glm46-bash-textbook-traces / trainer_state.json
penfever's picture
End of training
865f515 verified
{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 7.0,
"eval_steps": 500,
"global_step": 2884,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.012135922330097087,
"grad_norm": 12.854040001394706,
"learning_rate": 5.536332179930796e-07,
"loss": 0.6964,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4155339002609253,
"step": 5,
"valid_targets_mean": 4542.5,
"valid_targets_min": 2267
},
{
"epoch": 0.024271844660194174,
"grad_norm": 12.263505819320088,
"learning_rate": 1.2456747404844292e-06,
"loss": 0.6879,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32189181447029114,
"step": 10,
"valid_targets_mean": 3697.0,
"valid_targets_min": 1235
},
{
"epoch": 0.03640776699029126,
"grad_norm": 8.655191989264384,
"learning_rate": 1.9377162629757786e-06,
"loss": 0.6731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27928075194358826,
"step": 15,
"valid_targets_mean": 3204.8,
"valid_targets_min": 679
},
{
"epoch": 0.04854368932038835,
"grad_norm": 6.053602144731899,
"learning_rate": 2.629757785467128e-06,
"loss": 0.6458,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25370243191719055,
"step": 20,
"valid_targets_mean": 3698.5,
"valid_targets_min": 1424
},
{
"epoch": 0.06067961165048544,
"grad_norm": 5.003464297392446,
"learning_rate": 3.3217993079584777e-06,
"loss": 0.612,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27288690209388733,
"step": 25,
"valid_targets_mean": 3452.8,
"valid_targets_min": 1295
},
{
"epoch": 0.07281553398058252,
"grad_norm": 4.021963230554392,
"learning_rate": 4.0138408304498275e-06,
"loss": 0.5331,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25928282737731934,
"step": 30,
"valid_targets_mean": 4078.8,
"valid_targets_min": 982
},
{
"epoch": 0.08495145631067962,
"grad_norm": 1.8655289131571513,
"learning_rate": 4.705882352941177e-06,
"loss": 0.4874,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26225602626800537,
"step": 35,
"valid_targets_mean": 4939.9,
"valid_targets_min": 1989
},
{
"epoch": 0.0970873786407767,
"grad_norm": 1.2623078257855396,
"learning_rate": 5.397923875432527e-06,
"loss": 0.4341,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19804851710796356,
"step": 40,
"valid_targets_mean": 3603.0,
"valid_targets_min": 963
},
{
"epoch": 0.10922330097087378,
"grad_norm": 1.016812149629259,
"learning_rate": 6.089965397923876e-06,
"loss": 0.4436,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2455306053161621,
"step": 45,
"valid_targets_mean": 4351.8,
"valid_targets_min": 1317
},
{
"epoch": 0.12135922330097088,
"grad_norm": 0.8190897088948215,
"learning_rate": 6.782006920415225e-06,
"loss": 0.4312,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14470702409744263,
"step": 50,
"valid_targets_mean": 2596.8,
"valid_targets_min": 1150
},
{
"epoch": 0.13349514563106796,
"grad_norm": 0.8992937986712327,
"learning_rate": 7.474048442906575e-06,
"loss": 0.392,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3150596022605896,
"step": 55,
"valid_targets_mean": 5262.4,
"valid_targets_min": 2103
},
{
"epoch": 0.14563106796116504,
"grad_norm": 0.7941579220121936,
"learning_rate": 8.166089965397924e-06,
"loss": 0.4005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25509214401245117,
"step": 60,
"valid_targets_mean": 3579.2,
"valid_targets_min": 881
},
{
"epoch": 0.15776699029126215,
"grad_norm": 0.6616371695567353,
"learning_rate": 8.858131487889274e-06,
"loss": 0.3801,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20132774114608765,
"step": 65,
"valid_targets_mean": 4061.9,
"valid_targets_min": 1846
},
{
"epoch": 0.16990291262135923,
"grad_norm": 0.7435073664433595,
"learning_rate": 9.550173010380624e-06,
"loss": 0.3712,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18138478696346283,
"step": 70,
"valid_targets_mean": 2758.0,
"valid_targets_min": 1574
},
{
"epoch": 0.1820388349514563,
"grad_norm": 0.6503373357699267,
"learning_rate": 1.0242214532871973e-05,
"loss": 0.3625,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13326121866703033,
"step": 75,
"valid_targets_mean": 2995.8,
"valid_targets_min": 1232
},
{
"epoch": 0.1941747572815534,
"grad_norm": 0.6195684210653057,
"learning_rate": 1.0934256055363323e-05,
"loss": 0.3576,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11990141868591309,
"step": 80,
"valid_targets_mean": 2627.5,
"valid_targets_min": 1076
},
{
"epoch": 0.20631067961165048,
"grad_norm": 0.6556518420141108,
"learning_rate": 1.1626297577854673e-05,
"loss": 0.3622,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13326506316661835,
"step": 85,
"valid_targets_mean": 2755.5,
"valid_targets_min": 963
},
{
"epoch": 0.21844660194174756,
"grad_norm": 0.6791831975620268,
"learning_rate": 1.2318339100346022e-05,
"loss": 0.3631,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17219778895378113,
"step": 90,
"valid_targets_mean": 2926.9,
"valid_targets_min": 744
},
{
"epoch": 0.23058252427184467,
"grad_norm": 0.5758089406951674,
"learning_rate": 1.301038062283737e-05,
"loss": 0.358,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16182292997837067,
"step": 95,
"valid_targets_mean": 4118.5,
"valid_targets_min": 1466
},
{
"epoch": 0.24271844660194175,
"grad_norm": 0.7168748580960569,
"learning_rate": 1.370242214532872e-05,
"loss": 0.341,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.190993070602417,
"step": 100,
"valid_targets_mean": 3162.9,
"valid_targets_min": 1068
},
{
"epoch": 0.25485436893203883,
"grad_norm": 0.6268658529522273,
"learning_rate": 1.439446366782007e-05,
"loss": 0.3404,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17415133118629456,
"step": 105,
"valid_targets_mean": 4034.0,
"valid_targets_min": 2304
},
{
"epoch": 0.2669902912621359,
"grad_norm": 0.5668274141102878,
"learning_rate": 1.5086505190311421e-05,
"loss": 0.3364,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15786170959472656,
"step": 110,
"valid_targets_mean": 4065.1,
"valid_targets_min": 2168
},
{
"epoch": 0.279126213592233,
"grad_norm": 0.7811512860803487,
"learning_rate": 1.577854671280277e-05,
"loss": 0.3295,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1751110851764679,
"step": 115,
"valid_targets_mean": 3848.4,
"valid_targets_min": 1979
},
{
"epoch": 0.2912621359223301,
"grad_norm": 0.7325732214372631,
"learning_rate": 1.647058823529412e-05,
"loss": 0.342,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22932222485542297,
"step": 120,
"valid_targets_mean": 3327.5,
"valid_targets_min": 1445
},
{
"epoch": 0.30339805825242716,
"grad_norm": 0.6234117604561062,
"learning_rate": 1.716262975778547e-05,
"loss": 0.3209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1070050299167633,
"step": 125,
"valid_targets_mean": 2585.1,
"valid_targets_min": 1787
},
{
"epoch": 0.3155339805825243,
"grad_norm": 0.8086481534437464,
"learning_rate": 1.785467128027682e-05,
"loss": 0.3122,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13294263184070587,
"step": 130,
"valid_targets_mean": 1800.2,
"valid_targets_min": 852
},
{
"epoch": 0.3276699029126214,
"grad_norm": 0.779627481503764,
"learning_rate": 1.8546712802768167e-05,
"loss": 0.3184,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16735510528087616,
"step": 135,
"valid_targets_mean": 2570.0,
"valid_targets_min": 154
},
{
"epoch": 0.33980582524271846,
"grad_norm": 0.6561696109611663,
"learning_rate": 1.9238754325259517e-05,
"loss": 0.3057,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13328135013580322,
"step": 140,
"valid_targets_mean": 3109.8,
"valid_targets_min": 1702
},
{
"epoch": 0.35194174757281554,
"grad_norm": 0.6768518880134367,
"learning_rate": 1.9930795847750867e-05,
"loss": 0.3047,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18371689319610596,
"step": 145,
"valid_targets_mean": 4016.6,
"valid_targets_min": 1318
},
{
"epoch": 0.3640776699029126,
"grad_norm": 1.0956881404346224,
"learning_rate": 2.0622837370242218e-05,
"loss": 0.3139,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15525943040847778,
"step": 150,
"valid_targets_mean": 3546.4,
"valid_targets_min": 1141
},
{
"epoch": 0.3762135922330097,
"grad_norm": 0.7171940974026795,
"learning_rate": 2.1314878892733564e-05,
"loss": 0.3155,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1355014443397522,
"step": 155,
"valid_targets_mean": 3003.1,
"valid_targets_min": 1359
},
{
"epoch": 0.3883495145631068,
"grad_norm": 0.6728120751122655,
"learning_rate": 2.2006920415224915e-05,
"loss": 0.3206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1952664852142334,
"step": 160,
"valid_targets_mean": 3901.9,
"valid_targets_min": 1730
},
{
"epoch": 0.40048543689320387,
"grad_norm": 0.6161226727639294,
"learning_rate": 2.2698961937716265e-05,
"loss": 0.3166,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12122677266597748,
"step": 165,
"valid_targets_mean": 2647.9,
"valid_targets_min": 1490
},
{
"epoch": 0.41262135922330095,
"grad_norm": 0.6469585037118063,
"learning_rate": 2.339100346020761e-05,
"loss": 0.313,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19482526183128357,
"step": 170,
"valid_targets_mean": 4572.6,
"valid_targets_min": 2111
},
{
"epoch": 0.42475728155339804,
"grad_norm": 0.9410877866180319,
"learning_rate": 2.4083044982698965e-05,
"loss": 0.3092,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15275007486343384,
"step": 175,
"valid_targets_mean": 3254.4,
"valid_targets_min": 1179
},
{
"epoch": 0.4368932038834951,
"grad_norm": 0.6765430895918028,
"learning_rate": 2.4775086505190315e-05,
"loss": 0.2915,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10032335668802261,
"step": 180,
"valid_targets_mean": 2602.9,
"valid_targets_min": 950
},
{
"epoch": 0.44902912621359226,
"grad_norm": 0.654267067299084,
"learning_rate": 2.5467128027681662e-05,
"loss": 0.309,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1599547266960144,
"step": 185,
"valid_targets_mean": 3938.4,
"valid_targets_min": 2050
},
{
"epoch": 0.46116504854368934,
"grad_norm": 0.5650941683809592,
"learning_rate": 2.6159169550173012e-05,
"loss": 0.2829,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16773730516433716,
"step": 190,
"valid_targets_mean": 4922.8,
"valid_targets_min": 1698
},
{
"epoch": 0.4733009708737864,
"grad_norm": 0.6931328735750193,
"learning_rate": 2.685121107266436e-05,
"loss": 0.3158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15911036729812622,
"step": 195,
"valid_targets_mean": 3501.1,
"valid_targets_min": 1842
},
{
"epoch": 0.4854368932038835,
"grad_norm": 0.7443197401450684,
"learning_rate": 2.7543252595155713e-05,
"loss": 0.2954,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1301368772983551,
"step": 200,
"valid_targets_mean": 2486.9,
"valid_targets_min": 951
},
{
"epoch": 0.4975728155339806,
"grad_norm": 0.5909860805645347,
"learning_rate": 2.8235294117647063e-05,
"loss": 0.3014,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11716020852327347,
"step": 205,
"valid_targets_mean": 2947.4,
"valid_targets_min": 1474
},
{
"epoch": 0.5097087378640777,
"grad_norm": 0.7400619110961242,
"learning_rate": 2.892733564013841e-05,
"loss": 0.2974,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1818861961364746,
"step": 210,
"valid_targets_mean": 3435.0,
"valid_targets_min": 1232
},
{
"epoch": 0.5218446601941747,
"grad_norm": 0.5566261435396108,
"learning_rate": 2.961937716262976e-05,
"loss": 0.2768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14670011401176453,
"step": 215,
"valid_targets_mean": 4573.8,
"valid_targets_min": 1491
},
{
"epoch": 0.5339805825242718,
"grad_norm": 0.6491616411301813,
"learning_rate": 3.0311418685121107e-05,
"loss": 0.3005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15393364429473877,
"step": 220,
"valid_targets_mean": 4069.5,
"valid_targets_min": 1213
},
{
"epoch": 0.5461165048543689,
"grad_norm": 0.651227385714233,
"learning_rate": 3.100346020761246e-05,
"loss": 0.2932,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1222854033112526,
"step": 225,
"valid_targets_mean": 2875.4,
"valid_targets_min": 1986
},
{
"epoch": 0.558252427184466,
"grad_norm": 0.7201639930513787,
"learning_rate": 3.169550173010381e-05,
"loss": 0.2929,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1631743609905243,
"step": 230,
"valid_targets_mean": 3297.5,
"valid_targets_min": 1943
},
{
"epoch": 0.5703883495145631,
"grad_norm": 0.6862592865265383,
"learning_rate": 3.238754325259516e-05,
"loss": 0.2844,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1268606185913086,
"step": 235,
"valid_targets_mean": 3565.4,
"valid_targets_min": 1040
},
{
"epoch": 0.5825242718446602,
"grad_norm": 0.6048275275492347,
"learning_rate": 3.307958477508651e-05,
"loss": 0.3013,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.162528395652771,
"step": 240,
"valid_targets_mean": 4108.6,
"valid_targets_min": 2390
},
{
"epoch": 0.5946601941747572,
"grad_norm": 0.6156091154175889,
"learning_rate": 3.377162629757786e-05,
"loss": 0.2822,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.140919029712677,
"step": 245,
"valid_targets_mean": 3538.2,
"valid_targets_min": 923
},
{
"epoch": 0.6067961165048543,
"grad_norm": 0.6400483190574742,
"learning_rate": 3.446366782006921e-05,
"loss": 0.2775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13898254930973053,
"step": 250,
"valid_targets_mean": 3269.1,
"valid_targets_min": 1421
},
{
"epoch": 0.6189320388349514,
"grad_norm": 0.6970894027738842,
"learning_rate": 3.515570934256056e-05,
"loss": 0.2912,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17165809869766235,
"step": 255,
"valid_targets_mean": 3104.5,
"valid_targets_min": 1410
},
{
"epoch": 0.6310679611650486,
"grad_norm": 0.5939002555284147,
"learning_rate": 3.58477508650519e-05,
"loss": 0.2649,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12092805653810501,
"step": 260,
"valid_targets_mean": 3420.2,
"valid_targets_min": 1898
},
{
"epoch": 0.6432038834951457,
"grad_norm": 0.6528504544546815,
"learning_rate": 3.653979238754326e-05,
"loss": 0.28,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1366952359676361,
"step": 265,
"valid_targets_mean": 3324.0,
"valid_targets_min": 1728
},
{
"epoch": 0.6553398058252428,
"grad_norm": 0.6944728275522337,
"learning_rate": 3.723183391003461e-05,
"loss": 0.2774,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15160223841667175,
"step": 270,
"valid_targets_mean": 3373.8,
"valid_targets_min": 1483
},
{
"epoch": 0.6674757281553398,
"grad_norm": 0.6262629949685379,
"learning_rate": 3.792387543252595e-05,
"loss": 0.268,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15413989126682281,
"step": 275,
"valid_targets_mean": 5052.5,
"valid_targets_min": 1657
},
{
"epoch": 0.6796116504854369,
"grad_norm": 0.6052569249622719,
"learning_rate": 3.86159169550173e-05,
"loss": 0.2799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13474929332733154,
"step": 280,
"valid_targets_mean": 3708.5,
"valid_targets_min": 1614
},
{
"epoch": 0.691747572815534,
"grad_norm": 0.6981366071065592,
"learning_rate": 3.930795847750865e-05,
"loss": 0.2723,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1354372501373291,
"step": 285,
"valid_targets_mean": 2741.2,
"valid_targets_min": 1129
},
{
"epoch": 0.7038834951456311,
"grad_norm": 0.6002483395830216,
"learning_rate": 4e-05,
"loss": 0.2709,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11007172614336014,
"step": 290,
"valid_targets_mean": 3100.9,
"valid_targets_min": 1340
},
{
"epoch": 0.7160194174757282,
"grad_norm": 0.6825989706442029,
"learning_rate": 3.9999633593049284e-05,
"loss": 0.2746,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12751522660255432,
"step": 295,
"valid_targets_mean": 3690.2,
"valid_targets_min": 1666
},
{
"epoch": 0.7281553398058253,
"grad_norm": 0.6576525961646845,
"learning_rate": 3.999853438562254e-05,
"loss": 0.2801,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12925320863723755,
"step": 300,
"valid_targets_mean": 3314.1,
"valid_targets_min": 1471
},
{
"epoch": 0.7402912621359223,
"grad_norm": 0.6792178177035387,
"learning_rate": 3.999670241799547e-05,
"loss": 0.2764,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1715441644191742,
"step": 305,
"valid_targets_mean": 3274.6,
"valid_targets_min": 1216
},
{
"epoch": 0.7524271844660194,
"grad_norm": 0.5121260619365018,
"learning_rate": 3.9994137757292664e-05,
"loss": 0.2706,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09343647956848145,
"step": 310,
"valid_targets_mean": 3678.0,
"valid_targets_min": 1387
},
{
"epoch": 0.7645631067961165,
"grad_norm": 0.5279179315814312,
"learning_rate": 3.999084049748507e-05,
"loss": 0.2704,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1317557543516159,
"step": 315,
"valid_targets_mean": 4906.8,
"valid_targets_min": 1740
},
{
"epoch": 0.7766990291262136,
"grad_norm": 0.5206447170940441,
"learning_rate": 3.998681075938658e-05,
"loss": 0.2619,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1034955084323883,
"step": 320,
"valid_targets_mean": 4226.9,
"valid_targets_min": 1559
},
{
"epoch": 0.7888349514563107,
"grad_norm": 0.5980580937400733,
"learning_rate": 3.998204869064959e-05,
"loss": 0.2871,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12027301639318466,
"step": 325,
"valid_targets_mean": 3074.5,
"valid_targets_min": 1835
},
{
"epoch": 0.8009708737864077,
"grad_norm": 0.6559797958502117,
"learning_rate": 3.997655446575961e-05,
"loss": 0.2799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10785755515098572,
"step": 330,
"valid_targets_mean": 2978.2,
"valid_targets_min": 1134
},
{
"epoch": 0.8131067961165048,
"grad_norm": 0.67670826710767,
"learning_rate": 3.9970328286028866e-05,
"loss": 0.2925,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17468750476837158,
"step": 335,
"valid_targets_mean": 3400.9,
"valid_targets_min": 1934
},
{
"epoch": 0.8252427184466019,
"grad_norm": 0.7159467239533798,
"learning_rate": 3.9963370379588914e-05,
"loss": 0.2713,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1383286565542221,
"step": 340,
"valid_targets_mean": 3279.6,
"valid_targets_min": 2148
},
{
"epoch": 0.837378640776699,
"grad_norm": 0.6267928861956277,
"learning_rate": 3.9955681001382265e-05,
"loss": 0.2848,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15330195426940918,
"step": 345,
"valid_targets_mean": 3971.2,
"valid_targets_min": 1364
},
{
"epoch": 0.8495145631067961,
"grad_norm": 0.6688473921995564,
"learning_rate": 3.994726043315309e-05,
"loss": 0.2767,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17813362181186676,
"step": 350,
"valid_targets_mean": 3752.0,
"valid_targets_min": 1521
},
{
"epoch": 0.8616504854368932,
"grad_norm": 0.6286454777476559,
"learning_rate": 3.9938108983436875e-05,
"loss": 0.2624,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11453067511320114,
"step": 355,
"valid_targets_mean": 3157.2,
"valid_targets_min": 1535
},
{
"epoch": 0.8737864077669902,
"grad_norm": 0.6255661412757746,
"learning_rate": 3.992822698754908e-05,
"loss": 0.2575,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.147567480802536,
"step": 360,
"valid_targets_mean": 4401.5,
"valid_targets_min": 2076
},
{
"epoch": 0.8859223300970874,
"grad_norm": 0.6882879135052699,
"learning_rate": 3.9917614807572905e-05,
"loss": 0.2758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17819637060165405,
"step": 365,
"valid_targets_mean": 3818.5,
"valid_targets_min": 1420
},
{
"epoch": 0.8980582524271845,
"grad_norm": 0.6712737418433733,
"learning_rate": 3.990627283234601e-05,
"loss": 0.2704,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16544793546199799,
"step": 370,
"valid_targets_mean": 4163.1,
"valid_targets_min": 1569
},
{
"epoch": 0.9101941747572816,
"grad_norm": 0.6617855355506335,
"learning_rate": 3.989420147744625e-05,
"loss": 0.2742,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13588759303092957,
"step": 375,
"valid_targets_mean": 2627.9,
"valid_targets_min": 1237
},
{
"epoch": 0.9223300970873787,
"grad_norm": 0.6177844878907712,
"learning_rate": 3.9881401185176456e-05,
"loss": 0.2691,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10162941366434097,
"step": 380,
"valid_targets_mean": 3267.5,
"valid_targets_min": 1613
},
{
"epoch": 0.9344660194174758,
"grad_norm": 0.6435029752740142,
"learning_rate": 3.986787242454823e-05,
"loss": 0.271,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14436334371566772,
"step": 385,
"valid_targets_mean": 3619.1,
"valid_targets_min": 1432
},
{
"epoch": 0.9466019417475728,
"grad_norm": 0.6370382737956126,
"learning_rate": 3.985361569126477e-05,
"loss": 0.2674,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12925313413143158,
"step": 390,
"valid_targets_mean": 3150.9,
"valid_targets_min": 1150
},
{
"epoch": 0.9587378640776699,
"grad_norm": 0.6116349728364452,
"learning_rate": 3.9838631507702685e-05,
"loss": 0.2639,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.140370711684227,
"step": 395,
"valid_targets_mean": 3446.0,
"valid_targets_min": 1174
},
{
"epoch": 0.970873786407767,
"grad_norm": 0.6474641948868104,
"learning_rate": 3.982292042289289e-05,
"loss": 0.2712,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1495288759469986,
"step": 400,
"valid_targets_mean": 3898.5,
"valid_targets_min": 1171
},
{
"epoch": 0.9830097087378641,
"grad_norm": 0.9458537674723543,
"learning_rate": 3.980648301250044e-05,
"loss": 0.2647,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12278160452842712,
"step": 405,
"valid_targets_mean": 2943.9,
"valid_targets_min": 1064
},
{
"epoch": 0.9951456310679612,
"grad_norm": 0.6350271715765894,
"learning_rate": 3.978931987880348e-05,
"loss": 0.2718,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15371200442314148,
"step": 410,
"valid_targets_mean": 3921.5,
"valid_targets_min": 1317
},
{
"epoch": 1.0072815533980584,
"grad_norm": 0.5379550038615714,
"learning_rate": 3.977143165067117e-05,
"loss": 0.2551,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11607427150011063,
"step": 415,
"valid_targets_mean": 3544.9,
"valid_targets_min": 1435
},
{
"epoch": 1.0194174757281553,
"grad_norm": 0.7332424499117026,
"learning_rate": 3.97528189835406e-05,
"loss": 0.2575,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11831533908843994,
"step": 420,
"valid_targets_mean": 3280.1,
"valid_targets_min": 1803
},
{
"epoch": 1.0315533980582525,
"grad_norm": 0.5868621643179659,
"learning_rate": 3.973348255939284e-05,
"loss": 0.2384,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13912534713745117,
"step": 425,
"valid_targets_mean": 4546.8,
"valid_targets_min": 1723
},
{
"epoch": 1.0436893203883495,
"grad_norm": 0.6197835822461427,
"learning_rate": 3.971342308672792e-05,
"loss": 0.2408,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11320110410451889,
"step": 430,
"valid_targets_mean": 3838.4,
"valid_targets_min": 1281
},
{
"epoch": 1.0558252427184467,
"grad_norm": 0.7136890568777918,
"learning_rate": 3.969264130053886e-05,
"loss": 0.2496,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14417457580566406,
"step": 435,
"valid_targets_mean": 3970.9,
"valid_targets_min": 1210
},
{
"epoch": 1.0679611650485437,
"grad_norm": 0.7450500002555064,
"learning_rate": 3.967113796228475e-05,
"loss": 0.2568,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14854758977890015,
"step": 440,
"valid_targets_mean": 3500.0,
"valid_targets_min": 1602
},
{
"epoch": 1.0800970873786409,
"grad_norm": 0.5678938077200287,
"learning_rate": 3.9648913859862835e-05,
"loss": 0.2336,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10903868079185486,
"step": 445,
"valid_targets_mean": 4461.2,
"valid_targets_min": 1197
},
{
"epoch": 1.0922330097087378,
"grad_norm": 0.6365732190894927,
"learning_rate": 3.962596980757969e-05,
"loss": 0.2571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10738103836774826,
"step": 450,
"valid_targets_mean": 2641.8,
"valid_targets_min": 1456
},
{
"epoch": 1.104368932038835,
"grad_norm": 0.6065028994817468,
"learning_rate": 3.9602306646121335e-05,
"loss": 0.2473,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14354068040847778,
"step": 455,
"valid_targets_mean": 4470.8,
"valid_targets_min": 2039
},
{
"epoch": 1.116504854368932,
"grad_norm": 0.7081738009243782,
"learning_rate": 3.957792524252245e-05,
"loss": 0.2546,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15602517127990723,
"step": 460,
"valid_targets_mean": 2730.9,
"valid_targets_min": 1141
},
{
"epoch": 1.1286407766990292,
"grad_norm": 0.6375872090776061,
"learning_rate": 3.9552826490134615e-05,
"loss": 0.2344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09666715562343597,
"step": 465,
"valid_targets_mean": 2714.4,
"valid_targets_min": 1930
},
{
"epoch": 1.1407766990291262,
"grad_norm": 0.6944751140755465,
"learning_rate": 3.9527011308593564e-05,
"loss": 0.2311,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.109642893075943,
"step": 470,
"valid_targets_mean": 2836.0,
"valid_targets_min": 1296
},
{
"epoch": 1.1529126213592233,
"grad_norm": 0.8145259627397033,
"learning_rate": 3.9500480643785485e-05,
"loss": 0.2527,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11052658408880234,
"step": 475,
"valid_targets_mean": 2766.4,
"valid_targets_min": 1059
},
{
"epoch": 1.1650485436893203,
"grad_norm": 0.6135951782513367,
"learning_rate": 3.947323546781238e-05,
"loss": 0.249,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15463319420814514,
"step": 480,
"valid_targets_mean": 4576.8,
"valid_targets_min": 1915
},
{
"epoch": 1.1771844660194175,
"grad_norm": 0.6545374176329989,
"learning_rate": 3.9445276778956436e-05,
"loss": 0.2498,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11262653768062592,
"step": 485,
"valid_targets_mean": 3131.9,
"valid_targets_min": 1855
},
{
"epoch": 1.1893203883495145,
"grad_norm": 0.621487816464075,
"learning_rate": 3.941660560164345e-05,
"loss": 0.2486,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1143549233675003,
"step": 490,
"valid_targets_mean": 2956.8,
"valid_targets_min": 1541
},
{
"epoch": 1.2014563106796117,
"grad_norm": 0.6833728236907811,
"learning_rate": 3.9387222986405275e-05,
"loss": 0.2572,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.084357351064682,
"step": 495,
"valid_targets_mean": 3084.0,
"valid_targets_min": 1812
},
{
"epoch": 1.2135922330097086,
"grad_norm": 0.6078117462532137,
"learning_rate": 3.935713000984136e-05,
"loss": 0.2455,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13067331910133362,
"step": 500,
"valid_targets_mean": 3735.4,
"valid_targets_min": 877
},
{
"epoch": 1.2257281553398058,
"grad_norm": 0.6609742832243334,
"learning_rate": 3.93263277745793e-05,
"loss": 0.2557,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11562322080135345,
"step": 505,
"valid_targets_mean": 2854.5,
"valid_targets_min": 1417
},
{
"epoch": 1.237864077669903,
"grad_norm": 0.6415955411027615,
"learning_rate": 3.929481740923438e-05,
"loss": 0.2426,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09550823271274567,
"step": 510,
"valid_targets_mean": 2475.8,
"valid_targets_min": 1020
},
{
"epoch": 1.25,
"grad_norm": 0.643175441795958,
"learning_rate": 3.926260006836831e-05,
"loss": 0.2531,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1402314007282257,
"step": 515,
"valid_targets_mean": 4919.4,
"valid_targets_min": 1414
},
{
"epoch": 1.262135922330097,
"grad_norm": 0.5931837308838739,
"learning_rate": 3.922967693244683e-05,
"loss": 0.2407,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17234575748443604,
"step": 520,
"valid_targets_mean": 5562.9,
"valid_targets_min": 2284
},
{
"epoch": 1.2742718446601942,
"grad_norm": 0.662132650722406,
"learning_rate": 3.9196049207796545e-05,
"loss": 0.2589,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12815213203430176,
"step": 525,
"valid_targets_mean": 3004.4,
"valid_targets_min": 1424
},
{
"epoch": 1.2864077669902914,
"grad_norm": 0.5202391686725419,
"learning_rate": 3.916171812656065e-05,
"loss": 0.2353,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11894293129444122,
"step": 530,
"valid_targets_mean": 5745.9,
"valid_targets_min": 1094
},
{
"epoch": 1.2985436893203883,
"grad_norm": 0.9126442312432708,
"learning_rate": 3.912668494665383e-05,
"loss": 0.2459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15907669067382812,
"step": 535,
"valid_targets_mean": 3221.5,
"valid_targets_min": 1261
},
{
"epoch": 1.3106796116504853,
"grad_norm": 0.5916121776720732,
"learning_rate": 3.9090950951716135e-05,
"loss": 0.2365,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11489912867546082,
"step": 540,
"valid_targets_mean": 3476.2,
"valid_targets_min": 2223
},
{
"epoch": 1.3228155339805825,
"grad_norm": 0.6564628349585573,
"learning_rate": 3.905451745106598e-05,
"loss": 0.2324,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09989440441131592,
"step": 545,
"valid_targets_mean": 2514.9,
"valid_targets_min": 1312
},
{
"epoch": 1.3349514563106797,
"grad_norm": 0.6433511514109358,
"learning_rate": 3.901738577965217e-05,
"loss": 0.2346,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12087219953536987,
"step": 550,
"valid_targets_mean": 3374.9,
"valid_targets_min": 1405
},
{
"epoch": 1.3470873786407767,
"grad_norm": 0.5312994333749906,
"learning_rate": 3.897955729800494e-05,
"loss": 0.2538,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12850449979305267,
"step": 555,
"valid_targets_mean": 4245.9,
"valid_targets_min": 2328
},
{
"epoch": 1.3592233009708738,
"grad_norm": 0.6609285657527223,
"learning_rate": 3.894103339218615e-05,
"loss": 0.2381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10774380713701248,
"step": 560,
"valid_targets_mean": 3120.5,
"valid_targets_min": 1212
},
{
"epoch": 1.3713592233009708,
"grad_norm": 0.6646470913478599,
"learning_rate": 3.890181547373849e-05,
"loss": 0.2347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1032709926366806,
"step": 565,
"valid_targets_mean": 2867.5,
"valid_targets_min": 1950
},
{
"epoch": 1.383495145631068,
"grad_norm": 0.6944474836548723,
"learning_rate": 3.886190497963375e-05,
"loss": 0.2308,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11551807820796967,
"step": 570,
"valid_targets_mean": 3114.9,
"valid_targets_min": 1512
},
{
"epoch": 1.395631067961165,
"grad_norm": 0.5968757161301088,
"learning_rate": 3.882130337222018e-05,
"loss": 0.2433,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14402076601982117,
"step": 575,
"valid_targets_mean": 3520.2,
"valid_targets_min": 507
},
{
"epoch": 1.4077669902912622,
"grad_norm": 0.575928084082454,
"learning_rate": 3.8780012139168886e-05,
"loss": 0.2544,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14746049046516418,
"step": 580,
"valid_targets_mean": 4501.2,
"valid_targets_min": 2242
},
{
"epoch": 1.4199029126213591,
"grad_norm": 0.6571542413220807,
"learning_rate": 3.873803279341936e-05,
"loss": 0.2267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15570998191833496,
"step": 585,
"valid_targets_mean": 3834.4,
"valid_targets_min": 1332
},
{
"epoch": 1.4320388349514563,
"grad_norm": 0.564408924201247,
"learning_rate": 3.8695366873124e-05,
"loss": 0.2539,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1202673390507698,
"step": 590,
"valid_targets_mean": 3465.2,
"valid_targets_min": 1380
},
{
"epoch": 1.4441747572815533,
"grad_norm": 0.5946594447318396,
"learning_rate": 3.865201594159178e-05,
"loss": 0.2491,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10651735216379166,
"step": 595,
"valid_targets_mean": 3241.0,
"valid_targets_min": 1524
},
{
"epoch": 1.4563106796116505,
"grad_norm": 0.5559127875758053,
"learning_rate": 3.860798158723097e-05,
"loss": 0.2481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1417837291955948,
"step": 600,
"valid_targets_mean": 4477.0,
"valid_targets_min": 1146
},
{
"epoch": 1.4684466019417477,
"grad_norm": 0.5541626383906009,
"learning_rate": 3.856326542349092e-05,
"loss": 0.2535,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.108732670545578,
"step": 605,
"valid_targets_mean": 3920.4,
"valid_targets_min": 1690
},
{
"epoch": 1.4805825242718447,
"grad_norm": 0.5674606556742555,
"learning_rate": 3.851786908880295e-05,
"loss": 0.2476,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14006434381008148,
"step": 610,
"valid_targets_mean": 3864.9,
"valid_targets_min": 2580
},
{
"epoch": 1.4927184466019416,
"grad_norm": 0.5549766555611617,
"learning_rate": 3.847179424652031e-05,
"loss": 0.2344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1859729140996933,
"step": 615,
"valid_targets_mean": 5408.5,
"valid_targets_min": 1940
},
{
"epoch": 1.5048543689320388,
"grad_norm": 0.5124904185652867,
"learning_rate": 3.842504258485726e-05,
"loss": 0.2246,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12066035717725754,
"step": 620,
"valid_targets_mean": 4723.6,
"valid_targets_min": 1718
},
{
"epoch": 1.516990291262136,
"grad_norm": 0.5509552950817731,
"learning_rate": 3.837761581682716e-05,
"loss": 0.2411,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09093599021434784,
"step": 625,
"valid_targets_mean": 3244.4,
"valid_targets_min": 1160
},
{
"epoch": 1.529126213592233,
"grad_norm": 0.5993582762966828,
"learning_rate": 3.8329515680179776e-05,
"loss": 0.2487,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1373687982559204,
"step": 630,
"valid_targets_mean": 3561.0,
"valid_targets_min": 679
},
{
"epoch": 1.54126213592233,
"grad_norm": 0.5585412679417914,
"learning_rate": 3.8280743937337534e-05,
"loss": 0.2395,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1317381113767624,
"step": 635,
"valid_targets_mean": 3862.8,
"valid_targets_min": 1182
},
{
"epoch": 1.5533980582524272,
"grad_norm": 0.6733352411768975,
"learning_rate": 3.823130237533099e-05,
"loss": 0.2361,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11223067343235016,
"step": 640,
"valid_targets_mean": 2959.6,
"valid_targets_min": 1662
},
{
"epoch": 1.5655339805825244,
"grad_norm": 0.46144023572813947,
"learning_rate": 3.818119280573336e-05,
"loss": 0.2342,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07549016177654266,
"step": 645,
"valid_targets_mean": 2909.5,
"valid_targets_min": 1426
},
{
"epoch": 1.5776699029126213,
"grad_norm": 0.5324533914053096,
"learning_rate": 3.8130417064594085e-05,
"loss": 0.2359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15126053988933563,
"step": 650,
"valid_targets_mean": 6172.4,
"valid_targets_min": 1770
},
{
"epoch": 1.5898058252427183,
"grad_norm": 0.6195468936105147,
"learning_rate": 3.807897701237162e-05,
"loss": 0.2474,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15350914001464844,
"step": 655,
"valid_targets_mean": 4238.2,
"valid_targets_min": 1940
},
{
"epoch": 1.6019417475728155,
"grad_norm": 0.76454679368284,
"learning_rate": 3.802687453386523e-05,
"loss": 0.2344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11093638092279434,
"step": 660,
"valid_targets_mean": 2401.6,
"valid_targets_min": 836
},
{
"epoch": 1.6140776699029127,
"grad_norm": 0.5400951747141338,
"learning_rate": 3.797411153814595e-05,
"loss": 0.2555,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12915802001953125,
"step": 665,
"valid_targets_mean": 4380.1,
"valid_targets_min": 2515
},
{
"epoch": 1.6262135922330097,
"grad_norm": 0.5278459050868829,
"learning_rate": 3.79206899584866e-05,
"loss": 0.2342,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09788914024829865,
"step": 670,
"valid_targets_mean": 3473.1,
"valid_targets_min": 1474
},
{
"epoch": 1.6383495145631068,
"grad_norm": 0.6348044871165122,
"learning_rate": 3.7866611752291016e-05,
"loss": 0.2439,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12198871374130249,
"step": 675,
"valid_targets_mean": 3115.9,
"valid_targets_min": 1225
},
{
"epoch": 1.650485436893204,
"grad_norm": 0.59219872937863,
"learning_rate": 3.7811878901022234e-05,
"loss": 0.2339,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10915888100862503,
"step": 680,
"valid_targets_mean": 3712.9,
"valid_targets_min": 1935
},
{
"epoch": 1.662621359223301,
"grad_norm": 0.6904197037633991,
"learning_rate": 3.775649341012999e-05,
"loss": 0.2521,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13143035769462585,
"step": 685,
"valid_targets_mean": 2945.6,
"valid_targets_min": 723
},
{
"epoch": 1.674757281553398,
"grad_norm": 0.5436113134969518,
"learning_rate": 3.770045730897716e-05,
"loss": 0.2354,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11741673201322556,
"step": 690,
"valid_targets_mean": 3804.9,
"valid_targets_min": 1681
},
{
"epoch": 1.6868932038834952,
"grad_norm": 0.6063591912564006,
"learning_rate": 3.7643772650765436e-05,
"loss": 0.2274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09935449063777924,
"step": 695,
"valid_targets_mean": 3038.4,
"valid_targets_min": 1979
},
{
"epoch": 1.6990291262135924,
"grad_norm": 0.5345891774927867,
"learning_rate": 3.75864415124601e-05,
"loss": 0.2352,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10488158464431763,
"step": 700,
"valid_targets_mean": 3789.5,
"valid_targets_min": 2112
},
{
"epoch": 1.7111650485436893,
"grad_norm": 0.4676880706242841,
"learning_rate": 3.75284659947139e-05,
"loss": 0.227,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15096917748451233,
"step": 705,
"valid_targets_mean": 6318.2,
"valid_targets_min": 1627
},
{
"epoch": 1.7233009708737863,
"grad_norm": 0.6571534969047385,
"learning_rate": 3.746984822179011e-05,
"loss": 0.2359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11676843464374542,
"step": 710,
"valid_targets_mean": 3594.9,
"valid_targets_min": 1108
},
{
"epoch": 1.7354368932038835,
"grad_norm": 0.613009025693255,
"learning_rate": 3.741059034148468e-05,
"loss": 0.2434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13570618629455566,
"step": 715,
"valid_targets_mean": 3184.4,
"valid_targets_min": 1887
},
{
"epoch": 1.7475728155339807,
"grad_norm": 0.5555217845339221,
"learning_rate": 3.735069452504751e-05,
"loss": 0.2266,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07074729353189468,
"step": 720,
"valid_targets_mean": 2201.2,
"valid_targets_min": 1084
},
{
"epoch": 1.7597087378640777,
"grad_norm": 0.5237513453569695,
"learning_rate": 3.729016296710298e-05,
"loss": 0.2229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09680597484111786,
"step": 725,
"valid_targets_mean": 4383.0,
"valid_targets_min": 1303
},
{
"epoch": 1.7718446601941746,
"grad_norm": 0.6298115302030642,
"learning_rate": 3.722899788556942e-05,
"loss": 0.2357,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12231288850307465,
"step": 730,
"valid_targets_mean": 3147.4,
"valid_targets_min": 1594
},
{
"epoch": 1.7839805825242718,
"grad_norm": 0.6587050708268605,
"learning_rate": 3.7167201521577934e-05,
"loss": 0.2389,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09437572211027145,
"step": 735,
"valid_targets_mean": 2561.0,
"valid_targets_min": 1557
},
{
"epoch": 1.796116504854369,
"grad_norm": 0.6653530316509757,
"learning_rate": 3.710477613939026e-05,
"loss": 0.2435,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12413933873176575,
"step": 740,
"valid_targets_mean": 3271.1,
"valid_targets_min": 1299
},
{
"epoch": 1.808252427184466,
"grad_norm": 0.5886871648583835,
"learning_rate": 3.7041724026315796e-05,
"loss": 0.2298,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13463392853736877,
"step": 745,
"valid_targets_mean": 4542.9,
"valid_targets_min": 1486
},
{
"epoch": 1.820388349514563,
"grad_norm": 0.659491946253881,
"learning_rate": 3.697804749262778e-05,
"loss": 0.2297,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12109142541885376,
"step": 750,
"valid_targets_mean": 3197.5,
"valid_targets_min": 2294
},
{
"epoch": 1.8325242718446602,
"grad_norm": 0.5496460281209354,
"learning_rate": 3.6913748871478674e-05,
"loss": 0.2314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1468179076910019,
"step": 755,
"valid_targets_mean": 5731.5,
"valid_targets_min": 2318
},
{
"epoch": 1.8446601941747574,
"grad_norm": 0.5923558134337653,
"learning_rate": 3.684883051881463e-05,
"loss": 0.2338,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1461079716682434,
"step": 760,
"valid_targets_mean": 4946.4,
"valid_targets_min": 1328
},
{
"epoch": 1.8567961165048543,
"grad_norm": 0.5834483166458697,
"learning_rate": 3.678329481328924e-05,
"loss": 0.2406,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09497156739234924,
"step": 765,
"valid_targets_mean": 3247.4,
"valid_targets_min": 1267
},
{
"epoch": 1.8689320388349513,
"grad_norm": 0.6617719598355142,
"learning_rate": 3.671714415617629e-05,
"loss": 0.2192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13364149630069733,
"step": 770,
"valid_targets_mean": 3441.9,
"valid_targets_min": 1194
},
{
"epoch": 1.8810679611650487,
"grad_norm": 0.5795659094415618,
"learning_rate": 3.665038097128185e-05,
"loss": 0.2358,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11587332934141159,
"step": 775,
"valid_targets_mean": 3492.6,
"valid_targets_min": 1090
},
{
"epoch": 1.8932038834951457,
"grad_norm": 0.5668623418672875,
"learning_rate": 3.6583007704855395e-05,
"loss": 0.2336,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14234060049057007,
"step": 780,
"valid_targets_mean": 4155.0,
"valid_targets_min": 2346
},
{
"epoch": 1.9053398058252426,
"grad_norm": 0.6190538984516589,
"learning_rate": 3.651502682550026e-05,
"loss": 0.2389,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10980303585529327,
"step": 785,
"valid_targets_mean": 3097.0,
"valid_targets_min": 1472
},
{
"epoch": 1.9174757281553398,
"grad_norm": 0.6615820988326352,
"learning_rate": 3.6446440824083105e-05,
"loss": 0.2399,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1562778502702713,
"step": 790,
"valid_targets_mean": 3828.0,
"valid_targets_min": 1250
},
{
"epoch": 1.929611650485437,
"grad_norm": 0.5779506637324582,
"learning_rate": 3.637725221364269e-05,
"loss": 0.2266,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14356032013893127,
"step": 795,
"valid_targets_mean": 4774.2,
"valid_targets_min": 1354
},
{
"epoch": 1.941747572815534,
"grad_norm": 0.6231523563187834,
"learning_rate": 3.6307463529297814e-05,
"loss": 0.2385,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12037024646997452,
"step": 800,
"valid_targets_mean": 2992.2,
"valid_targets_min": 971
},
{
"epoch": 1.953883495145631,
"grad_norm": 0.591718001654194,
"learning_rate": 3.623707732815435e-05,
"loss": 0.2335,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11665384471416473,
"step": 805,
"valid_targets_mean": 3620.1,
"valid_targets_min": 2003
},
{
"epoch": 1.9660194174757282,
"grad_norm": 0.549811911272467,
"learning_rate": 3.616609618921165e-05,
"loss": 0.2397,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13667896389961243,
"step": 810,
"valid_targets_mean": 4200.8,
"valid_targets_min": 1549
},
{
"epoch": 1.9781553398058254,
"grad_norm": 0.7347899806379934,
"learning_rate": 3.609452271326798e-05,
"loss": 0.2301,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11621441692113876,
"step": 815,
"valid_targets_mean": 2682.1,
"valid_targets_min": 1270
},
{
"epoch": 1.9902912621359223,
"grad_norm": 0.5922313661938818,
"learning_rate": 3.602235952282524e-05,
"loss": 0.2352,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09103728830814362,
"step": 820,
"valid_targets_mean": 3236.0,
"valid_targets_min": 1540
},
{
"epoch": 2.0024271844660193,
"grad_norm": 0.8265974740601438,
"learning_rate": 3.59496092619929e-05,
"loss": 0.2349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.071834996342659,
"step": 825,
"valid_targets_mean": 2064.2,
"valid_targets_min": 689
},
{
"epoch": 2.0145631067961167,
"grad_norm": 0.669371768067199,
"learning_rate": 3.587627459639106e-05,
"loss": 0.205,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09449409693479538,
"step": 830,
"valid_targets_mean": 3118.1,
"valid_targets_min": 1435
},
{
"epoch": 2.0266990291262137,
"grad_norm": 0.5992705928707142,
"learning_rate": 3.5802358213052876e-05,
"loss": 0.2073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08361292630434036,
"step": 835,
"valid_targets_mean": 3604.1,
"valid_targets_min": 1693
},
{
"epoch": 2.0388349514563107,
"grad_norm": 0.5725959797963897,
"learning_rate": 3.5727862820325975e-05,
"loss": 0.2056,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1300264596939087,
"step": 840,
"valid_targets_mean": 4659.9,
"valid_targets_min": 1188
},
{
"epoch": 2.0509708737864076,
"grad_norm": 0.6110376030142648,
"learning_rate": 3.565279114777334e-05,
"loss": 0.2203,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11453216522932053,
"step": 845,
"valid_targets_mean": 3787.2,
"valid_targets_min": 1199
},
{
"epoch": 2.063106796116505,
"grad_norm": 0.6595318854214649,
"learning_rate": 3.557714594607324e-05,
"loss": 0.2003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1011197566986084,
"step": 850,
"valid_targets_mean": 2687.2,
"valid_targets_min": 1645
},
{
"epoch": 2.075242718446602,
"grad_norm": 0.7396083436756492,
"learning_rate": 3.550092998691844e-05,
"loss": 0.2138,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09144885838031769,
"step": 855,
"valid_targets_mean": 3008.8,
"valid_targets_min": 1667
},
{
"epoch": 2.087378640776699,
"grad_norm": 0.6556055829897129,
"learning_rate": 3.5424146062914664e-05,
"loss": 0.2125,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10240662097930908,
"step": 860,
"valid_targets_mean": 3578.4,
"valid_targets_min": 1919
},
{
"epoch": 2.099514563106796,
"grad_norm": 0.6126429094535576,
"learning_rate": 3.534679698747824e-05,
"loss": 0.208,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09693218767642975,
"step": 865,
"valid_targets_mean": 3090.9,
"valid_targets_min": 1587
},
{
"epoch": 2.1116504854368934,
"grad_norm": 0.62362313853087,
"learning_rate": 3.5268885594733074e-05,
"loss": 0.2074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08325746655464172,
"step": 870,
"valid_targets_mean": 2677.8,
"valid_targets_min": 1686
},
{
"epoch": 2.1237864077669903,
"grad_norm": 0.6259527154133824,
"learning_rate": 3.519041473940674e-05,
"loss": 0.2114,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09583015739917755,
"step": 875,
"valid_targets_mean": 3664.4,
"valid_targets_min": 1969
},
{
"epoch": 2.1359223300970873,
"grad_norm": 0.6478641168572616,
"learning_rate": 3.511138729672592e-05,
"loss": 0.2024,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06377410888671875,
"step": 880,
"valid_targets_mean": 1970.8,
"valid_targets_min": 1108
},
{
"epoch": 2.1480582524271843,
"grad_norm": 0.6878182220032585,
"learning_rate": 3.503180616231105e-05,
"loss": 0.2083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09734027087688446,
"step": 885,
"valid_targets_mean": 3070.2,
"valid_targets_min": 1148
},
{
"epoch": 2.1601941747572817,
"grad_norm": 0.6782442739400362,
"learning_rate": 3.495167425207021e-05,
"loss": 0.2238,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1031256914138794,
"step": 890,
"valid_targets_mean": 3148.2,
"valid_targets_min": 1690
},
{
"epoch": 2.1723300970873787,
"grad_norm": 0.555148254185791,
"learning_rate": 3.487099450209228e-05,
"loss": 0.2234,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15578101575374603,
"step": 895,
"valid_targets_mean": 5298.6,
"valid_targets_min": 1538
},
{
"epoch": 2.1844660194174756,
"grad_norm": 0.8139654856060342,
"learning_rate": 3.478976986853939e-05,
"loss": 0.2125,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12132055312395096,
"step": 900,
"valid_targets_mean": 4041.8,
"valid_targets_min": 1883
},
{
"epoch": 2.1966019417475726,
"grad_norm": 0.6729341475316266,
"learning_rate": 3.4708003327538556e-05,
"loss": 0.2129,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1465618759393692,
"step": 905,
"valid_targets_mean": 3630.5,
"valid_targets_min": 1731
},
{
"epoch": 2.20873786407767,
"grad_norm": 0.6556753176427049,
"learning_rate": 3.462569787507268e-05,
"loss": 0.1996,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10640613734722137,
"step": 910,
"valid_targets_mean": 4469.9,
"valid_targets_min": 2755
},
{
"epoch": 2.220873786407767,
"grad_norm": 0.5999956305203342,
"learning_rate": 3.454285652687075e-05,
"loss": 0.2132,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09524613618850708,
"step": 915,
"valid_targets_mean": 3620.0,
"valid_targets_min": 1665
},
{
"epoch": 2.233009708737864,
"grad_norm": 0.589337519898445,
"learning_rate": 3.445948231829735e-05,
"loss": 0.2088,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08690285682678223,
"step": 920,
"valid_targets_mean": 2898.5,
"valid_targets_min": 722
},
{
"epoch": 2.2451456310679614,
"grad_norm": 0.6808562269069073,
"learning_rate": 3.4375578304241424e-05,
"loss": 0.2032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13871456682682037,
"step": 925,
"valid_targets_mean": 4055.0,
"valid_targets_min": 1436
},
{
"epoch": 2.2572815533980584,
"grad_norm": 0.5917088371530682,
"learning_rate": 3.429114755900437e-05,
"loss": 0.2013,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08324424922466278,
"step": 930,
"valid_targets_mean": 3107.5,
"valid_targets_min": 1389
},
{
"epoch": 2.2694174757281553,
"grad_norm": 0.7186628368159883,
"learning_rate": 3.420619317618737e-05,
"loss": 0.2131,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10876597464084625,
"step": 935,
"valid_targets_mean": 2816.9,
"valid_targets_min": 1442
},
{
"epoch": 2.2815533980582523,
"grad_norm": 0.5851678494613404,
"learning_rate": 3.412071826857808e-05,
"loss": 0.2031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1166793555021286,
"step": 940,
"valid_targets_mean": 4515.9,
"valid_targets_min": 2239
},
{
"epoch": 2.2936893203883493,
"grad_norm": 0.6313278564162049,
"learning_rate": 3.403472596803651e-05,
"loss": 0.2023,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13072362542152405,
"step": 945,
"valid_targets_mean": 4345.6,
"valid_targets_min": 1304
},
{
"epoch": 2.3058252427184467,
"grad_norm": 0.6754081524719817,
"learning_rate": 3.3948219425380326e-05,
"loss": 0.2096,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15190935134887695,
"step": 950,
"valid_targets_mean": 4081.5,
"valid_targets_min": 1785
},
{
"epoch": 2.3179611650485437,
"grad_norm": 0.7282378039686259,
"learning_rate": 3.3861201810269376e-05,
"loss": 0.2058,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10873965173959732,
"step": 955,
"valid_targets_mean": 2639.2,
"valid_targets_min": 1072
},
{
"epoch": 2.3300970873786406,
"grad_norm": 0.595680867160716,
"learning_rate": 3.377367631108957e-05,
"loss": 0.222,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10408247262239456,
"step": 960,
"valid_targets_mean": 3947.5,
"valid_targets_min": 1285
},
{
"epoch": 2.342233009708738,
"grad_norm": 0.5804568018879114,
"learning_rate": 3.368564613483604e-05,
"loss": 0.2206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09195692837238312,
"step": 965,
"valid_targets_mean": 3713.5,
"valid_targets_min": 2418
},
{
"epoch": 2.354368932038835,
"grad_norm": 0.550283188819811,
"learning_rate": 3.3597114506995605e-05,
"loss": 0.209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1009785532951355,
"step": 970,
"valid_targets_mean": 4098.9,
"valid_targets_min": 1040
},
{
"epoch": 2.366504854368932,
"grad_norm": 0.6838989758732019,
"learning_rate": 3.350808467142867e-05,
"loss": 0.2074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.089122474193573,
"step": 975,
"valid_targets_mean": 2784.0,
"valid_targets_min": 1395
},
{
"epoch": 2.378640776699029,
"grad_norm": 0.5245250979248586,
"learning_rate": 3.341855989025028e-05,
"loss": 0.2046,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07528170943260193,
"step": 980,
"valid_targets_mean": 2959.4,
"valid_targets_min": 2285
},
{
"epoch": 2.3907766990291264,
"grad_norm": 0.5825786649927605,
"learning_rate": 3.332854344371065e-05,
"loss": 0.1994,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.061905454844236374,
"step": 985,
"valid_targets_mean": 2363.6,
"valid_targets_min": 1432
},
{
"epoch": 2.4029126213592233,
"grad_norm": 0.6400360884778853,
"learning_rate": 3.3238038630074944e-05,
"loss": 0.2024,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09072259068489075,
"step": 990,
"valid_targets_mean": 2682.4,
"valid_targets_min": 1405
},
{
"epoch": 2.4150485436893203,
"grad_norm": 0.5157231572500409,
"learning_rate": 3.314704876550244e-05,
"loss": 0.2033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12075865268707275,
"step": 995,
"valid_targets_mean": 5359.8,
"valid_targets_min": 2334
},
{
"epoch": 2.4271844660194173,
"grad_norm": 0.6456780093180492,
"learning_rate": 3.305557718392503e-05,
"loss": 0.2047,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07853558659553528,
"step": 1000,
"valid_targets_mean": 3140.8,
"valid_targets_min": 1964
},
{
"epoch": 2.4393203883495147,
"grad_norm": 0.6815848527399921,
"learning_rate": 3.296362723692503e-05,
"loss": 0.2102,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08463180065155029,
"step": 1005,
"valid_targets_mean": 2608.5,
"valid_targets_min": 1179
},
{
"epoch": 2.4514563106796117,
"grad_norm": 0.696437382100364,
"learning_rate": 3.287120229361242e-05,
"loss": 0.2115,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1109643429517746,
"step": 1010,
"valid_targets_mean": 2854.5,
"valid_targets_min": 1626
},
{
"epoch": 2.4635922330097086,
"grad_norm": 0.5899964966996672,
"learning_rate": 3.2778305740501354e-05,
"loss": 0.2005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0780857652425766,
"step": 1015,
"valid_targets_mean": 2638.8,
"valid_targets_min": 1150
},
{
"epoch": 2.475728155339806,
"grad_norm": 0.6022235573558729,
"learning_rate": 3.268494098138612e-05,
"loss": 0.2142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1379772126674652,
"step": 1020,
"valid_targets_mean": 4332.5,
"valid_targets_min": 2061
},
{
"epoch": 2.487864077669903,
"grad_norm": 0.6197416033254599,
"learning_rate": 3.259111143721638e-05,
"loss": 0.2013,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11153020709753036,
"step": 1025,
"valid_targets_mean": 3978.9,
"valid_targets_min": 1803
},
{
"epoch": 2.5,
"grad_norm": 0.5676417474896228,
"learning_rate": 3.2496820545971856e-05,
"loss": 0.1957,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15110592544078827,
"step": 1030,
"valid_targets_mean": 5153.1,
"valid_targets_min": 1307
},
{
"epoch": 2.512135922330097,
"grad_norm": 0.5646670369137284,
"learning_rate": 3.2402071762536335e-05,
"loss": 0.1958,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09453196823596954,
"step": 1035,
"valid_targets_mean": 3770.9,
"valid_targets_min": 1932
},
{
"epoch": 2.524271844660194,
"grad_norm": 0.4800658121358941,
"learning_rate": 3.23068685585711e-05,
"loss": 0.2061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0763217955827713,
"step": 1040,
"valid_targets_mean": 4163.5,
"valid_targets_min": 1441
},
{
"epoch": 2.5364077669902914,
"grad_norm": 0.6090481605500978,
"learning_rate": 3.2211214422387723e-05,
"loss": 0.2075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11317304521799088,
"step": 1045,
"valid_targets_mean": 4364.5,
"valid_targets_min": 1087
},
{
"epoch": 2.5485436893203883,
"grad_norm": 0.601397158926294,
"learning_rate": 3.211511285882024e-05,
"loss": 0.2141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12892407178878784,
"step": 1050,
"valid_targets_mean": 3863.4,
"valid_targets_min": 2405
},
{
"epoch": 2.5606796116504853,
"grad_norm": 0.6090221774074198,
"learning_rate": 3.201856738909673e-05,
"loss": 0.2107,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16123273968696594,
"step": 1055,
"valid_targets_mean": 4166.6,
"valid_targets_min": 1949
},
{
"epoch": 2.5728155339805827,
"grad_norm": 0.6965919936635202,
"learning_rate": 3.192158155071032e-05,
"loss": 0.2093,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12045931816101074,
"step": 1060,
"valid_targets_mean": 3225.8,
"valid_targets_min": 1613
},
{
"epoch": 2.5849514563106797,
"grad_norm": 0.6018373187578538,
"learning_rate": 3.182415889728953e-05,
"loss": 0.2095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08410439640283585,
"step": 1065,
"valid_targets_mean": 3278.5,
"valid_targets_min": 1251
},
{
"epoch": 2.5970873786407767,
"grad_norm": 0.5369035092697442,
"learning_rate": 3.172630299846811e-05,
"loss": 0.2009,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09620429575443268,
"step": 1070,
"valid_targets_mean": 3479.0,
"valid_targets_min": 893
},
{
"epoch": 2.6092233009708736,
"grad_norm": 0.6937192957864611,
"learning_rate": 3.162801743975419e-05,
"loss": 0.2029,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.102183498442173,
"step": 1075,
"valid_targets_mean": 3027.8,
"valid_targets_min": 1254
},
{
"epoch": 2.6213592233009706,
"grad_norm": 0.699252941121941,
"learning_rate": 3.1529305822398976e-05,
"loss": 0.2069,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12127447128295898,
"step": 1080,
"valid_targets_mean": 3735.1,
"valid_targets_min": 1169
},
{
"epoch": 2.633495145631068,
"grad_norm": 0.6136531232396643,
"learning_rate": 3.143017176326473e-05,
"loss": 0.2088,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12307693809270859,
"step": 1085,
"valid_targets_mean": 4301.5,
"valid_targets_min": 2832
},
{
"epoch": 2.645631067961165,
"grad_norm": 0.5928021303397657,
"learning_rate": 3.133061889469229e-05,
"loss": 0.1874,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09699161350727081,
"step": 1090,
"valid_targets_mean": 4019.6,
"valid_targets_min": 1667
},
{
"epoch": 2.657766990291262,
"grad_norm": 0.45261747080420606,
"learning_rate": 3.123065086436795e-05,
"loss": 0.2097,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0772111788392067,
"step": 1095,
"valid_targets_mean": 4814.9,
"valid_targets_min": 2769
},
{
"epoch": 2.6699029126213594,
"grad_norm": 0.5895465036953527,
"learning_rate": 3.113027133518983e-05,
"loss": 0.2058,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12790998816490173,
"step": 1100,
"valid_targets_mean": 5211.4,
"valid_targets_min": 1431
},
{
"epoch": 2.6820388349514563,
"grad_norm": 0.6783253403819054,
"learning_rate": 3.102948398513365e-05,
"loss": 0.2026,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09133732318878174,
"step": 1105,
"valid_targets_mean": 2532.2,
"valid_targets_min": 968
},
{
"epoch": 2.6941747572815533,
"grad_norm": 0.7702823241622098,
"learning_rate": 3.092829250711797e-05,
"loss": 0.2116,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09099040180444717,
"step": 1110,
"valid_targets_mean": 2986.5,
"valid_targets_min": 1376
},
{
"epoch": 2.7063106796116507,
"grad_norm": 0.6236690656497922,
"learning_rate": 3.082670060886888e-05,
"loss": 0.2133,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11725375801324844,
"step": 1115,
"valid_targets_mean": 3339.4,
"valid_targets_min": 1897
},
{
"epoch": 2.7184466019417477,
"grad_norm": 0.5568614042905151,
"learning_rate": 3.0724712012784147e-05,
"loss": 0.2075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11161457747220993,
"step": 1120,
"valid_targets_mean": 4007.6,
"valid_targets_min": 1657
},
{
"epoch": 2.7305825242718447,
"grad_norm": 0.5182856611185835,
"learning_rate": 3.062233045579681e-05,
"loss": 0.2112,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07638495415449142,
"step": 1125,
"valid_targets_mean": 3582.0,
"valid_targets_min": 1918
},
{
"epoch": 2.7427184466019416,
"grad_norm": 0.5599101394403653,
"learning_rate": 3.0519559689238305e-05,
"loss": 0.1989,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07720982283353806,
"step": 1130,
"valid_targets_mean": 2618.5,
"valid_targets_min": 775
},
{
"epoch": 2.7548543689320386,
"grad_norm": 0.5316704528665577,
"learning_rate": 3.041640347870092e-05,
"loss": 0.1965,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08207693696022034,
"step": 1135,
"valid_targets_mean": 2982.9,
"valid_targets_min": 1645
},
{
"epoch": 2.766990291262136,
"grad_norm": 0.5651276016168141,
"learning_rate": 3.0312865603899943e-05,
"loss": 0.2147,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08095473051071167,
"step": 1140,
"valid_targets_mean": 3067.9,
"valid_targets_min": 724
},
{
"epoch": 2.779126213592233,
"grad_norm": 0.6300326882361118,
"learning_rate": 3.020894985853504e-05,
"loss": 0.2063,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10498537123203278,
"step": 1145,
"valid_targets_mean": 3194.8,
"valid_targets_min": 1961
},
{
"epoch": 2.79126213592233,
"grad_norm": 0.544868186505999,
"learning_rate": 3.0104660050151375e-05,
"loss": 0.1956,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13778439164161682,
"step": 1150,
"valid_targets_mean": 4388.9,
"valid_targets_min": 1048
},
{
"epoch": 2.8033980582524274,
"grad_norm": 0.5875875350547773,
"learning_rate": 3.0000000000000004e-05,
"loss": 0.2085,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08588963747024536,
"step": 1155,
"valid_targets_mean": 3546.6,
"valid_targets_min": 679
},
{
"epoch": 2.8155339805825244,
"grad_norm": 0.5695878831968436,
"learning_rate": 2.989497354289791e-05,
"loss": 0.2103,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12203121185302734,
"step": 1160,
"valid_targets_mean": 3824.6,
"valid_targets_min": 1620
},
{
"epoch": 2.8276699029126213,
"grad_norm": 0.6013250631763066,
"learning_rate": 2.9789584527087488e-05,
"loss": 0.2094,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08212064206600189,
"step": 1165,
"valid_targets_mean": 2810.9,
"valid_targets_min": 1866
},
{
"epoch": 2.8398058252427183,
"grad_norm": 0.597609840351872,
"learning_rate": 2.9683836814095533e-05,
"loss": 0.2054,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13834813237190247,
"step": 1170,
"valid_targets_mean": 4261.8,
"valid_targets_min": 1395
},
{
"epoch": 2.8519417475728153,
"grad_norm": 0.6114713624604186,
"learning_rate": 2.957773427859174e-05,
"loss": 0.2086,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10041432082653046,
"step": 1175,
"valid_targets_mean": 3442.9,
"valid_targets_min": 2231
},
{
"epoch": 2.8640776699029127,
"grad_norm": 0.5667265076909537,
"learning_rate": 2.947128080824677e-05,
"loss": 0.2151,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1092417985200882,
"step": 1180,
"valid_targets_mean": 4111.5,
"valid_targets_min": 1759
},
{
"epoch": 2.8762135922330097,
"grad_norm": 0.597339005981577,
"learning_rate": 2.9364480303589763e-05,
"loss": 0.2163,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08976036310195923,
"step": 1185,
"valid_targets_mean": 2977.1,
"valid_targets_min": 1962
},
{
"epoch": 2.8883495145631066,
"grad_norm": 0.6266567960967807,
"learning_rate": 2.9257336677865442e-05,
"loss": 0.2053,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09315776824951172,
"step": 1190,
"valid_targets_mean": 3013.2,
"valid_targets_min": 1181
},
{
"epoch": 2.900485436893204,
"grad_norm": 0.5900254054109899,
"learning_rate": 2.9149853856890732e-05,
"loss": 0.2092,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11730660498142242,
"step": 1195,
"valid_targets_mean": 3533.0,
"valid_targets_min": 1059
},
{
"epoch": 2.912621359223301,
"grad_norm": 0.5241497225875399,
"learning_rate": 2.9042035778910898e-05,
"loss": 0.2058,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10431572794914246,
"step": 1200,
"valid_targets_mean": 4257.8,
"valid_targets_min": 1484
},
{
"epoch": 2.924757281553398,
"grad_norm": 0.5585633108222432,
"learning_rate": 2.893388639445526e-05,
"loss": 0.1996,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08963444828987122,
"step": 1205,
"valid_targets_mean": 3510.6,
"valid_targets_min": 1991
},
{
"epoch": 2.9368932038834954,
"grad_norm": 0.5171469901256968,
"learning_rate": 2.882540966619243e-05,
"loss": 0.2115,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08788655698299408,
"step": 1210,
"valid_targets_mean": 3734.1,
"valid_targets_min": 1541
},
{
"epoch": 2.9490291262135924,
"grad_norm": 0.5428745451114342,
"learning_rate": 2.8716609568785138e-05,
"loss": 0.2123,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11415513604879379,
"step": 1215,
"valid_targets_mean": 4197.1,
"valid_targets_min": 1783
},
{
"epoch": 2.9611650485436893,
"grad_norm": 0.6268445631115539,
"learning_rate": 2.8607490088744587e-05,
"loss": 0.2051,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13960544764995575,
"step": 1220,
"valid_targets_mean": 4845.2,
"valid_targets_min": 1972
},
{
"epoch": 2.9733009708737863,
"grad_norm": 0.5534380459387299,
"learning_rate": 2.8498055224284347e-05,
"loss": 0.2103,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07392726838588715,
"step": 1225,
"valid_targets_mean": 2749.5,
"valid_targets_min": 1308
},
{
"epoch": 2.9854368932038833,
"grad_norm": 0.6140556781656503,
"learning_rate": 2.8388308985173934e-05,
"loss": 0.1962,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1277606338262558,
"step": 1230,
"valid_targets_mean": 3432.5,
"valid_targets_min": 1874
},
{
"epoch": 2.9975728155339807,
"grad_norm": 0.5273204588114058,
"learning_rate": 2.8278255392591826e-05,
"loss": 0.2028,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11101394891738892,
"step": 1235,
"valid_targets_mean": 4800.6,
"valid_targets_min": 2455
},
{
"epoch": 3.0097087378640777,
"grad_norm": 0.5913147521191209,
"learning_rate": 2.8167898478978154e-05,
"loss": 0.1861,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1095903143286705,
"step": 1240,
"valid_targets_mean": 4048.1,
"valid_targets_min": 1932
},
{
"epoch": 3.0218446601941746,
"grad_norm": 0.7302288780186236,
"learning_rate": 2.8057242287886927e-05,
"loss": 0.1798,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09420832991600037,
"step": 1245,
"valid_targets_mean": 3082.8,
"valid_targets_min": 1157
},
{
"epoch": 3.033980582524272,
"grad_norm": 0.598469535096793,
"learning_rate": 2.7946290873837914e-05,
"loss": 0.1853,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07562121003866196,
"step": 1250,
"valid_targets_mean": 3080.2,
"valid_targets_min": 1717
},
{
"epoch": 3.046116504854369,
"grad_norm": 0.5862293560940997,
"learning_rate": 2.7835048302168038e-05,
"loss": 0.1816,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0946711003780365,
"step": 1255,
"valid_targets_mean": 3453.5,
"valid_targets_min": 1669
},
{
"epoch": 3.058252427184466,
"grad_norm": 0.6916347999805793,
"learning_rate": 2.772351864888245e-05,
"loss": 0.1815,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07562961429357529,
"step": 1260,
"valid_targets_mean": 2563.5,
"valid_targets_min": 1150
},
{
"epoch": 3.070388349514563,
"grad_norm": 0.6575638628931207,
"learning_rate": 2.761170600050517e-05,
"loss": 0.1796,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07140407711267471,
"step": 1265,
"valid_targets_mean": 2628.6,
"valid_targets_min": 1261
},
{
"epoch": 3.0825242718446604,
"grad_norm": 0.6445143910542973,
"learning_rate": 2.749961445392934e-05,
"loss": 0.176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06978870928287506,
"step": 1270,
"valid_targets_mean": 2225.8,
"valid_targets_min": 1477
},
{
"epoch": 3.0946601941747574,
"grad_norm": 0.6169827534821816,
"learning_rate": 2.738724811626715e-05,
"loss": 0.1871,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09161722660064697,
"step": 1275,
"valid_targets_mean": 4036.2,
"valid_targets_min": 199
},
{
"epoch": 3.1067961165048543,
"grad_norm": 0.691440862638205,
"learning_rate": 2.727461110469931e-05,
"loss": 0.1859,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08448947966098785,
"step": 1280,
"valid_targets_mean": 3493.6,
"valid_targets_min": 2157
},
{
"epoch": 3.1189320388349513,
"grad_norm": 0.6804296655092491,
"learning_rate": 2.716170754632422e-05,
"loss": 0.1824,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09447147697210312,
"step": 1285,
"valid_targets_mean": 2978.0,
"valid_targets_min": 1024
},
{
"epoch": 3.1310679611650487,
"grad_norm": 0.5347768965809608,
"learning_rate": 2.7048541578006724e-05,
"loss": 0.1801,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07182887196540833,
"step": 1290,
"valid_targets_mean": 3241.4,
"valid_targets_min": 1211
},
{
"epoch": 3.1432038834951457,
"grad_norm": 2.450357718388632,
"learning_rate": 2.6935117346226573e-05,
"loss": 0.1855,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07918355613946915,
"step": 1295,
"valid_targets_mean": 2544.1,
"valid_targets_min": 1506
},
{
"epoch": 3.1553398058252426,
"grad_norm": 0.623432599085737,
"learning_rate": 2.682143900692645e-05,
"loss": 0.1799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09146401286125183,
"step": 1300,
"valid_targets_mean": 3998.4,
"valid_targets_min": 1097
},
{
"epoch": 3.1674757281553396,
"grad_norm": 0.5909345760687255,
"learning_rate": 2.6707510725359712e-05,
"loss": 0.1655,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09626485407352448,
"step": 1305,
"valid_targets_mean": 4271.4,
"valid_targets_min": 1740
},
{
"epoch": 3.179611650485437,
"grad_norm": 0.5260309917010774,
"learning_rate": 2.6593336675937808e-05,
"loss": 0.1761,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0687076523900032,
"step": 1310,
"valid_targets_mean": 3186.8,
"valid_targets_min": 1394
},
{
"epoch": 3.191747572815534,
"grad_norm": 0.5687769408518127,
"learning_rate": 2.647892104207725e-05,
"loss": 0.1842,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0814671590924263,
"step": 1315,
"valid_targets_mean": 3339.4,
"valid_targets_min": 1674
},
{
"epoch": 3.203883495145631,
"grad_norm": 0.5749512417220419,
"learning_rate": 2.6364268016046396e-05,
"loss": 0.1811,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09464827179908752,
"step": 1320,
"valid_targets_mean": 4326.9,
"valid_targets_min": 2030
},
{
"epoch": 3.216019417475728,
"grad_norm": 0.6001248567918428,
"learning_rate": 2.6249381798811807e-05,
"loss": 0.1741,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08275163173675537,
"step": 1325,
"valid_targets_mean": 2866.8,
"valid_targets_min": 1779
},
{
"epoch": 3.2281553398058254,
"grad_norm": 0.721622252671796,
"learning_rate": 2.6134266599884344e-05,
"loss": 0.1913,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09183996170759201,
"step": 1330,
"valid_targets_mean": 2902.8,
"valid_targets_min": 1400
},
{
"epoch": 3.2402912621359223,
"grad_norm": 0.6836600741488431,
"learning_rate": 2.6018926637164895e-05,
"loss": 0.1841,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06310693919658661,
"step": 1335,
"valid_targets_mean": 2040.1,
"valid_targets_min": 1144
},
{
"epoch": 3.2524271844660193,
"grad_norm": 0.6377407271749633,
"learning_rate": 2.5903366136789882e-05,
"loss": 0.1758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09916958212852478,
"step": 1340,
"valid_targets_mean": 3197.8,
"valid_targets_min": 1090
},
{
"epoch": 3.2645631067961167,
"grad_norm": 0.638693511747596,
"learning_rate": 2.578758933297635e-05,
"loss": 0.1919,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10669068992137909,
"step": 1345,
"valid_targets_mean": 3726.5,
"valid_targets_min": 1775
},
{
"epoch": 3.2766990291262137,
"grad_norm": 0.6876292836787211,
"learning_rate": 2.567160046786686e-05,
"loss": 0.1769,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09454578906297684,
"step": 1350,
"valid_targets_mean": 3223.4,
"valid_targets_min": 2196
},
{
"epoch": 3.2888349514563107,
"grad_norm": 0.6324928321100726,
"learning_rate": 2.5555403791374068e-05,
"loss": 0.1691,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10274048894643784,
"step": 1355,
"valid_targets_mean": 4538.9,
"valid_targets_min": 1679
},
{
"epoch": 3.3009708737864076,
"grad_norm": 0.618566895840547,
"learning_rate": 2.543900356102495e-05,
"loss": 0.1833,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09213300794363022,
"step": 1360,
"valid_targets_mean": 3491.9,
"valid_targets_min": 1245
},
{
"epoch": 3.313106796116505,
"grad_norm": 0.6519843929173121,
"learning_rate": 2.5322404041804865e-05,
"loss": 0.1904,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11786150932312012,
"step": 1365,
"valid_targets_mean": 3470.9,
"valid_targets_min": 1685
},
{
"epoch": 3.325242718446602,
"grad_norm": 0.526537215231363,
"learning_rate": 2.5205609506001222e-05,
"loss": 0.1761,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08241671323776245,
"step": 1370,
"valid_targets_mean": 3700.0,
"valid_targets_min": 1344
},
{
"epoch": 3.337378640776699,
"grad_norm": 0.7250466077689124,
"learning_rate": 2.5088624233047015e-05,
"loss": 0.1981,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07773889601230621,
"step": 1375,
"valid_targets_mean": 2372.9,
"valid_targets_min": 1156
},
{
"epoch": 3.349514563106796,
"grad_norm": 0.5958389264411624,
"learning_rate": 2.4971452509363942e-05,
"loss": 0.1797,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10091575980186462,
"step": 1380,
"valid_targets_mean": 3744.1,
"valid_targets_min": 1342
},
{
"epoch": 3.3616504854368934,
"grad_norm": 0.5648822992049304,
"learning_rate": 2.4854098628205415e-05,
"loss": 0.18,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09946310520172119,
"step": 1385,
"valid_targets_mean": 4884.4,
"valid_targets_min": 1567
},
{
"epoch": 3.3737864077669903,
"grad_norm": 0.6013851368925153,
"learning_rate": 2.4736566889499205e-05,
"loss": 0.1888,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09450529515743256,
"step": 1390,
"valid_targets_mean": 3243.0,
"valid_targets_min": 724
},
{
"epoch": 3.3859223300970873,
"grad_norm": 0.6005842252127365,
"learning_rate": 2.46188615996899e-05,
"loss": 0.1889,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07906970381736755,
"step": 1395,
"valid_targets_mean": 3335.0,
"valid_targets_min": 1742
},
{
"epoch": 3.3980582524271843,
"grad_norm": 0.6435300362525067,
"learning_rate": 2.4500987071581148e-05,
"loss": 0.1826,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0672961100935936,
"step": 1400,
"valid_targets_mean": 2638.9,
"valid_targets_min": 1155
},
{
"epoch": 3.4101941747572817,
"grad_norm": 0.6570382517711868,
"learning_rate": 2.438294762417758e-05,
"loss": 0.1785,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08356106281280518,
"step": 1405,
"valid_targets_mean": 3129.6,
"valid_targets_min": 1470
},
{
"epoch": 3.4223300970873787,
"grad_norm": 1.1769320216636936,
"learning_rate": 2.42647475825266e-05,
"loss": 0.1786,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07444003224372864,
"step": 1410,
"valid_targets_mean": 3383.8,
"valid_targets_min": 1031
},
{
"epoch": 3.4344660194174756,
"grad_norm": 0.5782257836635971,
"learning_rate": 2.4146391277559888e-05,
"loss": 0.1842,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06292670220136642,
"step": 1415,
"valid_targets_mean": 2569.1,
"valid_targets_min": 1051
},
{
"epoch": 3.4466019417475726,
"grad_norm": 0.6104145407137364,
"learning_rate": 2.402788304593473e-05,
"loss": 0.1836,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10125751048326492,
"step": 1420,
"valid_targets_mean": 3472.4,
"valid_targets_min": 986
},
{
"epoch": 3.45873786407767,
"grad_norm": 0.5761882962457345,
"learning_rate": 2.3909227229875098e-05,
"loss": 0.1832,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14404471218585968,
"step": 1425,
"valid_targets_mean": 5465.6,
"valid_targets_min": 1352
},
{
"epoch": 3.470873786407767,
"grad_norm": 0.49002822678077895,
"learning_rate": 2.3790428177012568e-05,
"loss": 0.1778,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08405841141939163,
"step": 1430,
"valid_targets_mean": 3978.2,
"valid_targets_min": 1684
},
{
"epoch": 3.483009708737864,
"grad_norm": 0.5929515003527267,
"learning_rate": 2.367149024022702e-05,
"loss": 0.1805,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08606874942779541,
"step": 1435,
"valid_targets_mean": 3651.9,
"valid_targets_min": 1685
},
{
"epoch": 3.4951456310679614,
"grad_norm": 0.6680545436984766,
"learning_rate": 2.355241777748711e-05,
"loss": 0.1881,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08092724531888962,
"step": 1440,
"valid_targets_mean": 2584.1,
"valid_targets_min": 1526
},
{
"epoch": 3.5072815533980584,
"grad_norm": 0.634699263382518,
"learning_rate": 2.3433215151690654e-05,
"loss": 0.1872,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0889577716588974,
"step": 1445,
"valid_targets_mean": 3179.8,
"valid_targets_min": 1505
},
{
"epoch": 3.5194174757281553,
"grad_norm": 0.5739195780418455,
"learning_rate": 2.3313886730504702e-05,
"loss": 0.1823,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10116564482450485,
"step": 1450,
"valid_targets_mean": 4469.4,
"valid_targets_min": 1705
},
{
"epoch": 3.5315533980582523,
"grad_norm": 0.6084191733395088,
"learning_rate": 2.319443688620556e-05,
"loss": 0.1798,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08992961049079895,
"step": 1455,
"valid_targets_mean": 3321.0,
"valid_targets_min": 1310
},
{
"epoch": 3.5436893203883493,
"grad_norm": 0.5193244853272002,
"learning_rate": 2.307486999551854e-05,
"loss": 0.196,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07257943600416183,
"step": 1460,
"valid_targets_mean": 3448.1,
"valid_targets_min": 2088
},
{
"epoch": 3.5558252427184467,
"grad_norm": 0.647300154790062,
"learning_rate": 2.2955190439457633e-05,
"loss": 0.1873,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09518633782863617,
"step": 1465,
"valid_targets_mean": 3557.0,
"valid_targets_min": 759
},
{
"epoch": 3.5679611650485437,
"grad_norm": 0.5560183013156753,
"learning_rate": 2.2835402603164952e-05,
"loss": 0.181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08485472202301025,
"step": 1470,
"valid_targets_mean": 3804.9,
"valid_targets_min": 1418
},
{
"epoch": 3.5800970873786406,
"grad_norm": 0.6420002820341509,
"learning_rate": 2.271551087575008e-05,
"loss": 0.1829,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12662814557552338,
"step": 1475,
"valid_targets_mean": 4294.1,
"valid_targets_min": 1800
},
{
"epoch": 3.592233009708738,
"grad_norm": 0.5628512567230222,
"learning_rate": 2.259551965012925e-05,
"loss": 0.1803,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08930374681949615,
"step": 1480,
"valid_targets_mean": 3868.1,
"valid_targets_min": 1605
},
{
"epoch": 3.604368932038835,
"grad_norm": 0.5942699922394248,
"learning_rate": 2.247543332286436e-05,
"loss": 0.1811,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.043789103627204895,
"step": 1485,
"valid_targets_mean": 1840.9,
"valid_targets_min": 1048
},
{
"epoch": 3.616504854368932,
"grad_norm": 0.6764635247239849,
"learning_rate": 2.2355256294001923e-05,
"loss": 0.192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0796380490064621,
"step": 1490,
"valid_targets_mean": 2909.2,
"valid_targets_min": 1579
},
{
"epoch": 3.6286407766990294,
"grad_norm": 0.6483791137954262,
"learning_rate": 2.2234992966911797e-05,
"loss": 0.1795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10145193338394165,
"step": 1495,
"valid_targets_mean": 3460.6,
"valid_targets_min": 1558
},
{
"epoch": 3.6407766990291264,
"grad_norm": 0.6522113723332937,
"learning_rate": 2.2114647748125884e-05,
"loss": 0.1761,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08800666034221649,
"step": 1500,
"valid_targets_mean": 2548.8,
"valid_targets_min": 1247
},
{
"epoch": 3.6529126213592233,
"grad_norm": 0.5500305777970501,
"learning_rate": 2.1994225047176644e-05,
"loss": 0.1732,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06895212829113007,
"step": 1505,
"valid_targets_mean": 2877.4,
"valid_targets_min": 1266
},
{
"epoch": 3.6650485436893203,
"grad_norm": 0.7123429888006033,
"learning_rate": 2.1873729276435555e-05,
"loss": 0.1801,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10481727123260498,
"step": 1510,
"valid_targets_mean": 3103.2,
"valid_targets_min": 982
},
{
"epoch": 3.6771844660194173,
"grad_norm": 0.6291764660443171,
"learning_rate": 2.1753164850951388e-05,
"loss": 0.1841,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11651508510112762,
"step": 1515,
"valid_targets_mean": 3804.6,
"valid_targets_min": 1710
},
{
"epoch": 3.6893203883495147,
"grad_norm": 0.6184525670689193,
"learning_rate": 2.1632536188288514e-05,
"loss": 0.1853,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09603258222341537,
"step": 1520,
"valid_targets_mean": 3642.8,
"valid_targets_min": 1650
},
{
"epoch": 3.7014563106796117,
"grad_norm": 0.6959373671584226,
"learning_rate": 2.1511847708364968e-05,
"loss": 0.1815,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07312964648008347,
"step": 1525,
"valid_targets_mean": 2316.8,
"valid_targets_min": 1146
},
{
"epoch": 3.7135922330097086,
"grad_norm": 0.5983315653496404,
"learning_rate": 2.1391103833290547e-05,
"loss": 0.1843,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07146932929754257,
"step": 1530,
"valid_targets_mean": 2917.2,
"valid_targets_min": 1902
},
{
"epoch": 3.725728155339806,
"grad_norm": 0.6049580609133546,
"learning_rate": 2.127030898720476e-05,
"loss": 0.1775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09873651713132858,
"step": 1535,
"valid_targets_mean": 4061.9,
"valid_targets_min": 1939
},
{
"epoch": 3.737864077669903,
"grad_norm": 0.539463192250893,
"learning_rate": 2.1149467596114718e-05,
"loss": 0.1866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07125517725944519,
"step": 1540,
"valid_targets_mean": 3048.1,
"valid_targets_min": 1761
},
{
"epoch": 3.75,
"grad_norm": 0.6228416964259874,
"learning_rate": 2.1028584087732996e-05,
"loss": 0.1851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10454161465167999,
"step": 1545,
"valid_targets_mean": 3733.9,
"valid_targets_min": 2098
},
{
"epoch": 3.762135922330097,
"grad_norm": 0.5668212572848379,
"learning_rate": 2.090766289131536e-05,
"loss": 0.1787,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07913221418857574,
"step": 1550,
"valid_targets_mean": 3194.4,
"valid_targets_min": 1549
},
{
"epoch": 3.774271844660194,
"grad_norm": 0.6440715401161596,
"learning_rate": 2.0786708437498495e-05,
"loss": 0.1884,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13528820872306824,
"step": 1555,
"valid_targets_mean": 3940.4,
"valid_targets_min": 1713
},
{
"epoch": 3.7864077669902914,
"grad_norm": 0.5294351757910775,
"learning_rate": 2.066572515813766e-05,
"loss": 0.1749,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09246683120727539,
"step": 1560,
"valid_targets_mean": 4199.2,
"valid_targets_min": 1969
},
{
"epoch": 3.7985436893203883,
"grad_norm": 0.6071394838184747,
"learning_rate": 2.0544717486144304e-05,
"loss": 0.187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1087888553738594,
"step": 1565,
"valid_targets_mean": 3913.0,
"valid_targets_min": 1666
},
{
"epoch": 3.8106796116504853,
"grad_norm": 0.6536536181259859,
"learning_rate": 2.042368985532364e-05,
"loss": 0.1901,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0904291570186615,
"step": 1570,
"valid_targets_mean": 3215.1,
"valid_targets_min": 1490
},
{
"epoch": 3.8228155339805827,
"grad_norm": 0.6138395154451999,
"learning_rate": 2.0302646700212176e-05,
"loss": 0.1829,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07491710782051086,
"step": 1575,
"valid_targets_mean": 2292.6,
"valid_targets_min": 1231
},
{
"epoch": 3.8349514563106797,
"grad_norm": 0.6376732356779239,
"learning_rate": 2.0181592455915265e-05,
"loss": 0.1859,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09693042933940887,
"step": 1580,
"valid_targets_mean": 3006.1,
"valid_targets_min": 1557
},
{
"epoch": 3.8470873786407767,
"grad_norm": 0.5375346908706248,
"learning_rate": 2.006053155794454e-05,
"loss": 0.1794,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09568825364112854,
"step": 1585,
"valid_targets_mean": 4550.6,
"valid_targets_min": 836
},
{
"epoch": 3.8592233009708736,
"grad_norm": 0.5868045602794454,
"learning_rate": 1.9939468442055464e-05,
"loss": 0.186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08628455549478531,
"step": 1590,
"valid_targets_mean": 3895.5,
"valid_targets_min": 1935
},
{
"epoch": 3.8713592233009706,
"grad_norm": 0.6044670299599346,
"learning_rate": 1.9818407544084738e-05,
"loss": 0.1861,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09561071544885635,
"step": 1595,
"valid_targets_mean": 3360.9,
"valid_targets_min": 1194
},
{
"epoch": 3.883495145631068,
"grad_norm": 0.6219913232811554,
"learning_rate": 1.9697353299787823e-05,
"loss": 0.1911,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09782769531011581,
"step": 1600,
"valid_targets_mean": 3633.2,
"valid_targets_min": 1742
},
{
"epoch": 3.895631067961165,
"grad_norm": 0.6553478693469604,
"learning_rate": 1.957631014467637e-05,
"loss": 0.1792,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11004262417554855,
"step": 1605,
"valid_targets_mean": 4486.5,
"valid_targets_min": 2033
},
{
"epoch": 3.907766990291262,
"grad_norm": 0.6149287642186397,
"learning_rate": 1.94552825138557e-05,
"loss": 0.1805,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07110429555177689,
"step": 1610,
"valid_targets_mean": 2959.8,
"valid_targets_min": 1111
},
{
"epoch": 3.9199029126213594,
"grad_norm": 0.6939900736792899,
"learning_rate": 1.9334274841862348e-05,
"loss": 0.1868,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0976545438170433,
"step": 1615,
"valid_targets_mean": 2351.5,
"valid_targets_min": 875
},
{
"epoch": 3.9320388349514563,
"grad_norm": 0.6077962073963812,
"learning_rate": 1.921329156250151e-05,
"loss": 0.185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10699484497308731,
"step": 1620,
"valid_targets_mean": 4106.9,
"valid_targets_min": 1060
},
{
"epoch": 3.9441747572815533,
"grad_norm": 0.7625026036100137,
"learning_rate": 1.9092337108684645e-05,
"loss": 0.1841,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09406739473342896,
"step": 1625,
"valid_targets_mean": 3242.8,
"valid_targets_min": 1673
},
{
"epoch": 3.9563106796116507,
"grad_norm": 0.6792620792485804,
"learning_rate": 1.8971415912267004e-05,
"loss": 0.174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10032247751951218,
"step": 1630,
"valid_targets_mean": 2993.8,
"valid_targets_min": 1683
},
{
"epoch": 3.9684466019417477,
"grad_norm": 0.5877117810006225,
"learning_rate": 1.8850532403885292e-05,
"loss": 0.1918,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10044887661933899,
"step": 1635,
"valid_targets_mean": 4671.0,
"valid_targets_min": 2346
},
{
"epoch": 3.9805825242718447,
"grad_norm": 0.6674443024989563,
"learning_rate": 1.872969101279525e-05,
"loss": 0.1851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08899429440498352,
"step": 1640,
"valid_targets_mean": 4927.5,
"valid_targets_min": 2941
},
{
"epoch": 3.9927184466019416,
"grad_norm": 0.6360221604527024,
"learning_rate": 1.860889616670946e-05,
"loss": 0.1807,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10733228176832199,
"step": 1645,
"valid_targets_mean": 4231.6,
"valid_targets_min": 2024
},
{
"epoch": 4.004854368932039,
"grad_norm": 0.6376305041106346,
"learning_rate": 1.8488152291635035e-05,
"loss": 0.1741,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0698719322681427,
"step": 1650,
"valid_targets_mean": 2299.9,
"valid_targets_min": 986
},
{
"epoch": 4.016990291262136,
"grad_norm": 0.6007981842561522,
"learning_rate": 1.8367463811711492e-05,
"loss": 0.1605,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08111803233623505,
"step": 1655,
"valid_targets_mean": 3804.0,
"valid_targets_min": 1731
},
{
"epoch": 4.029126213592233,
"grad_norm": 0.5901549735726199,
"learning_rate": 1.8246835149048615e-05,
"loss": 0.1618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0642840713262558,
"step": 1660,
"valid_targets_mean": 3591.2,
"valid_targets_min": 1592
},
{
"epoch": 4.04126213592233,
"grad_norm": 0.5919920312184711,
"learning_rate": 1.8126270723564455e-05,
"loss": 0.1625,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08251504600048065,
"step": 1665,
"valid_targets_mean": 4157.9,
"valid_targets_min": 1743
},
{
"epoch": 4.053398058252427,
"grad_norm": 0.7219194445693066,
"learning_rate": 1.8005774952823363e-05,
"loss": 0.1609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08370067179203033,
"step": 1670,
"valid_targets_mean": 2554.2,
"valid_targets_min": 885
},
{
"epoch": 4.065533980582524,
"grad_norm": 0.6050343829279111,
"learning_rate": 1.7885352251874126e-05,
"loss": 0.1652,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09770159423351288,
"step": 1675,
"valid_targets_mean": 4725.2,
"valid_targets_min": 2557
},
{
"epoch": 4.077669902912621,
"grad_norm": 0.6122677134201706,
"learning_rate": 1.776500703308821e-05,
"loss": 0.1627,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08404100686311722,
"step": 1680,
"valid_targets_mean": 3240.8,
"valid_targets_min": 1242
},
{
"epoch": 4.089805825242719,
"grad_norm": 1.0024615164533803,
"learning_rate": 1.764474370599808e-05,
"loss": 0.1582,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06985482573509216,
"step": 1685,
"valid_targets_mean": 2736.6,
"valid_targets_min": 1764
},
{
"epoch": 4.101941747572815,
"grad_norm": 0.6771468418726674,
"learning_rate": 1.7524566677135643e-05,
"loss": 0.1653,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10466563701629639,
"step": 1690,
"valid_targets_mean": 3743.9,
"valid_targets_min": 1453
},
{
"epoch": 4.114077669902913,
"grad_norm": 0.5823396929950576,
"learning_rate": 1.7404480349870755e-05,
"loss": 0.1593,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07395598292350769,
"step": 1695,
"valid_targets_mean": 4382.4,
"valid_targets_min": 1922
},
{
"epoch": 4.12621359223301,
"grad_norm": 0.6446244067916461,
"learning_rate": 1.7284489124249925e-05,
"loss": 0.1613,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07926598191261292,
"step": 1700,
"valid_targets_mean": 4340.9,
"valid_targets_min": 1533
},
{
"epoch": 4.138349514563107,
"grad_norm": 0.6168378593335033,
"learning_rate": 1.7164597396835058e-05,
"loss": 0.1728,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0865173414349556,
"step": 1705,
"valid_targets_mean": 4124.6,
"valid_targets_min": 1450
},
{
"epoch": 4.150485436893204,
"grad_norm": 0.6383232379942986,
"learning_rate": 1.7044809560542374e-05,
"loss": 0.1584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08923160284757614,
"step": 1710,
"valid_targets_mean": 3533.0,
"valid_targets_min": 1626
},
{
"epoch": 4.162621359223301,
"grad_norm": 0.6912681968459473,
"learning_rate": 1.6925130004481465e-05,
"loss": 0.1652,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08258406817913055,
"step": 1715,
"valid_targets_mean": 2640.1,
"valid_targets_min": 1303
},
{
"epoch": 4.174757281553398,
"grad_norm": 0.7126872237286653,
"learning_rate": 1.6805563113794444e-05,
"loss": 0.1674,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08104707300662994,
"step": 1720,
"valid_targets_mean": 2673.0,
"valid_targets_min": 1925
},
{
"epoch": 4.186893203883495,
"grad_norm": 0.6268106349111477,
"learning_rate": 1.66861132694953e-05,
"loss": 0.1647,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06212837994098663,
"step": 1725,
"valid_targets_mean": 2566.4,
"valid_targets_min": 1996
},
{
"epoch": 4.199029126213592,
"grad_norm": 0.7038366770653405,
"learning_rate": 1.656678484830936e-05,
"loss": 0.1705,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08673835545778275,
"step": 1730,
"valid_targets_mean": 2931.2,
"valid_targets_min": 852
},
{
"epoch": 4.211165048543689,
"grad_norm": 0.7327301618866306,
"learning_rate": 1.6447582222512895e-05,
"loss": 0.1575,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09165322035551071,
"step": 1735,
"valid_targets_mean": 3164.9,
"valid_targets_min": 1498
},
{
"epoch": 4.223300970873787,
"grad_norm": 0.5327241800055587,
"learning_rate": 1.632850975977299e-05,
"loss": 0.1649,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.062342334538698196,
"step": 1740,
"valid_targets_mean": 3985.8,
"valid_targets_min": 1285
},
{
"epoch": 4.235436893203883,
"grad_norm": 0.6538854123388836,
"learning_rate": 1.620957182298744e-05,
"loss": 0.1632,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08116279542446136,
"step": 1745,
"valid_targets_mean": 3399.4,
"valid_targets_min": 1615
},
{
"epoch": 4.247572815533981,
"grad_norm": 0.6617629182004681,
"learning_rate": 1.6090772770124905e-05,
"loss": 0.1715,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08452899008989334,
"step": 1750,
"valid_targets_mean": 2826.5,
"valid_targets_min": 1120
},
{
"epoch": 4.259708737864077,
"grad_norm": 0.5470542086227382,
"learning_rate": 1.5972116954065277e-05,
"loss": 0.1629,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07663094997406006,
"step": 1755,
"valid_targets_mean": 4436.6,
"valid_targets_min": 1458
},
{
"epoch": 4.271844660194175,
"grad_norm": 0.5993578996228709,
"learning_rate": 1.5853608722440112e-05,
"loss": 0.1633,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06837455928325653,
"step": 1760,
"valid_targets_mean": 3948.0,
"valid_targets_min": 1964
},
{
"epoch": 4.283980582524272,
"grad_norm": 0.5731052138674041,
"learning_rate": 1.5735252417473405e-05,
"loss": 0.1582,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05875806510448456,
"step": 1765,
"valid_targets_mean": 3568.8,
"valid_targets_min": 1103
},
{
"epoch": 4.296116504854369,
"grad_norm": 0.5717546980994862,
"learning_rate": 1.561705237582243e-05,
"loss": 0.1575,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06041959673166275,
"step": 1770,
"valid_targets_mean": 3666.8,
"valid_targets_min": 1782
},
{
"epoch": 4.308252427184466,
"grad_norm": 0.672825821159089,
"learning_rate": 1.5499012928418856e-05,
"loss": 0.1578,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08665572106838226,
"step": 1775,
"valid_targets_mean": 3513.6,
"valid_targets_min": 1124
},
{
"epoch": 4.320388349514563,
"grad_norm": 0.6249589334115596,
"learning_rate": 1.5381138400310103e-05,
"loss": 0.1618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07051187753677368,
"step": 1780,
"valid_targets_mean": 3477.1,
"valid_targets_min": 1914
},
{
"epoch": 4.33252427184466,
"grad_norm": 0.6570106040959265,
"learning_rate": 1.5263433110500802e-05,
"loss": 0.1706,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08447672426700592,
"step": 1785,
"valid_targets_mean": 3259.0,
"valid_targets_min": 1274
},
{
"epoch": 4.344660194174757,
"grad_norm": 0.6122632078981618,
"learning_rate": 1.514590137179459e-05,
"loss": 0.1666,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0870431587100029,
"step": 1790,
"valid_targets_mean": 3619.5,
"valid_targets_min": 1541
},
{
"epoch": 4.356796116504855,
"grad_norm": 0.5847849033222478,
"learning_rate": 1.5028547490636064e-05,
"loss": 0.1607,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0881710797548294,
"step": 1795,
"valid_targets_mean": 5009.1,
"valid_targets_min": 1361
},
{
"epoch": 4.368932038834951,
"grad_norm": 0.7505995641953497,
"learning_rate": 1.4911375766952993e-05,
"loss": 0.1648,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09135429561138153,
"step": 1800,
"valid_targets_mean": 2734.8,
"valid_targets_min": 1468
},
{
"epoch": 4.381067961165049,
"grad_norm": 0.6368510145976548,
"learning_rate": 1.4794390493998783e-05,
"loss": 0.1668,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08600610494613647,
"step": 1805,
"valid_targets_mean": 3418.2,
"valid_targets_min": 1986
},
{
"epoch": 4.393203883495145,
"grad_norm": 0.6312219911308268,
"learning_rate": 1.4677595958195144e-05,
"loss": 0.167,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.097256138920784,
"step": 1810,
"valid_targets_mean": 3581.1,
"valid_targets_min": 836
},
{
"epoch": 4.405339805825243,
"grad_norm": 0.6764938564803865,
"learning_rate": 1.4560996438975053e-05,
"loss": 0.1667,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10250940918922424,
"step": 1815,
"valid_targets_mean": 3384.1,
"valid_targets_min": 1170
},
{
"epoch": 4.41747572815534,
"grad_norm": 0.4918369833266944,
"learning_rate": 1.4444596208625932e-05,
"loss": 0.1552,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08474405109882355,
"step": 1820,
"valid_targets_mean": 5902.1,
"valid_targets_min": 2400
},
{
"epoch": 4.429611650485437,
"grad_norm": 0.6021602018323046,
"learning_rate": 1.4328399532133142e-05,
"loss": 0.1675,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08138144016265869,
"step": 1825,
"valid_targets_mean": 3603.5,
"valid_targets_min": 2501
},
{
"epoch": 4.441747572815534,
"grad_norm": 0.5993399151174874,
"learning_rate": 1.4212410667023659e-05,
"loss": 0.1539,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07491548359394073,
"step": 1830,
"valid_targets_mean": 3218.6,
"valid_targets_min": 1087
},
{
"epoch": 4.453883495145631,
"grad_norm": 0.5787747505641043,
"learning_rate": 1.4096633863210125e-05,
"loss": 0.1665,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08039519190788269,
"step": 1835,
"valid_targets_mean": 4166.8,
"valid_targets_min": 1904
},
{
"epoch": 4.466019417475728,
"grad_norm": 0.6382698005002889,
"learning_rate": 1.398107336283511e-05,
"loss": 0.1707,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08741675317287445,
"step": 1840,
"valid_targets_mean": 3511.0,
"valid_targets_min": 1714
},
{
"epoch": 4.478155339805825,
"grad_norm": 1.324102196522183,
"learning_rate": 1.3865733400115664e-05,
"loss": 0.1585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07260295748710632,
"step": 1845,
"valid_targets_mean": 2769.2,
"valid_targets_min": 1224
},
{
"epoch": 4.490291262135923,
"grad_norm": 0.5340287993354884,
"learning_rate": 1.3750618201188197e-05,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08613450825214386,
"step": 1850,
"valid_targets_mean": 4908.6,
"valid_targets_min": 1699
},
{
"epoch": 4.502427184466019,
"grad_norm": 0.5675741704382613,
"learning_rate": 1.3635731983953611e-05,
"loss": 0.1658,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07664692401885986,
"step": 1855,
"valid_targets_mean": 3896.5,
"valid_targets_min": 1739
},
{
"epoch": 4.514563106796117,
"grad_norm": 0.646327808417541,
"learning_rate": 1.352107895792276e-05,
"loss": 0.1688,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08643044531345367,
"step": 1860,
"valid_targets_mean": 3422.1,
"valid_targets_min": 1024
},
{
"epoch": 4.526699029126213,
"grad_norm": 0.607222668948478,
"learning_rate": 1.3406663324062197e-05,
"loss": 0.1555,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08579134941101074,
"step": 1865,
"valid_targets_mean": 4283.4,
"valid_targets_min": 1432
},
{
"epoch": 4.538834951456311,
"grad_norm": 0.678168899247409,
"learning_rate": 1.329248927464029e-05,
"loss": 0.1645,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08471087366342545,
"step": 1870,
"valid_targets_mean": 2757.2,
"valid_targets_min": 1763
},
{
"epoch": 4.550970873786408,
"grad_norm": 0.7091594797510936,
"learning_rate": 1.3178560993073557e-05,
"loss": 0.1629,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07870915532112122,
"step": 1875,
"valid_targets_mean": 2751.5,
"valid_targets_min": 1504
},
{
"epoch": 4.563106796116505,
"grad_norm": 0.6152835555029719,
"learning_rate": 1.3064882653773432e-05,
"loss": 0.1691,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09075906127691269,
"step": 1880,
"valid_targets_mean": 4003.4,
"valid_targets_min": 1048
},
{
"epoch": 4.575242718446602,
"grad_norm": 0.5947941583470783,
"learning_rate": 1.2951458421993274e-05,
"loss": 0.165,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05957557260990143,
"step": 1885,
"valid_targets_mean": 2733.8,
"valid_targets_min": 727
},
{
"epoch": 4.5873786407766985,
"grad_norm": 0.7036294916041566,
"learning_rate": 1.2838292453675788e-05,
"loss": 0.1648,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0706222802400589,
"step": 1890,
"valid_targets_mean": 2274.9,
"valid_targets_min": 782
},
{
"epoch": 4.599514563106796,
"grad_norm": 0.6089799750152856,
"learning_rate": 1.2725388895300697e-05,
"loss": 0.1648,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10470053553581238,
"step": 1895,
"valid_targets_mean": 4981.9,
"valid_targets_min": 1097
},
{
"epoch": 4.611650485436893,
"grad_norm": 0.5561455892664549,
"learning_rate": 1.2612751883732855e-05,
"loss": 0.158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.043489858508110046,
"step": 1900,
"valid_targets_mean": 2397.8,
"valid_targets_min": 1376
},
{
"epoch": 4.62378640776699,
"grad_norm": 0.6977809275100504,
"learning_rate": 1.2500385546070666e-05,
"loss": 0.1597,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07588768005371094,
"step": 1905,
"valid_targets_mean": 2811.4,
"valid_targets_min": 1853
},
{
"epoch": 4.635922330097087,
"grad_norm": 0.6324741335117167,
"learning_rate": 1.2388293999494838e-05,
"loss": 0.1597,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.061961568892002106,
"step": 1910,
"valid_targets_mean": 2950.4,
"valid_targets_min": 1448
},
{
"epoch": 4.648058252427185,
"grad_norm": 0.7614644553741854,
"learning_rate": 1.2276481351117555e-05,
"loss": 0.1619,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06363546848297119,
"step": 1915,
"valid_targets_mean": 2392.4,
"valid_targets_min": 1310
},
{
"epoch": 4.660194174757281,
"grad_norm": 0.6003178649589078,
"learning_rate": 1.216495169783197e-05,
"loss": 0.1641,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10143725574016571,
"step": 1920,
"valid_targets_mean": 4638.2,
"valid_targets_min": 2103
},
{
"epoch": 4.672330097087379,
"grad_norm": 0.6003661964218074,
"learning_rate": 1.2053709126162092e-05,
"loss": 0.1636,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0876988023519516,
"step": 1925,
"valid_targets_mean": 3837.8,
"valid_targets_min": 1979
},
{
"epoch": 4.684466019417476,
"grad_norm": 0.6922996969735212,
"learning_rate": 1.194275771211308e-05,
"loss": 0.1634,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06299655139446259,
"step": 1930,
"valid_targets_mean": 2526.6,
"valid_targets_min": 1122
},
{
"epoch": 4.696601941747573,
"grad_norm": 0.5846578429974645,
"learning_rate": 1.1832101521021856e-05,
"loss": 0.1659,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09790931642055511,
"step": 1935,
"valid_targets_mean": 4971.5,
"valid_targets_min": 2459
},
{
"epoch": 4.70873786407767,
"grad_norm": 0.5969391712768711,
"learning_rate": 1.1721744607408179e-05,
"loss": 0.1678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0635816901922226,
"step": 1940,
"valid_targets_mean": 3122.8,
"valid_targets_min": 1717
},
{
"epoch": 4.720873786407767,
"grad_norm": 0.47995347289803375,
"learning_rate": 1.1611691014826065e-05,
"loss": 0.1603,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07271182537078857,
"step": 1945,
"valid_targets_mean": 5810.5,
"valid_targets_min": 1698
},
{
"epoch": 4.733009708737864,
"grad_norm": 0.6457635207360032,
"learning_rate": 1.1501944775715654e-05,
"loss": 0.1603,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08773714303970337,
"step": 1950,
"valid_targets_mean": 3569.2,
"valid_targets_min": 1009
},
{
"epoch": 4.745145631067961,
"grad_norm": 0.6316691464441535,
"learning_rate": 1.1392509911255427e-05,
"loss": 0.1624,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07867784798145294,
"step": 1955,
"valid_targets_mean": 3457.5,
"valid_targets_min": 1546
},
{
"epoch": 4.757281553398058,
"grad_norm": 0.5142155187944049,
"learning_rate": 1.128339043121486e-05,
"loss": 0.1704,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09572489559650421,
"step": 1960,
"valid_targets_mean": 6349.5,
"valid_targets_min": 1020
},
{
"epoch": 4.769417475728155,
"grad_norm": 0.5627284295706056,
"learning_rate": 1.117459033380757e-05,
"loss": 0.164,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06830741465091705,
"step": 1965,
"valid_targets_mean": 3174.4,
"valid_targets_min": 1520
},
{
"epoch": 4.781553398058253,
"grad_norm": 0.6322304484972311,
"learning_rate": 1.1066113605544748e-05,
"loss": 0.1616,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07575127482414246,
"step": 1970,
"valid_targets_mean": 3068.0,
"valid_targets_min": 1530
},
{
"epoch": 4.793689320388349,
"grad_norm": 0.5744716583139726,
"learning_rate": 1.095796422108911e-05,
"loss": 0.1657,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10021036863327026,
"step": 1975,
"valid_targets_mean": 4126.1,
"valid_targets_min": 2032
},
{
"epoch": 4.805825242718447,
"grad_norm": 0.5897496398386591,
"learning_rate": 1.0850146143109268e-05,
"loss": 0.1638,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08345703780651093,
"step": 1980,
"valid_targets_mean": 4125.6,
"valid_targets_min": 1943
},
{
"epoch": 4.817961165048544,
"grad_norm": 0.5834807960696952,
"learning_rate": 1.0742663322134568e-05,
"loss": 0.1686,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0929785892367363,
"step": 1985,
"valid_targets_mean": 3861.1,
"valid_targets_min": 1808
},
{
"epoch": 4.830097087378641,
"grad_norm": 0.5310239123233369,
"learning_rate": 1.0635519696410247e-05,
"loss": 0.1556,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07360508292913437,
"step": 1990,
"valid_targets_mean": 4016.9,
"valid_targets_min": 1277
},
{
"epoch": 4.842233009708738,
"grad_norm": 0.8508797164260467,
"learning_rate": 1.0528719191753236e-05,
"loss": 0.1739,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08134840428829193,
"step": 1995,
"valid_targets_mean": 3629.8,
"valid_targets_min": 1638
},
{
"epoch": 4.854368932038835,
"grad_norm": 0.6108099167468786,
"learning_rate": 1.0422265721408262e-05,
"loss": 0.166,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10604679584503174,
"step": 2000,
"valid_targets_mean": 4434.2,
"valid_targets_min": 1602
},
{
"epoch": 4.866504854368932,
"grad_norm": 0.5430263362704955,
"learning_rate": 1.0316163185904474e-05,
"loss": 0.1604,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0801430195569992,
"step": 2005,
"valid_targets_mean": 4722.5,
"valid_targets_min": 2097
},
{
"epoch": 4.878640776699029,
"grad_norm": 0.672769577413632,
"learning_rate": 1.0210415472912514e-05,
"loss": 0.1595,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08061999082565308,
"step": 2010,
"valid_targets_mean": 2972.4,
"valid_targets_min": 931
},
{
"epoch": 4.890776699029126,
"grad_norm": 0.6332603787510136,
"learning_rate": 1.0105026457102096e-05,
"loss": 0.1581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10193298757076263,
"step": 2015,
"valid_targets_mean": 4878.4,
"valid_targets_min": 1214
},
{
"epoch": 4.902912621359223,
"grad_norm": 1.2157452844117738,
"learning_rate": 1.0000000000000006e-05,
"loss": 0.1704,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10513357818126678,
"step": 2020,
"valid_targets_mean": 3313.5,
"valid_targets_min": 1948
},
{
"epoch": 4.915048543689321,
"grad_norm": 0.5728099544633143,
"learning_rate": 9.895339949848635e-06,
"loss": 0.1693,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0697425827383995,
"step": 2025,
"valid_targets_mean": 3168.4,
"valid_targets_min": 1441
},
{
"epoch": 4.927184466019417,
"grad_norm": 0.5626737460000928,
"learning_rate": 9.79105014146496e-06,
"loss": 0.154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05919818580150604,
"step": 2030,
"valid_targets_mean": 2758.6,
"valid_targets_min": 1445
},
{
"epoch": 4.939320388349515,
"grad_norm": 0.6320376879389409,
"learning_rate": 9.687134396100062e-06,
"loss": 0.1634,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06882365047931671,
"step": 2035,
"valid_targets_mean": 2641.6,
"valid_targets_min": 1115
},
{
"epoch": 4.951456310679612,
"grad_norm": 0.6278607616651094,
"learning_rate": 9.583596521299077e-06,
"loss": 0.1645,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08043377101421356,
"step": 2040,
"valid_targets_mean": 2969.1,
"valid_targets_min": 1457
},
{
"epoch": 4.963592233009709,
"grad_norm": 0.5735892228632438,
"learning_rate": 9.4804403107617e-06,
"loss": 0.1579,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07562442123889923,
"step": 2045,
"valid_targets_mean": 3207.5,
"valid_targets_min": 1076
},
{
"epoch": 4.975728155339806,
"grad_norm": 0.6936606634767299,
"learning_rate": 9.377669544203187e-06,
"loss": 0.1618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07958224415779114,
"step": 2050,
"valid_targets_mean": 2781.1,
"valid_targets_min": 859
},
{
"epoch": 4.987864077669903,
"grad_norm": 0.6104900715500903,
"learning_rate": 9.275287987215857e-06,
"loss": 0.1606,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09229432046413422,
"step": 2055,
"valid_targets_mean": 3740.0,
"valid_targets_min": 1446
},
{
"epoch": 5.0,
"grad_norm": 0.6637694491776907,
"learning_rate": 9.173299391131126e-06,
"loss": 0.1678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08112204819917679,
"step": 2060,
"valid_targets_mean": 3364.8,
"valid_targets_min": 1075
},
{
"epoch": 5.012135922330097,
"grad_norm": 0.5694499299382454,
"learning_rate": 9.071707492882035e-06,
"loss": 0.1438,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07236331701278687,
"step": 2065,
"valid_targets_mean": 3204.6,
"valid_targets_min": 1869
},
{
"epoch": 5.024271844660194,
"grad_norm": 0.640940566866636,
"learning_rate": 8.97051601486635e-06,
"loss": 0.1432,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07630031555891037,
"step": 2070,
"valid_targets_mean": 3385.1,
"valid_targets_min": 1410
},
{
"epoch": 5.036407766990291,
"grad_norm": 0.6219352174108067,
"learning_rate": 8.86972866481017e-06,
"loss": 0.1558,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08140607923269272,
"step": 2075,
"valid_targets_mean": 3766.6,
"valid_targets_min": 2162
},
{
"epoch": 5.048543689320389,
"grad_norm": 0.709381004761882,
"learning_rate": 8.769349135632059e-06,
"loss": 0.1474,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09190673381090164,
"step": 2080,
"valid_targets_mean": 3932.1,
"valid_targets_min": 1246
},
{
"epoch": 5.060679611650485,
"grad_norm": 0.4871341052633552,
"learning_rate": 8.669381105307715e-06,
"loss": 0.1534,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07484342902898788,
"step": 2085,
"valid_targets_mean": 5967.4,
"valid_targets_min": 2101
},
{
"epoch": 5.072815533980583,
"grad_norm": 0.7404417548491246,
"learning_rate": 8.569828236735274e-06,
"loss": 0.1512,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08928266167640686,
"step": 2090,
"valid_targets_mean": 4752.4,
"valid_targets_min": 1672
},
{
"epoch": 5.084951456310679,
"grad_norm": 0.6828708192638967,
"learning_rate": 8.47069417760103e-06,
"loss": 0.1472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06445221602916718,
"step": 2095,
"valid_targets_mean": 2634.1,
"valid_targets_min": 1255
},
{
"epoch": 5.097087378640777,
"grad_norm": 0.6666631798146496,
"learning_rate": 8.371982560245817e-06,
"loss": 0.1443,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0968986302614212,
"step": 2100,
"valid_targets_mean": 4704.6,
"valid_targets_min": 1558
},
{
"epoch": 5.109223300970874,
"grad_norm": 0.6577213050896413,
"learning_rate": 8.273697001531897e-06,
"loss": 0.151,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08093103021383286,
"step": 2105,
"valid_targets_mean": 3558.5,
"valid_targets_min": 1613
},
{
"epoch": 5.121359223300971,
"grad_norm": 0.5670101026557458,
"learning_rate": 8.17584110271048e-06,
"loss": 0.1446,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0696868747472763,
"step": 2110,
"valid_targets_mean": 4481.8,
"valid_targets_min": 1128
},
{
"epoch": 5.133495145631068,
"grad_norm": 0.6959629067493437,
"learning_rate": 8.07841844928969e-06,
"loss": 0.1489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08995959162712097,
"step": 2115,
"valid_targets_mean": 3125.6,
"valid_targets_min": 1502
},
{
"epoch": 5.145631067961165,
"grad_norm": 0.6762645341760751,
"learning_rate": 7.981432610903272e-06,
"loss": 0.1527,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06880819797515869,
"step": 2120,
"valid_targets_mean": 2915.0,
"valid_targets_min": 1272
},
{
"epoch": 5.157766990291262,
"grad_norm": 0.6378908952513069,
"learning_rate": 7.884887141179764e-06,
"loss": 0.1512,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06967654824256897,
"step": 2125,
"valid_targets_mean": 3081.5,
"valid_targets_min": 1345
},
{
"epoch": 5.169902912621359,
"grad_norm": 0.6610188193789854,
"learning_rate": 7.788785577612282e-06,
"loss": 0.1527,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06627450883388519,
"step": 2130,
"valid_targets_mean": 2843.4,
"valid_targets_min": 1695
},
{
"epoch": 5.182038834951456,
"grad_norm": 0.6598136985625391,
"learning_rate": 7.693131441428902e-06,
"loss": 0.1455,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06188979744911194,
"step": 2135,
"valid_targets_mean": 3435.4,
"valid_targets_min": 1146
},
{
"epoch": 5.194174757281553,
"grad_norm": 3.373295214173741,
"learning_rate": 7.597928237463665e-06,
"loss": 0.1439,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06905145943164825,
"step": 2140,
"valid_targets_mean": 3131.8,
"valid_targets_min": 1302
},
{
"epoch": 5.206310679611651,
"grad_norm": 0.573929394646733,
"learning_rate": 7.503179454028151e-06,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0902794748544693,
"step": 2145,
"valid_targets_mean": 5315.5,
"valid_targets_min": 1953
},
{
"epoch": 5.218446601941747,
"grad_norm": 0.6791092537387124,
"learning_rate": 7.408888562783625e-06,
"loss": 0.1553,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0938718318939209,
"step": 2150,
"valid_targets_mean": 3589.5,
"valid_targets_min": 1810
},
{
"epoch": 5.230582524271845,
"grad_norm": 0.6763604265942801,
"learning_rate": 7.315059018613879e-06,
"loss": 0.1494,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10188092291355133,
"step": 2155,
"valid_targets_mean": 4453.9,
"valid_targets_min": 2242
},
{
"epoch": 5.242718446601942,
"grad_norm": 0.6629742124397665,
"learning_rate": 7.221694259498646e-06,
"loss": 0.1507,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0789945051074028,
"step": 2160,
"valid_targets_mean": 3783.9,
"valid_targets_min": 1912
},
{
"epoch": 5.254854368932039,
"grad_norm": 0.5976093798522817,
"learning_rate": 7.128797706387585e-06,
"loss": 0.1479,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07215683162212372,
"step": 2165,
"valid_targets_mean": 4004.4,
"valid_targets_min": 1367
},
{
"epoch": 5.266990291262136,
"grad_norm": 0.6634024447046827,
"learning_rate": 7.036372763074972e-06,
"loss": 0.1463,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.062130093574523926,
"step": 2170,
"valid_targets_mean": 2647.8,
"valid_targets_min": 1530
},
{
"epoch": 5.279126213592233,
"grad_norm": 0.6274673266739658,
"learning_rate": 6.944422816074976e-06,
"loss": 0.1549,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06876838207244873,
"step": 2175,
"valid_targets_mean": 3588.6,
"valid_targets_min": 1190
},
{
"epoch": 5.29126213592233,
"grad_norm": 0.6100537122859042,
"learning_rate": 6.852951234497563e-06,
"loss": 0.1525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0792982429265976,
"step": 2180,
"valid_targets_mean": 4009.8,
"valid_targets_min": 1579
},
{
"epoch": 5.303398058252427,
"grad_norm": 0.5234356440297586,
"learning_rate": 6.761961369925063e-06,
"loss": 0.1539,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08306777477264404,
"step": 2185,
"valid_targets_mean": 5989.4,
"valid_targets_min": 1856
},
{
"epoch": 5.315533980582524,
"grad_norm": 0.6400208916044893,
"learning_rate": 6.671456556289357e-06,
"loss": 0.146,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06951512396335602,
"step": 2190,
"valid_targets_mean": 3401.1,
"valid_targets_min": 1145
},
{
"epoch": 5.327669902912621,
"grad_norm": 0.6001918389244463,
"learning_rate": 6.58144010974972e-06,
"loss": 0.1496,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07771484553813934,
"step": 2195,
"valid_targets_mean": 4463.0,
"valid_targets_min": 1466
},
{
"epoch": 5.339805825242719,
"grad_norm": 0.7164999721587011,
"learning_rate": 6.491915328571332e-06,
"loss": 0.1492,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06512286514043808,
"step": 2200,
"valid_targets_mean": 2715.6,
"valid_targets_min": 990
},
{
"epoch": 5.351941747572815,
"grad_norm": 0.7164387442685589,
"learning_rate": 6.402885493004403e-06,
"loss": 0.1461,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09120684117078781,
"step": 2205,
"valid_targets_mean": 3654.6,
"valid_targets_min": 1937
},
{
"epoch": 5.364077669902913,
"grad_norm": 0.6969501735034505,
"learning_rate": 6.314353865163967e-06,
"loss": 0.1451,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09865804016590118,
"step": 2210,
"valid_targets_mean": 3678.6,
"valid_targets_min": 1839
},
{
"epoch": 5.37621359223301,
"grad_norm": 0.6236620265480889,
"learning_rate": 6.2263236889104295e-06,
"loss": 0.1586,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08302754163742065,
"step": 2215,
"valid_targets_mean": 3988.5,
"valid_targets_min": 1663
},
{
"epoch": 5.388349514563107,
"grad_norm": 0.5586806418562769,
"learning_rate": 6.138798189730624e-06,
"loss": 0.1532,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10435667634010315,
"step": 2220,
"valid_targets_mean": 5996.4,
"valid_targets_min": 1729
},
{
"epoch": 5.400485436893204,
"grad_norm": 0.847762422798948,
"learning_rate": 6.051780574619681e-06,
"loss": 0.14,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07793180644512177,
"step": 2225,
"valid_targets_mean": 4386.5,
"valid_targets_min": 2208
},
{
"epoch": 5.412621359223301,
"grad_norm": 0.8733909043378533,
"learning_rate": 5.965274031963493e-06,
"loss": 0.1525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09074176847934723,
"step": 2230,
"valid_targets_mean": 4193.4,
"valid_targets_min": 2012
},
{
"epoch": 5.424757281553398,
"grad_norm": 0.5530792491614996,
"learning_rate": 5.879281731421922e-06,
"loss": 0.1473,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08208553493022919,
"step": 2235,
"valid_targets_mean": 4456.9,
"valid_targets_min": 1585
},
{
"epoch": 5.436893203883495,
"grad_norm": 0.6521534798637539,
"learning_rate": 5.793806823812635e-06,
"loss": 0.1398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06144779920578003,
"step": 2240,
"valid_targets_mean": 2863.9,
"valid_targets_min": 1645
},
{
"epoch": 5.449029126213592,
"grad_norm": 0.5608212549314925,
"learning_rate": 5.7088524409956355e-06,
"loss": 0.1494,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07763676345348358,
"step": 2245,
"valid_targets_mean": 4621.2,
"valid_targets_min": 1267
},
{
"epoch": 5.461165048543689,
"grad_norm": 0.6192333002033894,
"learning_rate": 5.624421695758582e-06,
"loss": 0.1407,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08435680717229843,
"step": 2250,
"valid_targets_mean": 3748.1,
"valid_targets_min": 1481
},
{
"epoch": 5.473300970873787,
"grad_norm": 0.8347076210818524,
"learning_rate": 5.540517681702655e-06,
"loss": 0.1452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04964081197977066,
"step": 2255,
"valid_targets_mean": 3394.0,
"valid_targets_min": 1448
},
{
"epoch": 5.485436893203883,
"grad_norm": 0.666286106968509,
"learning_rate": 5.4571434731292516e-06,
"loss": 0.1509,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09312699735164642,
"step": 2260,
"valid_targets_mean": 3513.9,
"valid_targets_min": 1443
},
{
"epoch": 5.497572815533981,
"grad_norm": 0.621434081641305,
"learning_rate": 5.374302124927317e-06,
"loss": 0.1537,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09595674276351929,
"step": 2265,
"valid_targets_mean": 5372.8,
"valid_targets_min": 1717
},
{
"epoch": 5.509708737864077,
"grad_norm": 0.6993863559715406,
"learning_rate": 5.291996672461453e-06,
"loss": 0.1481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07010838389396667,
"step": 2270,
"valid_targets_mean": 3161.6,
"valid_targets_min": 1734
},
{
"epoch": 5.521844660194175,
"grad_norm": 0.6575558695023589,
"learning_rate": 5.2102301314606206e-06,
"loss": 0.1504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.056373730301856995,
"step": 2275,
"valid_targets_mean": 2388.8,
"valid_targets_min": 1129
},
{
"epoch": 5.533980582524272,
"grad_norm": 0.590747888216288,
"learning_rate": 5.1290054979077265e-06,
"loss": 0.1503,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07603400200605392,
"step": 2280,
"valid_targets_mean": 3713.2,
"valid_targets_min": 1312
},
{
"epoch": 5.546116504854369,
"grad_norm": 0.7792966112670023,
"learning_rate": 5.048325747929792e-06,
"loss": 0.1459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08462271094322205,
"step": 2285,
"valid_targets_mean": 5166.1,
"valid_targets_min": 1685
},
{
"epoch": 5.558252427184466,
"grad_norm": 0.6761593634804218,
"learning_rate": 4.968193837688951e-06,
"loss": 0.1576,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06802897900342941,
"step": 2290,
"valid_targets_mean": 3315.9,
"valid_targets_min": 1458
},
{
"epoch": 5.570388349514563,
"grad_norm": 0.6113496576741958,
"learning_rate": 4.888612703274083e-06,
"loss": 0.1498,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06313666701316833,
"step": 2295,
"valid_targets_mean": 3236.5,
"valid_targets_min": 1020
},
{
"epoch": 5.58252427184466,
"grad_norm": 0.6005006640946785,
"learning_rate": 4.809585260593266e-06,
"loss": 0.1485,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06968920677900314,
"step": 2300,
"valid_targets_mean": 3627.8,
"valid_targets_min": 1831
},
{
"epoch": 5.594660194174757,
"grad_norm": 0.6770840291864929,
"learning_rate": 4.731114405266935e-06,
"loss": 0.1502,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0816071480512619,
"step": 2305,
"valid_targets_mean": 3003.2,
"valid_targets_min": 1445
},
{
"epoch": 5.606796116504855,
"grad_norm": 0.7036708194120482,
"learning_rate": 4.653203012521767e-06,
"loss": 0.1433,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08601066470146179,
"step": 2310,
"valid_targets_mean": 3168.6,
"valid_targets_min": 1898
},
{
"epoch": 5.618932038834951,
"grad_norm": 0.6574914087946998,
"learning_rate": 4.5758539370853485e-06,
"loss": 0.1504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07015518844127655,
"step": 2315,
"valid_targets_mean": 3117.0,
"valid_targets_min": 1376
},
{
"epoch": 5.631067961165049,
"grad_norm": 0.6394684069252662,
"learning_rate": 4.499070013081559e-06,
"loss": 0.1575,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08622835576534271,
"step": 2320,
"valid_targets_mean": 3432.8,
"valid_targets_min": 1505
},
{
"epoch": 5.643203883495145,
"grad_norm": 0.6853231483147304,
"learning_rate": 4.42285405392676e-06,
"loss": 0.1492,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0739336758852005,
"step": 2325,
"valid_targets_mean": 2989.0,
"valid_targets_min": 1523
},
{
"epoch": 5.655339805825243,
"grad_norm": 0.742483184350521,
"learning_rate": 4.347208852226661e-06,
"loss": 0.1514,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.055889129638671875,
"step": 2330,
"valid_targets_mean": 2673.1,
"valid_targets_min": 2014
},
{
"epoch": 5.66747572815534,
"grad_norm": 0.7319081033437336,
"learning_rate": 4.2721371796740294e-06,
"loss": 0.1548,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.054273173213005066,
"step": 2335,
"valid_targets_mean": 2171.2,
"valid_targets_min": 1342
},
{
"epoch": 5.679611650485437,
"grad_norm": 0.5694929171086996,
"learning_rate": 4.197641786947133e-06,
"loss": 0.1486,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0792790949344635,
"step": 2340,
"valid_targets_mean": 4758.4,
"valid_targets_min": 2084
},
{
"epoch": 5.691747572815534,
"grad_norm": 0.6521872520098428,
"learning_rate": 4.12372540360894e-06,
"loss": 0.1524,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07627454400062561,
"step": 2345,
"valid_targets_mean": 3035.0,
"valid_targets_min": 2011
},
{
"epoch": 5.703883495145631,
"grad_norm": 0.7164127157502476,
"learning_rate": 4.050390738007109e-06,
"loss": 0.1493,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06955574452877045,
"step": 2350,
"valid_targets_mean": 2513.8,
"valid_targets_min": 802
},
{
"epoch": 5.716019417475728,
"grad_norm": 0.6649528822793993,
"learning_rate": 3.977640477174758e-06,
"loss": 0.1499,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08120199292898178,
"step": 2355,
"valid_targets_mean": 3231.2,
"valid_targets_min": 1686
},
{
"epoch": 5.728155339805825,
"grad_norm": 0.7284719176910501,
"learning_rate": 3.905477286732022e-06,
"loss": 0.1472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07665879279375076,
"step": 2360,
"valid_targets_mean": 2845.0,
"valid_targets_min": 2153
},
{
"epoch": 5.740291262135923,
"grad_norm": 0.6847212602064418,
"learning_rate": 3.833903810788357e-06,
"loss": 0.1468,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06536301970481873,
"step": 2365,
"valid_targets_mean": 2883.0,
"valid_targets_min": 1160
},
{
"epoch": 5.752427184466019,
"grad_norm": 0.6763841009989021,
"learning_rate": 3.7629226718456592e-06,
"loss": 0.1506,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07870544493198395,
"step": 2370,
"valid_targets_mean": 4191.6,
"valid_targets_min": 1700
},
{
"epoch": 5.764563106796117,
"grad_norm": 0.5773367550415236,
"learning_rate": 3.6925364707021947e-06,
"loss": 0.1518,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0766630619764328,
"step": 2375,
"valid_targets_mean": 4318.2,
"valid_targets_min": 1789
},
{
"epoch": 5.776699029126213,
"grad_norm": 0.8314691703234193,
"learning_rate": 3.6227477863573102e-06,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07838580757379532,
"step": 2380,
"valid_targets_mean": 2916.0,
"valid_targets_min": 1663
},
{
"epoch": 5.788834951456311,
"grad_norm": 0.5901858257990975,
"learning_rate": 3.553559175916901e-06,
"loss": 0.1439,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0642097145318985,
"step": 2385,
"valid_targets_mean": 3285.1,
"valid_targets_min": 982
},
{
"epoch": 5.800970873786408,
"grad_norm": 0.6621833654822529,
"learning_rate": 3.484973174499746e-06,
"loss": 0.1492,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06813094019889832,
"step": 2390,
"valid_targets_mean": 3111.0,
"valid_targets_min": 2070
},
{
"epoch": 5.813106796116505,
"grad_norm": 0.6030978802029339,
"learning_rate": 3.41699229514461e-06,
"loss": 0.1536,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08363445103168488,
"step": 2395,
"valid_targets_mean": 4574.1,
"valid_targets_min": 1607
},
{
"epoch": 5.825242718446602,
"grad_norm": 0.5840469190418897,
"learning_rate": 3.3496190287181608e-06,
"loss": 0.1573,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09648744761943817,
"step": 2400,
"valid_targets_mean": 5053.9,
"valid_targets_min": 2058
},
{
"epoch": 5.8373786407766985,
"grad_norm": 0.6571121152106498,
"learning_rate": 3.2828558438237137e-06,
"loss": 0.1526,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.059269074350595474,
"step": 2405,
"valid_targets_mean": 2840.1,
"valid_targets_min": 154
},
{
"epoch": 5.849514563106796,
"grad_norm": 0.551898856840062,
"learning_rate": 3.21670518671076e-06,
"loss": 0.1441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08503453433513641,
"step": 2410,
"valid_targets_mean": 4728.0,
"valid_targets_min": 1179
},
{
"epoch": 5.861650485436893,
"grad_norm": 0.6037149292849203,
"learning_rate": 3.15116948118537e-06,
"loss": 0.1463,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.059433773159980774,
"step": 2415,
"valid_targets_mean": 3881.9,
"valid_targets_min": 1575
},
{
"epoch": 5.87378640776699,
"grad_norm": 0.6893897603929113,
"learning_rate": 3.086251128521336e-06,
"loss": 0.1467,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07661645114421844,
"step": 2420,
"valid_targets_mean": 3307.5,
"valid_targets_min": 1303
},
{
"epoch": 5.885922330097087,
"grad_norm": 0.5716277703360539,
"learning_rate": 3.0219525073722255e-06,
"loss": 0.1506,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.039183855056762695,
"step": 2425,
"valid_targets_mean": 2702.4,
"valid_targets_min": 806
},
{
"epoch": 5.898058252427185,
"grad_norm": 0.6384719616394122,
"learning_rate": 2.9582759736842104e-06,
"loss": 0.1387,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07910118252038956,
"step": 2430,
"valid_targets_mean": 3609.9,
"valid_targets_min": 1580
},
{
"epoch": 5.910194174757281,
"grad_norm": 0.6901966069324146,
"learning_rate": 2.8952238606097417e-06,
"loss": 0.1547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07921449095010757,
"step": 2435,
"valid_targets_mean": 2894.8,
"valid_targets_min": 1541
},
{
"epoch": 5.922330097087379,
"grad_norm": 0.724488632530999,
"learning_rate": 2.832798478422072e-06,
"loss": 0.1492,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0726192370057106,
"step": 2440,
"valid_targets_mean": 3625.4,
"valid_targets_min": 2590
},
{
"epoch": 5.934466019417476,
"grad_norm": 0.6902288020190513,
"learning_rate": 2.7710021144305855e-06,
"loss": 0.1457,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06895279884338379,
"step": 2445,
"valid_targets_mean": 3310.1,
"valid_targets_min": 1599
},
{
"epoch": 5.946601941747573,
"grad_norm": 0.6835182619427295,
"learning_rate": 2.709837032897027e-06,
"loss": 0.1562,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09847630560398102,
"step": 2450,
"valid_targets_mean": 4339.9,
"valid_targets_min": 1614
},
{
"epoch": 5.95873786407767,
"grad_norm": 0.611210034754114,
"learning_rate": 2.649305474952488e-06,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07036708295345306,
"step": 2455,
"valid_targets_mean": 3745.5,
"valid_targets_min": 2641
},
{
"epoch": 5.970873786407767,
"grad_norm": 0.6680622326529521,
"learning_rate": 2.589409658515327e-06,
"loss": 0.1603,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07266347110271454,
"step": 2460,
"valid_targets_mean": 2796.6,
"valid_targets_min": 1484
},
{
"epoch": 5.983009708737864,
"grad_norm": 0.6433598851109712,
"learning_rate": 2.5301517782098928e-06,
"loss": 0.1502,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09710367023944855,
"step": 2465,
"valid_targets_mean": 4557.1,
"valid_targets_min": 1213
},
{
"epoch": 5.995145631067961,
"grad_norm": 1.00074274585472,
"learning_rate": 2.471534005286105e-06,
"loss": 0.1615,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07861033082008362,
"step": 2470,
"valid_targets_mean": 3341.9,
"valid_targets_min": 1954
},
{
"epoch": 6.007281553398058,
"grad_norm": 0.61522365712328,
"learning_rate": 2.4135584875399064e-06,
"loss": 0.1448,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06568078696727753,
"step": 2475,
"valid_targets_mean": 3456.9,
"valid_targets_min": 1789
},
{
"epoch": 6.019417475728155,
"grad_norm": 0.7095084127064099,
"learning_rate": 2.35622734923457e-06,
"loss": 0.1446,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06181590259075165,
"step": 2480,
"valid_targets_mean": 2576.9,
"valid_targets_min": 713
},
{
"epoch": 6.031553398058253,
"grad_norm": 0.6611164183178138,
"learning_rate": 2.2995426910228425e-06,
"loss": 0.1454,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08089602738618851,
"step": 2485,
"valid_targets_mean": 3604.8,
"valid_targets_min": 1750
},
{
"epoch": 6.043689320388349,
"grad_norm": 0.6340472729835719,
"learning_rate": 2.243506589870015e-06,
"loss": 0.1461,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07942962646484375,
"step": 2490,
"valid_targets_mean": 3629.8,
"valid_targets_min": 1520
},
{
"epoch": 6.055825242718447,
"grad_norm": 0.6449109263135766,
"learning_rate": 2.188121098977771e-06,
"loss": 0.1431,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09307433664798737,
"step": 2495,
"valid_targets_mean": 4451.9,
"valid_targets_min": 1934
},
{
"epoch": 6.067961165048544,
"grad_norm": 0.6393208307419251,
"learning_rate": 2.1333882477089916e-06,
"loss": 0.1406,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06869735568761826,
"step": 2500,
"valid_targets_mean": 3683.2,
"valid_targets_min": 2032
},
{
"epoch": 6.080097087378641,
"grad_norm": 0.6256073236645769,
"learning_rate": 2.079310041513403e-06,
"loss": 0.1464,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08035487681627274,
"step": 2505,
"valid_targets_mean": 4093.1,
"valid_targets_min": 1342
},
{
"epoch": 6.092233009708738,
"grad_norm": 0.5383972450058381,
"learning_rate": 2.025888461854062e-06,
"loss": 0.1535,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07836288213729858,
"step": 2510,
"valid_targets_mean": 5589.0,
"valid_targets_min": 1575
},
{
"epoch": 6.104368932038835,
"grad_norm": 0.8931658744554578,
"learning_rate": 1.973125466134778e-06,
"loss": 0.1429,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09864304959774017,
"step": 2515,
"valid_targets_mean": 4026.6,
"valid_targets_min": 1954
},
{
"epoch": 6.116504854368932,
"grad_norm": 0.6627815261302755,
"learning_rate": 1.921022987628387e-06,
"loss": 0.1411,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09288658201694489,
"step": 2520,
"valid_targets_mean": 4059.6,
"valid_targets_min": 1169
},
{
"epoch": 6.128640776699029,
"grad_norm": 0.7139910774564058,
"learning_rate": 1.8695829354059248e-06,
"loss": 0.1388,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.079840749502182,
"step": 2525,
"valid_targets_mean": 3227.5,
"valid_targets_min": 1718
},
{
"epoch": 6.140776699029126,
"grad_norm": 0.6172268313992102,
"learning_rate": 1.8188071942666497e-06,
"loss": 0.1444,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05144008621573448,
"step": 2530,
"valid_targets_mean": 2724.4,
"valid_targets_min": 1225
},
{
"epoch": 6.152912621359223,
"grad_norm": 0.6699120347201785,
"learning_rate": 1.768697624669009e-06,
"loss": 0.1434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04121575132012367,
"step": 2535,
"valid_targets_mean": 2290.6,
"valid_targets_min": 1326
},
{
"epoch": 6.165048543689321,
"grad_norm": 0.66636815986164,
"learning_rate": 1.719256062662471e-06,
"loss": 0.1417,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06704620271921158,
"step": 2540,
"valid_targets_mean": 2898.5,
"valid_targets_min": 1124
},
{
"epoch": 6.177184466019417,
"grad_norm": 0.5847545489624413,
"learning_rate": 1.6704843198202315e-06,
"loss": 0.1408,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07875080406665802,
"step": 2545,
"valid_targets_mean": 4575.8,
"valid_targets_min": 2056
},
{
"epoch": 6.189320388349515,
"grad_norm": 0.6158417794675092,
"learning_rate": 1.6223841831728426e-06,
"loss": 0.1432,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07072131335735321,
"step": 2550,
"valid_targets_mean": 3803.0,
"valid_targets_min": 1751
},
{
"epoch": 6.201456310679611,
"grad_norm": 0.7408615660931046,
"learning_rate": 1.5749574151427483e-06,
"loss": 0.1432,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07321467995643616,
"step": 2555,
"valid_targets_mean": 2858.1,
"valid_targets_min": 1361
},
{
"epoch": 6.213592233009709,
"grad_norm": 0.6440124685863008,
"learning_rate": 1.5282057534796945e-06,
"loss": 0.1419,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.059953607618808746,
"step": 2560,
"valid_targets_mean": 3239.6,
"valid_targets_min": 1376
},
{
"epoch": 6.225728155339806,
"grad_norm": 0.8379036899312501,
"learning_rate": 1.4821309111970573e-06,
"loss": 0.141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0646471306681633,
"step": 2565,
"valid_targets_mean": 2546.4,
"valid_targets_min": 1671
},
{
"epoch": 6.237864077669903,
"grad_norm": 0.6252500636859444,
"learning_rate": 1.4367345765090801e-06,
"loss": 0.1477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09526094794273376,
"step": 2570,
"valid_targets_mean": 4737.1,
"valid_targets_min": 1403
},
{
"epoch": 6.25,
"grad_norm": 0.5847749260081763,
"learning_rate": 1.3920184127690295e-06,
"loss": 0.1408,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05771882086992264,
"step": 2575,
"valid_targets_mean": 3475.0,
"valid_targets_min": 1394
},
{
"epoch": 6.262135922330097,
"grad_norm": 0.7252067642866787,
"learning_rate": 1.3479840584082204e-06,
"loss": 0.1465,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07907982915639877,
"step": 2580,
"valid_targets_mean": 2604.8,
"valid_targets_min": 1394
},
{
"epoch": 6.274271844660194,
"grad_norm": 0.6668816991108799,
"learning_rate": 1.3046331268760094e-06,
"loss": 0.1322,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0727609246969223,
"step": 2585,
"valid_targets_mean": 3145.1,
"valid_targets_min": 1734
},
{
"epoch": 6.286407766990291,
"grad_norm": 0.6579639991115882,
"learning_rate": 1.2619672065806455e-06,
"loss": 0.1505,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06705568730831146,
"step": 2590,
"valid_targets_mean": 4196.0,
"valid_targets_min": 982
},
{
"epoch": 6.298543689320388,
"grad_norm": 0.6740102969000507,
"learning_rate": 1.2199878608311155e-06,
"loss": 0.1382,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07435420155525208,
"step": 2595,
"valid_targets_mean": 3852.0,
"valid_targets_min": 2156
},
{
"epoch": 6.310679611650485,
"grad_norm": 0.6098825491635004,
"learning_rate": 1.178696627779825e-06,
"loss": 0.1377,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05692951753735542,
"step": 2600,
"valid_targets_mean": 3248.5,
"valid_targets_min": 1513
},
{
"epoch": 6.322815533980583,
"grad_norm": 0.6229483313555091,
"learning_rate": 1.1380950203662523e-06,
"loss": 0.1482,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06636616587638855,
"step": 2605,
"valid_targets_mean": 3370.1,
"valid_targets_min": 1415
},
{
"epoch": 6.334951456310679,
"grad_norm": 0.5787804773076026,
"learning_rate": 1.098184526261512e-06,
"loss": 0.1443,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06577181816101074,
"step": 2610,
"valid_targets_mean": 3861.2,
"valid_targets_min": 2004
},
{
"epoch": 6.347087378640777,
"grad_norm": 0.5614490842122992,
"learning_rate": 1.0589666078138539e-06,
"loss": 0.142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06613291054964066,
"step": 2615,
"valid_targets_mean": 4287.5,
"valid_targets_min": 2050
},
{
"epoch": 6.359223300970874,
"grad_norm": 0.5864568640047932,
"learning_rate": 1.020442701995068e-06,
"loss": 0.143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.058257006108760834,
"step": 2620,
"valid_targets_mean": 2756.1,
"valid_targets_min": 1641
},
{
"epoch": 6.371359223300971,
"grad_norm": 0.6627743134278709,
"learning_rate": 9.826142203478329e-07,
"loss": 0.145,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09011754393577576,
"step": 2625,
"valid_targets_mean": 4519.4,
"valid_targets_min": 1623
},
{
"epoch": 6.383495145631068,
"grad_norm": 2.34510705676269,
"learning_rate": 9.45482548934018e-07,
"loss": 0.1354,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08167122304439545,
"step": 2630,
"valid_targets_mean": 4168.9,
"valid_targets_min": 1451
},
{
"epoch": 6.395631067961165,
"grad_norm": 0.647067095363297,
"learning_rate": 9.090490482838699e-07,
"loss": 0.1477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06741274893283844,
"step": 2635,
"valid_targets_mean": 2769.5,
"valid_targets_min": 1308
},
{
"epoch": 6.407766990291262,
"grad_norm": 0.6749557578579873,
"learning_rate": 8.733150533461776e-07,
"loss": 0.1438,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0710570365190506,
"step": 2640,
"valid_targets_mean": 3083.4,
"valid_targets_min": 1648
},
{
"epoch": 6.419902912621359,
"grad_norm": 0.6960761992223283,
"learning_rate": 8.382818734393506e-07,
"loss": 0.1479,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08328501135110855,
"step": 2645,
"valid_targets_mean": 3413.6,
"valid_targets_min": 1341
},
{
"epoch": 6.432038834951456,
"grad_norm": 0.711865448453578,
"learning_rate": 8.039507922034584e-07,
"loss": 0.1406,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09096627682447433,
"step": 2650,
"valid_targets_mean": 3681.6,
"valid_targets_min": 2090
},
{
"epoch": 6.444174757281553,
"grad_norm": 0.5700389666545927,
"learning_rate": 7.703230675531715e-07,
"loss": 0.1365,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07652769237756729,
"step": 2655,
"valid_targets_mean": 4778.4,
"valid_targets_min": 1885
},
{
"epoch": 6.456310679611651,
"grad_norm": 0.6345626977967229,
"learning_rate": 7.373999316316949e-07,
"loss": 0.1451,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08243800699710846,
"step": 2660,
"valid_targets_mean": 4278.9,
"valid_targets_min": 1160
},
{
"epoch": 6.468446601941747,
"grad_norm": 0.7200544831084392,
"learning_rate": 7.051825907656207e-07,
"loss": 0.1482,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06946928054094315,
"step": 2665,
"valid_targets_mean": 3029.6,
"valid_targets_min": 1028
},
{
"epoch": 6.480582524271845,
"grad_norm": 0.7319929881418774,
"learning_rate": 6.736722254207051e-07,
"loss": 0.1462,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09311191737651825,
"step": 2670,
"valid_targets_mean": 3704.5,
"valid_targets_min": 1407
},
{
"epoch": 6.492718446601942,
"grad_norm": 0.7357800828338439,
"learning_rate": 6.428699901586366e-07,
"loss": 0.1425,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06931672990322113,
"step": 2675,
"valid_targets_mean": 2659.6,
"valid_targets_min": 1457
},
{
"epoch": 6.504854368932039,
"grad_norm": 0.6105271705759387,
"learning_rate": 6.127770135947297e-07,
"loss": 0.1417,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.043224647641181946,
"step": 2680,
"valid_targets_mean": 2228.5,
"valid_targets_min": 724
},
{
"epoch": 6.516990291262136,
"grad_norm": 0.6517081064818845,
"learning_rate": 5.83394398356556e-07,
"loss": 0.1391,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0976434201002121,
"step": 2685,
"valid_targets_mean": 4323.6,
"valid_targets_min": 1209
},
{
"epoch": 6.529126213592233,
"grad_norm": 0.7461683837431963,
"learning_rate": 5.547232210435672e-07,
"loss": 0.1429,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06791960448026657,
"step": 2690,
"valid_targets_mean": 2701.2,
"valid_targets_min": 1707
},
{
"epoch": 6.54126213592233,
"grad_norm": 0.6154391631008862,
"learning_rate": 5.267645321876224e-07,
"loss": 0.1338,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06627804040908813,
"step": 2695,
"valid_targets_mean": 3819.4,
"valid_targets_min": 1272
},
{
"epoch": 6.553398058252427,
"grad_norm": 0.6621111062541762,
"learning_rate": 4.995193562145173e-07,
"loss": 0.1466,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06474607437849045,
"step": 2700,
"valid_targets_mean": 3058.0,
"valid_targets_min": 1700
},
{
"epoch": 6.565533980582524,
"grad_norm": 0.6692864019480245,
"learning_rate": 4.7298869140643967e-07,
"loss": 0.1342,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07640602439641953,
"step": 2705,
"valid_targets_mean": 3494.9,
"valid_targets_min": 1440
},
{
"epoch": 6.577669902912621,
"grad_norm": 0.6757852839612576,
"learning_rate": 4.471735098653862e-07,
"loss": 0.1416,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07545896619558334,
"step": 2710,
"valid_targets_mean": 3133.0,
"valid_targets_min": 1412
},
{
"epoch": 6.589805825242719,
"grad_norm": 0.6606824660440535,
"learning_rate": 4.220747574775508e-07,
"loss": 0.1477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.048327356576919556,
"step": 2715,
"valid_targets_mean": 2134.6,
"valid_targets_min": 709
},
{
"epoch": 6.601941747572815,
"grad_norm": 0.6825078696335327,
"learning_rate": 3.976933538786698e-07,
"loss": 0.1405,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09242834895849228,
"step": 2720,
"valid_targets_mean": 3948.9,
"valid_targets_min": 1305
},
{
"epoch": 6.614077669902913,
"grad_norm": 0.6687503358919056,
"learning_rate": 3.740301924203138e-07,
"loss": 0.151,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.04534143954515457,
"step": 2725,
"valid_targets_mean": 2313.4,
"valid_targets_min": 1229
},
{
"epoch": 6.62621359223301,
"grad_norm": 0.5425668403706105,
"learning_rate": 3.51086140137169e-07,
"loss": 0.1481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06036786735057831,
"step": 2730,
"valid_targets_mean": 3309.1,
"valid_targets_min": 770
},
{
"epoch": 6.638349514563107,
"grad_norm": 0.7234834853134211,
"learning_rate": 3.2886203771525625e-07,
"loss": 0.1424,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0862589180469513,
"step": 2735,
"valid_targets_mean": 3368.8,
"valid_targets_min": 1196
},
{
"epoch": 6.650485436893204,
"grad_norm": 0.7348857742193243,
"learning_rate": 3.0735869946113994e-07,
"loss": 0.1472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11306288838386536,
"step": 2740,
"valid_targets_mean": 4025.1,
"valid_targets_min": 2143
},
{
"epoch": 6.6626213592233015,
"grad_norm": 0.574274475684817,
"learning_rate": 2.865769132720786e-07,
"loss": 0.145,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05526868999004364,
"step": 2745,
"valid_targets_mean": 2962.4,
"valid_targets_min": 1750
},
{
"epoch": 6.674757281553398,
"grad_norm": 0.6815690341104472,
"learning_rate": 2.665174406071591e-07,
"loss": 0.1452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08603289723396301,
"step": 2750,
"valid_targets_mean": 3753.0,
"valid_targets_min": 1907
},
{
"epoch": 6.686893203883495,
"grad_norm": 0.5720157458563386,
"learning_rate": 2.4718101645940574e-07,
"loss": 0.1437,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07757186889648438,
"step": 2755,
"valid_targets_mean": 5120.9,
"valid_targets_min": 1582
},
{
"epoch": 6.699029126213592,
"grad_norm": 0.6478158749013252,
"learning_rate": 2.2856834932884154e-07,
"loss": 0.1375,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06913763284683228,
"step": 2760,
"valid_targets_mean": 3226.5,
"valid_targets_min": 1325
},
{
"epoch": 6.711165048543689,
"grad_norm": 0.593286364179664,
"learning_rate": 2.106801211965226e-07,
"loss": 0.1352,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05656867101788521,
"step": 2765,
"valid_targets_mean": 3574.1,
"valid_targets_min": 957
},
{
"epoch": 6.723300970873787,
"grad_norm": 0.6015583765389638,
"learning_rate": 1.935169874995646e-07,
"loss": 0.1441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06670238077640533,
"step": 2770,
"valid_targets_mean": 3821.0,
"valid_targets_min": 1109
},
{
"epoch": 6.735436893203883,
"grad_norm": 0.6634794069113384,
"learning_rate": 1.7707957710711766e-07,
"loss": 0.1452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06610158085823059,
"step": 2775,
"valid_targets_mean": 3390.5,
"valid_targets_min": 1573
},
{
"epoch": 6.747572815533981,
"grad_norm": 0.7118167073805021,
"learning_rate": 1.61368492297318e-07,
"loss": 0.1382,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07803932577371597,
"step": 2780,
"valid_targets_mean": 3096.2,
"valid_targets_min": 1427
},
{
"epoch": 6.759708737864077,
"grad_norm": 0.7113816309993987,
"learning_rate": 1.463843087352368e-07,
"loss": 0.1501,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07603172957897186,
"step": 2785,
"valid_targets_mean": 3264.4,
"valid_targets_min": 1565
},
{
"epoch": 6.771844660194175,
"grad_norm": 0.4997666727420533,
"learning_rate": 1.321275754517748e-07,
"loss": 0.13,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.047423481941223145,
"step": 2790,
"valid_targets_mean": 2717.4,
"valid_targets_min": 1565
},
{
"epoch": 6.783980582524272,
"grad_norm": 0.7752576383887028,
"learning_rate": 1.1859881482354952e-07,
"loss": 0.1367,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06419270485639572,
"step": 2795,
"valid_targets_mean": 2384.8,
"valid_targets_min": 914
},
{
"epoch": 6.796116504854369,
"grad_norm": 0.6857112575163926,
"learning_rate": 1.0579852255375279e-07,
"loss": 0.1449,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0799475908279419,
"step": 2800,
"valid_targets_mean": 3656.0,
"valid_targets_min": 1260
},
{
"epoch": 6.808252427184466,
"grad_norm": 0.7670334504528521,
"learning_rate": 9.37271676539897e-08,
"loss": 0.142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06760217249393463,
"step": 2805,
"valid_targets_mean": 3008.2,
"valid_targets_min": 1288
},
{
"epoch": 6.820388349514563,
"grad_norm": 0.4943891378249372,
"learning_rate": 8.238519242709686e-08,
"loss": 0.1376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07551370561122894,
"step": 2810,
"valid_targets_mean": 6238.0,
"valid_targets_min": 1150
},
{
"epoch": 6.83252427184466,
"grad_norm": 0.6532748669438782,
"learning_rate": 7.177301245092639e-08,
"loss": 0.1459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07944832742214203,
"step": 2815,
"valid_targets_mean": 3733.1,
"valid_targets_min": 2267
},
{
"epoch": 6.844660194174757,
"grad_norm": 0.49783304947016654,
"learning_rate": 6.189101656313146e-08,
"loss": 0.1409,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.061214931309223175,
"step": 2820,
"valid_targets_mean": 4377.9,
"valid_targets_min": 1756
},
{
"epoch": 6.856796116504855,
"grad_norm": 0.5876157227953795,
"learning_rate": 5.273956684691106e-08,
"loss": 0.141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06784014403820038,
"step": 2825,
"valid_targets_mean": 3764.0,
"valid_targets_min": 1853
},
{
"epoch": 6.868932038834951,
"grad_norm": 0.6810625110302154,
"learning_rate": 4.431899861774058e-08,
"loss": 0.1489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09138327836990356,
"step": 2830,
"valid_targets_mean": 4906.6,
"valid_targets_min": 1563
},
{
"epoch": 6.881067961165049,
"grad_norm": 0.9684515554562491,
"learning_rate": 3.662962041109497e-08,
"loss": 0.1463,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06961645931005478,
"step": 2835,
"valid_targets_mean": 4573.1,
"valid_targets_min": 892
},
{
"epoch": 6.893203883495145,
"grad_norm": 0.622954766467944,
"learning_rate": 2.9671713971137773e-08,
"loss": 0.142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07675614953041077,
"step": 2840,
"valid_targets_mean": 3849.5,
"valid_targets_min": 1989
},
{
"epoch": 6.905339805825243,
"grad_norm": 0.6264139972858416,
"learning_rate": 2.3445534240393864e-08,
"loss": 0.1454,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05458412319421768,
"step": 2845,
"valid_targets_mean": 2797.8,
"valid_targets_min": 1432
},
{
"epoch": 6.91747572815534,
"grad_norm": 0.8154906698937269,
"learning_rate": 1.795130935041689e-08,
"loss": 0.1419,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07041016966104507,
"step": 2850,
"valid_targets_mean": 2487.6,
"valid_targets_min": 1219
},
{
"epoch": 6.929611650485437,
"grad_norm": 0.6151552275410459,
"learning_rate": 1.3189240613427079e-08,
"loss": 0.1383,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.060174889862537384,
"step": 2855,
"valid_targets_mean": 3283.4,
"valid_targets_min": 1345
},
{
"epoch": 6.941747572815534,
"grad_norm": 0.6883176452454132,
"learning_rate": 9.159502514932694e-09,
"loss": 0.1471,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07768334448337555,
"step": 2860,
"valid_targets_mean": 3149.6,
"valid_targets_min": 1492
},
{
"epoch": 6.953883495145631,
"grad_norm": 0.7446865949649802,
"learning_rate": 5.8622427073373735e-09,
"loss": 0.1462,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06840886175632477,
"step": 2865,
"valid_targets_mean": 2271.4,
"valid_targets_min": 1259
},
{
"epoch": 6.966019417475728,
"grad_norm": 0.5483440169168549,
"learning_rate": 3.2975820045333397e-09,
"loss": 0.1403,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06071193143725395,
"step": 2870,
"valid_targets_mean": 3820.0,
"valid_targets_min": 1526
},
{
"epoch": 6.978155339805825,
"grad_norm": 0.8690479433788378,
"learning_rate": 1.4656143774716136e-09,
"loss": 0.1466,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08116362988948822,
"step": 2875,
"valid_targets_mean": 3485.0,
"valid_targets_min": 1751
},
{
"epoch": 6.990291262135923,
"grad_norm": 0.6594548306711724,
"learning_rate": 3.6640695072032164e-10,
"loss": 0.1414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08794179558753967,
"step": 2880,
"valid_targets_mean": 4861.2,
"valid_targets_min": 2895
},
{
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.06831502914428711,
"step": 2884,
"total_flos": 7.580169136525804e+17,
"train_loss": 0.20268679274848034,
"train_runtime": 22655.0442,
"train_samples_per_second": 2.035,
"train_steps_per_second": 0.127,
"valid_targets_mean": 3534.1,
"valid_targets_min": 1389
}
],
"logging_steps": 5,
"max_steps": 2884,
"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": 7.580169136525804e+17,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}