Files
exp-uns-r2egym-8_4x_glm_4_7…/trainer_state.json
ModelHub XC 3eda8854bb 初始化项目,由ModelHub XC社区提供模型
Model: laion/exp-uns-r2egym-8_4x_glm_4_7_traces_jupiter
Source: Original Platform
2026-05-26 12:10:19 +08:00

9376 lines
261 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 7.0,
"eval_steps": 500,
"global_step": 4242,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.008257638315441783,
"grad_norm": 26.186356086160057,
"learning_rate": 3.764705882352941e-07,
"loss": 0.8754,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41719168424606323,
"step": 5,
"valid_targets_mean": 4119.4,
"valid_targets_min": 2593
},
{
"epoch": 0.016515276630883566,
"grad_norm": 25.76423946570163,
"learning_rate": 8.470588235294118e-07,
"loss": 0.8797,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.45204952359199524,
"step": 10,
"valid_targets_mean": 4033.1,
"valid_targets_min": 1431
},
{
"epoch": 0.02477291494632535,
"grad_norm": 21.451837643676473,
"learning_rate": 1.3176470588235296e-06,
"loss": 0.8466,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.40752485394477844,
"step": 15,
"valid_targets_mean": 4028.4,
"valid_targets_min": 2376
},
{
"epoch": 0.03303055326176713,
"grad_norm": 16.09584016764473,
"learning_rate": 1.7882352941176474e-06,
"loss": 0.8153,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.41522884368896484,
"step": 20,
"valid_targets_mean": 4668.5,
"valid_targets_min": 3321
},
{
"epoch": 0.04128819157720892,
"grad_norm": 8.349551435294561,
"learning_rate": 2.258823529411765e-06,
"loss": 0.7169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27859964966773987,
"step": 25,
"valid_targets_mean": 3812.1,
"valid_targets_min": 1777
},
{
"epoch": 0.0495458298926507,
"grad_norm": 4.721556664244327,
"learning_rate": 2.7294117647058825e-06,
"loss": 0.6578,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3394656777381897,
"step": 30,
"valid_targets_mean": 4272.2,
"valid_targets_min": 2674
},
{
"epoch": 0.057803468208092484,
"grad_norm": 2.5909296931132837,
"learning_rate": 3.2000000000000003e-06,
"loss": 0.6295,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2982783913612366,
"step": 35,
"valid_targets_mean": 4514.1,
"valid_targets_min": 3459
},
{
"epoch": 0.06606110652353427,
"grad_norm": 1.8166864624512067,
"learning_rate": 3.670588235294118e-06,
"loss": 0.5766,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2537887692451477,
"step": 40,
"valid_targets_mean": 3918.4,
"valid_targets_min": 1907
},
{
"epoch": 0.07431874483897605,
"grad_norm": 1.5788921200933257,
"learning_rate": 4.141176470588235e-06,
"loss": 0.5631,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.312985897064209,
"step": 45,
"valid_targets_mean": 4190.2,
"valid_targets_min": 2275
},
{
"epoch": 0.08257638315441784,
"grad_norm": 1.2342363684419146,
"learning_rate": 4.611764705882353e-06,
"loss": 0.5418,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27724453806877136,
"step": 50,
"valid_targets_mean": 4576.5,
"valid_targets_min": 3848
},
{
"epoch": 0.09083402146985962,
"grad_norm": 1.0843204570689695,
"learning_rate": 5.08235294117647e-06,
"loss": 0.5304,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2596698999404907,
"step": 55,
"valid_targets_mean": 4265.2,
"valid_targets_min": 3466
},
{
"epoch": 0.0990916597853014,
"grad_norm": 0.8368432129708421,
"learning_rate": 5.552941176470589e-06,
"loss": 0.5188,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2367270141839981,
"step": 60,
"valid_targets_mean": 4503.9,
"valid_targets_min": 4077
},
{
"epoch": 0.10734929810074319,
"grad_norm": 0.8062411862300828,
"learning_rate": 6.023529411764706e-06,
"loss": 0.4941,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24812698364257812,
"step": 65,
"valid_targets_mean": 4417.8,
"valid_targets_min": 3554
},
{
"epoch": 0.11560693641618497,
"grad_norm": 0.7849235596681929,
"learning_rate": 6.494117647058824e-06,
"loss": 0.48,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26946136355400085,
"step": 70,
"valid_targets_mean": 4754.4,
"valid_targets_min": 2575
},
{
"epoch": 0.12386457473162675,
"grad_norm": 0.7039288853864024,
"learning_rate": 6.964705882352941e-06,
"loss": 0.4493,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21798175573349,
"step": 75,
"valid_targets_mean": 4384.6,
"valid_targets_min": 3752
},
{
"epoch": 0.13212221304706853,
"grad_norm": 0.6786497369106139,
"learning_rate": 7.43529411764706e-06,
"loss": 0.4565,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2901225984096527,
"step": 80,
"valid_targets_mean": 4769.4,
"valid_targets_min": 4246
},
{
"epoch": 0.14037985136251033,
"grad_norm": 0.6696619550681839,
"learning_rate": 7.905882352941176e-06,
"loss": 0.4452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22058208286762238,
"step": 85,
"valid_targets_mean": 4338.8,
"valid_targets_min": 3742
},
{
"epoch": 0.1486374896779521,
"grad_norm": 0.6029919896127592,
"learning_rate": 8.376470588235295e-06,
"loss": 0.4177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1983250081539154,
"step": 90,
"valid_targets_mean": 4312.9,
"valid_targets_min": 2304
},
{
"epoch": 0.1568951279933939,
"grad_norm": 0.590669631002932,
"learning_rate": 8.847058823529413e-06,
"loss": 0.3958,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.200449138879776,
"step": 95,
"valid_targets_mean": 4457.8,
"valid_targets_min": 3511
},
{
"epoch": 0.16515276630883569,
"grad_norm": 0.60960492363256,
"learning_rate": 9.31764705882353e-06,
"loss": 0.3986,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18468210101127625,
"step": 100,
"valid_targets_mean": 4166.4,
"valid_targets_min": 3101
},
{
"epoch": 0.17341040462427745,
"grad_norm": 0.5706571115754389,
"learning_rate": 9.788235294117649e-06,
"loss": 0.399,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20624786615371704,
"step": 105,
"valid_targets_mean": 4492.2,
"valid_targets_min": 3867
},
{
"epoch": 0.18166804293971925,
"grad_norm": 0.594616504669444,
"learning_rate": 1.0258823529411766e-05,
"loss": 0.3923,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15550225973129272,
"step": 110,
"valid_targets_mean": 3577.6,
"valid_targets_min": 1728
},
{
"epoch": 0.18992568125516102,
"grad_norm": 0.6034858128527163,
"learning_rate": 1.0729411764705884e-05,
"loss": 0.3778,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13538461923599243,
"step": 115,
"valid_targets_mean": 3620.6,
"valid_targets_min": 669
},
{
"epoch": 0.1981833195706028,
"grad_norm": 0.5821129684937367,
"learning_rate": 1.1200000000000001e-05,
"loss": 0.3782,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16570690274238586,
"step": 120,
"valid_targets_mean": 3693.8,
"valid_targets_min": 1558
},
{
"epoch": 0.20644095788604458,
"grad_norm": 0.5811849048187077,
"learning_rate": 1.1670588235294118e-05,
"loss": 0.3699,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17445670068264008,
"step": 125,
"valid_targets_mean": 4332.0,
"valid_targets_min": 3546
},
{
"epoch": 0.21469859620148637,
"grad_norm": 0.6849463962289217,
"learning_rate": 1.2141176470588237e-05,
"loss": 0.3638,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17460374534130096,
"step": 130,
"valid_targets_mean": 4239.9,
"valid_targets_min": 2437
},
{
"epoch": 0.22295623451692817,
"grad_norm": 0.6183674655322359,
"learning_rate": 1.2611764705882354e-05,
"loss": 0.3598,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17198729515075684,
"step": 135,
"valid_targets_mean": 4256.5,
"valid_targets_min": 2362
},
{
"epoch": 0.23121387283236994,
"grad_norm": 0.5493967412692456,
"learning_rate": 1.3082352941176472e-05,
"loss": 0.3496,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15696118772029877,
"step": 140,
"valid_targets_mean": 4853.9,
"valid_targets_min": 3447
},
{
"epoch": 0.23947151114781173,
"grad_norm": 0.7257350590384588,
"learning_rate": 1.355294117647059e-05,
"loss": 0.3644,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1798969805240631,
"step": 145,
"valid_targets_mean": 4432.5,
"valid_targets_min": 3044
},
{
"epoch": 0.2477291494632535,
"grad_norm": 0.6475102092398726,
"learning_rate": 1.4023529411764706e-05,
"loss": 0.3529,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1613834649324417,
"step": 150,
"valid_targets_mean": 3535.4,
"valid_targets_min": 402
},
{
"epoch": 0.25598678777869527,
"grad_norm": 0.5669809660340177,
"learning_rate": 1.4494117647058825e-05,
"loss": 0.3416,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18542392551898956,
"step": 155,
"valid_targets_mean": 4411.4,
"valid_targets_min": 3978
},
{
"epoch": 0.26424442609413706,
"grad_norm": 0.610456321088563,
"learning_rate": 1.4964705882352943e-05,
"loss": 0.3444,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.158249169588089,
"step": 160,
"valid_targets_mean": 3969.1,
"valid_targets_min": 1883
},
{
"epoch": 0.27250206440957886,
"grad_norm": 0.614410687004008,
"learning_rate": 1.543529411764706e-05,
"loss": 0.3373,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1547265201807022,
"step": 165,
"valid_targets_mean": 3673.4,
"valid_targets_min": 2014
},
{
"epoch": 0.28075970272502065,
"grad_norm": 0.6281018886269379,
"learning_rate": 1.5905882352941177e-05,
"loss": 0.3343,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21055272221565247,
"step": 170,
"valid_targets_mean": 5029.5,
"valid_targets_min": 2942
},
{
"epoch": 0.28901734104046245,
"grad_norm": 0.5805048580106464,
"learning_rate": 1.6376470588235298e-05,
"loss": 0.3571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15478618443012238,
"step": 175,
"valid_targets_mean": 4390.4,
"valid_targets_min": 1839
},
{
"epoch": 0.2972749793559042,
"grad_norm": 0.6262930921617086,
"learning_rate": 1.6847058823529414e-05,
"loss": 0.3279,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15774977207183838,
"step": 180,
"valid_targets_mean": 4234.1,
"valid_targets_min": 3398
},
{
"epoch": 0.305532617671346,
"grad_norm": 0.6216788136756237,
"learning_rate": 1.731764705882353e-05,
"loss": 0.3419,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16977465152740479,
"step": 185,
"valid_targets_mean": 4396.0,
"valid_targets_min": 3300
},
{
"epoch": 0.3137902559867878,
"grad_norm": 0.6085697944195338,
"learning_rate": 1.778823529411765e-05,
"loss": 0.347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17286333441734314,
"step": 190,
"valid_targets_mean": 4309.9,
"valid_targets_min": 3650
},
{
"epoch": 0.3220478943022296,
"grad_norm": 0.5876917030423758,
"learning_rate": 1.8258823529411765e-05,
"loss": 0.3334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14240354299545288,
"step": 195,
"valid_targets_mean": 3771.1,
"valid_targets_min": 2686
},
{
"epoch": 0.33030553261767137,
"grad_norm": 0.6115250776385859,
"learning_rate": 1.8729411764705886e-05,
"loss": 0.3274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16405156254768372,
"step": 200,
"valid_targets_mean": 4564.9,
"valid_targets_min": 3826
},
{
"epoch": 0.3385631709331131,
"grad_norm": 0.5747263290547863,
"learning_rate": 1.9200000000000003e-05,
"loss": 0.3379,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21425801515579224,
"step": 205,
"valid_targets_mean": 5265.6,
"valid_targets_min": 3079
},
{
"epoch": 0.3468208092485549,
"grad_norm": 0.66060691293213,
"learning_rate": 1.967058823529412e-05,
"loss": 0.3291,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16240888833999634,
"step": 210,
"valid_targets_mean": 4253.0,
"valid_targets_min": 3479
},
{
"epoch": 0.3550784475639967,
"grad_norm": 0.6507541774806247,
"learning_rate": 2.0141176470588236e-05,
"loss": 0.3246,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17824137210845947,
"step": 215,
"valid_targets_mean": 4602.9,
"valid_targets_min": 3855
},
{
"epoch": 0.3633360858794385,
"grad_norm": 0.605845509674343,
"learning_rate": 2.0611764705882353e-05,
"loss": 0.3214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17864280939102173,
"step": 220,
"valid_targets_mean": 4649.8,
"valid_targets_min": 4367
},
{
"epoch": 0.37159372419488024,
"grad_norm": 0.5865789140074175,
"learning_rate": 2.1082352941176474e-05,
"loss": 0.3155,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1578708291053772,
"step": 225,
"valid_targets_mean": 4562.1,
"valid_targets_min": 3920
},
{
"epoch": 0.37985136251032203,
"grad_norm": 0.6521829724452352,
"learning_rate": 2.155294117647059e-05,
"loss": 0.3175,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1891648769378662,
"step": 230,
"valid_targets_mean": 4475.2,
"valid_targets_min": 3622
},
{
"epoch": 0.38810900082576383,
"grad_norm": 0.6362813829340013,
"learning_rate": 2.2023529411764707e-05,
"loss": 0.3243,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1538279801607132,
"step": 235,
"valid_targets_mean": 4105.6,
"valid_targets_min": 3270
},
{
"epoch": 0.3963666391412056,
"grad_norm": 0.5683505276319541,
"learning_rate": 2.2494117647058828e-05,
"loss": 0.3262,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17053112387657166,
"step": 240,
"valid_targets_mean": 4159.6,
"valid_targets_min": 1824
},
{
"epoch": 0.4046242774566474,
"grad_norm": 0.6044292377854251,
"learning_rate": 2.296470588235294e-05,
"loss": 0.3352,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18530462682247162,
"step": 245,
"valid_targets_mean": 4472.9,
"valid_targets_min": 2623
},
{
"epoch": 0.41288191577208916,
"grad_norm": 1.0306335022062794,
"learning_rate": 2.343529411764706e-05,
"loss": 0.3177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14189110696315765,
"step": 250,
"valid_targets_mean": 4193.1,
"valid_targets_min": 3067
},
{
"epoch": 0.42113955408753095,
"grad_norm": 0.7343237018853076,
"learning_rate": 2.390588235294118e-05,
"loss": 0.3113,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1477409303188324,
"step": 255,
"valid_targets_mean": 4398.6,
"valid_targets_min": 3993
},
{
"epoch": 0.42939719240297275,
"grad_norm": 0.656629873794441,
"learning_rate": 2.4376470588235296e-05,
"loss": 0.31,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13981479406356812,
"step": 260,
"valid_targets_mean": 3929.2,
"valid_targets_min": 1896
},
{
"epoch": 0.43765483071841454,
"grad_norm": 0.6042459120534337,
"learning_rate": 2.4847058823529416e-05,
"loss": 0.3117,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17179864645004272,
"step": 265,
"valid_targets_mean": 4701.2,
"valid_targets_min": 3885
},
{
"epoch": 0.44591246903385634,
"grad_norm": 0.7336198481761635,
"learning_rate": 2.5317647058823533e-05,
"loss": 0.3193,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11718828231096268,
"step": 270,
"valid_targets_mean": 3518.9,
"valid_targets_min": 1477
},
{
"epoch": 0.4541701073492981,
"grad_norm": 0.6049728789443988,
"learning_rate": 2.5788235294117646e-05,
"loss": 0.3125,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14982685446739197,
"step": 275,
"valid_targets_mean": 4643.0,
"valid_targets_min": 3800
},
{
"epoch": 0.4624277456647399,
"grad_norm": 0.6693542858188497,
"learning_rate": 2.6258823529411767e-05,
"loss": 0.3145,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14874421060085297,
"step": 280,
"valid_targets_mean": 3763.1,
"valid_targets_min": 2008
},
{
"epoch": 0.47068538398018167,
"grad_norm": 0.6226521528119886,
"learning_rate": 2.6729411764705884e-05,
"loss": 0.3229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15412822365760803,
"step": 285,
"valid_targets_mean": 4071.8,
"valid_targets_min": 1111
},
{
"epoch": 0.47894302229562347,
"grad_norm": 0.6037789203027593,
"learning_rate": 2.7200000000000004e-05,
"loss": 0.3214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15540811419487,
"step": 290,
"valid_targets_mean": 4198.5,
"valid_targets_min": 3295
},
{
"epoch": 0.48720066061106526,
"grad_norm": 0.5813994400197731,
"learning_rate": 2.767058823529412e-05,
"loss": 0.3077,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15485572814941406,
"step": 295,
"valid_targets_mean": 4370.9,
"valid_targets_min": 3989
},
{
"epoch": 0.495458298926507,
"grad_norm": 0.666355595659689,
"learning_rate": 2.8141176470588238e-05,
"loss": 0.304,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11822428554296494,
"step": 300,
"valid_targets_mean": 3068.1,
"valid_targets_min": 995
},
{
"epoch": 0.5037159372419489,
"grad_norm": 0.6215682191849712,
"learning_rate": 2.8611764705882358e-05,
"loss": 0.3177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.170493483543396,
"step": 305,
"valid_targets_mean": 4355.0,
"valid_targets_min": 3360
},
{
"epoch": 0.5119735755573905,
"grad_norm": 1.0498665785917747,
"learning_rate": 2.908235294117647e-05,
"loss": 0.3134,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17342056334018707,
"step": 310,
"valid_targets_mean": 4591.2,
"valid_targets_min": 4154
},
{
"epoch": 0.5202312138728323,
"grad_norm": 0.630253723588536,
"learning_rate": 2.955294117647059e-05,
"loss": 0.3032,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14559635519981384,
"step": 315,
"valid_targets_mean": 3953.0,
"valid_targets_min": 1782
},
{
"epoch": 0.5284888521882741,
"grad_norm": 0.5972753711513472,
"learning_rate": 3.002352941176471e-05,
"loss": 0.3003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1728266477584839,
"step": 320,
"valid_targets_mean": 4717.0,
"valid_targets_min": 3879
},
{
"epoch": 0.5367464905037159,
"grad_norm": 0.5910751651941389,
"learning_rate": 3.0494117647058826e-05,
"loss": 0.3041,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16499464213848114,
"step": 325,
"valid_targets_mean": 4895.4,
"valid_targets_min": 4193
},
{
"epoch": 0.5450041288191577,
"grad_norm": 0.5856277774624803,
"learning_rate": 3.0964705882352946e-05,
"loss": 0.3028,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14935341477394104,
"step": 330,
"valid_targets_mean": 4262.1,
"valid_targets_min": 2163
},
{
"epoch": 0.5532617671345995,
"grad_norm": 0.6347865959392145,
"learning_rate": 3.143529411764706e-05,
"loss": 0.2985,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14569245278835297,
"step": 335,
"valid_targets_mean": 4546.1,
"valid_targets_min": 2203
},
{
"epoch": 0.5615194054500413,
"grad_norm": 0.5823324049614251,
"learning_rate": 3.190588235294118e-05,
"loss": 0.3087,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15887591242790222,
"step": 340,
"valid_targets_mean": 4407.2,
"valid_targets_min": 3684
},
{
"epoch": 0.5697770437654831,
"grad_norm": 0.6003250210204946,
"learning_rate": 3.23764705882353e-05,
"loss": 0.311,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14752474427223206,
"step": 345,
"valid_targets_mean": 4430.6,
"valid_targets_min": 3348
},
{
"epoch": 0.5780346820809249,
"grad_norm": 0.5969177438468508,
"learning_rate": 3.2847058823529414e-05,
"loss": 0.3101,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17232048511505127,
"step": 350,
"valid_targets_mean": 4459.9,
"valid_targets_min": 3330
},
{
"epoch": 0.5862923203963666,
"grad_norm": 0.6538575533682767,
"learning_rate": 3.331764705882353e-05,
"loss": 0.3077,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16106244921684265,
"step": 355,
"valid_targets_mean": 4526.4,
"valid_targets_min": 3556
},
{
"epoch": 0.5945499587118084,
"grad_norm": 0.6533574014469453,
"learning_rate": 3.378823529411765e-05,
"loss": 0.3052,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1500328928232193,
"step": 360,
"valid_targets_mean": 4197.2,
"valid_targets_min": 2860
},
{
"epoch": 0.6028075970272502,
"grad_norm": 0.5966011855735348,
"learning_rate": 3.425882352941177e-05,
"loss": 0.2946,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16375574469566345,
"step": 365,
"valid_targets_mean": 5028.6,
"valid_targets_min": 3754
},
{
"epoch": 0.611065235342692,
"grad_norm": 0.5951525244049103,
"learning_rate": 3.472941176470589e-05,
"loss": 0.3079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14389994740486145,
"step": 370,
"valid_targets_mean": 3986.2,
"valid_targets_min": 1727
},
{
"epoch": 0.6193228736581338,
"grad_norm": 0.5798790733286899,
"learning_rate": 3.52e-05,
"loss": 0.3048,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12721113860607147,
"step": 375,
"valid_targets_mean": 4538.0,
"valid_targets_min": 3557
},
{
"epoch": 0.6275805119735756,
"grad_norm": 0.6268723001684364,
"learning_rate": 3.567058823529412e-05,
"loss": 0.2906,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14850017428398132,
"step": 380,
"valid_targets_mean": 4565.5,
"valid_targets_min": 3495
},
{
"epoch": 0.6358381502890174,
"grad_norm": 0.5554322896776761,
"learning_rate": 3.614117647058824e-05,
"loss": 0.2932,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14433631300926208,
"step": 385,
"valid_targets_mean": 4520.1,
"valid_targets_min": 3828
},
{
"epoch": 0.6440957886044592,
"grad_norm": 0.6080247267252055,
"learning_rate": 3.6611764705882356e-05,
"loss": 0.2886,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14060217142105103,
"step": 390,
"valid_targets_mean": 4163.4,
"valid_targets_min": 3448
},
{
"epoch": 0.652353426919901,
"grad_norm": 0.5753963785972113,
"learning_rate": 3.708235294117647e-05,
"loss": 0.2992,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14996804296970367,
"step": 395,
"valid_targets_mean": 4634.0,
"valid_targets_min": 3700
},
{
"epoch": 0.6606110652353427,
"grad_norm": 0.7241917342625542,
"learning_rate": 3.755294117647059e-05,
"loss": 0.2985,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15200330317020416,
"step": 400,
"valid_targets_mean": 4604.9,
"valid_targets_min": 2797
},
{
"epoch": 0.6688687035507844,
"grad_norm": 0.5754861829078405,
"learning_rate": 3.802352941176471e-05,
"loss": 0.2945,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14571356773376465,
"step": 405,
"valid_targets_mean": 4339.1,
"valid_targets_min": 3894
},
{
"epoch": 0.6771263418662262,
"grad_norm": 0.538164966710115,
"learning_rate": 3.849411764705883e-05,
"loss": 0.2947,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.141482412815094,
"step": 410,
"valid_targets_mean": 4422.2,
"valid_targets_min": 3840
},
{
"epoch": 0.685383980181668,
"grad_norm": 0.5713695449792673,
"learning_rate": 3.8964705882352944e-05,
"loss": 0.2898,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1560671627521515,
"step": 415,
"valid_targets_mean": 4909.4,
"valid_targets_min": 4070
},
{
"epoch": 0.6936416184971098,
"grad_norm": 0.6037753427441735,
"learning_rate": 3.943529411764706e-05,
"loss": 0.2916,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14155718684196472,
"step": 420,
"valid_targets_mean": 4338.1,
"valid_targets_min": 3621
},
{
"epoch": 0.7018992568125516,
"grad_norm": 0.6041318715249562,
"learning_rate": 3.990588235294118e-05,
"loss": 0.2843,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13594946265220642,
"step": 425,
"valid_targets_mean": 3771.1,
"valid_targets_min": 1780
},
{
"epoch": 0.7101568951279934,
"grad_norm": 0.530436548869443,
"learning_rate": 3.9999891613544045e-05,
"loss": 0.2947,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1394493132829666,
"step": 430,
"valid_targets_mean": 4361.0,
"valid_targets_min": 3383
},
{
"epoch": 0.7184145334434352,
"grad_norm": 0.6543019115407284,
"learning_rate": 3.999945129558009e-05,
"loss": 0.2995,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14561928808689117,
"step": 435,
"valid_targets_mean": 4153.9,
"valid_targets_min": 2318
},
{
"epoch": 0.726672171758877,
"grad_norm": 0.5577829586031012,
"learning_rate": 3.9998672279405856e-05,
"loss": 0.2995,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11069948971271515,
"step": 440,
"valid_targets_mean": 3995.2,
"valid_targets_min": 3303
},
{
"epoch": 0.7349298100743188,
"grad_norm": 0.5633189088273238,
"learning_rate": 3.999755457821429e-05,
"loss": 0.2915,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14281581342220306,
"step": 445,
"valid_targets_mean": 4197.6,
"valid_targets_min": 3096
},
{
"epoch": 0.7431874483897605,
"grad_norm": 0.5765018970661508,
"learning_rate": 3.999609821093405e-05,
"loss": 0.2942,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1458982229232788,
"step": 450,
"valid_targets_mean": 4230.1,
"valid_targets_min": 2961
},
{
"epoch": 0.7514450867052023,
"grad_norm": 0.6145081435690086,
"learning_rate": 3.9994303202229285e-05,
"loss": 0.2988,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17397338151931763,
"step": 455,
"valid_targets_mean": 4965.8,
"valid_targets_min": 3641
},
{
"epoch": 0.7597027250206441,
"grad_norm": 0.7873051614931993,
"learning_rate": 3.999216958249914e-05,
"loss": 0.2923,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13865745067596436,
"step": 460,
"valid_targets_mean": 4106.9,
"valid_targets_min": 1269
},
{
"epoch": 0.7679603633360859,
"grad_norm": 0.5666368626504148,
"learning_rate": 3.9989697387877264e-05,
"loss": 0.2942,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14277788996696472,
"step": 465,
"valid_targets_mean": 4504.1,
"valid_targets_min": 3606
},
{
"epoch": 0.7762180016515277,
"grad_norm": 0.5729132320453892,
"learning_rate": 3.9986886660231184e-05,
"loss": 0.2848,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14766156673431396,
"step": 470,
"valid_targets_mean": 4833.9,
"valid_targets_min": 2544
},
{
"epoch": 0.7844756399669695,
"grad_norm": 0.6442273666440981,
"learning_rate": 3.998373744716165e-05,
"loss": 0.2772,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12925535440444946,
"step": 475,
"valid_targets_mean": 4583.6,
"valid_targets_min": 3298
},
{
"epoch": 0.7927332782824112,
"grad_norm": 0.703738315931259,
"learning_rate": 3.9980249802001756e-05,
"loss": 0.2922,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1205299124121666,
"step": 480,
"valid_targets_mean": 4004.1,
"valid_targets_min": 2201
},
{
"epoch": 0.800990916597853,
"grad_norm": 0.5421068724240508,
"learning_rate": 3.997642378381608e-05,
"loss": 0.2849,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13460440933704376,
"step": 485,
"valid_targets_mean": 4209.1,
"valid_targets_min": 1974
},
{
"epoch": 0.8092485549132948,
"grad_norm": 0.6557443686009133,
"learning_rate": 3.997225945739968e-05,
"loss": 0.291,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11940011382102966,
"step": 490,
"valid_targets_mean": 4061.6,
"valid_targets_min": 3155
},
{
"epoch": 0.8175061932287366,
"grad_norm": 0.5717077717590093,
"learning_rate": 3.996775689327698e-05,
"loss": 0.2893,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13545432686805725,
"step": 495,
"valid_targets_mean": 3886.5,
"valid_targets_min": 2835
},
{
"epoch": 0.8257638315441783,
"grad_norm": 0.5626132976802045,
"learning_rate": 3.99629161677006e-05,
"loss": 0.2887,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15409281849861145,
"step": 500,
"valid_targets_mean": 4993.1,
"valid_targets_min": 3793
},
{
"epoch": 0.8340214698596201,
"grad_norm": 0.5894813856763842,
"learning_rate": 3.9957737362650034e-05,
"loss": 0.2798,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12317012250423431,
"step": 505,
"valid_targets_mean": 3887.1,
"valid_targets_min": 1915
},
{
"epoch": 0.8422791081750619,
"grad_norm": 0.5094534125706007,
"learning_rate": 3.99522205658303e-05,
"loss": 0.286,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1142788752913475,
"step": 510,
"valid_targets_mean": 3940.6,
"valid_targets_min": 1523
},
{
"epoch": 0.8505367464905037,
"grad_norm": 0.5761894461521411,
"learning_rate": 3.9946365870670417e-05,
"loss": 0.2896,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1313384771347046,
"step": 515,
"valid_targets_mean": 4390.0,
"valid_targets_min": 3619
},
{
"epoch": 0.8587943848059455,
"grad_norm": 0.592872129637009,
"learning_rate": 3.994017337632184e-05,
"loss": 0.2828,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12142593413591385,
"step": 520,
"valid_targets_mean": 3592.5,
"valid_targets_min": 2292
},
{
"epoch": 0.8670520231213873,
"grad_norm": 0.587626646180235,
"learning_rate": 3.9933643187656786e-05,
"loss": 0.2847,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.177256777882576,
"step": 525,
"valid_targets_mean": 4611.9,
"valid_targets_min": 3488
},
{
"epoch": 0.8753096614368291,
"grad_norm": 0.533989725973806,
"learning_rate": 3.992677541526645e-05,
"loss": 0.2891,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13841873407363892,
"step": 530,
"valid_targets_mean": 4837.2,
"valid_targets_min": 4013
},
{
"epoch": 0.8835672997522709,
"grad_norm": 0.5480739998340078,
"learning_rate": 3.9919570175459136e-05,
"loss": 0.2837,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12190024554729462,
"step": 535,
"valid_targets_mean": 4241.9,
"valid_targets_min": 2501
},
{
"epoch": 0.8918249380677127,
"grad_norm": 0.535061280283479,
"learning_rate": 3.991202759025828e-05,
"loss": 0.2852,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14882051944732666,
"step": 540,
"valid_targets_mean": 4843.4,
"valid_targets_min": 3537
},
{
"epoch": 0.9000825763831544,
"grad_norm": 0.5565806715554168,
"learning_rate": 3.990414778740038e-05,
"loss": 0.2809,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1481572687625885,
"step": 545,
"valid_targets_mean": 4560.6,
"valid_targets_min": 3965
},
{
"epoch": 0.9083402146985962,
"grad_norm": 0.573864841360069,
"learning_rate": 3.989593090033288e-05,
"loss": 0.2795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15470004081726074,
"step": 550,
"valid_targets_mean": 4821.4,
"valid_targets_min": 3723
},
{
"epoch": 0.916597853014038,
"grad_norm": 0.5189863655272903,
"learning_rate": 3.988737706821184e-05,
"loss": 0.2768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14931809902191162,
"step": 555,
"valid_targets_mean": 4658.0,
"valid_targets_min": 3932
},
{
"epoch": 0.9248554913294798,
"grad_norm": 0.5023290454666544,
"learning_rate": 3.987848643589959e-05,
"loss": 0.2885,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16054192185401917,
"step": 560,
"valid_targets_mean": 5291.6,
"valid_targets_min": 3225
},
{
"epoch": 0.9331131296449215,
"grad_norm": 0.5822792139831601,
"learning_rate": 3.986925915396235e-05,
"loss": 0.2928,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14044755697250366,
"step": 565,
"valid_targets_mean": 4809.6,
"valid_targets_min": 3989
},
{
"epoch": 0.9413707679603633,
"grad_norm": 0.6505229114376824,
"learning_rate": 3.9859695378667584e-05,
"loss": 0.2769,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15827780961990356,
"step": 570,
"valid_targets_mean": 4538.4,
"valid_targets_min": 3589
},
{
"epoch": 0.9496284062758051,
"grad_norm": 0.5474962446696429,
"learning_rate": 3.984979527198143e-05,
"loss": 0.2724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13671788573265076,
"step": 575,
"valid_targets_mean": 4300.5,
"valid_targets_min": 3745
},
{
"epoch": 0.9578860445912469,
"grad_norm": 0.5481918762835275,
"learning_rate": 3.98395590015659e-05,
"loss": 0.2811,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14504030346870422,
"step": 580,
"valid_targets_mean": 4552.9,
"valid_targets_min": 2986
},
{
"epoch": 0.9661436829066887,
"grad_norm": 0.5156744523249689,
"learning_rate": 3.982898674077607e-05,
"loss": 0.2757,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12438619136810303,
"step": 585,
"valid_targets_mean": 3868.4,
"valid_targets_min": 1519
},
{
"epoch": 0.9744013212221305,
"grad_norm": 0.5433063101284344,
"learning_rate": 3.981807866865715e-05,
"loss": 0.2844,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14425161480903625,
"step": 590,
"valid_targets_mean": 4044.4,
"valid_targets_min": 1415
},
{
"epoch": 0.9826589595375722,
"grad_norm": 0.5809382790402149,
"learning_rate": 3.980683496994143e-05,
"loss": 0.2859,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1343752145767212,
"step": 595,
"valid_targets_mean": 4221.9,
"valid_targets_min": 3500
},
{
"epoch": 0.990916597853014,
"grad_norm": 0.5174489797026075,
"learning_rate": 3.9795255835045166e-05,
"loss": 0.2931,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1453278362751007,
"step": 600,
"valid_targets_mean": 4489.8,
"valid_targets_min": 3621
},
{
"epoch": 0.9991742361684558,
"grad_norm": 0.525396668668542,
"learning_rate": 3.9783341460065343e-05,
"loss": 0.281,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15435823798179626,
"step": 605,
"valid_targets_mean": 4458.6,
"valid_targets_min": 3457
},
{
"epoch": 1.0066061106523534,
"grad_norm": 0.5377285807988771,
"learning_rate": 3.977109204677639e-05,
"loss": 0.2681,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.123194620013237,
"step": 610,
"valid_targets_mean": 4110.0,
"valid_targets_min": 2001
},
{
"epoch": 1.0148637489677952,
"grad_norm": 0.5434405094382597,
"learning_rate": 3.9758507802626704e-05,
"loss": 0.2727,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1259099841117859,
"step": 615,
"valid_targets_mean": 4031.2,
"valid_targets_min": 1904
},
{
"epoch": 1.023121387283237,
"grad_norm": 0.5853900624345613,
"learning_rate": 3.9745588940735176e-05,
"loss": 0.2768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13724340498447418,
"step": 620,
"valid_targets_mean": 4493.4,
"valid_targets_min": 4306
},
{
"epoch": 1.0313790255986788,
"grad_norm": 0.5927708558459054,
"learning_rate": 3.973233567988759e-05,
"loss": 0.266,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1297268271446228,
"step": 625,
"valid_targets_mean": 3983.6,
"valid_targets_min": 1280
},
{
"epoch": 1.0396366639141206,
"grad_norm": 0.5439698699641045,
"learning_rate": 3.9718748244532883e-05,
"loss": 0.2811,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13257832825183868,
"step": 630,
"valid_targets_mean": 4244.6,
"valid_targets_min": 2905
},
{
"epoch": 1.0478943022295624,
"grad_norm": 0.5654156605592788,
"learning_rate": 3.970482686477937e-05,
"loss": 0.2664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16105100512504578,
"step": 635,
"valid_targets_mean": 4653.4,
"valid_targets_min": 3583
},
{
"epoch": 1.0561519405450042,
"grad_norm": 0.5111995569028033,
"learning_rate": 3.969057177639084e-05,
"loss": 0.2707,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1316896677017212,
"step": 640,
"valid_targets_mean": 4790.2,
"valid_targets_min": 3776
},
{
"epoch": 1.064409578860446,
"grad_norm": 0.5740777040779244,
"learning_rate": 3.967598322078257e-05,
"loss": 0.2779,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1341586709022522,
"step": 645,
"valid_targets_mean": 4339.4,
"valid_targets_min": 3568
},
{
"epoch": 1.0726672171758878,
"grad_norm": 0.5973535625101994,
"learning_rate": 3.966106144501721e-05,
"loss": 0.2687,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12249834090471268,
"step": 650,
"valid_targets_mean": 3968.8,
"valid_targets_min": 2351
},
{
"epoch": 1.0809248554913296,
"grad_norm": 0.4853112125894759,
"learning_rate": 3.964580670180063e-05,
"loss": 0.2664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12538987398147583,
"step": 655,
"valid_targets_mean": 4294.0,
"valid_targets_min": 3643
},
{
"epoch": 1.0891824938067713,
"grad_norm": 0.5488639577576694,
"learning_rate": 3.9630219249477655e-05,
"loss": 0.2715,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14338180422782898,
"step": 660,
"valid_targets_mean": 4154.4,
"valid_targets_min": 2950
},
{
"epoch": 1.0974401321222131,
"grad_norm": 0.5164322941110284,
"learning_rate": 3.96142993520276e-05,
"loss": 0.2758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15259218215942383,
"step": 665,
"valid_targets_mean": 4890.5,
"valid_targets_min": 3372
},
{
"epoch": 1.1056977704376547,
"grad_norm": 0.5602264570375924,
"learning_rate": 3.959804727905992e-05,
"loss": 0.2664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1278294026851654,
"step": 670,
"valid_targets_mean": 4131.5,
"valid_targets_min": 3420
},
{
"epoch": 1.1139554087530965,
"grad_norm": 0.5249022508043477,
"learning_rate": 3.9581463305809576e-05,
"loss": 0.2775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11901842057704926,
"step": 675,
"valid_targets_mean": 4134.4,
"valid_targets_min": 3312
},
{
"epoch": 1.1222130470685383,
"grad_norm": 0.5122945796009134,
"learning_rate": 3.956454771313236e-05,
"loss": 0.2614,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13165123760700226,
"step": 680,
"valid_targets_mean": 4184.8,
"valid_targets_min": 3427
},
{
"epoch": 1.13047068538398,
"grad_norm": 0.5129554054285936,
"learning_rate": 3.954730078750018e-05,
"loss": 0.2731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11850960552692413,
"step": 685,
"valid_targets_mean": 4370.5,
"valid_targets_min": 3812
},
{
"epoch": 1.138728323699422,
"grad_norm": 0.5283644926267766,
"learning_rate": 3.95297228209962e-05,
"loss": 0.2762,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13979145884513855,
"step": 690,
"valid_targets_mean": 4110.9,
"valid_targets_min": 3291
},
{
"epoch": 1.1469859620148637,
"grad_norm": 0.5202182858162632,
"learning_rate": 3.9511814111309875e-05,
"loss": 0.2692,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12166427075862885,
"step": 695,
"valid_targets_mean": 3915.0,
"valid_targets_min": 2463
},
{
"epoch": 1.1552436003303055,
"grad_norm": 0.5225622459843526,
"learning_rate": 3.949357496173192e-05,
"loss": 0.2667,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14088931679725647,
"step": 700,
"valid_targets_mean": 4847.1,
"valid_targets_min": 4036
},
{
"epoch": 1.1635012386457473,
"grad_norm": 0.5134483932606695,
"learning_rate": 3.947500568114918e-05,
"loss": 0.2724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13691473007202148,
"step": 705,
"valid_targets_mean": 4700.0,
"valid_targets_min": 4022
},
{
"epoch": 1.171758876961189,
"grad_norm": 0.5463536560238659,
"learning_rate": 3.945610658403938e-05,
"loss": 0.2698,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12929245829582214,
"step": 710,
"valid_targets_mean": 4458.9,
"valid_targets_min": 4049
},
{
"epoch": 1.1800165152766309,
"grad_norm": 0.49537987364644726,
"learning_rate": 3.9436877990465826e-05,
"loss": 0.2683,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13547129929065704,
"step": 715,
"valid_targets_mean": 4233.1,
"valid_targets_min": 3565
},
{
"epoch": 1.1882741535920727,
"grad_norm": 0.5393839195675705,
"learning_rate": 3.9417320226071956e-05,
"loss": 0.271,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15521205961704254,
"step": 720,
"valid_targets_mean": 4115.5,
"valid_targets_min": 781
},
{
"epoch": 1.1965317919075145,
"grad_norm": 0.49224363644153335,
"learning_rate": 3.939743362207586e-05,
"loss": 0.2644,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13274291157722473,
"step": 725,
"valid_targets_mean": 4692.2,
"valid_targets_min": 3721
},
{
"epoch": 1.2047894302229563,
"grad_norm": 0.48075378903775595,
"learning_rate": 3.937721851526462e-05,
"loss": 0.2581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12308897078037262,
"step": 730,
"valid_targets_mean": 4324.1,
"valid_targets_min": 3255
},
{
"epoch": 1.213047068538398,
"grad_norm": 0.4993336191456028,
"learning_rate": 3.935667524798866e-05,
"loss": 0.2721,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13531145453453064,
"step": 735,
"valid_targets_mean": 4808.1,
"valid_targets_min": 4293
},
{
"epoch": 1.2213047068538398,
"grad_norm": 0.48209866825286096,
"learning_rate": 3.933580416815593e-05,
"loss": 0.2607,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12871845066547394,
"step": 740,
"valid_targets_mean": 4806.1,
"valid_targets_min": 4317
},
{
"epoch": 1.2295623451692816,
"grad_norm": 0.5093088687299417,
"learning_rate": 3.931460562922598e-05,
"loss": 0.2729,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13394972681999207,
"step": 745,
"valid_targets_mean": 4142.1,
"valid_targets_min": 3440
},
{
"epoch": 1.2378199834847234,
"grad_norm": 0.5317082271183605,
"learning_rate": 3.9293079990204034e-05,
"loss": 0.2767,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11608986556529999,
"step": 750,
"valid_targets_mean": 3686.0,
"valid_targets_min": 1261
},
{
"epoch": 1.2460776218001652,
"grad_norm": 0.5338158595418969,
"learning_rate": 3.927122761563485e-05,
"loss": 0.2726,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13573448359966278,
"step": 755,
"valid_targets_mean": 3960.1,
"valid_targets_min": 2208
},
{
"epoch": 1.254335260115607,
"grad_norm": 0.4949707560512785,
"learning_rate": 3.9249048875596593e-05,
"loss": 0.278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13002905249595642,
"step": 760,
"valid_targets_mean": 4323.4,
"valid_targets_min": 2014
},
{
"epoch": 1.2625928984310488,
"grad_norm": 0.522721264427305,
"learning_rate": 3.922654414569455e-05,
"loss": 0.2699,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12254970520734787,
"step": 765,
"valid_targets_mean": 4026.4,
"valid_targets_min": 2069
},
{
"epoch": 1.2708505367464906,
"grad_norm": 0.4874481210534528,
"learning_rate": 3.9203713807054755e-05,
"loss": 0.2631,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1266925036907196,
"step": 770,
"valid_targets_mean": 4711.5,
"valid_targets_min": 2739
},
{
"epoch": 1.2791081750619324,
"grad_norm": 0.48441216519686536,
"learning_rate": 3.918055824631757e-05,
"loss": 0.2585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10997404158115387,
"step": 775,
"valid_targets_mean": 3919.9,
"valid_targets_min": 3097
},
{
"epoch": 1.287365813377374,
"grad_norm": 0.48651108153223493,
"learning_rate": 3.915707785563109e-05,
"loss": 0.269,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10929493606090546,
"step": 780,
"valid_targets_mean": 3999.6,
"valid_targets_min": 3205
},
{
"epoch": 1.2956234516928158,
"grad_norm": 0.4690418979959962,
"learning_rate": 3.913327303264456e-05,
"loss": 0.2716,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15855711698532104,
"step": 785,
"valid_targets_mean": 5348.9,
"valid_targets_min": 3471
},
{
"epoch": 1.3038810900082576,
"grad_norm": 0.5091692667324333,
"learning_rate": 3.9109144180501564e-05,
"loss": 0.266,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15191052854061127,
"step": 790,
"valid_targets_mean": 4555.6,
"valid_targets_min": 3867
},
{
"epoch": 1.3121387283236994,
"grad_norm": 0.4861540542484402,
"learning_rate": 3.908469170783328e-05,
"loss": 0.2651,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14065532386302948,
"step": 795,
"valid_targets_mean": 4531.2,
"valid_targets_min": 3990
},
{
"epoch": 1.3203963666391412,
"grad_norm": 0.5361323197019161,
"learning_rate": 3.9059916028751496e-05,
"loss": 0.2722,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1507452428340912,
"step": 800,
"valid_targets_mean": 4768.9,
"valid_targets_min": 2720
},
{
"epoch": 1.328654004954583,
"grad_norm": 0.4724411289276464,
"learning_rate": 3.903481756284164e-05,
"loss": 0.2591,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11483338475227356,
"step": 805,
"valid_targets_mean": 4334.9,
"valid_targets_min": 3622
},
{
"epoch": 1.3369116432700248,
"grad_norm": 0.4983151966336367,
"learning_rate": 3.900939673515564e-05,
"loss": 0.2707,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15205402672290802,
"step": 810,
"valid_targets_mean": 4622.8,
"valid_targets_min": 3759
},
{
"epoch": 1.3451692815854666,
"grad_norm": 0.4509378616820213,
"learning_rate": 3.898365397620475e-05,
"loss": 0.2659,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13023245334625244,
"step": 815,
"valid_targets_mean": 4614.8,
"valid_targets_min": 3761
},
{
"epoch": 1.3534269199009084,
"grad_norm": 0.47778439401950346,
"learning_rate": 3.8957589721952254e-05,
"loss": 0.2689,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13775765895843506,
"step": 820,
"valid_targets_mean": 4986.0,
"valid_targets_min": 3824
},
{
"epoch": 1.3616845582163501,
"grad_norm": 0.5085444924521247,
"learning_rate": 3.8931204413806076e-05,
"loss": 0.2762,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14558349549770355,
"step": 825,
"valid_targets_mean": 4562.6,
"valid_targets_min": 3828
},
{
"epoch": 1.369942196531792,
"grad_norm": 0.488521503161596,
"learning_rate": 3.8904498498611306e-05,
"loss": 0.2653,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.153921440243721,
"step": 830,
"valid_targets_mean": 4524.0,
"valid_targets_min": 3486
},
{
"epoch": 1.3781998348472337,
"grad_norm": 0.4805774895972684,
"learning_rate": 3.8877472428642634e-05,
"loss": 0.2698,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12913748621940613,
"step": 835,
"valid_targets_mean": 4389.9,
"valid_targets_min": 2766
},
{
"epoch": 1.3864574731626755,
"grad_norm": 0.48480759061687784,
"learning_rate": 3.885012666159669e-05,
"loss": 0.2614,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12615995109081268,
"step": 840,
"valid_targets_mean": 4357.5,
"valid_targets_min": 3352
},
{
"epoch": 1.3947151114781173,
"grad_norm": 0.4989342275174894,
"learning_rate": 3.8822461660584315e-05,
"loss": 0.2575,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14129380881786346,
"step": 845,
"valid_targets_mean": 4315.8,
"valid_targets_min": 3221
},
{
"epoch": 1.402972749793559,
"grad_norm": 0.49777588910404974,
"learning_rate": 3.8794477894122666e-05,
"loss": 0.2724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12208463251590729,
"step": 850,
"valid_targets_mean": 4067.4,
"valid_targets_min": 3201
},
{
"epoch": 1.4112303881090007,
"grad_norm": 0.4605303981782248,
"learning_rate": 3.8766175836127323e-05,
"loss": 0.2687,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13396482169628143,
"step": 855,
"valid_targets_mean": 4450.1,
"valid_targets_min": 3832
},
{
"epoch": 1.4194880264244425,
"grad_norm": 0.467896435210906,
"learning_rate": 3.873755596590426e-05,
"loss": 0.2611,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11965446174144745,
"step": 860,
"valid_targets_mean": 3988.8,
"valid_targets_min": 1563
},
{
"epoch": 1.4277456647398843,
"grad_norm": 0.46479791896075456,
"learning_rate": 3.8708618768141715e-05,
"loss": 0.2593,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12830713391304016,
"step": 865,
"valid_targets_mean": 4201.5,
"valid_targets_min": 3527
},
{
"epoch": 1.436003303055326,
"grad_norm": 0.4693831641307115,
"learning_rate": 3.867936473290199e-05,
"loss": 0.2724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13127730786800385,
"step": 870,
"valid_targets_mean": 4881.4,
"valid_targets_min": 3896
},
{
"epoch": 1.4442609413707679,
"grad_norm": 0.46985456686845517,
"learning_rate": 3.8649794355613136e-05,
"loss": 0.2647,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12574949860572815,
"step": 875,
"valid_targets_mean": 4206.5,
"valid_targets_min": 2691
},
{
"epoch": 1.4525185796862097,
"grad_norm": 0.45226300326417596,
"learning_rate": 3.8619908137060605e-05,
"loss": 0.2676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11677674949169159,
"step": 880,
"valid_targets_mean": 4291.6,
"valid_targets_min": 3566
},
{
"epoch": 1.4607762180016515,
"grad_norm": 0.49969791082084947,
"learning_rate": 3.85897065833787e-05,
"loss": 0.2622,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12426415085792542,
"step": 885,
"valid_targets_mean": 4524.9,
"valid_targets_min": 4104
},
{
"epoch": 1.4690338563170933,
"grad_norm": 0.4824969924662978,
"learning_rate": 3.855919020604207e-05,
"loss": 0.2728,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16476695239543915,
"step": 890,
"valid_targets_mean": 5433.5,
"valid_targets_min": 777
},
{
"epoch": 1.477291494632535,
"grad_norm": 0.4869394052854892,
"learning_rate": 3.852835952185702e-05,
"loss": 0.2694,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12065634876489639,
"step": 895,
"valid_targets_mean": 4755.6,
"valid_targets_min": 4445
},
{
"epoch": 1.4855491329479769,
"grad_norm": 0.47107408516497784,
"learning_rate": 3.849721505295274e-05,
"loss": 0.2659,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11350315809249878,
"step": 900,
"valid_targets_mean": 4294.0,
"valid_targets_min": 3464
},
{
"epoch": 1.4938067712634187,
"grad_norm": 0.511407254658733,
"learning_rate": 3.8465757326772507e-05,
"loss": 0.2638,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1495528370141983,
"step": 905,
"valid_targets_mean": 4677.5,
"valid_targets_min": 2858
},
{
"epoch": 1.5020644095788604,
"grad_norm": 0.47993269566521135,
"learning_rate": 3.84339868760647e-05,
"loss": 0.2659,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13788801431655884,
"step": 910,
"valid_targets_mean": 4710.1,
"valid_targets_min": 3730
},
{
"epoch": 1.5103220478943022,
"grad_norm": 0.5293268363049909,
"learning_rate": 3.840190423887383e-05,
"loss": 0.2609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11736133694648743,
"step": 915,
"valid_targets_mean": 3748.9,
"valid_targets_min": 2649
},
{
"epoch": 1.518579686209744,
"grad_norm": 0.4748886141794393,
"learning_rate": 3.8369509958531394e-05,
"loss": 0.2776,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13339635729789734,
"step": 920,
"valid_targets_mean": 4496.0,
"valid_targets_min": 2767
},
{
"epoch": 1.5268373245251858,
"grad_norm": 0.4788492145141905,
"learning_rate": 3.833680458364668e-05,
"loss": 0.2565,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12580786645412445,
"step": 925,
"valid_targets_mean": 4554.0,
"valid_targets_min": 3775
},
{
"epoch": 1.5350949628406276,
"grad_norm": 0.491063097020191,
"learning_rate": 3.8303788668097486e-05,
"loss": 0.2573,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13013118505477905,
"step": 930,
"valid_targets_mean": 4519.9,
"valid_targets_min": 3948
},
{
"epoch": 1.5433526011560694,
"grad_norm": 0.4552684112083814,
"learning_rate": 3.8270462771020714e-05,
"loss": 0.2645,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12426851689815521,
"step": 935,
"valid_targets_mean": 4385.4,
"valid_targets_min": 3829
},
{
"epoch": 1.5516102394715112,
"grad_norm": 0.4484556157896209,
"learning_rate": 3.8236827456802944e-05,
"loss": 0.2702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11891467124223709,
"step": 940,
"valid_targets_mean": 3905.4,
"valid_targets_min": 1712
},
{
"epoch": 1.559867877786953,
"grad_norm": 0.49253950025833915,
"learning_rate": 3.820288329507083e-05,
"loss": 0.2728,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12675052881240845,
"step": 945,
"valid_targets_mean": 4208.2,
"valid_targets_min": 3062
},
{
"epoch": 1.5681255161023948,
"grad_norm": 0.48077778511212665,
"learning_rate": 3.81686308606815e-05,
"loss": 0.265,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13709363341331482,
"step": 950,
"valid_targets_mean": 4661.5,
"valid_targets_min": 4008
},
{
"epoch": 1.5763831544178366,
"grad_norm": 0.46953264796819677,
"learning_rate": 3.813407073371277e-05,
"loss": 0.2628,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1242109090089798,
"step": 955,
"valid_targets_mean": 4375.1,
"valid_targets_min": 3475
},
{
"epoch": 1.5846407927332784,
"grad_norm": 0.47971429057912757,
"learning_rate": 3.8099203499453354e-05,
"loss": 0.265,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1679193377494812,
"step": 960,
"valid_targets_mean": 6077.4,
"valid_targets_min": 3700
},
{
"epoch": 1.5928984310487202,
"grad_norm": 0.4459745468559393,
"learning_rate": 3.806402974839295e-05,
"loss": 0.2647,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10789842903614044,
"step": 965,
"valid_targets_mean": 4299.5,
"valid_targets_min": 3635
},
{
"epoch": 1.601156069364162,
"grad_norm": 0.44895785011313855,
"learning_rate": 3.8028550076212223e-05,
"loss": 0.2708,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13352727890014648,
"step": 970,
"valid_targets_mean": 4810.8,
"valid_targets_min": 3937
},
{
"epoch": 1.6094137076796038,
"grad_norm": 0.457083721181934,
"learning_rate": 3.799276508377273e-05,
"loss": 0.2664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12350866943597794,
"step": 975,
"valid_targets_mean": 4557.5,
"valid_targets_min": 3907
},
{
"epoch": 1.6176713459950454,
"grad_norm": 0.484456565451039,
"learning_rate": 3.7956675377106746e-05,
"loss": 0.2669,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12426236271858215,
"step": 980,
"valid_targets_mean": 4355.1,
"valid_targets_min": 3673
},
{
"epoch": 1.6259289843104872,
"grad_norm": 0.4412065607279786,
"learning_rate": 3.792028156740699e-05,
"loss": 0.2672,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11076773703098297,
"step": 985,
"valid_targets_mean": 4217.6,
"valid_targets_min": 3336
},
{
"epoch": 1.634186622625929,
"grad_norm": 0.5000554567249035,
"learning_rate": 3.788358427101629e-05,
"loss": 0.264,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12345005571842194,
"step": 990,
"valid_targets_mean": 4113.9,
"valid_targets_min": 3537
},
{
"epoch": 1.6424442609413707,
"grad_norm": 0.47377715700161044,
"learning_rate": 3.784658410941711e-05,
"loss": 0.253,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.143669992685318,
"step": 995,
"valid_targets_mean": 4877.9,
"valid_targets_min": 3732
},
{
"epoch": 1.6507018992568125,
"grad_norm": 0.46966238655861103,
"learning_rate": 3.780928170922108e-05,
"loss": 0.2726,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1153988391160965,
"step": 1000,
"valid_targets_mean": 4118.8,
"valid_targets_min": 2148
},
{
"epoch": 1.6589595375722543,
"grad_norm": 0.49164516002546615,
"learning_rate": 3.777167770215834e-05,
"loss": 0.2634,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13667413592338562,
"step": 1005,
"valid_targets_mean": 4633.8,
"valid_targets_min": 3374
},
{
"epoch": 1.6672171758876961,
"grad_norm": 0.45680795399285584,
"learning_rate": 3.773377272506685e-05,
"loss": 0.2655,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14256805181503296,
"step": 1010,
"valid_targets_mean": 4927.6,
"valid_targets_min": 3709
},
{
"epoch": 1.675474814203138,
"grad_norm": 0.5361973016378934,
"learning_rate": 3.769556741988163e-05,
"loss": 0.2564,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12212878465652466,
"step": 1015,
"valid_targets_mean": 4202.5,
"valid_targets_min": 2504
},
{
"epoch": 1.6837324525185797,
"grad_norm": 0.47713366103511345,
"learning_rate": 3.7657062433623825e-05,
"loss": 0.2696,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15070876479148865,
"step": 1020,
"valid_targets_mean": 5206.8,
"valid_targets_min": 4256
},
{
"epoch": 1.6919900908340215,
"grad_norm": 0.8459999411503897,
"learning_rate": 3.761825841838986e-05,
"loss": 0.2581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14952445030212402,
"step": 1025,
"valid_targets_mean": 4900.5,
"valid_targets_min": 3526
},
{
"epoch": 1.700247729149463,
"grad_norm": 0.46502111478478436,
"learning_rate": 3.7579156031340274e-05,
"loss": 0.268,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11956256628036499,
"step": 1030,
"valid_targets_mean": 4228.4,
"valid_targets_min": 2962
},
{
"epoch": 1.708505367464905,
"grad_norm": 0.540757549905523,
"learning_rate": 3.753975593468865e-05,
"loss": 0.2578,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11560507863759995,
"step": 1035,
"valid_targets_mean": 3385.9,
"valid_targets_min": 1780
},
{
"epoch": 1.7167630057803467,
"grad_norm": 0.4939475771211517,
"learning_rate": 3.750005879569043e-05,
"loss": 0.2616,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1276017427444458,
"step": 1040,
"valid_targets_mean": 4448.8,
"valid_targets_min": 3058
},
{
"epoch": 1.7250206440957885,
"grad_norm": 0.4605312771093914,
"learning_rate": 3.7460065286631526e-05,
"loss": 0.2615,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1260099709033966,
"step": 1045,
"valid_targets_mean": 4711.2,
"valid_targets_min": 3701
},
{
"epoch": 1.7332782824112303,
"grad_norm": 0.4794655266534176,
"learning_rate": 3.741977608481704e-05,
"loss": 0.2624,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11960376799106598,
"step": 1050,
"valid_targets_mean": 4035.0,
"valid_targets_min": 1570
},
{
"epoch": 1.741535920726672,
"grad_norm": 0.4946904249393211,
"learning_rate": 3.73791918725597e-05,
"loss": 0.2568,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13107861578464508,
"step": 1055,
"valid_targets_mean": 4312.9,
"valid_targets_min": 996
},
{
"epoch": 1.7497935590421139,
"grad_norm": 0.4419379441213521,
"learning_rate": 3.733831333716838e-05,
"loss": 0.2637,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13335686922073364,
"step": 1060,
"valid_targets_mean": 4650.4,
"valid_targets_min": 4003
},
{
"epoch": 1.7580511973575557,
"grad_norm": 0.4561095081645458,
"learning_rate": 3.729714117093641e-05,
"loss": 0.2694,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13015910983085632,
"step": 1065,
"valid_targets_mean": 4510.6,
"valid_targets_min": 3562
},
{
"epoch": 1.7663088356729975,
"grad_norm": 0.5216734794531762,
"learning_rate": 3.725567607112987e-05,
"loss": 0.2728,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13285207748413086,
"step": 1070,
"valid_targets_mean": 4412.8,
"valid_targets_min": 3574
},
{
"epoch": 1.7745664739884393,
"grad_norm": 0.43450642038300163,
"learning_rate": 3.721391873997577e-05,
"loss": 0.2613,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12838274240493774,
"step": 1075,
"valid_targets_mean": 4546.5,
"valid_targets_min": 3060
},
{
"epoch": 1.782824112303881,
"grad_norm": 0.42862589975418464,
"learning_rate": 3.7171869884650186e-05,
"loss": 0.2638,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12460225075483322,
"step": 1080,
"valid_targets_mean": 4534.8,
"valid_targets_min": 3923
},
{
"epoch": 1.7910817506193228,
"grad_norm": 0.5109533118794043,
"learning_rate": 3.712953021726626e-05,
"loss": 0.2757,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1391299068927765,
"step": 1085,
"valid_targets_mean": 3927.8,
"valid_targets_min": 2108
},
{
"epoch": 1.7993393889347646,
"grad_norm": 0.47332776910648194,
"learning_rate": 3.708690045486214e-05,
"loss": 0.2574,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12305203080177307,
"step": 1090,
"valid_targets_mean": 4466.9,
"valid_targets_min": 3495
},
{
"epoch": 1.8075970272502064,
"grad_norm": 0.46837964308184676,
"learning_rate": 3.7043981319388844e-05,
"loss": 0.251,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12698081135749817,
"step": 1095,
"valid_targets_mean": 4123.2,
"valid_targets_min": 1775
},
{
"epoch": 1.8158546655656482,
"grad_norm": 0.4770365312933627,
"learning_rate": 3.700077353769803e-05,
"loss": 0.2598,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1248985230922699,
"step": 1100,
"valid_targets_mean": 4500.5,
"valid_targets_min": 1726
},
{
"epoch": 1.82411230388109,
"grad_norm": 0.44869876267200837,
"learning_rate": 3.69572778415297e-05,
"loss": 0.2688,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11670802533626556,
"step": 1105,
"valid_targets_mean": 4032.9,
"valid_targets_min": 1005
},
{
"epoch": 1.8323699421965318,
"grad_norm": 0.4480957760289604,
"learning_rate": 3.691349496749977e-05,
"loss": 0.263,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13353858888149261,
"step": 1110,
"valid_targets_mean": 4485.8,
"valid_targets_min": 2954
},
{
"epoch": 1.8406275805119736,
"grad_norm": 0.44078232054773303,
"learning_rate": 3.686942565708765e-05,
"loss": 0.2597,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17375579476356506,
"step": 1115,
"valid_targets_mean": 5642.4,
"valid_targets_min": 1601
},
{
"epoch": 1.8488852188274154,
"grad_norm": 0.44691775227672,
"learning_rate": 3.6825070656623626e-05,
"loss": 0.2597,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13711009919643402,
"step": 1120,
"valid_targets_mean": 4354.9,
"valid_targets_min": 2242
},
{
"epoch": 1.8571428571428572,
"grad_norm": 0.4353519219514731,
"learning_rate": 3.6780430717276295e-05,
"loss": 0.2753,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1522749811410904,
"step": 1125,
"valid_targets_mean": 5484.5,
"valid_targets_min": 4154
},
{
"epoch": 1.865400495458299,
"grad_norm": 0.46659282770462546,
"learning_rate": 3.673550659503975e-05,
"loss": 0.256,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11492359638214111,
"step": 1130,
"valid_targets_mean": 4188.6,
"valid_targets_min": 3368
},
{
"epoch": 1.8736581337737408,
"grad_norm": 0.45398686832758106,
"learning_rate": 3.669029905072087e-05,
"loss": 0.256,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11406615376472473,
"step": 1135,
"valid_targets_mean": 4217.8,
"valid_targets_min": 2162
},
{
"epoch": 1.8819157720891826,
"grad_norm": 0.5152788233575248,
"learning_rate": 3.664480884992638e-05,
"loss": 0.254,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13111890852451324,
"step": 1140,
"valid_targets_mean": 4628.8,
"valid_targets_min": 4068
},
{
"epoch": 1.8901734104046244,
"grad_norm": 0.4497895521580664,
"learning_rate": 3.6599036763049886e-05,
"loss": 0.2615,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13089975714683533,
"step": 1145,
"valid_targets_mean": 5063.1,
"valid_targets_min": 3861
},
{
"epoch": 1.8984310487200662,
"grad_norm": 0.4572192803600719,
"learning_rate": 3.655298356525885e-05,
"loss": 0.2645,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14287543296813965,
"step": 1150,
"valid_targets_mean": 4357.4,
"valid_targets_min": 3619
},
{
"epoch": 1.906688687035508,
"grad_norm": 0.47539023670459546,
"learning_rate": 3.650665003648147e-05,
"loss": 0.2632,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1281719207763672,
"step": 1155,
"valid_targets_mean": 3928.0,
"valid_targets_min": 1415
},
{
"epoch": 1.9149463253509498,
"grad_norm": 0.47540039807951257,
"learning_rate": 3.646003696139341e-05,
"loss": 0.2577,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12878680229187012,
"step": 1160,
"valid_targets_mean": 4181.5,
"valid_targets_min": 2237
},
{
"epoch": 1.9232039636663916,
"grad_norm": 0.44687555133933315,
"learning_rate": 3.64131451294046e-05,
"loss": 0.2655,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12017180025577545,
"step": 1165,
"valid_targets_mean": 3800.1,
"valid_targets_min": 2437
},
{
"epoch": 1.9314616019818331,
"grad_norm": 0.4654441653957933,
"learning_rate": 3.6365975334645806e-05,
"loss": 0.2587,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13186921179294586,
"step": 1170,
"valid_targets_mean": 4312.8,
"valid_targets_min": 3109
},
{
"epoch": 1.939719240297275,
"grad_norm": 0.46245350666390495,
"learning_rate": 3.63185283759552e-05,
"loss": 0.26,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14636358618736267,
"step": 1175,
"valid_targets_mean": 4637.4,
"valid_targets_min": 3604
},
{
"epoch": 1.9479768786127167,
"grad_norm": 0.47576207592671826,
"learning_rate": 3.627080505686481e-05,
"loss": 0.2618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13235077261924744,
"step": 1180,
"valid_targets_mean": 4069.9,
"valid_targets_min": 2186
},
{
"epoch": 1.9562345169281585,
"grad_norm": 0.4674465346836783,
"learning_rate": 3.622280618558696e-05,
"loss": 0.2671,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14810220897197723,
"step": 1185,
"valid_targets_mean": 4419.2,
"valid_targets_min": 3523
},
{
"epoch": 1.9644921552436003,
"grad_norm": 0.469610809945553,
"learning_rate": 3.617453257500055e-05,
"loss": 0.2803,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2076636552810669,
"step": 1190,
"valid_targets_mean": 6090.9,
"valid_targets_min": 4039
},
{
"epoch": 1.9727497935590421,
"grad_norm": 0.5052453558652513,
"learning_rate": 3.6125985042637265e-05,
"loss": 0.2602,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13445237278938293,
"step": 1195,
"valid_targets_mean": 3941.2,
"valid_targets_min": 402
},
{
"epoch": 1.981007431874484,
"grad_norm": 0.4490177992033094,
"learning_rate": 3.6077164410667786e-05,
"loss": 0.2627,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1545429676771164,
"step": 1200,
"valid_targets_mean": 5046.8,
"valid_targets_min": 3898
},
{
"epoch": 1.9892650701899257,
"grad_norm": 0.43375464569622274,
"learning_rate": 3.602807150588784e-05,
"loss": 0.2553,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12218572199344635,
"step": 1205,
"valid_targets_mean": 4313.9,
"valid_targets_min": 3267
},
{
"epoch": 1.9975227085053675,
"grad_norm": 0.4893291876845755,
"learning_rate": 3.597870715970417e-05,
"loss": 0.2546,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14900383353233337,
"step": 1210,
"valid_targets_mean": 4348.5,
"valid_targets_min": 3603
},
{
"epoch": 2.004954582989265,
"grad_norm": 0.43397097336059837,
"learning_rate": 3.592907220812051e-05,
"loss": 0.2451,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11573085188865662,
"step": 1215,
"valid_targets_mean": 3938.4,
"valid_targets_min": 1974
},
{
"epoch": 2.013212221304707,
"grad_norm": 0.45501770540007197,
"learning_rate": 3.587916749172338e-05,
"loss": 0.2509,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15736833214759827,
"step": 1220,
"valid_targets_mean": 5122.4,
"valid_targets_min": 4484
},
{
"epoch": 2.0214698596201486,
"grad_norm": 0.46460853885766024,
"learning_rate": 3.582899385566787e-05,
"loss": 0.25,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13113336265087128,
"step": 1225,
"valid_targets_mean": 4416.5,
"valid_targets_min": 3927
},
{
"epoch": 2.0297274979355904,
"grad_norm": 0.4959147061937609,
"learning_rate": 3.577855214966333e-05,
"loss": 0.2422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09888025373220444,
"step": 1230,
"valid_targets_mean": 3427.9,
"valid_targets_min": 2063
},
{
"epoch": 2.037985136251032,
"grad_norm": 0.485617306351796,
"learning_rate": 3.572784322795898e-05,
"loss": 0.2539,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11776536703109741,
"step": 1235,
"valid_targets_mean": 4142.1,
"valid_targets_min": 1766
},
{
"epoch": 2.046242774566474,
"grad_norm": 0.45826238421212523,
"learning_rate": 3.567686794932943e-05,
"loss": 0.2542,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13118301331996918,
"step": 1240,
"valid_targets_mean": 4503.8,
"valid_targets_min": 3502
},
{
"epoch": 2.0545004128819158,
"grad_norm": 0.44986676345766696,
"learning_rate": 3.5625627177060136e-05,
"loss": 0.2543,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12356361001729965,
"step": 1245,
"valid_targets_mean": 4630.6,
"valid_targets_min": 3907
},
{
"epoch": 2.0627580511973576,
"grad_norm": 0.45431876731426335,
"learning_rate": 3.557412177893281e-05,
"loss": 0.2457,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12505778670310974,
"step": 1250,
"valid_targets_mean": 4530.9,
"valid_targets_min": 3641
},
{
"epoch": 2.0710156895127994,
"grad_norm": 0.5276431416989884,
"learning_rate": 3.5522352627210685e-05,
"loss": 0.25,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09214842319488525,
"step": 1255,
"valid_targets_mean": 3728.6,
"valid_targets_min": 2269
},
{
"epoch": 2.079273327828241,
"grad_norm": 0.43844120085744087,
"learning_rate": 3.5470320598623784e-05,
"loss": 0.2573,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1284557729959488,
"step": 1260,
"valid_targets_mean": 4392.9,
"valid_targets_min": 2186
},
{
"epoch": 2.087530966143683,
"grad_norm": 0.46585862918408655,
"learning_rate": 3.5418026574354e-05,
"loss": 0.2583,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14499947428703308,
"step": 1265,
"valid_targets_mean": 4538.5,
"valid_targets_min": 3274
},
{
"epoch": 2.0957886044591247,
"grad_norm": 0.5845812827215499,
"learning_rate": 3.536547144002027e-05,
"loss": 0.2383,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12081082165241241,
"step": 1270,
"valid_targets_mean": 4693.5,
"valid_targets_min": 3811
},
{
"epoch": 2.1040462427745665,
"grad_norm": 0.5743138176345955,
"learning_rate": 3.531265608566351e-05,
"loss": 0.2531,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1258414089679718,
"step": 1275,
"valid_targets_mean": 4232.6,
"valid_targets_min": 3462
},
{
"epoch": 2.1123038810900083,
"grad_norm": 0.4625130255351799,
"learning_rate": 3.525958140573155e-05,
"loss": 0.2489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11214599013328552,
"step": 1280,
"valid_targets_mean": 4151.6,
"valid_targets_min": 1355
},
{
"epoch": 2.12056151940545,
"grad_norm": 0.4982959105794377,
"learning_rate": 3.5206248299064016e-05,
"loss": 0.2477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11039195209741592,
"step": 1285,
"valid_targets_mean": 4729.5,
"valid_targets_min": 3418
},
{
"epoch": 2.128819157720892,
"grad_norm": 0.5102078163912296,
"learning_rate": 3.5152657668877065e-05,
"loss": 0.2497,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11829587817192078,
"step": 1290,
"valid_targets_mean": 3910.5,
"valid_targets_min": 3201
},
{
"epoch": 2.1370767960363337,
"grad_norm": 0.4404683881255341,
"learning_rate": 3.5098810422748144e-05,
"loss": 0.2466,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10727163404226303,
"step": 1295,
"valid_targets_mean": 4287.6,
"valid_targets_min": 3397
},
{
"epoch": 2.1453344343517755,
"grad_norm": 0.442342251285477,
"learning_rate": 3.504470747260054e-05,
"loss": 0.2431,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12597107887268066,
"step": 1300,
"valid_targets_mean": 4568.2,
"valid_targets_min": 3669
},
{
"epoch": 2.1535920726672173,
"grad_norm": 0.4502385454781581,
"learning_rate": 3.499034973468802e-05,
"loss": 0.252,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13838830590248108,
"step": 1305,
"valid_targets_mean": 5067.6,
"valid_targets_min": 2180
},
{
"epoch": 2.161849710982659,
"grad_norm": 0.4682919209025986,
"learning_rate": 3.493573812957927e-05,
"loss": 0.2423,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10592959821224213,
"step": 1310,
"valid_targets_mean": 3769.1,
"valid_targets_min": 1041
},
{
"epoch": 2.170107349298101,
"grad_norm": 0.4925438873901821,
"learning_rate": 3.488087358214232e-05,
"loss": 0.2504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1760067641735077,
"step": 1315,
"valid_targets_mean": 4923.5,
"valid_targets_min": 2337
},
{
"epoch": 2.1783649876135427,
"grad_norm": 0.5049514830504129,
"learning_rate": 3.4825757021528844e-05,
"loss": 0.2468,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11750254034996033,
"step": 1320,
"valid_targets_mean": 4303.8,
"valid_targets_min": 3700
},
{
"epoch": 2.1866226259289845,
"grad_norm": 0.44548811603464666,
"learning_rate": 3.477038938115848e-05,
"loss": 0.2486,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11112725734710693,
"step": 1325,
"valid_targets_mean": 3649.2,
"valid_targets_min": 1931
},
{
"epoch": 2.1948802642444263,
"grad_norm": 0.4747995390565894,
"learning_rate": 3.471477159870299e-05,
"loss": 0.2457,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11855631321668625,
"step": 1330,
"valid_targets_mean": 4241.5,
"valid_targets_min": 3365
},
{
"epoch": 2.203137902559868,
"grad_norm": 0.4770273667014847,
"learning_rate": 3.465890461607038e-05,
"loss": 0.2591,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11744537949562073,
"step": 1335,
"valid_targets_mean": 4378.0,
"valid_targets_min": 2839
},
{
"epoch": 2.2113955408753094,
"grad_norm": 0.47853379376490063,
"learning_rate": 3.460278937938896e-05,
"loss": 0.2468,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12426887452602386,
"step": 1340,
"valid_targets_mean": 4228.0,
"valid_targets_min": 3550
},
{
"epoch": 2.2196531791907512,
"grad_norm": 0.4692158680802995,
"learning_rate": 3.454642683899132e-05,
"loss": 0.2532,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12839433550834656,
"step": 1345,
"valid_targets_mean": 4168.0,
"valid_targets_min": 2376
},
{
"epoch": 2.227910817506193,
"grad_norm": 0.4600459985453449,
"learning_rate": 3.4489817949398224e-05,
"loss": 0.2521,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13021403551101685,
"step": 1350,
"valid_targets_mean": 4738.4,
"valid_targets_min": 3016
},
{
"epoch": 2.236168455821635,
"grad_norm": 0.5002844596895054,
"learning_rate": 3.443296366930244e-05,
"loss": 0.254,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11979164928197861,
"step": 1355,
"valid_targets_mean": 4388.4,
"valid_targets_min": 2044
},
{
"epoch": 2.2444260941370766,
"grad_norm": 0.4529249522967833,
"learning_rate": 3.4375864961552546e-05,
"loss": 0.2569,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13094031810760498,
"step": 1360,
"valid_targets_mean": 4926.1,
"valid_targets_min": 3847
},
{
"epoch": 2.2526837324525184,
"grad_norm": 0.4571613173658892,
"learning_rate": 3.431852279313657e-05,
"loss": 0.2476,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12817558646202087,
"step": 1365,
"valid_targets_mean": 4638.9,
"valid_targets_min": 4294
},
{
"epoch": 2.26094137076796,
"grad_norm": 0.4539232415645253,
"learning_rate": 3.426093813516565e-05,
"loss": 0.2444,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12236069142818451,
"step": 1370,
"valid_targets_mean": 4569.5,
"valid_targets_min": 4104
},
{
"epoch": 2.269199009083402,
"grad_norm": 0.47193617242287783,
"learning_rate": 3.420311196285757e-05,
"loss": 0.2493,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1482844203710556,
"step": 1375,
"valid_targets_mean": 4683.5,
"valid_targets_min": 3205
},
{
"epoch": 2.277456647398844,
"grad_norm": 0.4830761121604635,
"learning_rate": 3.4145045255520244e-05,
"loss": 0.249,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11896125972270966,
"step": 1380,
"valid_targets_mean": 4559.6,
"valid_targets_min": 4053
},
{
"epoch": 2.2857142857142856,
"grad_norm": 0.48390888898451884,
"learning_rate": 3.408673899653515e-05,
"loss": 0.2514,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14629939198493958,
"step": 1385,
"valid_targets_mean": 4711.5,
"valid_targets_min": 3719
},
{
"epoch": 2.2939719240297274,
"grad_norm": 0.4341495739988123,
"learning_rate": 3.4028194173340655e-05,
"loss": 0.2511,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1253558099269867,
"step": 1390,
"valid_targets_mean": 4275.4,
"valid_targets_min": 3637
},
{
"epoch": 2.302229562345169,
"grad_norm": 0.4518837312860524,
"learning_rate": 3.396941177741531e-05,
"loss": 0.2466,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11686206609010696,
"step": 1395,
"valid_targets_mean": 4058.5,
"valid_targets_min": 3195
},
{
"epoch": 2.310487200660611,
"grad_norm": 0.4840426059978692,
"learning_rate": 3.391039280426103e-05,
"loss": 0.2504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12521852552890778,
"step": 1400,
"valid_targets_mean": 4404.0,
"valid_targets_min": 3966
},
{
"epoch": 2.3187448389760528,
"grad_norm": 0.4699925522760237,
"learning_rate": 3.385113825338627e-05,
"loss": 0.2554,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13816365599632263,
"step": 1405,
"valid_targets_mean": 4963.6,
"valid_targets_min": 3984
},
{
"epoch": 2.3270024772914946,
"grad_norm": 0.4783185959305509,
"learning_rate": 3.379164912828908e-05,
"loss": 0.249,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12407098710536957,
"step": 1410,
"valid_targets_mean": 4064.8,
"valid_targets_min": 1932
},
{
"epoch": 2.3352601156069364,
"grad_norm": 0.44361959606080625,
"learning_rate": 3.373192643644011e-05,
"loss": 0.25,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12075299024581909,
"step": 1415,
"valid_targets_mean": 4214.1,
"valid_targets_min": 3373
},
{
"epoch": 2.343517753922378,
"grad_norm": 0.4517438261953717,
"learning_rate": 3.3671971189265554e-05,
"loss": 0.2479,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13503974676132202,
"step": 1420,
"valid_targets_mean": 4809.0,
"valid_targets_min": 3565
},
{
"epoch": 2.35177539223782,
"grad_norm": 0.4312339154840659,
"learning_rate": 3.3611784402129995e-05,
"loss": 0.2462,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14392805099487305,
"step": 1425,
"valid_targets_mean": 4905.9,
"valid_targets_min": 3635
},
{
"epoch": 2.3600330305532617,
"grad_norm": 0.42980826026277213,
"learning_rate": 3.3551367094319275e-05,
"loss": 0.2522,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12983693182468414,
"step": 1430,
"valid_targets_mean": 4739.5,
"valid_targets_min": 3173
},
{
"epoch": 2.3682906688687035,
"grad_norm": 0.4319114267487943,
"learning_rate": 3.349072028902316e-05,
"loss": 0.2392,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14828899502754211,
"step": 1435,
"valid_targets_mean": 5494.0,
"valid_targets_min": 3602
},
{
"epoch": 2.3765483071841453,
"grad_norm": 0.42771523828942015,
"learning_rate": 3.342984501331805e-05,
"loss": 0.2413,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13541395962238312,
"step": 1440,
"valid_targets_mean": 4619.9,
"valid_targets_min": 3880
},
{
"epoch": 2.384805945499587,
"grad_norm": 0.44811265476945117,
"learning_rate": 3.3368742298149566e-05,
"loss": 0.2457,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1396416425704956,
"step": 1445,
"valid_targets_mean": 4563.0,
"valid_targets_min": 3434
},
{
"epoch": 2.393063583815029,
"grad_norm": 0.44155790602613754,
"learning_rate": 3.330741317831512e-05,
"loss": 0.2441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12792906165122986,
"step": 1450,
"valid_targets_mean": 4292.2,
"valid_targets_min": 3094
},
{
"epoch": 2.4013212221304707,
"grad_norm": 0.4280961238423992,
"learning_rate": 3.324585869244636e-05,
"loss": 0.2427,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10621410608291626,
"step": 1455,
"valid_targets_mean": 4434.9,
"valid_targets_min": 3370
},
{
"epoch": 2.4095788604459125,
"grad_norm": 0.4527679333983257,
"learning_rate": 3.3184079882991606e-05,
"loss": 0.2472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11382193118333817,
"step": 1460,
"valid_targets_mean": 3837.6,
"valid_targets_min": 1527
},
{
"epoch": 2.4178364987613543,
"grad_norm": 0.4440672601972065,
"learning_rate": 3.312207779619815e-05,
"loss": 0.2449,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1207650825381279,
"step": 1465,
"valid_targets_mean": 4512.2,
"valid_targets_min": 4106
},
{
"epoch": 2.426094137076796,
"grad_norm": 0.4580358981725213,
"learning_rate": 3.305985348209462e-05,
"loss": 0.2592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13219891488552094,
"step": 1470,
"valid_targets_mean": 4382.5,
"valid_targets_min": 2795
},
{
"epoch": 2.434351775392238,
"grad_norm": 0.4370763208517525,
"learning_rate": 3.2997407994473095e-05,
"loss": 0.2398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11910690367221832,
"step": 1475,
"valid_targets_mean": 4532.6,
"valid_targets_min": 3830
},
{
"epoch": 2.4426094137076797,
"grad_norm": 0.4556388501913592,
"learning_rate": 3.293474239087134e-05,
"loss": 0.2521,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13865631818771362,
"step": 1480,
"valid_targets_mean": 5303.8,
"valid_targets_min": 2566
},
{
"epoch": 2.4508670520231215,
"grad_norm": 0.43109345180568515,
"learning_rate": 3.2871857732554854e-05,
"loss": 0.2444,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14210563898086548,
"step": 1485,
"valid_targets_mean": 4955.0,
"valid_targets_min": 3968
},
{
"epoch": 2.4591246903385633,
"grad_norm": 0.43566912602460367,
"learning_rate": 3.2808755084498936e-05,
"loss": 0.249,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10723206400871277,
"step": 1490,
"valid_targets_mean": 4011.1,
"valid_targets_min": 3245
},
{
"epoch": 2.467382328654005,
"grad_norm": 0.4280764345599046,
"learning_rate": 3.2745435515370585e-05,
"loss": 0.2394,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11878250539302826,
"step": 1495,
"valid_targets_mean": 4584.2,
"valid_targets_min": 2875
},
{
"epoch": 2.475639966969447,
"grad_norm": 0.4564628988191037,
"learning_rate": 3.268190009751046e-05,
"loss": 0.247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11330521106719971,
"step": 1500,
"valid_targets_mean": 4282.5,
"valid_targets_min": 3577
},
{
"epoch": 2.4838976052848887,
"grad_norm": 0.4702182813218408,
"learning_rate": 3.261814990691471e-05,
"loss": 0.2429,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1347694993019104,
"step": 1505,
"valid_targets_mean": 4631.5,
"valid_targets_min": 4001
},
{
"epoch": 2.4921552436003305,
"grad_norm": 0.44540108560215913,
"learning_rate": 3.2554186023216715e-05,
"loss": 0.2547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13759279251098633,
"step": 1510,
"valid_targets_mean": 4346.0,
"valid_targets_min": 3640
},
{
"epoch": 2.5004128819157723,
"grad_norm": 0.4446916687297095,
"learning_rate": 3.2490009529668845e-05,
"loss": 0.2544,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13435958325862885,
"step": 1515,
"valid_targets_mean": 5134.4,
"valid_targets_min": 4087
},
{
"epoch": 2.508670520231214,
"grad_norm": 0.4267308486996638,
"learning_rate": 3.24256215131241e-05,
"loss": 0.2547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1179559975862503,
"step": 1520,
"valid_targets_mean": 4386.2,
"valid_targets_min": 2200
},
{
"epoch": 2.516928158546656,
"grad_norm": 0.4965568082809046,
"learning_rate": 3.236102306401767e-05,
"loss": 0.2349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10542820394039154,
"step": 1525,
"valid_targets_mean": 4006.8,
"valid_targets_min": 3116
},
{
"epoch": 2.5251857968620977,
"grad_norm": 0.4779725134173445,
"learning_rate": 3.229621527634855e-05,
"loss": 0.2396,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12921521067619324,
"step": 1530,
"valid_targets_mean": 4189.4,
"valid_targets_min": 3182
},
{
"epoch": 2.5334434351775394,
"grad_norm": 0.43643436355919185,
"learning_rate": 3.223119924766093e-05,
"loss": 0.2468,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13907399773597717,
"step": 1535,
"valid_targets_mean": 4950.4,
"valid_targets_min": 4034
},
{
"epoch": 2.5417010734929812,
"grad_norm": 0.5007879447541802,
"learning_rate": 3.2165976079025644e-05,
"loss": 0.241,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11765088886022568,
"step": 1540,
"valid_targets_mean": 4159.8,
"valid_targets_min": 3653
},
{
"epoch": 2.549958711808423,
"grad_norm": 0.4374581708549872,
"learning_rate": 3.210054687502152e-05,
"loss": 0.2471,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1014547199010849,
"step": 1545,
"valid_targets_mean": 4044.4,
"valid_targets_min": 2314
},
{
"epoch": 2.558216350123865,
"grad_norm": 0.4376031180971691,
"learning_rate": 3.2034912743716666e-05,
"loss": 0.2407,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10798119008541107,
"step": 1550,
"valid_targets_mean": 4236.2,
"valid_targets_min": 2504
},
{
"epoch": 2.5664739884393066,
"grad_norm": 0.43150567512577714,
"learning_rate": 3.1969074796649734e-05,
"loss": 0.2392,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11575426161289215,
"step": 1555,
"valid_targets_mean": 4428.0,
"valid_targets_min": 3854
},
{
"epoch": 2.574731626754748,
"grad_norm": 0.47651070063030754,
"learning_rate": 3.190303414881105e-05,
"loss": 0.247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14137311279773712,
"step": 1560,
"valid_targets_mean": 4742.5,
"valid_targets_min": 4119
},
{
"epoch": 2.5829892650701898,
"grad_norm": 0.44884881688366846,
"learning_rate": 3.183679191862375e-05,
"loss": 0.2453,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12973275780677795,
"step": 1565,
"valid_targets_mean": 4427.5,
"valid_targets_min": 3267
},
{
"epoch": 2.5912469033856316,
"grad_norm": 0.4257771335337423,
"learning_rate": 3.1770349227924854e-05,
"loss": 0.2498,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13654743134975433,
"step": 1570,
"valid_targets_mean": 4834.1,
"valid_targets_min": 3959
},
{
"epoch": 2.5995045417010734,
"grad_norm": 0.44476448960434495,
"learning_rate": 3.170370720194626e-05,
"loss": 0.247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13154035806655884,
"step": 1575,
"valid_targets_mean": 4475.2,
"valid_targets_min": 3847
},
{
"epoch": 2.607762180016515,
"grad_norm": 0.5302074949456399,
"learning_rate": 3.1636866969295684e-05,
"loss": 0.2447,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12269929051399231,
"step": 1580,
"valid_targets_mean": 4144.6,
"valid_targets_min": 3193
},
{
"epoch": 2.616019818331957,
"grad_norm": 0.4400008398719134,
"learning_rate": 3.156982966193753e-05,
"loss": 0.2494,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11068129539489746,
"step": 1585,
"valid_targets_mean": 4297.6,
"valid_targets_min": 2042
},
{
"epoch": 2.6242774566473988,
"grad_norm": 0.3975716092034701,
"learning_rate": 3.150259641517375e-05,
"loss": 0.2349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10300234705209732,
"step": 1590,
"valid_targets_mean": 4020.9,
"valid_targets_min": 2613
},
{
"epoch": 2.6325350949628405,
"grad_norm": 0.45262861152384215,
"learning_rate": 3.14351683676246e-05,
"loss": 0.2478,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11385839432477951,
"step": 1595,
"valid_targets_mean": 4200.1,
"valid_targets_min": 2701
},
{
"epoch": 2.6407927332782823,
"grad_norm": 0.4406015658188927,
"learning_rate": 3.1367546661209355e-05,
"loss": 0.2536,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12662234902381897,
"step": 1600,
"valid_targets_mean": 5081.5,
"valid_targets_min": 3245
},
{
"epoch": 2.649050371593724,
"grad_norm": 0.479500261650723,
"learning_rate": 3.1299732441126995e-05,
"loss": 0.2558,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.125592440366745,
"step": 1605,
"valid_targets_mean": 4554.6,
"valid_targets_min": 3083
},
{
"epoch": 2.657308009909166,
"grad_norm": 0.43407005034057244,
"learning_rate": 3.123172685583676e-05,
"loss": 0.2398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11420107632875443,
"step": 1610,
"valid_targets_mean": 4498.0,
"valid_targets_min": 3475
},
{
"epoch": 2.6655656482246077,
"grad_norm": 0.4694188458851959,
"learning_rate": 3.116353105703876e-05,
"loss": 0.245,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1340787410736084,
"step": 1615,
"valid_targets_mean": 4633.9,
"valid_targets_min": 1669
},
{
"epoch": 2.6738232865400495,
"grad_norm": 0.471328364981702,
"learning_rate": 3.1095146199654426e-05,
"loss": 0.2489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13994908332824707,
"step": 1620,
"valid_targets_mean": 4625.1,
"valid_targets_min": 2969
},
{
"epoch": 2.6820809248554913,
"grad_norm": 0.46452694818887164,
"learning_rate": 3.1026573441806976e-05,
"loss": 0.2477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11189450323581696,
"step": 1625,
"valid_targets_mean": 4207.9,
"valid_targets_min": 3437
},
{
"epoch": 2.690338563170933,
"grad_norm": 0.4235118073756756,
"learning_rate": 3.095781394480177e-05,
"loss": 0.2486,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09240743517875671,
"step": 1630,
"valid_targets_mean": 4034.9,
"valid_targets_min": 1812
},
{
"epoch": 2.698596201486375,
"grad_norm": 0.42738671611878837,
"learning_rate": 3.088886887310671e-05,
"loss": 0.2477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1039619967341423,
"step": 1635,
"valid_targets_mean": 3920.1,
"valid_targets_min": 1327
},
{
"epoch": 2.7068538398018167,
"grad_norm": 0.45760723678327386,
"learning_rate": 3.081973939433244e-05,
"loss": 0.2441,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12901362776756287,
"step": 1640,
"valid_targets_mean": 4613.4,
"valid_targets_min": 4083
},
{
"epoch": 2.7151114781172585,
"grad_norm": 0.4395431025326516,
"learning_rate": 3.0750426679212614e-05,
"loss": 0.2477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11939951777458191,
"step": 1645,
"valid_targets_mean": 4244.9,
"valid_targets_min": 996
},
{
"epoch": 2.7233691164327003,
"grad_norm": 0.42388931218545417,
"learning_rate": 3.068093190158406e-05,
"loss": 0.2533,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11903703957796097,
"step": 1650,
"valid_targets_mean": 4652.9,
"valid_targets_min": 3466
},
{
"epoch": 2.731626754748142,
"grad_norm": 0.47132721643052794,
"learning_rate": 3.061125623836692e-05,
"loss": 0.2452,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13952422142028809,
"step": 1655,
"valid_targets_mean": 4893.6,
"valid_targets_min": 3648
},
{
"epoch": 2.739884393063584,
"grad_norm": 0.45395809194439946,
"learning_rate": 3.054140086954466e-05,
"loss": 0.2515,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12727044522762299,
"step": 1660,
"valid_targets_mean": 4271.8,
"valid_targets_min": 2738
},
{
"epoch": 2.7481420313790257,
"grad_norm": 0.4318670127447187,
"learning_rate": 3.04713669781442e-05,
"loss": 0.2401,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12868531048297882,
"step": 1665,
"valid_targets_mean": 4715.8,
"valid_targets_min": 4123
},
{
"epoch": 2.7563996696944675,
"grad_norm": 0.4253777391200916,
"learning_rate": 3.0401155750215733e-05,
"loss": 0.2438,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1321411430835724,
"step": 1670,
"valid_targets_mean": 4755.1,
"valid_targets_min": 3133
},
{
"epoch": 2.7646573080099093,
"grad_norm": 0.44191019628430306,
"learning_rate": 3.033076837481275e-05,
"loss": 0.2458,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1255328357219696,
"step": 1675,
"valid_targets_mean": 4326.5,
"valid_targets_min": 3497
},
{
"epoch": 2.772914946325351,
"grad_norm": 0.46245826259586514,
"learning_rate": 3.0260206043971857e-05,
"loss": 0.2548,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11325007677078247,
"step": 1680,
"valid_targets_mean": 4255.4,
"valid_targets_min": 1997
},
{
"epoch": 2.781172584640793,
"grad_norm": 0.4343187912150736,
"learning_rate": 3.0189469952692608e-05,
"loss": 0.2478,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12894544005393982,
"step": 1685,
"valid_targets_mean": 4514.1,
"valid_targets_min": 3541
},
{
"epoch": 2.7894302229562347,
"grad_norm": 0.452134166308327,
"learning_rate": 3.011856129891723e-05,
"loss": 0.2442,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11136821657419205,
"step": 1690,
"valid_targets_mean": 4088.6,
"valid_targets_min": 3280
},
{
"epoch": 2.7976878612716765,
"grad_norm": 0.4559061066699252,
"learning_rate": 3.004748128351038e-05,
"loss": 0.2448,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12184464931488037,
"step": 1695,
"valid_targets_mean": 4389.0,
"valid_targets_min": 3694
},
{
"epoch": 2.805945499587118,
"grad_norm": 0.4308914145191253,
"learning_rate": 2.997623111023879e-05,
"loss": 0.2355,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1059965118765831,
"step": 1700,
"valid_targets_mean": 4182.0,
"valid_targets_min": 1356
},
{
"epoch": 2.8142031379025596,
"grad_norm": 0.43939786174762097,
"learning_rate": 2.9904811985750868e-05,
"loss": 0.2428,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12782078981399536,
"step": 1705,
"valid_targets_mean": 4254.9,
"valid_targets_min": 2278
},
{
"epoch": 2.8224607762180014,
"grad_norm": 0.42836920759059627,
"learning_rate": 2.9833225119556277e-05,
"loss": 0.245,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11180959641933441,
"step": 1710,
"valid_targets_mean": 4051.9,
"valid_targets_min": 2718
},
{
"epoch": 2.830718414533443,
"grad_norm": 0.4707088014164877,
"learning_rate": 2.9761471724005457e-05,
"loss": 0.2424,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11054940521717072,
"step": 1715,
"valid_targets_mean": 4029.9,
"valid_targets_min": 2858
},
{
"epoch": 2.838976052848885,
"grad_norm": 0.46090613924986656,
"learning_rate": 2.968955301426908e-05,
"loss": 0.2312,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11227643489837646,
"step": 1720,
"valid_targets_mean": 4101.4,
"valid_targets_min": 3398
},
{
"epoch": 2.847233691164327,
"grad_norm": 0.47896895836941583,
"learning_rate": 2.9617470208317467e-05,
"loss": 0.244,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11067643761634827,
"step": 1725,
"valid_targets_mean": 4037.0,
"valid_targets_min": 1939
},
{
"epoch": 2.8554913294797686,
"grad_norm": 0.40820934151354926,
"learning_rate": 2.954522452689998e-05,
"loss": 0.2393,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11375854164361954,
"step": 1730,
"valid_targets_mean": 4573.1,
"valid_targets_min": 3824
},
{
"epoch": 2.8637489677952104,
"grad_norm": 0.4282984841635347,
"learning_rate": 2.947281719352434e-05,
"loss": 0.2362,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1166783794760704,
"step": 1735,
"valid_targets_mean": 4829.0,
"valid_targets_min": 4210
},
{
"epoch": 2.872006606110652,
"grad_norm": 0.43835985255727206,
"learning_rate": 2.9400249434435905e-05,
"loss": 0.2465,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10834118723869324,
"step": 1740,
"valid_targets_mean": 4046.0,
"valid_targets_min": 1558
},
{
"epoch": 2.880264244426094,
"grad_norm": 0.44533523694524346,
"learning_rate": 2.9327522478596885e-05,
"loss": 0.2358,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10838210582733154,
"step": 1745,
"valid_targets_mean": 4570.5,
"valid_targets_min": 3647
},
{
"epoch": 2.8885218827415358,
"grad_norm": 0.4546436471705934,
"learning_rate": 2.9254637557665565e-05,
"loss": 0.2381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1312817633152008,
"step": 1750,
"valid_targets_mean": 4484.9,
"valid_targets_min": 2739
},
{
"epoch": 2.8967795210569776,
"grad_norm": 0.4321044544272027,
"learning_rate": 2.9181595905975434e-05,
"loss": 0.2499,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11854548752307892,
"step": 1755,
"valid_targets_mean": 4121.9,
"valid_targets_min": 1902
},
{
"epoch": 2.9050371593724194,
"grad_norm": 0.48792686540393154,
"learning_rate": 2.9108398760514246e-05,
"loss": 0.2475,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1243300586938858,
"step": 1760,
"valid_targets_mean": 4803.1,
"valid_targets_min": 4334
},
{
"epoch": 2.913294797687861,
"grad_norm": 0.4302216302476997,
"learning_rate": 2.903504736090313e-05,
"loss": 0.2442,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12560029327869415,
"step": 1765,
"valid_targets_mean": 4504.1,
"valid_targets_min": 3291
},
{
"epoch": 2.921552436003303,
"grad_norm": 0.46856300990849903,
"learning_rate": 2.8961542949375556e-05,
"loss": 0.241,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11492963880300522,
"step": 1770,
"valid_targets_mean": 3906.8,
"valid_targets_min": 2450
},
{
"epoch": 2.9298100743187447,
"grad_norm": 0.4537015906707524,
"learning_rate": 2.8887886770756302e-05,
"loss": 0.2467,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12422479689121246,
"step": 1775,
"valid_targets_mean": 3935.0,
"valid_targets_min": 2170
},
{
"epoch": 2.9380677126341865,
"grad_norm": 0.4428745796930995,
"learning_rate": 2.881408007244039e-05,
"loss": 0.2516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12604756653308868,
"step": 1780,
"valid_targets_mean": 4049.2,
"valid_targets_min": 1537
},
{
"epoch": 2.9463253509496283,
"grad_norm": 0.424803624139159,
"learning_rate": 2.8740124104371937e-05,
"loss": 0.2493,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11889474093914032,
"step": 1785,
"valid_targets_mean": 4502.0,
"valid_targets_min": 3731
},
{
"epoch": 2.95458298926507,
"grad_norm": 0.43998745628720526,
"learning_rate": 2.866602011902301e-05,
"loss": 0.2469,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10292455554008484,
"step": 1790,
"valid_targets_mean": 4129.2,
"valid_targets_min": 3526
},
{
"epoch": 2.962840627580512,
"grad_norm": 0.4260567219248242,
"learning_rate": 2.8591769371372405e-05,
"loss": 0.2454,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12749522924423218,
"step": 1795,
"valid_targets_mean": 5166.5,
"valid_targets_min": 3962
},
{
"epoch": 2.9710982658959537,
"grad_norm": 0.41268008401326123,
"learning_rate": 2.851737311888438e-05,
"loss": 0.2446,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10959450900554657,
"step": 1800,
"valid_targets_mean": 4624.6,
"valid_targets_min": 3844
},
{
"epoch": 2.9793559042113955,
"grad_norm": 0.42720051287209215,
"learning_rate": 2.8442832621487385e-05,
"loss": 0.2398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12152857333421707,
"step": 1805,
"valid_targets_mean": 4435.2,
"valid_targets_min": 3624
},
{
"epoch": 2.9876135425268373,
"grad_norm": 0.4492819164658997,
"learning_rate": 2.8368149141552698e-05,
"loss": 0.2454,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14166270196437836,
"step": 1810,
"valid_targets_mean": 4482.9,
"valid_targets_min": 680
},
{
"epoch": 2.995871180842279,
"grad_norm": 0.4351224233051691,
"learning_rate": 2.8293323943873077e-05,
"loss": 0.2559,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12363643944263458,
"step": 1815,
"valid_targets_mean": 4691.1,
"valid_targets_min": 3958
},
{
"epoch": 3.003303055326177,
"grad_norm": 0.40833317666316626,
"learning_rate": 2.8218358295641315e-05,
"loss": 0.2381,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11818711459636688,
"step": 1820,
"valid_targets_mean": 4326.6,
"valid_targets_min": 3275
},
{
"epoch": 3.0115606936416186,
"grad_norm": 0.43033169982763675,
"learning_rate": 2.8143253466428782e-05,
"loss": 0.2284,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11558468639850616,
"step": 1825,
"valid_targets_mean": 4240.1,
"valid_targets_min": 3343
},
{
"epoch": 3.0198183319570604,
"grad_norm": 0.441977227753935,
"learning_rate": 2.8068010728163942e-05,
"loss": 0.2344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11249755322933197,
"step": 1830,
"valid_targets_mean": 4645.4,
"valid_targets_min": 2379
},
{
"epoch": 3.028075970272502,
"grad_norm": 0.47500489275269636,
"learning_rate": 2.7992631355110786e-05,
"loss": 0.2321,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12710264325141907,
"step": 1835,
"valid_targets_mean": 4751.5,
"valid_targets_min": 3826
},
{
"epoch": 3.036333608587944,
"grad_norm": 0.43810403451820834,
"learning_rate": 2.7917116623847285e-05,
"loss": 0.2426,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10873642563819885,
"step": 1840,
"valid_targets_mean": 4615.5,
"valid_targets_min": 3790
},
{
"epoch": 3.044591246903386,
"grad_norm": 0.502801889716308,
"learning_rate": 2.7841467813243723e-05,
"loss": 0.2415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13157841563224792,
"step": 1845,
"valid_targets_mean": 3915.2,
"valid_targets_min": 2844
},
{
"epoch": 3.0528488852188276,
"grad_norm": 0.4539148542598535,
"learning_rate": 2.7765686204441108e-05,
"loss": 0.2285,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11095467209815979,
"step": 1850,
"valid_targets_mean": 4383.2,
"valid_targets_min": 3239
},
{
"epoch": 3.0611065235342694,
"grad_norm": 0.4786353016926269,
"learning_rate": 2.7689773080829406e-05,
"loss": 0.2401,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1073569804430008,
"step": 1855,
"valid_targets_mean": 3942.9,
"valid_targets_min": 2062
},
{
"epoch": 3.069364161849711,
"grad_norm": 0.44455138253196164,
"learning_rate": 2.7613729728025837e-05,
"loss": 0.2299,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11056707799434662,
"step": 1860,
"valid_targets_mean": 4147.0,
"valid_targets_min": 2892
},
{
"epoch": 3.077621800165153,
"grad_norm": 0.42954762154853665,
"learning_rate": 2.7537557433853116e-05,
"loss": 0.2176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12161555141210556,
"step": 1865,
"valid_targets_mean": 4496.2,
"valid_targets_min": 3460
},
{
"epoch": 3.0858794384805948,
"grad_norm": 0.48475164862460735,
"learning_rate": 2.7461257488317614e-05,
"loss": 0.2341,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10821182280778885,
"step": 1870,
"valid_targets_mean": 3925.4,
"valid_targets_min": 1962
},
{
"epoch": 3.094137076796036,
"grad_norm": 0.41987459239962854,
"learning_rate": 2.738483118358753e-05,
"loss": 0.2336,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12882305681705475,
"step": 1875,
"valid_targets_mean": 5562.9,
"valid_targets_min": 1932
},
{
"epoch": 3.102394715111478,
"grad_norm": 0.42937926308937285,
"learning_rate": 2.7308279813971022e-05,
"loss": 0.2276,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09914682805538177,
"step": 1880,
"valid_targets_mean": 3814.0,
"valid_targets_min": 2016
},
{
"epoch": 3.1106523534269197,
"grad_norm": 0.435724172880294,
"learning_rate": 2.7231604675894226e-05,
"loss": 0.2389,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10737501084804535,
"step": 1885,
"valid_targets_mean": 4369.2,
"valid_targets_min": 3437
},
{
"epoch": 3.1189099917423615,
"grad_norm": 0.485083299602065,
"learning_rate": 2.715480706787939e-05,
"loss": 0.2346,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1658203899860382,
"step": 1890,
"valid_targets_mean": 5167.6,
"valid_targets_min": 3897
},
{
"epoch": 3.1271676300578033,
"grad_norm": 0.4383181344424069,
"learning_rate": 2.70778882905228e-05,
"loss": 0.2253,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10339805483818054,
"step": 1895,
"valid_targets_mean": 4332.9,
"valid_targets_min": 3479
},
{
"epoch": 3.135425268373245,
"grad_norm": 0.4627057085482623,
"learning_rate": 2.7000849646472826e-05,
"loss": 0.2353,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10575290024280548,
"step": 1900,
"valid_targets_mean": 4403.1,
"valid_targets_min": 3662
},
{
"epoch": 3.143682906688687,
"grad_norm": 0.4590406349237048,
"learning_rate": 2.6923692440407784e-05,
"loss": 0.2274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12150724232196808,
"step": 1905,
"valid_targets_mean": 4793.5,
"valid_targets_min": 2747
},
{
"epoch": 3.1519405450041287,
"grad_norm": 0.46304211180998905,
"learning_rate": 2.6846417979013915e-05,
"loss": 0.2347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1174415796995163,
"step": 1910,
"valid_targets_mean": 4188.9,
"valid_targets_min": 2760
},
{
"epoch": 3.1601981833195705,
"grad_norm": 0.44247381520294427,
"learning_rate": 2.676902757096321e-05,
"loss": 0.2298,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10154491662979126,
"step": 1915,
"valid_targets_mean": 4116.2,
"valid_targets_min": 3468
},
{
"epoch": 3.1684558216350123,
"grad_norm": 0.45573804782342364,
"learning_rate": 2.6691522526891258e-05,
"loss": 0.2344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1369427889585495,
"step": 1920,
"valid_targets_mean": 5139.9,
"valid_targets_min": 3663
},
{
"epoch": 3.176713459950454,
"grad_norm": 0.4752873075189136,
"learning_rate": 2.661390415937506e-05,
"loss": 0.2331,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11559513956308365,
"step": 1925,
"valid_targets_mean": 4564.6,
"valid_targets_min": 2649
},
{
"epoch": 3.184971098265896,
"grad_norm": 0.45055693798145796,
"learning_rate": 2.6536173782910782e-05,
"loss": 0.2321,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12754902243614197,
"step": 1930,
"valid_targets_mean": 4952.0,
"valid_targets_min": 3330
},
{
"epoch": 3.1932287365813377,
"grad_norm": 0.44058202022577664,
"learning_rate": 2.645833271389152e-05,
"loss": 0.2347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12682124972343445,
"step": 1935,
"valid_targets_mean": 4670.0,
"valid_targets_min": 4267
},
{
"epoch": 3.2014863748967795,
"grad_norm": 0.44765290394813595,
"learning_rate": 2.6380382270584966e-05,
"loss": 0.2344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11415952444076538,
"step": 1940,
"valid_targets_mean": 4371.9,
"valid_targets_min": 1945
},
{
"epoch": 3.2097440132122212,
"grad_norm": 0.4430500343374567,
"learning_rate": 2.630232377311113e-05,
"loss": 0.2364,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09972745180130005,
"step": 1945,
"valid_targets_mean": 4212.8,
"valid_targets_min": 2661
},
{
"epoch": 3.218001651527663,
"grad_norm": 0.426298399892698,
"learning_rate": 2.622415854341994e-05,
"loss": 0.2292,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11920993030071259,
"step": 1950,
"valid_targets_mean": 4668.5,
"valid_targets_min": 2598
},
{
"epoch": 3.226259289843105,
"grad_norm": 0.4579836516337324,
"learning_rate": 2.6145887905268893e-05,
"loss": 0.2303,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11307698488235474,
"step": 1955,
"valid_targets_mean": 4477.2,
"valid_targets_min": 3236
},
{
"epoch": 3.2345169281585466,
"grad_norm": 0.4468915039745297,
"learning_rate": 2.60675131842006e-05,
"loss": 0.2306,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10588641464710236,
"step": 1960,
"valid_targets_mean": 4326.8,
"valid_targets_min": 3910
},
{
"epoch": 3.2427745664739884,
"grad_norm": 0.45420935740896473,
"learning_rate": 2.5989035707520374e-05,
"loss": 0.2376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1129479706287384,
"step": 1965,
"valid_targets_mean": 4080.2,
"valid_targets_min": 1481
},
{
"epoch": 3.2510322047894302,
"grad_norm": 0.5080682047776788,
"learning_rate": 2.591045680427371e-05,
"loss": 0.2314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11931029707193375,
"step": 1970,
"valid_targets_mean": 4332.1,
"valid_targets_min": 996
},
{
"epoch": 3.259289843104872,
"grad_norm": 0.5735357108744691,
"learning_rate": 2.583177780522382e-05,
"loss": 0.2328,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12322189658880234,
"step": 1975,
"valid_targets_mean": 4229.8,
"valid_targets_min": 3666
},
{
"epoch": 3.267547481420314,
"grad_norm": 0.4697806798669391,
"learning_rate": 2.5753000042829078e-05,
"loss": 0.2492,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1277894228696823,
"step": 1980,
"valid_targets_mean": 4767.2,
"valid_targets_min": 3197
},
{
"epoch": 3.2758051197357556,
"grad_norm": 0.46034534234191765,
"learning_rate": 2.5674124851220422e-05,
"loss": 0.2358,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12803256511688232,
"step": 1985,
"valid_targets_mean": 4699.6,
"valid_targets_min": 4016
},
{
"epoch": 3.2840627580511974,
"grad_norm": 0.4236207117621651,
"learning_rate": 2.5595153566178824e-05,
"loss": 0.2287,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12320274114608765,
"step": 1990,
"valid_targets_mean": 5754.2,
"valid_targets_min": 3561
},
{
"epoch": 3.292320396366639,
"grad_norm": 0.443700768540824,
"learning_rate": 2.5516087525112623e-05,
"loss": 0.2235,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11062918603420258,
"step": 1995,
"valid_targets_mean": 4299.0,
"valid_targets_min": 3877
},
{
"epoch": 3.300578034682081,
"grad_norm": 0.45382805280093996,
"learning_rate": 2.5436928067034876e-05,
"loss": 0.2364,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11447031050920486,
"step": 2000,
"valid_targets_mean": 4136.5,
"valid_targets_min": 1003
},
{
"epoch": 3.308835672997523,
"grad_norm": 0.4400652292692582,
"learning_rate": 2.535767653254071e-05,
"loss": 0.2355,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11634854972362518,
"step": 2005,
"valid_targets_mean": 4331.9,
"valid_targets_min": 2948
},
{
"epoch": 3.3170933113129646,
"grad_norm": 0.4600868318531102,
"learning_rate": 2.5278334263784587e-05,
"loss": 0.2407,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14061662554740906,
"step": 2010,
"valid_targets_mean": 4567.4,
"valid_targets_min": 2671
},
{
"epoch": 3.3253509496284064,
"grad_norm": 0.4372646523259602,
"learning_rate": 2.5198902604457594e-05,
"loss": 0.2387,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10162105411291122,
"step": 2015,
"valid_targets_mean": 4110.0,
"valid_targets_min": 3578
},
{
"epoch": 3.333608587943848,
"grad_norm": 0.45822041157884785,
"learning_rate": 2.511938289976468e-05,
"loss": 0.2291,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10923411697149277,
"step": 2020,
"valid_targets_mean": 3755.9,
"valid_targets_min": 1867
},
{
"epoch": 3.34186622625929,
"grad_norm": 0.4659698505638632,
"learning_rate": 2.503977649640188e-05,
"loss": 0.2311,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10929447412490845,
"step": 2025,
"valid_targets_mean": 3799.4,
"valid_targets_min": 1014
},
{
"epoch": 3.3501238645747318,
"grad_norm": 0.4548067566681893,
"learning_rate": 2.496008474253349e-05,
"loss": 0.2334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1208266019821167,
"step": 2030,
"valid_targets_mean": 4024.5,
"valid_targets_min": 1570
},
{
"epoch": 3.3583815028901736,
"grad_norm": 0.47244784639340753,
"learning_rate": 2.4880308987769262e-05,
"loss": 0.2329,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12616991996765137,
"step": 2035,
"valid_targets_mean": 4285.4,
"valid_targets_min": 2850
},
{
"epoch": 3.3666391412056154,
"grad_norm": 0.45622648080895406,
"learning_rate": 2.4800450583141527e-05,
"loss": 0.2356,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13009721040725708,
"step": 2040,
"valid_targets_mean": 5128.6,
"valid_targets_min": 3750
},
{
"epoch": 3.374896779521057,
"grad_norm": 0.4318829283011845,
"learning_rate": 2.472051088108233e-05,
"loss": 0.232,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1202012151479721,
"step": 2045,
"valid_targets_mean": 4602.5,
"valid_targets_min": 4202
},
{
"epoch": 3.383154417836499,
"grad_norm": 0.42721744760184277,
"learning_rate": 2.4640491235400513e-05,
"loss": 0.2387,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10404972732067108,
"step": 2050,
"valid_targets_mean": 4112.0,
"valid_targets_min": 2031
},
{
"epoch": 3.3914120561519407,
"grad_norm": 0.4557743657214354,
"learning_rate": 2.4560393001258786e-05,
"loss": 0.2379,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1105554848909378,
"step": 2055,
"valid_targets_mean": 4970.5,
"valid_targets_min": 3946
},
{
"epoch": 3.3996696944673825,
"grad_norm": 0.4242529195437994,
"learning_rate": 2.44802175351508e-05,
"loss": 0.24,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11852426826953888,
"step": 2060,
"valid_targets_mean": 4395.9,
"valid_targets_min": 3083
},
{
"epoch": 3.4079273327828243,
"grad_norm": 0.4694069076925552,
"learning_rate": 2.4399966194878158e-05,
"loss": 0.2471,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10787539184093475,
"step": 2065,
"valid_targets_mean": 4131.9,
"valid_targets_min": 2623
},
{
"epoch": 3.416184971098266,
"grad_norm": 0.4343677396166447,
"learning_rate": 2.4319640339527393e-05,
"loss": 0.2308,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10990211367607117,
"step": 2070,
"valid_targets_mean": 4351.2,
"valid_targets_min": 3001
},
{
"epoch": 3.424442609413708,
"grad_norm": 0.4555397957938699,
"learning_rate": 2.4239241329447016e-05,
"loss": 0.2468,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12912330031394958,
"step": 2075,
"valid_targets_mean": 4504.5,
"valid_targets_min": 3793
},
{
"epoch": 3.4327002477291493,
"grad_norm": 0.4462129205169679,
"learning_rate": 2.4158770526224417e-05,
"loss": 0.2289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1056344211101532,
"step": 2080,
"valid_targets_mean": 4155.9,
"valid_targets_min": 3365
},
{
"epoch": 3.440957886044591,
"grad_norm": 0.45279490785318854,
"learning_rate": 2.4078229292662835e-05,
"loss": 0.2321,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11056367307901382,
"step": 2085,
"valid_targets_mean": 4394.0,
"valid_targets_min": 3801
},
{
"epoch": 3.449215524360033,
"grad_norm": 0.4722989521224819,
"learning_rate": 2.399761899275828e-05,
"loss": 0.2328,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13840241730213165,
"step": 2090,
"valid_targets_mean": 4714.4,
"valid_targets_min": 4083
},
{
"epoch": 3.4574731626754747,
"grad_norm": 0.44778262534622487,
"learning_rate": 2.3916940991676417e-05,
"loss": 0.2368,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13277274370193481,
"step": 2095,
"valid_targets_mean": 4692.0,
"valid_targets_min": 3981
},
{
"epoch": 3.4657308009909165,
"grad_norm": 0.4236558697343389,
"learning_rate": 2.3836196655729458e-05,
"loss": 0.2349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10841336846351624,
"step": 2100,
"valid_targets_mean": 4227.6,
"valid_targets_min": 658
},
{
"epoch": 3.4739884393063583,
"grad_norm": 0.43317859134641157,
"learning_rate": 2.375538735235302e-05,
"loss": 0.2362,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11688166111707687,
"step": 2105,
"valid_targets_mean": 4281.4,
"valid_targets_min": 2289
},
{
"epoch": 3.4822460776218,
"grad_norm": 0.43067688568336915,
"learning_rate": 2.3674514450082984e-05,
"loss": 0.2285,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10368673503398895,
"step": 2110,
"valid_targets_mean": 4367.4,
"valid_targets_min": 3834
},
{
"epoch": 3.490503715937242,
"grad_norm": 0.4523816589515863,
"learning_rate": 2.359357931853228e-05,
"loss": 0.2244,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1045714020729065,
"step": 2115,
"valid_targets_mean": 3924.6,
"valid_targets_min": 2008
},
{
"epoch": 3.4987613542526836,
"grad_norm": 0.41409993411194235,
"learning_rate": 2.3512583328367717e-05,
"loss": 0.2397,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14192454516887665,
"step": 2120,
"valid_targets_mean": 5231.1,
"valid_targets_min": 3682
},
{
"epoch": 3.5070189925681254,
"grad_norm": 0.4587051416703706,
"learning_rate": 2.3431527851286782e-05,
"loss": 0.2347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1290726512670517,
"step": 2125,
"valid_targets_mean": 4158.8,
"valid_targets_min": 2242
},
{
"epoch": 3.5152766308835672,
"grad_norm": 0.44555236532467357,
"learning_rate": 2.3350414259994382e-05,
"loss": 0.2361,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12022750824689865,
"step": 2130,
"valid_targets_mean": 3926.9,
"valid_targets_min": 1837
},
{
"epoch": 3.523534269199009,
"grad_norm": 0.46890045880688624,
"learning_rate": 2.326924392817962e-05,
"loss": 0.2433,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15420836210250854,
"step": 2135,
"valid_targets_mean": 5153.0,
"valid_targets_min": 2270
},
{
"epoch": 3.531791907514451,
"grad_norm": 0.46153470295840066,
"learning_rate": 2.318801823049251e-05,
"loss": 0.2433,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13352477550506592,
"step": 2140,
"valid_targets_mean": 4608.5,
"valid_targets_min": 1651
},
{
"epoch": 3.5400495458298926,
"grad_norm": 0.4616450788046373,
"learning_rate": 2.310673854252071e-05,
"loss": 0.2303,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12886768579483032,
"step": 2145,
"valid_targets_mean": 4585.6,
"valid_targets_min": 3824
},
{
"epoch": 3.5483071841453344,
"grad_norm": 0.46958636201080095,
"learning_rate": 2.3025406240766233e-05,
"loss": 0.2307,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12544748187065125,
"step": 2150,
"valid_targets_mean": 4569.6,
"valid_targets_min": 3287
},
{
"epoch": 3.556564822460776,
"grad_norm": 0.4287242434754252,
"learning_rate": 2.2944022702622117e-05,
"loss": 0.2285,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11762858182191849,
"step": 2155,
"valid_targets_mean": 4026.4,
"valid_targets_min": 2237
},
{
"epoch": 3.564822460776218,
"grad_norm": 0.42744548591281417,
"learning_rate": 2.286258930634912e-05,
"loss": 0.2289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11696900427341461,
"step": 2160,
"valid_targets_mean": 4624.0,
"valid_targets_min": 3621
},
{
"epoch": 3.57308009909166,
"grad_norm": 0.4472936403601587,
"learning_rate": 2.2781107431052346e-05,
"loss": 0.2321,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11836177855730057,
"step": 2165,
"valid_targets_mean": 4132.6,
"valid_targets_min": 2309
},
{
"epoch": 3.5813377374071016,
"grad_norm": 0.45205098078595574,
"learning_rate": 2.269957845665792e-05,
"loss": 0.2289,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12267406284809113,
"step": 2170,
"valid_targets_mean": 4613.2,
"valid_targets_min": 4315
},
{
"epoch": 3.5895953757225434,
"grad_norm": 0.4309072717435429,
"learning_rate": 2.261800376388962e-05,
"loss": 0.2266,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10620898753404617,
"step": 2175,
"valid_targets_mean": 4163.1,
"valid_targets_min": 2679
},
{
"epoch": 3.597853014037985,
"grad_norm": 0.4480518343275279,
"learning_rate": 2.2536384734245455e-05,
"loss": 0.2374,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12154752016067505,
"step": 2180,
"valid_targets_mean": 4303.8,
"valid_targets_min": 3561
},
{
"epoch": 3.606110652353427,
"grad_norm": 0.43134054038522407,
"learning_rate": 2.2454722749974315e-05,
"loss": 0.2408,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10952987521886826,
"step": 2185,
"valid_targets_mean": 4234.9,
"valid_targets_min": 3838
},
{
"epoch": 3.6143682906688688,
"grad_norm": 0.43239207162100946,
"learning_rate": 2.237301919405255e-05,
"loss": 0.2334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11773309111595154,
"step": 2190,
"valid_targets_mean": 4306.6,
"valid_targets_min": 3081
},
{
"epoch": 3.6226259289843106,
"grad_norm": 0.4380698590240578,
"learning_rate": 2.229127545016051e-05,
"loss": 0.2299,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1039758175611496,
"step": 2195,
"valid_targets_mean": 4134.0,
"valid_targets_min": 2693
},
{
"epoch": 3.6308835672997524,
"grad_norm": 0.4357978874617094,
"learning_rate": 2.2209492902659183e-05,
"loss": 0.2251,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09786414355039597,
"step": 2200,
"valid_targets_mean": 4353.0,
"valid_targets_min": 2483
},
{
"epoch": 3.639141205615194,
"grad_norm": 0.4286061308285143,
"learning_rate": 2.2127672936566676e-05,
"loss": 0.2342,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11065396666526794,
"step": 2205,
"valid_targets_mean": 4108.2,
"valid_targets_min": 2952
},
{
"epoch": 3.647398843930636,
"grad_norm": 0.44347779646072083,
"learning_rate": 2.204581693753481e-05,
"loss": 0.235,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11825980246067047,
"step": 2210,
"valid_targets_mean": 4439.8,
"valid_targets_min": 3500
},
{
"epoch": 3.6556564822460778,
"grad_norm": 0.45982558047235966,
"learning_rate": 2.196392629182565e-05,
"loss": 0.2318,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10631947219371796,
"step": 2215,
"valid_targets_mean": 4083.6,
"valid_targets_min": 3155
},
{
"epoch": 3.6639141205615195,
"grad_norm": 0.42955303831506925,
"learning_rate": 2.1882002386287983e-05,
"loss": 0.2302,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11716745793819427,
"step": 2220,
"valid_targets_mean": 4808.9,
"valid_targets_min": 3990
},
{
"epoch": 3.6721717588769613,
"grad_norm": 0.43659325327930176,
"learning_rate": 2.1800046608333893e-05,
"loss": 0.2295,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10952255129814148,
"step": 2225,
"valid_targets_mean": 4725.5,
"valid_targets_min": 3779
},
{
"epoch": 3.6804293971924027,
"grad_norm": 0.4353073886336334,
"learning_rate": 2.171806034591522e-05,
"loss": 0.2367,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13252434134483337,
"step": 2230,
"valid_targets_mean": 4644.6,
"valid_targets_min": 3968
},
{
"epoch": 3.6886870355078445,
"grad_norm": 0.4299606139667525,
"learning_rate": 2.163604498750008e-05,
"loss": 0.2384,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11505083739757538,
"step": 2235,
"valid_targets_mean": 4733.6,
"valid_targets_min": 3561
},
{
"epoch": 3.6969446738232863,
"grad_norm": 0.42356916458591837,
"learning_rate": 2.1554001922049333e-05,
"loss": 0.241,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10846732556819916,
"step": 2240,
"valid_targets_mean": 4430.6,
"valid_targets_min": 2558
},
{
"epoch": 3.705202312138728,
"grad_norm": 0.4349502342670851,
"learning_rate": 2.1471932538993063e-05,
"loss": 0.2387,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1261318027973175,
"step": 2245,
"valid_targets_mean": 4698.2,
"valid_targets_min": 2362
},
{
"epoch": 3.71345995045417,
"grad_norm": 0.44357154426300277,
"learning_rate": 2.1389838228207068e-05,
"loss": 0.2306,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10098519921302795,
"step": 2250,
"valid_targets_mean": 3919.2,
"valid_targets_min": 1827
},
{
"epoch": 3.7217175887696117,
"grad_norm": 0.46577747716275053,
"learning_rate": 2.130772037998929e-05,
"loss": 0.2401,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1240251287817955,
"step": 2255,
"valid_targets_mean": 4159.9,
"valid_targets_min": 3425
},
{
"epoch": 3.7299752270850535,
"grad_norm": 0.4792400748746058,
"learning_rate": 2.122558038503631e-05,
"loss": 0.2274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1068907231092453,
"step": 2260,
"valid_targets_mean": 3794.4,
"valid_targets_min": 310
},
{
"epoch": 3.7382328654004953,
"grad_norm": 0.46289372698197473,
"learning_rate": 2.114341963441974e-05,
"loss": 0.2255,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11028460413217545,
"step": 2265,
"valid_targets_mean": 4166.8,
"valid_targets_min": 3243
},
{
"epoch": 3.746490503715937,
"grad_norm": 0.4334196841294762,
"learning_rate": 2.106123951956271e-05,
"loss": 0.2345,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10762996226549149,
"step": 2270,
"valid_targets_mean": 4156.4,
"valid_targets_min": 2593
},
{
"epoch": 3.754748142031379,
"grad_norm": 0.42119163960390404,
"learning_rate": 2.0979041432216318e-05,
"loss": 0.2286,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09685902297496796,
"step": 2275,
"valid_targets_mean": 4116.4,
"valid_targets_min": 3182
},
{
"epoch": 3.7630057803468207,
"grad_norm": 0.45844328154285025,
"learning_rate": 2.0896826764435984e-05,
"loss": 0.2429,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12231326103210449,
"step": 2280,
"valid_targets_mean": 4393.4,
"valid_targets_min": 3766
},
{
"epoch": 3.7712634186622624,
"grad_norm": 0.45329351178250943,
"learning_rate": 2.0814596908557966e-05,
"loss": 0.2347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12290851771831512,
"step": 2285,
"valid_targets_mean": 4443.5,
"valid_targets_min": 1918
},
{
"epoch": 3.7795210569777042,
"grad_norm": 0.41680800085795866,
"learning_rate": 2.073235325717571e-05,
"loss": 0.2268,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10138364136219025,
"step": 2290,
"valid_targets_mean": 4234.6,
"valid_targets_min": 2858
},
{
"epoch": 3.787778695293146,
"grad_norm": 0.46648417002289044,
"learning_rate": 2.0650097203116308e-05,
"loss": 0.2316,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11494442820549011,
"step": 2295,
"valid_targets_mean": 4481.9,
"valid_targets_min": 3077
},
{
"epoch": 3.796036333608588,
"grad_norm": 0.4524735573711148,
"learning_rate": 2.0567830139416895e-05,
"loss": 0.2433,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1235344260931015,
"step": 2300,
"valid_targets_mean": 4136.2,
"valid_targets_min": 2485
},
{
"epoch": 3.8042939719240296,
"grad_norm": 0.4145053625244723,
"learning_rate": 2.0485553459301058e-05,
"loss": 0.2328,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11331503838300705,
"step": 2305,
"valid_targets_mean": 4390.1,
"valid_targets_min": 3705
},
{
"epoch": 3.8125516102394714,
"grad_norm": 0.42652547444847605,
"learning_rate": 2.0403268556155237e-05,
"loss": 0.2332,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12065534293651581,
"step": 2310,
"valid_targets_mean": 4324.2,
"valid_targets_min": 2995
},
{
"epoch": 3.820809248554913,
"grad_norm": 0.43971818618897096,
"learning_rate": 2.0320976823505135e-05,
"loss": 0.229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12459652125835419,
"step": 2315,
"valid_targets_mean": 4709.0,
"valid_targets_min": 3656
},
{
"epoch": 3.829066886870355,
"grad_norm": 0.5418474339863071,
"learning_rate": 2.0238679654992115e-05,
"loss": 0.2341,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12130506336688995,
"step": 2320,
"valid_targets_mean": 3935.0,
"valid_targets_min": 1931
},
{
"epoch": 3.837324525185797,
"grad_norm": 0.42667829831083975,
"learning_rate": 2.0156378444349597e-05,
"loss": 0.2274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10672591626644135,
"step": 2325,
"valid_targets_mean": 4143.2,
"valid_targets_min": 3445
},
{
"epoch": 3.8455821635012386,
"grad_norm": 0.46462321820398467,
"learning_rate": 2.0074074585379466e-05,
"loss": 0.2331,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11428795754909515,
"step": 2330,
"valid_targets_mean": 4783.6,
"valid_targets_min": 3962
},
{
"epoch": 3.8538398018166804,
"grad_norm": 0.4307322408516787,
"learning_rate": 1.999176947192844e-05,
"loss": 0.2335,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10569777339696884,
"step": 2335,
"valid_targets_mean": 4360.4,
"valid_targets_min": 3655
},
{
"epoch": 3.862097440132122,
"grad_norm": 0.45991717970431834,
"learning_rate": 1.9909464497864487e-05,
"loss": 0.2371,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11744770407676697,
"step": 2340,
"valid_targets_mean": 4432.9,
"valid_targets_min": 3721
},
{
"epoch": 3.870355078447564,
"grad_norm": 0.44244657286280636,
"learning_rate": 1.9827161057053245e-05,
"loss": 0.227,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11885372549295425,
"step": 2345,
"valid_targets_mean": 4550.2,
"valid_targets_min": 3705
},
{
"epoch": 3.878612716763006,
"grad_norm": 0.429137784693554,
"learning_rate": 1.9744860543334324e-05,
"loss": 0.2294,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10367788374423981,
"step": 2350,
"valid_targets_mean": 4173.2,
"valid_targets_min": 3414
},
{
"epoch": 3.8868703550784476,
"grad_norm": 0.44075963541618607,
"learning_rate": 1.966256435049782e-05,
"loss": 0.2326,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11907249689102173,
"step": 2355,
"valid_targets_mean": 4532.9,
"valid_targets_min": 3717
},
{
"epoch": 3.8951279933938894,
"grad_norm": 0.40226326370628873,
"learning_rate": 1.9580273872260623e-05,
"loss": 0.226,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1086915135383606,
"step": 2360,
"valid_targets_mean": 4166.6,
"valid_targets_min": 3727
},
{
"epoch": 3.903385631709331,
"grad_norm": 0.41907620445907073,
"learning_rate": 1.949799050224286e-05,
"loss": 0.2331,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12478712201118469,
"step": 2365,
"valid_targets_mean": 4591.5,
"valid_targets_min": 3654
},
{
"epoch": 3.911643270024773,
"grad_norm": 0.4321443643342046,
"learning_rate": 1.9415715633944264e-05,
"loss": 0.238,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11555442214012146,
"step": 2370,
"valid_targets_mean": 4362.8,
"valid_targets_min": 3300
},
{
"epoch": 3.9199009083402148,
"grad_norm": 0.42891020586167805,
"learning_rate": 1.933345066072059e-05,
"loss": 0.2347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1171942949295044,
"step": 2375,
"valid_targets_mean": 4651.1,
"valid_targets_min": 2919
},
{
"epoch": 3.9281585466556566,
"grad_norm": 0.4521886258284723,
"learning_rate": 1.9251196975760036e-05,
"loss": 0.2369,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11669059097766876,
"step": 2380,
"valid_targets_mean": 3741.4,
"valid_targets_min": 1735
},
{
"epoch": 3.9364161849710984,
"grad_norm": 0.4162914061511045,
"learning_rate": 1.9168955972059597e-05,
"loss": 0.2216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10332491993904114,
"step": 2385,
"valid_targets_mean": 4432.1,
"valid_targets_min": 3731
},
{
"epoch": 3.94467382328654,
"grad_norm": 0.4301184178514016,
"learning_rate": 1.9086729042401525e-05,
"loss": 0.2354,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12041283398866653,
"step": 2390,
"valid_targets_mean": 4266.0,
"valid_targets_min": 3735
},
{
"epoch": 3.952931461601982,
"grad_norm": 0.43548860705017434,
"learning_rate": 1.900451757932973e-05,
"loss": 0.2274,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11648362874984741,
"step": 2395,
"valid_targets_mean": 4482.6,
"valid_targets_min": 3674
},
{
"epoch": 3.9611890999174237,
"grad_norm": 0.49226120355453196,
"learning_rate": 1.8922322975126172e-05,
"loss": 0.2276,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10247261822223663,
"step": 2400,
"valid_targets_mean": 4260.0,
"valid_targets_min": 3418
},
{
"epoch": 3.9694467382328655,
"grad_norm": 0.48011005467390555,
"learning_rate": 1.884014662178731e-05,
"loss": 0.229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1003057062625885,
"step": 2405,
"valid_targets_mean": 3035.4,
"valid_targets_min": 680
},
{
"epoch": 3.9777043765483073,
"grad_norm": 0.49626914710856807,
"learning_rate": 1.8757989911000512e-05,
"loss": 0.2262,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11136093735694885,
"step": 2410,
"valid_targets_mean": 3901.8,
"valid_targets_min": 2390
},
{
"epoch": 3.985962014863749,
"grad_norm": 0.42775545009854876,
"learning_rate": 1.8675854234120506e-05,
"loss": 0.2365,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11673128604888916,
"step": 2415,
"valid_targets_mean": 4471.6,
"valid_targets_min": 4106
},
{
"epoch": 3.994219653179191,
"grad_norm": 0.43897438033166525,
"learning_rate": 1.8593740982145775e-05,
"loss": 0.233,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1342780888080597,
"step": 2420,
"valid_targets_mean": 4895.6,
"valid_targets_min": 4365
},
{
"epoch": 4.001651527663088,
"grad_norm": 0.43138655213413063,
"learning_rate": 1.8511651545695057e-05,
"loss": 0.224,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11240030825138092,
"step": 2425,
"valid_targets_mean": 4366.8,
"valid_targets_min": 3517
},
{
"epoch": 4.00990916597853,
"grad_norm": 0.42432731377937705,
"learning_rate": 1.842958731498376e-05,
"loss": 0.2297,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11508003622293472,
"step": 2430,
"valid_targets_mean": 4424.1,
"valid_targets_min": 3012
},
{
"epoch": 4.018166804293972,
"grad_norm": 0.4653900092446504,
"learning_rate": 1.8347549679800397e-05,
"loss": 0.2292,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1034216433763504,
"step": 2435,
"valid_targets_mean": 4068.9,
"valid_targets_min": 3201
},
{
"epoch": 4.026424442609414,
"grad_norm": 0.43375080384700787,
"learning_rate": 1.826554002948311e-05,
"loss": 0.2185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11905544996261597,
"step": 2440,
"valid_targets_mean": 4631.2,
"valid_targets_min": 3675
},
{
"epoch": 4.034682080924855,
"grad_norm": 0.4502351260619515,
"learning_rate": 1.818355975289608e-05,
"loss": 0.2209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10532905161380768,
"step": 2445,
"valid_targets_mean": 4504.5,
"valid_targets_min": 1836
},
{
"epoch": 4.042939719240297,
"grad_norm": 0.47829412298584406,
"learning_rate": 1.810161023840607e-05,
"loss": 0.2263,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1118680089712143,
"step": 2450,
"valid_targets_mean": 4594.2,
"valid_targets_min": 3466
},
{
"epoch": 4.051197357555739,
"grad_norm": 0.49253311626613555,
"learning_rate": 1.8019692873858824e-05,
"loss": 0.2244,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09949852526187897,
"step": 2455,
"valid_targets_mean": 4149.1,
"valid_targets_min": 3326
},
{
"epoch": 4.059454995871181,
"grad_norm": 0.4425596584377973,
"learning_rate": 1.793780904655565e-05,
"loss": 0.2211,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10728386789560318,
"step": 2460,
"valid_targets_mean": 4114.6,
"valid_targets_min": 3248
},
{
"epoch": 4.0677126341866225,
"grad_norm": 0.45790124247020314,
"learning_rate": 1.785596014322989e-05,
"loss": 0.2166,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11399378627538681,
"step": 2465,
"valid_targets_mean": 4518.5,
"valid_targets_min": 3838
},
{
"epoch": 4.075970272502064,
"grad_norm": 0.508368371744764,
"learning_rate": 1.7774147550023414e-05,
"loss": 0.2207,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10383783280849457,
"step": 2470,
"valid_targets_mean": 3627.2,
"valid_targets_min": 1661
},
{
"epoch": 4.084227910817506,
"grad_norm": 0.4581960078981182,
"learning_rate": 1.7692372652463178e-05,
"loss": 0.2174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0977829173207283,
"step": 2475,
"valid_targets_mean": 4031.9,
"valid_targets_min": 2477
},
{
"epoch": 4.092485549132948,
"grad_norm": 0.4730284720091334,
"learning_rate": 1.7610636835437763e-05,
"loss": 0.2219,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12772221863269806,
"step": 2480,
"valid_targets_mean": 4351.8,
"valid_targets_min": 3070
},
{
"epoch": 4.10074318744839,
"grad_norm": 0.5380904257464577,
"learning_rate": 1.752894148317388e-05,
"loss": 0.2198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12358734011650085,
"step": 2485,
"valid_targets_mean": 4417.6,
"valid_targets_min": 2844
},
{
"epoch": 4.1090008257638315,
"grad_norm": 0.4351070430802422,
"learning_rate": 1.744728797921297e-05,
"loss": 0.2227,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12527084350585938,
"step": 2490,
"valid_targets_mean": 4680.2,
"valid_targets_min": 3758
},
{
"epoch": 4.117258464079273,
"grad_norm": 0.4325960822324002,
"learning_rate": 1.7365677706387757e-05,
"loss": 0.2224,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11105137318372726,
"step": 2495,
"valid_targets_mean": 4439.4,
"valid_targets_min": 3067
},
{
"epoch": 4.125516102394715,
"grad_norm": 0.48174653676956386,
"learning_rate": 1.7284112046798833e-05,
"loss": 0.2199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10959590971469879,
"step": 2500,
"valid_targets_mean": 5102.6,
"valid_targets_min": 3298
},
{
"epoch": 4.133773740710157,
"grad_norm": 0.5221397794128658,
"learning_rate": 1.7202592381791222e-05,
"loss": 0.2211,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11155468970537186,
"step": 2505,
"valid_targets_mean": 4185.4,
"valid_targets_min": 1607
},
{
"epoch": 4.142031379025599,
"grad_norm": 0.4738165947262898,
"learning_rate": 1.712112009193105e-05,
"loss": 0.2192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1028067022562027,
"step": 2510,
"valid_targets_mean": 3804.9,
"valid_targets_min": 1481
},
{
"epoch": 4.1502890173410405,
"grad_norm": 0.43404080170052983,
"learning_rate": 1.7039696556982105e-05,
"loss": 0.2311,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13499866425991058,
"step": 2515,
"valid_targets_mean": 4913.2,
"valid_targets_min": 3666
},
{
"epoch": 4.158546655656482,
"grad_norm": 0.47049004496066815,
"learning_rate": 1.6958323155882485e-05,
"loss": 0.2278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12356163561344147,
"step": 2520,
"valid_targets_mean": 4724.2,
"valid_targets_min": 3529
},
{
"epoch": 4.166804293971924,
"grad_norm": 0.46703234912476077,
"learning_rate": 1.687700126672128e-05,
"loss": 0.2253,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09205446392297745,
"step": 2525,
"valid_targets_mean": 3819.2,
"valid_targets_min": 2128
},
{
"epoch": 4.175061932287366,
"grad_norm": 0.4781620300592793,
"learning_rate": 1.6795732266715188e-05,
"loss": 0.2175,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12023566663265228,
"step": 2530,
"valid_targets_mean": 4254.9,
"valid_targets_min": 3499
},
{
"epoch": 4.183319570602808,
"grad_norm": 0.462167285274948,
"learning_rate": 1.67145175321852e-05,
"loss": 0.2199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09703060984611511,
"step": 2535,
"valid_targets_mean": 4193.9,
"valid_targets_min": 3011
},
{
"epoch": 4.1915772089182495,
"grad_norm": 0.4432752153783257,
"learning_rate": 1.6633358438533318e-05,
"loss": 0.2206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1219254583120346,
"step": 2540,
"valid_targets_mean": 4617.9,
"valid_targets_min": 3978
},
{
"epoch": 4.199834847233691,
"grad_norm": 0.46438818911841734,
"learning_rate": 1.655225636021924e-05,
"loss": 0.2278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12295781075954437,
"step": 2545,
"valid_targets_mean": 4195.4,
"valid_targets_min": 2967
},
{
"epoch": 4.208092485549133,
"grad_norm": 0.4639169958518085,
"learning_rate": 1.6471212670737092e-05,
"loss": 0.2146,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11131924390792847,
"step": 2550,
"valid_targets_mean": 3854.0,
"valid_targets_min": 2975
},
{
"epoch": 4.216350123864575,
"grad_norm": 0.4929084913870907,
"learning_rate": 1.6390228742592153e-05,
"loss": 0.23,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1357860118150711,
"step": 2555,
"valid_targets_mean": 4844.0,
"valid_targets_min": 3060
},
{
"epoch": 4.224607762180017,
"grad_norm": 0.4793882745149162,
"learning_rate": 1.630930594727762e-05,
"loss": 0.2225,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1064787209033966,
"step": 2560,
"valid_targets_mean": 4284.2,
"valid_targets_min": 2905
},
{
"epoch": 4.2328654004954585,
"grad_norm": 0.42999246058421575,
"learning_rate": 1.6228445655251405e-05,
"loss": 0.2222,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11736476421356201,
"step": 2565,
"valid_targets_mean": 5082.6,
"valid_targets_min": 3365
},
{
"epoch": 4.2411230388109,
"grad_norm": 0.46522370822280235,
"learning_rate": 1.6147649235912874e-05,
"loss": 0.2208,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10605399310588837,
"step": 2570,
"valid_targets_mean": 4093.4,
"valid_targets_min": 2130
},
{
"epoch": 4.249380677126342,
"grad_norm": 0.4160563893418732,
"learning_rate": 1.6066918057579693e-05,
"loss": 0.2202,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10428132116794586,
"step": 2575,
"valid_targets_mean": 4302.2,
"valid_targets_min": 3350
},
{
"epoch": 4.257638315441784,
"grad_norm": 0.43779549968283665,
"learning_rate": 1.5986253487464674e-05,
"loss": 0.2198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1027291864156723,
"step": 2580,
"valid_targets_mean": 4795.8,
"valid_targets_min": 4126
},
{
"epoch": 4.265895953757226,
"grad_norm": 0.4615658324369947,
"learning_rate": 1.5905656891652555e-05,
"loss": 0.2216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10856552422046661,
"step": 2585,
"valid_targets_mean": 4111.1,
"valid_targets_min": 3352
},
{
"epoch": 4.274153592072667,
"grad_norm": 0.44678949995378664,
"learning_rate": 1.5825129635076923e-05,
"loss": 0.221,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10645823180675507,
"step": 2590,
"valid_targets_mean": 4201.5,
"valid_targets_min": 3585
},
{
"epoch": 4.282411230388109,
"grad_norm": 0.4416358495944346,
"learning_rate": 1.5744673081497083e-05,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11892388015985489,
"step": 2595,
"valid_targets_mean": 5106.5,
"valid_targets_min": 4042
},
{
"epoch": 4.290668868703551,
"grad_norm": 0.41651874151376506,
"learning_rate": 1.5664288593474955e-05,
"loss": 0.2248,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11502254009246826,
"step": 2600,
"valid_targets_mean": 4494.4,
"valid_targets_min": 3694
},
{
"epoch": 4.298926507018993,
"grad_norm": 0.47432752097864367,
"learning_rate": 1.558397753235198e-05,
"loss": 0.2148,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11153976619243622,
"step": 2605,
"valid_targets_mean": 4518.8,
"valid_targets_min": 3529
},
{
"epoch": 4.307184145334435,
"grad_norm": 0.45737440215312447,
"learning_rate": 1.550374125822613e-05,
"loss": 0.2178,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12691405415534973,
"step": 2610,
"valid_targets_mean": 5188.0,
"valid_targets_min": 3468
},
{
"epoch": 4.315441783649876,
"grad_norm": 0.4727255872582784,
"learning_rate": 1.54235811299288e-05,
"loss": 0.2216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15137368440628052,
"step": 2615,
"valid_targets_mean": 5478.6,
"valid_targets_min": 4232
},
{
"epoch": 4.323699421965318,
"grad_norm": 0.4433399163605833,
"learning_rate": 1.534349850500182e-05,
"loss": 0.2235,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11273278295993805,
"step": 2620,
"valid_targets_mean": 4209.6,
"valid_targets_min": 3300
},
{
"epoch": 4.33195706028076,
"grad_norm": 0.43665317661238773,
"learning_rate": 1.52634947396745e-05,
"loss": 0.2193,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09157262742519379,
"step": 2625,
"valid_targets_mean": 3358.9,
"valid_targets_min": 1042
},
{
"epoch": 4.340214698596202,
"grad_norm": 0.4744904293962483,
"learning_rate": 1.5183571188840622e-05,
"loss": 0.2241,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10983486473560333,
"step": 2630,
"valid_targets_mean": 4229.0,
"valid_targets_min": 2069
},
{
"epoch": 4.348472336911644,
"grad_norm": 0.44280004806121537,
"learning_rate": 1.5103729206035482e-05,
"loss": 0.2225,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08512726426124573,
"step": 2635,
"valid_targets_mean": 4101.0,
"valid_targets_min": 3811
},
{
"epoch": 4.356729975227085,
"grad_norm": 0.4534040062889216,
"learning_rate": 1.5023970143413029e-05,
"loss": 0.216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0986454039812088,
"step": 2640,
"valid_targets_mean": 4242.5,
"valid_targets_min": 3314
},
{
"epoch": 4.364987613542527,
"grad_norm": 0.4562639868504877,
"learning_rate": 1.4944295351722898e-05,
"loss": 0.2212,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1408693492412567,
"step": 2645,
"valid_targets_mean": 5239.1,
"valid_targets_min": 3538
},
{
"epoch": 4.373245251857969,
"grad_norm": 0.43448409977432206,
"learning_rate": 1.486470618028759e-05,
"loss": 0.2179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.107313372194767,
"step": 2650,
"valid_targets_mean": 4680.8,
"valid_targets_min": 4084
},
{
"epoch": 4.381502890173411,
"grad_norm": 0.4341198000261488,
"learning_rate": 1.4785203976979565e-05,
"loss": 0.2204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10691449791193008,
"step": 2655,
"valid_targets_mean": 4219.4,
"valid_targets_min": 2737
},
{
"epoch": 4.389760528488853,
"grad_norm": 0.471178966716307,
"learning_rate": 1.4705790088198461e-05,
"loss": 0.2176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1182539239525795,
"step": 2660,
"valid_targets_mean": 4362.5,
"valid_targets_min": 3834
},
{
"epoch": 4.398018166804294,
"grad_norm": 0.45468901217649416,
"learning_rate": 1.4626465858848293e-05,
"loss": 0.2238,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14260879158973694,
"step": 2665,
"valid_targets_mean": 5426.1,
"valid_targets_min": 4223
},
{
"epoch": 4.406275805119736,
"grad_norm": 0.45479980291464306,
"learning_rate": 1.4547232632314624e-05,
"loss": 0.2267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1140911877155304,
"step": 2670,
"valid_targets_mean": 4752.0,
"valid_targets_min": 4069
},
{
"epoch": 4.414533443435178,
"grad_norm": 0.46381999126218804,
"learning_rate": 1.4468091750441875e-05,
"loss": 0.2294,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11255607008934021,
"step": 2675,
"valid_targets_mean": 4597.6,
"valid_targets_min": 3821
},
{
"epoch": 4.422791081750619,
"grad_norm": 0.4427551259618707,
"learning_rate": 1.4389044553510572e-05,
"loss": 0.2165,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10426762700080872,
"step": 2680,
"valid_targets_mean": 4264.2,
"valid_targets_min": 3296
},
{
"epoch": 4.431048720066061,
"grad_norm": 0.45937106851442844,
"learning_rate": 1.431009238021465e-05,
"loss": 0.2249,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12401307374238968,
"step": 2685,
"valid_targets_mean": 4556.6,
"valid_targets_min": 3607
},
{
"epoch": 4.4393063583815024,
"grad_norm": 0.4253698245312328,
"learning_rate": 1.423123656763877e-05,
"loss": 0.217,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11196266114711761,
"step": 2690,
"valid_targets_mean": 4481.2,
"valid_targets_min": 3861
},
{
"epoch": 4.447563996696944,
"grad_norm": 0.4917414152447796,
"learning_rate": 1.4152478451235717e-05,
"loss": 0.2174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11584701389074326,
"step": 2695,
"valid_targets_mean": 4419.8,
"valid_targets_min": 3562
},
{
"epoch": 4.455821635012386,
"grad_norm": 0.4731160195869125,
"learning_rate": 1.4073819364803729e-05,
"loss": 0.2235,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1256760209798813,
"step": 2700,
"valid_targets_mean": 4355.6,
"valid_targets_min": 3689
},
{
"epoch": 4.464079273327828,
"grad_norm": 0.4397341766270852,
"learning_rate": 1.399526064046394e-05,
"loss": 0.2343,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10881868004798889,
"step": 2705,
"valid_targets_mean": 4300.5,
"valid_targets_min": 3532
},
{
"epoch": 4.47233691164327,
"grad_norm": 0.45845456142202395,
"learning_rate": 1.3916803608637818e-05,
"loss": 0.2098,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1284886598587036,
"step": 2710,
"valid_targets_mean": 4722.2,
"valid_targets_min": 2790
},
{
"epoch": 4.480594549958711,
"grad_norm": 0.4569492298111738,
"learning_rate": 1.383844959802464e-05,
"loss": 0.224,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0997026264667511,
"step": 2715,
"valid_targets_mean": 4096.5,
"valid_targets_min": 3003
},
{
"epoch": 4.488852188274153,
"grad_norm": 0.44996819861311854,
"learning_rate": 1.376019993557895e-05,
"loss": 0.2281,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11038827896118164,
"step": 2720,
"valid_targets_mean": 4620.6,
"valid_targets_min": 4029
},
{
"epoch": 4.497109826589595,
"grad_norm": 0.4586506993779925,
"learning_rate": 1.3682055946488153e-05,
"loss": 0.2174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12795566022396088,
"step": 2725,
"valid_targets_mean": 4619.1,
"valid_targets_min": 3735
},
{
"epoch": 4.505367464905037,
"grad_norm": 0.4405743669653405,
"learning_rate": 1.360401895415001e-05,
"loss": 0.2239,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10197857022285461,
"step": 2730,
"valid_targets_mean": 4325.9,
"valid_targets_min": 3920
},
{
"epoch": 4.513625103220479,
"grad_norm": 0.4537330141999022,
"learning_rate": 1.352609028015027e-05,
"loss": 0.2191,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12227954715490341,
"step": 2735,
"valid_targets_mean": 4891.4,
"valid_targets_min": 3835
},
{
"epoch": 4.52188274153592,
"grad_norm": 0.4765935959395887,
"learning_rate": 1.3448271244240252e-05,
"loss": 0.2157,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11824171245098114,
"step": 2740,
"valid_targets_mean": 4496.6,
"valid_targets_min": 2163
},
{
"epoch": 4.530140379851362,
"grad_norm": 0.44379988605478604,
"learning_rate": 1.3370563164314523e-05,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09580877423286438,
"step": 2745,
"valid_targets_mean": 4023.5,
"valid_targets_min": 2612
},
{
"epoch": 4.538398018166804,
"grad_norm": 0.5080483331476907,
"learning_rate": 1.3292967356388578e-05,
"loss": 0.2282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11642448604106903,
"step": 2750,
"valid_targets_mean": 4632.1,
"valid_targets_min": 3346
},
{
"epoch": 4.546655656482246,
"grad_norm": 0.4268925399255378,
"learning_rate": 1.321548513457652e-05,
"loss": 0.2279,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10125703364610672,
"step": 2755,
"valid_targets_mean": 4425.1,
"valid_targets_min": 3511
},
{
"epoch": 4.554913294797688,
"grad_norm": 0.47323191498483536,
"learning_rate": 1.3138117811068845e-05,
"loss": 0.2195,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12184475362300873,
"step": 2760,
"valid_targets_mean": 4379.4,
"valid_targets_min": 2253
},
{
"epoch": 4.563170933113129,
"grad_norm": 0.4428165117259587,
"learning_rate": 1.3060866696110213e-05,
"loss": 0.2215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10011693835258484,
"step": 2765,
"valid_targets_mean": 4029.9,
"valid_targets_min": 1604
},
{
"epoch": 4.571428571428571,
"grad_norm": 0.4333274524314617,
"learning_rate": 1.298373309797722e-05,
"loss": 0.2319,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11478620767593384,
"step": 2770,
"valid_targets_mean": 4555.0,
"valid_targets_min": 3760
},
{
"epoch": 4.579686209744013,
"grad_norm": 0.44326881317262595,
"learning_rate": 1.290671832295629e-05,
"loss": 0.2214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12019691616296768,
"step": 2775,
"valid_targets_mean": 5197.5,
"valid_targets_min": 3719
},
{
"epoch": 4.587943848059455,
"grad_norm": 0.449310966227182,
"learning_rate": 1.2829823675321535e-05,
"loss": 0.2279,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10975726693868637,
"step": 2780,
"valid_targets_mean": 4447.4,
"valid_targets_min": 3984
},
{
"epoch": 4.596201486374897,
"grad_norm": 0.4535122958868912,
"learning_rate": 1.275305045731266e-05,
"loss": 0.2157,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11039043217897415,
"step": 2785,
"valid_targets_mean": 4720.9,
"valid_targets_min": 3062
},
{
"epoch": 4.604459124690338,
"grad_norm": 0.4523172027845556,
"learning_rate": 1.26763999691129e-05,
"loss": 0.2206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09925751388072968,
"step": 2790,
"valid_targets_mean": 4322.6,
"valid_targets_min": 3155
},
{
"epoch": 4.61271676300578,
"grad_norm": 0.45631290850693274,
"learning_rate": 1.259987350882704e-05,
"loss": 0.2253,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10459814965724945,
"step": 2795,
"valid_targets_mean": 4116.1,
"valid_targets_min": 2475
},
{
"epoch": 4.620974401321222,
"grad_norm": 0.4480923640476862,
"learning_rate": 1.252347237245939e-05,
"loss": 0.2167,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10976381599903107,
"step": 2800,
"valid_targets_mean": 4304.0,
"valid_targets_min": 3791
},
{
"epoch": 4.629232039636664,
"grad_norm": 0.4572100355980429,
"learning_rate": 1.2447197853891848e-05,
"loss": 0.221,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10427606850862503,
"step": 2805,
"valid_targets_mean": 4567.8,
"valid_targets_min": 3271
},
{
"epoch": 4.6374896779521055,
"grad_norm": 0.450626835284948,
"learning_rate": 1.237105124486201e-05,
"loss": 0.2234,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13489267230033875,
"step": 2810,
"valid_targets_mean": 5020.4,
"valid_targets_min": 3026
},
{
"epoch": 4.645747316267547,
"grad_norm": 0.44374024837093445,
"learning_rate": 1.2295033834941266e-05,
"loss": 0.2297,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0997210294008255,
"step": 2815,
"valid_targets_mean": 4172.2,
"valid_targets_min": 2148
},
{
"epoch": 4.654004954582989,
"grad_norm": 0.4244780804843001,
"learning_rate": 1.2219146911512958e-05,
"loss": 0.2252,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09132840484380722,
"step": 2820,
"valid_targets_mean": 4299.9,
"valid_targets_min": 2784
},
{
"epoch": 4.662262592898431,
"grad_norm": 0.4704454232519091,
"learning_rate": 1.2143391759750607e-05,
"loss": 0.2201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11313526332378387,
"step": 2825,
"valid_targets_mean": 3925.9,
"valid_targets_min": 781
},
{
"epoch": 4.670520231213873,
"grad_norm": 0.4411033014624,
"learning_rate": 1.206776966259613e-05,
"loss": 0.2229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10199315845966339,
"step": 2830,
"valid_targets_mean": 4343.1,
"valid_targets_min": 3845
},
{
"epoch": 4.6787778695293145,
"grad_norm": 0.44738043799105326,
"learning_rate": 1.1992281900738121e-05,
"loss": 0.2247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10181967169046402,
"step": 2835,
"valid_targets_mean": 4275.5,
"valid_targets_min": 2369
},
{
"epoch": 4.687035507844756,
"grad_norm": 0.4518968894518391,
"learning_rate": 1.1916929752590126e-05,
"loss": 0.222,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1144866943359375,
"step": 2840,
"valid_targets_mean": 4326.1,
"valid_targets_min": 3842
},
{
"epoch": 4.695293146160198,
"grad_norm": 0.4309390352843624,
"learning_rate": 1.1841714494269051e-05,
"loss": 0.226,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09399577975273132,
"step": 2845,
"valid_targets_mean": 4177.0,
"valid_targets_min": 2275
},
{
"epoch": 4.70355078447564,
"grad_norm": 0.431834592555752,
"learning_rate": 1.1766637399573517e-05,
"loss": 0.2167,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11416701972484589,
"step": 2850,
"valid_targets_mean": 4389.6,
"valid_targets_min": 2918
},
{
"epoch": 4.711808422791082,
"grad_norm": 0.4721205685454713,
"learning_rate": 1.1691699739962275e-05,
"loss": 0.2304,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13181836903095245,
"step": 2855,
"valid_targets_mean": 4444.1,
"valid_targets_min": 3536
},
{
"epoch": 4.7200660611065235,
"grad_norm": 0.43826928979328167,
"learning_rate": 1.1616902784532711e-05,
"loss": 0.226,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10028581321239471,
"step": 2860,
"valid_targets_mean": 4279.0,
"valid_targets_min": 3314
},
{
"epoch": 4.728323699421965,
"grad_norm": 0.4639641582468062,
"learning_rate": 1.1542247799999328e-05,
"loss": 0.2181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09799149632453918,
"step": 2865,
"valid_targets_mean": 4085.8,
"valid_targets_min": 1962
},
{
"epoch": 4.736581337737407,
"grad_norm": 0.4490454765485913,
"learning_rate": 1.146773605067228e-05,
"loss": 0.2219,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09673963487148285,
"step": 2870,
"valid_targets_mean": 3926.6,
"valid_targets_min": 2186
},
{
"epoch": 4.744838976052849,
"grad_norm": 0.43121537078305067,
"learning_rate": 1.1393368798436007e-05,
"loss": 0.2113,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09363870322704315,
"step": 2875,
"valid_targets_mean": 4113.8,
"valid_targets_min": 2016
},
{
"epoch": 4.753096614368291,
"grad_norm": 0.46731668889547245,
"learning_rate": 1.1319147302727827e-05,
"loss": 0.2276,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08646662533283234,
"step": 2880,
"valid_targets_mean": 3764.1,
"valid_targets_min": 2966
},
{
"epoch": 4.7613542526837325,
"grad_norm": 0.45238305902266057,
"learning_rate": 1.1245072820516606e-05,
"loss": 0.2243,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12130767852067947,
"step": 2885,
"valid_targets_mean": 5117.2,
"valid_targets_min": 4171
},
{
"epoch": 4.769611890999174,
"grad_norm": 0.4285269625583215,
"learning_rate": 1.1171146606281482e-05,
"loss": 0.221,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10104897618293762,
"step": 2890,
"valid_targets_mean": 4003.9,
"valid_targets_min": 1915
},
{
"epoch": 4.777869529314616,
"grad_norm": 0.46807997245038174,
"learning_rate": 1.1097369911990637e-05,
"loss": 0.229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11494791507720947,
"step": 2895,
"valid_targets_mean": 5001.0,
"valid_targets_min": 3933
},
{
"epoch": 4.786127167630058,
"grad_norm": 0.43120045123293865,
"learning_rate": 1.1023743987080064e-05,
"loss": 0.2237,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10903510451316833,
"step": 2900,
"valid_targets_mean": 4232.1,
"valid_targets_min": 3601
},
{
"epoch": 4.7943848059455,
"grad_norm": 0.4338993550822451,
"learning_rate": 1.0950270078432412e-05,
"loss": 0.2258,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09605984389781952,
"step": 2905,
"valid_targets_mean": 4176.2,
"valid_targets_min": 3555
},
{
"epoch": 4.8026424442609414,
"grad_norm": 0.4432216410804114,
"learning_rate": 1.0876949430355904e-05,
"loss": 0.2254,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13977304100990295,
"step": 2910,
"valid_targets_mean": 6112.8,
"valid_targets_min": 3892
},
{
"epoch": 4.810900082576383,
"grad_norm": 0.45630817915186134,
"learning_rate": 1.08037832845632e-05,
"loss": 0.2218,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10191337764263153,
"step": 2915,
"valid_targets_mean": 4542.1,
"valid_targets_min": 4084
},
{
"epoch": 4.819157720891825,
"grad_norm": 0.48250171720955,
"learning_rate": 1.0730772880150445e-05,
"loss": 0.2209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10146736353635788,
"step": 2920,
"valid_targets_mean": 4839.4,
"valid_targets_min": 3823
},
{
"epoch": 4.827415359207267,
"grad_norm": 0.47769385657387303,
"learning_rate": 1.0657919453576213e-05,
"loss": 0.2248,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12479188293218613,
"step": 2925,
"valid_targets_mean": 4133.6,
"valid_targets_min": 2940
},
{
"epoch": 4.835672997522709,
"grad_norm": 0.42123468302775374,
"learning_rate": 1.0585224238640619e-05,
"loss": 0.2186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10465064644813538,
"step": 2930,
"valid_targets_mean": 4299.5,
"valid_targets_min": 2705
},
{
"epoch": 4.84393063583815,
"grad_norm": 0.4516934358593685,
"learning_rate": 1.0512688466464404e-05,
"loss": 0.2214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10467134416103363,
"step": 2935,
"valid_targets_mean": 4315.9,
"valid_targets_min": 3300
},
{
"epoch": 4.852188274153592,
"grad_norm": 0.47169166624893866,
"learning_rate": 1.0440313365468077e-05,
"loss": 0.2228,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10604281723499298,
"step": 2940,
"valid_targets_mean": 4332.1,
"valid_targets_min": 3609
},
{
"epoch": 4.860445912469034,
"grad_norm": 0.4349590370474785,
"learning_rate": 1.0368100161351116e-05,
"loss": 0.2247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10916487872600555,
"step": 2945,
"valid_targets_mean": 4540.6,
"valid_targets_min": 4231
},
{
"epoch": 4.868703550784476,
"grad_norm": 0.42139858058025015,
"learning_rate": 1.0296050077071238e-05,
"loss": 0.2164,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09808946400880814,
"step": 2950,
"valid_targets_mean": 4303.0,
"valid_targets_min": 2390
},
{
"epoch": 4.876961189099918,
"grad_norm": 0.4590357802679242,
"learning_rate": 1.0224164332823632e-05,
"loss": 0.2166,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09969616681337357,
"step": 2955,
"valid_targets_mean": 4432.9,
"valid_targets_min": 3616
},
{
"epoch": 4.885218827415359,
"grad_norm": 0.5203050571632374,
"learning_rate": 1.015244414602035e-05,
"loss": 0.2223,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13236811757087708,
"step": 2960,
"valid_targets_mean": 4192.0,
"valid_targets_min": 3405
},
{
"epoch": 4.893476465730801,
"grad_norm": 0.46288514473950576,
"learning_rate": 1.0080890731269674e-05,
"loss": 0.2225,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10135738551616669,
"step": 2965,
"valid_targets_mean": 4363.2,
"valid_targets_min": 3811
},
{
"epoch": 4.901734104046243,
"grad_norm": 0.4394218883769587,
"learning_rate": 1.0009505300355515e-05,
"loss": 0.2229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10720351338386536,
"step": 2970,
"valid_targets_mean": 4491.4,
"valid_targets_min": 3859
},
{
"epoch": 4.909991742361685,
"grad_norm": 0.49675390212976744,
"learning_rate": 9.938289062216916e-06,
"loss": 0.2223,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09742853045463562,
"step": 2975,
"valid_targets_mean": 4197.4,
"valid_targets_min": 1548
},
{
"epoch": 4.918249380677127,
"grad_norm": 0.4314334506009114,
"learning_rate": 9.867243222927593e-06,
"loss": 0.2237,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10945402085781097,
"step": 2980,
"valid_targets_mean": 4346.4,
"valid_targets_min": 3497
},
{
"epoch": 4.926507018992568,
"grad_norm": 0.4593520972197408,
"learning_rate": 9.796368985675497e-06,
"loss": 0.2269,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12122351676225662,
"step": 2985,
"valid_targets_mean": 4893.9,
"valid_targets_min": 3373
},
{
"epoch": 4.93476465730801,
"grad_norm": 0.42808980971446214,
"learning_rate": 9.725667550742408e-06,
"loss": 0.2172,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11039487272500992,
"step": 2990,
"valid_targets_mean": 4374.4,
"valid_targets_min": 2495
},
{
"epoch": 4.943022295623452,
"grad_norm": 0.4471756703739772,
"learning_rate": 9.655140115483663e-06,
"loss": 0.2267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10563766956329346,
"step": 2995,
"valid_targets_mean": 4002.4,
"valid_targets_min": 3093
},
{
"epoch": 4.951279933938894,
"grad_norm": 0.43034612712846293,
"learning_rate": 9.584787874307828e-06,
"loss": 0.2216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12290877848863602,
"step": 3000,
"valid_targets_mean": 5446.5,
"valid_targets_min": 3628
},
{
"epoch": 4.959537572254336,
"grad_norm": 0.46855986980259284,
"learning_rate": 9.514612018656493e-06,
"loss": 0.2259,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10534742474555969,
"step": 3005,
"valid_targets_mean": 4322.4,
"valid_targets_min": 3477
},
{
"epoch": 4.967795210569777,
"grad_norm": 0.4277441977319072,
"learning_rate": 9.444613736984107e-06,
"loss": 0.2195,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09368419647216797,
"step": 3010,
"valid_targets_mean": 3994.0,
"valid_targets_min": 1663
},
{
"epoch": 4.976052848885219,
"grad_norm": 0.43980767149640193,
"learning_rate": 9.374794214737828e-06,
"loss": 0.2232,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11430057883262634,
"step": 3015,
"valid_targets_mean": 4725.9,
"valid_targets_min": 3606
},
{
"epoch": 4.984310487200661,
"grad_norm": 0.44007158013904724,
"learning_rate": 9.305154634337466e-06,
"loss": 0.2218,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12239345908164978,
"step": 3020,
"valid_targets_mean": 4698.4,
"valid_targets_min": 3494
},
{
"epoch": 4.992568125516103,
"grad_norm": 0.4627687094036946,
"learning_rate": 9.235696175155429e-06,
"loss": 0.23,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11731258034706116,
"step": 3025,
"valid_targets_mean": 4486.1,
"valid_targets_min": 1651
},
{
"epoch": 5.0,
"grad_norm": 0.6140878222189191,
"learning_rate": 9.166420013496778e-06,
"loss": 0.2217,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21208547055721283,
"step": 3030,
"valid_targets_mean": 4393.9,
"valid_targets_min": 3341
},
{
"epoch": 5.008257638315442,
"grad_norm": 0.433940518348012,
"learning_rate": 9.097327322579309e-06,
"loss": 0.2083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11628393828868866,
"step": 3035,
"valid_targets_mean": 4502.4,
"valid_targets_min": 3846
},
{
"epoch": 5.016515276630884,
"grad_norm": 0.525028418083287,
"learning_rate": 9.028419272513649e-06,
"loss": 0.2081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11277692019939423,
"step": 3040,
"valid_targets_mean": 4509.2,
"valid_targets_min": 3837
},
{
"epoch": 5.024772914946325,
"grad_norm": 0.4829913906649403,
"learning_rate": 8.959697030283483e-06,
"loss": 0.2159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10610626637935638,
"step": 3045,
"valid_targets_mean": 4020.4,
"valid_targets_min": 3204
},
{
"epoch": 5.033030553261767,
"grad_norm": 0.47483441071365123,
"learning_rate": 8.891161759725772e-06,
"loss": 0.2155,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10142424702644348,
"step": 3050,
"valid_targets_mean": 3921.4,
"valid_targets_min": 1727
},
{
"epoch": 5.041288191577209,
"grad_norm": 0.437963316032041,
"learning_rate": 8.822814621511026e-06,
"loss": 0.2076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10212666541337967,
"step": 3055,
"valid_targets_mean": 4785.2,
"valid_targets_min": 3331
},
{
"epoch": 5.049545829892651,
"grad_norm": 0.4658363442600916,
"learning_rate": 8.754656773123662e-06,
"loss": 0.216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10734973102807999,
"step": 3060,
"valid_targets_mean": 4257.8,
"valid_targets_min": 2593
},
{
"epoch": 5.057803468208093,
"grad_norm": 0.44278694067092916,
"learning_rate": 8.686689368842419e-06,
"loss": 0.2112,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10486813634634018,
"step": 3065,
"valid_targets_mean": 4485.4,
"valid_targets_min": 3830
},
{
"epoch": 5.066061106523534,
"grad_norm": 0.45534950024089654,
"learning_rate": 8.61891355972079e-06,
"loss": 0.2099,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10280141234397888,
"step": 3070,
"valid_targets_mean": 4188.5,
"valid_targets_min": 2165
},
{
"epoch": 5.074318744838976,
"grad_norm": 0.5083233088137198,
"learning_rate": 8.551330493567517e-06,
"loss": 0.219,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1061544269323349,
"step": 3075,
"valid_targets_mean": 3882.6,
"valid_targets_min": 2408
},
{
"epoch": 5.082576383154418,
"grad_norm": 0.436187194383888,
"learning_rate": 8.483941314927193e-06,
"loss": 0.2244,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13365265727043152,
"step": 3080,
"valid_targets_mean": 4938.1,
"valid_targets_min": 2612
},
{
"epoch": 5.09083402146986,
"grad_norm": 0.4748117730283424,
"learning_rate": 8.41674716506083e-06,
"loss": 0.2088,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09705816209316254,
"step": 3085,
"valid_targets_mean": 4370.9,
"valid_targets_min": 3049
},
{
"epoch": 5.0990916597853015,
"grad_norm": 0.46231383579733526,
"learning_rate": 8.349749181926556e-06,
"loss": 0.2134,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1041770949959755,
"step": 3090,
"valid_targets_mean": 4285.1,
"valid_targets_min": 3631
},
{
"epoch": 5.107349298100743,
"grad_norm": 0.49124546543437186,
"learning_rate": 8.28294850016036e-06,
"loss": 0.2183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1132897287607193,
"step": 3095,
"valid_targets_mean": 4232.6,
"valid_targets_min": 3445
},
{
"epoch": 5.115606936416185,
"grad_norm": 0.47928313808611667,
"learning_rate": 8.216346251056846e-06,
"loss": 0.2159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10141552984714508,
"step": 3100,
"valid_targets_mean": 4335.1,
"valid_targets_min": 3719
},
{
"epoch": 5.123864574731627,
"grad_norm": 0.443398000644317,
"learning_rate": 8.14994356255008e-06,
"loss": 0.2081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10616658627986908,
"step": 3105,
"valid_targets_mean": 4583.2,
"valid_targets_min": 3067
},
{
"epoch": 5.132122213047069,
"grad_norm": 0.5233217149350713,
"learning_rate": 8.083741559194515e-06,
"loss": 0.2203,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11892595887184143,
"step": 3110,
"valid_targets_mean": 4245.5,
"valid_targets_min": 2966
},
{
"epoch": 5.1403798513625105,
"grad_norm": 0.4630064327317817,
"learning_rate": 8.017741362145899e-06,
"loss": 0.2106,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11951257288455963,
"step": 3115,
"valid_targets_mean": 5296.9,
"valid_targets_min": 3555
},
{
"epoch": 5.148637489677952,
"grad_norm": 0.45388687326124455,
"learning_rate": 7.95194408914234e-06,
"loss": 0.2173,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12584248185157776,
"step": 3120,
"valid_targets_mean": 4379.4,
"valid_targets_min": 3311
},
{
"epoch": 5.156895127993394,
"grad_norm": 0.4494932509856417,
"learning_rate": 7.886350854485329e-06,
"loss": 0.2249,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1212790235877037,
"step": 3125,
"valid_targets_mean": 4483.2,
"valid_targets_min": 3939
},
{
"epoch": 5.165152766308836,
"grad_norm": 0.43368049702066025,
"learning_rate": 7.820962769020906e-06,
"loss": 0.2179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10778104513883591,
"step": 3130,
"valid_targets_mean": 4651.6,
"valid_targets_min": 3680
},
{
"epoch": 5.173410404624278,
"grad_norm": 0.4559544980546319,
"learning_rate": 7.755780940120836e-06,
"loss": 0.2213,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1204538494348526,
"step": 3135,
"valid_targets_mean": 4499.2,
"valid_targets_min": 692
},
{
"epoch": 5.1816680429397195,
"grad_norm": 0.4871550663909911,
"learning_rate": 7.690806471663836e-06,
"loss": 0.2187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10757050663232803,
"step": 3140,
"valid_targets_mean": 4197.0,
"valid_targets_min": 1726
},
{
"epoch": 5.189925681255161,
"grad_norm": 0.4650848378004337,
"learning_rate": 7.626040464016897e-06,
"loss": 0.2118,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09520135819911957,
"step": 3145,
"valid_targets_mean": 3959.6,
"valid_targets_min": 2463
},
{
"epoch": 5.198183319570603,
"grad_norm": 0.45137678195265213,
"learning_rate": 7.561484014016665e-06,
"loss": 0.2037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10195222496986389,
"step": 3150,
"valid_targets_mean": 3955.5,
"valid_targets_min": 1527
},
{
"epoch": 5.206440957886045,
"grad_norm": 0.44676395239635347,
"learning_rate": 7.497138214950839e-06,
"loss": 0.2148,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09813147783279419,
"step": 3155,
"valid_targets_mean": 4102.6,
"valid_targets_min": 2955
},
{
"epoch": 5.214698596201487,
"grad_norm": 0.5670034217682519,
"learning_rate": 7.433004156539656e-06,
"loss": 0.2146,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10359115153551102,
"step": 3160,
"valid_targets_mean": 3889.1,
"valid_targets_min": 1997
},
{
"epoch": 5.2229562345169285,
"grad_norm": 0.4731384780709025,
"learning_rate": 7.369082924917465e-06,
"loss": 0.2118,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11328642815351486,
"step": 3165,
"valid_targets_mean": 4499.9,
"valid_targets_min": 3155
},
{
"epoch": 5.23121387283237,
"grad_norm": 0.47889673616606887,
"learning_rate": 7.305375602614299e-06,
"loss": 0.2084,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11517007648944855,
"step": 3170,
"valid_targets_mean": 4523.9,
"valid_targets_min": 3209
},
{
"epoch": 5.239471511147812,
"grad_norm": 0.48844080913191323,
"learning_rate": 7.2418832685375525e-06,
"loss": 0.2182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10912356525659561,
"step": 3175,
"valid_targets_mean": 4496.1,
"valid_targets_min": 1381
},
{
"epoch": 5.247729149463254,
"grad_norm": 0.45043108657164277,
"learning_rate": 7.178606997953728e-06,
"loss": 0.2143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11774206161499023,
"step": 3180,
"valid_targets_mean": 4810.1,
"valid_targets_min": 4240
},
{
"epoch": 5.255986787778696,
"grad_norm": 0.49582080199377604,
"learning_rate": 7.115547862470211e-06,
"loss": 0.2115,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10136029869318008,
"step": 3185,
"valid_targets_mean": 4158.4,
"valid_targets_min": 3427
},
{
"epoch": 5.2642444260941375,
"grad_norm": 0.44077382174619784,
"learning_rate": 7.052706930017106e-06,
"loss": 0.2138,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10543619096279144,
"step": 3190,
"valid_targets_mean": 4536.4,
"valid_targets_min": 3485
},
{
"epoch": 5.272502064409579,
"grad_norm": 0.49352676335176354,
"learning_rate": 6.9900852648291895e-06,
"loss": 0.2065,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13009649515151978,
"step": 3195,
"valid_targets_mean": 4724.1,
"valid_targets_min": 3064
},
{
"epoch": 5.280759702725021,
"grad_norm": 0.4680415598626916,
"learning_rate": 6.927683927427842e-06,
"loss": 0.2062,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10420256853103638,
"step": 3200,
"valid_targets_mean": 3684.4,
"valid_targets_min": 1556
},
{
"epoch": 5.289017341040463,
"grad_norm": 0.44584067778478936,
"learning_rate": 6.8655039746031315e-06,
"loss": 0.2174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09919565916061401,
"step": 3205,
"valid_targets_mean": 4111.1,
"valid_targets_min": 1334
},
{
"epoch": 5.297274979355905,
"grad_norm": 0.46420230274212004,
"learning_rate": 6.803546459395873e-06,
"loss": 0.221,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11786352097988129,
"step": 3210,
"valid_targets_mean": 5017.5,
"valid_targets_min": 3570
},
{
"epoch": 5.305532617671346,
"grad_norm": 0.43419800369315525,
"learning_rate": 6.741812431079839e-06,
"loss": 0.2084,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09311288595199585,
"step": 3215,
"valid_targets_mean": 3898.5,
"valid_targets_min": 2483
},
{
"epoch": 5.313790255986788,
"grad_norm": 0.4421323033624433,
"learning_rate": 6.680302935143963e-06,
"loss": 0.2075,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09455594420433044,
"step": 3220,
"valid_targets_mean": 4532.8,
"valid_targets_min": 3732
},
{
"epoch": 5.32204789430223,
"grad_norm": 0.48853543735036054,
"learning_rate": 6.619019013274626e-06,
"loss": 0.2154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08839322626590729,
"step": 3225,
"valid_targets_mean": 3923.1,
"valid_targets_min": 1570
},
{
"epoch": 5.330305532617672,
"grad_norm": 0.46256545631056417,
"learning_rate": 6.557961703338027e-06,
"loss": 0.2168,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10854192823171616,
"step": 3230,
"valid_targets_mean": 4357.9,
"valid_targets_min": 2520
},
{
"epoch": 5.338563170933113,
"grad_norm": 0.47524064221826606,
"learning_rate": 6.49713203936263e-06,
"loss": 0.2195,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1096210926771164,
"step": 3235,
"valid_targets_mean": 4091.8,
"valid_targets_min": 1980
},
{
"epoch": 5.3468208092485545,
"grad_norm": 0.5160067136616678,
"learning_rate": 6.4365310515216e-06,
"loss": 0.2183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10057702660560608,
"step": 3240,
"valid_targets_mean": 4586.5,
"valid_targets_min": 4025
},
{
"epoch": 5.355078447563996,
"grad_norm": 0.5154543742439444,
"learning_rate": 6.376159766115408e-06,
"loss": 0.2192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10424821078777313,
"step": 3245,
"valid_targets_mean": 4353.2,
"valid_targets_min": 3828
},
{
"epoch": 5.363336085879438,
"grad_norm": 0.48030628709647966,
"learning_rate": 6.316019205554425e-06,
"loss": 0.2138,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11591741442680359,
"step": 3250,
"valid_targets_mean": 4475.1,
"valid_targets_min": 2855
},
{
"epoch": 5.37159372419488,
"grad_norm": 0.49766744913108296,
"learning_rate": 6.256110388341597e-06,
"loss": 0.2109,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.07996664941310883,
"step": 3255,
"valid_targets_mean": 3990.6,
"valid_targets_min": 2676
},
{
"epoch": 5.379851362510322,
"grad_norm": 0.482387714051572,
"learning_rate": 6.196434329055214e-06,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11598837375640869,
"step": 3260,
"valid_targets_mean": 4076.8,
"valid_targets_min": 2320
},
{
"epoch": 5.3881090008257635,
"grad_norm": 0.5786740321339716,
"learning_rate": 6.136992038331735e-06,
"loss": 0.2148,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10620634257793427,
"step": 3265,
"valid_targets_mean": 4564.9,
"valid_targets_min": 2767
},
{
"epoch": 5.396366639141205,
"grad_norm": 0.4519832103521578,
"learning_rate": 6.077784522848653e-06,
"loss": 0.2143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10976023972034454,
"step": 3270,
"valid_targets_mean": 4915.6,
"valid_targets_min": 3613
},
{
"epoch": 5.404624277456647,
"grad_norm": 0.4564810942220355,
"learning_rate": 6.018812785307447e-06,
"loss": 0.2141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11017479002475739,
"step": 3275,
"valid_targets_mean": 4556.0,
"valid_targets_min": 3575
},
{
"epoch": 5.412881915772089,
"grad_norm": 0.4842138101713595,
"learning_rate": 5.960077824416623e-06,
"loss": 0.2043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1073116734623909,
"step": 3280,
"valid_targets_mean": 4100.9,
"valid_targets_min": 880
},
{
"epoch": 5.421139554087531,
"grad_norm": 0.43260278672112945,
"learning_rate": 5.901580634874775e-06,
"loss": 0.2143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09420692920684814,
"step": 3285,
"valid_targets_mean": 4059.6,
"valid_targets_min": 3259
},
{
"epoch": 5.4293971924029725,
"grad_norm": 0.46488713322313135,
"learning_rate": 5.843322207353746e-06,
"loss": 0.2203,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12504369020462036,
"step": 3290,
"valid_targets_mean": 4143.1,
"valid_targets_min": 3094
},
{
"epoch": 5.437654830718414,
"grad_norm": 0.4542721719842577,
"learning_rate": 5.785303528481862e-06,
"loss": 0.2155,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11584823578596115,
"step": 3295,
"valid_targets_mean": 4252.9,
"valid_targets_min": 1800
},
{
"epoch": 5.445912469033856,
"grad_norm": 0.45043440864208956,
"learning_rate": 5.72752558082722e-06,
"loss": 0.2115,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10430656373500824,
"step": 3300,
"valid_targets_mean": 4326.1,
"valid_targets_min": 3422
},
{
"epoch": 5.454170107349298,
"grad_norm": 0.4422717929946044,
"learning_rate": 5.66998934288105e-06,
"loss": 0.2132,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09787015616893768,
"step": 3305,
"valid_targets_mean": 4297.0,
"valid_targets_min": 3423
},
{
"epoch": 5.46242774566474,
"grad_norm": 0.4690467521956276,
"learning_rate": 5.6126957890411e-06,
"loss": 0.2197,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10723280161619186,
"step": 3310,
"valid_targets_mean": 4467.9,
"valid_targets_min": 3401
},
{
"epoch": 5.4706853839801814,
"grad_norm": 0.4861248649611912,
"learning_rate": 5.5556458895952115e-06,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09763900190591812,
"step": 3315,
"valid_targets_mean": 4087.8,
"valid_targets_min": 2200
},
{
"epoch": 5.478943022295623,
"grad_norm": 0.4581842931862147,
"learning_rate": 5.498840610704837e-06,
"loss": 0.205,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09928775578737259,
"step": 3320,
"valid_targets_mean": 4049.9,
"valid_targets_min": 2905
},
{
"epoch": 5.487200660611065,
"grad_norm": 0.44242388066112215,
"learning_rate": 5.442280914388673e-06,
"loss": 0.212,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09497937560081482,
"step": 3325,
"valid_targets_mean": 4040.6,
"valid_targets_min": 3395
},
{
"epoch": 5.495458298926507,
"grad_norm": 0.49050857708703044,
"learning_rate": 5.385967758506407e-06,
"loss": 0.2099,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10977080464363098,
"step": 3330,
"valid_targets_mean": 4397.2,
"valid_targets_min": 3855
},
{
"epoch": 5.503715937241949,
"grad_norm": 0.44679720676471235,
"learning_rate": 5.329902096742452e-06,
"loss": 0.2157,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10305120050907135,
"step": 3335,
"valid_targets_mean": 4637.1,
"valid_targets_min": 3364
},
{
"epoch": 5.51197357555739,
"grad_norm": 0.4421017319679382,
"learning_rate": 5.274084878589818e-06,
"loss": 0.2204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11496944725513458,
"step": 3340,
"valid_targets_mean": 4419.6,
"valid_targets_min": 3116
},
{
"epoch": 5.520231213872832,
"grad_norm": 0.4732363787775278,
"learning_rate": 5.21851704933404e-06,
"loss": 0.2127,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11920087784528732,
"step": 3345,
"valid_targets_mean": 4477.0,
"valid_targets_min": 1971
},
{
"epoch": 5.528488852188274,
"grad_norm": 0.4678112580814567,
"learning_rate": 5.16319955003715e-06,
"loss": 0.2107,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10399312525987625,
"step": 3350,
"valid_targets_mean": 3845.6,
"valid_targets_min": 2869
},
{
"epoch": 5.536746490503716,
"grad_norm": 0.4512473780934103,
"learning_rate": 5.108133317521757e-06,
"loss": 0.2145,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09832729399204254,
"step": 3355,
"valid_targets_mean": 4442.4,
"valid_targets_min": 3452
},
{
"epoch": 5.545004128819158,
"grad_norm": 0.4303922235957302,
"learning_rate": 5.053319284355162e-06,
"loss": 0.2185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12026438117027283,
"step": 3360,
"valid_targets_mean": 4972.1,
"valid_targets_min": 4195
},
{
"epoch": 5.553261767134599,
"grad_norm": 0.46141285338180854,
"learning_rate": 4.99875837883357e-06,
"loss": 0.2126,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09322738647460938,
"step": 3365,
"valid_targets_mean": 4315.5,
"valid_targets_min": 3438
},
{
"epoch": 5.561519405450041,
"grad_norm": 0.45451403895275383,
"learning_rate": 4.944451524966401e-06,
"loss": 0.2057,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10366357862949371,
"step": 3370,
"valid_targets_mean": 4912.9,
"valid_targets_min": 3891
},
{
"epoch": 5.569777043765483,
"grad_norm": 0.46270724439983885,
"learning_rate": 4.890399642460582e-06,
"loss": 0.2142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09599032253026962,
"step": 3375,
"valid_targets_mean": 4006.9,
"valid_targets_min": 2163
},
{
"epoch": 5.578034682080925,
"grad_norm": 0.4464845459734285,
"learning_rate": 4.836603646705027e-06,
"loss": 0.224,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11775711923837662,
"step": 3380,
"valid_targets_mean": 4297.1,
"valid_targets_min": 3175
},
{
"epoch": 5.586292320396367,
"grad_norm": 0.49228359377846115,
"learning_rate": 4.783064448755113e-06,
"loss": 0.215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10951682925224304,
"step": 3385,
"valid_targets_mean": 4558.8,
"valid_targets_min": 3643
},
{
"epoch": 5.594549958711808,
"grad_norm": 0.5056204530947794,
"learning_rate": 4.729782955317233e-06,
"loss": 0.2133,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11983240395784378,
"step": 3390,
"valid_targets_mean": 4430.6,
"valid_targets_min": 3689
},
{
"epoch": 5.60280759702725,
"grad_norm": 0.45820874577446424,
"learning_rate": 4.676760068733461e-06,
"loss": 0.2088,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1002374142408371,
"step": 3395,
"valid_targets_mean": 4278.8,
"valid_targets_min": 3095
},
{
"epoch": 5.611065235342692,
"grad_norm": 0.48504221589740126,
"learning_rate": 4.623996686966279e-06,
"loss": 0.2182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10616739094257355,
"step": 3400,
"valid_targets_mean": 4287.4,
"valid_targets_min": 2661
},
{
"epoch": 5.619322873658134,
"grad_norm": 0.44979511185546367,
"learning_rate": 4.571493703583358e-06,
"loss": 0.2162,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09403562545776367,
"step": 3405,
"valid_targets_mean": 4353.4,
"valid_targets_min": 1155
},
{
"epoch": 5.627580511973576,
"grad_norm": 0.4761522658645116,
"learning_rate": 4.519252007742405e-06,
"loss": 0.2174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09915102273225784,
"step": 3410,
"valid_targets_mean": 4135.2,
"valid_targets_min": 2669
},
{
"epoch": 5.635838150289017,
"grad_norm": 0.5771727867210109,
"learning_rate": 4.467272484176146e-06,
"loss": 0.2083,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09634391963481903,
"step": 3415,
"valid_targets_mean": 4479.5,
"valid_targets_min": 2948
},
{
"epoch": 5.644095788604459,
"grad_norm": 0.46998022522182964,
"learning_rate": 4.415556013177311e-06,
"loss": 0.2223,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12644150853157043,
"step": 3420,
"valid_targets_mean": 4870.4,
"valid_targets_min": 4063
},
{
"epoch": 5.652353426919901,
"grad_norm": 0.4737489351510268,
"learning_rate": 4.364103470583729e-06,
"loss": 0.2234,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11984790116548538,
"step": 3425,
"valid_targets_mean": 4521.5,
"valid_targets_min": 4041
},
{
"epoch": 5.660611065235343,
"grad_norm": 0.46532637794833676,
"learning_rate": 4.312915727763516e-06,
"loss": 0.2158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13450059294700623,
"step": 3430,
"valid_targets_mean": 5368.4,
"valid_targets_min": 4165
},
{
"epoch": 5.6688687035507845,
"grad_norm": 0.4488667605081453,
"learning_rate": 4.2619936516003e-06,
"loss": 0.2137,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10450101643800735,
"step": 3435,
"valid_targets_mean": 4721.9,
"valid_targets_min": 3992
},
{
"epoch": 5.677126341866226,
"grad_norm": 0.4460088837280221,
"learning_rate": 4.211338104478548e-06,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0902666226029396,
"step": 3440,
"valid_targets_mean": 4095.6,
"valid_targets_min": 3566
},
{
"epoch": 5.685383980181668,
"grad_norm": 0.4706653432593833,
"learning_rate": 4.16094994426895e-06,
"loss": 0.2199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10886543989181519,
"step": 3445,
"valid_targets_mean": 4380.5,
"valid_targets_min": 3438
},
{
"epoch": 5.69364161849711,
"grad_norm": 0.4518821505277068,
"learning_rate": 4.1108300243138945e-06,
"loss": 0.204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09984904527664185,
"step": 3450,
"valid_targets_mean": 4621.9,
"valid_targets_min": 2275
},
{
"epoch": 5.701899256812552,
"grad_norm": 0.4530699314996465,
"learning_rate": 4.060979193413041e-06,
"loss": 0.2142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10747627168893814,
"step": 3455,
"valid_targets_mean": 4327.6,
"valid_targets_min": 3454
},
{
"epoch": 5.7101568951279935,
"grad_norm": 0.4585354416932689,
"learning_rate": 4.011398295808899e-06,
"loss": 0.2163,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12132354080677032,
"step": 3460,
"valid_targets_mean": 4376.5,
"valid_targets_min": 3426
},
{
"epoch": 5.718414533443435,
"grad_norm": 0.42608117634517756,
"learning_rate": 3.962088171172574e-06,
"loss": 0.2164,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09466895461082458,
"step": 3465,
"valid_targets_mean": 4103.2,
"valid_targets_min": 1570
},
{
"epoch": 5.726672171758877,
"grad_norm": 0.46326994818543826,
"learning_rate": 3.913049654589531e-06,
"loss": 0.2168,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10534656047821045,
"step": 3470,
"valid_targets_mean": 4593.9,
"valid_targets_min": 3209
},
{
"epoch": 5.734929810074319,
"grad_norm": 0.4242602533657284,
"learning_rate": 3.864283576545442e-06,
"loss": 0.2077,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1007297933101654,
"step": 3475,
"valid_targets_mean": 4213.5,
"valid_targets_min": 3002
},
{
"epoch": 5.743187448389761,
"grad_norm": 0.4558281708655785,
"learning_rate": 3.815790762912124e-06,
"loss": 0.2152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09895782917737961,
"step": 3480,
"valid_targets_mean": 4342.6,
"valid_targets_min": 3561
},
{
"epoch": 5.7514450867052025,
"grad_norm": 0.44172747181203537,
"learning_rate": 3.767572034933573e-06,
"loss": 0.2122,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12205647677183151,
"step": 3485,
"valid_targets_mean": 4847.6,
"valid_targets_min": 3416
},
{
"epoch": 5.759702725020644,
"grad_norm": 0.4842224672318322,
"learning_rate": 3.719628209212043e-06,
"loss": 0.2197,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11489449441432953,
"step": 3490,
"valid_targets_mean": 4433.9,
"valid_targets_min": 3423
},
{
"epoch": 5.767960363336086,
"grad_norm": 0.4571172310371479,
"learning_rate": 3.671960097694196e-06,
"loss": 0.2121,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09982895851135254,
"step": 3495,
"valid_targets_mean": 4416.8,
"valid_targets_min": 4143
},
{
"epoch": 5.776218001651528,
"grad_norm": 0.4449900886737973,
"learning_rate": 3.6245685076573956e-06,
"loss": 0.2095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10301372408866882,
"step": 3500,
"valid_targets_mean": 4787.8,
"valid_targets_min": 3730
},
{
"epoch": 5.78447563996697,
"grad_norm": 0.4746623854022696,
"learning_rate": 3.577454241695988e-06,
"loss": 0.2222,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11077150702476501,
"step": 3505,
"valid_targets_mean": 4349.9,
"valid_targets_min": 3349
},
{
"epoch": 5.7927332782824115,
"grad_norm": 0.4527138131054143,
"learning_rate": 3.530618097707743e-06,
"loss": 0.2067,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08911068737506866,
"step": 3510,
"valid_targets_mean": 4057.4,
"valid_targets_min": 1902
},
{
"epoch": 5.800990916597853,
"grad_norm": 0.48036469508912494,
"learning_rate": 3.484060868880328e-06,
"loss": 0.2177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09339256584644318,
"step": 3515,
"valid_targets_mean": 3877.1,
"valid_targets_min": 1768
},
{
"epoch": 5.809248554913295,
"grad_norm": 0.4807227448882868,
"learning_rate": 3.4377833436778874e-06,
"loss": 0.2182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10168130695819855,
"step": 3520,
"valid_targets_mean": 3859.2,
"valid_targets_min": 3012
},
{
"epoch": 5.817506193228737,
"grad_norm": 0.4660642406487113,
"learning_rate": 3.39178630582766e-06,
"loss": 0.215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11255677789449692,
"step": 3525,
"valid_targets_mean": 4868.5,
"valid_targets_min": 3616
},
{
"epoch": 5.825763831544179,
"grad_norm": 0.4621509519217746,
"learning_rate": 3.3460705343067467e-06,
"loss": 0.209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10139954090118408,
"step": 3530,
"valid_targets_mean": 4411.0,
"valid_targets_min": 3855
},
{
"epoch": 5.8340214698596204,
"grad_norm": 0.4831499052879207,
"learning_rate": 3.3006368033288783e-06,
"loss": 0.2059,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11212857067584991,
"step": 3535,
"valid_targets_mean": 4318.1,
"valid_targets_min": 1356
},
{
"epoch": 5.842279108175062,
"grad_norm": 0.45403032654783493,
"learning_rate": 3.2554858823313417e-06,
"loss": 0.2156,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0986095666885376,
"step": 3540,
"valid_targets_mean": 4567.9,
"valid_targets_min": 3426
},
{
"epoch": 5.850536746490504,
"grad_norm": 0.46917314805250077,
"learning_rate": 3.210618535961916e-06,
"loss": 0.2054,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10373765230178833,
"step": 3545,
"valid_targets_mean": 4259.8,
"valid_targets_min": 3705
},
{
"epoch": 5.858794384805946,
"grad_norm": 0.4637092061295093,
"learning_rate": 3.1660355240659423e-06,
"loss": 0.2136,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09390874952077866,
"step": 3550,
"valid_targets_mean": 4078.2,
"valid_targets_min": 3277
},
{
"epoch": 5.867052023121388,
"grad_norm": 0.49080897786176625,
"learning_rate": 3.1217376016734624e-06,
"loss": 0.2102,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09790024906396866,
"step": 3555,
"valid_targets_mean": 4338.8,
"valid_targets_min": 3275
},
{
"epoch": 5.875309661436829,
"grad_norm": 0.48744460013728824,
"learning_rate": 3.077725518986401e-06,
"loss": 0.2145,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09616672992706299,
"step": 3560,
"valid_targets_mean": 4311.4,
"valid_targets_min": 3250
},
{
"epoch": 5.883567299752271,
"grad_norm": 0.45793796387301444,
"learning_rate": 3.0340000213658882e-06,
"loss": 0.2097,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12131049484014511,
"step": 3565,
"valid_targets_mean": 4731.5,
"valid_targets_min": 3934
},
{
"epoch": 5.891824938067713,
"grad_norm": 0.47482238852255926,
"learning_rate": 2.9905618493196353e-06,
"loss": 0.2199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09843767434358597,
"step": 3570,
"valid_targets_mean": 4274.6,
"valid_targets_min": 3127
},
{
"epoch": 5.900082576383154,
"grad_norm": 0.47338493683628696,
"learning_rate": 2.947411738489374e-06,
"loss": 0.21,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10318023711442947,
"step": 3575,
"valid_targets_mean": 4090.4,
"valid_targets_min": 3201
},
{
"epoch": 5.908340214698596,
"grad_norm": 0.47605970178437684,
"learning_rate": 2.904550419638421e-06,
"loss": 0.2171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10486222803592682,
"step": 3580,
"valid_targets_mean": 3951.4,
"valid_targets_min": 2254
},
{
"epoch": 5.9165978530140375,
"grad_norm": 0.44974078106928367,
"learning_rate": 2.8619786186392986e-06,
"loss": 0.2154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11146186292171478,
"step": 3585,
"valid_targets_mean": 4450.1,
"valid_targets_min": 2738
},
{
"epoch": 5.924855491329479,
"grad_norm": 0.4902172072045995,
"learning_rate": 2.8196970564614167e-06,
"loss": 0.2171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10667402297258377,
"step": 3590,
"valid_targets_mean": 4575.5,
"valid_targets_min": 3969
},
{
"epoch": 5.933113129644921,
"grad_norm": 0.45896868597930146,
"learning_rate": 2.7777064491588946e-06,
"loss": 0.2174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11253248900175095,
"step": 3595,
"valid_targets_mean": 4760.4,
"valid_targets_min": 4078
},
{
"epoch": 5.941370767960363,
"grad_norm": 0.46885916917842,
"learning_rate": 2.736007507858418e-06,
"loss": 0.2098,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11156116425991058,
"step": 3600,
"valid_targets_mean": 4450.4,
"valid_targets_min": 2239
},
{
"epoch": 5.949628406275805,
"grad_norm": 0.46359239174233846,
"learning_rate": 2.6946009387472074e-06,
"loss": 0.2136,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09594880044460297,
"step": 3605,
"valid_targets_mean": 4537.2,
"valid_targets_min": 3384
},
{
"epoch": 5.9578860445912465,
"grad_norm": 0.4560620201383562,
"learning_rate": 2.653487443061036e-06,
"loss": 0.21,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09522822499275208,
"step": 3610,
"valid_targets_mean": 4389.1,
"valid_targets_min": 2639
},
{
"epoch": 5.966143682906688,
"grad_norm": 0.44912905718365526,
"learning_rate": 2.6126677170723837e-06,
"loss": 0.215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11366581171751022,
"step": 3615,
"valid_targets_mean": 4548.1,
"valid_targets_min": 3374
},
{
"epoch": 5.97440132122213,
"grad_norm": 0.47910206923757565,
"learning_rate": 2.5721424520786163e-06,
"loss": 0.2172,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11073582619428635,
"step": 3620,
"valid_targets_mean": 4261.0,
"valid_targets_min": 3507
},
{
"epoch": 5.982658959537572,
"grad_norm": 0.46356315226875466,
"learning_rate": 2.531912334390307e-06,
"loss": 0.2126,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10287211835384369,
"step": 3625,
"valid_targets_mean": 4633.6,
"valid_targets_min": 4089
},
{
"epoch": 5.990916597853014,
"grad_norm": 0.4529757277041409,
"learning_rate": 2.4919780453195808e-06,
"loss": 0.215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12963898479938507,
"step": 3630,
"valid_targets_mean": 5128.9,
"valid_targets_min": 1651
},
{
"epoch": 5.9991742361684555,
"grad_norm": 0.453718098938784,
"learning_rate": 2.452340261168611e-06,
"loss": 0.2171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13191549479961395,
"step": 3635,
"valid_targets_mean": 5213.5,
"valid_targets_min": 4418
},
{
"epoch": 6.006606110652354,
"grad_norm": 0.443329098814205,
"learning_rate": 2.4129996532181423e-06,
"loss": 0.2003,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0950363278388977,
"step": 3640,
"valid_targets_mean": 4223.4,
"valid_targets_min": 1664
},
{
"epoch": 6.014863748967795,
"grad_norm": 0.46743012386952604,
"learning_rate": 2.3739568877161266e-06,
"loss": 0.2112,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11123231053352356,
"step": 3645,
"valid_targets_mean": 4458.4,
"valid_targets_min": 3063
},
{
"epoch": 6.023121387283237,
"grad_norm": 0.45937987831960836,
"learning_rate": 2.335212625866443e-06,
"loss": 0.2192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1077718734741211,
"step": 3650,
"valid_targets_mean": 4372.9,
"valid_targets_min": 2055
},
{
"epoch": 6.031379025598679,
"grad_norm": 0.42682010913883395,
"learning_rate": 2.296767523817702e-06,
"loss": 0.2067,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10394793748855591,
"step": 3655,
"valid_targets_mean": 4401.9,
"valid_targets_min": 2869
},
{
"epoch": 6.039636663914121,
"grad_norm": 0.44228450735772534,
"learning_rate": 2.2586222326521277e-06,
"loss": 0.2099,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11123920232057571,
"step": 3660,
"valid_targets_mean": 4688.6,
"valid_targets_min": 3910
},
{
"epoch": 6.047894302229563,
"grad_norm": 0.4401421693217935,
"learning_rate": 2.220777398374534e-06,
"loss": 0.2204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11513420194387436,
"step": 3665,
"valid_targets_mean": 4955.8,
"valid_targets_min": 2106
},
{
"epoch": 6.056151940545004,
"grad_norm": 0.4565522390589867,
"learning_rate": 2.183233661901396e-06,
"loss": 0.2038,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09502358734607697,
"step": 3670,
"valid_targets_mean": 4045.6,
"valid_targets_min": 2255
},
{
"epoch": 6.064409578860446,
"grad_norm": 0.4501664664042855,
"learning_rate": 2.1459916590499663e-06,
"loss": 0.2111,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10309629142284393,
"step": 3675,
"valid_targets_mean": 4159.5,
"valid_targets_min": 1860
},
{
"epoch": 6.072667217175888,
"grad_norm": 0.45753392862206754,
"learning_rate": 2.1090520205275333e-06,
"loss": 0.209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1050737053155899,
"step": 3680,
"valid_targets_mean": 4147.8,
"valid_targets_min": 3049
},
{
"epoch": 6.08092485549133,
"grad_norm": 0.462563245763296,
"learning_rate": 2.072415371920735e-06,
"loss": 0.2053,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11134891211986542,
"step": 3685,
"valid_targets_mean": 4303.1,
"valid_targets_min": 3499
},
{
"epoch": 6.089182493806772,
"grad_norm": 0.5055408418668056,
"learning_rate": 2.0360823336849634e-06,
"loss": 0.2048,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10344287008047104,
"step": 3690,
"valid_targets_mean": 4089.4,
"valid_targets_min": 2160
},
{
"epoch": 6.097440132122213,
"grad_norm": 0.44324625798896394,
"learning_rate": 2.0000535211338447e-06,
"loss": 0.2108,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0899750292301178,
"step": 3695,
"valid_targets_mean": 4273.6,
"valid_targets_min": 3675
},
{
"epoch": 6.105697770437655,
"grad_norm": 0.4576770159973642,
"learning_rate": 1.964329544428842e-06,
"loss": 0.2084,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10486738383769989,
"step": 3700,
"valid_targets_mean": 4416.0,
"valid_targets_min": 2558
},
{
"epoch": 6.113955408753097,
"grad_norm": 0.470682891657836,
"learning_rate": 1.928911008568899e-06,
"loss": 0.2103,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09307803213596344,
"step": 3705,
"valid_targets_mean": 3923.8,
"valid_targets_min": 2309
},
{
"epoch": 6.122213047068539,
"grad_norm": 0.4707505718745452,
"learning_rate": 1.8937985133802028e-06,
"loss": 0.2111,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09149274230003357,
"step": 3710,
"valid_targets_mean": 3968.2,
"valid_targets_min": 2055
},
{
"epoch": 6.1304706853839805,
"grad_norm": 0.4759153066513328,
"learning_rate": 1.8589926535060353e-06,
"loss": 0.213,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10827469825744629,
"step": 3715,
"valid_targets_mean": 4217.4,
"valid_targets_min": 2270
},
{
"epoch": 6.138728323699422,
"grad_norm": 0.4383686130452567,
"learning_rate": 1.824494018396692e-06,
"loss": 0.2158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10896565020084381,
"step": 3720,
"valid_targets_mean": 4798.6,
"valid_targets_min": 4125
},
{
"epoch": 6.146985962014864,
"grad_norm": 0.4582470398200648,
"learning_rate": 1.7903031922995007e-06,
"loss": 0.2128,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1139170378446579,
"step": 3725,
"valid_targets_mean": 4305.6,
"valid_targets_min": 3369
},
{
"epoch": 6.155243600330306,
"grad_norm": 0.4752654398789045,
"learning_rate": 1.7564207542489287e-06,
"loss": 0.2094,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10397236049175262,
"step": 3730,
"valid_targets_mean": 4595.1,
"valid_targets_min": 2858
},
{
"epoch": 6.163501238645748,
"grad_norm": 0.4571715012938738,
"learning_rate": 1.7228472780567673e-06,
"loss": 0.2114,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1044120341539383,
"step": 3735,
"valid_targets_mean": 4330.8,
"valid_targets_min": 3274
},
{
"epoch": 6.1717588769611895,
"grad_norm": 0.44042515317003217,
"learning_rate": 1.6895833323024403e-06,
"loss": 0.2103,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10458225011825562,
"step": 3740,
"valid_targets_mean": 4358.6,
"valid_targets_min": 3300
},
{
"epoch": 6.18001651527663,
"grad_norm": 0.48549820956519946,
"learning_rate": 1.6566294803233374e-06,
"loss": 0.2157,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12184527516365051,
"step": 3745,
"valid_targets_mean": 4790.2,
"valid_targets_min": 4206
},
{
"epoch": 6.188274153592072,
"grad_norm": 0.4837047338848428,
"learning_rate": 1.623986280205312e-06,
"loss": 0.2049,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09763555973768234,
"step": 3750,
"valid_targets_mean": 4086.9,
"valid_targets_min": 1581
},
{
"epoch": 6.196531791907514,
"grad_norm": 0.4570939152863252,
"learning_rate": 1.5916542847732076e-06,
"loss": 0.2091,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10981544852256775,
"step": 3755,
"valid_targets_mean": 4709.1,
"valid_targets_min": 3865
},
{
"epoch": 6.204789430222956,
"grad_norm": 0.45964680181987666,
"learning_rate": 1.5596340415814837e-06,
"loss": 0.2111,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11334769427776337,
"step": 3760,
"valid_targets_mean": 4803.5,
"valid_targets_min": 3942
},
{
"epoch": 6.213047068538398,
"grad_norm": 0.4896542198934198,
"learning_rate": 1.5279260929049766e-06,
"loss": 0.2081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10671227425336838,
"step": 3765,
"valid_targets_mean": 4592.5,
"valid_targets_min": 3752
},
{
"epoch": 6.221304706853839,
"grad_norm": 0.4762017277765319,
"learning_rate": 1.496530975729693e-06,
"loss": 0.2092,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11180372536182404,
"step": 3770,
"valid_targets_mean": 4405.1,
"valid_targets_min": 3435
},
{
"epoch": 6.229562345169281,
"grad_norm": 0.45748948302727704,
"learning_rate": 1.4654492217437222e-06,
"loss": 0.2121,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10895584523677826,
"step": 3775,
"valid_targets_mean": 4588.4,
"valid_targets_min": 3326
},
{
"epoch": 6.237819983484723,
"grad_norm": 0.44185708794256573,
"learning_rate": 1.434681357328227e-06,
"loss": 0.2043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1165788397192955,
"step": 3780,
"valid_targets_mean": 4862.5,
"valid_targets_min": 4068
},
{
"epoch": 6.246077621800165,
"grad_norm": 0.4858436395275709,
"learning_rate": 1.4042279035485251e-06,
"loss": 0.2136,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10854051262140274,
"step": 3785,
"valid_targets_mean": 4182.9,
"valid_targets_min": 3155
},
{
"epoch": 6.254335260115607,
"grad_norm": 0.45019175097080494,
"learning_rate": 1.3740893761452934e-06,
"loss": 0.2089,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10347172617912292,
"step": 3790,
"valid_targets_mean": 4602.8,
"valid_targets_min": 2746
},
{
"epoch": 6.262592898431048,
"grad_norm": 0.46537819768983907,
"learning_rate": 1.3442662855257883e-06,
"loss": 0.2238,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10508914291858673,
"step": 3795,
"valid_targets_mean": 4528.5,
"valid_targets_min": 2148
},
{
"epoch": 6.27085053674649,
"grad_norm": 0.45588668025459866,
"learning_rate": 1.3147591367552416e-06,
"loss": 0.2064,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10580870509147644,
"step": 3800,
"valid_targets_mean": 4356.2,
"valid_targets_min": 3973
},
{
"epoch": 6.279108175061932,
"grad_norm": 0.4792824685439443,
"learning_rate": 1.2855684295482918e-06,
"loss": 0.2046,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11273561418056488,
"step": 3805,
"valid_targets_mean": 4207.0,
"valid_targets_min": 2527
},
{
"epoch": 6.287365813377374,
"grad_norm": 0.48411199817039346,
"learning_rate": 1.2566946582605133e-06,
"loss": 0.204,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.110837422311306,
"step": 3810,
"valid_targets_mean": 4461.5,
"valid_targets_min": 3277
},
{
"epoch": 6.295623451692816,
"grad_norm": 0.476947448517145,
"learning_rate": 1.2281383118800472e-06,
"loss": 0.2051,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12245800346136093,
"step": 3815,
"valid_targets_mean": 4570.2,
"valid_targets_min": 2598
},
{
"epoch": 6.303881090008257,
"grad_norm": 0.47916508524756907,
"learning_rate": 1.1998998740193413e-06,
"loss": 0.2166,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11055730283260345,
"step": 3820,
"valid_targets_mean": 4556.1,
"valid_targets_min": 3816
},
{
"epoch": 6.312138728323699,
"grad_norm": 0.47800225354356507,
"learning_rate": 1.1719798229069324e-06,
"loss": 0.2116,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11608046293258667,
"step": 3825,
"valid_targets_mean": 4250.5,
"valid_targets_min": 2501
},
{
"epoch": 6.320396366639141,
"grad_norm": 0.49094193416937654,
"learning_rate": 1.1443786313793548e-06,
"loss": 0.2086,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11911673843860626,
"step": 3830,
"valid_targets_mean": 4348.6,
"valid_targets_min": 1212
},
{
"epoch": 6.328654004954583,
"grad_norm": 0.45547394722697837,
"learning_rate": 1.117096766873149e-06,
"loss": 0.2035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0916033610701561,
"step": 3835,
"valid_targets_mean": 4373.0,
"valid_targets_min": 3730
},
{
"epoch": 6.3369116432700245,
"grad_norm": 0.45137448648227896,
"learning_rate": 1.0901346914169197e-06,
"loss": 0.2123,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0887003242969513,
"step": 3840,
"valid_targets_mean": 4375.1,
"valid_targets_min": 3113
},
{
"epoch": 6.345169281585466,
"grad_norm": 0.4746701587711895,
"learning_rate": 1.0634928616235273e-06,
"loss": 0.2099,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10210265964269638,
"step": 3845,
"valid_targets_mean": 4313.0,
"valid_targets_min": 3162
},
{
"epoch": 6.353426919900908,
"grad_norm": 0.5014487466786011,
"learning_rate": 1.0371717286823601e-06,
"loss": 0.21,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10156911611557007,
"step": 3850,
"valid_targets_mean": 4364.9,
"valid_targets_min": 1888
},
{
"epoch": 6.36168455821635,
"grad_norm": 0.5006253109722267,
"learning_rate": 1.0111717383516728e-06,
"loss": 0.2102,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10288563370704651,
"step": 3855,
"valid_targets_mean": 4469.6,
"valid_targets_min": 3557
},
{
"epoch": 6.369942196531792,
"grad_norm": 0.45253458534419433,
"learning_rate": 9.854933309510618e-07,
"loss": 0.2031,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11036263406276703,
"step": 3860,
"valid_targets_mean": 4698.4,
"valid_targets_min": 4358
},
{
"epoch": 6.3781998348472335,
"grad_norm": 0.4458825392785625,
"learning_rate": 9.60136941353984e-07,
"loss": 0.2106,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11198890209197998,
"step": 3865,
"valid_targets_mean": 4588.4,
"valid_targets_min": 3461
},
{
"epoch": 6.386457473162675,
"grad_norm": 0.49673943782713165,
"learning_rate": 9.351029989804106e-07,
"loss": 0.2086,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10539565980434418,
"step": 3870,
"valid_targets_mean": 4306.5,
"valid_targets_min": 3044
},
{
"epoch": 6.394715111478117,
"grad_norm": 0.44601725759784355,
"learning_rate": 9.103919277895468e-07,
"loss": 0.2085,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09746614843606949,
"step": 3875,
"valid_targets_mean": 4567.8,
"valid_targets_min": 3620
},
{
"epoch": 6.402972749793559,
"grad_norm": 0.4574646116667568,
"learning_rate": 8.860041462726543e-07,
"loss": 0.2052,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10204243659973145,
"step": 3880,
"valid_targets_mean": 4731.1,
"valid_targets_min": 3891
},
{
"epoch": 6.411230388109001,
"grad_norm": 0.4789879181066168,
"learning_rate": 8.619400674459589e-07,
"loss": 0.2078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09731446951627731,
"step": 3885,
"valid_targets_mean": 3661.9,
"valid_targets_min": 622
},
{
"epoch": 6.4194880264244425,
"grad_norm": 0.4617914682239047,
"learning_rate": 8.38200098843669e-07,
"loss": 0.2158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10870667546987534,
"step": 3890,
"valid_targets_mean": 4445.2,
"valid_targets_min": 2737
},
{
"epoch": 6.427745664739884,
"grad_norm": 0.5158531671876079,
"learning_rate": 8.14784642511055e-07,
"loss": 0.2049,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09793481230735779,
"step": 3895,
"valid_targets_mean": 4374.4,
"valid_targets_min": 3969
},
{
"epoch": 6.436003303055326,
"grad_norm": 0.5136943340362236,
"learning_rate": 7.916940949976526e-07,
"loss": 0.2143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10914270579814911,
"step": 3900,
"valid_targets_mean": 4438.2,
"valid_targets_min": 3209
},
{
"epoch": 6.444260941370768,
"grad_norm": 0.4470643625267753,
"learning_rate": 7.689288473505474e-07,
"loss": 0.2132,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09543461352586746,
"step": 3905,
"valid_targets_mean": 4312.8,
"valid_targets_min": 2528
},
{
"epoch": 6.45251857968621,
"grad_norm": 0.46425979055264205,
"learning_rate": 7.464892851077499e-07,
"loss": 0.2112,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10784029215574265,
"step": 3910,
"valid_targets_mean": 4370.2,
"valid_targets_min": 2016
},
{
"epoch": 6.4607762180016515,
"grad_norm": 0.4489860976833304,
"learning_rate": 7.243757882916624e-07,
"loss": 0.2139,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10391361266374588,
"step": 3915,
"valid_targets_mean": 3928.5,
"valid_targets_min": 1663
},
{
"epoch": 6.469033856317093,
"grad_norm": 0.4593123891844981,
"learning_rate": 7.025887314026513e-07,
"loss": 0.2069,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11030080169439316,
"step": 3920,
"valid_targets_mean": 4337.5,
"valid_targets_min": 3786
},
{
"epoch": 6.477291494632535,
"grad_norm": 0.43752265525654976,
"learning_rate": 6.811284834126963e-07,
"loss": 0.2066,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10311497747898102,
"step": 3925,
"valid_targets_mean": 4161.2,
"valid_targets_min": 2408
},
{
"epoch": 6.485549132947977,
"grad_norm": 0.43879318763576247,
"learning_rate": 6.599954077591464e-07,
"loss": 0.2043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10299549996852875,
"step": 3930,
"valid_targets_mean": 4404.0,
"valid_targets_min": 3939
},
{
"epoch": 6.493806771263419,
"grad_norm": 0.4809863129872432,
"learning_rate": 6.391898623385695e-07,
"loss": 0.2187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09846025705337524,
"step": 3935,
"valid_targets_mean": 4055.9,
"valid_targets_min": 1497
},
{
"epoch": 6.5020644095788604,
"grad_norm": 0.4545468335526449,
"learning_rate": 6.187121995006817e-07,
"loss": 0.2086,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10218028724193573,
"step": 3940,
"valid_targets_mean": 4510.0,
"valid_targets_min": 3774
},
{
"epoch": 6.510322047894302,
"grad_norm": 0.4601713003151688,
"learning_rate": 5.98562766042381e-07,
"loss": 0.2072,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11521266400814056,
"step": 3945,
"valid_targets_mean": 5056.4,
"valid_targets_min": 4248
},
{
"epoch": 6.518579686209744,
"grad_norm": 0.4825411092899984,
"learning_rate": 5.78741903201887e-07,
"loss": 0.2061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10463918745517731,
"step": 3950,
"valid_targets_mean": 4606.4,
"valid_targets_min": 3306
},
{
"epoch": 6.526837324525186,
"grad_norm": 0.4724346445489711,
"learning_rate": 5.592499466529445e-07,
"loss": 0.2081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11009636521339417,
"step": 3955,
"valid_targets_mean": 4472.9,
"valid_targets_min": 3897
},
{
"epoch": 6.535094962840628,
"grad_norm": 0.47715700504044883,
"learning_rate": 5.400872264991508e-07,
"loss": 0.2098,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11740729957818985,
"step": 3960,
"valid_targets_mean": 4494.6,
"valid_targets_min": 3859
},
{
"epoch": 6.543352601156069,
"grad_norm": 0.461209846576918,
"learning_rate": 5.212540672683575e-07,
"loss": 0.2103,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12106248736381531,
"step": 3965,
"valid_targets_mean": 4647.2,
"valid_targets_min": 3545
},
{
"epoch": 6.551610239471511,
"grad_norm": 0.476073957208728,
"learning_rate": 5.027507879071869e-07,
"loss": 0.1998,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0960778295993805,
"step": 3970,
"valid_targets_mean": 4646.2,
"valid_targets_min": 3614
},
{
"epoch": 6.559867877786953,
"grad_norm": 0.45682873048403283,
"learning_rate": 4.845777017756126e-07,
"loss": 0.216,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1089847981929779,
"step": 3975,
"valid_targets_mean": 4550.6,
"valid_targets_min": 3621
},
{
"epoch": 6.568125516102395,
"grad_norm": 1.4081138193181622,
"learning_rate": 4.667351166416678e-07,
"loss": 0.2086,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08985770493745804,
"step": 3980,
"valid_targets_mean": 4168.5,
"valid_targets_min": 2524
},
{
"epoch": 6.576383154417837,
"grad_norm": 0.4763412844198788,
"learning_rate": 4.4922333467622316e-07,
"loss": 0.2067,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09986129403114319,
"step": 3985,
"valid_targets_mean": 4318.6,
"valid_targets_min": 3494
},
{
"epoch": 6.584640792733278,
"grad_norm": 0.43305137156838286,
"learning_rate": 4.320426524478749e-07,
"loss": 0.2102,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08903586119413376,
"step": 3990,
"valid_targets_mean": 4401.9,
"valid_targets_min": 3524
},
{
"epoch": 6.59289843104872,
"grad_norm": 0.47550998100773834,
"learning_rate": 4.1519336091792263e-07,
"loss": 0.2092,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11374057829380035,
"step": 3995,
"valid_targets_mean": 4725.8,
"valid_targets_min": 3753
},
{
"epoch": 6.601156069364162,
"grad_norm": 0.43748710749795694,
"learning_rate": 3.9867574543544174e-07,
"loss": 0.21,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09821216762065887,
"step": 4000,
"valid_targets_mean": 4954.0,
"valid_targets_min": 3796
},
{
"epoch": 6.609413707679604,
"grad_norm": 0.48736868817977974,
"learning_rate": 3.824900857324432e-07,
"loss": 0.2092,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08940805494785309,
"step": 4005,
"valid_targets_mean": 4183.1,
"valid_targets_min": 3075
},
{
"epoch": 6.617671345995046,
"grad_norm": 0.4700029876843552,
"learning_rate": 3.6663665591915033e-07,
"loss": 0.2107,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10451637208461761,
"step": 4010,
"valid_targets_mean": 4126.1,
"valid_targets_min": 2975
},
{
"epoch": 6.625928984310487,
"grad_norm": 0.4681474747997738,
"learning_rate": 3.511157244793384e-07,
"loss": 0.2081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12119415402412415,
"step": 4015,
"valid_targets_mean": 4749.9,
"valid_targets_min": 4256
},
{
"epoch": 6.634186622625929,
"grad_norm": 0.4582949979156882,
"learning_rate": 3.359275542658069e-07,
"loss": 0.2153,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10983027517795563,
"step": 4020,
"valid_targets_mean": 4142.0,
"valid_targets_min": 797
},
{
"epoch": 6.642444260941371,
"grad_norm": 0.44381098418418435,
"learning_rate": 3.2107240249591887e-07,
"loss": 0.2137,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11540737748146057,
"step": 4025,
"valid_targets_mean": 4457.8,
"valid_targets_min": 3948
},
{
"epoch": 6.650701899256813,
"grad_norm": 0.4868086644302285,
"learning_rate": 3.0655052074723747e-07,
"loss": 0.2005,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10091888904571533,
"step": 4030,
"valid_targets_mean": 4441.6,
"valid_targets_min": 3376
},
{
"epoch": 6.658959537572255,
"grad_norm": 0.45823702182669723,
"learning_rate": 2.9236215495328067e-07,
"loss": 0.2117,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10041674226522446,
"step": 4035,
"valid_targets_mean": 4329.6,
"valid_targets_min": 3214
},
{
"epoch": 6.667217175887696,
"grad_norm": 0.4537952704741499,
"learning_rate": 2.785075453993469e-07,
"loss": 0.2063,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0971527099609375,
"step": 4040,
"valid_targets_mean": 4205.5,
"valid_targets_min": 3556
},
{
"epoch": 6.675474814203138,
"grad_norm": 0.4690460617014493,
"learning_rate": 2.649869267184402e-07,
"loss": 0.2101,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12965399026870728,
"step": 4045,
"valid_targets_mean": 5110.6,
"valid_targets_min": 2455
},
{
"epoch": 6.68373245251858,
"grad_norm": 0.506281179695536,
"learning_rate": 2.518005278873159e-07,
"loss": 0.2043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12673425674438477,
"step": 4050,
"valid_targets_mean": 4649.1,
"valid_targets_min": 3982
},
{
"epoch": 6.691990090834022,
"grad_norm": 0.4427903814369831,
"learning_rate": 2.389485722225837e-07,
"loss": 0.2095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10964621603488922,
"step": 4055,
"valid_targets_mean": 4922.0,
"valid_targets_min": 3194
},
{
"epoch": 6.7002477291494635,
"grad_norm": 0.44276905835578534,
"learning_rate": 2.2643127737693503e-07,
"loss": 0.2073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10598443448543549,
"step": 4060,
"valid_targets_mean": 4640.4,
"valid_targets_min": 3585
},
{
"epoch": 6.708505367464905,
"grad_norm": 0.4699901743967069,
"learning_rate": 2.1424885533545269e-07,
"loss": 0.2094,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10731061547994614,
"step": 4065,
"valid_targets_mean": 4614.5,
"valid_targets_min": 4172
},
{
"epoch": 6.716763005780347,
"grad_norm": 0.4449097656891119,
"learning_rate": 2.0240151241202265e-07,
"loss": 0.2113,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10288381576538086,
"step": 4070,
"valid_targets_mean": 4325.4,
"valid_targets_min": 1280
},
{
"epoch": 6.725020644095789,
"grad_norm": 0.4739566939740566,
"learning_rate": 1.9088944924584572e-07,
"loss": 0.214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10443182289600372,
"step": 4075,
"valid_targets_mean": 4395.5,
"valid_targets_min": 2693
},
{
"epoch": 6.733278282411231,
"grad_norm": 0.4772601069279297,
"learning_rate": 1.7971286079802474e-07,
"loss": 0.2152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10051245987415314,
"step": 4080,
"valid_targets_mean": 3800.5,
"valid_targets_min": 2203
},
{
"epoch": 6.7415359207266725,
"grad_norm": 0.49020107661823686,
"learning_rate": 1.6887193634828048e-07,
"loss": 0.2148,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09047770500183105,
"step": 4085,
"valid_targets_mean": 4388.6,
"valid_targets_min": 3139
},
{
"epoch": 6.749793559042114,
"grad_norm": 0.4819390590838297,
"learning_rate": 1.5836685949173648e-07,
"loss": 0.2035,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12125623226165771,
"step": 4090,
"valid_targets_mean": 5068.4,
"valid_targets_min": 3379
},
{
"epoch": 6.758051197357556,
"grad_norm": 0.4445923586702321,
"learning_rate": 1.481978081358104e-07,
"loss": 0.206,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0934639573097229,
"step": 4095,
"valid_targets_mean": 4460.9,
"valid_targets_min": 3897
},
{
"epoch": 6.766308835672998,
"grad_norm": 0.4642864304956839,
"learning_rate": 1.3836495449719878e-07,
"loss": 0.2157,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11159105598926544,
"step": 4100,
"valid_targets_mean": 4308.4,
"valid_targets_min": 3249
},
{
"epoch": 6.77456647398844,
"grad_norm": 0.45803230484756524,
"learning_rate": 1.28868465098968e-07,
"loss": 0.2059,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10908431559801102,
"step": 4105,
"valid_targets_mean": 4308.9,
"valid_targets_min": 3380
},
{
"epoch": 6.7828241123038815,
"grad_norm": 0.45426920673362553,
"learning_rate": 1.1970850076773234e-07,
"loss": 0.2144,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11683052778244019,
"step": 4110,
"valid_targets_mean": 4576.9,
"valid_targets_min": 3651
},
{
"epoch": 6.791081750619323,
"grad_norm": 0.5015754489784501,
"learning_rate": 1.1088521663091823e-07,
"loss": 0.2091,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09520648419857025,
"step": 4115,
"valid_targets_mean": 3644.6,
"valid_targets_min": 1381
},
{
"epoch": 6.799339388934765,
"grad_norm": 0.45884889184363076,
"learning_rate": 1.0239876211415533e-07,
"loss": 0.2113,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09719965606927872,
"step": 4120,
"valid_targets_mean": 4378.5,
"valid_targets_min": 3453
},
{
"epoch": 6.807597027250207,
"grad_norm": 0.47934430399214695,
"learning_rate": 9.424928093873409e-08,
"loss": 0.2163,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11145424842834473,
"step": 4125,
"valid_targets_mean": 4738.4,
"valid_targets_min": 3774
},
{
"epoch": 6.815854665565649,
"grad_norm": 0.4523964889908379,
"learning_rate": 8.643691111917652e-08,
"loss": 0.2095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09790743887424469,
"step": 4130,
"valid_targets_mean": 4394.9,
"valid_targets_min": 3894
},
{
"epoch": 6.8241123038810905,
"grad_norm": 0.44214206646031545,
"learning_rate": 7.896178496089368e-08,
"loss": 0.2149,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12457379698753357,
"step": 4135,
"valid_targets_mean": 5277.6,
"valid_targets_min": 2644
},
{
"epoch": 6.832369942196532,
"grad_norm": 0.46496701389167644,
"learning_rate": 7.182402905795194e-08,
"loss": 0.2088,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11708584427833557,
"step": 4140,
"valid_targets_mean": 4939.9,
"valid_targets_min": 3894
},
{
"epoch": 6.840627580511974,
"grad_norm": 0.48893855804217057,
"learning_rate": 6.502376429092794e-08,
"loss": 0.2104,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09200628846883774,
"step": 4145,
"valid_targets_mean": 4517.6,
"valid_targets_min": 4128
},
{
"epoch": 6.848885218827416,
"grad_norm": 0.4460256402035902,
"learning_rate": 5.856110582485475e-08,
"loss": 0.2082,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09956889599561691,
"step": 4150,
"valid_targets_mean": 4091.9,
"valid_targets_min": 3170
},
{
"epoch": 6.857142857142857,
"grad_norm": 0.4572235238681599,
"learning_rate": 5.2436163107276726e-08,
"loss": 0.2038,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10769429802894592,
"step": 4155,
"valid_targets_mean": 4496.5,
"valid_targets_min": 3766
},
{
"epoch": 6.865400495458299,
"grad_norm": 0.4779799274730323,
"learning_rate": 4.6649039866399904e-08,
"loss": 0.2128,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10232293605804443,
"step": 4160,
"valid_targets_mean": 4342.2,
"valid_targets_min": 2839
},
{
"epoch": 6.87365813377374,
"grad_norm": 0.45348117194408033,
"learning_rate": 4.1199834109322266e-08,
"loss": 0.215,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09970030188560486,
"step": 4165,
"valid_targets_mean": 4795.4,
"valid_targets_min": 2883
},
{
"epoch": 6.881915772089182,
"grad_norm": 0.4721452975187866,
"learning_rate": 3.608863812039065e-08,
"loss": 0.2121,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10781273245811462,
"step": 4170,
"valid_targets_mean": 4330.2,
"valid_targets_min": 2980
},
{
"epoch": 6.890173410404624,
"grad_norm": 0.4801557283415626,
"learning_rate": 3.131553845962643e-08,
"loss": 0.2101,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.08960318565368652,
"step": 4175,
"valid_targets_mean": 3620.2,
"valid_targets_min": 1766
},
{
"epoch": 6.898431048720066,
"grad_norm": 0.48788556292480284,
"learning_rate": 2.6880615961264456e-08,
"loss": 0.2078,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10963458567857742,
"step": 4180,
"valid_targets_mean": 4450.4,
"valid_targets_min": 3490
},
{
"epoch": 6.9066886870355075,
"grad_norm": 0.47657779870651035,
"learning_rate": 2.278394573237641e-08,
"loss": 0.2067,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09093846380710602,
"step": 4185,
"valid_targets_mean": 4248.6,
"valid_targets_min": 3743
},
{
"epoch": 6.914946325350949,
"grad_norm": 0.47477031835203715,
"learning_rate": 1.9025597151614006e-08,
"loss": 0.2143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09622550010681152,
"step": 4190,
"valid_targets_mean": 4264.9,
"valid_targets_min": 3707
},
{
"epoch": 6.923203963666391,
"grad_norm": 0.4755784709006041,
"learning_rate": 1.5605633868018833e-08,
"loss": 0.2122,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09107546508312225,
"step": 4195,
"valid_targets_mean": 4052.6,
"valid_targets_min": 2044
},
{
"epoch": 6.931461601981833,
"grad_norm": 0.4657787461520029,
"learning_rate": 1.2524113799949888e-08,
"loss": 0.2033,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09151225537061691,
"step": 4200,
"valid_targets_mean": 4014.9,
"valid_targets_min": 3221
},
{
"epoch": 6.939719240297275,
"grad_norm": 0.46532756631825184,
"learning_rate": 9.781089134108801e-09,
"loss": 0.2142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.0972677543759346,
"step": 4205,
"valid_targets_mean": 4152.4,
"valid_targets_min": 867
},
{
"epoch": 6.9479768786127165,
"grad_norm": 0.5218672628521883,
"learning_rate": 7.376606324644986e-09,
"loss": 0.2111,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12868523597717285,
"step": 4210,
"valid_targets_mean": 4516.1,
"valid_targets_min": 2004
},
{
"epoch": 6.956234516928158,
"grad_norm": 0.4388699940574022,
"learning_rate": 5.310706092378493e-09,
"loss": 0.2112,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09615793824195862,
"step": 4215,
"valid_targets_mean": 4452.1,
"valid_targets_min": 3642
},
{
"epoch": 6.9644921552436,
"grad_norm": 0.4635429887568062,
"learning_rate": 3.5834234241050037e-09,
"loss": 0.2098,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1175151914358139,
"step": 4220,
"valid_targets_mean": 5389.9,
"valid_targets_min": 4205
},
{
"epoch": 6.972749793559042,
"grad_norm": 0.5025208995888553,
"learning_rate": 2.194787572000756e-09,
"loss": 0.2116,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1054515689611435,
"step": 4225,
"valid_targets_mean": 4109.4,
"valid_targets_min": 3087
},
{
"epoch": 6.981007431874484,
"grad_norm": 0.46742749625209445,
"learning_rate": 1.1448220531407039e-09,
"loss": 0.2101,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10800908505916595,
"step": 4230,
"valid_targets_mean": 4283.4,
"valid_targets_min": 2593
},
{
"epoch": 6.9892650701899255,
"grad_norm": 0.4565839219950087,
"learning_rate": 4.3354464907885417e-10,
"loss": 0.2062,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10034844279289246,
"step": 4235,
"valid_targets_mean": 3978.0,
"valid_targets_min": 2225
},
{
"epoch": 6.997522708505367,
"grad_norm": 0.4647571756194701,
"learning_rate": 6.096740556849057e-11,
"loss": 0.2019,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.096927210688591,
"step": 4240,
"valid_targets_mean": 4383.1,
"valid_targets_min": 3058
},
{
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17524641752243042,
"step": 4242,
"total_flos": 2.557069172618035e+18,
"train_loss": 0.2496959186285973,
"train_runtime": 54891.1457,
"train_samples_per_second": 1.235,
"train_steps_per_second": 0.077,
"valid_targets_mean": 4057.6,
"valid_targets_min": 3531
}
],
"logging_steps": 5,
"max_steps": 4242,
"num_input_tokens_seen": 0,
"num_train_epochs": 7,
"save_steps": 1500,
"stateful_callbacks": {
"TrainerControl": {
"args": {
"should_epoch_stop": false,
"should_evaluate": false,
"should_log": false,
"should_save": true,
"should_training_stop": true
},
"attributes": {}
}
},
"total_flos": 2.557069172618035e+18,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}