid_targets_mean": 5172.5,
"valid_targets_min": 1986
},
{
"epoch": 1.3506711409395973,
"grad_norm": 0.5139257968050499,
"learning_rate": 3.8965558264938215e-05,
"loss": 0.3695,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3865506947040558,
"step": 805,
"valid_targets_mean": 5646.0,
"valid_targets_min": 2711
},
{
"epoch": 1.3590604026845639,
"grad_norm": 0.49640559975442256,
"learning_rate": 3.893882674635671e-05,
"loss": 0.3732,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3451784551143646,
"step": 810,
"valid_targets_mean": 5899.6,
"valid_targets_min": 2783
},
{
"epoch": 1.3674496644295302,
"grad_norm": 0.5472589862344993,
"learning_rate": 3.8911763636072407e-05,
"loss": 0.3845,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4126671254634857,
"step": 815,
"valid_targets_mean": 5326.4,
"valid_targets_min": 2250
},
{
"epoch": 1.3758389261744965,
"grad_norm": 0.5161484905741421,
"learning_rate": 3.888436940792156e-05,
"loss": 0.3541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35769641399383545,
"step": 820,
"valid_targets_mean": 5666.2,
"valid_targets_min": 2615
},
{
"epoch": 1.3842281879194631,
"grad_norm": 0.6251770756641752,
"learning_rate": 3.885664454153785e-05,
"loss": 0.3614,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41325899958610535,
"step": 825,
"valid_targets_mean": 5925.5,
"valid_targets_min": 2441
},
{
"epoch": 1.3926174496644295,
"grad_norm": 0.5768409575573539,
"learning_rate": 3.882858952234391e-05,
"loss": 0.3654,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4020509421825409,
"step": 830,
"valid_targets_mean": 5128.5,
"valid_targets_min": 2734
},
{
"epoch": 1.401006711409396,
"grad_norm": 0.5092759848211265,
"learning_rate": 3.880020484154292e-05,
"loss": 0.3863,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35542160272598267,
"step": 835,
"valid_targets_mean": 6427.1,
"valid_targets_min": 2195
},
{
"epoch": 1.4093959731543624,
"grad_norm": 0.49681951586818196,
"learning_rate": 3.8771490996109945e-05,
"loss": 0.3715,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35615652799606323,
"step": 840,
"valid_targets_mean": 5481.1,
"valid_targets_min": 2577
},
{
"epoch": 1.4177852348993287,
"grad_norm": 0.558628488347687,
"learning_rate": 3.8742448488783244e-05,
"loss": 0.3699,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3594701886177063,
"step": 845,
"valid_targets_mean": 5730.9,
"valid_targets_min": 2656
},
{
"epoch": 1.4261744966442953,
"grad_norm": 0.5800936470135034,
"learning_rate": 3.8713077828055476e-05,
"loss": 0.376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4405136704444885,
"step": 850,
"valid_targets_mean": 5378.9,
"valid_targets_min": 2528
},
{
"epoch": 1.4345637583892619,
"grad_norm": 0.6254593304977412,
"learning_rate": 3.868337952816481e-05,
"loss": 0.3672,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3616484999656677,
"step": 855,
"valid_targets_mean": 4310.3,
"valid_targets_min": 2349
},
{
"epoch": 1.4429530201342282,
"grad_norm": 0.5137402379247945,
"learning_rate": 3.865335410908589e-05,
"loss": 0.3669,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3925308883190155,
"step": 860,
"valid_targets_mean": 6220.2,
"valid_targets_min": 2448
},
{
"epoch": 1.4513422818791946,
"grad_norm": 0.5258674268005596,
"learning_rate": 3.8623002096520764e-05,
"loss": 0.3773,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38936707377433777,
"step": 865,
"valid_targets_mean": 5707.8,
"valid_targets_min": 2331
},
{
"epoch": 1.4597315436241611,
"grad_norm": 0.5834875983165838,
"learning_rate": 3.859232402188965e-05,
"loss": 0.3591,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3603193163871765,
"step": 870,
"valid_targets_mean": 5051.4,
"valid_targets_min": 2703
},
{
"epoch": 1.4681208053691275,
"grad_norm": 0.5825621031797557,
"learning_rate": 3.8561320422321636e-05,
"loss": 0.3749,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.378108948469162,
"step": 875,
"valid_targets_mean": 5654.9,
"valid_targets_min": 2931
},
{
"epoch": 1.476510067114094,
"grad_norm": 0.5488901678483015,
"learning_rate": 3.8529991840645324e-05,
"loss": 0.3675,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3602268695831299,
"step": 880,
"valid_targets_mean": 5810.9,
"valid_targets_min": 2138
},
{
"epoch": 1.4848993288590604,
"grad_norm": 0.5219479264561664,
"learning_rate": 3.849833882537924e-05,
"loss": 0.3717,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35310423374176025,
"step": 885,
"valid_targets_mean": 5527.7,
"valid_targets_min": 1927
},
{
"epoch": 1.4932885906040267,
"grad_norm": 0.5496723311386894,
"learning_rate": 3.846636193072232e-05,
"loss": 0.3756,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3754619359970093,
"step": 890,
"valid_targets_mean": 4609.9,
"valid_targets_min": 1738
},
{
"epoch": 1.5016778523489933,
"grad_norm": 0.4791401644411036,
"learning_rate": 3.8434061716544125e-05,
"loss": 0.3751,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3771435618400574,
"step": 895,
"valid_targets_mean": 6249.0,
"valid_targets_min": 2300
},
{
"epoch": 1.5100671140939599,
"grad_norm": 0.4912481560476231,
"learning_rate": 3.840143874837511e-05,
"loss": 0.3594,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.389949232339859,
"step": 900,
"valid_targets_mean": 6485.4,
"valid_targets_min": 3271
},
{
"epoch": 1.5184563758389262,
"grad_norm": 0.5381446544638994,
"learning_rate": 3.836849359739665e-05,
"loss": 0.3569,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.430573046207428,
"step": 905,
"valid_targets_mean": 5474.5,
"valid_targets_min": 2357
},
{
"epoch": 1.5268456375838926,
"grad_norm": 0.5510872012236998,
"learning_rate": 3.833522684043113e-05,
"loss": 0.3688,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38888856768608093,
"step": 910,
"valid_targets_mean": 4406.9,
"valid_targets_min": 2371
},
{
"epoch": 1.535234899328859,
"grad_norm": 0.5351087164429011,
"learning_rate": 3.8301639059931755e-05,
"loss": 0.3634,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3681836724281311,
"step": 915,
"valid_targets_mean": 5347.7,
"valid_targets_min": 1977
},
{
"epoch": 1.5436241610738255,
"grad_norm": 0.658345246126234,
"learning_rate": 3.82677308439724e-05,
"loss": 0.3519,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4065844416618347,
"step": 920,
"valid_targets_mean": 5232.4,
"valid_targets_min": 1318
},
{
"epoch": 1.552013422818792,
"grad_norm": 0.580190492631664,
"learning_rate": 3.823350278623733e-05,
"loss": 0.3484,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35968106985092163,
"step": 925,
"valid_targets_mean": 4733.3,
"valid_targets_min": 2417
},
{
"epoch": 1.5604026845637584,
"grad_norm": 0.48939790257645266,
"learning_rate": 3.819895548601074e-05,
"loss": 0.3695,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3777046799659729,
"step": 930,
"valid_targets_mean": 6044.8,
"valid_targets_min": 3505
},
{
"epoch": 1.5687919463087248,
"grad_norm": 0.5882236315118772,
"learning_rate": 3.816408954816634e-05,
"loss": 0.3723,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41846126317977905,
"step": 935,
"valid_targets_mean": 4460.7,
"valid_targets_min": 2447
},
{
"epoch": 1.5771812080536913,
"grad_norm": 0.5922482030056567,
"learning_rate": 3.8128905583156706e-05,
"loss": 0.3718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38410115242004395,
"step": 940,
"valid_targets_mean": 4022.8,
"valid_targets_min": 2295
},
{
"epoch": 1.5855704697986577,
"grad_norm": 0.488867420498041,
"learning_rate": 3.809340420700262e-05,
"loss": 0.3467,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35235142707824707,
"step": 945,
"valid_targets_mean": 6063.1,
"valid_targets_min": 2559
},
{
"epoch": 1.5939597315436242,
"grad_norm": 0.5915949260916269,
"learning_rate": 3.805758604128228e-05,
"loss": 0.3666,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40172410011291504,
"step": 950,
"valid_targets_mean": 4791.2,
"valid_targets_min": 2263
},
{
"epoch": 1.6023489932885906,
"grad_norm": 0.48300758195294397,
"learning_rate": 3.8021451713120414e-05,
"loss": 0.3707,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36421895027160645,
"step": 955,
"valid_targets_mean": 6213.2,
"valid_targets_min": 1136
},
{
"epoch": 1.610738255033557,
"grad_norm": 0.5504338399140267,
"learning_rate": 3.7985001855177304e-05,
"loss": 0.3763,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3736567795276642,
"step": 960,
"valid_targets_mean": 5184.2,
"valid_targets_min": 2136
},
{
"epoch": 1.6191275167785235,
"grad_norm": 0.4978807777353653,
"learning_rate": 3.794823710563769e-05,
"loss": 0.3657,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33235692977905273,
"step": 965,
"valid_targets_mean": 5770.5,
"valid_targets_min": 2282
},
{
"epoch": 1.62751677852349,
"grad_norm": 0.5144516992188022,
"learning_rate": 3.791115810819964e-05,
"loss": 0.3543,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35421377420425415,
"step": 970,
"valid_targets_mean": 5229.9,
"valid_targets_min": 1734
},
{
"epoch": 1.6359060402684564,
"grad_norm": 0.565262620362482,
"learning_rate": 3.7873765512063225e-05,
"loss": 0.3549,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31595155596733093,
"step": 975,
"valid_targets_mean": 4759.6,
"valid_targets_min": 2816
},
{
"epoch": 1.6442953020134228,
"grad_norm": 0.5396227988448086,
"learning_rate": 3.78360599719192e-05,
"loss": 0.3485,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3592430353164673,
"step": 980,
"valid_targets_mean": 4967.5,
"valid_targets_min": 2596
},
{
"epoch": 1.6526845637583891,
"grad_norm": 0.4996506969758093,
"learning_rate": 3.7798042147937504e-05,
"loss": 0.3571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3265230059623718,
"step": 985,
"valid_targets_mean": 5883.9,
"valid_targets_min": 1924
},
{
"epoch": 1.6610738255033557,
"grad_norm": 0.5141480191166163,
"learning_rate": 3.775971270575574e-05,
"loss": 0.3625,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34632036089897156,
"step": 990,
"valid_targets_mean": 5984.4,
"valid_targets_min": 2377
},
{
"epoch": 1.6694630872483223,
"grad_norm": 0.5306926973941407,
"learning_rate": 3.772107231646748e-05,
"loss": 0.3386,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33942991495132446,
"step": 995,
"valid_targets_mean": 5250.2,
"valid_targets_min": 2070
},
{
"epoch": 1.6778523489932886,
"grad_norm": 0.4996568933812205,
"learning_rate": 3.768212165661053e-05,
"loss": 0.3667,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32338327169418335,
"step": 1000,
"valid_targets_mean": 6141.6,
"valid_targets_min": 2431
},
{
"epoch": 1.686241610738255,
"grad_norm": 0.5850818510185796,
"learning_rate": 3.764286140815511e-05,
"loss": 0.3495,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3676605224609375,
"step": 1005,
"valid_targets_mean": 5132.9,
"valid_targets_min": 3068
},
{
"epoch": 1.6946308724832215,
"grad_norm": 0.6344409516901052,
"learning_rate": 3.760329225849188e-05,
"loss": 0.3488,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38461560010910034,
"step": 1010,
"valid_targets_mean": 3993.8,
"valid_targets_min": 2800
},
{
"epoch": 1.7030201342281879,
"grad_norm": 0.523773448095904,
"learning_rate": 3.7563414900419915e-05,
"loss": 0.3532,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3306989073753357,
"step": 1015,
"valid_targets_mean": 5226.9,
"valid_targets_min": 1976
},
{
"epoch": 1.7114093959731544,
"grad_norm": 0.5421527936391247,
"learning_rate": 3.7523230032134576e-05,
"loss": 0.3745,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38302117586135864,
"step": 1020,
"valid_targets_mean": 5426.8,
"valid_targets_min": 2434
},
{
"epoch": 1.7197986577181208,
"grad_norm": 0.6283186088693279,
"learning_rate": 3.7482738357215305e-05,
"loss": 0.3692,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34678956866264343,
"step": 1025,
"valid_targets_mean": 4664.9,
"valid_targets_min": 2044
},
{
"epoch": 1.7281879194630871,
"grad_norm": 0.5987903080913546,
"learning_rate": 3.744194058461327e-05,
"loss": 0.3619,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40563148260116577,
"step": 1030,
"valid_targets_mean": 4845.0,
"valid_targets_min": 2604
},
{
"epoch": 1.7365771812080537,
"grad_norm": 0.5387213495568653,
"learning_rate": 3.7400837428638974e-05,
"loss": 0.3665,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3455734848976135,
"step": 1035,
"valid_targets_mean": 5003.0,
"valid_targets_min": 2237
},
{
"epoch": 1.7449664429530203,
"grad_norm": 0.5730068728812344,
"learning_rate": 3.735942960894973e-05,
"loss": 0.3105,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2724571228027344,
"step": 1040,
"valid_targets_mean": 5693.6,
"valid_targets_min": 1613
},
{
"epoch": 1.7533557046979866,
"grad_norm": 0.4557401404728535,
"learning_rate": 3.7317717850537126e-05,
"loss": 0.2572,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2845574617385864,
"step": 1045,
"valid_targets_mean": 6052.1,
"valid_targets_min": 3569
},
{
"epoch": 1.761744966442953,
"grad_norm": 0.490941511593462,
"learning_rate": 3.727570288371422e-05,
"loss": 0.2566,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27951961755752563,
"step": 1050,
"valid_targets_mean": 5929.9,
"valid_targets_min": 3249
},
{
"epoch": 1.7701342281879193,
"grad_norm": 0.39912571844584904,
"learning_rate": 3.723338544410285e-05,
"loss": 0.2415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24245810508728027,
"step": 1055,
"valid_targets_mean": 7443.8,
"valid_targets_min": 3135
},
{
"epoch": 1.778523489932886,
"grad_norm": 0.4369536002137865,
"learning_rate": 3.7190766272620715e-05,
"loss": 0.2504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23756088316440582,
"step": 1060,
"valid_targets_mean": 5697.2,
"valid_targets_min": 3549
},
{
"epoch": 1.7869127516778525,
"grad_norm": 0.46317933698906005,
"learning_rate": 3.71478461154684e-05,
"loss": 0.2471,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24878591299057007,
"step": 1065,
"valid_targets_mean": 5807.6,
"valid_targets_min": 3662
},
{
"epoch": 1.7953020134228188,
"grad_norm": 0.43470481316057275,
"learning_rate": 3.710462572411632e-05,
"loss": 0.2485,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27754926681518555,
"step": 1070,
"valid_targets_mean": 6232.0,
"valid_targets_min": 3426
},
{
"epoch": 1.8036912751677852,
"grad_norm": 0.45247043120323754,
"learning_rate": 3.706110585529156e-05,
"loss": 0.2592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2804315984249115,
"step": 1075,
"valid_targets_mean": 6666.6,
"valid_targets_min": 4053
},
{
"epoch": 1.8120805369127517,
"grad_norm": 0.4987278015238315,
"learning_rate": 3.7017287270964634e-05,
"loss": 0.2512,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2920111417770386,
"step": 1080,
"valid_targets_mean": 5846.2,
"valid_targets_min": 3798
},
{
"epoch": 1.820469798657718,
"grad_norm": 0.4216048767738071,
"learning_rate": 3.697317073833611e-05,
"loss": 0.2531,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22008225321769714,
"step": 1085,
"valid_targets_mean": 6118.9,
"valid_targets_min": 3260
},
{
"epoch": 1.8288590604026846,
"grad_norm": 0.4281254602547522,
"learning_rate": 3.6928757029823244e-05,
"loss": 0.2616,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23976412415504456,
"step": 1090,
"valid_targets_mean": 5992.8,
"valid_targets_min": 3387
},
{
"epoch": 1.837248322147651,
"grad_norm": 0.4674475054728659,
"learning_rate": 3.6884046923046376e-05,
"loss": 0.2399,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23339003324508667,
"step": 1095,
"valid_targets_mean": 6847.2,
"valid_targets_min": 3521
},
{
"epoch": 1.8456375838926173,
"grad_norm": 0.4057445322670464,
"learning_rate": 3.683904120081538e-05,
"loss": 0.2459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23576325178146362,
"step": 1100,
"valid_targets_mean": 6886.6,
"valid_targets_min": 3878
},
{
"epoch": 1.854026845637584,
"grad_norm": 0.4681252086760783,
"learning_rate": 3.679374065111593e-05,
"loss": 0.2414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25929751992225647,
"step": 1105,
"valid_targets_mean": 5832.1,
"valid_targets_min": 3646
},
{
"epoch": 1.8624161073825505,
"grad_norm": 0.4335110886632023,
"learning_rate": 3.6748146067095696e-05,
"loss": 0.2592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24445495009422302,
"step": 1110,
"valid_targets_mean": 6038.9,
"valid_targets_min": 1245
},
{
"epoch": 1.8708053691275168,
"grad_norm": 0.48454690692771296,
"learning_rate": 3.670225824705047e-05,
"loss": 0.2408,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2542775869369507,
"step": 1115,
"valid_targets_mean": 6231.9,
"valid_targets_min": 3422
},
{
"epoch": 1.8791946308724832,
"grad_norm": 0.42148519483491925,
"learning_rate": 3.66560779944102e-05,
"loss": 0.2434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2250690758228302,
"step": 1120,
"valid_targets_mean": 6255.6,
"valid_targets_min": 4194
},
{
"epoch": 1.8875838926174495,
"grad_norm": 0.4437674672529351,
"learning_rate": 3.660960611772489e-05,
"loss": 0.2399,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23289281129837036,
"step": 1125,
"valid_targets_mean": 6671.9,
"valid_targets_min": 4183
},
{
"epoch": 1.895973154362416,
"grad_norm": 0.44806539129694345,
"learning_rate": 3.656284343065047e-05,
"loss": 0.2475,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26682284474372864,
"step": 1130,
"valid_targets_mean": 6685.4,
"valid_targets_min": 2159
},
{
"epoch": 1.9043624161073827,
"grad_norm": 0.4491732362005353,
"learning_rate": 3.651579075193455e-05,
"loss": 0.2472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24047374725341797,
"step": 1135,
"valid_targets_mean": 5939.6,
"valid_targets_min": 3334
},
{
"epoch": 1.912751677852349,
"grad_norm": 0.4029334214175894,
"learning_rate": 3.646844890540207e-05,
"loss": 0.2441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2124040126800537,
"step": 1140,
"valid_targets_mean": 6672.1,
"valid_targets_min": 4159
},
{
"epoch": 1.9211409395973154,
"grad_norm": 0.4096211852913776,
"learning_rate": 3.6420818719940896e-05,
"loss": 0.2405,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20634448528289795,
"step": 1145,
"valid_targets_mean": 6304.5,
"valid_targets_min": 3844
},
{
"epoch": 1.929530201342282,
"grad_norm": 0.509142197733445,
"learning_rate": 3.637290102948727e-05,
"loss": 0.2399,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23059163987636566,
"step": 1150,
"valid_targets_mean": 6762.9,
"valid_targets_min": 2156
},
{
"epoch": 1.9379194630872483,
"grad_norm": 0.37918884185680446,
"learning_rate": 3.632469667301126e-05,
"loss": 0.2507,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21625059843063354,
"step": 1155,
"valid_targets_mean": 7273.2,
"valid_targets_min": 5220
},
{
"epoch": 1.9463087248322148,
"grad_norm": 0.4808560827748866,
"learning_rate": 3.627620649450203e-05,
"loss": 0.2459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23380225896835327,
"step": 1160,
"valid_targets_mean": 6628.9,
"valid_targets_min": 3077
},
{
"epoch": 1.9546979865771812,
"grad_norm": 0.45732174037653883,
"learning_rate": 3.6227431342953086e-05,
"loss": 0.2252,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2154151201248169,
"step": 1165,
"valid_targets_mean": 5498.8,
"valid_targets_min": 3451
},
{
"epoch": 1.9630872483221475,
"grad_norm": 0.4967601460003687,
"learning_rate": 3.6178372072347404e-05,
"loss": 0.2459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24176403880119324,
"step": 1170,
"valid_targets_mean": 5527.3,
"valid_targets_min": 3368
},
{
"epoch": 1.971476510067114,
"grad_norm": 0.4407962352710594,
"learning_rate": 3.612902954164247e-05,
"loss": 0.2584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23315542936325073,
"step": 1175,
"valid_targets_mean": 5724.3,
"valid_targets_min": 2579
},
{
"epoch": 1.9798657718120807,
"grad_norm": 0.40896304985315307,
"learning_rate": 3.607940461475524e-05,
"loss": 0.2565,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2468375861644745,
"step": 1180,
"valid_targets_mean": 6724.3,
"valid_targets_min": 3695
},
{
"epoch": 1.988255033557047,
"grad_norm": 0.42187613625785797,
"learning_rate": 3.6029498160547044e-05,
"loss": 0.2578,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2743101716041565,
"step": 1185,
"valid_targets_mean": 6980.3,
"valid_targets_min": 3782
},
{
"epoch": 1.9966442953020134,
"grad_norm": 0.4073947575969795,
"learning_rate": 3.597931105280836e-05,
"loss": 0.2353,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2240162193775177,
"step": 1190,
"valid_targets_mean": 6677.1,
"valid_targets_min": 3670
},
{
"epoch": 2.0050335570469797,
"grad_norm": 0.453140442704113,
"learning_rate": 3.592884417024346e-05,
"loss": 0.3111,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3438984751701355,
"step": 1195,
"valid_targets_mean": 7762.6,
"valid_targets_min": 4256
},
{
"epoch": 2.0134228187919465,
"grad_norm": 0.45341280143818996,
"learning_rate": 3.5878098396455105e-05,
"loss": 0.3291,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28538778424263,
"step": 1200,
"valid_targets_mean": 6290.4,
"valid_targets_min": 2571
},
{
"epoch": 2.021812080536913,
"grad_norm": 0.4581633641665778,
"learning_rate": 3.582707461992906e-05,
"loss": 0.3412,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32423585653305054,
"step": 1205,
"valid_targets_mean": 6965.4,
"valid_targets_min": 2769
},
{
"epoch": 2.030201342281879,
"grad_norm": 0.46660712344014627,
"learning_rate": 3.577577373401847e-05,
"loss": 0.3353,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.368762731552124,
"step": 1210,
"valid_targets_mean": 6936.1,
"valid_targets_min": 2974
},
{
"epoch": 2.0385906040268456,
"grad_norm": 0.46814653120085054,
"learning_rate": 3.572419663692831e-05,
"loss": 0.3266,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32211530208587646,
"step": 1215,
"valid_targets_mean": 6466.9,
"valid_targets_min": 1417
},
{
"epoch": 2.046979865771812,
"grad_norm": 0.39343469842039014,
"learning_rate": 3.5672344231699596e-05,
"loss": 0.3274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30910724401474,
"step": 1220,
"valid_targets_mean": 8162.1,
"valid_targets_min": 3446
},
{
"epoch": 2.0553691275167787,
"grad_norm": 0.46827202121954786,
"learning_rate": 3.56202174261936e-05,
"loss": 0.3245,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32270243763923645,
"step": 1225,
"valid_targets_mean": 6284.0,
"valid_targets_min": 2526
},
{
"epoch": 2.063758389261745,
"grad_norm": 0.4423032808271841,
"learning_rate": 3.556781713307595e-05,
"loss": 0.3431,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33364373445510864,
"step": 1230,
"valid_targets_mean": 7554.5,
"valid_targets_min": 3408
},
{
"epoch": 2.0721476510067114,
"grad_norm": 0.4872083515401277,
"learning_rate": 3.5515144269800656e-05,
"loss": 0.3179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2982063889503479,
"step": 1235,
"valid_targets_mean": 6332.1,
"valid_targets_min": 3229
},
{
"epoch": 2.0805369127516777,
"grad_norm": 0.4156249500175109,
"learning_rate": 3.546219975859403e-05,
"loss": 0.319,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32445910573005676,
"step": 1240,
"valid_targets_mean": 7520.4,
"valid_targets_min": 3432
},
{
"epoch": 2.088926174496644,
"grad_norm": 0.4189383774068392,
"learning_rate": 3.5408984526438555e-05,
"loss": 0.3328,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32531315088272095,
"step": 1245,
"valid_targets_mean": 7121.8,
"valid_targets_min": 4887
},
{
"epoch": 2.097315436241611,
"grad_norm": 0.4473660009234783,
"learning_rate": 3.535549950505664e-05,
"loss": 0.3266,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39161813259124756,
"step": 1250,
"valid_targets_mean": 7733.2,
"valid_targets_min": 3272
},
{
"epoch": 2.1057046979865772,
"grad_norm": 0.4481249663257016,
"learning_rate": 3.5301745630894335e-05,
"loss": 0.322,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31673017144203186,
"step": 1255,
"valid_targets_mean": 6697.4,
"valid_targets_min": 3148
},
{
"epoch": 2.1140939597315436,
"grad_norm": 0.4625274988899184,
"learning_rate": 3.52477238451049e-05,
"loss": 0.3407,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3583962619304657,
"step": 1260,
"valid_targets_mean": 7746.1,
"valid_targets_min": 2988
},
{
"epoch": 2.12248322147651,
"grad_norm": 0.5172213228460174,
"learning_rate": 3.519343509353236e-05,
"loss": 0.3154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3001815378665924,
"step": 1265,
"valid_targets_mean": 7451.6,
"valid_targets_min": 2703
},
{
"epoch": 2.1308724832214767,
"grad_norm": 0.40316022275127994,
"learning_rate": 3.513888032669491e-05,
"loss": 0.3018,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32507652044296265,
"step": 1270,
"valid_targets_mean": 8174.7,
"valid_targets_min": 3910
},
{
"epoch": 2.139261744966443,
"grad_norm": 0.4768029209801054,
"learning_rate": 3.5084060499768306e-05,
"loss": 0.3377,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4111858010292053,
"step": 1275,
"valid_targets_mean": 7502.3,
"valid_targets_min": 2392
},
{
"epoch": 2.1476510067114094,
"grad_norm": 0.45513706136100934,
"learning_rate": 3.5028976572569106e-05,
"loss": 0.2927,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3249369263648987,
"step": 1280,
"valid_targets_mean": 7557.0,
"valid_targets_min": 3073
},
{
"epoch": 2.1560402684563758,
"grad_norm": 0.4145970961311004,
"learning_rate": 3.497362950953792e-05,
"loss": 0.3246,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30537134408950806,
"step": 1285,
"valid_targets_mean": 7468.0,
"valid_targets_min": 4054
},
{
"epoch": 2.164429530201342,
"grad_norm": 0.4150633170155709,
"learning_rate": 3.4918020279722454e-05,
"loss": 0.3009,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2901865839958191,
"step": 1290,
"valid_targets_mean": 7112.4,
"valid_targets_min": 2080
},
{
"epoch": 2.172818791946309,
"grad_norm": 0.49272676592663855,
"learning_rate": 3.486214985676059e-05,
"loss": 0.3191,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36454761028289795,
"step": 1295,
"valid_targets_mean": 7546.1,
"valid_targets_min": 3564
},
{
"epoch": 2.1812080536912752,
"grad_norm": 0.46374231405417815,
"learning_rate": 3.4806019218863336e-05,
"loss": 0.3126,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27287721633911133,
"step": 1300,
"valid_targets_mean": 7521.9,
"valid_targets_min": 3040
},
{
"epoch": 2.1895973154362416,
"grad_norm": 0.4597971994199139,
"learning_rate": 3.474962934879769e-05,
"loss": 0.3178,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3161107301712036,
"step": 1305,
"valid_targets_mean": 6784.7,
"valid_targets_min": 2822
},
{
"epoch": 2.197986577181208,
"grad_norm": 0.46020007013121705,
"learning_rate": 3.469298123386942e-05,
"loss": 0.3037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3112471401691437,
"step": 1310,
"valid_targets_mean": 6906.2,
"valid_targets_min": 2470
},
{
"epoch": 2.2063758389261743,
"grad_norm": 0.492958857992654,
"learning_rate": 3.4636075865905794e-05,
"loss": 0.2968,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30589598417282104,
"step": 1315,
"valid_targets_mean": 7868.9,
"valid_targets_min": 5007
},
{
"epoch": 2.214765100671141,
"grad_norm": 0.5710433872243388,
"learning_rate": 3.457891424123823e-05,
"loss": 0.3759,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38415563106536865,
"step": 1320,
"valid_targets_mean": 5229.1,
"valid_targets_min": 2397
},
{
"epoch": 2.2231543624161074,
"grad_norm": 0.7935791016843832,
"learning_rate": 3.4521497360684806e-05,
"loss": 0.4225,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.43013817071914673,
"step": 1325,
"valid_targets_mean": 3945.6,
"valid_targets_min": 2194
},
{
"epoch": 2.2315436241610738,
"grad_norm": 0.6669523753063487,
"learning_rate": 3.4463826229532794e-05,
"loss": 0.4129,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3834550976753235,
"step": 1330,
"valid_targets_mean": 4037.4,
"valid_targets_min": 1497
},
{
"epoch": 2.23993288590604,
"grad_norm": 0.7716561296284383,
"learning_rate": 3.4405901857521e-05,
"loss": 0.4248,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.414074182510376,
"step": 1335,
"valid_targets_mean": 5008.7,
"valid_targets_min": 2932
},
{
"epoch": 2.248322147651007,
"grad_norm": 0.5506109732136797,
"learning_rate": 3.4347725258822143e-05,
"loss": 0.4115,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4239192008972168,
"step": 1340,
"valid_targets_mean": 5910.4,
"valid_targets_min": 2367
},
{
"epoch": 2.2567114093959733,
"grad_norm": 0.5982912817837929,
"learning_rate": 3.4289297452025044e-05,
"loss": 0.3964,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4122527837753296,
"step": 1345,
"valid_targets_mean": 4245.2,
"valid_targets_min": 2401
},
{
"epoch": 2.2651006711409396,
"grad_norm": 0.5810879635226942,
"learning_rate": 3.423061946011684e-05,
"loss": 0.4081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37277913093566895,
"step": 1350,
"valid_targets_mean": 4953.4,
"valid_targets_min": 2166
},
{
"epoch": 2.273489932885906,
"grad_norm": 0.6082515302140243,
"learning_rate": 3.417169231046503e-05,
"loss": 0.4055,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.384243369102478,
"step": 1355,
"valid_targets_mean": 4656.1,
"valid_targets_min": 2314
},
{
"epoch": 2.2818791946308723,
"grad_norm": 0.7161321577324711,
"learning_rate": 3.411251703479952e-05,
"loss": 0.4028,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39829784631729126,
"step": 1360,
"valid_targets_mean": 4454.9,
"valid_targets_min": 1801
},
{
"epoch": 2.290268456375839,
"grad_norm": 0.6618368360922882,
"learning_rate": 3.405309466919454e-05,
"loss": 0.391,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3945760726928711,
"step": 1365,
"valid_targets_mean": 3714.0,
"valid_targets_min": 1819
},
{
"epoch": 2.2986577181208054,
"grad_norm": 0.6374297039673902,
"learning_rate": 3.3993426254050534e-05,
"loss": 0.3974,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4257645308971405,
"step": 1370,
"valid_targets_mean": 4132.1,
"valid_targets_min": 2114
},
{
"epoch": 2.307046979865772,
"grad_norm": 0.6145397461152713,
"learning_rate": 3.393351283407588e-05,
"loss": 0.4084,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36819249391555786,
"step": 1375,
"valid_targets_mean": 4337.7,
"valid_targets_min": 2174
},
{
"epoch": 2.315436241610738,
"grad_norm": 0.5809229241749594,
"learning_rate": 3.387335545826867e-05,
"loss": 0.3879,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.356969952583313,
"step": 1380,
"valid_targets_mean": 4287.4,
"valid_targets_min": 2148
},
{
"epoch": 2.323825503355705,
"grad_norm": 0.5675018196487998,
"learning_rate": 3.3812955179898305e-05,
"loss": 0.3883,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36231207847595215,
"step": 1385,
"valid_targets_mean": 5713.9,
"valid_targets_min": 1871
},
{
"epoch": 2.3322147651006713,
"grad_norm": 0.546838535455084,
"learning_rate": 3.3752313056487056e-05,
"loss": 0.3493,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3755466341972351,
"step": 1390,
"valid_targets_mean": 5019.2,
"valid_targets_min": 2007
},
{
"epoch": 2.3406040268456376,
"grad_norm": 0.535635390940889,
"learning_rate": 3.369143014979156e-05,
"loss": 0.3619,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3481118381023407,
"step": 1395,
"valid_targets_mean": 5106.4,
"valid_targets_min": 1711
},
{
"epoch": 2.348993288590604,
"grad_norm": 0.5452383044456509,
"learning_rate": 3.3630307525784224e-05,
"loss": 0.332,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3285505771636963,
"step": 1400,
"valid_targets_mean": 4400.2,
"valid_targets_min": 2244
},
{
"epoch": 2.3573825503355703,
"grad_norm": 0.6285622660379921,
"learning_rate": 3.356894625463456e-05,
"loss": 0.3454,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3412173092365265,
"step": 1405,
"valid_targets_mean": 4533.4,
"valid_targets_min": 1844
},
{
"epoch": 2.365771812080537,
"grad_norm": 0.5812892813114495,
"learning_rate": 3.350734741069045e-05,
"loss": 0.3309,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3514842987060547,
"step": 1410,
"valid_targets_mean": 4403.6,
"valid_targets_min": 1981
},
{
"epoch": 2.3741610738255035,
"grad_norm": 0.5495705106539454,
"learning_rate": 3.344551207245933e-05,
"loss": 0.3282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31175684928894043,
"step": 1415,
"valid_targets_mean": 5373.3,
"valid_targets_min": 2135
},
{
"epoch": 2.38255033557047,
"grad_norm": 0.4990043682913641,
"learning_rate": 3.3383441322589315e-05,
"loss": 0.3144,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.308898001909256,
"step": 1420,
"valid_targets_mean": 5323.1,
"valid_targets_min": 2232
},
{
"epoch": 2.390939597315436,
"grad_norm": 0.6302503170489214,
"learning_rate": 3.332113624785025e-05,
"loss": 0.3319,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3533690571784973,
"step": 1425,
"valid_targets_mean": 5083.4,
"valid_targets_min": 2802
},
{
"epoch": 2.3993288590604025,
"grad_norm": 0.9366298498480529,
"learning_rate": 3.325859793911465e-05,
"loss": 0.3571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3332177400588989,
"step": 1430,
"valid_targets_mean": 5944.1,
"valid_targets_min": 2963
},
{
"epoch": 2.4077181208053693,
"grad_norm": 0.5715778498762676,
"learning_rate": 3.319582749133866e-05,
"loss": 0.3332,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3450821042060852,
"step": 1435,
"valid_targets_mean": 5382.9,
"valid_targets_min": 2507
},
{
"epoch": 2.4161073825503356,
"grad_norm": 0.5820358441715864,
"learning_rate": 3.313282600354279e-05,
"loss": 0.3318,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38198742270469666,
"step": 1440,
"valid_targets_mean": 4676.9,
"valid_targets_min": 2346
},
{
"epoch": 2.424496644295302,
"grad_norm": 0.5381040915088271,
"learning_rate": 3.3069594578792795e-05,
"loss": 0.3247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32292282581329346,
"step": 1445,
"valid_targets_mean": 4919.2,
"valid_targets_min": 2405
},
{
"epoch": 2.4328859060402683,
"grad_norm": 0.5464755127598416,
"learning_rate": 3.3006134324180246e-05,
"loss": 0.3453,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34378236532211304,
"step": 1450,
"valid_targets_mean": 4644.6,
"valid_targets_min": 2987
},
{
"epoch": 2.4412751677852347,
"grad_norm": 0.5778247505646406,
"learning_rate": 3.294244635080322e-05,
"loss": 0.323,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3166796565055847,
"step": 1455,
"valid_targets_mean": 4582.3,
"valid_targets_min": 2019
},
{
"epoch": 2.4496644295302015,
"grad_norm": 0.5438572036946208,
"learning_rate": 3.2878531773746815e-05,
"loss": 0.343,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32977181673049927,
"step": 1460,
"valid_targets_mean": 5231.9,
"valid_targets_min": 2798
},
{
"epoch": 2.458053691275168,
"grad_norm": 0.5481029808369472,
"learning_rate": 3.2814391712063645e-05,
"loss": 0.3293,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2937307059764862,
"step": 1465,
"valid_targets_mean": 5156.6,
"valid_targets_min": 2080
},
{
"epoch": 2.466442953020134,
"grad_norm": 0.5334750424683744,
"learning_rate": 3.275002728875423e-05,
"loss": 0.3354,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33928877115249634,
"step": 1470,
"valid_targets_mean": 4651.2,
"valid_targets_min": 2329
},
{
"epoch": 2.4748322147651005,
"grad_norm": 0.5873695229900598,
"learning_rate": 3.268543963074732e-05,
"loss": 0.3363,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35795867443084717,
"step": 1475,
"valid_targets_mean": 4338.4,
"valid_targets_min": 2375
},
{
"epoch": 2.4832214765100673,
"grad_norm": 0.7139681301099575,
"learning_rate": 3.2620629868880234e-05,
"loss": 0.3379,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33694660663604736,
"step": 1480,
"valid_targets_mean": 5411.5,
"valid_targets_min": 2320
},
{
"epoch": 2.4916107382550337,
"grad_norm": 0.5186540713707555,
"learning_rate": 3.255559913787896e-05,
"loss": 0.3374,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3617810606956482,
"step": 1485,
"valid_targets_mean": 6352.4,
"valid_targets_min": 2255
},
{
"epoch": 2.5,
"grad_norm": 0.5811315874862599,
"learning_rate": 3.249034857633837e-05,
"loss": 0.3388,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39030152559280396,
"step": 1490,
"valid_targets_mean": 4852.2,
"valid_targets_min": 2408
},
{
"epoch": 2.5083892617449663,
"grad_norm": 0.5355377816240061,
"learning_rate": 3.242487932670224e-05,
"loss": 0.3225,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3067181706428528,
"step": 1495,
"valid_targets_mean": 6067.6,
"valid_targets_min": 3288
},
{
"epoch": 2.5167785234899327,
"grad_norm": 0.5033326491273779,
"learning_rate": 3.2359192535243255e-05,
"loss": 0.3157,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27847611904144287,
"step": 1500,
"valid_targets_mean": 5254.0,
"valid_targets_min": 1880
},
{
"epoch": 2.5251677852348995,
"grad_norm": 0.567141960655738,
"learning_rate": 3.2293289352042974e-05,
"loss": 0.3438,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33290085196495056,
"step": 1505,
"valid_targets_mean": 5129.2,
"valid_targets_min": 2704
},
{
"epoch": 2.533557046979866,
"grad_norm": 0.537752086775137,
"learning_rate": 3.2227170930971635e-05,
"loss": 0.3336,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3089789152145386,
"step": 1510,
"valid_targets_mean": 5251.9,
"valid_targets_min": 2324
},
{
"epoch": 2.541946308724832,
"grad_norm": 0.46371799158922394,
"learning_rate": 3.216083842966799e-05,
"loss": 0.3116,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28042498230934143,
"step": 1515,
"valid_targets_mean": 5800.2,
"valid_targets_min": 2357
},
{
"epoch": 2.5503355704697985,
"grad_norm": 0.49906778637915816,
"learning_rate": 3.209429300951905e-05,
"loss": 0.3257,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3296702802181244,
"step": 1520,
"valid_targets_mean": 5885.9,
"valid_targets_min": 2532
},
{
"epoch": 2.5587248322147653,
"grad_norm": 0.5901564252008695,
"learning_rate": 3.202753583563969e-05,
"loss": 0.3338,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3533482551574707,
"step": 1525,
"valid_targets_mean": 4611.0,
"valid_targets_min": 2247
},
{
"epoch": 2.5671140939597317,
"grad_norm": 0.6049425090356672,
"learning_rate": 3.196056807685233e-05,
"loss": 0.331,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34530556201934814,
"step": 1530,
"valid_targets_mean": 4014.8,
"valid_targets_min": 1920
},
{
"epoch": 2.575503355704698,
"grad_norm": 0.5463985915202139,
"learning_rate": 3.189339090566639e-05,
"loss": 0.344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37530845403671265,
"step": 1535,
"valid_targets_mean": 5276.6,
"valid_targets_min": 2612
},
{
"epoch": 2.5838926174496644,
"grad_norm": 0.553966804293357,
"learning_rate": 3.182600549825785e-05,
"loss": 0.3181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32243046164512634,
"step": 1540,
"valid_targets_mean": 4850.3,
"valid_targets_min": 2301
},
{
"epoch": 2.5922818791946307,
"grad_norm": 0.5931536075290793,
"learning_rate": 3.175841303444856e-05,
"loss": 0.3245,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3262476623058319,
"step": 1545,
"valid_targets_mean": 5194.4,
"valid_targets_min": 2371
},
{
"epoch": 2.600671140939597,
"grad_norm": 0.5344464504713267,
"learning_rate": 3.169061469768566e-05,
"loss": 0.3455,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3294913172721863,
"step": 1550,
"valid_targets_mean": 5431.9,
"valid_targets_min": 2159
},
{
"epoch": 2.609060402684564,
"grad_norm": 0.5735165104061004,
"learning_rate": 3.162261167502081e-05,
"loss": 0.3414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3188868761062622,
"step": 1555,
"valid_targets_mean": 4606.6,
"valid_targets_min": 2096
},
{
"epoch": 2.61744966442953,
"grad_norm": 0.510957083138931,
"learning_rate": 3.155440515708945e-05,
"loss": 0.3415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3340016305446625,
"step": 1560,
"valid_targets_mean": 5406.6,
"valid_targets_min": 2219
},
{
"epoch": 2.6258389261744965,
"grad_norm": 0.5105944754114194,
"learning_rate": 3.1485996338089906e-05,
"loss": 0.3174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2784973382949829,
"step": 1565,
"valid_targets_mean": 5233.4,
"valid_targets_min": 1930
},
{
"epoch": 2.6342281879194633,
"grad_norm": 0.5701812279281744,
"learning_rate": 3.141738641576254e-05,
"loss": 0.3292,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3298957943916321,
"step": 1570,
"valid_targets_mean": 5110.7,
"valid_targets_min": 1339
},
{
"epoch": 2.6426174496644297,
"grad_norm": 0.5373578069205611,
"learning_rate": 3.134857659136869e-05,
"loss": 0.3093,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.317375123500824,
"step": 1575,
"valid_targets_mean": 5667.8,
"valid_targets_min": 2598
},
{
"epoch": 2.651006711409396,
"grad_norm": 0.5289643876533299,
"learning_rate": 3.1279568069669744e-05,
"loss": 0.3318,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3362849950790405,
"step": 1580,
"valid_targets_mean": 4838.7,
"valid_targets_min": 2379
},
{
"epoch": 2.6593959731543624,
"grad_norm": 0.5301873536727923,
"learning_rate": 3.121036205890598e-05,
"loss": 0.3289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3280187249183655,
"step": 1585,
"valid_targets_mean": 5387.4,
"valid_targets_min": 2435
},
{
"epoch": 2.6677852348993287,
"grad_norm": 0.4773658826796727,
"learning_rate": 3.114095977077542e-05,
"loss": 0.3117,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33107709884643555,
"step": 1590,
"valid_targets_mean": 6580.9,
"valid_targets_min": 3096
},
{
"epoch": 2.676174496644295,
"grad_norm": 0.5127675328178021,
"learning_rate": 3.1071362420412604e-05,
"loss": 0.3395,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28227123618125916,
"step": 1595,
"valid_targets_mean": 5331.3,
"valid_targets_min": 2607
},
{
"epoch": 2.684563758389262,
"grad_norm": 0.6809931288855824,
"learning_rate": 3.100157122636738e-05,
"loss": 0.3117,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31957268714904785,
"step": 1600,
"valid_targets_mean": 4802.6,
"valid_targets_min": 1843
},
{
"epoch": 2.692953020134228,
"grad_norm": 0.5468454656103353,
"learning_rate": 3.093158741058348e-05,
"loss": 0.3158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3136354088783264,
"step": 1605,
"valid_targets_mean": 5199.5,
"valid_targets_min": 2283
},
{
"epoch": 2.7013422818791946,
"grad_norm": 0.5878026558771516,
"learning_rate": 3.086141219837718e-05,
"loss": 0.3332,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35434913635253906,
"step": 1610,
"valid_targets_mean": 4519.5,
"valid_targets_min": 2306
},
{
"epoch": 2.709731543624161,
"grad_norm": 0.5567037217282692,
"learning_rate": 3.079104681841583e-05,
"loss": 0.3334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3579016923904419,
"step": 1615,
"valid_targets_mean": 5361.8,
"valid_targets_min": 2040
},
{
"epoch": 2.7181208053691277,
"grad_norm": 0.5318120910443703,
"learning_rate": 3.072049250269634e-05,
"loss": 0.3477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32555705308914185,
"step": 1620,
"valid_targets_mean": 6091.1,
"valid_targets_min": 2118
},
{
"epoch": 2.726510067114094,
"grad_norm": 0.564359774938993,
"learning_rate": 3.064975048652364e-05,
"loss": 0.3223,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3219876289367676,
"step": 1625,
"valid_targets_mean": 4456.3,
"valid_targets_min": 2053
},
{
"epoch": 2.7348993288590604,
"grad_norm": 0.5741402906706731,
"learning_rate": 3.0578822008488984e-05,
"loss": 0.3475,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36692801117897034,
"step": 1630,
"valid_targets_mean": 4881.3,
"valid_targets_min": 1094
},
{
"epoch": 2.7432885906040267,
"grad_norm": 0.47687725725615143,
"learning_rate": 3.050770831044835e-05,
"loss": 0.2969,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21516865491867065,
"step": 1635,
"valid_targets_mean": 7170.1,
"valid_targets_min": 4264
},
{
"epoch": 2.751677852348993,
"grad_norm": 0.4546711410215575,
"learning_rate": 3.0436410637500624e-05,
"loss": 0.2285,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23095494508743286,
"step": 1640,
"valid_targets_mean": 6591.3,
"valid_targets_min": 3146
},
{
"epoch": 2.76006711409396,
"grad_norm": 0.47189522703744397,
"learning_rate": 3.036493023796583e-05,
"loss": 0.2312,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23911446332931519,
"step": 1645,
"valid_targets_mean": 5952.6,
"valid_targets_min": 3344
},
{
"epoch": 2.7684563758389262,
"grad_norm": 0.4541596865622402,
"learning_rate": 3.0293268363363276e-05,
"loss": 0.2228,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21723681688308716,
"step": 1650,
"valid_targets_mean": 6554.4,
"valid_targets_min": 3676
},
{
"epoch": 2.7768456375838926,
"grad_norm": 0.46028976426643436,
"learning_rate": 3.0221426268389663e-05,
"loss": 0.2258,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2234461009502411,
"step": 1655,
"valid_targets_mean": 5764.8,
"valid_targets_min": 3725
},
{
"epoch": 2.785234899328859,
"grad_norm": 0.4678718404389343,
"learning_rate": 3.014940521089705e-05,
"loss": 0.2198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21628400683403015,
"step": 1660,
"valid_targets_mean": 6907.3,
"valid_targets_min": 4340
},
{
"epoch": 2.7936241610738257,
"grad_norm": 0.4498709718134225,
"learning_rate": 3.00772064518709e-05,
"loss": 0.2171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1899103969335556,
"step": 1665,
"valid_targets_mean": 5534.2,
"valid_targets_min": 2727
},
{
"epoch": 2.802013422818792,
"grad_norm": 0.4418082266156402,
"learning_rate": 3.000483125540796e-05,
"loss": 0.232,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2425849437713623,
"step": 1670,
"valid_targets_mean": 5948.1,
"valid_targets_min": 3544
},
{
"epoch": 2.8104026845637584,
"grad_norm": 0.44388047405272696,
"learning_rate": 2.9932280888694168e-05,
"loss": 0.2255,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23246996104717255,
"step": 1675,
"valid_targets_mean": 6550.1,
"valid_targets_min": 3124
},
{
"epoch": 2.8187919463087248,
"grad_norm": 0.4442756641577408,
"learning_rate": 2.985955662198242e-05,
"loss": 0.2429,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28180423378944397,
"step": 1680,
"valid_targets_mean": 6668.4,
"valid_targets_min": 3364
},
{
"epoch": 2.827181208053691,
"grad_norm": 0.45460177829337356,
"learning_rate": 2.978665972857035e-05,
"loss": 0.2336,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2684323787689209,
"step": 1685,
"valid_targets_mean": 6537.2,
"valid_targets_min": 2860
},
{
"epoch": 2.8355704697986575,
"grad_norm": 0.3957006959045116,
"learning_rate": 2.9713591484778055e-05,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19664597511291504,
"step": 1690,
"valid_targets_mean": 6501.0,
"valid_targets_min": 3368
},
{
"epoch": 2.8439597315436242,
"grad_norm": 0.4254197395693919,
"learning_rate": 2.9640353169925715e-05,
"loss": 0.221,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17788392305374146,
"step": 1695,
"valid_targets_mean": 6252.9,
"valid_targets_min": 2384
},
{
"epoch": 2.8523489932885906,
"grad_norm": 0.48175756674110215,
"learning_rate": 2.956694606631123e-05,
"loss": 0.2137,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23316311836242676,
"step": 1700,
"valid_targets_mean": 7136.6,
"valid_targets_min": 3722
},
{
"epoch": 2.860738255033557,
"grad_norm": 0.4777776571998629,
"learning_rate": 2.9493371459187723e-05,
"loss": 0.2377,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2757401764392853,
"step": 1705,
"valid_targets_mean": 6547.6,
"valid_targets_min": 3433
},
{
"epoch": 2.8691275167785237,
"grad_norm": 0.42669637388101245,
"learning_rate": 2.9419630636741088e-05,
"loss": 0.2147,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19498570263385773,
"step": 1710,
"valid_targets_mean": 7310.3,
"valid_targets_min": 3731
},
{
"epoch": 2.87751677852349,
"grad_norm": 0.4693843718327286,
"learning_rate": 2.9345724890067386e-05,
"loss": 0.2254,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25868386030197144,
"step": 1715,
"valid_targets_mean": 6397.4,
"valid_targets_min": 3861
},
{
"epoch": 2.8859060402684564,
"grad_norm": 0.45185240695761386,
"learning_rate": 2.9271655513150293e-05,
"loss": 0.2166,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2129001021385193,
"step": 1720,
"valid_targets_mean": 6480.8,
"valid_targets_min": 2753
},
{
"epoch": 2.8942953020134228,
"grad_norm": 0.5352926202412781,
"learning_rate": 2.9197423802838396e-05,
"loss": 0.2182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20774738490581512,
"step": 1725,
"valid_targets_mean": 5949.8,
"valid_targets_min": 2908
},
{
"epoch": 2.902684563758389,
"grad_norm": 0.4729386598413279,
"learning_rate": 2.912303105882251e-05,
"loss": 0.2289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23421156406402588,
"step": 1730,
"valid_targets_mean": 6394.9,
"valid_targets_min": 3994
},
{
"epoch": 2.9110738255033555,
"grad_norm": 0.43333103318513644,
"learning_rate": 2.9048478583612932e-05,
"loss": 0.2271,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2326471507549286,
"step": 1735,
"valid_targets_mean": 6311.9,
"valid_targets_min": 2320
},
{
"epoch": 2.9194630872483223,
"grad_norm": 0.4405427141855079,
"learning_rate": 2.897376768251661e-05,
"loss": 0.2181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24898266792297363,
"step": 1740,
"valid_targets_mean": 6375.8,
"valid_targets_min": 4649
},
{
"epoch": 2.9278523489932886,
"grad_norm": 0.502641565407119,
"learning_rate": 2.8898899663614294e-05,
"loss": 0.2132,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21118290722370148,
"step": 1745,
"valid_targets_mean": 5890.0,
"valid_targets_min": 3165
},
{
"epoch": 2.936241610738255,
"grad_norm": 0.5654771704862958,
"learning_rate": 2.8823875837737664e-05,
"loss": 0.2325,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27783122658729553,
"step": 1750,
"valid_targets_mean": 6355.1,
"valid_targets_min": 3117
},
{
"epoch": 2.9446308724832218,
"grad_norm": 0.45101362296422326,
"learning_rate": 2.8748697518446343e-05,
"loss": 0.2208,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21551084518432617,
"step": 1755,
"valid_targets_mean": 6151.1,
"valid_targets_min": 3470
},
{
"epoch": 2.953020134228188,
"grad_norm": 0.4493212317585719,
"learning_rate": 2.867336602200492e-05,
"loss": 0.207,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1713525801897049,
"step": 1760,
"valid_targets_mean": 5871.4,
"valid_targets_min": 3001
},
{
"epoch": 2.9614093959731544,
"grad_norm": 0.4441136022460391,
"learning_rate": 2.8597882667359878e-05,
"loss": 0.2188,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18627864122390747,
"step": 1765,
"valid_targets_mean": 6023.6,
"valid_targets_min": 3882
},
{
"epoch": 2.969798657718121,
"grad_norm": 0.45978694258850117,
"learning_rate": 2.8522248776116544e-05,
"loss": 0.2366,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23736673593521118,
"step": 1770,
"valid_targets_mean": 6363.8,
"valid_targets_min": 4007
},
{
"epoch": 2.978187919463087,
"grad_norm": 0.5702066521743531,
"learning_rate": 2.8446465672515917e-05,
"loss": 0.2311,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.303191602230072,
"step": 1775,
"valid_targets_mean": 7757.6,
"valid_targets_min": 3037
},
{
"epoch": 2.9865771812080535,
"grad_norm": 0.48418902835942623,
"learning_rate": 2.8370534683411476e-05,
"loss": 0.231,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2507019639015198,
"step": 1780,
"valid_targets_mean": 6457.0,
"valid_targets_min": 3503
},
{
"epoch": 2.9949664429530203,
"grad_norm": 0.521598833215232,
"learning_rate": 2.829445713824598e-05,
"loss": 0.224,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1831042319536209,
"step": 1785,
"valid_targets_mean": 6649.4,
"valid_targets_min": 3589
},
{
"epoch": 3.0033557046979866,
"grad_norm": 0.5240397225363642,
"learning_rate": 2.8218234369028165e-05,
"loss": 0.2616,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3380255103111267,
"step": 1790,
"valid_targets_mean": 7485.6,
"valid_targets_min": 3105
},
{
"epoch": 3.011744966442953,
"grad_norm": 0.4221493227056485,
"learning_rate": 2.8141867710309435e-05,
"loss": 0.3159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2678639888763428,
"step": 1795,
"valid_targets_mean": 7863.2,
"valid_targets_min": 2267
},
{
"epoch": 3.0201342281879193,
"grad_norm": 0.42544867122179214,
"learning_rate": 2.8065358499160503e-05,
"loss": 0.3084,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3196631669998169,
"step": 1800,
"valid_targets_mean": 7926.6,
"valid_targets_min": 2905
},
{
"epoch": 3.028523489932886,
"grad_norm": 0.4701487483809963,
"learning_rate": 2.7988708075147954e-05,
"loss": 0.3023,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3164515197277069,
"step": 1805,
"valid_targets_mean": 8084.8,
"valid_targets_min": 2815
},
{
"epoch": 3.0369127516778525,
"grad_norm": 0.4650218396536915,
"learning_rate": 2.791191778031081e-05,
"loss": 0.3107,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3165399432182312,
"step": 1810,
"valid_targets_mean": 7303.1,
"valid_targets_min": 4407
},
{
"epoch": 3.045302013422819,
"grad_norm": 0.4955122809928877,
"learning_rate": 2.783498895913705e-05,
"loss": 0.3074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32211869955062866,
"step": 1815,
"valid_targets_mean": 8275.1,
"valid_targets_min": 4626
},
{
"epoch": 3.053691275167785,
"grad_norm": 0.4561818700403147,
"learning_rate": 2.775792295854002e-05,
"loss": 0.2989,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27344971895217896,
"step": 1820,
"valid_targets_mean": 7222.4,
"valid_targets_min": 3425
},
{
"epoch": 3.0620805369127515,
"grad_norm": 0.4704791947585593,
"learning_rate": 2.7680721127834912e-05,
"loss": 0.317,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37310269474983215,
"step": 1825,
"valid_targets_mean": 6871.7,
"valid_targets_min": 3063
},
{
"epoch": 3.0704697986577183,
"grad_norm": 0.49148705636382106,
"learning_rate": 2.760338481871508e-05,
"loss": 0.3021,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29893338680267334,
"step": 1830,
"valid_targets_mean": 6952.8,
"valid_targets_min": 3356
},
{
"epoch": 3.0788590604026846,
"grad_norm": 0.455318197558492,
"learning_rate": 2.7525915385228424e-05,
"loss": 0.2909,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3230130076408386,
"step": 1835,
"valid_targets_mean": 7839.2,
"valid_targets_min": 3062
},
{
"epoch": 3.087248322147651,
"grad_norm": 0.4383606924135849,
"learning_rate": 2.744831418375365e-05,
"loss": 0.3101,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2872515022754669,
"step": 1840,
"valid_targets_mean": 7526.3,
"valid_targets_min": 3746
},
{
"epoch": 3.0956375838926173,
"grad_norm": 0.4254691657270492,
"learning_rate": 2.7370582572976538e-05,
"loss": 0.2914,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27996861934661865,
"step": 1845,
"valid_targets_mean": 7435.4,
"valid_targets_min": 3537
},
{
"epoch": 3.1040268456375837,
"grad_norm": 0.45519104952926065,
"learning_rate": 2.7292721913866142e-05,
"loss": 0.3146,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2808360755443573,
"step": 1850,
"valid_targets_mean": 6994.8,
"valid_targets_min": 2802
},
{
"epoch": 3.1124161073825505,
"grad_norm": 0.6100969526136583,
"learning_rate": 2.7214733569650967e-05,
"loss": 0.3102,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28686290979385376,
"step": 1855,
"valid_targets_mean": 6349.4,
"valid_targets_min": 2371
},
{
"epoch": 3.120805369127517,
"grad_norm": 0.4316755246345198,
"learning_rate": 2.7136618905795112e-05,
"loss": 0.3045,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2808920443058014,
"step": 1860,
"valid_targets_mean": 6795.1,
"valid_targets_min": 1736
},
{
"epoch": 3.129194630872483,
"grad_norm": 0.40621092637250056,
"learning_rate": 2.7058379289974337e-05,
"loss": 0.2768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24867378175258636,
"step": 1865,
"valid_targets_mean": 7306.8,
"valid_targets_min": 2278
},
{
"epoch": 3.1375838926174495,
"grad_norm": 0.431813624631536,
"learning_rate": 2.698001609205215e-05,
"loss": 0.2987,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28059113025665283,
"step": 1870,
"valid_targets_mean": 7654.8,
"valid_targets_min": 3182
},
{
"epoch": 3.1459731543624163,
"grad_norm": 0.4513769489616382,
"learning_rate": 2.6901530684055784e-05,
"loss": 0.2885,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25672537088394165,
"step": 1875,
"valid_targets_mean": 6007.9,
"valid_targets_min": 2568
},
{
"epoch": 3.1543624161073827,
"grad_norm": 0.4256790978295847,
"learning_rate": 2.6822924440152203e-05,
"loss": 0.3069,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30021923780441284,
"step": 1880,
"valid_targets_mean": 7589.7,
"valid_targets_min": 2205
},
{
"epoch": 3.162751677852349,
"grad_norm": 0.43316319810075854,
"learning_rate": 2.6744198736624037e-05,
"loss": 0.2845,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2516406178474426,
"step": 1885,
"valid_targets_mean": 6713.5,
"valid_targets_min": 2733
},
{
"epoch": 3.1711409395973154,
"grad_norm": 0.4875409054454595,
"learning_rate": 2.6665354951845493e-05,
"loss": 0.2837,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32117947936058044,
"step": 1890,
"valid_targets_mean": 8127.4,
"valid_targets_min": 4010
},
{
"epoch": 3.1795302013422817,
"grad_norm": 0.4368370641874567,
"learning_rate": 2.6586394466258183e-05,
"loss": 0.3102,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32355719804763794,
"step": 1895,
"valid_targets_mean": 7846.9,
"valid_targets_min": 4184
},
{
"epoch": 3.1879194630872485,
"grad_norm": 0.4449816059461504,
"learning_rate": 2.6507318662347014e-05,
"loss": 0.2901,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3063253164291382,
"step": 1900,
"valid_targets_mean": 7408.3,
"valid_targets_min": 3979
},
{
"epoch": 3.196308724832215,
"grad_norm": 0.44951028807579263,
"learning_rate": 2.642812892461593e-05,
"loss": 0.2857,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2589150667190552,
"step": 1905,
"valid_targets_mean": 6944.8,
"valid_targets_min": 2285
},
{
"epoch": 3.204697986577181,
"grad_norm": 0.41413535877537183,
"learning_rate": 2.634882663956371e-05,
"loss": 0.2788,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2686363458633423,
"step": 1910,
"valid_targets_mean": 8719.4,
"valid_targets_min": 5440
},
{
"epoch": 3.2130872483221475,
"grad_norm": 0.6990227504042995,
"learning_rate": 2.6269413195659656e-05,
"loss": 0.332,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40405958890914917,
"step": 1915,
"valid_targets_mean": 3794.5,
"valid_targets_min": 2267
},
{
"epoch": 3.221476510067114,
"grad_norm": 0.6428158519527247,
"learning_rate": 2.6189889983319313e-05,
"loss": 0.3795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38603779673576355,
"step": 1920,
"valid_targets_mean": 4651.1,
"valid_targets_min": 2605
},
{
"epoch": 3.2298657718120807,
"grad_norm": 0.684014374995605,
"learning_rate": 2.6110258394880114e-05,
"loss": 0.385,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3808562755584717,
"step": 1925,
"valid_targets_mean": 3829.8,
"valid_targets_min": 2042
},
{
"epoch": 3.238255033557047,
"grad_norm": 0.64282943969701,
"learning_rate": 2.6030519824577002e-05,
"loss": 0.3831,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4083046317100525,
"step": 1930,
"valid_targets_mean": 5641.4,
"valid_targets_min": 2506
},
{
"epoch": 3.2466442953020134,
"grad_norm": 0.5748425031062804,
"learning_rate": 2.595067566851801e-05,
"loss": 0.3771,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39053720235824585,
"step": 1935,
"valid_targets_mean": 5179.0,
"valid_targets_min": 1579
},
{
"epoch": 3.2550335570469797,
"grad_norm": 0.6277330109492906,
"learning_rate": 2.5870727324659845e-05,
"loss": 0.3647,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4026775360107422,
"step": 1940,
"valid_targets_mean": 4710.8,
"valid_targets_min": 2127
},
{
"epoch": 3.2634228187919465,
"grad_norm": 0.6242211857959218,
"learning_rate": 2.579067619278338e-05,
"loss": 0.3801,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3968504071235657,
"step": 1945,
"valid_targets_mean": 4436.9,
"valid_targets_min": 2477
},
{
"epoch": 3.271812080536913,
"grad_norm": 0.6444616917512285,
"learning_rate": 2.5710523674469158e-05,
"loss": 0.3706,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3324239253997803,
"step": 1950,
"valid_targets_mean": 4587.4,
"valid_targets_min": 2920
},
{
"epoch": 3.280201342281879,
"grad_norm": 0.5556072300490296,
"learning_rate": 2.5630271173072858e-05,
"loss": 0.3654,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35132896900177,
"step": 1955,
"valid_targets_mean": 5464.1,
"valid_targets_min": 2240
},
{
"epoch": 3.2885906040268456,
"grad_norm": 0.594948617124457,
"learning_rate": 2.5549920093700715e-05,
"loss": 0.3585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37158772349357605,
"step": 1960,
"valid_targets_mean": 5314.4,
"valid_targets_min": 2299
},
{
"epoch": 3.296979865771812,
"grad_norm": 0.6840441239223596,
"learning_rate": 2.5469471843184925e-05,
"loss": 0.3563,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.37231218814849854,
"step": 1965,
"valid_targets_mean": 4179.4,
"valid_targets_min": 1855
},
{
"epoch": 3.3053691275167787,
"grad_norm": 0.6157766127078667,
"learning_rate": 2.5388927830059014e-05,
"loss": 0.3833,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35373762249946594,
"step": 1970,
"valid_targets_mean": 4273.2,
"valid_targets_min": 2650
},
{
"epoch": 3.313758389261745,
"grad_norm": 0.6149200620242575,
"learning_rate": 2.5308289464533173e-05,
"loss": 0.3563,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3691704571247101,
"step": 1975,
"valid_targets_mean": 4569.9,
"valid_targets_min": 2279
},
{
"epoch": 3.3221476510067114,
"grad_norm": 0.6167940191503856,
"learning_rate": 2.5227558158469565e-05,
"loss": 0.3546,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36069685220718384,
"step": 1980,
"valid_targets_mean": 5083.7,
"valid_targets_min": 2842
},
{
"epoch": 3.3305369127516777,
"grad_norm": 0.5420102438794782,
"learning_rate": 2.5146735325357613e-05,
"loss": 0.3191,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33579573035240173,
"step": 1985,
"valid_targets_mean": 5135.8,
"valid_targets_min": 1910
},
{
"epoch": 3.3389261744966445,
"grad_norm": 0.5015149384862749,
"learning_rate": 2.506582238028924e-05,
"loss": 0.3384,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3535152077674866,
"step": 1990,
"valid_targets_mean": 6253.7,
"valid_targets_min": 2707
},
{
"epoch": 3.347315436241611,
"grad_norm": 0.5719330778189297,
"learning_rate": 2.4984820739934118e-05,
"loss": 0.3103,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3111467957496643,
"step": 1995,
"valid_targets_mean": 5710.1,
"valid_targets_min": 2364
},
{
"epoch": 3.3557046979865772,
"grad_norm": 0.5970853065564492,
"learning_rate": 2.4903731822514814e-05,
"loss": 0.3159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32904163002967834,
"step": 2000,
"valid_targets_mean": 5014.8,
"valid_targets_min": 2351
},
{
"epoch": 3.3640939597315436,
"grad_norm": 0.6189306792110061,
"learning_rate": 2.4822557047782024e-05,
"loss": 0.3014,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3221728503704071,
"step": 2005,
"valid_targets_mean": 3701.6,
"valid_targets_min": 2426
},
{
"epoch": 3.37248322147651,
"grad_norm": 0.5151530786183165,
"learning_rate": 2.474129783698967e-05,
"loss": 0.3094,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2851989269256592,
"step": 2010,
"valid_targets_mean": 5461.2,
"valid_targets_min": 2245
},
{
"epoch": 3.3808724832214763,
"grad_norm": 0.5725900023340356,
"learning_rate": 2.4659955612870016e-05,
"loss": 0.2912,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2894130349159241,
"step": 2015,
"valid_targets_mean": 4493.8,
"valid_targets_min": 2697
},
{
"epoch": 3.389261744966443,
"grad_norm": 0.5153769358354738,
"learning_rate": 2.4578531799608792e-05,
"loss": 0.2975,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2715255916118622,
"step": 2020,
"valid_targets_mean": 5229.2,
"valid_targets_min": 2171
},
{
"epoch": 3.3976510067114094,
"grad_norm": 0.5745685888306168,
"learning_rate": 2.4497027822820226e-05,
"loss": 0.3339,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3185679018497467,
"step": 2025,
"valid_targets_mean": 4552.8,
"valid_targets_min": 1400
},
{
"epoch": 3.4060402684563758,
"grad_norm": 0.5470525083770996,
"learning_rate": 2.4415445109522095e-05,
"loss": 0.3059,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30468571186065674,
"step": 2030,
"valid_targets_mean": 4832.4,
"valid_targets_min": 2323
},
{
"epoch": 3.414429530201342,
"grad_norm": 0.5589974331861296,
"learning_rate": 2.433378508811074e-05,
"loss": 0.3017,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30940282344818115,
"step": 2035,
"valid_targets_mean": 5444.1,
"valid_targets_min": 2298
},
{
"epoch": 3.422818791946309,
"grad_norm": 0.5122909712097828,
"learning_rate": 2.4252049188336056e-05,
"loss": 0.3114,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30228179693222046,
"step": 2040,
"valid_targets_mean": 5076.4,
"valid_targets_min": 2796
},
{
"epoch": 3.4312080536912752,
"grad_norm": 0.5022232913745113,
"learning_rate": 2.4170238841276455e-05,
"loss": 0.3148,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2891295552253723,
"step": 2045,
"valid_targets_mean": 5921.9,
"valid_targets_min": 3442
},
{
"epoch": 3.4395973154362416,
"grad_norm": 0.554287853662016,
"learning_rate": 2.4088355479313822e-05,
"loss": 0.3005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29244285821914673,
"step": 2050,
"valid_targets_mean": 4365.0,
"valid_targets_min": 1975
},
{
"epoch": 3.447986577181208,
"grad_norm": 0.5260550414333021,
"learning_rate": 2.4006400536108427e-05,
"loss": 0.3136,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30590301752090454,
"step": 2055,
"valid_targets_mean": 5672.2,
"valid_targets_min": 2949
},
{
"epoch": 3.4563758389261743,
"grad_norm": 0.612793177345669,
"learning_rate": 2.3924375446573817e-05,
"loss": 0.3115,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3439445197582245,
"step": 2060,
"valid_targets_mean": 5140.5,
"valid_targets_min": 1948
},
{
"epoch": 3.464765100671141,
"grad_norm": 0.583035039668138,
"learning_rate": 2.3842281646851708e-05,
"loss": 0.301,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3229281008243561,
"step": 2065,
"valid_targets_mean": 4894.6,
"valid_targets_min": 2458
},
{
"epoch": 3.4731543624161074,
"grad_norm": 0.4933811710349901,
"learning_rate": 2.376012057428683e-05,
"loss": 0.3067,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2852545976638794,
"step": 2070,
"valid_targets_mean": 6132.2,
"valid_targets_min": 2886
},
{
"epoch": 3.4815436241610738,
"grad_norm": 0.540786845418613,
"learning_rate": 2.3677893667401772e-05,
"loss": 0.3152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31584304571151733,
"step": 2075,
"valid_targets_mean": 4967.4,
"valid_targets_min": 2392
},
{
"epoch": 3.48993288590604,
"grad_norm": 0.5076718942455755,
"learning_rate": 2.3595602365871776e-05,
"loss": 0.3089,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30375421047210693,
"step": 2080,
"valid_targets_mean": 5562.6,
"valid_targets_min": 1743
},
{
"epoch": 3.498322147651007,
"grad_norm": 0.5290450321960141,
"learning_rate": 2.3513248110499546e-05,
"loss": 0.3071,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3160179853439331,
"step": 2085,
"valid_targets_mean": 5659.8,
"valid_targets_min": 2278
},
{
"epoch": 3.5067114093959733,
"grad_norm": 0.523206758490836,
"learning_rate": 2.343083234319002e-05,
"loss": 0.3138,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27302926778793335,
"step": 2090,
"valid_targets_mean": 5796.7,
"valid_targets_min": 2246
},
{
"epoch": 3.5151006711409396,
"grad_norm": 0.5626998760053522,
"learning_rate": 2.3348356506925117e-05,
"loss": 0.2968,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3022427260875702,
"step": 2095,
"valid_targets_mean": 5033.8,
"valid_targets_min": 2728
},
{
"epoch": 3.523489932885906,
"grad_norm": 0.5266867911032614,
"learning_rate": 2.326582204573848e-05,
"loss": 0.3088,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3060499429702759,
"step": 2100,
"valid_targets_mean": 5743.9,
"valid_targets_min": 2507
},
{
"epoch": 3.5318791946308723,
"grad_norm": 0.5505459812430243,
"learning_rate": 2.3183230404690195e-05,
"loss": 0.3129,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2995159327983856,
"step": 2105,
"valid_targets_mean": 4891.2,
"valid_targets_min": 1922
},
{
"epoch": 3.540268456375839,
"grad_norm": 0.7088219065192544,
"learning_rate": 2.310058302984148e-05,
"loss": 0.2952,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2909039258956909,
"step": 2110,
"valid_targets_mean": 4275.5,
"valid_targets_min": 1993
},
{
"epoch": 3.5486577181208054,
"grad_norm": 0.5198473815132388,
"learning_rate": 2.301788136822937e-05,
"loss": 0.2924,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28495410084724426,
"step": 2115,
"valid_targets_mean": 5320.4,
"valid_targets_min": 2884
},
{
"epoch": 3.557046979865772,
"grad_norm": 0.5643858174679637,
"learning_rate": 2.2935126867841394e-05,
"loss": 0.3061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31023505330085754,
"step": 2120,
"valid_targets_mean": 5626.6,
"valid_targets_min": 2507
},
{
"epoch": 3.565436241610738,
"grad_norm": 0.5874311824212268,
"learning_rate": 2.2852320977590194e-05,
"loss": 0.3075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.280315101146698,
"step": 2125,
"valid_targets_mean": 4510.8,
"valid_targets_min": 2588
},
{
"epoch": 3.573825503355705,
"grad_norm": 0.5522976423655489,
"learning_rate": 2.2769465147288195e-05,
"loss": 0.3111,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27648425102233887,
"step": 2130,
"valid_targets_mean": 4664.8,
"valid_targets_min": 2411
},
{
"epoch": 3.5822147651006713,
"grad_norm": 0.6146463939230231,
"learning_rate": 2.2686560827622186e-05,
"loss": 0.3023,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2761026620864868,
"step": 2135,
"valid_targets_mean": 4296.9,
"valid_targets_min": 1992
},
{
"epoch": 3.5906040268456376,
"grad_norm": 0.5779331367222186,
"learning_rate": 2.2603609470127937e-05,
"loss": 0.3004,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27538245916366577,
"step": 2140,
"valid_targets_mean": 4394.1,
"valid_targets_min": 1874
},
{
"epoch": 3.598993288590604,
"grad_norm": 0.5792185895022344,
"learning_rate": 2.252061252716479e-05,
"loss": 0.3196,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34570086002349854,
"step": 2145,
"valid_targets_mean": 5718.8,
"valid_targets_min": 1804
},
{
"epoch": 3.6073825503355703,
"grad_norm": 0.5572373085343388,
"learning_rate": 2.2437571451890218e-05,
"loss": 0.3183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3412210941314697,
"step": 2150,
"valid_targets_mean": 5650.8,
"valid_targets_min": 2021
},
{
"epoch": 3.6157718120805367,
"grad_norm": 0.5841416387237853,
"learning_rate": 2.2354487698234394e-05,
"loss": 0.3143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34343183040618896,
"step": 2155,
"valid_targets_mean": 4891.8,
"valid_targets_min": 1752
},
{
"epoch": 3.6241610738255035,
"grad_norm": 0.5909359957259901,
"learning_rate": 2.2271362720874716e-05,
"loss": 0.3049,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3036544919013977,
"step": 2160,
"valid_targets_mean": 4496.9,
"valid_targets_min": 2057
},
{
"epoch": 3.63255033557047,
"grad_norm": 0.4745289480046454,
"learning_rate": 2.2188197975210364e-05,
"loss": 0.2942,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2647721767425537,
"step": 2165,
"valid_targets_mean": 6389.7,
"valid_targets_min": 2059
},
{
"epoch": 3.640939597315436,
"grad_norm": 0.5674392657651663,
"learning_rate": 2.2104994917336793e-05,
"loss": 0.2883,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30682075023651123,
"step": 2170,
"valid_targets_mean": 5512.1,
"valid_targets_min": 2473
},
{
"epoch": 3.649328859060403,
"grad_norm": 0.5370100126324117,
"learning_rate": 2.2021755004020252e-05,
"loss": 0.3048,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3067323565483093,
"step": 2175,
"valid_targets_mean": 5682.4,
"valid_targets_min": 2422
},
{
"epoch": 3.6577181208053693,
"grad_norm": 0.5380653490198266,
"learning_rate": 2.1938479692672288e-05,
"loss": 0.3054,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3057229518890381,
"step": 2180,
"valid_targets_mean": 5224.4,
"valid_targets_min": 2639
},
{
"epoch": 3.6661073825503356,
"grad_norm": 0.557532343065383,
"learning_rate": 2.185517044132419e-05,
"loss": 0.2881,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2966899871826172,
"step": 2185,
"valid_targets_mean": 4967.1,
"valid_targets_min": 2410
},
{
"epoch": 3.674496644295302,
"grad_norm": 0.6531169523201229,
"learning_rate": 2.1771828708601523e-05,
"loss": 0.327,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.361418217420578,
"step": 2190,
"valid_targets_mean": 5254.4,
"valid_targets_min": 2342
},
{
"epoch": 3.6828859060402683,
"grad_norm": 0.5013956441848538,
"learning_rate": 2.1688455953698523e-05,
"loss": 0.2818,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2977539896965027,
"step": 2195,
"valid_targets_mean": 5616.4,
"valid_targets_min": 2254
},
{
"epoch": 3.6912751677852347,
"grad_norm": 0.4878613362914435,
"learning_rate": 2.16050536363526e-05,
"loss": 0.2944,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2884175479412079,
"step": 2200,
"valid_targets_mean": 6386.1,
"valid_targets_min": 2194
},
{
"epoch": 3.6996644295302015,
"grad_norm": 0.558669875278882,
"learning_rate": 2.1521623216818747e-05,
"loss": 0.3009,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3179645538330078,
"step": 2205,
"valid_targets_mean": 4951.0,
"valid_targets_min": 2099
},
{
"epoch": 3.708053691275168,
"grad_norm": 0.5413587794144671,
"learning_rate": 2.1438166155843997e-05,
"loss": 0.3075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2814542055130005,
"step": 2210,
"valid_targets_mean": 5224.6,
"valid_targets_min": 2575
},
{
"epoch": 3.716442953020134,
"grad_norm": 0.6485887200779454,
"learning_rate": 2.1354683914641824e-05,
"loss": 0.3311,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32827144861221313,
"step": 2215,
"valid_targets_mean": 5500.0,
"valid_targets_min": 2371
},
{
"epoch": 3.7248322147651005,
"grad_norm": 0.5971733846726407,
"learning_rate": 2.127117795486658e-05,
"loss": 0.3001,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30677330493927,
"step": 2220,
"valid_targets_mean": 4529.9,
"valid_targets_min": 2239
},
{
"epoch": 3.7332214765100673,
"grad_norm": 0.558542387100517,
"learning_rate": 2.1187649738587903e-05,
"loss": 0.3146,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3089045286178589,
"step": 2225,
"valid_targets_mean": 5529.1,
"valid_targets_min": 2939
},
{
"epoch": 3.7416107382550337,
"grad_norm": 0.5461897459365828,
"learning_rate": 2.1104100728265094e-05,
"loss": 0.3048,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3009169101715088,
"step": 2230,
"valid_targets_mean": 6049.8,
"valid_targets_min": 2085
},
{
"epoch": 3.75,
"grad_norm": 0.48054771549470876,
"learning_rate": 2.102053238672154e-05,
"loss": 0.2061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19221694767475128,
"step": 2235,
"valid_targets_mean": 6055.4,
"valid_targets_min": 3996
},
{
"epoch": 3.7583892617449663,
"grad_norm": 0.4068560698574333,
"learning_rate": 2.0936946177119077e-05,
"loss": 0.2107,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1731725037097931,
"step": 2240,
"valid_targets_mean": 6946.6,
"valid_targets_min": 3690
},
{
"epoch": 3.7667785234899327,
"grad_norm": 0.47274143784004097,
"learning_rate": 2.08533435629324e-05,
"loss": 0.2081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20812168717384338,
"step": 2245,
"valid_targets_mean": 6285.7,
"valid_targets_min": 3701
},
{
"epoch": 3.7751677852348995,
"grad_norm": 0.8350425165034582,
"learning_rate": 2.0769726007923424e-05,
"loss": 0.2057,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2108156383037567,
"step": 2250,
"valid_targets_mean": 6577.6,
"valid_targets_min": 4191
},
{
"epoch": 3.783557046979866,
"grad_norm": 0.46422963536533474,
"learning_rate": 2.0686094976115637e-05,
"loss": 0.2033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25461485981941223,
"step": 2255,
"valid_targets_mean": 6968.4,
"valid_targets_min": 4204
},
{
"epoch": 3.791946308724832,
"grad_norm": 0.46851246392123524,
"learning_rate": 2.0602451931768504e-05,
"loss": 0.2033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19045130908489227,
"step": 2260,
"valid_targets_mean": 6070.0,
"valid_targets_min": 2665
},
{
"epoch": 3.8003355704697985,
"grad_norm": 0.4383500350948894,
"learning_rate": 2.051879833935182e-05,
"loss": 0.2016,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18891248106956482,
"step": 2265,
"valid_targets_mean": 6610.8,
"valid_targets_min": 3854
},
{
"epoch": 3.8087248322147653,
"grad_norm": 0.4948811419733198,
"learning_rate": 2.0435135663520033e-05,
"loss": 0.2095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17553368210792542,
"step": 2270,
"valid_targets_mean": 5562.7,
"valid_targets_min": 3428
},
{
"epoch": 3.8171140939597317,
"grad_norm": 0.4466830372228152,
"learning_rate": 2.0351465369086652e-05,
"loss": 0.2148,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20308974385261536,
"step": 2275,
"valid_targets_mean": 6741.6,
"valid_targets_min": 4103
},
{
"epoch": 3.825503355704698,
"grad_norm": 0.4773305850888652,
"learning_rate": 2.0267788920998575e-05,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23688244819641113,
"step": 2280,
"valid_targets_mean": 6210.7,
"valid_targets_min": 3054
},
{
"epoch": 3.8338926174496644,
"grad_norm": 0.5017879051242469,
"learning_rate": 2.0184107784310437e-05,
"loss": 0.2129,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20417118072509766,
"step": 2285,
"valid_targets_mean": 5449.9,
"valid_targets_min": 3429
},
{
"epoch": 3.8422818791946307,
"grad_norm": 0.42374438677742804,
"learning_rate": 2.0100423424158953e-05,
"loss": 0.2057,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20194853842258453,
"step": 2290,
"valid_targets_mean": 6796.1,
"valid_targets_min": 2154
},
{
"epoch": 3.850671140939597,
"grad_norm": 0.45596342356306857,
"learning_rate": 2.0016737305737303e-05,
"loss": 0.1849,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21212635934352875,
"step": 2295,
"valid_targets_mean": 6757.6,
"valid_targets_min": 2843
},
{
"epoch": 3.859060402684564,
"grad_norm": 0.46144727477534486,
"learning_rate": 1.993305089426942e-05,
"loss": 0.2118,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1891375184059143,
"step": 2300,
"valid_targets_mean": 5986.9,
"valid_targets_min": 2902
},
{
"epoch": 3.86744966442953,
"grad_norm": 0.43792305663443815,
"learning_rate": 1.984936565498439e-05,
"loss": 0.213,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19872517883777618,
"step": 2305,
"valid_targets_mean": 6473.8,
"valid_targets_min": 3805
},
{
"epoch": 3.8758389261744965,
"grad_norm": 0.48330819739590886,
"learning_rate": 1.976568305309077e-05,
"loss": 0.1956,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22943609952926636,
"step": 2310,
"valid_targets_mean": 6195.6,
"valid_targets_min": 3360
},
{
"epoch": 3.8842281879194633,
"grad_norm": 0.5000813764898681,
"learning_rate": 1.9682004553750943e-05,
"loss": 0.2088,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1994246244430542,
"step": 2315,
"valid_targets_mean": 5994.8,
"valid_targets_min": 3939
},
{
"epoch": 3.8926174496644297,
"grad_norm": 0.4627526863228376,
"learning_rate": 1.9598331622055447e-05,
"loss": 0.202,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23649166524410248,
"step": 2320,
"valid_targets_mean": 6188.6,
"valid_targets_min": 3302
},
{
"epoch": 3.901006711409396,
"grad_norm": 0.5424967303101933,
"learning_rate": 1.9514665722997365e-05,
"loss": 0.2061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17539843916893005,
"step": 2325,
"valid_targets_mean": 5927.9,
"valid_targets_min": 2692
},
{
"epoch": 3.9093959731543624,
"grad_norm": 0.4503420456669695,
"learning_rate": 1.9431008321446618e-05,
"loss": 0.2095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24055157601833344,
"step": 2330,
"valid_targets_mean": 7060.2,
"valid_targets_min": 3374
},
{
"epoch": 3.9177852348993287,
"grad_norm": 0.4689053016163275,
"learning_rate": 1.934736088212437e-05,
"loss": 0.1964,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18544931709766388,
"step": 2335,
"valid_targets_mean": 6187.8,
"valid_targets_min": 3148
},
{
"epoch": 3.926174496644295,
"grad_norm": 0.48588249258057403,
"learning_rate": 1.926372486957736e-05,
"loss": 0.2018,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22153879702091217,
"step": 2340,
"valid_targets_mean": 6664.8,
"valid_targets_min": 3316
},
{
"epoch": 3.934563758389262,
"grad_norm": 0.4276253010886047,
"learning_rate": 1.9180101748152255e-05,
"loss": 0.2021,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20937304198741913,
"step": 2345,
"valid_targets_mean": 6878.2,
"valid_targets_min": 3191
},
{
"epoch": 3.942953020134228,
"grad_norm": 0.48088975282234525,
"learning_rate": 1.909649298197e-05,
"loss": 0.2161,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2606477737426758,
"step": 2350,
"valid_targets_mean": 6769.5,
"valid_targets_min": 3970
},
{
"epoch": 3.9513422818791946,
"grad_norm": 0.48838778153152856,
"learning_rate": 1.9012900034900233e-05,
"loss": 0.198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19132721424102783,
"step": 2355,
"valid_targets_mean": 5951.7,
"valid_targets_min": 3512
},
{
"epoch": 3.959731543624161,
"grad_norm": 0.48444225690242276,
"learning_rate": 1.8929324370535604e-05,
"loss": 0.198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28127771615982056,
"step": 2360,
"valid_targets_mean": 6932.3,
"valid_targets_min": 3594
},
{
"epoch": 3.9681208053691277,
"grad_norm": 0.4748710452643287,
"learning_rate": 1.884576745216616e-05,
"loss": 0.2083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2094576060771942,
"step": 2365,
"valid_targets_mean": 7217.1,
"valid_targets_min": 4701
},
{
"epoch": 3.976510067114094,
"grad_norm": 0.5317500598985517,
"learning_rate": 1.8762230742753752e-05,
"loss": 0.2,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1885465383529663,
"step": 2370,
"valid_targets_mean": 6264.2,
"valid_targets_min": 4019
},
{
"epoch": 3.9848993288590604,
"grad_norm": 0.5010958802858826,
"learning_rate": 1.867871570490638e-05,
"loss": 0.2253,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21963798999786377,
"step": 2375,
"valid_targets_mean": 6582.8,
"valid_targets_min": 4515
},
{
"epoch": 3.9932885906040267,
"grad_norm": 0.4340304982368214,
"learning_rate": 1.8595223800852634e-05,
"loss": 0.2197,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19284960627555847,
"step": 2380,
"valid_targets_mean": 6617.6,
"valid_targets_min": 4115
},
{
"epoch": 4.001677852348993,
"grad_norm": 0.4853147164676359,
"learning_rate": 1.851175649241603e-05,
"loss": 0.214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2716200351715088,
"step": 2385,
"valid_targets_mean": 7599.1,
"valid_targets_min": 3331
},
{
"epoch": 4.010067114093959,
"grad_norm": 0.514721826011119,
"learning_rate": 1.842831524098948e-05,
"loss": 0.3124,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3376513123512268,
"step": 2390,
"valid_targets_mean": 7885.1,
"valid_targets_min": 4066
},
{
"epoch": 4.018456375838926,
"grad_norm": 0.43561423867279725,
"learning_rate": 1.8344901507509657e-05,
"loss": 0.2808,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27150431275367737,
"step": 2395,
"valid_targets_mean": 7081.2,
"valid_targets_min": 2778
},
{
"epoch": 4.026845637583893,
"grad_norm": 0.445177272510332,
"learning_rate": 1.8261516752431444e-05,
"loss": 0.286,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3342096507549286,
"step": 2400,
"valid_targets_mean": 8086.1,
"valid_targets_min": 2779
},
{
"epoch": 4.035234899328859,
"grad_norm": 0.44166296572836483,
"learning_rate": 1.8178162435702345e-05,
"loss": 0.2931,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28056544065475464,
"step": 2405,
"valid_targets_mean": 7863.1,
"valid_targets_min": 2100
},
{
"epoch": 4.043624161073826,
"grad_norm": 0.4607161191637732,
"learning_rate": 1.8094840016736935e-05,
"loss": 0.2888,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2680191397666931,
"step": 2410,
"valid_targets_mean": 7043.5,
"valid_targets_min": 3773
},
{
"epoch": 4.052013422818792,
"grad_norm": 0.4991969342796842,
"learning_rate": 1.8011550954391316e-05,
"loss": 0.293,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32698434591293335,
"step": 2415,
"valid_targets_mean": 8309.9,
"valid_targets_min": 3755
},
{
"epoch": 4.060402684563758,
"grad_norm": 0.45254746440080984,
"learning_rate": 1.7928296706937535e-05,
"loss": 0.2805,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31133538484573364,
"step": 2420,
"valid_targets_mean": 7775.0,
"valid_targets_min": 3195
},
{
"epoch": 4.068791946308725,
"grad_norm": 0.43160334905718295,
"learning_rate": 1.7845078732038127e-05,
"loss": 0.2998,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2710891664028168,
"step": 2425,
"valid_targets_mean": 7110.2,
"valid_targets_min": 2464
},
{
"epoch": 4.077181208053691,
"grad_norm": 0.4300383596969744,
"learning_rate": 1.7761898486720488e-05,
"loss": 0.2695,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26848047971725464,
"step": 2430,
"valid_targets_mean": 7550.1,
"valid_targets_min": 4712
},
{
"epoch": 4.0855704697986575,
"grad_norm": 0.46327444209541724,
"learning_rate": 1.7678757427351483e-05,
"loss": 0.2999,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30594420433044434,
"step": 2435,
"valid_targets_mean": 7295.6,
"valid_targets_min": 2285
},
{
"epoch": 4.093959731543624,
"grad_norm": 0.4770264420708371,
"learning_rate": 1.7595657009611832e-05,
"loss": 0.2766,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26230883598327637,
"step": 2440,
"valid_targets_mean": 6619.6,
"valid_targets_min": 4150
},
{
"epoch": 4.102348993288591,
"grad_norm": 0.46606607923929133,
"learning_rate": 1.7512598688470728e-05,
"loss": 0.2975,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30441442131996155,
"step": 2445,
"valid_targets_mean": 7391.6,
"valid_targets_min": 1417
},
{
"epoch": 4.110738255033557,
"grad_norm": 0.474648830741764,
"learning_rate": 1.742958391816028e-05,
"loss": 0.2921,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2997223734855652,
"step": 2450,
"valid_targets_mean": 7259.9,
"valid_targets_min": 3704
},
{
"epoch": 4.119127516778524,
"grad_norm": 0.43714707376109374,
"learning_rate": 1.734661415215008e-05,
"loss": 0.289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3029475808143616,
"step": 2455,
"valid_targets_mean": 7824.6,
"valid_targets_min": 2130
},
{
"epoch": 4.12751677852349,
"grad_norm": 0.4373469363456547,
"learning_rate": 1.7263690843121798e-05,
"loss": 0.2671,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.242117777466774,
"step": 2460,
"valid_targets_mean": 7064.0,
"valid_targets_min": 2854
},
{
"epoch": 4.135906040268456,
"grad_norm": 0.43413896957206854,
"learning_rate": 1.718081544294366e-05,
"loss": 0.2764,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.260306179523468,
"step": 2465,
"valid_targets_mean": 7091.0,
"valid_targets_min": 3606
},
{
"epoch": 4.144295302013423,
"grad_norm": 0.5416281298955139,
"learning_rate": 1.7097989402645115e-05,
"loss": 0.2771,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24207207560539246,
"step": 2470,
"valid_targets_mean": 6497.4,
"valid_targets_min": 2404
},
{
"epoch": 4.152684563758389,
"grad_norm": 0.42404304774144197,
"learning_rate": 1.7015214172391375e-05,
"loss": 0.282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29856228828430176,
"step": 2475,
"valid_targets_mean": 8269.1,
"valid_targets_min": 3133
},
{
"epoch": 4.1610738255033555,
"grad_norm": 0.4660327424434792,
"learning_rate": 1.6932491201458045e-05,
"loss": 0.2785,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26031380891799927,
"step": 2480,
"valid_targets_mean": 6791.9,
"valid_targets_min": 3109
},
{
"epoch": 4.169463087248322,
"grad_norm": 0.4619711988508796,
"learning_rate": 1.6849821938205737e-05,
"loss": 0.2541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2729356586933136,
"step": 2485,
"valid_targets_mean": 7605.9,
"valid_targets_min": 3880
},
{
"epoch": 4.177852348993288,
"grad_norm": 0.48261656010138115,
"learning_rate": 1.6767207830054717e-05,
"loss": 0.2937,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28516456484794617,
"step": 2490,
"valid_targets_mean": 7001.3,
"valid_targets_min": 2993
},
{
"epoch": 4.186241610738255,
"grad_norm": 0.435676417744988,
"learning_rate": 1.6684650323459585e-05,
"loss": 0.2786,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28370705246925354,
"step": 2495,
"valid_targets_mean": 8054.1,
"valid_targets_min": 2317
},
{
"epoch": 4.194630872483222,
"grad_norm": 0.4629944442481324,
"learning_rate": 1.66021508638839e-05,
"loss": 0.2802,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2606431841850281,
"step": 2500,
"valid_targets_mean": 6553.3,
"valid_targets_min": 2285
},
{
"epoch": 4.203020134228188,
"grad_norm": 0.5008388089430402,
"learning_rate": 1.6519710895774934e-05,
"loss": 0.2624,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24585726857185364,
"step": 2505,
"valid_targets_mean": 6143.4,
"valid_targets_min": 1991
},
{
"epoch": 4.2114093959731544,
"grad_norm": 0.7370006280619841,
"learning_rate": 1.643733186253831e-05,
"loss": 0.2875,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3383622169494629,
"step": 2510,
"valid_targets_mean": 4226.8,
"valid_targets_min": 1969
},
{
"epoch": 4.219798657718121,
"grad_norm": 0.6726870035574875,
"learning_rate": 1.635501520651281e-05,
"loss": 0.3571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.350715696811676,
"step": 2515,
"valid_targets_mean": 4737.2,
"valid_targets_min": 2343
},
{
"epoch": 4.228187919463087,
"grad_norm": 0.608201519648772,
"learning_rate": 1.6272762368945035e-05,
"loss": 0.3587,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3379996716976166,
"step": 2520,
"valid_targets_mean": 4954.1,
"valid_targets_min": 1754
},
{
"epoch": 4.2365771812080535,
"grad_norm": 0.6748912222477133,
"learning_rate": 1.6190574789964268e-05,
"loss": 0.3501,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34926557540893555,
"step": 2525,
"valid_targets_mean": 4583.1,
"valid_targets_min": 1374
},
{
"epoch": 4.24496644295302,
"grad_norm": 0.5908991732830243,
"learning_rate": 1.6108453908557163e-05,
"loss": 0.3558,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3555208444595337,
"step": 2530,
"valid_targets_mean": 5342.4,
"valid_targets_min": 1817
},
{
"epoch": 4.253355704697986,
"grad_norm": 0.6207911202257151,
"learning_rate": 1.602640116254261e-05,
"loss": 0.3364,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3539900779724121,
"step": 2535,
"valid_targets_mean": 4714.9,
"valid_targets_min": 2533
},
{
"epoch": 4.261744966442953,
"grad_norm": 0.6071187733101508,
"learning_rate": 1.5944417988546565e-05,
"loss": 0.3544,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35557985305786133,
"step": 2540,
"valid_targets_mean": 5150.1,
"valid_targets_min": 2383
},
{
"epoch": 4.27013422818792,
"grad_norm": 0.7303817142219611,
"learning_rate": 1.586250582197685e-05,
"loss": 0.356,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38117295503616333,
"step": 2545,
"valid_targets_mean": 4036.1,
"valid_targets_min": 1598
},
{
"epoch": 4.278523489932886,
"grad_norm": 0.7049488213843446,
"learning_rate": 1.578066609699806e-05,
"loss": 0.3349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3740962743759155,
"step": 2550,
"valid_targets_mean": 4233.7,
"valid_targets_min": 2148
},
{
"epoch": 4.2869127516778525,
"grad_norm": 0.6777203583683766,
"learning_rate": 1.569890024650644e-05,
"loss": 0.3274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34254148602485657,
"step": 2555,
"valid_targets_mean": 4274.1,
"valid_targets_min": 2663
},
{
"epoch": 4.295302013422819,
"grad_norm": 0.6329670857941676,
"learning_rate": 1.561720970210481e-05,
"loss": 0.3287,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32293078303337097,
"step": 2560,
"valid_targets_mean": 4559.8,
"valid_targets_min": 2353
},
{
"epoch": 4.303691275167785,
"grad_norm": 0.6816570333423395,
"learning_rate": 1.5535595894077463e-05,
"loss": 0.359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3947603702545166,
"step": 2565,
"valid_targets_mean": 4805.6,
"valid_targets_min": 1504
},
{
"epoch": 4.3120805369127515,
"grad_norm": 0.6914550996312198,
"learning_rate": 1.545406025136518e-05,
"loss": 0.327,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33923694491386414,
"step": 2570,
"valid_targets_mean": 4224.2,
"valid_targets_min": 2211
},
{
"epoch": 4.320469798657718,
"grad_norm": 0.6783245481279488,
"learning_rate": 1.537260420154016e-05,
"loss": 0.3292,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.359600692987442,
"step": 2575,
"valid_targets_mean": 4393.8,
"valid_targets_min": 2106
},
{
"epoch": 4.328859060402684,
"grad_norm": 0.5897648767880393,
"learning_rate": 1.5291229170781035e-05,
"loss": 0.3028,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28855106234550476,
"step": 2580,
"valid_targets_mean": 5054.9,
"valid_targets_min": 1993
},
{
"epoch": 4.337248322147651,
"grad_norm": 0.608332943102806,
"learning_rate": 1.520993658384794e-05,
"loss": 0.3144,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3093000650405884,
"step": 2585,
"valid_targets_mean": 5129.9,
"valid_targets_min": 1606
},
{
"epoch": 4.345637583892618,
"grad_norm": 0.5820052281954958,
"learning_rate": 1.5128727864057502e-05,
"loss": 0.3005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2772139608860016,
"step": 2590,
"valid_targets_mean": 5087.9,
"valid_targets_min": 1717
},
{
"epoch": 4.354026845637584,
"grad_norm": 0.4872210806517851,
"learning_rate": 1.504760443325799e-05,
"loss": 0.2939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2565154433250427,
"step": 2595,
"valid_targets_mean": 5956.1,
"valid_targets_min": 1968
},
{
"epoch": 4.3624161073825505,
"grad_norm": 0.7519622190322426,
"learning_rate": 1.4966567711804349e-05,
"loss": 0.2825,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2745606601238251,
"step": 2600,
"valid_targets_mean": 5401.3,
"valid_targets_min": 2613
},
{
"epoch": 4.370805369127517,
"grad_norm": 0.4983961960196886,
"learning_rate": 1.4885619118533396e-05,
"loss": 0.2959,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27974072098731995,
"step": 2605,
"valid_targets_mean": 6632.1,
"valid_targets_min": 3364
},
{
"epoch": 4.379194630872483,
"grad_norm": 0.5693873688495318,
"learning_rate": 1.4804760070738935e-05,
"loss": 0.2718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2834015190601349,
"step": 2610,
"valid_targets_mean": 5541.6,
"valid_targets_min": 2548
},
{
"epoch": 4.3875838926174495,
"grad_norm": 0.5204763621636481,
"learning_rate": 1.4723991984146967e-05,
"loss": 0.2821,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25561192631721497,
"step": 2615,
"valid_targets_mean": 5736.2,
"valid_targets_min": 2270
},
{
"epoch": 4.395973154362416,
"grad_norm": 0.5632914480821383,
"learning_rate": 1.464331627289088e-05,
"loss": 0.3043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3166077733039856,
"step": 2620,
"valid_targets_mean": 5236.0,
"valid_targets_min": 1941
},
{
"epoch": 4.404362416107382,
"grad_norm": 0.5748388246599184,
"learning_rate": 1.4562734349486709e-05,
"loss": 0.2884,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.299784779548645,
"step": 2625,
"valid_targets_mean": 5328.9,
"valid_targets_min": 2667
},
{
"epoch": 4.412751677852349,
"grad_norm": 0.8351688187349301,
"learning_rate": 1.4482247624808406e-05,
"loss": 0.2809,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2888963222503662,
"step": 2630,
"valid_targets_mean": 5435.7,
"valid_targets_min": 2181
},
{
"epoch": 4.421140939597316,
"grad_norm": 0.5226679124112104,
"learning_rate": 1.440185750806311e-05,
"loss": 0.2929,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2874293923377991,
"step": 2635,
"valid_targets_mean": 5936.7,
"valid_targets_min": 2860
},
{
"epoch": 4.429530201342282,
"grad_norm": 0.5852102312790867,
"learning_rate": 1.432156540676652e-05,
"loss": 0.297,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28217846155166626,
"step": 2640,
"valid_targets_mean": 4615.2,
"valid_targets_min": 2117
},
{
"epoch": 4.4379194630872485,
"grad_norm": 0.5912206423088883,
"learning_rate": 1.42413727267182e-05,
"loss": 0.2799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.275153249502182,
"step": 2645,
"valid_targets_mean": 4607.4,
"valid_targets_min": 2001
},
{
"epoch": 4.446308724832215,
"grad_norm": 0.5510547672202477,
"learning_rate": 1.416128087197701e-05,
"loss": 0.2906,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2669302821159363,
"step": 2650,
"valid_targets_mean": 4905.5,
"valid_targets_min": 1481
},
{
"epoch": 4.454697986577181,
"grad_norm": 0.5795809906690529,
"learning_rate": 1.4081291244836495e-05,
"loss": 0.2853,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2650591731071472,
"step": 2655,
"valid_targets_mean": 5034.6,
"valid_targets_min": 2853
},
{
"epoch": 4.4630872483221475,
"grad_norm": 0.5795788725358288,
"learning_rate": 1.4001405245800323e-05,
"loss": 0.286,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2940075397491455,
"step": 2660,
"valid_targets_mean": 4866.4,
"valid_targets_min": 2434
},
{
"epoch": 4.471476510067114,
"grad_norm": 0.5556124220513181,
"learning_rate": 1.3921624273557828e-05,
"loss": 0.2948,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2876970171928406,
"step": 2665,
"valid_targets_mean": 5490.1,
"valid_targets_min": 2224
},
{
"epoch": 4.47986577181208,
"grad_norm": 0.6269495521357358,
"learning_rate": 1.3841949724959422e-05,
"loss": 0.29,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30386286973953247,
"step": 2670,
"valid_targets_mean": 4621.5,
"valid_targets_min": 2806
},
{
"epoch": 4.488255033557047,
"grad_norm": 0.5742943619358468,
"learning_rate": 1.376238299499224e-05,
"loss": 0.2919,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2706253230571747,
"step": 2675,
"valid_targets_mean": 4650.4,
"valid_targets_min": 1988
},
{
"epoch": 4.496644295302014,
"grad_norm": 0.5642358371187647,
"learning_rate": 1.3682925476755633e-05,
"loss": 0.2881,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25260767340660095,
"step": 2680,
"valid_targets_mean": 4461.3,
"valid_targets_min": 2329
},
{
"epoch": 4.50503355704698,
"grad_norm": 0.6129367263474906,
"learning_rate": 1.3603578561436827e-05,
"loss": 0.3037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2913645803928375,
"step": 2685,
"valid_targets_mean": 4773.5,
"valid_targets_min": 2442
},
{
"epoch": 4.5134228187919465,
"grad_norm": 0.551861148819051,
"learning_rate": 1.3524343638286537e-05,
"loss": 0.2734,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25387758016586304,
"step": 2690,
"valid_targets_mean": 4992.1,
"valid_targets_min": 2397
},
{
"epoch": 4.521812080536913,
"grad_norm": 0.5887236812131857,
"learning_rate": 1.344522209459468e-05,
"loss": 0.2913,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30325213074684143,
"step": 2695,
"valid_targets_mean": 5113.8,
"valid_targets_min": 2660
},
{
"epoch": 4.530201342281879,
"grad_norm": 0.5217532063399144,
"learning_rate": 1.3366215315666034e-05,
"loss": 0.2952,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2804284989833832,
"step": 2700,
"valid_targets_mean": 5891.8,
"valid_targets_min": 2865
},
{
"epoch": 4.5385906040268456,
"grad_norm": 0.5193635384780518,
"learning_rate": 1.3287324684796019e-05,
"loss": 0.2785,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25956568121910095,
"step": 2705,
"valid_targets_mean": 5862.6,
"valid_targets_min": 2547
},
{
"epoch": 4.546979865771812,
"grad_norm": 0.5489831534061368,
"learning_rate": 1.3208551583246484e-05,
"loss": 0.2768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28311023116111755,
"step": 2710,
"valid_targets_mean": 5806.4,
"valid_targets_min": 2441
},
{
"epoch": 4.555369127516778,
"grad_norm": 0.540199644443618,
"learning_rate": 1.3129897390221485e-05,
"loss": 0.2831,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2677020728588104,
"step": 2715,
"valid_targets_mean": 5587.1,
"valid_targets_min": 2632
},
{
"epoch": 4.563758389261745,
"grad_norm": 0.5436125412994021,
"learning_rate": 1.3051363482843182e-05,
"loss": 0.296,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28627437353134155,
"step": 2720,
"valid_targets_mean": 5415.5,
"valid_targets_min": 1939
},
{
"epoch": 4.572147651006711,
"grad_norm": 0.5747523793137329,
"learning_rate": 1.2972951236127682e-05,
"loss": 0.2911,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27813655138015747,
"step": 2725,
"valid_targets_mean": 5290.0,
"valid_targets_min": 2625
},
{
"epoch": 4.580536912751678,
"grad_norm": 0.5910680967683857,
"learning_rate": 1.289466202296102e-05,
"loss": 0.2827,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2696879506111145,
"step": 2730,
"valid_targets_mean": 4538.6,
"valid_targets_min": 2657
},
{
"epoch": 4.5889261744966445,
"grad_norm": 0.5730597779184343,
"learning_rate": 1.2816497214075057e-05,
"loss": 0.282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2722139060497284,
"step": 2735,
"valid_targets_mean": 5449.6,
"valid_targets_min": 2145
},
{
"epoch": 4.597315436241611,
"grad_norm": 0.8398089754647258,
"learning_rate": 1.2738458178023527e-05,
"loss": 0.2862,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2623152434825897,
"step": 2740,
"valid_targets_mean": 4971.4,
"valid_targets_min": 1896
},
{
"epoch": 4.605704697986577,
"grad_norm": 0.6553716482187634,
"learning_rate": 1.2660546281158075e-05,
"loss": 0.3006,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30956971645355225,
"step": 2745,
"valid_targets_mean": 4423.3,
"valid_targets_min": 2343
},
{
"epoch": 4.614093959731544,
"grad_norm": 0.5509008667253257,
"learning_rate": 1.2582762887604298e-05,
"loss": 0.2951,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26514390110969543,
"step": 2750,
"valid_targets_mean": 4928.9,
"valid_targets_min": 2671
},
{
"epoch": 4.62248322147651,
"grad_norm": 0.5951253558076498,
"learning_rate": 1.2505109359237897e-05,
"loss": 0.2947,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3166443109512329,
"step": 2755,
"valid_targets_mean": 5017.6,
"valid_targets_min": 2167
},
{
"epoch": 4.630872483221476,
"grad_norm": 0.5566438080906776,
"learning_rate": 1.2427587055660819e-05,
"loss": 0.2829,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29906171560287476,
"step": 2760,
"valid_targets_mean": 5451.3,
"valid_targets_min": 1939
},
{
"epoch": 4.639261744966443,
"grad_norm": 0.5910283570343577,
"learning_rate": 1.2350197334177458e-05,
"loss": 0.2635,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28604206442832947,
"step": 2765,
"valid_targets_mean": 4653.0,
"valid_targets_min": 2312
},
{
"epoch": 4.64765100671141,
"grad_norm": 0.5559346129501184,
"learning_rate": 1.2272941549770864e-05,
"loss": 0.2864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2756231129169464,
"step": 2770,
"valid_targets_mean": 5461.8,
"valid_targets_min": 2815
},
{
"epoch": 4.656040268456376,
"grad_norm": 0.6208879977678844,
"learning_rate": 1.2195821055079078e-05,
"loss": 0.287,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2915377616882324,
"step": 2775,
"valid_targets_mean": 4882.7,
"valid_targets_min": 2137
},
{
"epoch": 4.6644295302013425,
"grad_norm": 0.5534164665840433,
"learning_rate": 1.2118837200371372e-05,
"loss": 0.2724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2411075234413147,
"step": 2780,
"valid_targets_mean": 5331.1,
"valid_targets_min": 2099
},
{
"epoch": 4.672818791946309,
"grad_norm": 0.6010404234076611,
"learning_rate": 1.204199133352468e-05,
"loss": 0.2968,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3226349949836731,
"step": 2785,
"valid_targets_mean": 5406.6,
"valid_targets_min": 2166
},
{
"epoch": 4.681208053691275,
"grad_norm": 0.656975187159069,
"learning_rate": 1.1965284799999947e-05,
"loss": 0.2778,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27804136276245117,
"step": 2790,
"valid_targets_mean": 5404.9,
"valid_targets_min": 1919
},
{
"epoch": 4.689597315436242,
"grad_norm": 0.5629538790995963,
"learning_rate": 1.1888718942818587e-05,
"loss": 0.2787,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28015273809432983,
"step": 2795,
"valid_targets_mean": 4785.9,
"valid_targets_min": 2309
},
{
"epoch": 4.697986577181208,
"grad_norm": 0.527420142626763,
"learning_rate": 1.1812295102539002e-05,
"loss": 0.2768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.280708372592926,
"step": 2800,
"valid_targets_mean": 6101.4,
"valid_targets_min": 2722
},
{
"epoch": 4.706375838926174,
"grad_norm": 0.5516311276083228,
"learning_rate": 1.1736014617233047e-05,
"loss": 0.2951,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3039126396179199,
"step": 2805,
"valid_targets_mean": 5812.8,
"valid_targets_min": 2733
},
{
"epoch": 4.714765100671141,
"grad_norm": 0.656745158068306,
"learning_rate": 1.165987882246267e-05,
"loss": 0.3041,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30918699502944946,
"step": 2810,
"valid_targets_mean": 5479.2,
"valid_targets_min": 2503
},
{
"epoch": 4.723154362416107,
"grad_norm": 0.5401018450335546,
"learning_rate": 1.1583889051256448e-05,
"loss": 0.2884,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2784702777862549,
"step": 2815,
"valid_targets_mean": 5206.1,
"valid_targets_min": 2613
},
{
"epoch": 4.731543624161074,
"grad_norm": 0.5781014291806919,
"learning_rate": 1.150804663408636e-05,
"loss": 0.2957,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3208625912666321,
"step": 2820,
"valid_targets_mean": 5432.6,
"valid_targets_min": 2352
},
{
"epoch": 4.739932885906041,
"grad_norm": 0.6368261954922712,
"learning_rate": 1.143235289884437e-05,
"loss": 0.2897,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2843908667564392,
"step": 2825,
"valid_targets_mean": 4625.9,
"valid_targets_min": 1904
},
{
"epoch": 4.748322147651007,
"grad_norm": 0.704140637848927,
"learning_rate": 1.1356809170819258e-05,
"loss": 0.2144,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1777072250843048,
"step": 2830,
"valid_targets_mean": 6397.6,
"valid_targets_min": 4315
},
{
"epoch": 4.756711409395973,
"grad_norm": 0.47549952432795023,
"learning_rate": 1.1281416772673394e-05,
"loss": 0.2009,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22065138816833496,
"step": 2835,
"valid_targets_mean": 6245.4,
"valid_targets_min": 1507
},
{
"epoch": 4.76510067114094,
"grad_norm": 0.4449784773697147,
"learning_rate": 1.1206177024419577e-05,
"loss": 0.1882,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19456003606319427,
"step": 2840,
"valid_targets_mean": 6353.1,
"valid_targets_min": 4320
},
{
"epoch": 4.773489932885906,
"grad_norm": 0.5019031110364914,
"learning_rate": 1.1131091243397924e-05,
"loss": 0.1914,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17940054833889008,
"step": 2845,
"valid_targets_mean": 5759.1,
"valid_targets_min": 3512
},
{
"epoch": 4.781879194630872,
"grad_norm": 0.47849909148730146,
"learning_rate": 1.1056160744252802e-05,
"loss": 0.1823,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17156977951526642,
"step": 2850,
"valid_targets_mean": 6142.6,
"valid_targets_min": 3520
},
{
"epoch": 4.790268456375839,
"grad_norm": 0.4505396927291348,
"learning_rate": 1.0981386838909825e-05,
"loss": 0.2027,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1775951087474823,
"step": 2855,
"valid_targets_mean": 6468.2,
"valid_targets_min": 2613
},
{
"epoch": 4.798657718120805,
"grad_norm": 1.0647002336131273,
"learning_rate": 1.0906770836552864e-05,
"loss": 0.1882,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2161630094051361,
"step": 2860,
"valid_targets_mean": 5752.1,
"valid_targets_min": 3901
},
{
"epoch": 4.807046979865772,
"grad_norm": 0.4594281535594725,
"learning_rate": 1.0832314043601151e-05,
"loss": 0.1999,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17830264568328857,
"step": 2865,
"valid_targets_mean": 6393.5,
"valid_targets_min": 3846
},
{
"epoch": 4.815436241610739,
"grad_norm": 0.46894295972395944,
"learning_rate": 1.075801776368636e-05,
"loss": 0.1962,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18615929782390594,
"step": 2870,
"valid_targets_mean": 6567.8,
"valid_targets_min": 3515
},
{
"epoch": 4.823825503355705,
"grad_norm": 0.5262962678608996,
"learning_rate": 1.068388329762984e-05,
"loss": 0.1974,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21721166372299194,
"step": 2875,
"valid_targets_mean": 6412.0,
"valid_targets_min": 3455
},
{
"epoch": 4.832214765100671,
"grad_norm": 0.4360845572105429,
"learning_rate": 1.06099119434198e-05,
"loss": 0.2059,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14607548713684082,
"step": 2880,
"valid_targets_mean": 6533.2,
"valid_targets_min": 3761
},
{
"epoch": 4.840604026845638,
"grad_norm": 0.47637015612750755,
"learning_rate": 1.0536104996188597e-05,
"loss": 0.1926,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17691192030906677,
"step": 2885,
"valid_targets_mean": 6240.8,
"valid_targets_min": 4018
},
{
"epoch": 4.848993288590604,
"grad_norm": 0.4721048496934634,
"learning_rate": 1.0462463748190053e-05,
"loss": 0.1702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13999581336975098,
"step": 2890,
"valid_targets_mean": 5873.0,
"valid_targets_min": 2984
},
{
"epoch": 4.85738255033557,
"grad_norm": 0.47818853140757095,
"learning_rate": 1.0388989488776842e-05,
"loss": 0.2029,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2173914760351181,
"step": 2895,
"valid_targets_mean": 6636.1,
"valid_targets_min": 1250
},
{
"epoch": 4.865771812080537,
"grad_norm": 0.4933586427426549,
"learning_rate": 1.03156835043779e-05,
"loss": 0.1979,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16754987835884094,
"step": 2900,
"valid_targets_mean": 5647.1,
"valid_targets_min": 3552
},
{
"epoch": 4.874161073825503,
"grad_norm": 0.4832947658452345,
"learning_rate": 1.0242547078475913e-05,
"loss": 0.1762,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1728580892086029,
"step": 2905,
"valid_targets_mean": 5428.6,
"valid_targets_min": 3757
},
{
"epoch": 4.882550335570469,
"grad_norm": 0.4496864870451458,
"learning_rate": 1.0169581491584834e-05,
"loss": 0.2015,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21508343517780304,
"step": 2910,
"valid_targets_mean": 8308.4,
"valid_targets_min": 5460
},
{
"epoch": 4.890939597315437,
"grad_norm": 0.4704527561762588,
"learning_rate": 1.0096788021227491e-05,
"loss": 0.1814,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19710998237133026,
"step": 2915,
"valid_targets_mean": 6912.1,
"valid_targets_min": 3715
},
{
"epoch": 4.899328859060403,
"grad_norm": 0.6239210413556465,
"learning_rate": 1.0024167941913161e-05,
"loss": 0.2051,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2420748770236969,
"step": 2920,
"valid_targets_mean": 6459.6,
"valid_targets_min": 3426
},
{
"epoch": 4.907718120805369,
"grad_norm": 0.5048108498592602,
"learning_rate": 9.951722525115325e-06,
"loss": 0.184,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19225165247917175,
"step": 2925,
"valid_targets_mean": 6065.3,
"valid_targets_min": 3891
},
{
"epoch": 4.916107382550336,
"grad_norm": 0.5190670849455624,
"learning_rate": 9.879453039249358e-06,
"loss": 0.1943,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21929067373275757,
"step": 2930,
"valid_targets_mean": 6779.8,
"valid_targets_min": 3424
},
{
"epoch": 4.924496644295302,
"grad_norm": 0.47699843154532523,
"learning_rate": 9.807360749650346e-06,
"loss": 0.181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17991751432418823,
"step": 2935,
"valid_targets_mean": 6102.3,
"valid_targets_min": 2016
},
{
"epoch": 4.932885906040268,
"grad_norm": 0.43225391604515273,
"learning_rate": 9.735446918550915e-06,
"loss": 0.192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18601876497268677,
"step": 2940,
"valid_targets_mean": 7915.7,
"valid_targets_min": 3023
},
{
"epoch": 4.941275167785235,
"grad_norm": 0.4618978801317599,
"learning_rate": 9.663712805059142e-06,
"loss": 0.1936,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16046935319900513,
"step": 2945,
"valid_targets_mean": 6680.5,
"valid_targets_min": 3246
},
{
"epoch": 4.949664429530201,
"grad_norm": 0.4720343532225163,
"learning_rate": 9.592159665136514e-06,
"loss": 0.1994,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21626628935337067,
"step": 2950,
"valid_targets_mean": 7108.4,
"valid_targets_min": 2988
},
{
"epoch": 4.958053691275168,
"grad_norm": 0.45250084098148113,
"learning_rate": 9.520788751575896e-06,
"loss": 0.168,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16823607683181763,
"step": 2955,
"valid_targets_mean": 6721.4,
"valid_targets_min": 3967
},
{
"epoch": 4.966442953020135,
"grad_norm": 0.49025570586568806,
"learning_rate": 9.44960131397969e-06,
"loss": 0.2111,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.256847620010376,
"step": 2960,
"valid_targets_mean": 7286.5,
"valid_targets_min": 2785
},
{
"epoch": 4.974832214765101,
"grad_norm": 0.4735867595052831,
"learning_rate": 9.37859859873783e-06,
"loss": 0.1918,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17443475127220154,
"step": 2965,
"valid_targets_mean": 5466.8,
"valid_targets_min": 3332
},
{
"epoch": 4.983221476510067,
"grad_norm": 0.6798764788425822,
"learning_rate": 9.307781849006094e-06,
"loss": 0.2068,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2139081507921219,
"step": 2970,
"valid_targets_mean": 7009.6,
"valid_targets_min": 3915
},
{
"epoch": 4.991610738255034,
"grad_norm": 0.5195061284963551,
"learning_rate": 9.237152304684203e-06,
"loss": 0.2119,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22304928302764893,
"step": 2975,
"valid_targets_mean": 5836.3,
"valid_targets_min": 1642
},
{
"epoch": 5.0,
"grad_norm": 0.47197583081561484,
"learning_rate": 9.166711202394243e-06,
"loss": 0.1851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21409043669700623,
"step": 2980,
"valid_targets_mean": 6589.3,
"valid_targets_min": 3933
},
{
"epoch": 5.008389261744966,
"grad_norm": 0.47870448609101784,
"learning_rate": 9.096459775458898e-06,
"loss": 0.2866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2664000988006592,
"step": 2985,
"valid_targets_mean": 8496.2,
"valid_targets_min": 3717
},
{
"epoch": 5.016778523489933,
"grad_norm": 0.4739912398355983,
"learning_rate": 9.026399253879941e-06,
"loss": 0.2826,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32267603278160095,
"step": 2990,
"valid_targets_mean": 8275.7,
"valid_targets_min": 5484
},
{
"epoch": 5.025167785234899,
"grad_norm": 0.46279022009864834,
"learning_rate": 8.956530864316653e-06,
"loss": 0.2617,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2048969566822052,
"step": 2995,
"valid_targets_mean": 7025.2,
"valid_targets_min": 1165
},
{
"epoch": 5.033557046979865,
"grad_norm": 0.46033239201695125,
"learning_rate": 8.88685583006436e-06,
"loss": 0.2917,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25695496797561646,
"step": 3000,
"valid_targets_mean": 6943.1,
"valid_targets_min": 2898
},
{
"epoch": 5.041946308724833,
"grad_norm": 0.45448076696826567,
"learning_rate": 8.817375371033026e-06,
"loss": 0.2794,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28961265087127686,
"step": 3005,
"valid_targets_mean": 8003.8,
"valid_targets_min": 3472
},
{
"epoch": 5.050335570469799,
"grad_norm": 0.4390274600464881,
"learning_rate": 8.748090703725846e-06,
"loss": 0.2698,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2670999765396118,
"step": 3010,
"valid_targets_mean": 7904.6,
"valid_targets_min": 3065
},
{
"epoch": 5.058724832214765,
"grad_norm": 0.4309997894016485,
"learning_rate": 8.679003041218039e-06,
"loss": 0.2726,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.256050705909729,
"step": 3015,
"valid_targets_mean": 7352.1,
"valid_targets_min": 3473
},
{
"epoch": 5.067114093959732,
"grad_norm": 0.4305909045993248,
"learning_rate": 8.610113593135492e-06,
"loss": 0.2956,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27161574363708496,
"step": 3020,
"valid_targets_mean": 8099.6,
"valid_targets_min": 3148
},
{
"epoch": 5.075503355704698,
"grad_norm": 0.4485275201330663,
"learning_rate": 8.5414235656337e-06,
"loss": 0.2584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23827433586120605,
"step": 3025,
"valid_targets_mean": 7346.4,
"valid_targets_min": 4070
},
{
"epoch": 5.083892617449664,
"grad_norm": 0.5572830362540551,
"learning_rate": 8.472934161376529e-06,
"loss": 0.2805,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26340413093566895,
"step": 3030,
"valid_targets_mean": 7365.1,
"valid_targets_min": 3180
},
{
"epoch": 5.092281879194631,
"grad_norm": 0.4704814613721344,
"learning_rate": 8.40464657951528e-06,
"loss": 0.2738,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2828633785247803,
"step": 3035,
"valid_targets_mean": 7270.7,
"valid_targets_min": 3439
},
{
"epoch": 5.100671140939597,
"grad_norm": 0.49678173562729816,
"learning_rate": 8.336562015667587e-06,
"loss": 0.2771,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27936238050460815,
"step": 3040,
"valid_targets_mean": 6689.4,
"valid_targets_min": 2870
},
{
"epoch": 5.109060402684563,
"grad_norm": 0.730343574954326,
"learning_rate": 8.268681661896547e-06,
"loss": 0.2814,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3053019940853119,
"step": 3045,
"valid_targets_mean": 7673.3,
"valid_targets_min": 3397
},
{
"epoch": 5.117449664429531,
"grad_norm": 0.44616181799630433,
"learning_rate": 8.201006706689838e-06,
"loss": 0.2783,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2513405680656433,
"step": 3050,
"valid_targets_mean": 7722.0,
"valid_targets_min": 4597
},
{
"epoch": 5.125838926174497,
"grad_norm": 0.43161004718512397,
"learning_rate": 8.13353833493889e-06,
"loss": 0.2677,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2722732424736023,
"step": 3055,
"valid_targets_mean": 7659.4,
"valid_targets_min": 2432
},
{
"epoch": 5.134228187919463,
"grad_norm": 0.45615272518939565,
"learning_rate": 8.066277727918166e-06,
"loss": 0.2615,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3073369264602661,
"step": 3060,
"valid_targets_mean": 7931.2,
"valid_targets_min": 3733
},
{
"epoch": 5.14261744966443,
"grad_norm": 0.45437577972785165,
"learning_rate": 7.999226063264456e-06,
"loss": 0.2692,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2379218339920044,
"step": 3065,
"valid_targets_mean": 7204.1,
"valid_targets_min": 3650
},
{
"epoch": 5.151006711409396,
"grad_norm": 0.45449479054366987,
"learning_rate": 7.932384514956282e-06,
"loss": 0.2594,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2663235068321228,
"step": 3070,
"valid_targets_mean": 6685.4,
"valid_targets_min": 2822
},
{
"epoch": 5.159395973154362,
"grad_norm": 0.47392972988772686,
"learning_rate": 7.865754253293298e-06,
"loss": 0.2746,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26842206716537476,
"step": 3075,
"valid_targets_mean": 7153.2,
"valid_targets_min": 2650
},
{
"epoch": 5.167785234899329,
"grad_norm": 0.4754855905699106,
"learning_rate": 7.799336444875885e-06,
"loss": 0.2405,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2305595576763153,
"step": 3080,
"valid_targets_mean": 6273.4,
"valid_targets_min": 2485
},
{
"epoch": 5.176174496644295,
"grad_norm": 0.5460104429522009,
"learning_rate": 7.733132252584616e-06,
"loss": 0.2802,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2933194637298584,
"step": 3085,
"valid_targets_mean": 6593.8,
"valid_targets_min": 3230
},
{
"epoch": 5.184563758389261,
"grad_norm": 0.44097726754756655,
"learning_rate": 7.66714283555999e-06,
"loss": 0.2684,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26890644431114197,
"step": 3090,
"valid_targets_mean": 7739.9,
"valid_targets_min": 2932
},
{
"epoch": 5.192953020134228,
"grad_norm": 0.4577779960367475,
"learning_rate": 7.6013693491820836e-06,
"loss": 0.2739,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2969372868537903,
"step": 3095,
"valid_targets_mean": 8366.8,
"valid_targets_min": 2430
},
{
"epoch": 5.201342281879195,
"grad_norm": 0.46292516018769403,
"learning_rate": 7.535812945050345e-06,
"loss": 0.2542,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2700284719467163,
"step": 3100,
"valid_targets_mean": 7350.1,
"valid_targets_min": 4405
},
{
"epoch": 5.209731543624161,
"grad_norm": 0.6821491770532897,
"learning_rate": 7.4704747709634115e-06,
"loss": 0.258,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.321647047996521,
"step": 3105,
"valid_targets_mean": 5855.0,
"valid_targets_min": 2012
},
{
"epoch": 5.218120805369128,
"grad_norm": 0.7178775314566408,
"learning_rate": 7.4053559708990395e-06,
"loss": 0.3382,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36472243070602417,
"step": 3110,
"valid_targets_mean": 4784.8,
"valid_targets_min": 2094
},
{
"epoch": 5.226510067114094,
"grad_norm": 1.673129983685166,
"learning_rate": 7.340457684994054e-06,
"loss": 0.3427,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35261544585227966,
"step": 3115,
"valid_targets_mean": 3623.2,
"valid_targets_min": 1395
},
{
"epoch": 5.23489932885906,
"grad_norm": 0.7340301707191419,
"learning_rate": 7.275781049524396e-06,
"loss": 0.3308,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.36930322647094727,
"step": 3120,
"valid_targets_mean": 4250.2,
"valid_targets_min": 2286
},
{
"epoch": 5.243288590604027,
"grad_norm": 0.6157367497270916,
"learning_rate": 7.211327196885203e-06,
"loss": 0.3377,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34948423504829407,
"step": 3125,
"valid_targets_mean": 5208.2,
"valid_targets_min": 2893
},
{
"epoch": 5.251677852348993,
"grad_norm": 0.6230716394450224,
"learning_rate": 7.1470972555710514e-06,
"loss": 0.3212,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28309696912765503,
"step": 3130,
"valid_targets_mean": 4094.2,
"valid_targets_min": 2428
},
{
"epoch": 5.260067114093959,
"grad_norm": 0.7455011889084798,
"learning_rate": 7.083092350156107e-06,
"loss": 0.3357,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3154565095901489,
"step": 3135,
"valid_targets_mean": 3976.8,
"valid_targets_min": 2196
},
{
"epoch": 5.268456375838926,
"grad_norm": 0.6455017235416665,
"learning_rate": 7.0193136012745e-06,
"loss": 0.333,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.332882285118103,
"step": 3140,
"valid_targets_mean": 4607.0,
"valid_targets_min": 2387
},
{
"epoch": 5.276845637583893,
"grad_norm": 0.6479087847373363,
"learning_rate": 6.95576212560068e-06,
"loss": 0.3171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3014754056930542,
"step": 3145,
"valid_targets_mean": 3891.6,
"valid_targets_min": 2396
},
{
"epoch": 5.285234899328859,
"grad_norm": 0.6232554711769066,
"learning_rate": 6.892439035829872e-06,
"loss": 0.3176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29828381538391113,
"step": 3150,
"valid_targets_mean": 4769.2,
"valid_targets_min": 2532
},
{
"epoch": 5.293624161073826,
"grad_norm": 0.6436196576063646,
"learning_rate": 6.8293454406585856e-06,
"loss": 0.3145,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2895636260509491,
"step": 3155,
"valid_targets_mean": 4180.4,
"valid_targets_min": 2412
},
{
"epoch": 5.302013422818792,
"grad_norm": 0.689976694149992,
"learning_rate": 6.766482444765214e-06,
"loss": 0.3263,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32714468240737915,
"step": 3160,
"valid_targets_mean": 4524.7,
"valid_targets_min": 1851
},
{
"epoch": 5.310402684563758,
"grad_norm": 0.6752591725405911,
"learning_rate": 6.703851148790692e-06,
"loss": 0.3199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3028419613838196,
"step": 3165,
"valid_targets_mean": 4274.9,
"valid_targets_min": 2163
},
{
"epoch": 5.318791946308725,
"grad_norm": 0.693517481253681,
"learning_rate": 6.641452649319194e-06,
"loss": 0.3064,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31211239099502563,
"step": 3170,
"valid_targets_mean": 3913.9,
"valid_targets_min": 2006
},
{
"epoch": 5.327181208053691,
"grad_norm": 0.6371822082542182,
"learning_rate": 6.5792880388590065e-06,
"loss": 0.3016,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2866162359714508,
"step": 3175,
"valid_targets_mean": 5175.9,
"valid_targets_min": 2342
},
{
"epoch": 5.3355704697986575,
"grad_norm": 0.6558028064378439,
"learning_rate": 6.517358405823306e-06,
"loss": 0.2963,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.292873352766037,
"step": 3180,
"valid_targets_mean": 4147.4,
"valid_targets_min": 2188
},
{
"epoch": 5.343959731543624,
"grad_norm": 0.5179265719146311,
"learning_rate": 6.455664834511201e-06,
"loss": 0.2947,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2651475667953491,
"step": 3185,
"valid_targets_mean": 6421.8,
"valid_targets_min": 2865
},
{
"epoch": 5.35234899328859,
"grad_norm": 0.5569622147942522,
"learning_rate": 6.394208405088649e-06,
"loss": 0.2852,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32521840929985046,
"step": 3190,
"valid_targets_mean": 6181.7,
"valid_targets_min": 2653
},
{
"epoch": 5.360738255033557,
"grad_norm": 0.573028215588994,
"learning_rate": 6.332990193569637e-06,
"loss": 0.2658,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26033419370651245,
"step": 3195,
"valid_targets_mean": 4921.6,
"valid_targets_min": 1130
},
{
"epoch": 5.369127516778524,
"grad_norm": 0.5493305346532782,
"learning_rate": 6.272011271797261e-06,
"loss": 0.2818,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2660631239414215,
"step": 3200,
"valid_targets_mean": 4941.5,
"valid_targets_min": 2248
},
{
"epoch": 5.37751677852349,
"grad_norm": 0.5567962399066856,
"learning_rate": 6.211272707425015e-06,
"loss": 0.2597,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2474927008152008,
"step": 3205,
"valid_targets_mean": 5334.0,
"valid_targets_min": 2202
},
{
"epoch": 5.385906040268456,
"grad_norm": 0.5510079696959509,
"learning_rate": 6.150775563898084e-06,
"loss": 0.2752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27622801065444946,
"step": 3210,
"valid_targets_mean": 5717.2,
"valid_targets_min": 1693
},
{
"epoch": 5.394295302013423,
"grad_norm": 0.5827697433258731,
"learning_rate": 6.090520900434706e-06,
"loss": 0.2794,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3159923255443573,
"step": 3215,
"valid_targets_mean": 5333.2,
"valid_targets_min": 2446
},
{
"epoch": 5.402684563758389,
"grad_norm": 0.6222624506584661,
"learning_rate": 6.030509772007662e-06,
"loss": 0.2792,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25103989243507385,
"step": 3220,
"valid_targets_mean": 4057.7,
"valid_targets_min": 2037
},
{
"epoch": 5.4110738255033555,
"grad_norm": 0.5566658371187673,
"learning_rate": 5.970743229325746e-06,
"loss": 0.2706,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24668127298355103,
"step": 3225,
"valid_targets_mean": 5323.5,
"valid_targets_min": 2568
},
{
"epoch": 5.419463087248322,
"grad_norm": 0.5136283372772473,
"learning_rate": 5.911222318815455e-06,
"loss": 0.282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26495736837387085,
"step": 3230,
"valid_targets_mean": 6049.9,
"valid_targets_min": 2894
},
{
"epoch": 5.427852348993289,
"grad_norm": 0.5982640347495186,
"learning_rate": 5.851948082602572e-06,
"loss": 0.2856,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2754901647567749,
"step": 3235,
"valid_targets_mean": 4810.2,
"valid_targets_min": 2273
},
{
"epoch": 5.436241610738255,
"grad_norm": 0.5905717948749144,
"learning_rate": 5.792921558494011e-06,
"loss": 0.2679,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2733975648880005,
"step": 3240,
"valid_targets_mean": 5604.8,
"valid_targets_min": 1967
},
{
"epoch": 5.444630872483222,
"grad_norm": 0.6570229798399173,
"learning_rate": 5.7341437799595574e-06,
"loss": 0.279,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3204910159111023,
"step": 3245,
"valid_targets_mean": 4861.6,
"valid_targets_min": 2047
},
{
"epoch": 5.453020134228188,
"grad_norm": 0.5228660349927623,
"learning_rate": 5.675615776113863e-06,
"loss": 0.2736,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26201581954956055,
"step": 3250,
"valid_targets_mean": 5708.2,
"valid_targets_min": 2660
},
{
"epoch": 5.4614093959731544,
"grad_norm": 0.6123767341953513,
"learning_rate": 5.617338571698343e-06,
"loss": 0.2693,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2594669461250305,
"step": 3255,
"valid_targets_mean": 5425.9,
"valid_targets_min": 1784
},
{
"epoch": 5.469798657718121,
"grad_norm": 0.613782611644556,
"learning_rate": 5.559313187063298e-06,
"loss": 0.283,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2767484486103058,
"step": 3260,
"valid_targets_mean": 4510.3,
"valid_targets_min": 2328
},
{
"epoch": 5.478187919463087,
"grad_norm": 0.5575730640709154,
"learning_rate": 5.5015406381500205e-06,
"loss": 0.2754,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2704010605812073,
"step": 3265,
"valid_targets_mean": 5093.3,
"valid_targets_min": 2668
},
{
"epoch": 5.4865771812080535,
"grad_norm": 0.8604661424911019,
"learning_rate": 5.444021936473008e-06,
"loss": 0.2856,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3080679774284363,
"step": 3270,
"valid_targets_mean": 5909.3,
"valid_targets_min": 2398
},
{
"epoch": 5.49496644295302,
"grad_norm": 0.5840707811766466,
"learning_rate": 5.386758089102266e-06,
"loss": 0.2796,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2685091197490692,
"step": 3275,
"valid_targets_mean": 5225.2,
"valid_targets_min": 2706
},
{
"epoch": 5.503355704697986,
"grad_norm": 0.8636651414120872,
"learning_rate": 5.329750098645645e-06,
"loss": 0.2832,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26322051882743835,
"step": 3280,
"valid_targets_mean": 4588.1,
"valid_targets_min": 2204
},
{
"epoch": 5.5117449664429525,
"grad_norm": 0.5860517348406411,
"learning_rate": 5.272998963231344e-06,
"loss": 0.2691,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2648066282272339,
"step": 3285,
"valid_targets_mean": 5412.2,
"valid_targets_min": 2473
},
{
"epoch": 5.52013422818792,
"grad_norm": 0.5272694115119618,
"learning_rate": 5.216505676490357e-06,
"loss": 0.2689,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2624067962169647,
"step": 3290,
"valid_targets_mean": 5886.4,
"valid_targets_min": 2693
},
{
"epoch": 5.528523489932886,
"grad_norm": 0.6240160287071939,
"learning_rate": 5.160271227539149e-06,
"loss": 0.287,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30568933486938477,
"step": 3295,
"valid_targets_mean": 5643.7,
"valid_targets_min": 2087
},
{
"epoch": 5.5369127516778525,
"grad_norm": 0.5396336101437137,
"learning_rate": 5.104296600962293e-06,
"loss": 0.2718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2812108099460602,
"step": 3300,
"valid_targets_mean": 5636.5,
"valid_targets_min": 2342
},
{
"epoch": 5.545302013422819,
"grad_norm": 0.5271077766258576,
"learning_rate": 5.048582776795243e-06,
"loss": 0.2603,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24015697836875916,
"step": 3305,
"valid_targets_mean": 5548.0,
"valid_targets_min": 2174
},
{
"epoch": 5.553691275167785,
"grad_norm": 0.5321830811059617,
"learning_rate": 4.9931307305071855e-06,
"loss": 0.2746,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29764196276664734,
"step": 3310,
"valid_targets_mean": 5774.5,
"valid_targets_min": 2188
},
{
"epoch": 5.5620805369127515,
"grad_norm": 0.5550944856458149,
"learning_rate": 4.937941432983944e-06,
"loss": 0.2805,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.278522253036499,
"step": 3315,
"valid_targets_mean": 5900.6,
"valid_targets_min": 2320
},
{
"epoch": 5.570469798657718,
"grad_norm": 0.5829907995697335,
"learning_rate": 4.883015850510992e-06,
"loss": 0.2793,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28701257705688477,
"step": 3320,
"valid_targets_mean": 5101.9,
"valid_targets_min": 1893
},
{
"epoch": 5.578859060402684,
"grad_norm": 0.5567254079395402,
"learning_rate": 4.828354944756528e-06,
"loss": 0.2721,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25332432985305786,
"step": 3325,
"valid_targets_mean": 5646.4,
"valid_targets_min": 1766
},
{
"epoch": 5.587248322147651,
"grad_norm": 0.66148945698542,
"learning_rate": 4.77395967275464e-06,
"loss": 0.2702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3146895468235016,
"step": 3330,
"valid_targets_mean": 6108.6,
"valid_targets_min": 1998
},
{
"epoch": 5.595637583892618,
"grad_norm": 0.581068718679316,
"learning_rate": 4.719830986888563e-06,
"loss": 0.2752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2924078404903412,
"step": 3335,
"valid_targets_mean": 5481.2,
"valid_targets_min": 2120
},
{
"epoch": 5.604026845637584,
"grad_norm": 0.606534691738802,
"learning_rate": 4.665969834873962e-06,
"loss": 0.2803,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2749243378639221,
"step": 3340,
"valid_targets_mean": 5890.2,
"valid_targets_min": 3253
},
{
"epoch": 5.6124161073825505,
"grad_norm": 0.6025828438194316,
"learning_rate": 4.612377159742394e-06,
"loss": 0.2918,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3006375730037689,
"step": 3345,
"valid_targets_mean": 4708.5,
"valid_targets_min": 1846
},
{
"epoch": 5.620805369127517,
"grad_norm": 0.531150977265223,
"learning_rate": 4.559053899824759e-06,
"loss": 0.2722,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2626258134841919,
"step": 3350,
"valid_targets_mean": 5491.2,
"valid_targets_min": 1920
},
{
"epoch": 5.629194630872483,
"grad_norm": 0.5768486615852984,
"learning_rate": 4.506000988734891e-06,
"loss": 0.2739,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30473679304122925,
"step": 3355,
"valid_targets_mean": 4956.1,
"valid_targets_min": 2345
},
{
"epoch": 5.6375838926174495,
"grad_norm": 0.5123514854207768,
"learning_rate": 4.453219355353196e-06,
"loss": 0.2543,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23201391100883484,
"step": 3360,
"valid_targets_mean": 5131.4,
"valid_targets_min": 1719
},
{
"epoch": 5.645973154362416,
"grad_norm": 0.6006963742896815,
"learning_rate": 4.4007099238104e-06,
"loss": 0.2764,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29507288336753845,
"step": 3365,
"valid_targets_mean": 4511.2,
"valid_targets_min": 2213
},
{
"epoch": 5.654362416107382,
"grad_norm": 0.6003991113770719,
"learning_rate": 4.34847361347136e-06,
"loss": 0.2721,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29218244552612305,
"step": 3370,
"valid_targets_mean": 5346.6,
"valid_targets_min": 2552
},
{
"epoch": 5.662751677852349,
"grad_norm": 0.6073430367710112,
"learning_rate": 4.2965113389189805e-06,
"loss": 0.2708,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2581964135169983,
"step": 3375,
"valid_targets_mean": 5379.6,
"valid_targets_min": 2402
},
{
"epoch": 5.671140939597316,
"grad_norm": 0.5571360677531385,
"learning_rate": 4.244824009938193e-06,
"loss": 0.2697,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31131523847579956,
"step": 3380,
"valid_targets_mean": 6561.0,
"valid_targets_min": 2063
},
{
"epoch": 5.679530201342282,
"grad_norm": 0.5504870859548427,
"learning_rate": 4.193412531500003e-06,
"loss": 0.2757,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25597238540649414,
"step": 3385,
"valid_targets_mean": 5516.9,
"valid_targets_min": 2094
},
{
"epoch": 5.6879194630872485,
"grad_norm": 0.4758922019113617,
"learning_rate": 4.142277803745711e-06,
"loss": 0.2671,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25087445974349976,
"step": 3390,
"valid_targets_mean": 6917.1,
"valid_targets_min": 2649
},
{
"epoch": 5.696308724832215,
"grad_norm": 0.5698014081403312,
"learning_rate": 4.09142072197106e-06,
"loss": 0.2649,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2483270764350891,
"step": 3395,
"valid_targets_mean": 5088.7,
"valid_targets_min": 2515
},
{
"epoch": 5.704697986577181,
"grad_norm": 0.7218408377099315,
"learning_rate": 4.040842176610662e-06,
"loss": 0.2788,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2839174270629883,
"step": 3400,
"valid_targets_mean": 3442.6,
"valid_targets_min": 2112
},
{
"epoch": 5.7130872483221475,
"grad_norm": 0.5873346576030452,
"learning_rate": 3.990543053222309e-06,
"loss": 0.2917,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30810725688934326,
"step": 3405,
"valid_targets_mean": 5594.4,
"valid_targets_min": 2832
},
{
"epoch": 5.721476510067114,
"grad_norm": 0.6198439223665871,
"learning_rate": 3.940524232471543e-06,
"loss": 0.2836,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.270698219537735,
"step": 3410,
"valid_targets_mean": 5105.4,
"valid_targets_min": 2265
},
{
"epoch": 5.72986577181208,
"grad_norm": 0.5974454589212128,
"learning_rate": 3.8907865901162e-06,
"loss": 0.2752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25877535343170166,
"step": 3415,
"valid_targets_mean": 5383.6,
"valid_targets_min": 2041
},
{
"epoch": 5.7382550335570475,
"grad_norm": 0.6183517738913075,
"learning_rate": 3.841330996991086e-06,
"loss": 0.286,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26947277784347534,
"step": 3420,
"valid_targets_mean": 6037.8,
"valid_targets_min": 2956
},
{
"epoch": 5.746644295302014,
"grad_norm": 0.5483479179128886,
"learning_rate": 3.792158318992738e-06,
"loss": 0.2263,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2083623856306076,
"step": 3425,
"valid_targets_mean": 6712.6,
"valid_targets_min": 3122
},
{
"epoch": 5.75503355704698,
"grad_norm": 0.49190989676005326,
"learning_rate": 3.743269417064226e-06,
"loss": 0.1845,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1769864559173584,
"step": 3430,
"valid_targets_mean": 5881.0,
"valid_targets_min": 1614
},
{
"epoch": 5.7634228187919465,
"grad_norm": 0.561912659237006,
"learning_rate": 3.694665147180152e-06,
"loss": 0.1852,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15704482793807983,
"step": 3435,
"valid_targets_mean": 5759.3,
"valid_targets_min": 2511
},
{
"epoch": 5.771812080536913,
"grad_norm": 0.47890066858081604,
"learning_rate": 3.6463463603315716e-06,
"loss": 0.1866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20538030564785004,
"step": 3440,
"valid_targets_mean": 6722.5,
"valid_targets_min": 3416
},
{
"epoch": 5.780201342281879,
"grad_norm": 0.4673386231786179,
"learning_rate": 3.5983139025111926e-06,
"loss": 0.1758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15834662318229675,
"step": 3445,
"valid_targets_mean": 6207.9,
"valid_targets_min": 3333
},
{
"epoch": 5.7885906040268456,
"grad_norm": 0.5024522767814739,
"learning_rate": 3.5505686146984577e-06,
"loss": 0.1939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2148260623216629,
"step": 3450,
"valid_targets_mean": 6317.9,
"valid_targets_min": 3749
},
{
"epoch": 5.796979865771812,
"grad_norm": 0.5206086016821205,
"learning_rate": 3.503111332844924e-06,
"loss": 0.173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16531062126159668,
"step": 3455,
"valid_targets_mean": 5869.5,
"valid_targets_min": 3804
},
{
"epoch": 5.805369127516778,
"grad_norm": 0.5481635904856844,
"learning_rate": 3.4559428878595424e-06,
"loss": 0.2008,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19838261604309082,
"step": 3460,
"valid_targets_mean": 6835.0,
"valid_targets_min": 3131
},
{
"epoch": 5.813758389261745,
"grad_norm": 1.0680530657889125,
"learning_rate": 3.4090641055941574e-06,
"loss": 0.1868,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1862424612045288,
"step": 3465,
"valid_targets_mean": 6305.6,
"valid_targets_min": 3498
},
{
"epoch": 5.822147651006711,
"grad_norm": 0.4593441329958182,
"learning_rate": 3.3624758068290395e-06,
"loss": 0.1844,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1599089503288269,
"step": 3470,
"valid_targets_mean": 7068.8,
"valid_targets_min": 4193
},
{
"epoch": 5.830536912751678,
"grad_norm": 0.561376863461526,
"learning_rate": 3.3161788072584988e-06,
"loss": 0.2121,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23653791844844818,
"step": 3475,
"valid_targets_mean": 5792.0,
"valid_targets_min": 1274
},
{
"epoch": 5.8389261744966445,
"grad_norm": 0.47725966160547145,
"learning_rate": 3.2701739174766222e-06,
"loss": 0.1785,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2512332797050476,
"step": 3480,
"valid_targets_mean": 7094.4,
"valid_targets_min": 4609
},
{
"epoch": 5.847315436241611,
"grad_norm": 0.5068504526072317,
"learning_rate": 3.224461942963071e-06,
"loss": 0.1701,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18296070396900177,
"step": 3485,
"valid_targets_mean": 7215.2,
"valid_targets_min": 3897
},
{
"epoch": 5.855704697986577,
"grad_norm": 0.46608300571713396,
"learning_rate": 3.1790436840689833e-06,
"loss": 0.1804,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18268756568431854,
"step": 3490,
"valid_targets_mean": 6558.3,
"valid_targets_min": 3351
},
{
"epoch": 5.864093959731544,
"grad_norm": 0.48271906884505544,
"learning_rate": 3.133919936002938e-06,
"loss": 0.2009,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21029271185398102,
"step": 3495,
"valid_targets_mean": 6824.3,
"valid_targets_min": 2436
},
{
"epoch": 5.87248322147651,
"grad_norm": 0.49227218427406577,
"learning_rate": 3.0890914888170885e-06,
"loss": 0.1678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1478789895772934,
"step": 3500,
"valid_targets_mean": 6103.6,
"valid_targets_min": 3898
},
{
"epoch": 5.880872483221476,
"grad_norm": 0.4512599500381554,
"learning_rate": 3.0445591273932563e-06,
"loss": 0.1856,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17004519701004028,
"step": 3505,
"valid_targets_mean": 6659.6,
"valid_targets_min": 3256
},
{
"epoch": 5.889261744966443,
"grad_norm": 0.4798981488967879,
"learning_rate": 3.0003236314292494e-06,
"loss": 0.1778,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1538674384355545,
"step": 3510,
"valid_targets_mean": 5699.6,
"valid_targets_min": 3248
},
{
"epoch": 5.89765100671141,
"grad_norm": 0.42409822884403503,
"learning_rate": 2.9563857754251766e-06,
"loss": 0.1889,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1613140106201172,
"step": 3515,
"valid_targets_mean": 7132.4,
"valid_targets_min": 3729
},
{
"epoch": 5.906040268456376,
"grad_norm": 0.5120995048382582,
"learning_rate": 2.912746328669902e-06,
"loss": 0.1868,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17431381344795227,
"step": 3520,
"valid_targets_mean": 5948.7,
"valid_targets_min": 2287
},
{
"epoch": 5.9144295302013425,
"grad_norm": 0.4613739475931782,
"learning_rate": 2.8694060552275703e-06,
"loss": 0.1821,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15326857566833496,
"step": 3525,
"valid_targets_mean": 5960.7,
"valid_targets_min": 1667
},
{
"epoch": 5.922818791946309,
"grad_norm": 0.535321688669644,
"learning_rate": 2.8263657139242305e-06,
"loss": 0.1814,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18127524852752686,
"step": 3530,
"valid_targets_mean": 6103.1,
"valid_targets_min": 4121
},
{
"epoch": 5.931208053691275,
"grad_norm": 0.5082073584586247,
"learning_rate": 2.78362605833455e-06,
"loss": 0.1831,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1950998306274414,
"step": 3535,
"valid_targets_mean": 7280.8,
"valid_targets_min": 4421
},
{
"epoch": 5.939597315436242,
"grad_norm": 0.4986802905405172,
"learning_rate": 2.7411878367686172e-06,
"loss": 0.1921,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18825015425682068,
"step": 3540,
"valid_targets_mean": 5570.1,
"valid_targets_min": 3241
},
{
"epoch": 5.947986577181208,
"grad_norm": 0.48454919031908744,
"learning_rate": 2.699051792258851e-06,
"loss": 0.1812,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15350499749183655,
"step": 3545,
"valid_targets_mean": 5923.2,
"valid_targets_min": 1037
},
{
"epoch": 5.956375838926174,
"grad_norm": 0.47865792564718573,
"learning_rate": 2.6572186625469786e-06,
"loss": 0.1702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1787925809621811,
"step": 3550,
"valid_targets_mean": 6194.7,
"valid_targets_min": 3559
},
{
"epoch": 5.964765100671141,
"grad_norm": 0.4747837842628126,
"learning_rate": 2.6156891800711147e-06,
"loss": 0.1864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17221690714359283,
"step": 3555,
"valid_targets_mean": 5676.1,
"valid_targets_min": 3539
},
{
"epoch": 5.973154362416107,
"grad_norm": 0.47000003408021024,
"learning_rate": 2.5744640719529623e-06,
"loss": 0.2012,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19577236473560333,
"step": 3560,
"valid_targets_mean": 6225.4,
"valid_targets_min": 3347
},
{
"epoch": 5.981543624161074,
"grad_norm": 0.4786796168997607,
"learning_rate": 2.5335440599850534e-06,
"loss": 0.1913,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16504432260990143,
"step": 3565,
"valid_targets_mean": 6355.5,
"valid_targets_min": 3652
},
{
"epoch": 5.989932885906041,
"grad_norm": 0.4571524104469842,
"learning_rate": 2.492929860618132e-06,
"loss": 0.2034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.183242529630661,
"step": 3570,
"valid_targets_mean": 6742.0,
"valid_targets_min": 4016
},
{
"epoch": 5.998322147651007,
"grad_norm": 0.4470555887790181,
"learning_rate": 2.4526221849485965e-06,
"loss": 0.1807,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1936855912208557,
"step": 3575,
"valid_targets_mean": 7195.1,
"valid_targets_min": 3331
},
{
"epoch": 6.006711409395973,
"grad_norm": 0.5744620239017322,
"learning_rate": 2.412621738706058e-06,
"loss": 0.2689,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3078378140926361,
"step": 3580,
"valid_targets_mean": 7649.6,
"valid_targets_min": 3784
},
{
"epoch": 6.01510067114094,
"grad_norm": 0.47827601773734446,
"learning_rate": 2.3729292222409784e-06,
"loss": 0.2653,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2858749032020569,
"step": 3585,
"valid_targets_mean": 8470.1,
"valid_targets_min": 2859
},
{
"epoch": 6.023489932885906,
"grad_norm": 0.478504417434516,
"learning_rate": 2.3335453305124122e-06,
"loss": 0.2788,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27858269214630127,
"step": 3590,
"valid_targets_mean": 7579.4,
"valid_targets_min": 2671
},
{
"epoch": 6.031879194630872,
"grad_norm": 0.4866406482441478,
"learning_rate": 2.2944707530758415e-06,
"loss": 0.2753,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2762504816055298,
"step": 3595,
"valid_targets_mean": 7706.4,
"valid_targets_min": 2385
},
{
"epoch": 6.040268456375839,
"grad_norm": 0.505236632176921,
"learning_rate": 2.2557061740710817e-06,
"loss": 0.2673,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.276828408241272,
"step": 3600,
"valid_targets_mean": 6949.2,
"valid_targets_min": 2196
},
{
"epoch": 6.048657718120805,
"grad_norm": 0.4561878390226166,
"learning_rate": 2.2172522722103484e-06,
"loss": 0.2684,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26320117712020874,
"step": 3605,
"valid_targets_mean": 7776.6,
"valid_targets_min": 3797
},
{
"epoch": 6.057046979865772,
"grad_norm": 0.49022817722052164,
"learning_rate": 2.1791097207663193e-06,
"loss": 0.2686,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2686783969402313,
"step": 3610,
"valid_targets_mean": 7734.9,
"valid_targets_min": 3678
},
{
"epoch": 6.065436241610739,
"grad_norm": 0.4955716668844715,
"learning_rate": 2.1412791875603857e-06,
"loss": 0.2862,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28073349595069885,
"step": 3615,
"valid_targets_mean": 6914.5,
"valid_targets_min": 2870
},
{
"epoch": 6.073825503355705,
"grad_norm": 0.5009805384042235,
"learning_rate": 2.1037613349509446e-06,
"loss": 0.2591,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2736015319824219,
"step": 3620,
"valid_targets_mean": 7338.3,
"valid_targets_min": 1373
},
{
"epoch": 6.082214765100671,
"grad_norm": 0.5049224672448991,
"learning_rate": 2.0665568198218032e-06,
"loss": 0.2695,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3070971965789795,
"step": 3625,
"valid_targets_mean": 7915.4,
"valid_targets_min": 2332
},
{
"epoch": 6.090604026845638,
"grad_norm": 0.46053044718311015,
"learning_rate": 2.0296662935706823e-06,
"loss": 0.264,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2544174790382385,
"step": 3630,
"valid_targets_mean": 7457.8,
"valid_targets_min": 2888
},
{
"epoch": 6.098993288590604,
"grad_norm": 0.475355952968601,
"learning_rate": 1.9930904020978015e-06,
"loss": 0.2719,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2597702741622925,
"step": 3635,
"valid_targets_mean": 6991.1,
"valid_targets_min": 2394
},
{
"epoch": 6.10738255033557,
"grad_norm": 0.47613769026423275,
"learning_rate": 1.956829785794585e-06,
"loss": 0.27,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2842886447906494,
"step": 3640,
"valid_targets_mean": 7398.9,
"valid_targets_min": 3358
},
{
"epoch": 6.115771812080537,
"grad_norm": 0.7016306327476952,
"learning_rate": 1.920885079532426e-06,
"loss": 0.2835,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.280631959438324,
"step": 3645,
"valid_targets_mean": 6854.7,
"valid_targets_min": 1726
},
{
"epoch": 6.124161073825503,
"grad_norm": 0.5781023685680786,
"learning_rate": 1.8852569126516097e-06,
"loss": 0.2575,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26035580039024353,
"step": 3650,
"valid_targets_mean": 6910.1,
"valid_targets_min": 3314
},
{
"epoch": 6.132550335570469,
"grad_norm": 0.4944156968192024,
"learning_rate": 1.8499459089502369e-06,
"loss": 0.2486,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2630462646484375,
"step": 3655,
"valid_targets_mean": 6882.6,
"valid_targets_min": 3063
},
{
"epoch": 6.140939597315437,
"grad_norm": 0.45476461537488766,
"learning_rate": 1.8149526866733769e-06,
"loss": 0.2772,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24438373744487762,
"step": 3660,
"valid_targets_mean": 7284.4,
"valid_targets_min": 1161
},
{
"epoch": 6.149328859060403,
"grad_norm": 0.5107069833520144,
"learning_rate": 1.7802778585021619e-06,
"loss": 0.2476,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28608813881874084,
"step": 3665,
"valid_targets_mean": 6990.2,
"valid_targets_min": 1791
},
{
"epoch": 6.157718120805369,
"grad_norm": 0.5015668482985232,
"learning_rate": 1.7459220315431324e-06,
"loss": 0.2682,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27950435876846313,
"step": 3670,
"valid_targets_mean": 7225.6,
"valid_targets_min": 3609
},
{
"epoch": 6.166107382550336,
"grad_norm": 0.46969046556834126,
"learning_rate": 1.7118858073175505e-06,
"loss": 0.2417,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24972212314605713,
"step": 3675,
"valid_targets_mean": 7060.9,
"valid_targets_min": 2549
},
{
"epoch": 6.174496644295302,
"grad_norm": 0.4913191686390605,
"learning_rate": 1.6781697817509092e-06,
"loss": 0.2616,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22810117900371552,
"step": 3680,
"valid_targets_mean": 6822.6,
"valid_targets_min": 2505
},
{
"epoch": 6.182885906040268,
"grad_norm": 0.4619700619771902,
"learning_rate": 1.6447745451624709e-06,
"loss": 0.2672,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26616352796554565,
"step": 3685,
"valid_targets_mean": 8112.8,
"valid_targets_min": 1985
},
{
"epoch": 6.191275167785235,
"grad_norm": 0.4729276556009845,
"learning_rate": 1.6117006822549509e-06,
"loss": 0.2623,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24359706044197083,
"step": 3690,
"valid_targets_mean": 7395.3,
"valid_targets_min": 3697
},
{
"epoch": 6.199664429530201,
"grad_norm": 0.5338622699034223,
"learning_rate": 1.578948772104274e-06,
"loss": 0.2538,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24764780700206757,
"step": 3695,
"valid_targets_mean": 6519.2,
"valid_targets_min": 1969
},
{
"epoch": 6.208053691275167,
"grad_norm": 0.4638131388421407,
"learning_rate": 1.5465193881494145e-06,
"loss": 0.2419,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22132614254951477,
"step": 3700,
"valid_targets_mean": 8276.1,
"valid_targets_min": 6762
},
{
"epoch": 6.216442953020135,
"grad_norm": 0.7911958091543346,
"learning_rate": 1.5144130981824078e-06,
"loss": 0.3212,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3236525356769562,
"step": 3705,
"valid_targets_mean": 4914.0,
"valid_targets_min": 2880
},
{
"epoch": 6.224832214765101,
"grad_norm": 0.7744445491078928,
"learning_rate": 1.4826304643383416e-06,
"loss": 0.3373,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33008724451065063,
"step": 3710,
"valid_targets_mean": 4679.4,
"valid_targets_min": 1761
},
{
"epoch": 6.233221476510067,
"grad_norm": 0.7051402300212682,
"learning_rate": 1.4511720430855824e-06,
"loss": 0.3204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3102956712245941,
"step": 3715,
"valid_targets_mean": 4310.0,
"valid_targets_min": 2294
},
{
"epoch": 6.241610738255034,
"grad_norm": 0.6609633072335865,
"learning_rate": 1.4200383852159694e-06,
"loss": 0.3342,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2929164469242096,
"step": 3720,
"valid_targets_mean": 4812.1,
"valid_targets_min": 2737
},
{
"epoch": 6.25,
"grad_norm": 0.6491330592160737,
"learning_rate": 1.3892300358352206e-06,
"loss": 0.3275,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2806921601295471,
"step": 3725,
"valid_targets_mean": 4642.6,
"valid_targets_min": 2195
},
{
"epoch": 6.258389261744966,
"grad_norm": 0.7348633217997382,
"learning_rate": 1.3587475343533619e-06,
"loss": 0.3217,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3333975672721863,
"step": 3730,
"valid_targets_mean": 4054.5,
"valid_targets_min": 2282
},
{
"epoch": 6.266778523489933,
"grad_norm": 0.6537026474778903,
"learning_rate": 1.3285914144752887e-06,
"loss": 0.3224,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3399960398674011,
"step": 3735,
"valid_targets_mean": 5353.2,
"valid_targets_min": 1488
},
{
"epoch": 6.275167785234899,
"grad_norm": 0.6376830707408451,
"learning_rate": 1.2987622041914261e-06,
"loss": 0.316,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3232038617134094,
"step": 3740,
"valid_targets_mean": 5255.4,
"valid_targets_min": 2195
},
{
"epoch": 6.283557046979865,
"grad_norm": 0.69140529240397,
"learning_rate": 1.2692604257684749e-06,
"loss": 0.3098,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28808358311653137,
"step": 3745,
"valid_targets_mean": 4070.5,
"valid_targets_min": 2035
},
{
"epoch": 6.291946308724833,
"grad_norm": 0.6928546144544808,
"learning_rate": 1.2400865957402797e-06,
"loss": 0.3081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3154190182685852,
"step": 3750,
"valid_targets_mean": 4327.9,
"valid_targets_min": 2313
},
{
"epoch": 6.300335570469799,
"grad_norm": 0.6779214563900702,
"learning_rate": 1.2112412248987737e-06,
"loss": 0.3104,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32815492153167725,
"step": 3755,
"valid_targets_mean": 4667.8,
"valid_targets_min": 1749
},
{
"epoch": 6.308724832214765,
"grad_norm": 0.6807826458254292,
"learning_rate": 1.1827248182850505e-06,
"loss": 0.3169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3172191381454468,
"step": 3760,
"valid_targets_mean": 4554.0,
"valid_targets_min": 1891
},
{
"epoch": 6.317114093959732,
"grad_norm": 0.6224605847137545,
"learning_rate": 1.1545378751804947e-06,
"loss": 0.2965,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.287032812833786,
"step": 3765,
"valid_targets_mean": 4446.8,
"valid_targets_min": 2098
},
{
"epoch": 6.325503355704698,
"grad_norm": 0.6807178371052428,
"learning_rate": 1.1266808890980728e-06,
"loss": 0.2991,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25281140208244324,
"step": 3770,
"valid_targets_mean": 4459.8,
"valid_targets_min": 2331
},
{
"epoch": 6.333892617449664,
"grad_norm": 0.6012763087634683,
"learning_rate": 1.0991543477736677e-06,
"loss": 0.2883,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30087077617645264,
"step": 3775,
"valid_targets_mean": 6275.2,
"valid_targets_min": 2533
},
{
"epoch": 6.342281879194631,
"grad_norm": 1.8182404552077396,
"learning_rate": 1.0719587331575566e-06,
"loss": 0.2939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29444316029548645,
"step": 3780,
"valid_targets_mean": 5172.5,
"valid_targets_min": 1986
},
{
"epoch": 6.350671140939597,
"grad_norm": 0.5446274995073764,
"learning_rate": 1.045094521405956e-06,
"loss": 0.2672,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2806011140346527,
"step": 3785,
"valid_targets_mean": 5646.0,
"valid_targets_min": 2711
},
{
"epoch": 6.359060402684563,
"grad_norm": 0.5498559495970365,
"learning_rate": 1.0185621828726977e-06,
"loss": 0.2731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24814215302467346,
"step": 3790,
"valid_targets_mean": 5899.6,
"valid_targets_min": 2783
},
{
"epoch": 6.367449664429531,
"grad_norm": 0.5919565649250172,
"learning_rate": 9.923621821009922e-07,
"loss": 0.2745,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3031330704689026,
"step": 3795,
"valid_targets_mean": 5326.4,
"valid_targets_min": 2250
},
{
"epoch": 6.375838926174497,
"grad_norm": 0.5536265680989166,
"learning_rate": 9.664949778152843e-07,
"loss": 0.2583,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26124638319015503,
"step": 3800,
"valid_targets_mean": 5666.2,
"valid_targets_min": 2615
},
{
"epoch": 6.384228187919463,
"grad_norm": 0.5395454464735167,
"learning_rate": 9.409610229132338e-07,
"loss": 0.264,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31404566764831543,
"step": 3805,
"valid_targets_mean": 5925.5,
"valid_targets_min": 2441
},
{
"epoch": 6.39261744966443,
"grad_norm": 0.6022961978629413,
"learning_rate": 9.157607644577871e-07,
"loss": 0.2663,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29775312542915344,
"step": 3810,
"valid_targets_mean": 5128.5,
"valid_targets_min": 2734
},
{
"epoch": 6.401006711409396,
"grad_norm": 0.5413477553753987,
"learning_rate": 8.908946436693289e-07,
"loss": 0.2866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26473984122276306,
"step": 3815,
"valid_targets_mean": 6427.1,
"valid_targets_min": 2195
},
{
"epoch": 6.409395973154362,
"grad_norm": 0.5418609809542219,
"learning_rate": 8.66363095917997e-07,
"loss": 0.2656,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2557719051837921,
"step": 3820,
"valid_targets_mean": 5481.1,
"valid_targets_min": 2577
},
{
"epoch": 6.417785234899329,
"grad_norm": 0.574362565687711,
"learning_rate": 8.421665507160059e-07,
"loss": 0.2726,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2666260004043579,
"step": 3825,
"valid_targets_mean": 5730.9,
"valid_targets_min": 2656
},
{
"epoch": 6.426174496644295,
"grad_norm": 0.5850353068125957,
"learning_rate": 8.183054317101801e-07,
"loss": 0.278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.33919697999954224,
"step": 3830,
"valid_targets_mean": 5378.9,
"valid_targets_min": 2528
},
{
"epoch": 6.434563758389261,
"grad_norm": 0.6058285149478984,
"learning_rate": 7.947801566744972e-07,
"loss": 0.2629,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24856190383434296,
"step": 3835,
"valid_targets_mean": 4310.3,
"valid_targets_min": 2349
},
{
"epoch": 6.442953020134228,
"grad_norm": 0.5742100513833732,
"learning_rate": 7.71591137502794e-07,
"loss": 0.2642,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2939155697822571,
"step": 3840,
"valid_targets_mean": 6220.2,
"valid_targets_min": 2448
},
{
"epoch": 6.451342281879195,
"grad_norm": 0.5346257041602569,
"learning_rate": 7.487387802015544e-07,
"loss": 0.28,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29794859886169434,
"step": 3845,
"valid_targets_mean": 5707.8,
"valid_targets_min": 2331
},
{
"epoch": 6.459731543624161,
"grad_norm": 0.5828300112809118,
"learning_rate": 7.262234848827821e-07,
"loss": 0.2648,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2682815492153168,
"step": 3850,
"valid_targets_mean": 5051.4,
"valid_targets_min": 2703
},
{
"epoch": 6.468120805369128,
"grad_norm": 0.575050722473611,
"learning_rate": 7.040456457570255e-07,
"loss": 0.2742,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28402549028396606,
"step": 3855,
"valid_targets_mean": 5654.9,
"valid_targets_min": 2931
},
{
"epoch": 6.476510067114094,
"grad_norm": 0.5317172160332462,
"learning_rate": 6.822056511264419e-07,
"loss": 0.2716,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27136534452438354,
"step": 3860,
"valid_targets_mean": 5810.9,
"valid_targets_min": 2138
},
{
"epoch": 6.48489932885906,
"grad_norm": 0.5721524618837316,
"learning_rate": 6.60703883378031e-07,
"loss": 0.2729,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.258329302072525,
"step": 3865,
"valid_targets_mean": 5527.7,
"valid_targets_min": 1927
},
{
"epoch": 6.493288590604027,
"grad_norm": 0.5842132721266479,
"learning_rate": 6.395407189769099e-07,
"loss": 0.2828,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2716050446033478,
"step": 3870,
"valid_targets_mean": 4609.9,
"valid_targets_min": 1738
},
{
"epoch": 6.501677852348993,
"grad_norm": 0.533890209714909,
"learning_rate": 6.187165284597551e-07,
"loss": 0.2796,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2933617830276489,
"step": 3875,
"valid_targets_mean": 6249.0,
"valid_targets_min": 2300
},
{
"epoch": 6.510067114093959,
"grad_norm": 0.52220489457869,
"learning_rate": 5.982316764282758e-07,
"loss": 0.2642,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3012363910675049,
"step": 3880,
"valid_targets_mean": 6485.4,
"valid_targets_min": 3271
},
{
"epoch": 6.518456375838926,
"grad_norm": 0.5935782644204488,
"learning_rate": 5.780865215428711e-07,
"loss": 0.2644,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.34011632204055786,
"step": 3885,
"valid_targets_mean": 5474.5,
"valid_targets_min": 2357
},
{
"epoch": 6.526845637583893,
"grad_norm": 0.6151496010574028,
"learning_rate": 5.582814165163109e-07,
"loss": 0.273,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27982866764068604,
"step": 3890,
"valid_targets_mean": 4406.9,
"valid_targets_min": 2371
},
{
"epoch": 6.535234899328859,
"grad_norm": 0.5887980527907611,
"learning_rate": 5.388167081075923e-07,
"loss": 0.2714,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27811765670776367,
"step": 3895,
"valid_targets_mean": 5347.7,
"valid_targets_min": 1977
},
{
"epoch": 6.543624161073826,
"grad_norm": 0.5880407834341802,
"learning_rate": 5.196927371158533e-07,
"loss": 0.2635,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3148224353790283,
"step": 3900,
"valid_targets_mean": 5232.4,
"valid_targets_min": 1318
},
{
"epoch": 6.552013422818792,
"grad_norm": 0.6006026304792391,
"learning_rate": 5.009098383744038e-07,
"loss": 0.2584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26570644974708557,
"step": 3905,
"valid_targets_mean": 4733.3,
"valid_targets_min": 2417
},
{
"epoch": 6.560402684563758,
"grad_norm": 0.5224193904069437,
"learning_rate": 4.824683407448749e-07,
"loss": 0.2791,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28725624084472656,
"step": 3910,
"valid_targets_mean": 6044.8,
"valid_targets_min": 3505
},
{
"epoch": 6.568791946308725,
"grad_norm": 0.6447586245356873,
"learning_rate": 4.64368567111444e-07,
"loss": 0.2723,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30462023615837097,
"step": 3915,
"valid_targets_mean": 4460.7,
"valid_targets_min": 2447
},
{
"epoch": 6.577181208053691,
"grad_norm": 0.6564426092689943,
"learning_rate": 4.4661083437520294e-07,
"loss": 0.2739,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2737728953361511,
"step": 3920,
"valid_targets_mean": 4022.8,
"valid_targets_min": 2295
},
{
"epoch": 6.5855704697986575,
"grad_norm": 0.5642864357746021,
"learning_rate": 4.291954534485854e-07,
"loss": 0.2534,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26668137311935425,
"step": 3925,
"valid_targets_mean": 6063.1,
"valid_targets_min": 2559
},
{
"epoch": 6.593959731543624,
"grad_norm": 0.6089569736952457,
"learning_rate": 4.121227292499486e-07,
"loss": 0.2752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3034970462322235,
"step": 3930,
"valid_targets_mean": 4791.2,
"valid_targets_min": 2263
},
{
"epoch": 6.60234899328859,
"grad_norm": 0.5393354458128405,
"learning_rate": 3.953929606982132e-07,
"loss": 0.2793,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27592217922210693,
"step": 3935,
"valid_targets_mean": 6213.2,
"valid_targets_min": 1136
},
{
"epoch": 6.610738255033557,
"grad_norm": 0.5925398764506686,
"learning_rate": 3.790064407076366e-07,
"loss": 0.2823,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27960777282714844,
"step": 3940,
"valid_targets_mean": 5184.2,
"valid_targets_min": 2136
},
{
"epoch": 6.619127516778524,
"grad_norm": 0.556331359870401,
"learning_rate": 3.629634561826945e-07,
"loss": 0.2757,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24813111126422882,
"step": 3945,
"valid_targets_mean": 5770.5,
"valid_targets_min": 2282
},
{
"epoch": 6.62751677852349,
"grad_norm": 0.7127109512244416,
"learning_rate": 3.472642880130406e-07,
"loss": 0.2614,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2581768333911896,
"step": 3950,
"valid_targets_mean": 5229.9,
"valid_targets_min": 1734
},
{
"epoch": 6.635906040268456,
"grad_norm": 0.5830084225170035,
"learning_rate": 3.3190921106860174e-07,
"loss": 0.2644,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2320040464401245,
"step": 3955,
"valid_targets_mean": 4759.6,
"valid_targets_min": 2816
},
{
"epoch": 6.644295302013423,
"grad_norm": 0.6052125591767035,
"learning_rate": 3.168984941947639e-07,
"loss": 0.2596,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2660321593284607,
"step": 3960,
"valid_targets_mean": 4967.5,
"valid_targets_min": 2596
},
{
"epoch": 6.652684563758389,
"grad_norm": 0.5096171355991038,
"learning_rate": 3.0223240020765155e-07,
"loss": 0.2682,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2473021000623703,
"step": 3965,
"valid_targets_mean": 5883.9,
"valid_targets_min": 1924
},
{
"epoch": 6.6610738255033555,
"grad_norm": 0.5727689596025739,
"learning_rate": 2.879111858895489e-07,
"loss": 0.273,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26150262355804443,
"step": 3970,
"valid_targets_mean": 5984.4,
"valid_targets_min": 2377
},
{
"epoch": 6.669463087248322,
"grad_norm": 0.5698461387298254,
"learning_rate": 2.7393510198438166e-07,
"loss": 0.2546,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25220152735710144,
"step": 3975,
"valid_targets_mean": 5250.2,
"valid_targets_min": 2070
},
{
"epoch": 6.677852348993289,
"grad_norm": 0.5270051573305511,
"learning_rate": 2.603043931933402e-07,
"loss": 0.2825,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2431250661611557,
"step": 3980,
"valid_targets_mean": 6141.6,
"valid_targets_min": 2431
},
{
"epoch": 6.686241610738255,
"grad_norm": 0.673697473824313,
"learning_rate": 2.4701929817059656e-07,
"loss": 0.2633,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2727034091949463,
"step": 3985,
"valid_targets_mean": 5132.9,
"valid_targets_min": 3068
},
{
"epoch": 6.694630872483222,
"grad_norm": 0.6780887776966226,
"learning_rate": 2.34080049519112e-07,
"loss": 0.261,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.282716304063797,
"step": 3990,
"valid_targets_mean": 3993.8,
"valid_targets_min": 2800
},
{
"epoch": 6.703020134228188,
"grad_norm": 0.5509057881590569,
"learning_rate": 2.2148687378658496e-07,
"loss": 0.267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2522139549255371,
"step": 3995,
"valid_targets_mean": 5226.9,
"valid_targets_min": 1976
},
{
"epoch": 6.7114093959731544,
"grad_norm": 0.7151036985671345,
"learning_rate": 2.0923999146146067e-07,
"loss": 0.2819,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.29253584146499634,
"step": 4000,
"valid_targets_mean": 5426.8,
"valid_targets_min": 2434
},
{
"epoch": 6.719798657718121,
"grad_norm": 0.6065604339329081,
"learning_rate": 1.973396169690922e-07,
"loss": 0.2864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26587897539138794,
"step": 4005,
"valid_targets_mean": 4664.9,
"valid_targets_min": 2044
},
{
"epoch": 6.728187919463087,
"grad_norm": 0.6571108007293199,
"learning_rate": 1.8578595866797667e-07,
"loss": 0.2724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3036031723022461,
"step": 4010,
"valid_targets_mean": 4845.0,
"valid_targets_min": 2604
},
{
"epoch": 6.7365771812080535,
"grad_norm": 0.5716603086343849,
"learning_rate": 1.7457921884610485e-07,
"loss": 0.2791,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25550198554992676,
"step": 4015,
"valid_targets_mean": 5003.0,
"valid_targets_min": 2237
},
{
"epoch": 6.74496644295302,
"grad_norm": 0.5272754256896639,
"learning_rate": 1.6371959371742186e-07,
"loss": 0.2339,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19655856490135193,
"step": 4020,
"valid_targets_mean": 5693.6,
"valid_targets_min": 1613
},
{
"epoch": 6.753355704697986,
"grad_norm": 0.521168845109171,
"learning_rate": 1.5320727341839424e-07,
"loss": 0.1869,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2061513513326645,
"step": 4025,
"valid_targets_mean": 6052.1,
"valid_targets_min": 3569
},
{
"epoch": 6.7617449664429525,
"grad_norm": 0.5920889932017265,
"learning_rate": 1.4304244200467944e-07,
"loss": 0.186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20252254605293274,
"step": 4030,
"valid_targets_mean": 5929.9,
"valid_targets_min": 3249
},
{
"epoch": 6.77013422818792,
"grad_norm": 0.486960723370642,
"learning_rate": 1.3322527744790147e-07,
"loss": 0.1746,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17718416452407837,
"step": 4035,
"valid_targets_mean": 7443.8,
"valid_targets_min": 3135
},
{
"epoch": 6.778523489932886,
"grad_norm": 0.5277762872045201,
"learning_rate": 1.2375595163253596e-07,
"loss": 0.1824,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1686478853225708,
"step": 4040,
"valid_targets_mean": 5697.2,
"valid_targets_min": 3549
},
{
"epoch": 6.7869127516778525,
"grad_norm": 0.5330888635454919,
"learning_rate": 1.146346303528989e-07,
"loss": 0.1797,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17880107462406158,
"step": 4045,
"valid_targets_mean": 5807.6,
"valid_targets_min": 3662
},
{
"epoch": 6.795302013422819,
"grad_norm": 0.5378053749104562,
"learning_rate": 1.0586147331025364e-07,
"loss": 0.1803,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20560482144355774,
"step": 4050,
"valid_targets_mean": 6232.0,
"valid_targets_min": 3426
},
{
"epoch": 6.803691275167785,
"grad_norm": 0.5139632522955673,
"learning_rate": 9.743663410999748e-08,
"loss": 0.1915,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2150823175907135,
"step": 4055,
"valid_targets_mean": 6666.6,
"valid_targets_min": 4053
},
{
"epoch": 6.8120805369127515,
"grad_norm": 0.5521484823495871,
"learning_rate": 8.936026025898603e-08,
"loss": 0.1872,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22691968083381653,
"step": 4060,
"valid_targets_mean": 5846.2,
"valid_targets_min": 3798
},
{
"epoch": 6.820469798657718,
"grad_norm": 0.5106761702258902,
"learning_rate": 8.163249316294642e-08,
"loss": 0.188,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1579877883195877,
"step": 4065,
"valid_targets_mean": 6118.9,
"valid_targets_min": 3260
},
{
"epoch": 6.828859060402684,
"grad_norm": 0.5261616901015815,
"learning_rate": 7.425346812400147e-08,
"loss": 0.1948,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1700320839881897,
"step": 4070,
"valid_targets_mean": 5992.8,
"valid_targets_min": 3387
},
{
"epoch": 6.837248322147651,
"grad_norm": 0.5158666363680571,
"learning_rate": 6.722331433830053e-08,
"loss": 0.1736,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16234266757965088,
"step": 4075,
"valid_targets_mean": 6847.2,
"valid_targets_min": 3521
},
{
"epoch": 6.845637583892618,
"grad_norm": 0.4868440686752683,
"learning_rate": 6.054215489375459e-08,
"loss": 0.182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1744418740272522,
"step": 4080,
"valid_targets_mean": 6886.6,
"valid_targets_min": 3878
},
{
"epoch": 6.854026845637584,
"grad_norm": 0.5246792710001337,
"learning_rate": 5.421010676788685e-08,
"loss": 0.1794,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19324232637882233,
"step": 4085,
"valid_targets_mean": 5832.1,
"valid_targets_min": 3646
},
{
"epoch": 6.8624161073825505,
"grad_norm": 0.5127901456438103,
"learning_rate": 4.8227280825783315e-08,
"loss": 0.1938,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1729700267314911,
"step": 4090,
"valid_targets_mean": 6038.9,
"valid_targets_min": 1245
},
{
"epoch": 6.870805369127517,
"grad_norm": 0.7002391609347778,
"learning_rate": 4.259378181814766e-08,
"loss": 0.1786,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19172480702400208,
"step": 4095,
"valid_targets_mean": 6231.9,
"valid_targets_min": 3422
},
{
"epoch": 6.879194630872483,
"grad_norm": 0.4713771213069829,
"learning_rate": 3.7309708379469345e-08,
"loss": 0.1802,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1667013019323349,
"step": 4100,
"valid_targets_mean": 6255.6,
"valid_targets_min": 4194
},
{
"epoch": 6.8875838926174495,
"grad_norm": 0.49532595263698936,
"learning_rate": 3.237515302630057e-08,
"loss": 0.1795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1756962537765503,
"step": 4105,
"valid_targets_mean": 6671.9,
"valid_targets_min": 4183
},
{
"epoch": 6.895973154362416,
"grad_norm": 0.49983599002888396,
"learning_rate": 2.7790202155626443e-08,
"loss": 0.1856,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20181216299533844,
"step": 4110,
"valid_targets_mean": 6685.4,
"valid_targets_min": 2159
},
{
"epoch": 6.904362416107382,
"grad_norm": 0.5318550218113627,
"learning_rate": 2.355493604336845e-08,
"loss": 0.1824,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17410992085933685,
"step": 4115,
"valid_targets_mean": 5939.6,
"valid_targets_min": 3334
},
{
"epoch": 6.912751677852349,
"grad_norm": 0.46295538858082336,
"learning_rate": 1.966942884295886e-08,
"loss": 0.185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15322157740592957,
"step": 4120,
"valid_targets_mean": 6672.1,
"valid_targets_min": 4159
},
{
"epoch": 6.921140939597316,
"grad_norm": 0.47037196527210495,
"learning_rate": 1.6133748584061805e-08,
"loss": 0.1745,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14129142463207245,
"step": 4125,
"valid_targets_mean": 6304.5,
"valid_targets_min": 3844
},
{
"epoch": 6.929530201342282,
"grad_norm": 0.5312339452079587,
"learning_rate": 1.2947957171365322e-08,
"loss": 0.1791,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17308609187602997,
"step": 4130,
"valid_targets_mean": 6762.9,
"valid_targets_min": 2156
},
{
"epoch": 6.9379194630872485,
"grad_norm": 0.46570277486483247,
"learning_rate": 1.0112110383513341e-08,
"loss": 0.1926,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16208329796791077,
"step": 4135,
"valid_targets_mean": 7273.2,
"valid_targets_min": 5220
},
{
"epoch": 6.946308724832215,
"grad_norm": 0.5974108217590955,
"learning_rate": 7.62625787211535e-09,
"loss": 0.187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17684219777584076,
"step": 4140,
"valid_targets_mean": 6628.9,
"valid_targets_min": 3077
},
{
"epoch": 6.954697986577181,
"grad_norm": 0.6064678087235988,
"learning_rate": 5.490443160887094e-09,
"loss": 0.1641,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15664657950401306,
"step": 4145,
"valid_targets_mean": 5498.8,
"valid_targets_min": 3451
},
{
"epoch": 6.9630872483221475,
"grad_norm": 0.7059314883538617,
"learning_rate": 3.7047036448778497e-09,
"loss": 0.1866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18920756876468658,
"step": 4150,
"valid_targets_mean": 5527.3,
"valid_targets_min": 3368
},
{
"epoch": 6.971476510067114,
"grad_norm": 0.5294657263902567,
"learning_rate": 2.269070589828726e-09,
"loss": 0.1951,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1674599051475525,
"step": 4155,
"valid_targets_mean": 5724.3,
"valid_targets_min": 2579
},
{
"epoch": 6.97986577181208,
"grad_norm": 0.5661217688292507,
"learning_rate": 1.1835691316153253e-09,
"loss": 0.1964,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18694236874580383,
"step": 4160,
"valid_targets_mean": 6724.3,
"valid_targets_min": 3695
},
{
"epoch": 6.9882550335570475,
"grad_norm": 0.4926191734702717,
"learning_rate": 4.482182758125397e-10,
"loss": 0.1986,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21235015988349915,
"step": 4165,
"valid_targets_mean": 6980.3,
"valid_targets_min": 3782
},
{
"epoch": 6.996644295302014,
"grad_norm": 0.532162824579163,
"learning_rate": 6.303089735926193e-11,
"loss": 0.1774,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16759955883026123,
"step": 4170,
"valid_targets_mean": 6677.1,
"valid_targets_min": 3670
},
{
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20417048037052155,
"step": 4172,
"total_flos": 1952534555328512.0,
"train_loss": 0.3077526548922005,
"train_runtime": 30255.504,
"train_samples_per_second": 2.205,
"train_steps_per_second": 0.138,
"valid_targets_mean": 6589.3,
"valid_targets_min": 3933
}
],
"logging_steps": 5,
"max_steps": 4172,
"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": 1952534555328512.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}