Files
a1-nemotron_pytest/trainer_state.json
ModelHub XC 8872701257 初始化项目,由ModelHub XC社区提供模型
Model: DCAgent/a1-nemotron_pytest
Source: Original Platform
2026-05-02 03:47:13 +08:00

9266 lines
258 KiB
JSON

{
"best_global_step": null,
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 7.0,
"eval_steps": 500,
"global_step": 4193,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.008347245409015025,
"grad_norm": 16.449758471449798,
"learning_rate": 3.80952380952381e-07,
"loss": 0.771,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7790951728820801,
"step": 5,
"valid_targets_mean": 4794.1,
"valid_targets_min": 2090
},
{
"epoch": 0.01669449081803005,
"grad_norm": 18.499364259031307,
"learning_rate": 8.571428571428572e-07,
"loss": 0.7457,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7398654222488403,
"step": 10,
"valid_targets_mean": 3437.5,
"valid_targets_min": 1552
},
{
"epoch": 0.025041736227045076,
"grad_norm": 13.721168031256521,
"learning_rate": 1.3333333333333334e-06,
"loss": 0.7487,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7121936678886414,
"step": 15,
"valid_targets_mean": 3884.1,
"valid_targets_min": 1573
},
{
"epoch": 0.0333889816360601,
"grad_norm": 11.786744858572096,
"learning_rate": 1.8095238095238097e-06,
"loss": 0.6962,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.7608169317245483,
"step": 20,
"valid_targets_mean": 3273.8,
"valid_targets_min": 1554
},
{
"epoch": 0.041736227045075125,
"grad_norm": 7.518407995602129,
"learning_rate": 2.285714285714286e-06,
"loss": 0.6181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.557266354560852,
"step": 25,
"valid_targets_mean": 4375.6,
"valid_targets_min": 1495
},
{
"epoch": 0.05008347245409015,
"grad_norm": 5.036039563500205,
"learning_rate": 2.7619047619047625e-06,
"loss": 0.5314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5265945792198181,
"step": 30,
"valid_targets_mean": 3589.2,
"valid_targets_min": 1394
},
{
"epoch": 0.05843071786310518,
"grad_norm": 2.3063294581378013,
"learning_rate": 3.2380952380952385e-06,
"loss": 0.5395,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5409011840820312,
"step": 35,
"valid_targets_mean": 5205.4,
"valid_targets_min": 1724
},
{
"epoch": 0.0667779632721202,
"grad_norm": 1.529477865502694,
"learning_rate": 3.7142857142857146e-06,
"loss": 0.4848,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.5050346851348877,
"step": 40,
"valid_targets_mean": 4925.6,
"valid_targets_min": 1340
},
{
"epoch": 0.07512520868113523,
"grad_norm": 1.216717479154899,
"learning_rate": 4.190476190476191e-06,
"loss": 0.4646,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3834213614463806,
"step": 45,
"valid_targets_mean": 3388.4,
"valid_targets_min": 1047
},
{
"epoch": 0.08347245409015025,
"grad_norm": 0.9899789336607441,
"learning_rate": 4.666666666666667e-06,
"loss": 0.4316,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.44761449098587036,
"step": 50,
"valid_targets_mean": 4702.2,
"valid_targets_min": 1416
},
{
"epoch": 0.09181969949916527,
"grad_norm": 0.9245176447241443,
"learning_rate": 5.142857142857142e-06,
"loss": 0.4305,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4472816586494446,
"step": 55,
"valid_targets_mean": 3888.0,
"valid_targets_min": 1370
},
{
"epoch": 0.1001669449081803,
"grad_norm": 0.8445583236484762,
"learning_rate": 5.619047619047619e-06,
"loss": 0.4358,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.393363356590271,
"step": 60,
"valid_targets_mean": 3676.4,
"valid_targets_min": 1367
},
{
"epoch": 0.10851419031719532,
"grad_norm": 0.938311524953121,
"learning_rate": 6.095238095238096e-06,
"loss": 0.4018,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.35671281814575195,
"step": 65,
"valid_targets_mean": 3521.1,
"valid_targets_min": 1449
},
{
"epoch": 0.11686143572621036,
"grad_norm": 0.7780942778568863,
"learning_rate": 6.571428571428572e-06,
"loss": 0.3901,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.39001137018203735,
"step": 70,
"valid_targets_mean": 3851.3,
"valid_targets_min": 1424
},
{
"epoch": 0.12520868113522537,
"grad_norm": 0.7824275389127349,
"learning_rate": 7.047619047619048e-06,
"loss": 0.4079,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3591366410255432,
"step": 75,
"valid_targets_mean": 3520.2,
"valid_targets_min": 1707
},
{
"epoch": 0.1335559265442404,
"grad_norm": 0.6539830798972561,
"learning_rate": 7.523809523809524e-06,
"loss": 0.3758,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4063882827758789,
"step": 80,
"valid_targets_mean": 3542.5,
"valid_targets_min": 1496
},
{
"epoch": 0.1419031719532554,
"grad_norm": 0.5848543693257306,
"learning_rate": 8.000000000000001e-06,
"loss": 0.419,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3979543447494507,
"step": 85,
"valid_targets_mean": 4294.7,
"valid_targets_min": 1430
},
{
"epoch": 0.15025041736227046,
"grad_norm": 0.5909348892197375,
"learning_rate": 8.476190476190477e-06,
"loss": 0.3738,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3587251305580139,
"step": 90,
"valid_targets_mean": 4082.9,
"valid_targets_min": 1486
},
{
"epoch": 0.15859766277128548,
"grad_norm": 0.5809796978248298,
"learning_rate": 8.952380952380953e-06,
"loss": 0.385,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3860463500022888,
"step": 95,
"valid_targets_mean": 4603.4,
"valid_targets_min": 2148
},
{
"epoch": 0.1669449081803005,
"grad_norm": 0.583544270082398,
"learning_rate": 9.42857142857143e-06,
"loss": 0.3311,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24891161918640137,
"step": 100,
"valid_targets_mean": 3401.2,
"valid_targets_min": 1484
},
{
"epoch": 0.17529215358931552,
"grad_norm": 0.548090730982247,
"learning_rate": 9.904761904761906e-06,
"loss": 0.339,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4126247465610504,
"step": 105,
"valid_targets_mean": 4525.1,
"valid_targets_min": 1370
},
{
"epoch": 0.18363939899833054,
"grad_norm": 0.5265473890988233,
"learning_rate": 1.0380952380952383e-05,
"loss": 0.3037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32239624857902527,
"step": 110,
"valid_targets_mean": 4051.6,
"valid_targets_min": 1334
},
{
"epoch": 0.19198664440734559,
"grad_norm": 0.6603057929937589,
"learning_rate": 1.0857142857142858e-05,
"loss": 0.3414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4038582444190979,
"step": 115,
"valid_targets_mean": 3750.9,
"valid_targets_min": 1448
},
{
"epoch": 0.2003338898163606,
"grad_norm": 0.6012115208695145,
"learning_rate": 1.1333333333333334e-05,
"loss": 0.3074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3197484612464905,
"step": 120,
"valid_targets_mean": 4089.4,
"valid_targets_min": 1800
},
{
"epoch": 0.20868113522537562,
"grad_norm": 1.7458395800289577,
"learning_rate": 1.180952380952381e-05,
"loss": 0.3271,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.30865028500556946,
"step": 125,
"valid_targets_mean": 4051.4,
"valid_targets_min": 1367
},
{
"epoch": 0.21702838063439064,
"grad_norm": 0.5452040221763657,
"learning_rate": 1.2285714285714288e-05,
"loss": 0.3296,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3391510248184204,
"step": 130,
"valid_targets_mean": 4346.8,
"valid_targets_min": 1438
},
{
"epoch": 0.22537562604340566,
"grad_norm": 0.9415286369100476,
"learning_rate": 1.2761904761904762e-05,
"loss": 0.296,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.32920509576797485,
"step": 135,
"valid_targets_mean": 4172.4,
"valid_targets_min": 1951
},
{
"epoch": 0.2337228714524207,
"grad_norm": 0.6842387685716974,
"learning_rate": 1.3238095238095238e-05,
"loss": 0.282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.253879189491272,
"step": 140,
"valid_targets_mean": 2951.6,
"valid_targets_min": 1386
},
{
"epoch": 0.24207011686143573,
"grad_norm": 0.5428335943460046,
"learning_rate": 1.3714285714285716e-05,
"loss": 0.2883,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31553781032562256,
"step": 145,
"valid_targets_mean": 4643.6,
"valid_targets_min": 1617
},
{
"epoch": 0.25041736227045075,
"grad_norm": 0.6724597835951847,
"learning_rate": 1.4190476190476192e-05,
"loss": 0.2663,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26506760716438293,
"step": 150,
"valid_targets_mean": 3644.8,
"valid_targets_min": 772
},
{
"epoch": 0.2587646076794658,
"grad_norm": 0.5208002017829473,
"learning_rate": 1.4666666666666666e-05,
"loss": 0.2876,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3080453872680664,
"step": 155,
"valid_targets_mean": 4450.4,
"valid_targets_min": 1362
},
{
"epoch": 0.2671118530884808,
"grad_norm": 0.6362864479443031,
"learning_rate": 1.5142857142857144e-05,
"loss": 0.3056,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24584180116653442,
"step": 160,
"valid_targets_mean": 3364.2,
"valid_targets_min": 1336
},
{
"epoch": 0.27545909849749584,
"grad_norm": 0.7315546020728038,
"learning_rate": 1.5619047619047622e-05,
"loss": 0.2583,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.220881849527359,
"step": 165,
"valid_targets_mean": 3647.0,
"valid_targets_min": 1862
},
{
"epoch": 0.2838063439065108,
"grad_norm": 0.5640979485294231,
"learning_rate": 1.6095238095238096e-05,
"loss": 0.2646,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27528372406959534,
"step": 170,
"valid_targets_mean": 4019.7,
"valid_targets_min": 1650
},
{
"epoch": 0.2921535893155259,
"grad_norm": 0.5692730266909324,
"learning_rate": 1.6571428571428574e-05,
"loss": 0.2465,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24535271525382996,
"step": 175,
"valid_targets_mean": 3414.9,
"valid_targets_min": 1581
},
{
"epoch": 0.3005008347245409,
"grad_norm": 0.687838002070504,
"learning_rate": 1.704761904761905e-05,
"loss": 0.2886,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2961692214012146,
"step": 180,
"valid_targets_mean": 3081.3,
"valid_targets_min": 718
},
{
"epoch": 0.3088480801335559,
"grad_norm": 0.5023153292940437,
"learning_rate": 1.7523809523809526e-05,
"loss": 0.2571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3332681655883789,
"step": 185,
"valid_targets_mean": 5757.7,
"valid_targets_min": 1686
},
{
"epoch": 0.31719532554257096,
"grad_norm": 0.6388958525954316,
"learning_rate": 1.8e-05,
"loss": 0.253,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3381766676902771,
"step": 190,
"valid_targets_mean": 4048.6,
"valid_targets_min": 1285
},
{
"epoch": 0.32554257095158595,
"grad_norm": 0.5156507606960894,
"learning_rate": 1.8476190476190478e-05,
"loss": 0.2646,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3153945207595825,
"step": 195,
"valid_targets_mean": 5093.8,
"valid_targets_min": 1894
},
{
"epoch": 0.333889816360601,
"grad_norm": 0.5522919560402644,
"learning_rate": 1.8952380952380953e-05,
"loss": 0.2776,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22514237463474274,
"step": 200,
"valid_targets_mean": 4099.3,
"valid_targets_min": 1450
},
{
"epoch": 0.34223706176961605,
"grad_norm": 0.560240899473144,
"learning_rate": 1.942857142857143e-05,
"loss": 0.2547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21474067866802216,
"step": 205,
"valid_targets_mean": 4329.1,
"valid_targets_min": 1318
},
{
"epoch": 0.35058430717863104,
"grad_norm": 0.7232272629915736,
"learning_rate": 1.9904761904761908e-05,
"loss": 0.2413,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19220605492591858,
"step": 210,
"valid_targets_mean": 2633.8,
"valid_targets_min": 1539
},
{
"epoch": 0.3589315525876461,
"grad_norm": 0.650023748545767,
"learning_rate": 2.0380952380952382e-05,
"loss": 0.2488,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20408010482788086,
"step": 215,
"valid_targets_mean": 3216.6,
"valid_targets_min": 1831
},
{
"epoch": 0.3672787979966611,
"grad_norm": 0.5882427826001565,
"learning_rate": 2.085714285714286e-05,
"loss": 0.231,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24231554567813873,
"step": 220,
"valid_targets_mean": 3868.1,
"valid_targets_min": 1512
},
{
"epoch": 0.3756260434056761,
"grad_norm": 0.6910962961139102,
"learning_rate": 2.1333333333333335e-05,
"loss": 0.2628,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2768801152706146,
"step": 225,
"valid_targets_mean": 3150.6,
"valid_targets_min": 1528
},
{
"epoch": 0.38397328881469117,
"grad_norm": 0.7092676414028564,
"learning_rate": 2.180952380952381e-05,
"loss": 0.2541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.215276837348938,
"step": 230,
"valid_targets_mean": 3448.4,
"valid_targets_min": 1426
},
{
"epoch": 0.39232053422370616,
"grad_norm": 0.6152281065448418,
"learning_rate": 2.2285714285714287e-05,
"loss": 0.2613,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21806620061397552,
"step": 235,
"valid_targets_mean": 3206.4,
"valid_targets_min": 1427
},
{
"epoch": 0.4006677796327212,
"grad_norm": 0.6012559719818651,
"learning_rate": 2.2761904761904765e-05,
"loss": 0.2473,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.31128859519958496,
"step": 240,
"valid_targets_mean": 4090.4,
"valid_targets_min": 1491
},
{
"epoch": 0.4090150250417362,
"grad_norm": 0.6841189490659898,
"learning_rate": 2.3238095238095242e-05,
"loss": 0.2443,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2116287648677826,
"step": 245,
"valid_targets_mean": 3225.9,
"valid_targets_min": 1518
},
{
"epoch": 0.41736227045075125,
"grad_norm": 0.5513899993312625,
"learning_rate": 2.3714285714285717e-05,
"loss": 0.2503,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28415653109550476,
"step": 250,
"valid_targets_mean": 4458.2,
"valid_targets_min": 1931
},
{
"epoch": 0.4257095158597663,
"grad_norm": 0.6018159042388,
"learning_rate": 2.419047619047619e-05,
"loss": 0.2773,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3083229660987854,
"step": 255,
"valid_targets_mean": 3830.6,
"valid_targets_min": 1622
},
{
"epoch": 0.4340567612687813,
"grad_norm": 0.6490907452773345,
"learning_rate": 2.466666666666667e-05,
"loss": 0.2548,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3152938485145569,
"step": 260,
"valid_targets_mean": 4199.6,
"valid_targets_min": 2064
},
{
"epoch": 0.44240400667779634,
"grad_norm": 0.5570684649448066,
"learning_rate": 2.5142857142857143e-05,
"loss": 0.2537,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19224336743354797,
"step": 265,
"valid_targets_mean": 4087.6,
"valid_targets_min": 2087
},
{
"epoch": 0.4507512520868113,
"grad_norm": 0.6469910330859645,
"learning_rate": 2.5619047619047618e-05,
"loss": 0.2246,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21363332867622375,
"step": 270,
"valid_targets_mean": 3349.8,
"valid_targets_min": 1496
},
{
"epoch": 0.4590984974958264,
"grad_norm": 0.62201338541621,
"learning_rate": 2.60952380952381e-05,
"loss": 0.2331,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2003246247768402,
"step": 275,
"valid_targets_mean": 3328.9,
"valid_targets_min": 2058
},
{
"epoch": 0.4674457429048414,
"grad_norm": 0.5348486583929786,
"learning_rate": 2.6571428571428573e-05,
"loss": 0.2299,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22155584394931793,
"step": 280,
"valid_targets_mean": 5291.0,
"valid_targets_min": 1563
},
{
"epoch": 0.4757929883138564,
"grad_norm": 0.5295963625289745,
"learning_rate": 2.704761904761905e-05,
"loss": 0.2239,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2738932967185974,
"step": 285,
"valid_targets_mean": 5086.9,
"valid_targets_min": 1337
},
{
"epoch": 0.48414023372287146,
"grad_norm": 0.578953712591871,
"learning_rate": 2.7523809523809525e-05,
"loss": 0.2349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22435756027698517,
"step": 290,
"valid_targets_mean": 3735.0,
"valid_targets_min": 1532
},
{
"epoch": 0.49248747913188645,
"grad_norm": 0.660802939458494,
"learning_rate": 2.8e-05,
"loss": 0.2271,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2040880024433136,
"step": 295,
"valid_targets_mean": 3545.1,
"valid_targets_min": 2099
},
{
"epoch": 0.5008347245409015,
"grad_norm": 0.6013315709895742,
"learning_rate": 2.847619047619048e-05,
"loss": 0.2267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.196973979473114,
"step": 300,
"valid_targets_mean": 3670.2,
"valid_targets_min": 1456
},
{
"epoch": 0.5091819699499165,
"grad_norm": 0.5794099021432261,
"learning_rate": 2.8952380952380955e-05,
"loss": 0.228,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2512788772583008,
"step": 305,
"valid_targets_mean": 3780.6,
"valid_targets_min": 1474
},
{
"epoch": 0.5175292153589316,
"grad_norm": 0.637924768013473,
"learning_rate": 2.9428571428571433e-05,
"loss": 0.2356,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2885277271270752,
"step": 310,
"valid_targets_mean": 5099.4,
"valid_targets_min": 1471
},
{
"epoch": 0.5258764607679466,
"grad_norm": 0.6293175139028003,
"learning_rate": 2.9904761904761907e-05,
"loss": 0.243,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20008844137191772,
"step": 315,
"valid_targets_mean": 3679.3,
"valid_targets_min": 1550
},
{
"epoch": 0.5342237061769616,
"grad_norm": 0.5393744599389656,
"learning_rate": 3.038095238095238e-05,
"loss": 0.2113,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1867285966873169,
"step": 320,
"valid_targets_mean": 3737.2,
"valid_targets_min": 1411
},
{
"epoch": 0.5425709515859767,
"grad_norm": 0.5636648107048134,
"learning_rate": 3.085714285714286e-05,
"loss": 0.2306,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1869732141494751,
"step": 325,
"valid_targets_mean": 3590.1,
"valid_targets_min": 1414
},
{
"epoch": 0.5509181969949917,
"grad_norm": 0.6740214000121728,
"learning_rate": 3.1333333333333334e-05,
"loss": 0.2304,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20452925562858582,
"step": 330,
"valid_targets_mean": 3557.4,
"valid_targets_min": 1364
},
{
"epoch": 0.5592654424040067,
"grad_norm": 0.608287565385674,
"learning_rate": 3.180952380952381e-05,
"loss": 0.2264,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22231003642082214,
"step": 335,
"valid_targets_mean": 3440.7,
"valid_targets_min": 1323
},
{
"epoch": 0.5676126878130217,
"grad_norm": 0.5703702079712171,
"learning_rate": 3.228571428571429e-05,
"loss": 0.236,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20021440088748932,
"step": 340,
"valid_targets_mean": 3883.1,
"valid_targets_min": 1514
},
{
"epoch": 0.5759599332220368,
"grad_norm": 0.5692629869123684,
"learning_rate": 3.276190476190477e-05,
"loss": 0.2293,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24576076865196228,
"step": 345,
"valid_targets_mean": 4225.1,
"valid_targets_min": 1489
},
{
"epoch": 0.5843071786310517,
"grad_norm": 0.9113146706371109,
"learning_rate": 3.3238095238095245e-05,
"loss": 0.2396,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17891007661819458,
"step": 350,
"valid_targets_mean": 3033.9,
"valid_targets_min": 1446
},
{
"epoch": 0.5926544240400667,
"grad_norm": 0.5495317395318315,
"learning_rate": 3.3714285714285716e-05,
"loss": 0.2054,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21951231360435486,
"step": 355,
"valid_targets_mean": 3824.1,
"valid_targets_min": 1983
},
{
"epoch": 0.6010016694490818,
"grad_norm": 0.6154540488434325,
"learning_rate": 3.4190476190476194e-05,
"loss": 0.2246,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21748147904872894,
"step": 360,
"valid_targets_mean": 3357.8,
"valid_targets_min": 1202
},
{
"epoch": 0.6093489148580968,
"grad_norm": 0.6778469089366,
"learning_rate": 3.466666666666667e-05,
"loss": 0.2244,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.176917165517807,
"step": 365,
"valid_targets_mean": 2836.6,
"valid_targets_min": 1432
},
{
"epoch": 0.6176961602671118,
"grad_norm": 0.4936120556487121,
"learning_rate": 3.514285714285714e-05,
"loss": 0.2257,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21924656629562378,
"step": 370,
"valid_targets_mean": 4530.2,
"valid_targets_min": 1428
},
{
"epoch": 0.6260434056761269,
"grad_norm": 0.6646613279624646,
"learning_rate": 3.561904761904762e-05,
"loss": 0.2467,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21542799472808838,
"step": 375,
"valid_targets_mean": 3270.3,
"valid_targets_min": 1428
},
{
"epoch": 0.6343906510851419,
"grad_norm": 0.6130429303265593,
"learning_rate": 3.60952380952381e-05,
"loss": 0.2081,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17093130946159363,
"step": 380,
"valid_targets_mean": 2974.3,
"valid_targets_min": 1383
},
{
"epoch": 0.6427378964941569,
"grad_norm": 0.5348286589553535,
"learning_rate": 3.6571428571428576e-05,
"loss": 0.2222,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2104872465133667,
"step": 385,
"valid_targets_mean": 3301.5,
"valid_targets_min": 1764
},
{
"epoch": 0.6510851419031719,
"grad_norm": 0.5327959335032387,
"learning_rate": 3.704761904761905e-05,
"loss": 0.2041,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19791337847709656,
"step": 390,
"valid_targets_mean": 4000.8,
"valid_targets_min": 1514
},
{
"epoch": 0.659432387312187,
"grad_norm": 0.5443518398719903,
"learning_rate": 3.7523809523809524e-05,
"loss": 0.2276,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18512150645256042,
"step": 395,
"valid_targets_mean": 3378.2,
"valid_targets_min": 1426
},
{
"epoch": 0.667779632721202,
"grad_norm": 0.5120896109446015,
"learning_rate": 3.8e-05,
"loss": 0.2394,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2590768337249756,
"step": 400,
"valid_targets_mean": 4461.8,
"valid_targets_min": 2380
},
{
"epoch": 0.676126878130217,
"grad_norm": 0.5328612149402605,
"learning_rate": 3.847619047619048e-05,
"loss": 0.2439,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21970009803771973,
"step": 405,
"valid_targets_mean": 4062.1,
"valid_targets_min": 1610
},
{
"epoch": 0.6844741235392321,
"grad_norm": 0.5544911293587242,
"learning_rate": 3.895238095238096e-05,
"loss": 0.2239,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18198083341121674,
"step": 410,
"valid_targets_mean": 3294.8,
"valid_targets_min": 1744
},
{
"epoch": 0.6928213689482471,
"grad_norm": 0.49880771464016166,
"learning_rate": 3.9428571428571435e-05,
"loss": 0.2089,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20763319730758667,
"step": 415,
"valid_targets_mean": 4266.3,
"valid_targets_min": 1425
},
{
"epoch": 0.7011686143572621,
"grad_norm": 0.5293891884748375,
"learning_rate": 3.9904761904761906e-05,
"loss": 0.2039,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17736777663230896,
"step": 420,
"valid_targets_mean": 4124.8,
"valid_targets_min": 1678
},
{
"epoch": 0.7095158597662772,
"grad_norm": 0.5180104218777899,
"learning_rate": 3.999988907084209e-05,
"loss": 0.2117,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18705101311206818,
"step": 425,
"valid_targets_mean": 3951.6,
"valid_targets_min": 1578
},
{
"epoch": 0.7178631051752922,
"grad_norm": 0.47591148563306207,
"learning_rate": 3.9999438423247035e-05,
"loss": 0.2301,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2004169225692749,
"step": 430,
"valid_targets_mean": 4376.9,
"valid_targets_min": 1455
},
{
"epoch": 0.7262103505843072,
"grad_norm": 0.5210833561106567,
"learning_rate": 3.999864113194738e-05,
"loss": 0.209,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20933908224105835,
"step": 435,
"valid_targets_mean": 4038.8,
"valid_targets_min": 2085
},
{
"epoch": 0.7345575959933222,
"grad_norm": 0.5225433946643954,
"learning_rate": 3.999749721076231e-05,
"loss": 0.2318,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22153371572494507,
"step": 440,
"valid_targets_mean": 4027.1,
"valid_targets_min": 1763
},
{
"epoch": 0.7429048414023373,
"grad_norm": 0.5234791164460791,
"learning_rate": 3.9996006679519054e-05,
"loss": 0.2019,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22157973051071167,
"step": 445,
"valid_targets_mean": 3944.8,
"valid_targets_min": 1439
},
{
"epoch": 0.7512520868113522,
"grad_norm": 0.4872218144147739,
"learning_rate": 3.9994169564052486e-05,
"loss": 0.2306,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21102100610733032,
"step": 450,
"valid_targets_mean": 4952.5,
"valid_targets_min": 2336
},
{
"epoch": 0.7595993322203672,
"grad_norm": 0.5010218195781705,
"learning_rate": 3.999198589620473e-05,
"loss": 0.2212,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24597254395484924,
"step": 455,
"valid_targets_mean": 4371.2,
"valid_targets_min": 1325
},
{
"epoch": 0.7679465776293823,
"grad_norm": 0.5383798379362373,
"learning_rate": 3.998945571382458e-05,
"loss": 0.2233,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20125095546245575,
"step": 460,
"valid_targets_mean": 4016.9,
"valid_targets_min": 1537
},
{
"epoch": 0.7762938230383973,
"grad_norm": 0.5050245208440932,
"learning_rate": 3.9986579060766866e-05,
"loss": 0.2208,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22594144940376282,
"step": 465,
"valid_targets_mean": 3609.8,
"valid_targets_min": 1809
},
{
"epoch": 0.7846410684474123,
"grad_norm": 0.5879419147770358,
"learning_rate": 3.9983355986891664e-05,
"loss": 0.192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19453193247318268,
"step": 470,
"valid_targets_mean": 3187.6,
"valid_targets_min": 1742
},
{
"epoch": 0.7929883138564274,
"grad_norm": 0.451217868223557,
"learning_rate": 3.9979786548063454e-05,
"loss": 0.2133,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21604406833648682,
"step": 475,
"valid_targets_mean": 4923.3,
"valid_targets_min": 1384
},
{
"epoch": 0.8013355592654424,
"grad_norm": 0.5388651431919386,
"learning_rate": 3.997587080615016e-05,
"loss": 0.2131,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1821407973766327,
"step": 480,
"valid_targets_mean": 3213.8,
"valid_targets_min": 1512
},
{
"epoch": 0.8096828046744574,
"grad_norm": 0.5456709602073617,
"learning_rate": 3.9971608829022036e-05,
"loss": 0.2074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22891031205654144,
"step": 485,
"valid_targets_mean": 3126.5,
"valid_targets_min": 1381
},
{
"epoch": 0.8180300500834724,
"grad_norm": 0.47054459318911074,
"learning_rate": 3.996700069055054e-05,
"loss": 0.229,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20265533030033112,
"step": 490,
"valid_targets_mean": 3801.8,
"valid_targets_min": 1506
},
{
"epoch": 0.8263772954924875,
"grad_norm": 0.47289778757291334,
"learning_rate": 3.9962046470607034e-05,
"loss": 0.2203,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2171851098537445,
"step": 495,
"valid_targets_mean": 3998.1,
"valid_targets_min": 1497
},
{
"epoch": 0.8347245409015025,
"grad_norm": 0.5261199604083402,
"learning_rate": 3.995674625506137e-05,
"loss": 0.2074,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22121337056159973,
"step": 500,
"valid_targets_mean": 3612.7,
"valid_targets_min": 1389
},
{
"epoch": 0.8430717863105175,
"grad_norm": 0.47799247311173443,
"learning_rate": 3.995110013578046e-05,
"loss": 0.1989,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18753355741500854,
"step": 505,
"valid_targets_mean": 3885.2,
"valid_targets_min": 2274
},
{
"epoch": 0.8514190317195326,
"grad_norm": 0.5594322261822097,
"learning_rate": 3.9945108210626635e-05,
"loss": 0.1943,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18138080835342407,
"step": 510,
"valid_targets_mean": 3085.6,
"valid_targets_min": 1530
},
{
"epoch": 0.8597662771285476,
"grad_norm": 0.49683446084717287,
"learning_rate": 3.9938770583455955e-05,
"loss": 0.1967,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18091800808906555,
"step": 515,
"valid_targets_mean": 3372.0,
"valid_targets_min": 1355
},
{
"epoch": 0.8681135225375626,
"grad_norm": 0.5303264522692365,
"learning_rate": 3.9932087364116446e-05,
"loss": 0.2028,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21672186255455017,
"step": 520,
"valid_targets_mean": 3241.2,
"valid_targets_min": 1632
},
{
"epoch": 0.8764607679465777,
"grad_norm": 0.4461510859408209,
"learning_rate": 3.992505866844615e-05,
"loss": 0.2171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1864897757768631,
"step": 525,
"valid_targets_mean": 4377.6,
"valid_targets_min": 2238
},
{
"epoch": 0.8848080133555927,
"grad_norm": 0.5485834333178916,
"learning_rate": 3.991768461827114e-05,
"loss": 0.2061,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23604360222816467,
"step": 530,
"valid_targets_mean": 3203.7,
"valid_targets_min": 1474
},
{
"epoch": 0.8931552587646077,
"grad_norm": 0.5849767322454961,
"learning_rate": 3.990996534140342e-05,
"loss": 0.2179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3368268609046936,
"step": 535,
"valid_targets_mean": 4832.6,
"valid_targets_min": 1440
},
{
"epoch": 0.9015025041736227,
"grad_norm": 0.44257486120238576,
"learning_rate": 3.990190097163867e-05,
"loss": 0.2073,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22880850732326508,
"step": 540,
"valid_targets_mean": 4451.9,
"valid_targets_min": 1710
},
{
"epoch": 0.9098497495826378,
"grad_norm": 0.4534765856781646,
"learning_rate": 3.989349164875397e-05,
"loss": 0.2211,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2534438371658325,
"step": 545,
"valid_targets_mean": 5441.2,
"valid_targets_min": 1907
},
{
"epoch": 0.9181969949916527,
"grad_norm": 0.5485303716789729,
"learning_rate": 3.988473751850536e-05,
"loss": 0.2186,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.3158420920372009,
"step": 550,
"valid_targets_mean": 4222.3,
"valid_targets_min": 1405
},
{
"epoch": 0.9265442404006677,
"grad_norm": 0.49962851467277347,
"learning_rate": 3.98756387326253e-05,
"loss": 0.2214,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25681281089782715,
"step": 555,
"valid_targets_mean": 4242.8,
"valid_targets_min": 605
},
{
"epoch": 0.9348914858096828,
"grad_norm": 0.43578816003157833,
"learning_rate": 3.9866195448820066e-05,
"loss": 0.2282,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25206494331359863,
"step": 560,
"valid_targets_mean": 5880.4,
"valid_targets_min": 1616
},
{
"epoch": 0.9432387312186978,
"grad_norm": 0.4612689276848329,
"learning_rate": 3.985640783076699e-05,
"loss": 0.2141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1919069141149521,
"step": 565,
"valid_targets_mean": 4366.2,
"valid_targets_min": 2134
},
{
"epoch": 0.9515859766277128,
"grad_norm": 0.5286196469332071,
"learning_rate": 3.984627604811166e-05,
"loss": 0.2055,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19821152091026306,
"step": 570,
"valid_targets_mean": 3329.8,
"valid_targets_min": 1644
},
{
"epoch": 0.9599332220367279,
"grad_norm": 0.4995461496470502,
"learning_rate": 3.983580027646492e-05,
"loss": 0.228,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23345303535461426,
"step": 575,
"valid_targets_mean": 3977.6,
"valid_targets_min": 871
},
{
"epoch": 0.9682804674457429,
"grad_norm": 0.4670087324716703,
"learning_rate": 3.9824980697399906e-05,
"loss": 0.2072,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2013092339038849,
"step": 580,
"valid_targets_mean": 4316.9,
"valid_targets_min": 1429
},
{
"epoch": 0.9766277128547579,
"grad_norm": 0.56734105404264,
"learning_rate": 3.981381749844882e-05,
"loss": 0.2113,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1934119313955307,
"step": 585,
"valid_targets_mean": 2784.1,
"valid_targets_min": 1405
},
{
"epoch": 0.9849749582637729,
"grad_norm": 0.4442718818171178,
"learning_rate": 3.980231087309971e-05,
"loss": 0.2065,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22155402600765228,
"step": 590,
"valid_targets_mean": 4288.9,
"valid_targets_min": 1524
},
{
"epoch": 0.993322203672788,
"grad_norm": 0.45421640540205843,
"learning_rate": 3.9790461020793166e-05,
"loss": 0.2151,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24775730073451996,
"step": 595,
"valid_targets_mean": 4101.6,
"valid_targets_min": 1342
},
{
"epoch": 1.001669449081803,
"grad_norm": 0.5417397437914816,
"learning_rate": 3.977826814691878e-05,
"loss": 0.2174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17313963174819946,
"step": 600,
"valid_targets_mean": 2918.8,
"valid_targets_min": 1393
},
{
"epoch": 1.010016694490818,
"grad_norm": 0.5265647737167619,
"learning_rate": 3.9765732462811625e-05,
"loss": 0.2126,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16617675125598907,
"step": 605,
"valid_targets_mean": 3102.7,
"valid_targets_min": 1049
},
{
"epoch": 1.018363939899833,
"grad_norm": 0.4616129084379421,
"learning_rate": 3.975285418574862e-05,
"loss": 0.2164,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1739726960659027,
"step": 610,
"valid_targets_mean": 3422.4,
"valid_targets_min": 1423
},
{
"epoch": 1.026711185308848,
"grad_norm": 0.5302491723498224,
"learning_rate": 3.97396335389447e-05,
"loss": 0.1917,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19165921211242676,
"step": 615,
"valid_targets_mean": 4196.4,
"valid_targets_min": 1436
},
{
"epoch": 1.0350584307178632,
"grad_norm": 0.4370541767471563,
"learning_rate": 3.972607075154901e-05,
"loss": 0.1905,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16564807295799255,
"step": 620,
"valid_targets_mean": 3967.0,
"valid_targets_min": 1355
},
{
"epoch": 1.0434056761268782,
"grad_norm": 0.43481970536197584,
"learning_rate": 3.971216605864087e-05,
"loss": 0.2164,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1917448341846466,
"step": 625,
"valid_targets_mean": 3972.4,
"valid_targets_min": 1526
},
{
"epoch": 1.0517529215358932,
"grad_norm": 0.4857624966767501,
"learning_rate": 3.969791970122579e-05,
"loss": 0.2225,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26058006286621094,
"step": 630,
"valid_targets_mean": 4080.7,
"valid_targets_min": 1606
},
{
"epoch": 1.0601001669449082,
"grad_norm": 0.39704839973046024,
"learning_rate": 3.96833319262312e-05,
"loss": 0.189,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1686481535434723,
"step": 635,
"valid_targets_mean": 4256.9,
"valid_targets_min": 1965
},
{
"epoch": 1.0684474123539232,
"grad_norm": 0.4452832494125616,
"learning_rate": 3.9668402986502214e-05,
"loss": 0.2101,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19080159068107605,
"step": 640,
"valid_targets_mean": 4267.8,
"valid_targets_min": 1580
},
{
"epoch": 1.0767946577629381,
"grad_norm": 0.4379546845492667,
"learning_rate": 3.9653133140797244e-05,
"loss": 0.1906,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18801546096801758,
"step": 645,
"valid_targets_mean": 3967.2,
"valid_targets_min": 1539
},
{
"epoch": 1.0851419031719534,
"grad_norm": 0.694824763186105,
"learning_rate": 3.963752265378352e-05,
"loss": 0.2259,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27527111768722534,
"step": 650,
"valid_targets_mean": 3967.6,
"valid_targets_min": 1458
},
{
"epoch": 1.0934891485809684,
"grad_norm": 0.4045268244592241,
"learning_rate": 3.962157179603249e-05,
"loss": 0.1934,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1786402463912964,
"step": 655,
"valid_targets_mean": 4905.0,
"valid_targets_min": 1376
},
{
"epoch": 1.1018363939899833,
"grad_norm": 0.51216574637076,
"learning_rate": 3.960528084401515e-05,
"loss": 0.2085,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27255719900131226,
"step": 660,
"valid_targets_mean": 4640.4,
"valid_targets_min": 1611
},
{
"epoch": 1.1101836393989983,
"grad_norm": 0.42269547903417154,
"learning_rate": 3.9588650080097196e-05,
"loss": 0.201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18487635254859924,
"step": 665,
"valid_targets_mean": 4188.8,
"valid_targets_min": 1411
},
{
"epoch": 1.1185308848080133,
"grad_norm": 0.5915571898115403,
"learning_rate": 3.957167979253424e-05,
"loss": 0.1928,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20450526475906372,
"step": 670,
"valid_targets_mean": 4005.5,
"valid_targets_min": 770
},
{
"epoch": 1.1268781302170283,
"grad_norm": 0.4838049120852178,
"learning_rate": 3.955437027546668e-05,
"loss": 0.2087,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24886548519134521,
"step": 675,
"valid_targets_mean": 4866.2,
"valid_targets_min": 1841
},
{
"epoch": 1.1352253756260433,
"grad_norm": 0.4161458941865741,
"learning_rate": 3.953672182891471e-05,
"loss": 0.2049,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1465485543012619,
"step": 680,
"valid_targets_mean": 4132.9,
"valid_targets_min": 1410
},
{
"epoch": 1.1435726210350585,
"grad_norm": 0.4843909073732772,
"learning_rate": 3.951873475877306e-05,
"loss": 0.1725,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14315181970596313,
"step": 685,
"valid_targets_mean": 2793.9,
"valid_targets_min": 1481
},
{
"epoch": 1.1519198664440735,
"grad_norm": 0.46350239762356077,
"learning_rate": 3.950040937680572e-05,
"loss": 0.2312,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23201894760131836,
"step": 690,
"valid_targets_mean": 4414.2,
"valid_targets_min": 1627
},
{
"epoch": 1.1602671118530885,
"grad_norm": 0.4328936699179332,
"learning_rate": 3.948174600064051e-05,
"loss": 0.1935,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22564777731895447,
"step": 695,
"valid_targets_mean": 4178.4,
"valid_targets_min": 2745
},
{
"epoch": 1.1686143572621035,
"grad_norm": 0.503147138264778,
"learning_rate": 3.946274495376362e-05,
"loss": 0.1786,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17849397659301758,
"step": 700,
"valid_targets_mean": 3454.8,
"valid_targets_min": 1365
},
{
"epoch": 1.1769616026711185,
"grad_norm": 0.42838666269817965,
"learning_rate": 3.9443406565513963e-05,
"loss": 0.2095,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18975511193275452,
"step": 705,
"valid_targets_mean": 4442.3,
"valid_targets_min": 1525
},
{
"epoch": 1.1853088480801335,
"grad_norm": 0.47176982351497787,
"learning_rate": 3.9423731171077465e-05,
"loss": 0.2142,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2033671885728836,
"step": 710,
"valid_targets_mean": 3734.4,
"valid_targets_min": 1405
},
{
"epoch": 1.1936560934891487,
"grad_norm": 0.3877259491244469,
"learning_rate": 3.9403719111481295e-05,
"loss": 0.2016,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19912946224212646,
"step": 715,
"valid_targets_mean": 4583.1,
"valid_targets_min": 2176
},
{
"epoch": 1.2020033388981637,
"grad_norm": 0.4069281118281998,
"learning_rate": 3.9383370733587905e-05,
"loss": 0.2076,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15315377712249756,
"step": 720,
"valid_targets_mean": 4168.5,
"valid_targets_min": 1442
},
{
"epoch": 1.2103505843071787,
"grad_norm": 0.42008543867930265,
"learning_rate": 3.936268639008906e-05,
"loss": 0.21,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15432016551494598,
"step": 725,
"valid_targets_mean": 4381.8,
"valid_targets_min": 2250
},
{
"epoch": 1.2186978297161937,
"grad_norm": 0.4825088942884119,
"learning_rate": 3.93416664394997e-05,
"loss": 0.201,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1441294252872467,
"step": 730,
"valid_targets_mean": 3178.2,
"valid_targets_min": 1324
},
{
"epoch": 1.2270450751252087,
"grad_norm": 0.4052908805889826,
"learning_rate": 3.932031124615172e-05,
"loss": 0.1904,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16092966496944427,
"step": 735,
"valid_targets_mean": 4391.7,
"valid_targets_min": 1442
},
{
"epoch": 1.2353923205342237,
"grad_norm": 0.4640360873735371,
"learning_rate": 3.92986211801877e-05,
"loss": 0.1863,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15035538375377655,
"step": 740,
"valid_targets_mean": 3316.4,
"valid_targets_min": 1390
},
{
"epoch": 1.2437395659432386,
"grad_norm": 0.4303584168757166,
"learning_rate": 3.927659661755442e-05,
"loss": 0.2092,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20159313082695007,
"step": 745,
"valid_targets_mean": 4263.1,
"valid_targets_min": 1746
},
{
"epoch": 1.2520868113522536,
"grad_norm": 0.46816222877220925,
"learning_rate": 3.925423793999641e-05,
"loss": 0.192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1846737116575241,
"step": 750,
"valid_targets_mean": 3817.5,
"valid_targets_min": 1793
},
{
"epoch": 1.2604340567612689,
"grad_norm": 0.5021475200575083,
"learning_rate": 3.923154553504929e-05,
"loss": 0.2114,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19226914644241333,
"step": 755,
"valid_targets_mean": 3526.4,
"valid_targets_min": 1584
},
{
"epoch": 1.2687813021702838,
"grad_norm": 0.45466743231077655,
"learning_rate": 3.920851979603306e-05,
"loss": 0.1875,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1986933946609497,
"step": 760,
"valid_targets_mean": 3466.7,
"valid_targets_min": 652
},
{
"epoch": 1.2771285475792988,
"grad_norm": 0.4583161241975255,
"learning_rate": 3.918516112204532e-05,
"loss": 0.1841,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2409411072731018,
"step": 765,
"valid_targets_mean": 4356.6,
"valid_targets_min": 1456
},
{
"epoch": 1.2854757929883138,
"grad_norm": 0.4399138828252771,
"learning_rate": 3.9161469917954273e-05,
"loss": 0.2004,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24284087121486664,
"step": 770,
"valid_targets_mean": 4298.3,
"valid_targets_min": 1903
},
{
"epoch": 1.2938230383973288,
"grad_norm": 0.4294986564270851,
"learning_rate": 3.913744659439181e-05,
"loss": 0.1902,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17483387887477875,
"step": 775,
"valid_targets_mean": 3675.4,
"valid_targets_min": 1588
},
{
"epoch": 1.302170283806344,
"grad_norm": 0.41352563034741097,
"learning_rate": 3.911309156774631e-05,
"loss": 0.1678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15267729759216309,
"step": 780,
"valid_targets_mean": 3254.7,
"valid_targets_min": 1354
},
{
"epoch": 1.310517529215359,
"grad_norm": 0.4262349832370848,
"learning_rate": 3.908840526015547e-05,
"loss": 0.2013,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20099084079265594,
"step": 785,
"valid_targets_mean": 4173.8,
"valid_targets_min": 2087
},
{
"epoch": 1.318864774624374,
"grad_norm": 0.4484784124181742,
"learning_rate": 3.906338809949893e-05,
"loss": 0.1979,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1672644317150116,
"step": 790,
"valid_targets_mean": 3427.0,
"valid_targets_min": 1883
},
{
"epoch": 1.327212020033389,
"grad_norm": 0.4379536871989578,
"learning_rate": 3.903804051939096e-05,
"loss": 0.2312,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22611889243125916,
"step": 795,
"valid_targets_mean": 3723.9,
"valid_targets_min": 1709
},
{
"epoch": 1.335559265442404,
"grad_norm": 0.40745223109878176,
"learning_rate": 3.9012362959172834e-05,
"loss": 0.1828,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20771396160125732,
"step": 800,
"valid_targets_mean": 4382.2,
"valid_targets_min": 1890
},
{
"epoch": 1.343906510851419,
"grad_norm": 0.431604165628237,
"learning_rate": 3.898635586390528e-05,
"loss": 0.185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16094966232776642,
"step": 805,
"valid_targets_mean": 3297.5,
"valid_targets_min": 1502
},
{
"epoch": 1.352253756260434,
"grad_norm": 0.4478646744840501,
"learning_rate": 3.8960019684360756e-05,
"loss": 0.1986,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22003845870494843,
"step": 810,
"valid_targets_mean": 4356.4,
"valid_targets_min": 1506
},
{
"epoch": 1.360601001669449,
"grad_norm": 0.4694556113932446,
"learning_rate": 3.8933354877015606e-05,
"loss": 0.2059,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21767646074295044,
"step": 815,
"valid_targets_mean": 3803.9,
"valid_targets_min": 1309
},
{
"epoch": 1.3689482470784642,
"grad_norm": 0.45512175600379035,
"learning_rate": 3.89063619040422e-05,
"loss": 0.1866,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20812076330184937,
"step": 820,
"valid_targets_mean": 4155.0,
"valid_targets_min": 1978
},
{
"epoch": 1.3772954924874792,
"grad_norm": 0.45611004239381786,
"learning_rate": 3.887904123330088e-05,
"loss": 0.1933,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17094948887825012,
"step": 825,
"valid_targets_mean": 3623.5,
"valid_targets_min": 1514
},
{
"epoch": 1.3856427378964942,
"grad_norm": 0.4414337853435407,
"learning_rate": 3.885139333833186e-05,
"loss": 0.1895,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19573955237865448,
"step": 830,
"valid_targets_mean": 3552.4,
"valid_targets_min": 1470
},
{
"epoch": 1.3939899833055092,
"grad_norm": 0.40273873690281814,
"learning_rate": 3.882341869834704e-05,
"loss": 0.1951,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18730872869491577,
"step": 835,
"valid_targets_mean": 4221.3,
"valid_targets_min": 2090
},
{
"epoch": 1.4023372287145242,
"grad_norm": 0.47324350221570133,
"learning_rate": 3.879511779822168e-05,
"loss": 0.2362,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2657076120376587,
"step": 840,
"valid_targets_mean": 3906.9,
"valid_targets_min": 1315
},
{
"epoch": 1.4106844741235394,
"grad_norm": 0.46159508203133287,
"learning_rate": 3.8766491128486e-05,
"loss": 0.1876,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21945779025554657,
"step": 845,
"valid_targets_mean": 4024.8,
"valid_targets_min": 1283
},
{
"epoch": 1.4190317195325544,
"grad_norm": 0.3996281241564718,
"learning_rate": 3.873753918531666e-05,
"loss": 0.2111,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2445368617773056,
"step": 850,
"valid_targets_mean": 5171.8,
"valid_targets_min": 1428
},
{
"epoch": 1.4273789649415694,
"grad_norm": 0.37118198743732866,
"learning_rate": 3.87082624705282e-05,
"loss": 0.1908,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17855234444141388,
"step": 855,
"valid_targets_mean": 5047.5,
"valid_targets_min": 1839
},
{
"epoch": 1.4357262103505843,
"grad_norm": 0.4766525343937952,
"learning_rate": 3.867866149156431e-05,
"loss": 0.2034,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20250694453716278,
"step": 860,
"valid_targets_mean": 4213.7,
"valid_targets_min": 2085
},
{
"epoch": 1.4440734557595993,
"grad_norm": 0.42205764465522955,
"learning_rate": 3.8648736761489043e-05,
"loss": 0.1754,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19850978255271912,
"step": 865,
"valid_targets_mean": 3934.6,
"valid_targets_min": 1362
},
{
"epoch": 1.4524207011686143,
"grad_norm": 0.3991661089811781,
"learning_rate": 3.861848879897794e-05,
"loss": 0.1815,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15859410166740417,
"step": 870,
"valid_targets_mean": 4325.4,
"valid_targets_min": 1671
},
{
"epoch": 1.4607679465776293,
"grad_norm": 0.411444665987959,
"learning_rate": 3.8587918128309e-05,
"loss": 0.1929,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19219332933425903,
"step": 875,
"valid_targets_mean": 4361.6,
"valid_targets_min": 1577
},
{
"epoch": 1.4691151919866443,
"grad_norm": 0.4221345200846809,
"learning_rate": 3.855702527935363e-05,
"loss": 0.1872,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17698226869106293,
"step": 880,
"valid_targets_mean": 3471.6,
"valid_targets_min": 1427
},
{
"epoch": 1.4774624373956593,
"grad_norm": 0.4527047348160603,
"learning_rate": 3.852581078756745e-05,
"loss": 0.2088,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2722078859806061,
"step": 885,
"valid_targets_mean": 4729.5,
"valid_targets_min": 1492
},
{
"epoch": 1.4858096828046745,
"grad_norm": 0.5330536154314943,
"learning_rate": 3.8494275193981e-05,
"loss": 0.2152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20743849873542786,
"step": 890,
"valid_targets_mean": 3607.9,
"valid_targets_min": 1982
},
{
"epoch": 1.4941569282136895,
"grad_norm": 0.4176999838658574,
"learning_rate": 3.8462419045190385e-05,
"loss": 0.1937,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2070760428905487,
"step": 895,
"valid_targets_mean": 4395.5,
"valid_targets_min": 1408
},
{
"epoch": 1.5025041736227045,
"grad_norm": 0.4064635644346735,
"learning_rate": 3.8430242893347765e-05,
"loss": 0.1869,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21248620748519897,
"step": 900,
"valid_targets_mean": 4216.4,
"valid_targets_min": 1376
},
{
"epoch": 1.5108514190317195,
"grad_norm": 0.44297335398255444,
"learning_rate": 3.839774729615184e-05,
"loss": 0.2024,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22493979334831238,
"step": 905,
"valid_targets_mean": 4159.1,
"valid_targets_min": 1503
},
{
"epoch": 1.5191986644407347,
"grad_norm": 0.41285289852226464,
"learning_rate": 3.8364932816838124e-05,
"loss": 0.1898,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16580085456371307,
"step": 910,
"valid_targets_mean": 4131.9,
"valid_targets_min": 1913
},
{
"epoch": 1.5275459098497497,
"grad_norm": 0.4352009341698042,
"learning_rate": 3.833180002416922e-05,
"loss": 0.2153,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23203584551811218,
"step": 915,
"valid_targets_mean": 4068.4,
"valid_targets_min": 1944
},
{
"epoch": 1.5358931552587647,
"grad_norm": 0.40397446961320027,
"learning_rate": 3.829834949242496e-05,
"loss": 0.1986,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15464527904987335,
"step": 920,
"valid_targets_mean": 3469.0,
"valid_targets_min": 1494
},
{
"epoch": 1.5442404006677797,
"grad_norm": 0.4618545420295146,
"learning_rate": 3.826458180139244e-05,
"loss": 0.217,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22266030311584473,
"step": 925,
"valid_targets_mean": 3303.1,
"valid_targets_min": 1426
},
{
"epoch": 1.5525876460767947,
"grad_norm": 0.40123900660820766,
"learning_rate": 3.823049753635596e-05,
"loss": 0.1754,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20035436749458313,
"step": 930,
"valid_targets_mean": 4314.1,
"valid_targets_min": 1339
},
{
"epoch": 1.5609348914858097,
"grad_norm": 0.4148259020417995,
"learning_rate": 3.81960972880869e-05,
"loss": 0.1735,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21289658546447754,
"step": 935,
"valid_targets_mean": 4126.2,
"valid_targets_min": 1833
},
{
"epoch": 1.5692821368948247,
"grad_norm": 0.39428417541635025,
"learning_rate": 3.8161381652833485e-05,
"loss": 0.1969,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21160581707954407,
"step": 940,
"valid_targets_mean": 4396.8,
"valid_targets_min": 2065
},
{
"epoch": 1.5776293823038396,
"grad_norm": 0.7722286738595375,
"learning_rate": 3.812635123231043e-05,
"loss": 0.1968,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20537340641021729,
"step": 945,
"valid_targets_mean": 3099.8,
"valid_targets_min": 1707
},
{
"epoch": 1.5859766277128546,
"grad_norm": 0.38050878204221195,
"learning_rate": 3.809100663368852e-05,
"loss": 0.183,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.162803053855896,
"step": 950,
"valid_targets_mean": 3694.8,
"valid_targets_min": 1593
},
{
"epoch": 1.5943238731218696,
"grad_norm": 0.40630472428747527,
"learning_rate": 3.805534846958408e-05,
"loss": 0.1727,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16229110956192017,
"step": 955,
"valid_targets_mean": 3921.1,
"valid_targets_min": 1363
},
{
"epoch": 1.6026711185308848,
"grad_norm": 0.39732456299196406,
"learning_rate": 3.801937735804838e-05,
"loss": 0.1843,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19792838394641876,
"step": 960,
"valid_targets_mean": 4172.2,
"valid_targets_min": 1649
},
{
"epoch": 1.6110183639398998,
"grad_norm": 0.44305148074660433,
"learning_rate": 3.79830939225569e-05,
"loss": 0.1737,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18473561108112335,
"step": 965,
"valid_targets_mean": 3244.9,
"valid_targets_min": 1464
},
{
"epoch": 1.6193656093489148,
"grad_norm": 0.39731672602454104,
"learning_rate": 3.79464987919985e-05,
"loss": 0.1904,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16850672662258148,
"step": 970,
"valid_targets_mean": 3465.9,
"valid_targets_min": 1362
},
{
"epoch": 1.62771285475793,
"grad_norm": 0.41964357597174584,
"learning_rate": 3.790959260066459e-05,
"loss": 0.1978,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17484863102436066,
"step": 975,
"valid_targets_mean": 3516.7,
"valid_targets_min": 1426
},
{
"epoch": 1.636060100166945,
"grad_norm": 0.396140280354335,
"learning_rate": 3.7872375988238075e-05,
"loss": 0.2182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1687312126159668,
"step": 980,
"valid_targets_mean": 3835.4,
"valid_targets_min": 1473
},
{
"epoch": 1.64440734557596,
"grad_norm": 0.43798134559400614,
"learning_rate": 3.783484959978228e-05,
"loss": 0.176,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18557807803153992,
"step": 985,
"valid_targets_mean": 3346.4,
"valid_targets_min": 1424
},
{
"epoch": 1.652754590984975,
"grad_norm": 0.4132221123654361,
"learning_rate": 3.7797014085729786e-05,
"loss": 0.1928,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24561835825443268,
"step": 990,
"valid_targets_mean": 4644.3,
"valid_targets_min": 1482
},
{
"epoch": 1.66110183639399,
"grad_norm": 0.4341492003334459,
"learning_rate": 3.7758870101871155e-05,
"loss": 0.187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14808312058448792,
"step": 995,
"valid_targets_mean": 2953.4,
"valid_targets_min": 1342
},
{
"epoch": 1.669449081803005,
"grad_norm": 0.4320784433206574,
"learning_rate": 3.772041830934352e-05,
"loss": 0.2159,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2128472626209259,
"step": 1000,
"valid_targets_mean": 5073.5,
"valid_targets_min": 1473
},
{
"epoch": 1.67779632721202,
"grad_norm": 0.40999065212676533,
"learning_rate": 3.7681659374619216e-05,
"loss": 0.2037,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26891839504241943,
"step": 1005,
"valid_targets_mean": 5264.4,
"valid_targets_min": 1554
},
{
"epoch": 1.686143572621035,
"grad_norm": 0.42264969189026164,
"learning_rate": 3.764259396949412e-05,
"loss": 0.1876,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20195063948631287,
"step": 1010,
"valid_targets_mean": 4184.3,
"valid_targets_min": 544
},
{
"epoch": 1.69449081803005,
"grad_norm": 0.46598281446435735,
"learning_rate": 3.7603222771076094e-05,
"loss": 0.2093,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17481181025505066,
"step": 1015,
"valid_targets_mean": 3585.2,
"valid_targets_min": 1901
},
{
"epoch": 1.702838063439065,
"grad_norm": 0.46915273916548544,
"learning_rate": 3.75635464617732e-05,
"loss": 0.199,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15742053091526031,
"step": 1020,
"valid_targets_mean": 3064.8,
"valid_targets_min": 1384
},
{
"epoch": 1.7111853088480802,
"grad_norm": 0.8067627739457696,
"learning_rate": 3.752356572928189e-05,
"loss": 0.2114,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17468111217021942,
"step": 1025,
"valid_targets_mean": 4547.9,
"valid_targets_min": 1349
},
{
"epoch": 1.7195325542570952,
"grad_norm": 0.4146011487610528,
"learning_rate": 3.748328126657508e-05,
"loss": 0.1724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19408324360847473,
"step": 1030,
"valid_targets_mean": 3705.2,
"valid_targets_min": 1516
},
{
"epoch": 1.7278797996661102,
"grad_norm": 0.40139276329876356,
"learning_rate": 3.7442693771890134e-05,
"loss": 0.2099,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20316094160079956,
"step": 1035,
"valid_targets_mean": 4225.6,
"valid_targets_min": 1414
},
{
"epoch": 1.7362270450751254,
"grad_norm": 0.43873505249612726,
"learning_rate": 3.740180394871681e-05,
"loss": 0.1905,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15976247191429138,
"step": 1040,
"valid_targets_mean": 3111.8,
"valid_targets_min": 1625
},
{
"epoch": 1.7445742904841404,
"grad_norm": 0.4448279397403342,
"learning_rate": 3.736061250578498e-05,
"loss": 0.1921,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21715202927589417,
"step": 1045,
"valid_targets_mean": 4017.9,
"valid_targets_min": 1430
},
{
"epoch": 1.7529215358931554,
"grad_norm": 0.3836356266617607,
"learning_rate": 3.7319120157052405e-05,
"loss": 0.1851,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18681854009628296,
"step": 1050,
"valid_targets_mean": 4413.0,
"valid_targets_min": 1582
},
{
"epoch": 1.7612687813021703,
"grad_norm": 0.4289127966268711,
"learning_rate": 3.727732762169236e-05,
"loss": 0.1977,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19718758761882782,
"step": 1055,
"valid_targets_mean": 3780.9,
"valid_targets_min": 1500
},
{
"epoch": 1.7696160267111853,
"grad_norm": 0.4144506736772134,
"learning_rate": 3.7235235624081144e-05,
"loss": 0.1767,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19311445951461792,
"step": 1060,
"valid_targets_mean": 3478.4,
"valid_targets_min": 1611
},
{
"epoch": 1.7779632721202003,
"grad_norm": 0.40901113325958294,
"learning_rate": 3.719284489378555e-05,
"loss": 0.1741,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14171263575553894,
"step": 1065,
"valid_targets_mean": 3169.3,
"valid_targets_min": 1455
},
{
"epoch": 1.7863105175292153,
"grad_norm": 0.4360232655047189,
"learning_rate": 3.715015616555019e-05,
"loss": 0.1651,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18953683972358704,
"step": 1070,
"valid_targets_mean": 3859.8,
"valid_targets_min": 1900
},
{
"epoch": 1.7946577629382303,
"grad_norm": 0.3982856347013278,
"learning_rate": 3.710717017928479e-05,
"loss": 0.2181,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23961114883422852,
"step": 1075,
"valid_targets_mean": 4795.3,
"valid_targets_min": 1427
},
{
"epoch": 1.8030050083472453,
"grad_norm": 0.43447421142062564,
"learning_rate": 3.7063887680051346e-05,
"loss": 0.1785,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17534184455871582,
"step": 1080,
"valid_targets_mean": 3499.0,
"valid_targets_min": 1467
},
{
"epoch": 1.8113522537562603,
"grad_norm": 0.37484687016486634,
"learning_rate": 3.702030941805122e-05,
"loss": 0.2043,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22565153241157532,
"step": 1085,
"valid_targets_mean": 4968.5,
"valid_targets_min": 1645
},
{
"epoch": 1.8196994991652755,
"grad_norm": 0.4253620491937088,
"learning_rate": 3.697643614861212e-05,
"loss": 0.1935,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24910342693328857,
"step": 1090,
"valid_targets_mean": 4526.3,
"valid_targets_min": 1421
},
{
"epoch": 1.8280467445742905,
"grad_norm": 0.3876416101052467,
"learning_rate": 3.6932268632175035e-05,
"loss": 0.1845,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1603732705116272,
"step": 1095,
"valid_targets_mean": 3542.6,
"valid_targets_min": 1388
},
{
"epoch": 1.8363939899833055,
"grad_norm": 0.41008061749501296,
"learning_rate": 3.6887807634281034e-05,
"loss": 0.2015,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19442768394947052,
"step": 1100,
"valid_targets_mean": 3616.4,
"valid_targets_min": 1362
},
{
"epoch": 1.8447412353923205,
"grad_norm": 0.38250220147931985,
"learning_rate": 3.684305392555802e-05,
"loss": 0.1864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16925008594989777,
"step": 1105,
"valid_targets_mean": 4450.6,
"valid_targets_min": 1438
},
{
"epoch": 1.8530884808013357,
"grad_norm": 0.4516520147121603,
"learning_rate": 3.6798008281707326e-05,
"loss": 0.1959,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1806880235671997,
"step": 1110,
"valid_targets_mean": 3044.1,
"valid_targets_min": 1402
},
{
"epoch": 1.8614357262103507,
"grad_norm": 0.43415677726665036,
"learning_rate": 3.675267148349033e-05,
"loss": 0.1873,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.190780907869339,
"step": 1115,
"valid_targets_mean": 3481.9,
"valid_targets_min": 1335
},
{
"epoch": 1.8697829716193657,
"grad_norm": 0.3852595590934276,
"learning_rate": 3.670704431671487e-05,
"loss": 0.2136,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2023189216852188,
"step": 1120,
"valid_targets_mean": 4188.6,
"valid_targets_min": 1315
},
{
"epoch": 1.8781302170283807,
"grad_norm": 0.40220713573380873,
"learning_rate": 3.6661127572221674e-05,
"loss": 0.1738,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14815208315849304,
"step": 1125,
"valid_targets_mean": 3508.1,
"valid_targets_min": 1427
},
{
"epoch": 1.8864774624373957,
"grad_norm": 0.3736795337861637,
"learning_rate": 3.661492204587059e-05,
"loss": 0.1874,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20401643216609955,
"step": 1130,
"valid_targets_mean": 5319.5,
"valid_targets_min": 2498
},
{
"epoch": 1.8948247078464107,
"grad_norm": 0.41354931322766325,
"learning_rate": 3.656842853852686e-05,
"loss": 0.1935,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22401690483093262,
"step": 1135,
"valid_targets_mean": 3935.3,
"valid_targets_min": 574
},
{
"epoch": 1.9031719532554257,
"grad_norm": 0.3754287464972552,
"learning_rate": 3.652164785604718e-05,
"loss": 0.2039,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22014948725700378,
"step": 1140,
"valid_targets_mean": 4917.1,
"valid_targets_min": 2484
},
{
"epoch": 1.9115191986644406,
"grad_norm": 0.3771865838798427,
"learning_rate": 3.647458080926579e-05,
"loss": 0.1939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2064426839351654,
"step": 1145,
"valid_targets_mean": 4613.6,
"valid_targets_min": 1531
},
{
"epoch": 1.9198664440734556,
"grad_norm": 0.4650972573565242,
"learning_rate": 3.642722821398036e-05,
"loss": 0.1745,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1807226538658142,
"step": 1150,
"valid_targets_mean": 3832.2,
"valid_targets_min": 1244
},
{
"epoch": 1.9282136894824706,
"grad_norm": 0.36140384362617156,
"learning_rate": 3.637959089093788e-05,
"loss": 0.1947,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17937152087688446,
"step": 1155,
"valid_targets_mean": 4637.2,
"valid_targets_min": 1426
},
{
"epoch": 1.9365609348914858,
"grad_norm": 0.382524851338302,
"learning_rate": 3.633166966582046e-05,
"loss": 0.1878,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17297901213169098,
"step": 1160,
"valid_targets_mean": 4146.6,
"valid_targets_min": 1352
},
{
"epoch": 1.9449081803005008,
"grad_norm": 0.40635470177143157,
"learning_rate": 3.6283465369230955e-05,
"loss": 0.1884,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22332972288131714,
"step": 1165,
"valid_targets_mean": 4250.2,
"valid_targets_min": 1775
},
{
"epoch": 1.9532554257095158,
"grad_norm": 0.43500848174626067,
"learning_rate": 3.6234978836678635e-05,
"loss": 0.2,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18507716059684753,
"step": 1170,
"valid_targets_mean": 3394.8,
"valid_targets_min": 1515
},
{
"epoch": 1.961602671118531,
"grad_norm": 0.3982834031952082,
"learning_rate": 3.618621090856465e-05,
"loss": 0.2026,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21530982851982117,
"step": 1175,
"valid_targets_mean": 4329.9,
"valid_targets_min": 1408
},
{
"epoch": 1.969949916527546,
"grad_norm": 0.36236979650290985,
"learning_rate": 3.6137162430167505e-05,
"loss": 0.1846,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15241621434688568,
"step": 1180,
"valid_targets_mean": 3629.4,
"valid_targets_min": 1497
},
{
"epoch": 1.978297161936561,
"grad_norm": 0.4130597977119614,
"learning_rate": 3.608783425162837e-05,
"loss": 0.1724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1333189308643341,
"step": 1185,
"valid_targets_mean": 2981.4,
"valid_targets_min": 1638
},
{
"epoch": 1.986644407345576,
"grad_norm": 0.4335053496330714,
"learning_rate": 3.603822722793641e-05,
"loss": 0.1802,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15315896272659302,
"step": 1190,
"valid_targets_mean": 2883.6,
"valid_targets_min": 1475
},
{
"epoch": 1.994991652754591,
"grad_norm": 0.4009985561626319,
"learning_rate": 3.598834221891386e-05,
"loss": 0.1958,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1875220239162445,
"step": 1195,
"valid_targets_mean": 4029.7,
"valid_targets_min": 1521
},
{
"epoch": 2.003338898163606,
"grad_norm": 0.3609913095429806,
"learning_rate": 3.5938180089201236e-05,
"loss": 0.1813,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18564942479133606,
"step": 1200,
"valid_targets_mean": 4209.9,
"valid_targets_min": 1363
},
{
"epoch": 2.011686143572621,
"grad_norm": 0.402444441098984,
"learning_rate": 3.588774170824225e-05,
"loss": 0.1961,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23691844940185547,
"step": 1205,
"valid_targets_mean": 4869.6,
"valid_targets_min": 1925
},
{
"epoch": 2.020033388981636,
"grad_norm": 0.39050077664902183,
"learning_rate": 3.5837027950268845e-05,
"loss": 0.1751,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19038726389408112,
"step": 1210,
"valid_targets_mean": 3901.4,
"valid_targets_min": 1891
},
{
"epoch": 2.028380634390651,
"grad_norm": 0.39314437944015984,
"learning_rate": 3.578603969428592e-05,
"loss": 0.192,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18340542912483215,
"step": 1215,
"valid_targets_mean": 4314.6,
"valid_targets_min": 1509
},
{
"epoch": 2.036727879799666,
"grad_norm": 0.3931119098050458,
"learning_rate": 3.57347778240562e-05,
"loss": 0.1786,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17078736424446106,
"step": 1220,
"valid_targets_mean": 4287.2,
"valid_targets_min": 1914
},
{
"epoch": 2.045075125208681,
"grad_norm": 0.3743482481969501,
"learning_rate": 3.568324322808486e-05,
"loss": 0.1883,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14049555361270905,
"step": 1225,
"valid_targets_mean": 4027.3,
"valid_targets_min": 1611
},
{
"epoch": 2.053422370617696,
"grad_norm": 0.4013199068276139,
"learning_rate": 3.563143679960415e-05,
"loss": 0.1845,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1572129875421524,
"step": 1230,
"valid_targets_mean": 3363.4,
"valid_targets_min": 1484
},
{
"epoch": 2.0617696160267114,
"grad_norm": 0.4219724238519255,
"learning_rate": 3.557935943655789e-05,
"loss": 0.208,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.211025670170784,
"step": 1235,
"valid_targets_mean": 4101.8,
"valid_targets_min": 1635
},
{
"epoch": 2.0701168614357264,
"grad_norm": 0.3854874839411631,
"learning_rate": 3.5527012041585936e-05,
"loss": 0.1678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16341739892959595,
"step": 1240,
"valid_targets_mean": 4231.6,
"valid_targets_min": 1796
},
{
"epoch": 2.0784641068447414,
"grad_norm": 0.42452746854104284,
"learning_rate": 3.5474395522008496e-05,
"loss": 0.1733,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15863069891929626,
"step": 1245,
"valid_targets_mean": 3204.3,
"valid_targets_min": 1846
},
{
"epoch": 2.0868113522537564,
"grad_norm": 0.40086387147096286,
"learning_rate": 3.542151078981046e-05,
"loss": 0.1579,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14610186219215393,
"step": 1250,
"valid_targets_mean": 3282.0,
"valid_targets_min": 1648
},
{
"epoch": 2.0951585976627713,
"grad_norm": 0.37932369321136467,
"learning_rate": 3.5368358761625514e-05,
"loss": 0.1755,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15776067972183228,
"step": 1255,
"valid_targets_mean": 3761.4,
"valid_targets_min": 1297
},
{
"epoch": 2.1035058430717863,
"grad_norm": 0.46571127472056717,
"learning_rate": 3.5314940358720356e-05,
"loss": 0.1893,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21181711554527283,
"step": 1260,
"valid_targets_mean": 3221.6,
"valid_targets_min": 1418
},
{
"epoch": 2.1118530884808013,
"grad_norm": 0.36828213655922665,
"learning_rate": 3.5261256506978615e-05,
"loss": 0.1822,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15022584795951843,
"step": 1265,
"valid_targets_mean": 4183.1,
"valid_targets_min": 1611
},
{
"epoch": 2.1202003338898163,
"grad_norm": 0.43354342450859784,
"learning_rate": 3.5207308136884894e-05,
"loss": 0.1889,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15850138664245605,
"step": 1270,
"valid_targets_mean": 3997.0,
"valid_targets_min": 2406
},
{
"epoch": 2.1285475792988313,
"grad_norm": 0.4034080365982173,
"learning_rate": 3.515309618350858e-05,
"loss": 0.158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17198872566223145,
"step": 1275,
"valid_targets_mean": 3938.7,
"valid_targets_min": 1506
},
{
"epoch": 2.1368948247078463,
"grad_norm": 0.4589149483651622,
"learning_rate": 3.509862158648768e-05,
"loss": 0.1939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23061206936836243,
"step": 1280,
"valid_targets_mean": 4169.5,
"valid_targets_min": 914
},
{
"epoch": 2.1452420701168613,
"grad_norm": 0.37991403157787446,
"learning_rate": 3.504388529001252e-05,
"loss": 0.1728,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13785114884376526,
"step": 1285,
"valid_targets_mean": 3620.9,
"valid_targets_min": 1484
},
{
"epoch": 2.1535893155258763,
"grad_norm": 0.40257735278160517,
"learning_rate": 3.498888824280935e-05,
"loss": 0.2019,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18923106789588928,
"step": 1290,
"valid_targets_mean": 4113.9,
"valid_targets_min": 1402
},
{
"epoch": 2.1619365609348913,
"grad_norm": 0.35267496735730713,
"learning_rate": 3.4933631398123956e-05,
"loss": 0.1788,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14884307980537415,
"step": 1295,
"valid_targets_mean": 4393.5,
"valid_targets_min": 1702
},
{
"epoch": 2.1702838063439067,
"grad_norm": 0.3875631912096111,
"learning_rate": 3.487811571370509e-05,
"loss": 0.2014,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19244906306266785,
"step": 1300,
"valid_targets_mean": 4315.8,
"valid_targets_min": 1703
},
{
"epoch": 2.1786310517529217,
"grad_norm": 0.5008686848118249,
"learning_rate": 3.482234215178791e-05,
"loss": 0.1716,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2607955038547516,
"step": 1305,
"valid_targets_mean": 4830.8,
"valid_targets_min": 1239
},
{
"epoch": 2.1869782971619367,
"grad_norm": 0.34968441674849415,
"learning_rate": 3.476631167907727e-05,
"loss": 0.1737,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19340205192565918,
"step": 1310,
"valid_targets_mean": 4876.8,
"valid_targets_min": 1610
},
{
"epoch": 2.1953255425709517,
"grad_norm": 0.4253672560213209,
"learning_rate": 3.471002526673094e-05,
"loss": 0.1675,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17603421211242676,
"step": 1315,
"valid_targets_mean": 3275.1,
"valid_targets_min": 1307
},
{
"epoch": 2.2036727879799667,
"grad_norm": 0.45350784105977754,
"learning_rate": 3.465348389034287e-05,
"loss": 0.1745,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1482657790184021,
"step": 1320,
"valid_targets_mean": 2981.2,
"valid_targets_min": 1426
},
{
"epoch": 2.2120200333889817,
"grad_norm": 0.40629535405371087,
"learning_rate": 3.459668852992617e-05,
"loss": 0.1877,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2098856270313263,
"step": 1325,
"valid_targets_mean": 4527.9,
"valid_targets_min": 1363
},
{
"epoch": 2.2203672787979967,
"grad_norm": 0.33934937183245395,
"learning_rate": 3.453964016989619e-05,
"loss": 0.167,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12940047681331635,
"step": 1330,
"valid_targets_mean": 3973.0,
"valid_targets_min": 1913
},
{
"epoch": 2.2287145242070117,
"grad_norm": 0.41518951613094857,
"learning_rate": 3.4482339799053444e-05,
"loss": 0.2106,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19499072432518005,
"step": 1335,
"valid_targets_mean": 4043.6,
"valid_targets_min": 1715
},
{
"epoch": 2.2370617696160267,
"grad_norm": 0.42937764752446256,
"learning_rate": 3.4424788410566455e-05,
"loss": 0.1637,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2001546174287796,
"step": 1340,
"valid_targets_mean": 3958.1,
"valid_targets_min": 1428
},
{
"epoch": 2.2454090150250416,
"grad_norm": 0.35021242362054383,
"learning_rate": 3.4366987001954555e-05,
"loss": 0.1722,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16036367416381836,
"step": 1345,
"valid_targets_mean": 4299.9,
"valid_targets_min": 1589
},
{
"epoch": 2.2537562604340566,
"grad_norm": 0.36637763427038916,
"learning_rate": 3.43089365750706e-05,
"loss": 0.1775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18065059185028076,
"step": 1350,
"valid_targets_mean": 4715.2,
"valid_targets_min": 2420
},
{
"epoch": 2.2621035058430716,
"grad_norm": 0.3640965084794924,
"learning_rate": 3.425063813608359e-05,
"loss": 0.1795,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18592235445976257,
"step": 1355,
"valid_targets_mean": 5197.2,
"valid_targets_min": 1514
},
{
"epoch": 2.2704507512520866,
"grad_norm": 0.34364181085181167,
"learning_rate": 3.419209269546125e-05,
"loss": 0.1731,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1647244095802307,
"step": 1360,
"valid_targets_mean": 4771.2,
"valid_targets_min": 1960
},
{
"epoch": 2.278797996661102,
"grad_norm": 0.3845436672345355,
"learning_rate": 3.4133301267952486e-05,
"loss": 0.1743,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18283039331436157,
"step": 1365,
"valid_targets_mean": 3974.3,
"valid_targets_min": 1987
},
{
"epoch": 2.287145242070117,
"grad_norm": 0.42668652294874804,
"learning_rate": 3.407426487256982e-05,
"loss": 0.1803,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2542681097984314,
"step": 1370,
"valid_targets_mean": 4809.9,
"valid_targets_min": 2101
},
{
"epoch": 2.295492487479132,
"grad_norm": 0.39718658556445346,
"learning_rate": 3.401498453257172e-05,
"loss": 0.1916,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19539326429367065,
"step": 1375,
"valid_targets_mean": 4438.9,
"valid_targets_min": 1699
},
{
"epoch": 2.303839732888147,
"grad_norm": 0.3601566054598613,
"learning_rate": 3.395546127544489e-05,
"loss": 0.1676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15501947700977325,
"step": 1380,
"valid_targets_mean": 4177.3,
"valid_targets_min": 1183
},
{
"epoch": 2.312186978297162,
"grad_norm": 0.379981010864199,
"learning_rate": 3.389569613288641e-05,
"loss": 0.1842,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17738771438598633,
"step": 1385,
"valid_targets_mean": 4147.7,
"valid_targets_min": 1405
},
{
"epoch": 2.320534223706177,
"grad_norm": 0.43253231266107883,
"learning_rate": 3.383569014078587e-05,
"loss": 0.2109,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2508506774902344,
"step": 1390,
"valid_targets_mean": 4498.2,
"valid_targets_min": 1531
},
{
"epoch": 2.328881469115192,
"grad_norm": 0.4045205462428933,
"learning_rate": 3.3775444339207465e-05,
"loss": 0.1763,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23198941349983215,
"step": 1395,
"valid_targets_mean": 4659.8,
"valid_targets_min": 1244
},
{
"epoch": 2.337228714524207,
"grad_norm": 0.4526577684588381,
"learning_rate": 3.3714959772371885e-05,
"loss": 0.1738,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17988790571689606,
"step": 1400,
"valid_targets_mean": 3236.9,
"valid_targets_min": 1624
},
{
"epoch": 2.345575959933222,
"grad_norm": 0.39924635137397385,
"learning_rate": 3.3654237488638306e-05,
"loss": 0.1864,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1388336420059204,
"step": 1405,
"valid_targets_mean": 3361.2,
"valid_targets_min": 1336
},
{
"epoch": 2.353923205342237,
"grad_norm": 0.36083578880495376,
"learning_rate": 3.3593278540486135e-05,
"loss": 0.1998,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18635602295398712,
"step": 1410,
"valid_targets_mean": 4485.7,
"valid_targets_min": 1942
},
{
"epoch": 2.362270450751252,
"grad_norm": 0.46843076228541314,
"learning_rate": 3.353208398449683e-05,
"loss": 0.1725,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15108218789100647,
"step": 1415,
"valid_targets_mean": 2438.8,
"valid_targets_min": 1355
},
{
"epoch": 2.370617696160267,
"grad_norm": 0.3881753591469688,
"learning_rate": 3.347065488133555e-05,
"loss": 0.1841,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20182552933692932,
"step": 1420,
"valid_targets_mean": 3880.5,
"valid_targets_min": 1462
},
{
"epoch": 2.378964941569282,
"grad_norm": 0.33103684775089787,
"learning_rate": 3.340899229573278e-05,
"loss": 0.1741,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16988861560821533,
"step": 1425,
"valid_targets_mean": 5135.4,
"valid_targets_min": 1957
},
{
"epoch": 2.3873121869782974,
"grad_norm": 0.3839853993740304,
"learning_rate": 3.334709729646589e-05,
"loss": 0.1762,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18751657009124756,
"step": 1430,
"valid_targets_mean": 3841.4,
"valid_targets_min": 1463
},
{
"epoch": 2.395659432387312,
"grad_norm": 0.42140756502663484,
"learning_rate": 3.32849709563406e-05,
"loss": 0.1652,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14289385080337524,
"step": 1435,
"valid_targets_mean": 3429.9,
"valid_targets_min": 1621
},
{
"epoch": 2.4040066777963274,
"grad_norm": 0.41185771845911473,
"learning_rate": 3.322261435217237e-05,
"loss": 0.1714,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23123914003372192,
"step": 1440,
"valid_targets_mean": 4574.2,
"valid_targets_min": 1514
},
{
"epoch": 2.4123539232053424,
"grad_norm": 0.3509324532885418,
"learning_rate": 3.316002856476776e-05,
"loss": 0.1909,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20444336533546448,
"step": 1445,
"valid_targets_mean": 4712.2,
"valid_targets_min": 1919
},
{
"epoch": 2.4207011686143574,
"grad_norm": 0.34885265819549843,
"learning_rate": 3.309721467890571e-05,
"loss": 0.1799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14365056157112122,
"step": 1450,
"valid_targets_mean": 4059.9,
"valid_targets_min": 2310
},
{
"epoch": 2.4290484140233723,
"grad_norm": 0.39384500044336085,
"learning_rate": 3.303417378331867e-05,
"loss": 0.1644,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15038427710533142,
"step": 1455,
"valid_targets_mean": 3667.2,
"valid_targets_min": 1678
},
{
"epoch": 2.4373956594323873,
"grad_norm": 0.37596336192206775,
"learning_rate": 3.2970906970673814e-05,
"loss": 0.1939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15225949883460999,
"step": 1460,
"valid_targets_mean": 4028.6,
"valid_targets_min": 2020
},
{
"epoch": 2.4457429048414023,
"grad_norm": 0.3996370086371849,
"learning_rate": 3.2907415337554045e-05,
"loss": 0.174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1824476420879364,
"step": 1465,
"valid_targets_mean": 3557.2,
"valid_targets_min": 1884
},
{
"epoch": 2.4540901502504173,
"grad_norm": 0.3811821926986149,
"learning_rate": 3.284369998443901e-05,
"loss": 0.1751,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13973477482795715,
"step": 1470,
"valid_targets_mean": 3632.4,
"valid_targets_min": 1515
},
{
"epoch": 2.4624373956594323,
"grad_norm": 0.3771615762931369,
"learning_rate": 3.277976201568603e-05,
"loss": 0.207,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1878414750099182,
"step": 1475,
"valid_targets_mean": 4590.6,
"valid_targets_min": 1847
},
{
"epoch": 2.4707846410684473,
"grad_norm": 0.38583537947825236,
"learning_rate": 3.2715602539510915e-05,
"loss": 0.1717,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13137011229991913,
"step": 1480,
"valid_targets_mean": 3166.2,
"valid_targets_min": 1507
},
{
"epoch": 2.4791318864774623,
"grad_norm": 0.41711885969961404,
"learning_rate": 3.265122266796884e-05,
"loss": 0.196,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19853655993938446,
"step": 1485,
"valid_targets_mean": 3741.7,
"valid_targets_min": 1501
},
{
"epoch": 2.4874791318864773,
"grad_norm": 0.38435960003244696,
"learning_rate": 3.258662351693498e-05,
"loss": 0.1796,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1852615475654602,
"step": 1490,
"valid_targets_mean": 4253.8,
"valid_targets_min": 1354
},
{
"epoch": 2.4958263772954927,
"grad_norm": 0.39689223158603776,
"learning_rate": 3.252180620608524e-05,
"loss": 0.1989,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.28923559188842773,
"step": 1495,
"valid_targets_mean": 5626.3,
"valid_targets_min": 1244
},
{
"epoch": 2.5041736227045073,
"grad_norm": 0.4867002320809955,
"learning_rate": 3.245677185887678e-05,
"loss": 0.1701,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19362148642539978,
"step": 1500,
"valid_targets_mean": 3091.3,
"valid_targets_min": 1425
},
{
"epoch": 2.5125208681135227,
"grad_norm": 0.3840135531557725,
"learning_rate": 3.2391521602528634e-05,
"loss": 0.1978,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16485507786273956,
"step": 1505,
"valid_targets_mean": 3793.5,
"valid_targets_min": 1439
},
{
"epoch": 2.5208681135225377,
"grad_norm": 0.3949626979816311,
"learning_rate": 3.232605656800207e-05,
"loss": 0.2141,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25324124097824097,
"step": 1510,
"valid_targets_mean": 5199.4,
"valid_targets_min": 2406
},
{
"epoch": 2.5292153589315527,
"grad_norm": 0.41371809654196057,
"learning_rate": 3.226037788998105e-05,
"loss": 0.1588,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17497789859771729,
"step": 1515,
"valid_targets_mean": 3453.5,
"valid_targets_min": 2109
},
{
"epoch": 2.5375626043405677,
"grad_norm": 0.41114279636013484,
"learning_rate": 3.219448670685256e-05,
"loss": 0.178,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16214880347251892,
"step": 1520,
"valid_targets_mean": 3575.6,
"valid_targets_min": 1718
},
{
"epoch": 2.5459098497495827,
"grad_norm": 0.36919347560605503,
"learning_rate": 3.212838416068685e-05,
"loss": 0.1779,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15680629014968872,
"step": 1525,
"valid_targets_mean": 3776.2,
"valid_targets_min": 1486
},
{
"epoch": 2.5542570951585977,
"grad_norm": 0.4138003969252916,
"learning_rate": 3.206207139721768e-05,
"loss": 0.1684,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17032107710838318,
"step": 1530,
"valid_targets_mean": 3267.6,
"valid_targets_min": 1473
},
{
"epoch": 2.5626043405676127,
"grad_norm": 0.42196706139878004,
"learning_rate": 3.199554956582241e-05,
"loss": 0.1889,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20715712010860443,
"step": 1535,
"valid_targets_mean": 3526.3,
"valid_targets_min": 1559
},
{
"epoch": 2.5709515859766277,
"grad_norm": 0.3722661515680046,
"learning_rate": 3.192881981950212e-05,
"loss": 0.1513,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1343880593776703,
"step": 1540,
"valid_targets_mean": 3572.8,
"valid_targets_min": 1768
},
{
"epoch": 2.5792988313856426,
"grad_norm": 0.45051136521852037,
"learning_rate": 3.1861883314861616e-05,
"loss": 0.1793,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15990979969501495,
"step": 1545,
"valid_targets_mean": 3307.1,
"valid_targets_min": 1374
},
{
"epoch": 2.5876460767946576,
"grad_norm": 0.35101532783011724,
"learning_rate": 3.179474121208937e-05,
"loss": 0.1759,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1429433524608612,
"step": 1550,
"valid_targets_mean": 3559.9,
"valid_targets_min": 1487
},
{
"epoch": 2.5959933222036726,
"grad_norm": 0.362509471383554,
"learning_rate": 3.172739467493741e-05,
"loss": 0.1562,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15836237370967865,
"step": 1555,
"valid_targets_mean": 3900.3,
"valid_targets_min": 2075
},
{
"epoch": 2.604340567612688,
"grad_norm": 0.38099936311161486,
"learning_rate": 3.165984487070118e-05,
"loss": 0.1901,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17195066809654236,
"step": 1560,
"valid_targets_mean": 4125.4,
"valid_targets_min": 1609
},
{
"epoch": 2.6126878130217026,
"grad_norm": 0.36249830667815447,
"learning_rate": 3.1592092970199266e-05,
"loss": 0.1755,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16423234343528748,
"step": 1565,
"valid_targets_mean": 4058.8,
"valid_targets_min": 1563
},
{
"epoch": 2.621035058430718,
"grad_norm": 0.38776283529242817,
"learning_rate": 3.152414014775315e-05,
"loss": 0.198,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21244670450687408,
"step": 1570,
"valid_targets_mean": 4376.8,
"valid_targets_min": 2040
},
{
"epoch": 2.629382303839733,
"grad_norm": 0.39888439800258313,
"learning_rate": 3.1455987581166784e-05,
"loss": 0.2133,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21948155760765076,
"step": 1575,
"valid_targets_mean": 4703.1,
"valid_targets_min": 1414
},
{
"epoch": 2.637729549248748,
"grad_norm": 0.4050837610287775,
"learning_rate": 3.138763645170626e-05,
"loss": 0.187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18739059567451477,
"step": 1580,
"valid_targets_mean": 3601.9,
"valid_targets_min": 1423
},
{
"epoch": 2.646076794657763,
"grad_norm": 0.4540117082903146,
"learning_rate": 3.1319087944079275e-05,
"loss": 0.1832,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.26165759563446045,
"step": 1585,
"valid_targets_mean": 4085.8,
"valid_targets_min": 1456
},
{
"epoch": 2.654424040066778,
"grad_norm": 0.3897875275227354,
"learning_rate": 3.125034324641462e-05,
"loss": 0.1667,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19020062685012817,
"step": 1590,
"valid_targets_mean": 4242.6,
"valid_targets_min": 1945
},
{
"epoch": 2.662771285475793,
"grad_norm": 0.3747972450813431,
"learning_rate": 3.118140355024159e-05,
"loss": 0.1585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14661559462547302,
"step": 1595,
"valid_targets_mean": 3469.0,
"valid_targets_min": 1730
},
{
"epoch": 2.671118530884808,
"grad_norm": 0.3912630941067351,
"learning_rate": 3.111227005046932e-05,
"loss": 0.1709,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18964293599128723,
"step": 1600,
"valid_targets_mean": 4400.2,
"valid_targets_min": 1610
},
{
"epoch": 2.679465776293823,
"grad_norm": 0.3752998749201782,
"learning_rate": 3.104294394536608e-05,
"loss": 0.1687,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1662491112947464,
"step": 1605,
"valid_targets_mean": 3868.1,
"valid_targets_min": 1425
},
{
"epoch": 2.687813021702838,
"grad_norm": 0.3638451758982178,
"learning_rate": 3.097342643653849e-05,
"loss": 0.1871,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13237513601779938,
"step": 1610,
"valid_targets_mean": 3775.5,
"valid_targets_min": 1252
},
{
"epoch": 2.696160267111853,
"grad_norm": 0.379813588301819,
"learning_rate": 3.090371872891074e-05,
"loss": 0.1939,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18782466650009155,
"step": 1615,
"valid_targets_mean": 4419.9,
"valid_targets_min": 1526
},
{
"epoch": 2.704507512520868,
"grad_norm": 0.42247127664008294,
"learning_rate": 3.083382203070365e-05,
"loss": 0.1578,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14077098667621613,
"step": 1620,
"valid_targets_mean": 2928.9,
"valid_targets_min": 1424
},
{
"epoch": 2.7128547579298834,
"grad_norm": 0.40519702773871696,
"learning_rate": 3.0763737553413766e-05,
"loss": 0.1843,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17491929233074188,
"step": 1625,
"valid_targets_mean": 3589.4,
"valid_targets_min": 1480
},
{
"epoch": 2.721202003338898,
"grad_norm": 0.3565728034617864,
"learning_rate": 3.069346651179233e-05,
"loss": 0.1693,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14302650094032288,
"step": 1630,
"valid_targets_mean": 3706.4,
"valid_targets_min": 1461
},
{
"epoch": 2.7295492487479134,
"grad_norm": 0.3551042853907041,
"learning_rate": 3.0623010123824245e-05,
"loss": 0.1662,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1554560512304306,
"step": 1635,
"valid_targets_mean": 3965.8,
"valid_targets_min": 1330
},
{
"epoch": 2.7378964941569284,
"grad_norm": 0.41497730093592444,
"learning_rate": 3.0552369610706985e-05,
"loss": 0.1782,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14315751194953918,
"step": 1640,
"valid_targets_mean": 3088.9,
"valid_targets_min": 1370
},
{
"epoch": 2.7462437395659434,
"grad_norm": 0.34080950122096165,
"learning_rate": 3.0481546196829375e-05,
"loss": 0.1909,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14897558093070984,
"step": 1645,
"valid_targets_mean": 4127.8,
"valid_targets_min": 1515
},
{
"epoch": 2.7545909849749584,
"grad_norm": 0.42473649987677536,
"learning_rate": 3.041054110975041e-05,
"loss": 0.1861,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2169104963541031,
"step": 1650,
"valid_targets_mean": 3800.3,
"valid_targets_min": 1556
},
{
"epoch": 2.7629382303839733,
"grad_norm": 0.3946586777993279,
"learning_rate": 3.033935558017797e-05,
"loss": 0.1856,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19707706570625305,
"step": 1655,
"valid_targets_mean": 4139.6,
"valid_targets_min": 1390
},
{
"epoch": 2.7712854757929883,
"grad_norm": 0.41140216305736166,
"learning_rate": 3.0267990841947492e-05,
"loss": 0.1665,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1477285474538803,
"step": 1660,
"valid_targets_mean": 3071.9,
"valid_targets_min": 1516
},
{
"epoch": 2.7796327212020033,
"grad_norm": 0.3670539040599514,
"learning_rate": 3.0196448132000563e-05,
"loss": 0.1743,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1892966777086258,
"step": 1665,
"valid_targets_mean": 4099.4,
"valid_targets_min": 1486
},
{
"epoch": 2.7879799666110183,
"grad_norm": 0.3804894809754149,
"learning_rate": 3.0124728690363504e-05,
"loss": 0.1793,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.181321918964386,
"step": 1670,
"valid_targets_mean": 4185.7,
"valid_targets_min": 1961
},
{
"epoch": 2.7963272120200333,
"grad_norm": 0.40144395625229207,
"learning_rate": 3.0052833760125864e-05,
"loss": 0.1682,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13178232312202454,
"step": 1675,
"valid_targets_mean": 3274.2,
"valid_targets_min": 1648
},
{
"epoch": 2.8046744574290483,
"grad_norm": 0.35649803939751235,
"learning_rate": 2.9980764587418885e-05,
"loss": 0.1859,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13898998498916626,
"step": 1680,
"valid_targets_mean": 3781.2,
"valid_targets_min": 1593
},
{
"epoch": 2.8130217028380633,
"grad_norm": 0.3942851927491884,
"learning_rate": 2.990852242139389e-05,
"loss": 0.2039,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20746630430221558,
"step": 1685,
"valid_targets_mean": 3834.1,
"valid_targets_min": 2211
},
{
"epoch": 2.8213689482470787,
"grad_norm": 0.37032432922943975,
"learning_rate": 2.983610851420064e-05,
"loss": 0.1701,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19588196277618408,
"step": 1690,
"valid_targets_mean": 4690.8,
"valid_targets_min": 1562
},
{
"epoch": 2.8297161936560933,
"grad_norm": 0.34562404981740413,
"learning_rate": 2.976352412096563e-05,
"loss": 0.1609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14124101400375366,
"step": 1695,
"valid_targets_mean": 3965.9,
"valid_targets_min": 1494
},
{
"epoch": 2.8380634390651087,
"grad_norm": 0.3616931031916152,
"learning_rate": 2.969077049977034e-05,
"loss": 0.1636,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19306877255439758,
"step": 1700,
"valid_targets_mean": 4813.1,
"valid_targets_min": 1705
},
{
"epoch": 2.8464106844741233,
"grad_norm": 0.37506220041510785,
"learning_rate": 2.9617848911629402e-05,
"loss": 0.1643,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1629461944103241,
"step": 1705,
"valid_targets_mean": 3513.9,
"valid_targets_min": 1421
},
{
"epoch": 2.8547579298831387,
"grad_norm": 0.39377736080444464,
"learning_rate": 2.9544760620468794e-05,
"loss": 0.195,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20144890248775482,
"step": 1710,
"valid_targets_mean": 4516.2,
"valid_targets_min": 1899
},
{
"epoch": 2.8631051752921537,
"grad_norm": 0.42090417381625955,
"learning_rate": 2.9471506893103883e-05,
"loss": 0.1816,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1861152946949005,
"step": 1715,
"valid_targets_mean": 3644.5,
"valid_targets_min": 1637
},
{
"epoch": 2.8714524207011687,
"grad_norm": 0.4271180840796994,
"learning_rate": 2.939808899921749e-05,
"loss": 0.1779,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22819042205810547,
"step": 1720,
"valid_targets_mean": 4635.6,
"valid_targets_min": 1542
},
{
"epoch": 2.8797996661101837,
"grad_norm": 0.4174935857105486,
"learning_rate": 2.932450821133788e-05,
"loss": 0.1903,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1691751480102539,
"step": 1725,
"valid_targets_mean": 3461.2,
"valid_targets_min": 1777
},
{
"epoch": 2.8881469115191987,
"grad_norm": 0.3850643570111332,
"learning_rate": 2.9250765804816712e-05,
"loss": 0.1904,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19057656824588776,
"step": 1730,
"valid_targets_mean": 4053.8,
"valid_targets_min": 1866
},
{
"epoch": 2.8964941569282137,
"grad_norm": 0.4032107394793199,
"learning_rate": 2.917686305780692e-05,
"loss": 0.1856,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2273217737674713,
"step": 1735,
"valid_targets_mean": 4329.3,
"valid_targets_min": 1386
},
{
"epoch": 2.9048414023372287,
"grad_norm": 0.3835287858006912,
"learning_rate": 2.9102801251240575e-05,
"loss": 0.1652,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15932297706604004,
"step": 1740,
"valid_targets_mean": 3465.6,
"valid_targets_min": 1384
},
{
"epoch": 2.9131886477462436,
"grad_norm": 0.39641810946841144,
"learning_rate": 2.902858166880667e-05,
"loss": 0.18,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.212120920419693,
"step": 1745,
"valid_targets_mean": 4468.6,
"valid_targets_min": 1863
},
{
"epoch": 2.9215358931552586,
"grad_norm": 0.45144937261721835,
"learning_rate": 2.8954205596928873e-05,
"loss": 0.1669,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21817569434642792,
"step": 1750,
"valid_targets_mean": 4008.8,
"valid_targets_min": 2230
},
{
"epoch": 2.9298831385642736,
"grad_norm": 0.3374028876731595,
"learning_rate": 2.8879674324743246e-05,
"loss": 0.1717,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1711035966873169,
"step": 1755,
"valid_targets_mean": 4836.6,
"valid_targets_min": 1134
},
{
"epoch": 2.9382303839732886,
"grad_norm": 0.4290703903958169,
"learning_rate": 2.880498914407587e-05,
"loss": 0.1814,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24183110892772675,
"step": 1760,
"valid_targets_mean": 4185.7,
"valid_targets_min": 1526
},
{
"epoch": 2.946577629382304,
"grad_norm": 0.4162550666800131,
"learning_rate": 2.8730151349420475e-05,
"loss": 0.1636,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17692837119102478,
"step": 1765,
"valid_targets_mean": 3091.3,
"valid_targets_min": 770
},
{
"epoch": 2.9549248747913186,
"grad_norm": 0.4040831857315516,
"learning_rate": 2.865516223791601e-05,
"loss": 0.1906,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22569309175014496,
"step": 1770,
"valid_targets_mean": 4042.6,
"valid_targets_min": 1364
},
{
"epoch": 2.963272120200334,
"grad_norm": 0.47408219736416474,
"learning_rate": 2.8580023109324137e-05,
"loss": 0.1645,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18150001764297485,
"step": 1775,
"valid_targets_mean": 3143.0,
"valid_targets_min": 1303
},
{
"epoch": 2.971619365609349,
"grad_norm": 0.36774294528090073,
"learning_rate": 2.8504735266006717e-05,
"loss": 0.1917,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1483282744884491,
"step": 1780,
"valid_targets_mean": 4298.1,
"valid_targets_min": 1272
},
{
"epoch": 2.979966611018364,
"grad_norm": 0.3717567087773957,
"learning_rate": 2.8429300012903245e-05,
"loss": 0.1811,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1875116527080536,
"step": 1785,
"valid_targets_mean": 4630.8,
"valid_targets_min": 1352
},
{
"epoch": 2.988313856427379,
"grad_norm": 0.41166317760784843,
"learning_rate": 2.8353718657508206e-05,
"loss": 0.1715,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18830318748950958,
"step": 1790,
"valid_targets_mean": 3810.6,
"valid_targets_min": 1838
},
{
"epoch": 2.996661101836394,
"grad_norm": 0.37316317488800915,
"learning_rate": 2.827799250984844e-05,
"loss": 0.1562,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1336875557899475,
"step": 1795,
"valid_targets_mean": 3665.7,
"valid_targets_min": 1584
},
{
"epoch": 3.005008347245409,
"grad_norm": 0.4063548789037882,
"learning_rate": 2.8202122882460418e-05,
"loss": 0.1741,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1809404045343399,
"step": 1800,
"valid_targets_mean": 3502.6,
"valid_targets_min": 1899
},
{
"epoch": 3.013355592654424,
"grad_norm": 0.4336551959152753,
"learning_rate": 2.81261110903675e-05,
"loss": 0.1603,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14447429776191711,
"step": 1805,
"valid_targets_mean": 2828.2,
"valid_targets_min": 1514
},
{
"epoch": 3.021702838063439,
"grad_norm": 0.36481871912093367,
"learning_rate": 2.804995845105714e-05,
"loss": 0.1742,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18342864513397217,
"step": 1810,
"valid_targets_mean": 4443.3,
"valid_targets_min": 1381
},
{
"epoch": 3.030050083472454,
"grad_norm": 0.37691446585353855,
"learning_rate": 2.7973666284458048e-05,
"loss": 0.1535,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1622544825077057,
"step": 1815,
"valid_targets_mean": 3952.1,
"valid_targets_min": 1869
},
{
"epoch": 3.038397328881469,
"grad_norm": 0.4022100122802466,
"learning_rate": 2.7897235912917318e-05,
"loss": 0.1552,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13804765045642853,
"step": 1820,
"valid_targets_mean": 3232.9,
"valid_targets_min": 1540
},
{
"epoch": 3.046744574290484,
"grad_norm": 0.4338966945529254,
"learning_rate": 2.7820668661177505e-05,
"loss": 0.1723,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13496927917003632,
"step": 1825,
"valid_targets_mean": 3257.6,
"valid_targets_min": 1984
},
{
"epoch": 3.0550918196994994,
"grad_norm": 0.421460512392483,
"learning_rate": 2.7743965856353666e-05,
"loss": 0.1504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11614467203617096,
"step": 1830,
"valid_targets_mean": 3389.6,
"valid_targets_min": 1405
},
{
"epoch": 3.0634390651085144,
"grad_norm": 0.38268929748837494,
"learning_rate": 2.7667128827910343e-05,
"loss": 0.1594,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12622597813606262,
"step": 1835,
"valid_targets_mean": 3840.2,
"valid_targets_min": 1653
},
{
"epoch": 3.0717863105175294,
"grad_norm": 0.38641056298745535,
"learning_rate": 2.7590158907638552e-05,
"loss": 0.1792,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14331111311912537,
"step": 1840,
"valid_targets_mean": 3908.7,
"valid_targets_min": 1496
},
{
"epoch": 3.0801335559265444,
"grad_norm": 0.391516861603331,
"learning_rate": 2.7513057429632656e-05,
"loss": 0.172,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12942743301391602,
"step": 1845,
"valid_targets_mean": 3237.4,
"valid_targets_min": 1461
},
{
"epoch": 3.0884808013355594,
"grad_norm": 0.4097059452528929,
"learning_rate": 2.7435825730267285e-05,
"loss": 0.1956,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2104695439338684,
"step": 1850,
"valid_targets_mean": 4672.2,
"valid_targets_min": 1134
},
{
"epoch": 3.0968280467445743,
"grad_norm": 0.4184849714104454,
"learning_rate": 2.7358465148174143e-05,
"loss": 0.1461,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1452220380306244,
"step": 1855,
"valid_targets_mean": 3005.0,
"valid_targets_min": 1429
},
{
"epoch": 3.1051752921535893,
"grad_norm": 0.3721084887961403,
"learning_rate": 2.728097702421882e-05,
"loss": 0.179,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18598805367946625,
"step": 1860,
"valid_targets_mean": 4447.5,
"valid_targets_min": 1512
},
{
"epoch": 3.1135225375626043,
"grad_norm": 0.3886671460796501,
"learning_rate": 2.720336270147754e-05,
"loss": 0.1674,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21523085236549377,
"step": 1865,
"valid_targets_mean": 4612.0,
"valid_targets_min": 1589
},
{
"epoch": 3.1218697829716193,
"grad_norm": 0.412037645711169,
"learning_rate": 2.71256235252139e-05,
"loss": 0.151,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13819587230682373,
"step": 1870,
"valid_targets_mean": 2930.2,
"valid_targets_min": 1714
},
{
"epoch": 3.1302170283806343,
"grad_norm": 0.4065631278058066,
"learning_rate": 2.7047760842855536e-05,
"loss": 0.1598,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13124097883701324,
"step": 1875,
"valid_targets_mean": 3087.6,
"valid_targets_min": 1475
},
{
"epoch": 3.1385642737896493,
"grad_norm": 0.38320463777177993,
"learning_rate": 2.6969776003970777e-05,
"loss": 0.1697,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1506544053554535,
"step": 1880,
"valid_targets_mean": 3822.8,
"valid_targets_min": 1484
},
{
"epoch": 3.1469115191986643,
"grad_norm": 0.36118220642464954,
"learning_rate": 2.6891670360245244e-05,
"loss": 0.1611,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12876233458518982,
"step": 1885,
"valid_targets_mean": 3769.2,
"valid_targets_min": 1491
},
{
"epoch": 3.1552587646076793,
"grad_norm": 0.43603726449006625,
"learning_rate": 2.6813445265458438e-05,
"loss": 0.172,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23195567727088928,
"step": 1890,
"valid_targets_mean": 3860.2,
"valid_targets_min": 1772
},
{
"epoch": 3.1636060100166947,
"grad_norm": 0.37254088444846367,
"learning_rate": 2.6735102075460257e-05,
"loss": 0.1527,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18866942822933197,
"step": 1895,
"valid_targets_mean": 4336.5,
"valid_targets_min": 1448
},
{
"epoch": 3.1719532554257097,
"grad_norm": 0.3527245570075822,
"learning_rate": 2.665664214814752e-05,
"loss": 0.1664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1749316155910492,
"step": 1900,
"valid_targets_mean": 4439.5,
"valid_targets_min": 1976
},
{
"epoch": 3.1803005008347247,
"grad_norm": 0.3741721242234238,
"learning_rate": 2.65780668434404e-05,
"loss": 0.1609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1178446114063263,
"step": 1905,
"valid_targets_mean": 3321.6,
"valid_targets_min": 1695
},
{
"epoch": 3.1886477462437397,
"grad_norm": 0.3557644578617192,
"learning_rate": 2.649937752325887e-05,
"loss": 0.182,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15885943174362183,
"step": 1910,
"valid_targets_mean": 4127.2,
"valid_targets_min": 1504
},
{
"epoch": 3.1969949916527547,
"grad_norm": 0.39493533441845896,
"learning_rate": 2.64205755514991e-05,
"loss": 0.1651,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13406017422676086,
"step": 1915,
"valid_targets_mean": 3264.3,
"valid_targets_min": 1372
},
{
"epoch": 3.2053422370617697,
"grad_norm": 0.3509469694391073,
"learning_rate": 2.634166229400982e-05,
"loss": 0.145,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1568133533000946,
"step": 1920,
"valid_targets_mean": 4733.9,
"valid_targets_min": 1977
},
{
"epoch": 3.2136894824707847,
"grad_norm": 0.3790815521393634,
"learning_rate": 2.626263911856863e-05,
"loss": 0.1746,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17264772951602936,
"step": 1925,
"valid_targets_mean": 4220.8,
"valid_targets_min": 1563
},
{
"epoch": 3.2220367278797997,
"grad_norm": 0.37642148833827255,
"learning_rate": 2.6183507394858296e-05,
"loss": 0.1969,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15538159012794495,
"step": 1930,
"valid_targets_mean": 3755.3,
"valid_targets_min": 1425
},
{
"epoch": 3.2303839732888147,
"grad_norm": 0.426221267866572,
"learning_rate": 2.6104268494443027e-05,
"loss": 0.1854,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17054955661296844,
"step": 1935,
"valid_targets_mean": 3506.8,
"valid_targets_min": 1642
},
{
"epoch": 3.2387312186978297,
"grad_norm": 0.371484980238735,
"learning_rate": 2.6024923790744686e-05,
"loss": 0.1775,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18081288039684296,
"step": 1940,
"valid_targets_mean": 4860.9,
"valid_targets_min": 1481
},
{
"epoch": 3.2470784641068446,
"grad_norm": 0.35304609695423167,
"learning_rate": 2.594547465901899e-05,
"loss": 0.1842,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18706491589546204,
"step": 1945,
"valid_targets_mean": 5291.2,
"valid_targets_min": 2097
},
{
"epoch": 3.2554257095158596,
"grad_norm": 0.39249732221568734,
"learning_rate": 2.5865922476331674e-05,
"loss": 0.1681,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17834143340587616,
"step": 1950,
"valid_targets_mean": 3747.8,
"valid_targets_min": 1414
},
{
"epoch": 3.2637729549248746,
"grad_norm": 0.4158339487101832,
"learning_rate": 2.5786268621534626e-05,
"loss": 0.1707,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2010677456855774,
"step": 1955,
"valid_targets_mean": 4371.7,
"valid_targets_min": 605
},
{
"epoch": 3.27212020033389,
"grad_norm": 0.3690944474645874,
"learning_rate": 2.5706514475241964e-05,
"loss": 0.1752,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1624496877193451,
"step": 1960,
"valid_targets_mean": 4484.4,
"valid_targets_min": 1313
},
{
"epoch": 3.2804674457429046,
"grad_norm": 0.34654217176754015,
"learning_rate": 2.5626661419806147e-05,
"loss": 0.1735,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14796465635299683,
"step": 1965,
"valid_targets_mean": 4165.5,
"valid_targets_min": 1381
},
{
"epoch": 3.28881469115192,
"grad_norm": 0.38630759381846,
"learning_rate": 2.5546710839293988e-05,
"loss": 0.1935,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18288516998291016,
"step": 1970,
"valid_targets_mean": 4361.8,
"valid_targets_min": 1411
},
{
"epoch": 3.297161936560935,
"grad_norm": 0.3898244129831731,
"learning_rate": 2.5466664119462667e-05,
"loss": 0.1726,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15863215923309326,
"step": 1975,
"valid_targets_mean": 3915.2,
"valid_targets_min": 1481
},
{
"epoch": 3.30550918196995,
"grad_norm": 0.3642717056752214,
"learning_rate": 2.5386522647735712e-05,
"loss": 0.1762,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15942087769508362,
"step": 1980,
"valid_targets_mean": 4849.8,
"valid_targets_min": 1796
},
{
"epoch": 3.313856427378965,
"grad_norm": 0.3460925560935459,
"learning_rate": 2.530628781317896e-05,
"loss": 0.1759,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1592680811882019,
"step": 1985,
"valid_targets_mean": 4602.9,
"valid_targets_min": 1364
},
{
"epoch": 3.32220367278798,
"grad_norm": 0.40797636271926757,
"learning_rate": 2.5225961006476484e-05,
"loss": 0.1805,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1347290724515915,
"step": 1990,
"valid_targets_mean": 3048.2,
"valid_targets_min": 1830
},
{
"epoch": 3.330550918196995,
"grad_norm": 0.3804223719640401,
"learning_rate": 2.5145543619906456e-05,
"loss": 0.1657,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14686168730258942,
"step": 1995,
"valid_targets_mean": 4163.6,
"valid_targets_min": 1405
},
{
"epoch": 3.33889816360601,
"grad_norm": 0.3490025920030248,
"learning_rate": 2.5065037047317066e-05,
"loss": 0.1442,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14989185333251953,
"step": 2000,
"valid_targets_mean": 4367.3,
"valid_targets_min": 1337
},
{
"epoch": 3.347245409015025,
"grad_norm": 0.3683089852500467,
"learning_rate": 2.4984442684102307e-05,
"loss": 0.1677,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1541152000427246,
"step": 2005,
"valid_targets_mean": 3845.2,
"valid_targets_min": 1783
},
{
"epoch": 3.35559265442404,
"grad_norm": 0.3704411504379215,
"learning_rate": 2.4903761927177853e-05,
"loss": 0.1592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19071197509765625,
"step": 2010,
"valid_targets_mean": 4362.1,
"valid_targets_min": 1466
},
{
"epoch": 3.363939899833055,
"grad_norm": 0.338722055336742,
"learning_rate": 2.482299617495678e-05,
"loss": 0.1545,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12962791323661804,
"step": 2015,
"valid_targets_mean": 4244.9,
"valid_targets_min": 1381
},
{
"epoch": 3.37228714524207,
"grad_norm": 0.39478500289782087,
"learning_rate": 2.474214682732538e-05,
"loss": 0.1793,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19913721084594727,
"step": 2020,
"valid_targets_mean": 4163.4,
"valid_targets_min": 1990
},
{
"epoch": 3.380634390651085,
"grad_norm": 0.38863850667827293,
"learning_rate": 2.466121528561887e-05,
"loss": 0.1759,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.132572203874588,
"step": 2025,
"valid_targets_mean": 3434.7,
"valid_targets_min": 1395
},
{
"epoch": 3.3889816360601,
"grad_norm": 0.4235567932378229,
"learning_rate": 2.4580202952597106e-05,
"loss": 0.1676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16671383380889893,
"step": 2030,
"valid_targets_mean": 3554.2,
"valid_targets_min": 1396
},
{
"epoch": 3.3973288814691154,
"grad_norm": 0.421763901302591,
"learning_rate": 2.4499111232420275e-05,
"loss": 0.1846,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2130683958530426,
"step": 2035,
"valid_targets_mean": 4445.9,
"valid_targets_min": 1352
},
{
"epoch": 3.4056761268781304,
"grad_norm": 0.3387478073736283,
"learning_rate": 2.441794153062457e-05,
"loss": 0.1768,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18954172730445862,
"step": 2040,
"valid_targets_mean": 5600.1,
"valid_targets_min": 2338
},
{
"epoch": 3.4140233722871454,
"grad_norm": 0.37126570579563545,
"learning_rate": 2.4336695254097782e-05,
"loss": 0.1501,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14153939485549927,
"step": 2045,
"valid_targets_mean": 4260.2,
"valid_targets_min": 2145
},
{
"epoch": 3.4223706176961604,
"grad_norm": 0.37863841685778366,
"learning_rate": 2.425537381105498e-05,
"loss": 0.174,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14031752943992615,
"step": 2050,
"valid_targets_mean": 3959.4,
"valid_targets_min": 1453
},
{
"epoch": 3.4307178631051753,
"grad_norm": 0.37263205538291166,
"learning_rate": 2.4173978611014053e-05,
"loss": 0.1616,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14423030614852905,
"step": 2055,
"valid_targets_mean": 4332.2,
"valid_targets_min": 2411
},
{
"epoch": 3.4390651085141903,
"grad_norm": 0.3769211186276375,
"learning_rate": 2.40925110647713e-05,
"loss": 0.1746,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14692571759223938,
"step": 2060,
"valid_targets_mean": 3559.0,
"valid_targets_min": 1552
},
{
"epoch": 3.4474123539232053,
"grad_norm": 0.3628988969465614,
"learning_rate": 2.4010972584376963e-05,
"loss": 0.1665,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16812914609909058,
"step": 2065,
"valid_targets_mean": 4515.1,
"valid_targets_min": 1981
},
{
"epoch": 3.4557595993322203,
"grad_norm": 0.3315799473254126,
"learning_rate": 2.392936458311078e-05,
"loss": 0.1872,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16034258902072906,
"step": 2070,
"valid_targets_mean": 4660.2,
"valid_targets_min": 1419
},
{
"epoch": 3.4641068447412353,
"grad_norm": 0.39977794009671996,
"learning_rate": 2.3847688475457455e-05,
"loss": 0.152,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14341315627098083,
"step": 2075,
"valid_targets_mean": 3301.4,
"valid_targets_min": 1473
},
{
"epoch": 3.4724540901502503,
"grad_norm": 0.4258292227392421,
"learning_rate": 2.3765945677082168e-05,
"loss": 0.1736,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15355074405670166,
"step": 2080,
"valid_targets_mean": 3067.2,
"valid_targets_min": 1652
},
{
"epoch": 3.4808013355592653,
"grad_norm": 0.34990017622725467,
"learning_rate": 2.368413760480603e-05,
"loss": 0.1762,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18550139665603638,
"step": 2085,
"valid_targets_mean": 5174.8,
"valid_targets_min": 1355
},
{
"epoch": 3.4891485809682803,
"grad_norm": 0.43741463678536596,
"learning_rate": 2.360226567658151e-05,
"loss": 0.1887,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23607659339904785,
"step": 2090,
"valid_targets_mean": 4738.4,
"valid_targets_min": 1560
},
{
"epoch": 3.4974958263772953,
"grad_norm": 0.36098318535126145,
"learning_rate": 2.3520331311467883e-05,
"loss": 0.1701,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13859453797340393,
"step": 2095,
"valid_targets_mean": 4226.0,
"valid_targets_min": 1984
},
{
"epoch": 3.5058430717863107,
"grad_norm": 0.4301289729769782,
"learning_rate": 2.3438335929606613e-05,
"loss": 0.143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16290982067584991,
"step": 2100,
"valid_targets_mean": 3329.2,
"valid_targets_min": 1283
},
{
"epoch": 3.5141903171953257,
"grad_norm": 0.3568191951818184,
"learning_rate": 2.3356280952196757e-05,
"loss": 0.1604,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1774396449327469,
"step": 2105,
"valid_targets_mean": 4704.4,
"valid_targets_min": 1949
},
{
"epoch": 3.5225375626043407,
"grad_norm": 0.41423710714049033,
"learning_rate": 2.3274167801470314e-05,
"loss": 0.1699,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19074732065200806,
"step": 2110,
"valid_targets_mean": 4236.9,
"valid_targets_min": 1529
},
{
"epoch": 3.5308848080133557,
"grad_norm": 0.39153286220848726,
"learning_rate": 2.3191997900667588e-05,
"loss": 0.166,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14745816588401794,
"step": 2115,
"valid_targets_mean": 3523.2,
"valid_targets_min": 1926
},
{
"epoch": 3.5392320534223707,
"grad_norm": 0.3941462680371825,
"learning_rate": 2.310977267401251e-05,
"loss": 0.1643,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14406254887580872,
"step": 2120,
"valid_targets_mean": 3569.7,
"valid_targets_min": 1869
},
{
"epoch": 3.5475792988313857,
"grad_norm": 0.4740609600816517,
"learning_rate": 2.302749354668795e-05,
"loss": 0.164,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16030162572860718,
"step": 2125,
"valid_targets_mean": 4044.6,
"valid_targets_min": 2171
},
{
"epoch": 3.5559265442404007,
"grad_norm": 0.38717906332991703,
"learning_rate": 2.2945161944811038e-05,
"loss": 0.1672,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1600584238767624,
"step": 2130,
"valid_targets_mean": 4036.8,
"valid_targets_min": 1937
},
{
"epoch": 3.5642737896494157,
"grad_norm": 0.3942248072998805,
"learning_rate": 2.2862779295408405e-05,
"loss": 0.1642,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13350114226341248,
"step": 2135,
"valid_targets_mean": 3245.2,
"valid_targets_min": 772
},
{
"epoch": 3.5726210350584306,
"grad_norm": 0.3785965227686558,
"learning_rate": 2.2780347026391495e-05,
"loss": 0.1836,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17680341005325317,
"step": 2140,
"valid_targets_mean": 3832.2,
"valid_targets_min": 1252
},
{
"epoch": 3.5809682804674456,
"grad_norm": 0.4062552089965508,
"learning_rate": 2.2697866566531775e-05,
"loss": 0.1652,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15948735177516937,
"step": 2145,
"valid_targets_mean": 3269.7,
"valid_targets_min": 1462
},
{
"epoch": 3.5893155258764606,
"grad_norm": 0.39689292362932294,
"learning_rate": 2.2615339345436e-05,
"loss": 0.1762,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14856375753879547,
"step": 2150,
"valid_targets_mean": 3517.3,
"valid_targets_min": 1882
},
{
"epoch": 3.597662771285476,
"grad_norm": 0.4099328299745972,
"learning_rate": 2.2532766793521413e-05,
"loss": 0.1755,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22073976695537567,
"step": 2155,
"valid_targets_mean": 4827.8,
"valid_targets_min": 1768
},
{
"epoch": 3.6060100166944906,
"grad_norm": 0.38534165803890724,
"learning_rate": 2.245015034199097e-05,
"loss": 0.1741,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20381030440330505,
"step": 2160,
"valid_targets_mean": 4272.9,
"valid_targets_min": 1875
},
{
"epoch": 3.614357262103506,
"grad_norm": 0.39124565933414396,
"learning_rate": 2.2367491422808514e-05,
"loss": 0.1814,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15277761220932007,
"step": 2165,
"valid_targets_mean": 3519.5,
"valid_targets_min": 1422
},
{
"epoch": 3.6227045075125206,
"grad_norm": 0.34314548777848675,
"learning_rate": 2.228479146867397e-05,
"loss": 0.1604,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15035606920719147,
"step": 2170,
"valid_targets_mean": 4359.1,
"valid_targets_min": 1462
},
{
"epoch": 3.631051752921536,
"grad_norm": 0.3987077578855071,
"learning_rate": 2.2202051912998516e-05,
"loss": 0.1695,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17159733176231384,
"step": 2175,
"valid_targets_mean": 3648.7,
"valid_targets_min": 1435
},
{
"epoch": 3.639398998330551,
"grad_norm": 0.3928712371169234,
"learning_rate": 2.2119274189879727e-05,
"loss": 0.1742,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15713515877723694,
"step": 2180,
"valid_targets_mean": 3639.8,
"valid_targets_min": 1441
},
{
"epoch": 3.647746243739566,
"grad_norm": 0.36049639331217803,
"learning_rate": 2.2036459734076715e-05,
"loss": 0.1587,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17143437266349792,
"step": 2185,
"valid_targets_mean": 4514.0,
"valid_targets_min": 1380
},
{
"epoch": 3.656093489148581,
"grad_norm": 0.41890205751392817,
"learning_rate": 2.1953609980985266e-05,
"loss": 0.1919,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19017371535301208,
"step": 2190,
"valid_targets_mean": 3717.8,
"valid_targets_min": 1407
},
{
"epoch": 3.664440734557596,
"grad_norm": 0.39444941209112105,
"learning_rate": 2.1870726366612978e-05,
"loss": 0.167,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1602620631456375,
"step": 2195,
"valid_targets_mean": 3927.1,
"valid_targets_min": 1394
},
{
"epoch": 3.672787979966611,
"grad_norm": 0.35926565619338585,
"learning_rate": 2.1787810327554345e-05,
"loss": 0.1599,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14117556810379028,
"step": 2200,
"valid_targets_mean": 3907.8,
"valid_targets_min": 2413
},
{
"epoch": 3.681135225375626,
"grad_norm": 0.3862780057443413,
"learning_rate": 2.170486330096586e-05,
"loss": 0.1573,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15667811036109924,
"step": 2205,
"valid_targets_mean": 4227.2,
"valid_targets_min": 2011
},
{
"epoch": 3.689482470784641,
"grad_norm": 0.37801749942623897,
"learning_rate": 2.1621886724541126e-05,
"loss": 0.1676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1503739356994629,
"step": 2210,
"valid_targets_mean": 3859.5,
"valid_targets_min": 1506
},
{
"epoch": 3.697829716193656,
"grad_norm": 0.3819200121905654,
"learning_rate": 2.1538882036485923e-05,
"loss": 0.1643,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13664531707763672,
"step": 2215,
"valid_targets_mean": 3466.6,
"valid_targets_min": 1425
},
{
"epoch": 3.706176961602671,
"grad_norm": 0.3914481721846573,
"learning_rate": 2.1455850675493267e-05,
"loss": 0.1534,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16868643462657928,
"step": 2220,
"valid_targets_mean": 4031.0,
"valid_targets_min": 1388
},
{
"epoch": 3.714524207011686,
"grad_norm": 0.3608402720784947,
"learning_rate": 2.1372794080718506e-05,
"loss": 0.1677,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16697438061237335,
"step": 2225,
"valid_targets_mean": 4327.9,
"valid_targets_min": 1627
},
{
"epoch": 3.7228714524207014,
"grad_norm": 0.4327178433338122,
"learning_rate": 2.1289713691754338e-05,
"loss": 0.1556,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18194326758384705,
"step": 2230,
"valid_targets_mean": 3355.5,
"valid_targets_min": 1362
},
{
"epoch": 3.731218697829716,
"grad_norm": 0.44715678173635187,
"learning_rate": 2.1206610948605894e-05,
"loss": 0.1764,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20626530051231384,
"step": 2235,
"valid_targets_mean": 3565.8,
"valid_targets_min": 1426
},
{
"epoch": 3.7395659432387314,
"grad_norm": 0.3604718932122772,
"learning_rate": 2.1123487291665753e-05,
"loss": 0.1664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14898714423179626,
"step": 2240,
"valid_targets_mean": 4825.9,
"valid_targets_min": 2001
},
{
"epoch": 3.7479131886477464,
"grad_norm": 0.3560806768881732,
"learning_rate": 2.104034416168899e-05,
"loss": 0.1662,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13929888606071472,
"step": 2245,
"valid_targets_mean": 3776.2,
"valid_targets_min": 1405
},
{
"epoch": 3.7562604340567614,
"grad_norm": 0.3776098424981539,
"learning_rate": 2.0957182999768198e-05,
"loss": 0.1631,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22549846768379211,
"step": 2250,
"valid_targets_mean": 4964.6,
"valid_targets_min": 1501
},
{
"epoch": 3.7646076794657763,
"grad_norm": 0.3574908549101836,
"learning_rate": 2.0874005247308512e-05,
"loss": 0.1619,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15582001209259033,
"step": 2255,
"valid_targets_mean": 4493.7,
"valid_targets_min": 1964
},
{
"epoch": 3.7729549248747913,
"grad_norm": 0.3869727245566632,
"learning_rate": 2.0790812346002626e-05,
"loss": 0.191,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18079110980033875,
"step": 2260,
"valid_targets_mean": 4029.4,
"valid_targets_min": 1660
},
{
"epoch": 3.7813021702838063,
"grad_norm": 0.3821422025597931,
"learning_rate": 2.0707605737805797e-05,
"loss": 0.1485,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17510807514190674,
"step": 2265,
"valid_targets_mean": 3819.0,
"valid_targets_min": 1490
},
{
"epoch": 3.7896494156928213,
"grad_norm": 0.3471735509183643,
"learning_rate": 2.0624386864910876e-05,
"loss": 0.1909,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15804417431354523,
"step": 2270,
"valid_targets_mean": 4479.2,
"valid_targets_min": 1755
},
{
"epoch": 3.7979966611018363,
"grad_norm": 0.3939626953333758,
"learning_rate": 2.054115716972328e-05,
"loss": 0.1376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10821995139122009,
"step": 2275,
"valid_targets_mean": 2679.1,
"valid_targets_min": 1467
},
{
"epoch": 3.8063439065108513,
"grad_norm": 0.3134251446212264,
"learning_rate": 2.045791809483601e-05,
"loss": 0.1732,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13288769125938416,
"step": 2280,
"valid_targets_mean": 4624.7,
"valid_targets_min": 2210
},
{
"epoch": 3.8146911519198663,
"grad_norm": 0.3668774498799034,
"learning_rate": 2.0374671083004642e-05,
"loss": 0.1608,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1607484668493271,
"step": 2285,
"valid_targets_mean": 3945.0,
"valid_targets_min": 1466
},
{
"epoch": 3.8230383973288813,
"grad_norm": 0.337367861038644,
"learning_rate": 2.0291417577122314e-05,
"loss": 0.1504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16019821166992188,
"step": 2290,
"valid_targets_mean": 5087.5,
"valid_targets_min": 1436
},
{
"epoch": 3.8313856427378967,
"grad_norm": 0.38693093789146166,
"learning_rate": 2.0208159020194734e-05,
"loss": 0.1609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17504851520061493,
"step": 2295,
"valid_targets_mean": 4586.9,
"valid_targets_min": 1365
},
{
"epoch": 3.8397328881469113,
"grad_norm": 0.36800307878803756,
"learning_rate": 2.012489685531515e-05,
"loss": 0.1572,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16387929022312164,
"step": 2300,
"valid_targets_mean": 4192.4,
"valid_targets_min": 1673
},
{
"epoch": 3.8480801335559267,
"grad_norm": 0.3493517983603417,
"learning_rate": 2.0041632525639356e-05,
"loss": 0.1674,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17332419753074646,
"step": 2305,
"valid_targets_mean": 4553.8,
"valid_targets_min": 1550
},
{
"epoch": 3.8564273789649417,
"grad_norm": 0.337680585615126,
"learning_rate": 1.9958367474360648e-05,
"loss": 0.1872,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1315561830997467,
"step": 2310,
"valid_targets_mean": 4040.8,
"valid_targets_min": 1281
},
{
"epoch": 3.8647746243739567,
"grad_norm": 0.3700061141519408,
"learning_rate": 1.9875103144684852e-05,
"loss": 0.1622,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15935084223747253,
"step": 2315,
"valid_targets_mean": 4163.8,
"valid_targets_min": 1503
},
{
"epoch": 3.8731218697829717,
"grad_norm": 0.34528220342283095,
"learning_rate": 1.979184097980527e-05,
"loss": 0.1713,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1959056258201599,
"step": 2320,
"valid_targets_mean": 5238.8,
"valid_targets_min": 1502
},
{
"epoch": 3.8814691151919867,
"grad_norm": 0.3876627732109439,
"learning_rate": 1.9708582422877693e-05,
"loss": 0.158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15763691067695618,
"step": 2325,
"valid_targets_mean": 3863.1,
"valid_targets_min": 1416
},
{
"epoch": 3.8898163606010017,
"grad_norm": 0.43726867923704305,
"learning_rate": 1.9625328916995365e-05,
"loss": 0.1676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22787757217884064,
"step": 2330,
"valid_targets_mean": 4212.8,
"valid_targets_min": 1724
},
{
"epoch": 3.8981636060100167,
"grad_norm": 0.3875585454729919,
"learning_rate": 1.9542081905163997e-05,
"loss": 0.1687,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15151071548461914,
"step": 2335,
"valid_targets_mean": 3840.8,
"valid_targets_min": 1783
},
{
"epoch": 3.9065108514190316,
"grad_norm": 0.3603098861884228,
"learning_rate": 1.9458842830276724e-05,
"loss": 0.1624,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13842938840389252,
"step": 2340,
"valid_targets_mean": 3935.2,
"valid_targets_min": 1469
},
{
"epoch": 3.9148580968280466,
"grad_norm": 0.3720126740283335,
"learning_rate": 1.937561313508913e-05,
"loss": 0.1707,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1920233815908432,
"step": 2345,
"valid_targets_mean": 4504.9,
"valid_targets_min": 1463
},
{
"epoch": 3.9232053422370616,
"grad_norm": 0.37888847999810993,
"learning_rate": 1.9292394262194213e-05,
"loss": 0.1584,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15109741687774658,
"step": 2350,
"valid_targets_mean": 3741.8,
"valid_targets_min": 1503
},
{
"epoch": 3.9315525876460766,
"grad_norm": 0.40056315238474244,
"learning_rate": 1.9209187653997385e-05,
"loss": 0.1596,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19897767901420593,
"step": 2355,
"valid_targets_mean": 3814.2,
"valid_targets_min": 1202
},
{
"epoch": 3.939899833055092,
"grad_norm": 0.35662050739233403,
"learning_rate": 1.91259947526915e-05,
"loss": 0.1906,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19056318700313568,
"step": 2360,
"valid_targets_mean": 4844.5,
"valid_targets_min": 1829
},
{
"epoch": 3.9482470784641066,
"grad_norm": 0.4119479222634502,
"learning_rate": 1.904281700023181e-05,
"loss": 0.2038,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20837408304214478,
"step": 2365,
"valid_targets_mean": 4211.3,
"valid_targets_min": 2359
},
{
"epoch": 3.956594323873122,
"grad_norm": 0.3687246315574956,
"learning_rate": 1.895965583831102e-05,
"loss": 0.1744,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14304903149604797,
"step": 2370,
"valid_targets_mean": 3908.1,
"valid_targets_min": 652
},
{
"epoch": 3.964941569282137,
"grad_norm": 0.35181309528376226,
"learning_rate": 1.887651270833425e-05,
"loss": 0.1581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16806970536708832,
"step": 2375,
"valid_targets_mean": 4548.2,
"valid_targets_min": 1463
},
{
"epoch": 3.973288814691152,
"grad_norm": 0.39553616431684224,
"learning_rate": 1.8793389051394116e-05,
"loss": 0.1622,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17902573943138123,
"step": 2380,
"valid_targets_mean": 4174.2,
"valid_targets_min": 1797
},
{
"epoch": 3.981636060100167,
"grad_norm": 0.35816057257959505,
"learning_rate": 1.8710286308245665e-05,
"loss": 0.1607,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12988346815109253,
"step": 2385,
"valid_targets_mean": 3944.4,
"valid_targets_min": 1516
},
{
"epoch": 3.989983305509182,
"grad_norm": 0.3436951405475913,
"learning_rate": 1.8627205919281507e-05,
"loss": 0.1662,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1546299010515213,
"step": 2390,
"valid_targets_mean": 4097.1,
"valid_targets_min": 1354
},
{
"epoch": 3.998330550918197,
"grad_norm": 0.3836783041847733,
"learning_rate": 1.8544149324506736e-05,
"loss": 0.1836,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16239479184150696,
"step": 2395,
"valid_targets_mean": 3659.1,
"valid_targets_min": 1321
},
{
"epoch": 4.006677796327212,
"grad_norm": 0.3335929558052769,
"learning_rate": 1.8461117963514087e-05,
"loss": 0.1665,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1180664449930191,
"step": 2400,
"valid_targets_mean": 3936.6,
"valid_targets_min": 1484
},
{
"epoch": 4.015025041736227,
"grad_norm": 0.39970812723937327,
"learning_rate": 1.8378113275458877e-05,
"loss": 0.1529,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12729769945144653,
"step": 2405,
"valid_targets_mean": 3840.1,
"valid_targets_min": 1681
},
{
"epoch": 4.023372287145242,
"grad_norm": 0.4118870593178704,
"learning_rate": 1.8295136699034152e-05,
"loss": 0.1471,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1741812378168106,
"step": 2410,
"valid_targets_mean": 3806.8,
"valid_targets_min": 1456
},
{
"epoch": 4.031719532554257,
"grad_norm": 0.3663930032834355,
"learning_rate": 1.8212189672445665e-05,
"loss": 0.1585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14877849817276,
"step": 2415,
"valid_targets_mean": 3980.6,
"valid_targets_min": 1984
},
{
"epoch": 4.040066777963272,
"grad_norm": 0.37493362328257307,
"learning_rate": 1.812927363338703e-05,
"loss": 0.1415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17160674929618835,
"step": 2420,
"valid_targets_mean": 4292.9,
"valid_targets_min": 1637
},
{
"epoch": 4.048414023372287,
"grad_norm": 0.3980479020276196,
"learning_rate": 1.8046390019014738e-05,
"loss": 0.1533,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14769533276557922,
"step": 2425,
"valid_targets_mean": 3776.4,
"valid_targets_min": 1563
},
{
"epoch": 4.056761268781302,
"grad_norm": 0.36476901335237366,
"learning_rate": 1.7963540265923298e-05,
"loss": 0.1667,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16194503009319305,
"step": 2430,
"valid_targets_mean": 4671.4,
"valid_targets_min": 1725
},
{
"epoch": 4.065108514190317,
"grad_norm": 0.38480050224002876,
"learning_rate": 1.788072581012028e-05,
"loss": 0.1608,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16978706419467926,
"step": 2435,
"valid_targets_mean": 4230.8,
"valid_targets_min": 1685
},
{
"epoch": 4.073455759599332,
"grad_norm": 0.35805339347115966,
"learning_rate": 1.779794808700149e-05,
"loss": 0.1644,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14665059745311737,
"step": 2440,
"valid_targets_mean": 4411.9,
"valid_targets_min": 2049
},
{
"epoch": 4.081803005008347,
"grad_norm": 0.43229906387409645,
"learning_rate": 1.7715208531326032e-05,
"loss": 0.1585,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17197956144809723,
"step": 2445,
"valid_targets_mean": 3183.1,
"valid_targets_min": 1496
},
{
"epoch": 4.090150250417362,
"grad_norm": 0.4498930695448525,
"learning_rate": 1.76325085771915e-05,
"loss": 0.177,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2522193193435669,
"step": 2450,
"valid_targets_mean": 4464.3,
"valid_targets_min": 1307
},
{
"epoch": 4.098497495826377,
"grad_norm": 0.4072062438143964,
"learning_rate": 1.7549849658009037e-05,
"loss": 0.1356,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14602962136268616,
"step": 2455,
"valid_targets_mean": 3844.2,
"valid_targets_min": 1983
},
{
"epoch": 4.106844741235392,
"grad_norm": 0.4056776097823568,
"learning_rate": 1.7467233206478597e-05,
"loss": 0.1719,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1761564314365387,
"step": 2460,
"valid_targets_mean": 4014.1,
"valid_targets_min": 1463
},
{
"epoch": 4.115191986644407,
"grad_norm": 0.3643392924990686,
"learning_rate": 1.7384660654564006e-05,
"loss": 0.1483,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1447259485721588,
"step": 2465,
"valid_targets_mean": 3887.2,
"valid_targets_min": 1411
},
{
"epoch": 4.123539232053423,
"grad_norm": 0.3646768430533948,
"learning_rate": 1.7302133433468232e-05,
"loss": 0.1463,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11564427614212036,
"step": 2470,
"valid_targets_mean": 3915.1,
"valid_targets_min": 1552
},
{
"epoch": 4.131886477462437,
"grad_norm": 0.3924072602133538,
"learning_rate": 1.7219652973608512e-05,
"loss": 0.1504,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14637045562267303,
"step": 2475,
"valid_targets_mean": 3627.2,
"valid_targets_min": 718
},
{
"epoch": 4.140233722871453,
"grad_norm": 0.3878311545856039,
"learning_rate": 1.7137220704591605e-05,
"loss": 0.1524,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15386143326759338,
"step": 2480,
"valid_targets_mean": 3538.0,
"valid_targets_min": 1697
},
{
"epoch": 4.148580968280467,
"grad_norm": 0.3880971988947329,
"learning_rate": 1.7054838055188972e-05,
"loss": 0.1483,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1420733630657196,
"step": 2485,
"valid_targets_mean": 3787.9,
"valid_targets_min": 1578
},
{
"epoch": 4.156928213689483,
"grad_norm": 0.395350731356636,
"learning_rate": 1.6972506453312057e-05,
"loss": 0.1564,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11656969785690308,
"step": 2490,
"valid_targets_mean": 3232.8,
"valid_targets_min": 1627
},
{
"epoch": 4.165275459098497,
"grad_norm": 0.37996597000557564,
"learning_rate": 1.6890227325987498e-05,
"loss": 0.1627,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14509516954421997,
"step": 2495,
"valid_targets_mean": 4232.1,
"valid_targets_min": 1730
},
{
"epoch": 4.173622704507513,
"grad_norm": 0.38827833877504675,
"learning_rate": 1.6808002099332422e-05,
"loss": 0.1676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1609729826450348,
"step": 2500,
"valid_targets_mean": 3878.6,
"valid_targets_min": 1839
},
{
"epoch": 4.181969949916527,
"grad_norm": 0.40520273989810945,
"learning_rate": 1.672583219852969e-05,
"loss": 0.1404,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1534803807735443,
"step": 2505,
"valid_targets_mean": 3784.2,
"valid_targets_min": 1530
},
{
"epoch": 4.190317195325543,
"grad_norm": 0.4097370504184848,
"learning_rate": 1.6643719047803243e-05,
"loss": 0.1533,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1326431781053543,
"step": 2510,
"valid_targets_mean": 3492.8,
"valid_targets_min": 1289
},
{
"epoch": 4.198664440734557,
"grad_norm": 0.3759352914481271,
"learning_rate": 1.656166407039339e-05,
"loss": 0.1678,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16762861609458923,
"step": 2515,
"valid_targets_mean": 4399.2,
"valid_targets_min": 1640
},
{
"epoch": 4.207011686143573,
"grad_norm": 0.3943740732741211,
"learning_rate": 1.647966868853212e-05,
"loss": 0.1467,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1504044383764267,
"step": 2520,
"valid_targets_mean": 4008.8,
"valid_targets_min": 1711
},
{
"epoch": 4.215358931552587,
"grad_norm": 0.4599779652429352,
"learning_rate": 1.6397734323418494e-05,
"loss": 0.1543,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14747872948646545,
"step": 2525,
"valid_targets_mean": 3330.1,
"valid_targets_min": 1991
},
{
"epoch": 4.223706176961603,
"grad_norm": 0.4136140040666242,
"learning_rate": 1.6315862395193972e-05,
"loss": 0.1572,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18149696290493011,
"step": 2530,
"valid_targets_mean": 3619.0,
"valid_targets_min": 1595
},
{
"epoch": 4.232053422370618,
"grad_norm": 0.3871962568935494,
"learning_rate": 1.6234054322917835e-05,
"loss": 0.1459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18514098227024078,
"step": 2535,
"valid_targets_mean": 4300.3,
"valid_targets_min": 1572
},
{
"epoch": 4.240400667779633,
"grad_norm": 0.39050587308728724,
"learning_rate": 1.615231152454255e-05,
"loss": 0.1668,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17684586346149445,
"step": 2540,
"valid_targets_mean": 4537.8,
"valid_targets_min": 1628
},
{
"epoch": 4.248747913188648,
"grad_norm": 0.3603847935820327,
"learning_rate": 1.6070635416889228e-05,
"loss": 0.1612,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13764050602912903,
"step": 2545,
"valid_targets_mean": 4523.0,
"valid_targets_min": 1446
},
{
"epoch": 4.257095158597663,
"grad_norm": 0.3703530630291026,
"learning_rate": 1.598902741562304e-05,
"loss": 0.1507,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15130406618118286,
"step": 2550,
"valid_targets_mean": 4445.7,
"valid_targets_min": 1714
},
{
"epoch": 4.265442404006678,
"grad_norm": 0.43139116321778126,
"learning_rate": 1.590748893522871e-05,
"loss": 0.1547,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1296728551387787,
"step": 2555,
"valid_targets_mean": 3089.1,
"valid_targets_min": 1387
},
{
"epoch": 4.273789649415693,
"grad_norm": 0.40689281882892736,
"learning_rate": 1.5826021388985947e-05,
"loss": 0.168,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16642817854881287,
"step": 2560,
"valid_targets_mean": 3739.2,
"valid_targets_min": 1281
},
{
"epoch": 4.282136894824708,
"grad_norm": 0.3844425561379755,
"learning_rate": 1.5744626188945023e-05,
"loss": 0.1646,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12353911995887756,
"step": 2565,
"valid_targets_mean": 3578.9,
"valid_targets_min": 2020
},
{
"epoch": 4.290484140233723,
"grad_norm": 0.4111499887097908,
"learning_rate": 1.5663304745902218e-05,
"loss": 0.1587,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17021812498569489,
"step": 2570,
"valid_targets_mean": 3618.8,
"valid_targets_min": 1283
},
{
"epoch": 4.298831385642738,
"grad_norm": 0.36435902044403873,
"learning_rate": 1.558205846937544e-05,
"loss": 0.1699,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17273738980293274,
"step": 2575,
"valid_targets_mean": 5047.9,
"valid_targets_min": 1622
},
{
"epoch": 4.307178631051753,
"grad_norm": 0.39352844605086823,
"learning_rate": 1.5500888767579722e-05,
"loss": 0.1692,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.22381192445755005,
"step": 2580,
"valid_targets_mean": 4765.4,
"valid_targets_min": 1456
},
{
"epoch": 4.315525876460768,
"grad_norm": 0.36467370352736916,
"learning_rate": 1.5419797047402897e-05,
"loss": 0.1278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12989157438278198,
"step": 2585,
"valid_targets_mean": 4107.6,
"valid_targets_min": 1446
},
{
"epoch": 4.323873121869783,
"grad_norm": 0.366117180070187,
"learning_rate": 1.533878471438113e-05,
"loss": 0.1495,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12719137966632843,
"step": 2590,
"valid_targets_mean": 4216.8,
"valid_targets_min": 1573
},
{
"epoch": 4.332220367278798,
"grad_norm": 0.3520734229205967,
"learning_rate": 1.5257853172674622e-05,
"loss": 0.1691,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17138662934303284,
"step": 2595,
"valid_targets_mean": 4692.9,
"valid_targets_min": 1460
},
{
"epoch": 4.340567612687813,
"grad_norm": 0.41546123687787023,
"learning_rate": 1.517700382504322e-05,
"loss": 0.1788,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18948113918304443,
"step": 2600,
"valid_targets_mean": 4023.4,
"valid_targets_min": 1515
},
{
"epoch": 4.348914858096828,
"grad_norm": 0.3990611092704054,
"learning_rate": 1.5096238072822153e-05,
"loss": 0.1494,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13814587891101837,
"step": 2605,
"valid_targets_mean": 4048.2,
"valid_targets_min": 2286
},
{
"epoch": 4.357262103505843,
"grad_norm": 0.44435565158377754,
"learning_rate": 1.5015557315897693e-05,
"loss": 0.1556,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15524575114250183,
"step": 2610,
"valid_targets_mean": 3159.7,
"valid_targets_min": 1470
},
{
"epoch": 4.365609348914858,
"grad_norm": 0.4137451635408681,
"learning_rate": 1.4934962952682943e-05,
"loss": 0.1609,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16444113850593567,
"step": 2615,
"valid_targets_mean": 3727.5,
"valid_targets_min": 1672
},
{
"epoch": 4.373956594323873,
"grad_norm": 0.3889186979770747,
"learning_rate": 1.4854456380093544e-05,
"loss": 0.1491,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13390201330184937,
"step": 2620,
"valid_targets_mean": 3448.1,
"valid_targets_min": 1550
},
{
"epoch": 4.382303839732888,
"grad_norm": 0.38191745994806764,
"learning_rate": 1.4774038993523523e-05,
"loss": 0.1777,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17715078592300415,
"step": 2625,
"valid_targets_mean": 4832.5,
"valid_targets_min": 1951
},
{
"epoch": 4.390651085141903,
"grad_norm": 0.423760919700067,
"learning_rate": 1.4693712186821039e-05,
"loss": 0.1476,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14304488897323608,
"step": 2630,
"valid_targets_mean": 3519.0,
"valid_targets_min": 1410
},
{
"epoch": 4.398998330550918,
"grad_norm": 0.4097921045469124,
"learning_rate": 1.4613477352264293e-05,
"loss": 0.1791,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16360926628112793,
"step": 2635,
"valid_targets_mean": 3793.3,
"valid_targets_min": 1710
},
{
"epoch": 4.407345575959933,
"grad_norm": 0.4289529976939456,
"learning_rate": 1.4533335880537336e-05,
"loss": 0.153,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17429673671722412,
"step": 2640,
"valid_targets_mean": 3553.5,
"valid_targets_min": 1558
},
{
"epoch": 4.415692821368948,
"grad_norm": 0.3606097258962959,
"learning_rate": 1.4453289160706017e-05,
"loss": 0.1786,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15936686098575592,
"step": 2645,
"valid_targets_mean": 4556.8,
"valid_targets_min": 2091
},
{
"epoch": 4.424040066777963,
"grad_norm": 0.414897016060369,
"learning_rate": 1.4373338580193853e-05,
"loss": 0.1696,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1937766969203949,
"step": 2650,
"valid_targets_mean": 4541.6,
"valid_targets_min": 2041
},
{
"epoch": 4.432387312186978,
"grad_norm": 0.3522501621506027,
"learning_rate": 1.4293485524758045e-05,
"loss": 0.1703,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1663854718208313,
"step": 2655,
"valid_targets_mean": 5143.9,
"valid_targets_min": 1791
},
{
"epoch": 4.440734557595993,
"grad_norm": 0.41929990150960866,
"learning_rate": 1.4213731378465379e-05,
"loss": 0.1532,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16678087413311005,
"step": 2660,
"valid_targets_mean": 3242.9,
"valid_targets_min": 1315
},
{
"epoch": 4.449081803005008,
"grad_norm": 0.34033913328125454,
"learning_rate": 1.4134077523668327e-05,
"loss": 0.1525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13038881123065948,
"step": 2665,
"valid_targets_mean": 4607.9,
"valid_targets_min": 1405
},
{
"epoch": 4.457429048414023,
"grad_norm": 0.36531481988736875,
"learning_rate": 1.405452534098101e-05,
"loss": 0.185,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15257573127746582,
"step": 2670,
"valid_targets_mean": 4429.6,
"valid_targets_min": 1405
},
{
"epoch": 4.465776293823039,
"grad_norm": 0.41248066262763994,
"learning_rate": 1.3975076209255321e-05,
"loss": 0.1713,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1533137559890747,
"step": 2675,
"valid_targets_mean": 3761.3,
"valid_targets_min": 1448
},
{
"epoch": 4.474123539232053,
"grad_norm": 0.38237365506297183,
"learning_rate": 1.3895731505556978e-05,
"loss": 0.1554,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1644631028175354,
"step": 2680,
"valid_targets_mean": 4533.9,
"valid_targets_min": 1354
},
{
"epoch": 4.482470784641069,
"grad_norm": 0.35062952226035243,
"learning_rate": 1.3816492605141712e-05,
"loss": 0.1566,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14511638879776,
"step": 2685,
"valid_targets_mean": 4546.5,
"valid_targets_min": 2198
},
{
"epoch": 4.490818030050083,
"grad_norm": 0.4353906417918772,
"learning_rate": 1.3737360881431374e-05,
"loss": 0.1635,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1806250512599945,
"step": 2690,
"valid_targets_mean": 3671.1,
"valid_targets_min": 1381
},
{
"epoch": 4.499165275459099,
"grad_norm": 0.3784606058447409,
"learning_rate": 1.3658337705990185e-05,
"loss": 0.1448,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14670738577842712,
"step": 2695,
"valid_targets_mean": 3712.1,
"valid_targets_min": 1342
},
{
"epoch": 4.507512520868113,
"grad_norm": 0.38538099131363107,
"learning_rate": 1.3579424448500901e-05,
"loss": 0.1414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1550869643688202,
"step": 2700,
"valid_targets_mean": 3695.3,
"valid_targets_min": 1875
},
{
"epoch": 4.515859766277129,
"grad_norm": 0.3975450030840865,
"learning_rate": 1.3500622476741142e-05,
"loss": 0.1781,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18774588406085968,
"step": 2705,
"valid_targets_mean": 4171.4,
"valid_targets_min": 1593
},
{
"epoch": 4.524207011686143,
"grad_norm": 0.45797382962695116,
"learning_rate": 1.3421933156559605e-05,
"loss": 0.1836,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2451082319021225,
"step": 2710,
"valid_targets_mean": 4464.2,
"valid_targets_min": 1384
},
{
"epoch": 4.532554257095159,
"grad_norm": 0.40694837027937425,
"learning_rate": 1.3343357851852488e-05,
"loss": 0.1704,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1796131730079651,
"step": 2715,
"valid_targets_mean": 3673.2,
"valid_targets_min": 1405
},
{
"epoch": 4.540901502504173,
"grad_norm": 0.36470277976309823,
"learning_rate": 1.3264897924539746e-05,
"loss": 0.1386,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14634127914905548,
"step": 2720,
"valid_targets_mean": 4269.3,
"valid_targets_min": 1375
},
{
"epoch": 4.549248747913189,
"grad_norm": 0.42775983397043676,
"learning_rate": 1.3186554734541574e-05,
"loss": 0.1665,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14799463748931885,
"step": 2725,
"valid_targets_mean": 3281.6,
"valid_targets_min": 1471
},
{
"epoch": 4.557595993322204,
"grad_norm": 0.37778486150135737,
"learning_rate": 1.3108329639754765e-05,
"loss": 0.1861,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15953314304351807,
"step": 2730,
"valid_targets_mean": 4123.1,
"valid_targets_min": 1677
},
{
"epoch": 4.565943238731219,
"grad_norm": 0.4011991994515831,
"learning_rate": 1.3030223996029237e-05,
"loss": 0.1567,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14625924825668335,
"step": 2735,
"valid_targets_mean": 3672.7,
"valid_targets_min": 1362
},
{
"epoch": 4.574290484140234,
"grad_norm": 0.4049534068315027,
"learning_rate": 1.295223915714447e-05,
"loss": 0.1706,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1741216480731964,
"step": 2740,
"valid_targets_mean": 4030.8,
"valid_targets_min": 1531
},
{
"epoch": 4.582637729549249,
"grad_norm": 0.3750960930861452,
"learning_rate": 1.287437647478611e-05,
"loss": 0.1606,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17940756678581238,
"step": 2745,
"valid_targets_mean": 5074.7,
"valid_targets_min": 1384
},
{
"epoch": 4.590984974958264,
"grad_norm": 0.38183642599914125,
"learning_rate": 1.2796637298522466e-05,
"loss": 0.1415,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13499857485294342,
"step": 2750,
"valid_targets_mean": 3520.5,
"valid_targets_min": 1494
},
{
"epoch": 4.599332220367279,
"grad_norm": 0.381406722988372,
"learning_rate": 1.271902297578119e-05,
"loss": 0.1523,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17434316873550415,
"step": 2755,
"valid_targets_mean": 4478.9,
"valid_targets_min": 1445
},
{
"epoch": 4.607679465776294,
"grad_norm": 0.37666672741716467,
"learning_rate": 1.2641534851825865e-05,
"loss": 0.157,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19380509853363037,
"step": 2760,
"valid_targets_mean": 4799.0,
"valid_targets_min": 1578
},
{
"epoch": 4.616026711185309,
"grad_norm": 0.37625310958958696,
"learning_rate": 1.256417426973272e-05,
"loss": 0.1613,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1542454957962036,
"step": 2765,
"valid_targets_mean": 4189.5,
"valid_targets_min": 1931
},
{
"epoch": 4.624373956594324,
"grad_norm": 0.41026835470638295,
"learning_rate": 1.248694257036735e-05,
"loss": 0.1459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15828590095043182,
"step": 2770,
"valid_targets_mean": 3639.0,
"valid_targets_min": 1592
},
{
"epoch": 4.632721202003339,
"grad_norm": 0.4207161762031595,
"learning_rate": 1.2409841092361457e-05,
"loss": 0.1694,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1780978888273239,
"step": 2775,
"valid_targets_mean": 3900.4,
"valid_targets_min": 1389
},
{
"epoch": 4.641068447412354,
"grad_norm": 0.4385093176125206,
"learning_rate": 1.233287117208966e-05,
"loss": 0.1977,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12772172689437866,
"step": 2780,
"valid_targets_mean": 2653.9,
"valid_targets_min": 1383
},
{
"epoch": 4.649415692821369,
"grad_norm": 0.3715540522713055,
"learning_rate": 1.2256034143646341e-05,
"loss": 0.188,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1914878785610199,
"step": 2785,
"valid_targets_mean": 4993.1,
"valid_targets_min": 1428
},
{
"epoch": 4.657762938230384,
"grad_norm": 0.4028481623543887,
"learning_rate": 1.2179331338822498e-05,
"loss": 0.1518,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17315779626369476,
"step": 2790,
"valid_targets_mean": 4060.1,
"valid_targets_min": 1524
},
{
"epoch": 4.666110183639399,
"grad_norm": 0.4041436958277061,
"learning_rate": 1.2102764087082685e-05,
"loss": 0.171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14328855276107788,
"step": 2795,
"valid_targets_mean": 3515.2,
"valid_targets_min": 1463
},
{
"epoch": 4.674457429048414,
"grad_norm": 0.37911549491366703,
"learning_rate": 1.2026333715541959e-05,
"loss": 0.1538,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13919568061828613,
"step": 2800,
"valid_targets_mean": 4206.1,
"valid_targets_min": 1367
},
{
"epoch": 4.682804674457429,
"grad_norm": 0.44419852779883967,
"learning_rate": 1.1950041548942867e-05,
"loss": 0.1661,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15678191184997559,
"step": 2805,
"valid_targets_mean": 3685.3,
"valid_targets_min": 1384
},
{
"epoch": 4.691151919866444,
"grad_norm": 0.40485894326217103,
"learning_rate": 1.1873888909632508e-05,
"loss": 0.1429,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16913247108459473,
"step": 2810,
"valid_targets_mean": 4086.8,
"valid_targets_min": 2246
},
{
"epoch": 4.699499165275459,
"grad_norm": 0.3791170891554961,
"learning_rate": 1.1797877117539587e-05,
"loss": 0.1557,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13671991229057312,
"step": 2815,
"valid_targets_mean": 3812.2,
"valid_targets_min": 1252
},
{
"epoch": 4.707846410684474,
"grad_norm": 0.3870882935371369,
"learning_rate": 1.1722007490151566e-05,
"loss": 0.1494,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1452493816614151,
"step": 2820,
"valid_targets_mean": 3973.8,
"valid_targets_min": 1644
},
{
"epoch": 4.716193656093489,
"grad_norm": 0.41178144995102794,
"learning_rate": 1.16462813424918e-05,
"loss": 0.1558,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13517948985099792,
"step": 2825,
"valid_targets_mean": 3647.9,
"valid_targets_min": 1658
},
{
"epoch": 4.724540901502504,
"grad_norm": 0.3765024026231588,
"learning_rate": 1.1570699987096763e-05,
"loss": 0.1516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18162378668785095,
"step": 2830,
"valid_targets_mean": 4538.9,
"valid_targets_min": 1625
},
{
"epoch": 4.732888146911519,
"grad_norm": 0.3760941628267728,
"learning_rate": 1.1495264733993288e-05,
"loss": 0.1582,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14754658937454224,
"step": 2835,
"valid_targets_mean": 4384.1,
"valid_targets_min": 1869
},
{
"epoch": 4.741235392320534,
"grad_norm": 0.39279266470665364,
"learning_rate": 1.141997689067587e-05,
"loss": 0.1724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.165314182639122,
"step": 2840,
"valid_targets_mean": 3989.9,
"valid_targets_min": 1345
},
{
"epoch": 4.749582637729549,
"grad_norm": 0.34882964633390773,
"learning_rate": 1.1344837762083997e-05,
"loss": 0.1735,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2033419907093048,
"step": 2845,
"valid_targets_mean": 5947.7,
"valid_targets_min": 1775
},
{
"epoch": 4.757929883138564,
"grad_norm": 0.4023623335563106,
"learning_rate": 1.1269848650579532e-05,
"loss": 0.1477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16237092018127441,
"step": 2850,
"valid_targets_mean": 3687.1,
"valid_targets_min": 1419
},
{
"epoch": 4.766277128547579,
"grad_norm": 0.37164989172432883,
"learning_rate": 1.1195010855924138e-05,
"loss": 0.1446,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1555103361606598,
"step": 2855,
"valid_targets_mean": 4158.9,
"valid_targets_min": 770
},
{
"epoch": 4.774624373956595,
"grad_norm": 0.3546257711362022,
"learning_rate": 1.112032567525676e-05,
"loss": 0.1483,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1440112292766571,
"step": 2860,
"valid_targets_mean": 4531.6,
"valid_targets_min": 1540
},
{
"epoch": 4.782971619365609,
"grad_norm": 0.3883480444034267,
"learning_rate": 1.1045794403071133e-05,
"loss": 0.1474,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11662937700748444,
"step": 2865,
"valid_targets_mean": 3583.9,
"valid_targets_min": 1485
},
{
"epoch": 4.791318864774624,
"grad_norm": 0.42521570723239477,
"learning_rate": 1.0971418331193337e-05,
"loss": 0.1645,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1642211526632309,
"step": 2870,
"valid_targets_mean": 3930.2,
"valid_targets_min": 1523
},
{
"epoch": 4.799666110183639,
"grad_norm": 0.37658945198790456,
"learning_rate": 1.0897198748759435e-05,
"loss": 0.1657,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16514746844768524,
"step": 2875,
"valid_targets_mean": 5020.5,
"valid_targets_min": 1631
},
{
"epoch": 4.808013355592655,
"grad_norm": 0.4065744794645929,
"learning_rate": 1.0823136942193089e-05,
"loss": 0.1533,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1544893980026245,
"step": 2880,
"valid_targets_mean": 3494.4,
"valid_targets_min": 1475
},
{
"epoch": 4.816360601001669,
"grad_norm": 0.4004861082508581,
"learning_rate": 1.07492341951833e-05,
"loss": 0.1672,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.165223628282547,
"step": 2885,
"valid_targets_mean": 4314.1,
"valid_targets_min": 1512
},
{
"epoch": 4.824707846410685,
"grad_norm": 0.37881093175793823,
"learning_rate": 1.0675491788662132e-05,
"loss": 0.1566,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15597029030323029,
"step": 2890,
"valid_targets_mean": 4086.5,
"valid_targets_min": 1819
},
{
"epoch": 4.833055091819699,
"grad_norm": 0.4164476162005396,
"learning_rate": 1.060191100078252e-05,
"loss": 0.1433,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1591774970293045,
"step": 2895,
"valid_targets_mean": 3379.1,
"valid_targets_min": 1702
},
{
"epoch": 4.841402337228715,
"grad_norm": 0.40505211425684223,
"learning_rate": 1.0528493106896126e-05,
"loss": 0.1653,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1540907472372055,
"step": 2900,
"valid_targets_mean": 3753.1,
"valid_targets_min": 871
},
{
"epoch": 4.849749582637729,
"grad_norm": 0.4028261281101974,
"learning_rate": 1.0455239379531213e-05,
"loss": 0.1512,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12867653369903564,
"step": 2905,
"valid_targets_mean": 3385.2,
"valid_targets_min": 1413
},
{
"epoch": 4.858096828046745,
"grad_norm": 0.35413368278335405,
"learning_rate": 1.0382151088370605e-05,
"loss": 0.1706,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15035101771354675,
"step": 2910,
"valid_targets_mean": 4878.9,
"valid_targets_min": 1367
},
{
"epoch": 4.866444073455759,
"grad_norm": 0.4417905948250191,
"learning_rate": 1.0309229500229665e-05,
"loss": 0.1659,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.189832866191864,
"step": 2915,
"valid_targets_mean": 3758.6,
"valid_targets_min": 1363
},
{
"epoch": 4.874791318864775,
"grad_norm": 0.3703216309586733,
"learning_rate": 1.023647587903438e-05,
"loss": 0.1445,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13962072134017944,
"step": 2920,
"valid_targets_mean": 3970.6,
"valid_targets_min": 1801
},
{
"epoch": 4.883138564273789,
"grad_norm": 0.3923032376215046,
"learning_rate": 1.0163891485799362e-05,
"loss": 0.1729,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1556883156299591,
"step": 2925,
"valid_targets_mean": 4068.7,
"valid_targets_min": 1692
},
{
"epoch": 4.891485809682805,
"grad_norm": 0.3943816007244313,
"learning_rate": 1.0091477578606121e-05,
"loss": 0.1556,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1305852085351944,
"step": 2930,
"valid_targets_mean": 3647.4,
"valid_targets_min": 1738
},
{
"epoch": 4.89983305509182,
"grad_norm": 0.4109406376130915,
"learning_rate": 1.0019235412581117e-05,
"loss": 0.1542,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17126506567001343,
"step": 2935,
"valid_targets_mean": 4006.1,
"valid_targets_min": 2178
},
{
"epoch": 4.908180300500835,
"grad_norm": 0.3495984561187602,
"learning_rate": 9.947166239874144e-06,
"loss": 0.1428,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14328116178512573,
"step": 2940,
"valid_targets_mean": 4676.8,
"valid_targets_min": 1946
},
{
"epoch": 4.91652754590985,
"grad_norm": 0.4025251801883529,
"learning_rate": 9.875271309636498e-06,
"loss": 0.1397,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14199863374233246,
"step": 2945,
"valid_targets_mean": 3569.0,
"valid_targets_min": 1354
},
{
"epoch": 4.924874791318865,
"grad_norm": 0.4279379142265408,
"learning_rate": 9.803551867999445e-06,
"loss": 0.15,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16999608278274536,
"step": 2950,
"valid_targets_mean": 3881.8,
"valid_targets_min": 1580
},
{
"epoch": 4.93322203672788,
"grad_norm": 0.37942793699499083,
"learning_rate": 9.732009158052508e-06,
"loss": 0.1556,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1422589123249054,
"step": 2955,
"valid_targets_mean": 3960.1,
"valid_targets_min": 1455
},
{
"epoch": 4.941569282136895,
"grad_norm": 0.5274164000830702,
"learning_rate": 9.660644419822037e-06,
"loss": 0.1613,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1878030002117157,
"step": 2960,
"valid_targets_mean": 4389.6,
"valid_targets_min": 1244
},
{
"epoch": 4.94991652754591,
"grad_norm": 0.4202707954271607,
"learning_rate": 9.589458890249595e-06,
"loss": 0.1692,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20435276627540588,
"step": 2965,
"valid_targets_mean": 4238.9,
"valid_targets_min": 1801
},
{
"epoch": 4.958263772954925,
"grad_norm": 0.42276135511100665,
"learning_rate": 9.518453803170637e-06,
"loss": 0.195,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20852920413017273,
"step": 2970,
"valid_targets_mean": 4104.8,
"valid_targets_min": 1526
},
{
"epoch": 4.96661101836394,
"grad_norm": 0.3754044527703142,
"learning_rate": 9.447630389293017e-06,
"loss": 0.1682,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12757155299186707,
"step": 2975,
"valid_targets_mean": 3852.2,
"valid_targets_min": 1466
},
{
"epoch": 4.974958263772955,
"grad_norm": 0.413775387764926,
"learning_rate": 9.37698987617576e-06,
"loss": 0.1344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10741505026817322,
"step": 2980,
"valid_targets_mean": 2660.2,
"valid_targets_min": 1364
},
{
"epoch": 4.98330550918197,
"grad_norm": 0.3979422390247439,
"learning_rate": 9.306533488207671e-06,
"loss": 0.1759,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17165054380893707,
"step": 2985,
"valid_targets_mean": 4162.2,
"valid_targets_min": 1581
},
{
"epoch": 4.9916527545909855,
"grad_norm": 0.40526883809594466,
"learning_rate": 9.236262446586239e-06,
"loss": 0.1437,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13515013456344604,
"step": 2990,
"valid_targets_mean": 3149.9,
"valid_targets_min": 1640
},
{
"epoch": 5.0,
"grad_norm": 0.3918834047017886,
"learning_rate": 9.166177969296343e-06,
"loss": 0.1479,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1336270272731781,
"step": 2995,
"valid_targets_mean": 3621.7,
"valid_targets_min": 1305
},
{
"epoch": 5.008347245409015,
"grad_norm": 0.32439992950178825,
"learning_rate": 9.096281271089264e-06,
"loss": 0.1495,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11996345967054367,
"step": 3000,
"valid_targets_mean": 4934.7,
"valid_targets_min": 1375
},
{
"epoch": 5.01669449081803,
"grad_norm": 0.40389904857806536,
"learning_rate": 9.02657356346151e-06,
"loss": 0.149,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16572311520576477,
"step": 3005,
"valid_targets_mean": 3859.5,
"valid_targets_min": 1376
},
{
"epoch": 5.025041736227045,
"grad_norm": 0.44339870180440955,
"learning_rate": 8.957056054633934e-06,
"loss": 0.1699,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13225135207176208,
"step": 3010,
"valid_targets_mean": 3370.6,
"valid_targets_min": 1640
},
{
"epoch": 5.03338898163606,
"grad_norm": 0.3797029496443868,
"learning_rate": 8.887729949530682e-06,
"loss": 0.1402,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10368698835372925,
"step": 3015,
"valid_targets_mean": 3671.1,
"valid_targets_min": 2133
},
{
"epoch": 5.041736227045075,
"grad_norm": 0.4018368181289492,
"learning_rate": 8.818596449758416e-06,
"loss": 0.1424,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12796178460121155,
"step": 3020,
"valid_targets_mean": 3804.5,
"valid_targets_min": 1488
},
{
"epoch": 5.05008347245409,
"grad_norm": 0.45529201782475176,
"learning_rate": 8.749656753585379e-06,
"loss": 0.158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15354162454605103,
"step": 3025,
"valid_targets_mean": 3396.1,
"valid_targets_min": 1455
},
{
"epoch": 5.058430717863105,
"grad_norm": 0.38433100778083057,
"learning_rate": 8.680912055920734e-06,
"loss": 0.1663,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1784001588821411,
"step": 3030,
"valid_targets_mean": 4655.1,
"valid_targets_min": 652
},
{
"epoch": 5.06677796327212,
"grad_norm": 0.3796123121769484,
"learning_rate": 8.612363548293744e-06,
"loss": 0.1434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11050169169902802,
"step": 3035,
"valid_targets_mean": 3595.1,
"valid_targets_min": 1793
},
{
"epoch": 5.075125208681135,
"grad_norm": 0.4646411179454276,
"learning_rate": 8.54401241883322e-06,
"loss": 0.1755,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.24500685930252075,
"step": 3040,
"valid_targets_mean": 4130.9,
"valid_targets_min": 1558
},
{
"epoch": 5.08347245409015,
"grad_norm": 0.3752728790933242,
"learning_rate": 8.475859852246854e-06,
"loss": 0.1436,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13120922446250916,
"step": 3045,
"valid_targets_mean": 4196.7,
"valid_targets_min": 1363
},
{
"epoch": 5.091819699499165,
"grad_norm": 0.4102518462062466,
"learning_rate": 8.407907029800732e-06,
"loss": 0.1639,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16653870046138763,
"step": 3050,
"valid_targets_mean": 3736.2,
"valid_targets_min": 1427
},
{
"epoch": 5.10016694490818,
"grad_norm": 0.41421937881275267,
"learning_rate": 8.340155129298824e-06,
"loss": 0.1564,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12992340326309204,
"step": 3055,
"valid_targets_mean": 3180.9,
"valid_targets_min": 1569
},
{
"epoch": 5.108514190317195,
"grad_norm": 0.3869634447599984,
"learning_rate": 8.272605325062595e-06,
"loss": 0.1602,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.147879958152771,
"step": 3060,
"valid_targets_mean": 4217.6,
"valid_targets_min": 1604
},
{
"epoch": 5.116861435726211,
"grad_norm": 0.38712230267068287,
"learning_rate": 8.205258787910636e-06,
"loss": 0.1516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13869859278202057,
"step": 3065,
"valid_targets_mean": 4031.9,
"valid_targets_min": 1597
},
{
"epoch": 5.125208681135225,
"grad_norm": 0.40907673079512274,
"learning_rate": 8.138116685138386e-06,
"loss": 0.1645,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16340908408164978,
"step": 3070,
"valid_targets_mean": 4099.2,
"valid_targets_min": 1307
},
{
"epoch": 5.133555926544241,
"grad_norm": 0.4472231144076031,
"learning_rate": 8.07118018049788e-06,
"loss": 0.1489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16141530871391296,
"step": 3075,
"valid_targets_mean": 3591.7,
"valid_targets_min": 2174
},
{
"epoch": 5.141903171953255,
"grad_norm": 0.4101929094668088,
"learning_rate": 8.00445043417759e-06,
"loss": 0.1442,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1607975959777832,
"step": 3080,
"valid_targets_mean": 3870.1,
"valid_targets_min": 1427
},
{
"epoch": 5.150250417362271,
"grad_norm": 0.4190558156655846,
"learning_rate": 7.93792860278232e-06,
"loss": 0.1445,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1311708390712738,
"step": 3085,
"valid_targets_mean": 3223.5,
"valid_targets_min": 1746
},
{
"epoch": 5.158597662771285,
"grad_norm": 0.43131893197016713,
"learning_rate": 7.871615839313147e-06,
"loss": 0.1477,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1534000039100647,
"step": 3090,
"valid_targets_mean": 3763.9,
"valid_targets_min": 1297
},
{
"epoch": 5.166944908180301,
"grad_norm": 0.40171418915149937,
"learning_rate": 7.805513293147441e-06,
"loss": 0.1267,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11731885373592377,
"step": 3095,
"valid_targets_mean": 3959.8,
"valid_targets_min": 2228
},
{
"epoch": 5.175292153589315,
"grad_norm": 0.4038598721263698,
"learning_rate": 7.739622110018951e-06,
"loss": 0.1887,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1757291555404663,
"step": 3100,
"valid_targets_mean": 4081.9,
"valid_targets_min": 2201
},
{
"epoch": 5.183639398998331,
"grad_norm": 0.3981173394009251,
"learning_rate": 7.673943431997935e-06,
"loss": 0.169,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17575743794441223,
"step": 3105,
"valid_targets_mean": 4207.0,
"valid_targets_min": 1720
},
{
"epoch": 5.191986644407345,
"grad_norm": 0.3881779223429235,
"learning_rate": 7.608478397471366e-06,
"loss": 0.1533,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1637006551027298,
"step": 3110,
"valid_targets_mean": 4346.9,
"valid_targets_min": 1632
},
{
"epoch": 5.200333889816361,
"grad_norm": 0.4259696227234815,
"learning_rate": 7.543228141123217e-06,
"loss": 0.149,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18132896721363068,
"step": 3115,
"valid_targets_mean": 3897.2,
"valid_targets_min": 1941
},
{
"epoch": 5.208681135225375,
"grad_norm": 0.367733074815043,
"learning_rate": 7.478193793914767e-06,
"loss": 0.1707,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14233815670013428,
"step": 3120,
"valid_targets_mean": 4717.5,
"valid_targets_min": 1532
},
{
"epoch": 5.217028380634391,
"grad_norm": 0.44111655062600313,
"learning_rate": 7.4133764830650246e-06,
"loss": 0.1583,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12742379307746887,
"step": 3125,
"valid_targets_mean": 2968.4,
"valid_targets_min": 1183
},
{
"epoch": 5.225375626043405,
"grad_norm": 0.43835279453053316,
"learning_rate": 7.348777332031168e-06,
"loss": 0.1765,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20351770520210266,
"step": 3130,
"valid_targets_mean": 4297.8,
"valid_targets_min": 1387
},
{
"epoch": 5.233722871452421,
"grad_norm": 0.38243043621304884,
"learning_rate": 7.28439746048909e-06,
"loss": 0.1518,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13077859580516815,
"step": 3135,
"valid_targets_mean": 4151.1,
"valid_targets_min": 1421
},
{
"epoch": 5.242070116861436,
"grad_norm": 0.4235540092170422,
"learning_rate": 7.22023798431398e-06,
"loss": 0.1488,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12467189878225327,
"step": 3140,
"valid_targets_mean": 3365.2,
"valid_targets_min": 1613
},
{
"epoch": 5.250417362270451,
"grad_norm": 0.43814595599937917,
"learning_rate": 7.156300015560993e-06,
"loss": 0.1637,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1741735339164734,
"step": 3145,
"valid_targets_mean": 4143.1,
"valid_targets_min": 1318
},
{
"epoch": 5.258764607679466,
"grad_norm": 0.3781697218471759,
"learning_rate": 7.092584662445958e-06,
"loss": 0.1466,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14791350066661835,
"step": 3150,
"valid_targets_mean": 4791.8,
"valid_targets_min": 1309
},
{
"epoch": 5.267111853088481,
"grad_norm": 0.3945710993506908,
"learning_rate": 7.029093029326191e-06,
"loss": 0.1459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14302855730056763,
"step": 3155,
"valid_targets_mean": 4068.8,
"valid_targets_min": 1598
},
{
"epoch": 5.275459098497496,
"grad_norm": 0.3972442868106146,
"learning_rate": 6.965826216681337e-06,
"loss": 0.1509,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14088527858257294,
"step": 3160,
"valid_targets_mean": 3831.9,
"valid_targets_min": 1538
},
{
"epoch": 5.283806343906511,
"grad_norm": 0.41359527058167767,
"learning_rate": 6.902785321094301e-06,
"loss": 0.1643,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13359525799751282,
"step": 3165,
"valid_targets_mean": 3728.0,
"valid_targets_min": 1529
},
{
"epoch": 5.292153589315526,
"grad_norm": 0.3681919637113868,
"learning_rate": 6.8399714352322424e-06,
"loss": 0.158,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17469653487205505,
"step": 3170,
"valid_targets_mean": 5309.1,
"valid_targets_min": 2356
},
{
"epoch": 5.300500834724541,
"grad_norm": 0.3893872294091121,
"learning_rate": 6.777385647827639e-06,
"loss": 0.1559,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15479931235313416,
"step": 3175,
"valid_targets_mean": 4566.4,
"valid_targets_min": 1408
},
{
"epoch": 5.308848080133556,
"grad_norm": 0.4699750608152581,
"learning_rate": 6.715029043659409e-06,
"loss": 0.1724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1684361696243286,
"step": 3180,
"valid_targets_mean": 3349.6,
"valid_targets_min": 1285
},
{
"epoch": 5.317195325542571,
"grad_norm": 0.3915156438932572,
"learning_rate": 6.652902703534114e-06,
"loss": 0.1483,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17017066478729248,
"step": 3185,
"valid_targets_mean": 4627.0,
"valid_targets_min": 1466
},
{
"epoch": 5.325542570951586,
"grad_norm": 0.40902369746588285,
"learning_rate": 6.5910077042672246e-06,
"loss": 0.1332,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12037716805934906,
"step": 3190,
"valid_targets_mean": 3666.2,
"valid_targets_min": 1436
},
{
"epoch": 5.333889816360601,
"grad_norm": 0.41335450620488573,
"learning_rate": 6.5293451186644566e-06,
"loss": 0.163,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17258934676647186,
"step": 3195,
"valid_targets_mean": 4244.1,
"valid_targets_min": 1417
},
{
"epoch": 5.342237061769616,
"grad_norm": 0.3848965371099251,
"learning_rate": 6.467916015503173e-06,
"loss": 0.1491,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15709856152534485,
"step": 3200,
"valid_targets_mean": 4322.3,
"valid_targets_min": 1920
},
{
"epoch": 5.350584307178631,
"grad_norm": 0.4064597821460649,
"learning_rate": 6.406721459513865e-06,
"loss": 0.1363,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13367027044296265,
"step": 3205,
"valid_targets_mean": 3733.5,
"valid_targets_min": 1362
},
{
"epoch": 5.358931552587646,
"grad_norm": 0.3725279320715027,
"learning_rate": 6.3457625113616995e-06,
"loss": 0.1448,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15729841589927673,
"step": 3210,
"valid_targets_mean": 5158.6,
"valid_targets_min": 2253
},
{
"epoch": 5.367278797996661,
"grad_norm": 0.5093284889857002,
"learning_rate": 6.2850402276281184e-06,
"loss": 0.1439,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16633614897727966,
"step": 3215,
"valid_targets_mean": 4270.2,
"valid_targets_min": 1442
},
{
"epoch": 5.375626043405676,
"grad_norm": 0.4287843753806089,
"learning_rate": 6.224555660792546e-06,
"loss": 0.1505,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1283961534500122,
"step": 3220,
"valid_targets_mean": 3212.4,
"valid_targets_min": 1393
},
{
"epoch": 5.383973288814691,
"grad_norm": 0.4387165321569552,
"learning_rate": 6.164309859214135e-06,
"loss": 0.1391,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20179302990436554,
"step": 3225,
"valid_targets_mean": 3816.3,
"valid_targets_min": 1506
},
{
"epoch": 5.392320534223706,
"grad_norm": 0.44177943202524567,
"learning_rate": 6.104303867113599e-06,
"loss": 0.136,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1586320400238037,
"step": 3230,
"valid_targets_mean": 3924.4,
"valid_targets_min": 1330
},
{
"epoch": 5.400667779632721,
"grad_norm": 0.4394515949810511,
"learning_rate": 6.044538724555109e-06,
"loss": 0.1412,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19394145905971527,
"step": 3235,
"valid_targets_mean": 3948.6,
"valid_targets_min": 1500
},
{
"epoch": 5.409015025041736,
"grad_norm": 0.338464867824139,
"learning_rate": 5.9850154674282766e-06,
"loss": 0.1757,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13745301961898804,
"step": 3240,
"valid_targets_mean": 4968.6,
"valid_targets_min": 2094
},
{
"epoch": 5.417362270450751,
"grad_norm": 0.4609810654309501,
"learning_rate": 5.925735127430186e-06,
"loss": 0.1683,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2042551189661026,
"step": 3245,
"valid_targets_mean": 3469.8,
"valid_targets_min": 1385
},
{
"epoch": 5.425709515859766,
"grad_norm": 0.3847064604219628,
"learning_rate": 5.866698732047522e-06,
"loss": 0.1413,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13322336971759796,
"step": 3250,
"valid_targets_mean": 4107.9,
"valid_targets_min": 1486
},
{
"epoch": 5.434056761268781,
"grad_norm": 0.40932145408526377,
"learning_rate": 5.8079073045387555e-06,
"loss": 0.1839,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2029389590024948,
"step": 3255,
"valid_targets_mean": 5278.1,
"valid_targets_min": 1367
},
{
"epoch": 5.442404006677796,
"grad_norm": 0.3880117364244024,
"learning_rate": 5.7493618639164115e-06,
"loss": 0.139,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11927919089794159,
"step": 3260,
"valid_targets_mean": 3425.2,
"valid_targets_min": 1550
},
{
"epoch": 5.450751252086811,
"grad_norm": 0.4249305653688245,
"learning_rate": 5.691063424929406e-06,
"loss": 0.1472,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1348227858543396,
"step": 3265,
"valid_targets_mean": 3846.4,
"valid_targets_min": 1449
},
{
"epoch": 5.459098497495827,
"grad_norm": 0.41479994105913315,
"learning_rate": 5.633012998045451e-06,
"loss": 0.1628,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21308234333992004,
"step": 3270,
"valid_targets_mean": 4622.9,
"valid_targets_min": 1656
},
{
"epoch": 5.467445742904841,
"grad_norm": 0.3698582714281757,
"learning_rate": 5.575211589433554e-06,
"loss": 0.1399,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.143967866897583,
"step": 3275,
"valid_targets_mean": 4229.7,
"valid_targets_min": 1272
},
{
"epoch": 5.475792988313857,
"grad_norm": 0.41612733729463747,
"learning_rate": 5.517660200946562e-06,
"loss": 0.1456,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15302608907222748,
"step": 3280,
"valid_targets_mean": 4032.1,
"valid_targets_min": 1515
},
{
"epoch": 5.484140233722871,
"grad_norm": 0.3934374273584147,
"learning_rate": 5.4603598301038145e-06,
"loss": 0.1561,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13985675573349,
"step": 3285,
"valid_targets_mean": 4294.1,
"valid_targets_min": 2234
},
{
"epoch": 5.492487479131887,
"grad_norm": 0.44993403517082425,
"learning_rate": 5.4033114700738375e-06,
"loss": 0.1373,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1476699858903885,
"step": 3290,
"valid_targets_mean": 3344.4,
"valid_targets_min": 1481
},
{
"epoch": 5.500834724540901,
"grad_norm": 0.3971533359449821,
"learning_rate": 5.346516109657136e-06,
"loss": 0.1432,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15139487385749817,
"step": 3295,
"valid_targets_mean": 4088.6,
"valid_targets_min": 1783
},
{
"epoch": 5.509181969949917,
"grad_norm": 0.3879355397956596,
"learning_rate": 5.289974733269063e-06,
"loss": 0.131,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13137929141521454,
"step": 3300,
"valid_targets_mean": 4284.2,
"valid_targets_min": 1392
},
{
"epoch": 5.517529215358931,
"grad_norm": 0.3785388044176155,
"learning_rate": 5.233688320922741e-06,
"loss": 0.1386,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13066166639328003,
"step": 3305,
"valid_targets_mean": 3775.4,
"valid_targets_min": 1483
},
{
"epoch": 5.525876460767947,
"grad_norm": 0.4453230344364764,
"learning_rate": 5.177657848212092e-06,
"loss": 0.1417,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12990698218345642,
"step": 3310,
"valid_targets_mean": 3210.2,
"valid_targets_min": 1381
},
{
"epoch": 5.534223706176961,
"grad_norm": 0.4132685547022122,
"learning_rate": 5.1218842862949115e-06,
"loss": 0.1473,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1465921849012375,
"step": 3315,
"valid_targets_mean": 4035.1,
"valid_targets_min": 1047
},
{
"epoch": 5.542570951585977,
"grad_norm": 0.4232032849976099,
"learning_rate": 5.066368601876048e-06,
"loss": 0.1655,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1861027479171753,
"step": 3320,
"valid_targets_mean": 3954.4,
"valid_targets_min": 1439
},
{
"epoch": 5.550918196994992,
"grad_norm": 0.4393672219297928,
"learning_rate": 5.011111757190661e-06,
"loss": 0.1454,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12021392583847046,
"step": 3325,
"valid_targets_mean": 3101.6,
"valid_targets_min": 1625
},
{
"epoch": 5.559265442404007,
"grad_norm": 0.3790793737380855,
"learning_rate": 4.956114709987488e-06,
"loss": 0.1603,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15145833790302277,
"step": 3330,
"valid_targets_mean": 4378.4,
"valid_targets_min": 1801
},
{
"epoch": 5.567612687813021,
"grad_norm": 0.4389137151970133,
"learning_rate": 4.901378413512325e-06,
"loss": 0.162,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2078292965888977,
"step": 3335,
"valid_targets_mean": 4065.6,
"valid_targets_min": 1665
},
{
"epoch": 5.575959933222037,
"grad_norm": 0.3893559030740735,
"learning_rate": 4.846903816491419e-06,
"loss": 0.1329,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.10496833175420761,
"step": 3340,
"valid_targets_mean": 3719.2,
"valid_targets_min": 1491
},
{
"epoch": 5.584307178631052,
"grad_norm": 0.3729039680176399,
"learning_rate": 4.792691863115113e-06,
"loss": 0.1389,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1308392584323883,
"step": 3345,
"valid_targets_mean": 4537.8,
"valid_targets_min": 1432
},
{
"epoch": 5.592654424040067,
"grad_norm": 0.3657430370204413,
"learning_rate": 4.738743493021383e-06,
"loss": 0.1557,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13987523317337036,
"step": 3350,
"valid_targets_mean": 4376.2,
"valid_targets_min": 1838
},
{
"epoch": 5.601001669449082,
"grad_norm": 0.39381748640941305,
"learning_rate": 4.68505964127965e-06,
"loss": 0.1494,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12407132983207703,
"step": 3355,
"valid_targets_mean": 4120.6,
"valid_targets_min": 1636
},
{
"epoch": 5.609348914858097,
"grad_norm": 0.4388676810567214,
"learning_rate": 4.631641238374482e-06,
"loss": 0.147,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12088508903980255,
"step": 3360,
"valid_targets_mean": 2881.7,
"valid_targets_min": 1476
},
{
"epoch": 5.617696160267112,
"grad_norm": 0.39617261584358665,
"learning_rate": 4.578489210189554e-06,
"loss": 0.1702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25199243426322937,
"step": 3365,
"valid_targets_mean": 6000.6,
"valid_targets_min": 1491
},
{
"epoch": 5.626043405676127,
"grad_norm": 0.374328452364103,
"learning_rate": 4.525604477991505e-06,
"loss": 0.1334,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15098735690116882,
"step": 3370,
"valid_targets_mean": 4393.1,
"valid_targets_min": 1639
},
{
"epoch": 5.634390651085142,
"grad_norm": 0.39542956076739977,
"learning_rate": 4.472987958414077e-06,
"loss": 0.1525,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18923836946487427,
"step": 3375,
"valid_targets_mean": 5107.1,
"valid_targets_min": 2207
},
{
"epoch": 5.642737896494157,
"grad_norm": 0.39294908862349665,
"learning_rate": 4.420640563442111e-06,
"loss": 0.1557,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14803722500801086,
"step": 3380,
"valid_targets_mean": 4307.4,
"valid_targets_min": 1646
},
{
"epoch": 5.651085141903172,
"grad_norm": 0.371641002140393,
"learning_rate": 4.368563200395859e-06,
"loss": 0.1597,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13434095680713654,
"step": 3385,
"valid_targets_mean": 4312.2,
"valid_targets_min": 1475
},
{
"epoch": 5.659432387312187,
"grad_norm": 0.39567117055950995,
"learning_rate": 4.31675677191514e-06,
"loss": 0.1618,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19538375735282898,
"step": 3390,
"valid_targets_mean": 4763.1,
"valid_targets_min": 1446
},
{
"epoch": 5.667779632721202,
"grad_norm": 0.41687465612635255,
"learning_rate": 4.265222175943808e-06,
"loss": 0.1634,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19268158078193665,
"step": 3395,
"valid_targets_mean": 4113.9,
"valid_targets_min": 1464
},
{
"epoch": 5.676126878130217,
"grad_norm": 0.41028334716721787,
"learning_rate": 4.213960305714082e-06,
"loss": 0.1434,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12247997522354126,
"step": 3400,
"valid_targets_mean": 3847.2,
"valid_targets_min": 1384
},
{
"epoch": 5.684474123539232,
"grad_norm": 0.3777454179600641,
"learning_rate": 4.162972049731164e-06,
"loss": 0.1389,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11124204844236374,
"step": 3405,
"valid_targets_mean": 4050.1,
"valid_targets_min": 1981
},
{
"epoch": 5.692821368948247,
"grad_norm": 0.40335065873622244,
"learning_rate": 4.112258291757747e-06,
"loss": 0.1505,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1639295220375061,
"step": 3410,
"valid_targets_mean": 4321.6,
"valid_targets_min": 1644
},
{
"epoch": 5.701168614357262,
"grad_norm": 0.4285112195748218,
"learning_rate": 4.061819910798777e-06,
"loss": 0.1391,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15136224031448364,
"step": 3415,
"valid_targets_mean": 3664.6,
"valid_targets_min": 1358
},
{
"epoch": 5.709515859766277,
"grad_norm": 0.40503496765642444,
"learning_rate": 4.01165778108614e-06,
"loss": 0.1579,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12341266870498657,
"step": 3420,
"valid_targets_mean": 3343.4,
"valid_targets_min": 1442
},
{
"epoch": 5.717863105175292,
"grad_norm": 0.39686418841734933,
"learning_rate": 3.961772772063599e-06,
"loss": 0.1445,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1693595051765442,
"step": 3425,
"valid_targets_mean": 4642.7,
"valid_targets_min": 1049
},
{
"epoch": 5.726210350584307,
"grad_norm": 0.633788958153466,
"learning_rate": 3.912165748371626e-06,
"loss": 0.1462,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12495502084493637,
"step": 3430,
"valid_targets_mean": 2631.0,
"valid_targets_min": 1244
},
{
"epoch": 5.734557595993322,
"grad_norm": 0.40025645150837613,
"learning_rate": 3.8628375698325045e-06,
"loss": 0.1783,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13537541031837463,
"step": 3435,
"valid_targets_mean": 3719.8,
"valid_targets_min": 1393
},
{
"epoch": 5.742904841402337,
"grad_norm": 0.3872216406354163,
"learning_rate": 3.8137890914353535e-06,
"loss": 0.1799,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18745312094688416,
"step": 3440,
"valid_targets_mean": 5563.6,
"valid_targets_min": 1920
},
{
"epoch": 5.751252086811352,
"grad_norm": 0.4680812080466845,
"learning_rate": 3.765021163321374e-06,
"loss": 0.1622,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1378217339515686,
"step": 3445,
"valid_targets_mean": 3020.4,
"valid_targets_min": 1411
},
{
"epoch": 5.759599332220367,
"grad_norm": 0.4357685731529475,
"learning_rate": 3.7165346307690466e-06,
"loss": 0.1702,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18366341292858124,
"step": 3450,
"valid_targets_mean": 3705.8,
"valid_targets_min": 1448
},
{
"epoch": 5.767946577629383,
"grad_norm": 0.404276768410918,
"learning_rate": 3.6683303341795483e-06,
"loss": 0.1424,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15461501479148865,
"step": 3455,
"valid_targets_mean": 4438.6,
"valid_targets_min": 1439
},
{
"epoch": 5.776293823038397,
"grad_norm": 0.4057692681381097,
"learning_rate": 3.6204091090621176e-06,
"loss": 0.1412,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13683287799358368,
"step": 3460,
"valid_targets_mean": 3771.2,
"valid_targets_min": 1478
},
{
"epoch": 5.784641068447412,
"grad_norm": 0.40637804373197595,
"learning_rate": 3.572771786019649e-06,
"loss": 0.1565,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19681739807128906,
"step": 3465,
"valid_targets_mean": 4143.6,
"valid_targets_min": 1482
},
{
"epoch": 5.792988313856427,
"grad_norm": 0.3551673866700703,
"learning_rate": 3.5254191907342117e-06,
"loss": 0.1507,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14093132317066193,
"step": 3470,
"valid_targets_mean": 4872.8,
"valid_targets_min": 1390
},
{
"epoch": 5.801335559265443,
"grad_norm": 0.42078909322481944,
"learning_rate": 3.4783521439528233e-06,
"loss": 0.1568,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1718795895576477,
"step": 3475,
"valid_targets_mean": 4379.8,
"valid_targets_min": 2113
},
{
"epoch": 5.809682804674457,
"grad_norm": 0.4171388137937807,
"learning_rate": 3.4315714614731467e-06,
"loss": 0.1322,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11956311762332916,
"step": 3480,
"valid_targets_mean": 3359.4,
"valid_targets_min": 1384
},
{
"epoch": 5.818030050083473,
"grad_norm": 0.42519206676803806,
"learning_rate": 3.385077954129421e-06,
"loss": 0.1414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1623392105102539,
"step": 3485,
"valid_targets_mean": 3458.9,
"valid_targets_min": 1573
},
{
"epoch": 5.826377295492487,
"grad_norm": 0.3765702654390368,
"learning_rate": 3.3388724277783347e-06,
"loss": 0.1502,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1616736650466919,
"step": 3490,
"valid_targets_mean": 4789.1,
"valid_targets_min": 1257
},
{
"epoch": 5.834724540901503,
"grad_norm": 0.4377820006235134,
"learning_rate": 3.292955683285135e-06,
"loss": 0.1552,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18910910189151764,
"step": 3495,
"valid_targets_mean": 3649.6,
"valid_targets_min": 1366
},
{
"epoch": 5.843071786310517,
"grad_norm": 0.4079980482123456,
"learning_rate": 3.2473285165096736e-06,
"loss": 0.1541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1576949506998062,
"step": 3500,
"valid_targets_mean": 4077.4,
"valid_targets_min": 1572
},
{
"epoch": 5.851419031719533,
"grad_norm": 0.40374064325638326,
"learning_rate": 3.2019917182926806e-06,
"loss": 0.154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1571791172027588,
"step": 3505,
"valid_targets_mean": 4139.4,
"valid_targets_min": 1413
},
{
"epoch": 5.859766277128547,
"grad_norm": 0.44008716128123,
"learning_rate": 3.156946074441982e-06,
"loss": 0.1538,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1598498523235321,
"step": 3510,
"valid_targets_mean": 3535.6,
"valid_targets_min": 1367
},
{
"epoch": 5.868113522537563,
"grad_norm": 0.4082815541872326,
"learning_rate": 3.112192365718969e-06,
"loss": 0.1508,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16287583112716675,
"step": 3515,
"valid_targets_mean": 3993.4,
"valid_targets_min": 1440
},
{
"epoch": 5.876460767946577,
"grad_norm": 0.4498704483336117,
"learning_rate": 3.067731367824969e-06,
"loss": 0.1579,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11932264268398285,
"step": 3520,
"valid_targets_mean": 4364.2,
"valid_targets_min": 1365
},
{
"epoch": 5.884808013355593,
"grad_norm": 0.38207236303683023,
"learning_rate": 3.023563851387885e-06,
"loss": 0.1724,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1368837058544159,
"step": 3525,
"valid_targets_mean": 4313.9,
"valid_targets_min": 2159
},
{
"epoch": 5.893155258764608,
"grad_norm": 0.3975415511553311,
"learning_rate": 2.979690581948784e-06,
"loss": 0.1385,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11136573553085327,
"step": 3530,
"valid_targets_mean": 3332.7,
"valid_targets_min": 1738
},
{
"epoch": 5.901502504173623,
"grad_norm": 0.4323676592801381,
"learning_rate": 2.936112319948654e-06,
"loss": 0.1489,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1412634253501892,
"step": 3535,
"valid_targets_mean": 3386.6,
"valid_targets_min": 1396
},
{
"epoch": 5.909849749582638,
"grad_norm": 0.3977766139791975,
"learning_rate": 2.892829820715208e-06,
"loss": 0.132,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13958236575126648,
"step": 3540,
"valid_targets_mean": 3752.4,
"valid_targets_min": 1574
},
{
"epoch": 5.918196994991653,
"grad_norm": 0.3857016215298441,
"learning_rate": 2.8498438344498103e-06,
"loss": 0.1365,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1721857488155365,
"step": 3545,
"valid_targets_mean": 4809.5,
"valid_targets_min": 1831
},
{
"epoch": 5.926544240400668,
"grad_norm": 0.4458704441991114,
"learning_rate": 2.8071551062144518e-06,
"loss": 0.1481,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11852441728115082,
"step": 3550,
"valid_targets_mean": 3085.1,
"valid_targets_min": 1388
},
{
"epoch": 5.934891485809683,
"grad_norm": 0.3870845184052212,
"learning_rate": 2.7647643759188557e-06,
"loss": 0.137,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1172284409403801,
"step": 3555,
"valid_targets_mean": 3655.8,
"valid_targets_min": 2389
},
{
"epoch": 5.943238731218698,
"grad_norm": 0.41972305491793754,
"learning_rate": 2.7226723783076447e-06,
"loss": 0.1511,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17211323976516724,
"step": 3560,
"valid_targets_mean": 4190.2,
"valid_targets_min": 1486
},
{
"epoch": 5.951585976627713,
"grad_norm": 0.39875299757615007,
"learning_rate": 2.680879842947601e-06,
"loss": 0.1496,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17626696825027466,
"step": 3565,
"valid_targets_mean": 4109.8,
"valid_targets_min": 1834
},
{
"epoch": 5.959933222036728,
"grad_norm": 0.3721950682095093,
"learning_rate": 2.6393874942150268e-06,
"loss": 0.1426,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1562456488609314,
"step": 3570,
"valid_targets_mean": 5186.9,
"valid_targets_min": 1377
},
{
"epoch": 5.968280467445743,
"grad_norm": 0.4402968978022113,
"learning_rate": 2.598196051283193e-06,
"loss": 0.1383,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14634525775909424,
"step": 3575,
"valid_targets_mean": 3270.2,
"valid_targets_min": 1600
},
{
"epoch": 5.976627712854758,
"grad_norm": 0.39856549155606713,
"learning_rate": 2.5573062281098638e-06,
"loss": 0.1701,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19217711687088013,
"step": 3580,
"valid_targets_mean": 5295.2,
"valid_targets_min": 1558
},
{
"epoch": 5.984974958263773,
"grad_norm": 0.40515122156641264,
"learning_rate": 2.5167187334249277e-06,
"loss": 0.1516,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13459628820419312,
"step": 3585,
"valid_targets_mean": 3885.8,
"valid_targets_min": 1621
},
{
"epoch": 5.993322203672788,
"grad_norm": 0.39100451002685,
"learning_rate": 2.4764342707181155e-06,
"loss": 0.1728,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15787723660469055,
"step": 3590,
"valid_targets_mean": 4636.8,
"valid_targets_min": 1559
},
{
"epoch": 6.001669449081803,
"grad_norm": 0.44544239697431576,
"learning_rate": 2.4364535382268017e-06,
"loss": 0.1534,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16798482835292816,
"step": 3595,
"valid_targets_mean": 3404.9,
"valid_targets_min": 1526
},
{
"epoch": 6.010016694490818,
"grad_norm": 0.4058188559942236,
"learning_rate": 2.3967772289239055e-06,
"loss": 0.1369,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14776292443275452,
"step": 3600,
"valid_targets_mean": 3633.4,
"valid_targets_min": 1396
},
{
"epoch": 6.018363939899833,
"grad_norm": 0.410470351046545,
"learning_rate": 2.357406030505878e-06,
"loss": 0.1571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11771170794963837,
"step": 3605,
"valid_targets_mean": 3415.6,
"valid_targets_min": 1376
},
{
"epoch": 6.026711185308848,
"grad_norm": 0.47068213298681094,
"learning_rate": 2.318340625380786e-06,
"loss": 0.1546,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20467709004878998,
"step": 3610,
"valid_targets_mean": 4296.9,
"valid_targets_min": 2078
},
{
"epoch": 6.035058430717863,
"grad_norm": 0.3836160543330413,
"learning_rate": 2.279581690656476e-06,
"loss": 0.1588,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1730513572692871,
"step": 3615,
"valid_targets_mean": 5380.2,
"valid_targets_min": 1767
},
{
"epoch": 6.043405676126878,
"grad_norm": 0.45720070328561174,
"learning_rate": 2.241129898128851e-06,
"loss": 0.1464,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13687875866889954,
"step": 3620,
"valid_targets_mean": 3152.5,
"valid_targets_min": 1335
},
{
"epoch": 6.051752921535893,
"grad_norm": 0.4357973677368353,
"learning_rate": 2.202985914270215e-06,
"loss": 0.1619,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11903496831655502,
"step": 3625,
"valid_targets_mean": 3008.2,
"valid_targets_min": 1450
},
{
"epoch": 6.060100166944908,
"grad_norm": 0.38672849101393547,
"learning_rate": 2.1651504002177236e-06,
"loss": 0.1446,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13006556034088135,
"step": 3630,
"valid_targets_mean": 4282.8,
"valid_targets_min": 1389
},
{
"epoch": 6.068447412353923,
"grad_norm": 0.3883009146282991,
"learning_rate": 2.1276240117619283e-06,
"loss": 0.148,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17265048623085022,
"step": 3635,
"valid_targets_mean": 4957.2,
"valid_targets_min": 2350
},
{
"epoch": 6.076794657762938,
"grad_norm": 0.4183046036965391,
"learning_rate": 2.09040739933541e-06,
"loss": 0.1319,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13943225145339966,
"step": 3640,
"valid_targets_mean": 3708.4,
"valid_targets_min": 1593
},
{
"epoch": 6.085141903171953,
"grad_norm": 0.3792833630233868,
"learning_rate": 2.0535012080015006e-06,
"loss": 0.1683,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1811470091342926,
"step": 3645,
"valid_targets_mean": 5005.7,
"valid_targets_min": 1487
},
{
"epoch": 6.093489148580968,
"grad_norm": 0.36631456947325636,
"learning_rate": 2.016906077443106e-06,
"loss": 0.1494,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1519770622253418,
"step": 3650,
"valid_targets_mean": 4946.0,
"valid_targets_min": 1893
},
{
"epoch": 6.101836393989983,
"grad_norm": 0.4161660858660546,
"learning_rate": 1.9806226419516195e-06,
"loss": 0.1672,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1819477379322052,
"step": 3655,
"valid_targets_mean": 4172.1,
"valid_targets_min": 1315
},
{
"epoch": 6.110183639398999,
"grad_norm": 0.4071274484788158,
"learning_rate": 1.9446515304159198e-06,
"loss": 0.1327,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1492326408624649,
"step": 3660,
"valid_targets_mean": 4035.2,
"valid_targets_min": 1470
},
{
"epoch": 6.118530884808013,
"grad_norm": 0.4281970891344757,
"learning_rate": 1.9089933663114868e-06,
"loss": 0.147,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14949047565460205,
"step": 3665,
"valid_targets_mean": 4121.2,
"valid_targets_min": 2069
},
{
"epoch": 6.126878130217029,
"grad_norm": 0.37803136426460476,
"learning_rate": 1.8736487676895754e-06,
"loss": 0.1561,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19181250035762787,
"step": 3670,
"valid_targets_mean": 5692.7,
"valid_targets_min": 1891
},
{
"epoch": 6.135225375626043,
"grad_norm": 0.4215046917484127,
"learning_rate": 1.8386183471665187e-06,
"loss": 0.1514,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.130486860871315,
"step": 3675,
"valid_targets_mean": 3343.8,
"valid_targets_min": 1281
},
{
"epoch": 6.143572621035059,
"grad_norm": 0.4209201948793024,
"learning_rate": 1.8039027119131057e-06,
"loss": 0.1499,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13366928696632385,
"step": 3680,
"valid_targets_mean": 3585.1,
"valid_targets_min": 1801
},
{
"epoch": 6.151919866444073,
"grad_norm": 0.41502314108422217,
"learning_rate": 1.7695024636440484e-06,
"loss": 0.143,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1650959998369217,
"step": 3685,
"valid_targets_mean": 3927.5,
"valid_targets_min": 1542
},
{
"epoch": 6.160267111853089,
"grad_norm": 0.40776160093015773,
"learning_rate": 1.7354181986075635e-06,
"loss": 0.1558,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16517841815948486,
"step": 3690,
"valid_targets_mean": 4263.7,
"valid_targets_min": 1583
},
{
"epoch": 6.168614357262103,
"grad_norm": 0.4262175507221063,
"learning_rate": 1.701650507575039e-06,
"loss": 0.1449,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13946987688541412,
"step": 3695,
"valid_targets_mean": 3732.7,
"valid_targets_min": 2033
},
{
"epoch": 6.176961602671119,
"grad_norm": 0.4189308085165393,
"learning_rate": 1.6681999758307799e-06,
"loss": 0.1571,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15114673972129822,
"step": 3700,
"valid_targets_mean": 3940.6,
"valid_targets_min": 1367
},
{
"epoch": 6.185308848080133,
"grad_norm": 0.3717758300400972,
"learning_rate": 1.6350671831618804e-06,
"loss": 0.1314,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13511475920677185,
"step": 3705,
"valid_targets_mean": 4325.9,
"valid_targets_min": 1411
},
{
"epoch": 6.193656093489149,
"grad_norm": 0.3864849515731443,
"learning_rate": 1.602252703848164e-06,
"loss": 0.1453,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19211497902870178,
"step": 3710,
"valid_targets_mean": 5274.4,
"valid_targets_min": 1893
},
{
"epoch": 6.202003338898163,
"grad_norm": 0.37573652566494325,
"learning_rate": 1.5697571066522321e-06,
"loss": 0.1471,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.20908550918102264,
"step": 3715,
"valid_targets_mean": 5656.1,
"valid_targets_min": 1809
},
{
"epoch": 6.210350584307179,
"grad_norm": 0.4199889018673102,
"learning_rate": 1.5375809548096187e-06,
"loss": 0.1587,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17829513549804688,
"step": 3720,
"valid_targets_mean": 4724.2,
"valid_targets_min": 1612
},
{
"epoch": 6.218697829716193,
"grad_norm": 0.42310944666747885,
"learning_rate": 1.5057248060189956e-06,
"loss": 0.1212,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1219317838549614,
"step": 3725,
"valid_targets_mean": 3595.0,
"valid_targets_min": 1710
},
{
"epoch": 6.227045075125209,
"grad_norm": 0.4287339432317377,
"learning_rate": 1.4741892124325508e-06,
"loss": 0.1751,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1763526201248169,
"step": 3730,
"valid_targets_mean": 4323.2,
"valid_targets_min": 1365
},
{
"epoch": 6.235392320534224,
"grad_norm": 0.388233110999744,
"learning_rate": 1.4429747206463662e-06,
"loss": 0.1376,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12875054776668549,
"step": 3735,
"valid_targets_mean": 4268.2,
"valid_targets_min": 1568
},
{
"epoch": 6.243739565943239,
"grad_norm": 0.4436182629228336,
"learning_rate": 1.4120818716910023e-06,
"loss": 0.1422,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1386677324771881,
"step": 3740,
"valid_targets_mean": 3435.2,
"valid_targets_min": 1442
},
{
"epoch": 6.252086811352254,
"grad_norm": 0.42061709494404725,
"learning_rate": 1.381511201022061e-06,
"loss": 0.1406,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1040327250957489,
"step": 3745,
"valid_targets_mean": 2783.2,
"valid_targets_min": 1352
},
{
"epoch": 6.260434056761269,
"grad_norm": 0.3967203483469124,
"learning_rate": 1.3512632385109582e-06,
"loss": 0.1414,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1614457666873932,
"step": 3750,
"valid_targets_mean": 4983.1,
"valid_targets_min": 1836
},
{
"epoch": 6.268781302170284,
"grad_norm": 0.39844684626650917,
"learning_rate": 1.3213385084356944e-06,
"loss": 0.1455,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1480848491191864,
"step": 3755,
"valid_targets_mean": 4147.2,
"valid_targets_min": 1996
},
{
"epoch": 6.277128547579299,
"grad_norm": 0.3588349607609908,
"learning_rate": 1.2917375294718083e-06,
"loss": 0.1475,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13978177309036255,
"step": 3760,
"valid_targets_mean": 4638.1,
"valid_targets_min": 1507
},
{
"epoch": 6.285475792988314,
"grad_norm": 0.4196349039830257,
"learning_rate": 1.262460814683344e-06,
"loss": 0.1551,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16789385676383972,
"step": 3765,
"valid_targets_mean": 3814.4,
"valid_targets_min": 1475
},
{
"epoch": 6.293823038397329,
"grad_norm": 0.4231575160689391,
"learning_rate": 1.2335088715140065e-06,
"loss": 0.1292,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12742555141448975,
"step": 3770,
"valid_targets_mean": 3498.1,
"valid_targets_min": 1550
},
{
"epoch": 6.302170283806344,
"grad_norm": 0.4104837077662595,
"learning_rate": 1.2048822017783168e-06,
"loss": 0.1412,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.144887775182724,
"step": 3775,
"valid_targets_mean": 3882.4,
"valid_targets_min": 1800
},
{
"epoch": 6.310517529215359,
"grad_norm": 0.4319166548098376,
"learning_rate": 1.1765813016529592e-06,
"loss": 0.1471,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1558988392353058,
"step": 3780,
"valid_targets_mean": 3503.1,
"valid_targets_min": 1392
},
{
"epoch": 6.318864774624374,
"grad_norm": 0.4484001149557676,
"learning_rate": 1.1486066616681413e-06,
"loss": 0.1432,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1359315812587738,
"step": 3785,
"valid_targets_mean": 3798.4,
"valid_targets_min": 1459
},
{
"epoch": 6.3272120200333895,
"grad_norm": 0.4054660585193615,
"learning_rate": 1.1209587666991273e-06,
"loss": 0.1298,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1483077108860016,
"step": 3790,
"valid_targets_mean": 4309.3,
"valid_targets_min": 1352
},
{
"epoch": 6.335559265442404,
"grad_norm": 0.4744823654792207,
"learning_rate": 1.093638095957803e-06,
"loss": 0.1513,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12355400621891022,
"step": 3795,
"valid_targets_mean": 3635.9,
"valid_targets_min": 1424
},
{
"epoch": 6.343906510851419,
"grad_norm": 0.4258123179029083,
"learning_rate": 1.0666451229844e-06,
"loss": 0.1581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13636405766010284,
"step": 3800,
"valid_targets_mean": 3672.7,
"valid_targets_min": 1441
},
{
"epoch": 6.352253756260434,
"grad_norm": 0.3603705759877712,
"learning_rate": 1.0399803156392507e-06,
"loss": 0.1347,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1230199858546257,
"step": 3805,
"valid_targets_mean": 4551.8,
"valid_targets_min": 1852
},
{
"epoch": 6.360601001669449,
"grad_norm": 0.40896792270063415,
"learning_rate": 1.0136441360947247e-06,
"loss": 0.1654,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.18940380215644836,
"step": 3810,
"valid_targets_mean": 4767.1,
"valid_targets_min": 1846
},
{
"epoch": 6.368948247078464,
"grad_norm": 0.4331911766699577,
"learning_rate": 9.876370408271675e-07,
"loss": 0.1427,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14649033546447754,
"step": 3815,
"valid_targets_mean": 3335.8,
"valid_targets_min": 1481
},
{
"epoch": 6.377295492487479,
"grad_norm": 0.3762009275253653,
"learning_rate": 9.619594806090449e-07,
"loss": 0.1403,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1219458281993866,
"step": 3820,
"valid_targets_mean": 3998.9,
"valid_targets_min": 1589
},
{
"epoch": 6.385642737896494,
"grad_norm": 0.4037998428863478,
"learning_rate": 9.366119005010699e-07,
"loss": 0.1344,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14254915714263916,
"step": 3825,
"valid_targets_mean": 4041.8,
"valid_targets_min": 1878
},
{
"epoch": 6.393989983305509,
"grad_norm": 0.44955374702390966,
"learning_rate": 9.115947398445413e-07,
"loss": 0.1582,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1691751778125763,
"step": 3830,
"valid_targets_mean": 3730.8,
"valid_targets_min": 1609
},
{
"epoch": 6.402337228714524,
"grad_norm": 0.4359028149352726,
"learning_rate": 8.869084322536881e-07,
"loss": 0.1468,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1457628309726715,
"step": 3835,
"valid_targets_mean": 2955.9,
"valid_targets_min": 1357
},
{
"epoch": 6.410684474123539,
"grad_norm": 0.37753688249607964,
"learning_rate": 8.62553405608193e-07,
"loss": 0.1363,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13241273164749146,
"step": 3840,
"valid_targets_mean": 4894.3,
"valid_targets_min": 1315
},
{
"epoch": 6.419031719532554,
"grad_norm": 0.40859124561690974,
"learning_rate": 8.385300820457276e-07,
"loss": 0.1537,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1553121656179428,
"step": 3845,
"valid_targets_mean": 5548.7,
"valid_targets_min": 1709
},
{
"epoch": 6.427378964941569,
"grad_norm": 0.4681961402327626,
"learning_rate": 8.148388779546912e-07,
"loss": 0.1187,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1499110460281372,
"step": 3850,
"valid_targets_mean": 3277.8,
"valid_targets_min": 1384
},
{
"epoch": 6.435726210350584,
"grad_norm": 0.43654365934466466,
"learning_rate": 7.914802039669412e-07,
"loss": 0.1348,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12845081090927124,
"step": 3855,
"valid_targets_mean": 3412.7,
"valid_targets_min": 1387
},
{
"epoch": 6.444073455759599,
"grad_norm": 0.4146474565017382,
"learning_rate": 7.684544649507164e-07,
"loss": 0.1476,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15668165683746338,
"step": 3860,
"valid_targets_mean": 4065.0,
"valid_targets_min": 1366
},
{
"epoch": 6.452420701168615,
"grad_norm": 0.46714447830500166,
"learning_rate": 7.457620600035898e-07,
"loss": 0.1684,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.2366269826889038,
"step": 3865,
"valid_targets_mean": 4201.5,
"valid_targets_min": 1467
},
{
"epoch": 6.460767946577629,
"grad_norm": 0.4160267184670818,
"learning_rate": 7.234033824455821e-07,
"loss": 0.1555,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1594693511724472,
"step": 3870,
"valid_targets_mean": 4366.9,
"valid_targets_min": 1476
},
{
"epoch": 6.469115191986645,
"grad_norm": 0.43462407178423773,
"learning_rate": 7.01378819812304e-07,
"loss": 0.1888,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.25142306089401245,
"step": 3875,
"valid_targets_mean": 4585.8,
"valid_targets_min": 1411
},
{
"epoch": 6.477462437395659,
"grad_norm": 0.41531946241268725,
"learning_rate": 6.796887538482821e-07,
"loss": 0.1541,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12880435585975647,
"step": 3880,
"valid_targets_mean": 4117.6,
"valid_targets_min": 1297
},
{
"epoch": 6.485809682804675,
"grad_norm": 0.39677387446784285,
"learning_rate": 6.583335605003083e-07,
"loss": 0.153,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1286594271659851,
"step": 3885,
"valid_targets_mean": 3851.5,
"valid_targets_min": 1460
},
{
"epoch": 6.494156928213689,
"grad_norm": 0.3829025794471524,
"learning_rate": 6.373136099109455e-07,
"loss": 0.1513,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15454237163066864,
"step": 3890,
"valid_targets_mean": 4839.5,
"valid_targets_min": 1964
},
{
"epoch": 6.502504173622705,
"grad_norm": 0.40920532965996803,
"learning_rate": 6.166292664120987e-07,
"loss": 0.1337,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17860062420368195,
"step": 3895,
"valid_targets_mean": 4343.2,
"valid_targets_min": 1557
},
{
"epoch": 6.510851419031719,
"grad_norm": 0.39183330876045247,
"learning_rate": 5.962808885187121e-07,
"loss": 0.1569,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13849595189094543,
"step": 3900,
"valid_targets_mean": 4484.2,
"valid_targets_min": 1428
},
{
"epoch": 6.519198664440735,
"grad_norm": 0.42154909123504963,
"learning_rate": 5.762688289225349e-07,
"loss": 0.1485,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13923391699790955,
"step": 3905,
"valid_targets_mean": 3983.9,
"valid_targets_min": 1831
},
{
"epoch": 6.527545909849749,
"grad_norm": 0.3982561822383007,
"learning_rate": 5.565934344860413e-07,
"loss": 0.1437,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17047888040542603,
"step": 3910,
"valid_targets_mean": 4670.4,
"valid_targets_min": 1384
},
{
"epoch": 6.535893155258765,
"grad_norm": 0.4365738032097916,
"learning_rate": 5.372550462363779e-07,
"loss": 0.1538,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17946776747703552,
"step": 3915,
"valid_targets_mean": 4282.5,
"valid_targets_min": 1326
},
{
"epoch": 6.54424040066778,
"grad_norm": 0.46867964845874105,
"learning_rate": 5.182539993594904e-07,
"loss": 0.1231,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12115922570228577,
"step": 3920,
"valid_targets_mean": 3614.4,
"valid_targets_min": 701
},
{
"epoch": 6.552587646076795,
"grad_norm": 0.41269099422849015,
"learning_rate": 4.99590623194286e-07,
"loss": 0.1536,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13953888416290283,
"step": 3925,
"valid_targets_mean": 3918.1,
"valid_targets_min": 1349
},
{
"epoch": 6.560934891485809,
"grad_norm": 0.40490045250158996,
"learning_rate": 4.812652412269448e-07,
"loss": 0.1523,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1728893518447876,
"step": 3930,
"valid_targets_mean": 4602.6,
"valid_targets_min": 1416
},
{
"epoch": 6.569282136894825,
"grad_norm": 0.4155800474911707,
"learning_rate": 4.632781710852929e-07,
"loss": 0.1458,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16982467472553253,
"step": 3935,
"valid_targets_mean": 4274.1,
"valid_targets_min": 1321
},
{
"epoch": 6.57762938230384,
"grad_norm": 0.3721789497670605,
"learning_rate": 4.456297245333252e-07,
"loss": 0.1549,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1757877618074417,
"step": 3940,
"valid_targets_mean": 5176.7,
"valid_targets_min": 1623
},
{
"epoch": 6.585976627712855,
"grad_norm": 0.4696274459711786,
"learning_rate": 4.2832020746576666e-07,
"loss": 0.1328,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15677687525749207,
"step": 3945,
"valid_targets_mean": 4947.6,
"valid_targets_min": 1376
},
{
"epoch": 6.59432387312187,
"grad_norm": 0.43280516580875666,
"learning_rate": 4.113499199028037e-07,
"loss": 0.1664,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17593613266944885,
"step": 3950,
"valid_targets_mean": 4264.0,
"valid_targets_min": 914
},
{
"epoch": 6.602671118530885,
"grad_norm": 0.4559545683708418,
"learning_rate": 3.9471915598485954e-07,
"loss": 0.1328,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14975443482398987,
"step": 3955,
"valid_targets_mean": 3412.4,
"valid_targets_min": 1361
},
{
"epoch": 6.6110183639399,
"grad_norm": 0.42978347968599523,
"learning_rate": 3.7842820396751134e-07,
"loss": 0.1428,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1132403016090393,
"step": 3960,
"valid_targets_mean": 3127.4,
"valid_targets_min": 1449
},
{
"epoch": 6.619365609348915,
"grad_norm": 0.3981563006915193,
"learning_rate": 3.62477346216481e-07,
"loss": 0.1445,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14268843829631805,
"step": 3965,
"valid_targets_mean": 4246.9,
"valid_targets_min": 1425
},
{
"epoch": 6.62771285475793,
"grad_norm": 0.3798788095773447,
"learning_rate": 3.468668592027613e-07,
"loss": 0.1626,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11680915951728821,
"step": 3970,
"valid_targets_mean": 4476.6,
"valid_targets_min": 1502
},
{
"epoch": 6.636060100166945,
"grad_norm": 0.4003365121405331,
"learning_rate": 3.3159701349779083e-07,
"loss": 0.1367,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19601371884346008,
"step": 3975,
"valid_targets_mean": 4777.7,
"valid_targets_min": 1974
},
{
"epoch": 6.64440734557596,
"grad_norm": 0.454340724951457,
"learning_rate": 3.1666807376880436e-07,
"loss": 0.1298,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11717614531517029,
"step": 3980,
"valid_targets_mean": 2865.9,
"valid_targets_min": 1491
},
{
"epoch": 6.652754590984975,
"grad_norm": 0.4387321374182813,
"learning_rate": 3.0208029877420996e-07,
"loss": 0.1464,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1354750692844391,
"step": 3985,
"valid_targets_mean": 3357.2,
"valid_targets_min": 1532
},
{
"epoch": 6.66110183639399,
"grad_norm": 0.40391798479712915,
"learning_rate": 2.8783394135913245e-07,
"loss": 0.1349,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13499857485294342,
"step": 3990,
"valid_targets_mean": 4215.4,
"valid_targets_min": 1466
},
{
"epoch": 6.669449081803005,
"grad_norm": 0.4308258977171262,
"learning_rate": 2.7392924845100364e-07,
"loss": 0.1354,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13604167103767395,
"step": 3995,
"valid_targets_mean": 3515.1,
"valid_targets_min": 1430
},
{
"epoch": 6.67779632721202,
"grad_norm": 0.41873026464679164,
"learning_rate": 2.6036646105530804e-07,
"loss": 0.1432,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12609747052192688,
"step": 4000,
"valid_targets_mean": 3619.0,
"valid_targets_min": 1443
},
{
"epoch": 6.686143572621035,
"grad_norm": 0.48015243674091945,
"learning_rate": 2.471458142513861e-07,
"loss": 0.1357,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11774508655071259,
"step": 4005,
"valid_targets_mean": 3475.4,
"valid_targets_min": 1381
},
{
"epoch": 6.69449081803005,
"grad_norm": 0.41204018081664034,
"learning_rate": 2.3426753718837735e-07,
"loss": 0.1581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16718773543834686,
"step": 4010,
"valid_targets_mean": 4271.9,
"valid_targets_min": 835
},
{
"epoch": 6.702838063439065,
"grad_norm": 0.4434689469703618,
"learning_rate": 2.2173185308122624e-07,
"loss": 0.1594,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19118532538414001,
"step": 4015,
"valid_targets_mean": 4180.0,
"valid_targets_min": 1364
},
{
"epoch": 6.71118530884808,
"grad_norm": 0.40636313851124267,
"learning_rate": 2.0953897920683807e-07,
"loss": 0.1459,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13011732697486877,
"step": 4020,
"valid_targets_mean": 3924.0,
"valid_targets_min": 1830
},
{
"epoch": 6.719532554257095,
"grad_norm": 0.4053672146915359,
"learning_rate": 1.976891269002934e-07,
"loss": 0.1579,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1793678253889084,
"step": 4025,
"valid_targets_mean": 4314.1,
"valid_targets_min": 871
},
{
"epoch": 6.72787979966611,
"grad_norm": 0.4284053916212973,
"learning_rate": 1.861825015511931e-07,
"loss": 0.1453,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15175093710422516,
"step": 4030,
"valid_targets_mean": 3522.3,
"valid_targets_min": 1409
},
{
"epoch": 6.736227045075125,
"grad_norm": 0.3762522430816129,
"learning_rate": 1.7501930260009902e-07,
"loss": 0.1281,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1307104229927063,
"step": 4035,
"valid_targets_mean": 4451.3,
"valid_targets_min": 1907
},
{
"epoch": 6.74457429048414,
"grad_norm": 0.3882826683314684,
"learning_rate": 1.6419972353507895e-07,
"loss": 0.1653,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12455746531486511,
"step": 4040,
"valid_targets_mean": 4246.4,
"valid_targets_min": 2036
},
{
"epoch": 6.752921535893155,
"grad_norm": 0.4503289550680519,
"learning_rate": 1.5372395188834265e-07,
"loss": 0.1427,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19019600749015808,
"step": 4045,
"valid_targets_mean": 3925.8,
"valid_targets_min": 1650
},
{
"epoch": 6.76126878130217,
"grad_norm": 0.3835593944135294,
"learning_rate": 1.435921692330089e-07,
"loss": 0.1398,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15802672505378723,
"step": 4050,
"valid_targets_mean": 5003.0,
"valid_targets_min": 2368
},
{
"epoch": 6.769616026711185,
"grad_norm": 0.44422407918422674,
"learning_rate": 1.3380455117993684e-07,
"loss": 0.1391,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17681658267974854,
"step": 4055,
"valid_targets_mean": 3981.9,
"valid_targets_min": 1672
},
{
"epoch": 6.7779632721202,
"grad_norm": 0.44315866198253373,
"learning_rate": 1.2436126737470189e-07,
"loss": 0.1654,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13663990795612335,
"step": 4060,
"valid_targets_mean": 3358.1,
"valid_targets_min": 1656
},
{
"epoch": 6.786310517529215,
"grad_norm": 0.36306926395424494,
"learning_rate": 1.1526248149464236e-07,
"loss": 0.1528,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13444079458713531,
"step": 4065,
"valid_targets_mean": 4604.8,
"valid_targets_min": 1467
},
{
"epoch": 6.794657762938231,
"grad_norm": 0.4084565570887492,
"learning_rate": 1.0650835124603076e-07,
"loss": 0.1645,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15772077441215515,
"step": 4070,
"valid_targets_mean": 4363.8,
"valid_targets_min": 652
},
{
"epoch": 6.803005008347245,
"grad_norm": 0.4315404536775707,
"learning_rate": 9.809902836133367e-08,
"loss": 0.1538,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15417814254760742,
"step": 4075,
"valid_targets_mean": 3842.4,
"valid_targets_min": 1645
},
{
"epoch": 6.811352253756261,
"grad_norm": 0.42073026460468893,
"learning_rate": 9.0034658596585e-08,
"loss": 0.1384,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.13165616989135742,
"step": 4080,
"valid_targets_mean": 3715.3,
"valid_targets_min": 1467
},
{
"epoch": 6.819699499165275,
"grad_norm": 0.4127142939173716,
"learning_rate": 8.231538172886133e-08,
"loss": 0.1278,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1386663019657135,
"step": 4085,
"valid_targets_mean": 3700.9,
"valid_targets_min": 1397
},
{
"epoch": 6.828046744574291,
"grad_norm": 0.4156856912915927,
"learning_rate": 7.494133155385497e-08,
"loss": 0.171,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.21586865186691284,
"step": 4090,
"valid_targets_mean": 5173.7,
"valid_targets_min": 2063
},
{
"epoch": 6.836393989983305,
"grad_norm": 0.40928084989141633,
"learning_rate": 6.791263588355801e-08,
"loss": 0.1316,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14061367511749268,
"step": 4095,
"valid_targets_mean": 3994.6,
"valid_targets_min": 1456
},
{
"epoch": 6.844741235392321,
"grad_norm": 0.37807248886780914,
"learning_rate": 6.122941654404635e-08,
"loss": 0.1577,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1406204104423523,
"step": 4100,
"valid_targets_mean": 4761.9,
"valid_targets_min": 1316
},
{
"epoch": 6.853088480801335,
"grad_norm": 0.41044512816122813,
"learning_rate": 5.489178937337025e-08,
"loss": 0.1592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1375463902950287,
"step": 4105,
"valid_targets_mean": 3717.1,
"valid_targets_min": 1402
},
{
"epoch": 6.861435726210351,
"grad_norm": 0.434379865931912,
"learning_rate": 4.88998642195404e-08,
"loss": 0.1581,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12036147713661194,
"step": 4110,
"valid_targets_mean": 3246.9,
"valid_targets_min": 1755
},
{
"epoch": 6.869782971619365,
"grad_norm": 0.4619912829547245,
"learning_rate": 4.325374493862944e-08,
"loss": 0.1503,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15554985404014587,
"step": 4115,
"valid_targets_mean": 3344.3,
"valid_targets_min": 1444
},
{
"epoch": 6.878130217028381,
"grad_norm": 0.4264549770755352,
"learning_rate": 3.795352939296892e-08,
"loss": 0.1542,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1536300778388977,
"step": 4120,
"valid_targets_mean": 3868.3,
"valid_targets_min": 2188
},
{
"epoch": 6.886477462437396,
"grad_norm": 0.419115218193265,
"learning_rate": 3.2999309449459616e-08,
"loss": 0.167,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14947623014450073,
"step": 4125,
"valid_targets_mean": 3898.5,
"valid_targets_min": 1134
},
{
"epoch": 6.894824707846411,
"grad_norm": 0.4147913074667173,
"learning_rate": 2.8391170977968287e-08,
"loss": 0.1427,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1693289577960968,
"step": 4130,
"valid_targets_mean": 4042.4,
"valid_targets_min": 1405
},
{
"epoch": 6.903171953255426,
"grad_norm": 0.41807102531928675,
"learning_rate": 2.4129193849848907e-08,
"loss": 0.1503,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15045379102230072,
"step": 4135,
"valid_targets_mean": 3773.2,
"valid_targets_min": 1307
},
{
"epoch": 6.911519198664441,
"grad_norm": 0.464162965798008,
"learning_rate": 2.0213451936550445e-08,
"loss": 0.1529,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16248679161071777,
"step": 4140,
"valid_targets_mean": 3229.6,
"valid_targets_min": 1494
},
{
"epoch": 6.919866444073456,
"grad_norm": 0.3707416632751337,
"learning_rate": 1.6644013108342294e-08,
"loss": 0.1661,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.11814869940280914,
"step": 4145,
"valid_targets_mean": 4293.8,
"valid_targets_min": 1202
},
{
"epoch": 6.928213689482471,
"grad_norm": 0.43047344272757615,
"learning_rate": 1.3420939233139696e-08,
"loss": 0.1676,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.23192083835601807,
"step": 4150,
"valid_targets_mean": 4701.8,
"valid_targets_min": 2086
},
{
"epoch": 6.936560934891486,
"grad_norm": 0.5020156740962824,
"learning_rate": 1.0544286175422358e-08,
"loss": 0.1247,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14178213477134705,
"step": 4155,
"valid_targets_mean": 3180.2,
"valid_targets_min": 1244
},
{
"epoch": 6.944908180300501,
"grad_norm": 0.3959293838607224,
"learning_rate": 8.01410379527301e-09,
"loss": 0.154,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16196832060813904,
"step": 4160,
"valid_targets_mean": 4417.8,
"valid_targets_min": 1364
},
{
"epoch": 6.953255425709516,
"grad_norm": 0.3928366141466874,
"learning_rate": 5.8304359475158665e-09,
"loss": 0.1498,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.15449389815330505,
"step": 4165,
"valid_targets_mean": 4523.5,
"valid_targets_min": 1503
},
{
"epoch": 6.961602671118531,
"grad_norm": 0.40180407014286684,
"learning_rate": 3.993320480946139e-09,
"loss": 0.1592,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1415395587682724,
"step": 4170,
"valid_targets_mean": 4812.2,
"valid_targets_min": 2024
},
{
"epoch": 6.969949916527546,
"grad_norm": 0.40776621065532875,
"learning_rate": 2.5027892376860984e-09,
"loss": 0.1405,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1601526141166687,
"step": 4175,
"valid_targets_mean": 4681.6,
"valid_targets_min": 1892
},
{
"epoch": 6.978297161936561,
"grad_norm": 0.45004499543329324,
"learning_rate": 1.3588680526255282e-09,
"loss": 0.1521,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17120622098445892,
"step": 4180,
"valid_targets_mean": 3262.2,
"valid_targets_min": 1581
},
{
"epoch": 6.986644407345576,
"grad_norm": 0.43064205577626485,
"learning_rate": 5.615767529709715e-10,
"loss": 0.136,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.12977677583694458,
"step": 4185,
"valid_targets_mean": 3346.2,
"valid_targets_min": 1580
},
{
"epoch": 6.994991652754591,
"grad_norm": 0.4560398512414008,
"learning_rate": 1.1092915791266479e-10,
"loss": 0.1359,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.14427489042282104,
"step": 4190,
"valid_targets_mean": 3283.8,
"valid_targets_min": 1395
},
{
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.1528853178024292,
"step": 4193,
"total_flos": 1136028581756928.0,
"train_loss": 0.1835220243889308,
"train_runtime": 19216.1811,
"train_samples_per_second": 3.487,
"train_steps_per_second": 0.218,
"valid_targets_mean": 3614.8,
"valid_targets_min": 1526
}
],
"logging_steps": 5,
"max_steps": 4193,
"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": 1136028581756928.0,
"train_batch_size": 1,
"trial_name": null,
"trial_params": null
}