{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 7.0,
  "eval_steps": 500,
  "global_step": 2310,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.015151515151515152,
      "grad_norm": 3.5849316120147705,
      "learning_rate": 6.926406926406927e-07,
      "loss": 0.7746,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.7759289741516113,
      "step": 5,
      "valid_targets_mean": 10894.1,
      "valid_targets_min": 1358
    },
    {
      "epoch": 0.030303030303030304,
      "grad_norm": 3.3887853622436523,
      "learning_rate": 1.5584415584415584e-06,
      "loss": 0.7775,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.7486324310302734,
      "step": 10,
      "valid_targets_mean": 10562.2,
      "valid_targets_min": 406
    },
    {
      "epoch": 0.045454545454545456,
      "grad_norm": 2.990579128265381,
      "learning_rate": 2.4242424242424244e-06,
      "loss": 0.7768,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.7580711245536804,
      "step": 15,
      "valid_targets_mean": 10265.7,
      "valid_targets_min": 1394
    },
    {
      "epoch": 0.06060606060606061,
      "grad_norm": 2.7737536430358887,
      "learning_rate": 3.2900432900432905e-06,
      "loss": 0.7692,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.7692350745201111,
      "step": 20,
      "valid_targets_mean": 10456.9,
      "valid_targets_min": 2441
    },
    {
      "epoch": 0.07575757575757576,
      "grad_norm": 1.9057916402816772,
      "learning_rate": 4.155844155844157e-06,
      "loss": 0.7403,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.731417179107666,
      "step": 25,
      "valid_targets_mean": 9558.7,
      "valid_targets_min": 322
    },
    {
      "epoch": 0.09090909090909091,
      "grad_norm": 1.2854702472686768,
      "learning_rate": 5.021645021645022e-06,
      "loss": 0.7163,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.6908976435661316,
      "step": 30,
      "valid_targets_mean": 10105.4,
      "valid_targets_min": 1608
    },
    {
      "epoch": 0.10606060606060606,
      "grad_norm": 1.2572232484817505,
      "learning_rate": 5.887445887445888e-06,
      "loss": 0.712,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.6961195468902588,
      "step": 35,
      "valid_targets_mean": 10464.2,
      "valid_targets_min": 1526
    },
    {
      "epoch": 0.12121212121212122,
      "grad_norm": 0.9280831813812256,
      "learning_rate": 6.753246753246754e-06,
      "loss": 0.7035,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.7104423642158508,
      "step": 40,
      "valid_targets_mean": 11346.2,
      "valid_targets_min": 1483
    },
    {
      "epoch": 0.13636363636363635,
      "grad_norm": 0.7417172193527222,
      "learning_rate": 7.61904761904762e-06,
      "loss": 0.6909,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.7006170153617859,
      "step": 45,
      "valid_targets_mean": 9975.1,
      "valid_targets_min": 1161
    },
    {
      "epoch": 0.15151515151515152,
      "grad_norm": 0.6055828332901001,
      "learning_rate": 8.484848484848486e-06,
      "loss": 0.6671,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.6549615263938904,
      "step": 50,
      "valid_targets_mean": 10963.4,
      "valid_targets_min": 2224
    },
    {
      "epoch": 0.16666666666666666,
      "grad_norm": 0.5373390913009644,
      "learning_rate": 9.350649350649352e-06,
      "loss": 0.647,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.6528773903846741,
      "step": 55,
      "valid_targets_mean": 11033.7,
      "valid_targets_min": 1901
    },
    {
      "epoch": 0.18181818181818182,
      "grad_norm": 0.4637213945388794,
      "learning_rate": 1.0216450216450218e-05,
      "loss": 0.6331,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.6208438873291016,
      "step": 60,
      "valid_targets_mean": 11376.1,
      "valid_targets_min": 1419
    },
    {
      "epoch": 0.19696969696969696,
      "grad_norm": 0.3759629428386688,
      "learning_rate": 1.1082251082251083e-05,
      "loss": 0.6304,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.6072301864624023,
      "step": 65,
      "valid_targets_mean": 11353.7,
      "valid_targets_min": 2013
    },
    {
      "epoch": 0.21212121212121213,
      "grad_norm": 0.3742857277393341,
      "learning_rate": 1.1948051948051949e-05,
      "loss": 0.615,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.6038457751274109,
      "step": 70,
      "valid_targets_mean": 9057.6,
      "valid_targets_min": 1241
    },
    {
      "epoch": 0.22727272727272727,
      "grad_norm": 0.35024023056030273,
      "learning_rate": 1.2813852813852813e-05,
      "loss": 0.5975,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5827907919883728,
      "step": 75,
      "valid_targets_mean": 9869.5,
      "valid_targets_min": 562
    },
    {
      "epoch": 0.24242424242424243,
      "grad_norm": 0.28136569261550903,
      "learning_rate": 1.3679653679653682e-05,
      "loss": 0.5938,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5833486914634705,
      "step": 80,
      "valid_targets_mean": 11238.7,
      "valid_targets_min": 1250
    },
    {
      "epoch": 0.25757575757575757,
      "grad_norm": 0.32320553064346313,
      "learning_rate": 1.4545454545454546e-05,
      "loss": 0.5793,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5890030860900879,
      "step": 85,
      "valid_targets_mean": 9924.8,
      "valid_targets_min": 643
    },
    {
      "epoch": 0.2727272727272727,
      "grad_norm": 0.3328065872192383,
      "learning_rate": 1.5411255411255414e-05,
      "loss": 0.5771,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5730295181274414,
      "step": 90,
      "valid_targets_mean": 11029.4,
      "valid_targets_min": 766
    },
    {
      "epoch": 0.2878787878787879,
      "grad_norm": 0.2945171594619751,
      "learning_rate": 1.6277056277056278e-05,
      "loss": 0.5776,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5663856863975525,
      "step": 95,
      "valid_targets_mean": 9861.0,
      "valid_targets_min": 290
    },
    {
      "epoch": 0.30303030303030304,
      "grad_norm": 0.35201555490493774,
      "learning_rate": 1.7142857142857142e-05,
      "loss": 0.5639,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5708717107772827,
      "step": 100,
      "valid_targets_mean": 10526.6,
      "valid_targets_min": 975
    },
    {
      "epoch": 0.3181818181818182,
      "grad_norm": 0.3862834870815277,
      "learning_rate": 1.800865800865801e-05,
      "loss": 0.552,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5564579963684082,
      "step": 105,
      "valid_targets_mean": 10206.7,
      "valid_targets_min": 777
    },
    {
      "epoch": 0.3333333333333333,
      "grad_norm": 0.2635056972503662,
      "learning_rate": 1.8874458874458877e-05,
      "loss": 0.5478,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5560527443885803,
      "step": 110,
      "valid_targets_mean": 10152.2,
      "valid_targets_min": 421
    },
    {
      "epoch": 0.3484848484848485,
      "grad_norm": 0.31485870480537415,
      "learning_rate": 1.974025974025974e-05,
      "loss": 0.5581,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5637703537940979,
      "step": 115,
      "valid_targets_mean": 11096.5,
      "valid_targets_min": 1760
    },
    {
      "epoch": 0.36363636363636365,
      "grad_norm": 0.2724863886833191,
      "learning_rate": 2.0606060606060608e-05,
      "loss": 0.5357,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5307426452636719,
      "step": 120,
      "valid_targets_mean": 10088.6,
      "valid_targets_min": 1757
    },
    {
      "epoch": 0.3787878787878788,
      "grad_norm": 0.32219961285591125,
      "learning_rate": 2.1471861471861476e-05,
      "loss": 0.5402,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5429307818412781,
      "step": 125,
      "valid_targets_mean": 10678.2,
      "valid_targets_min": 1876
    },
    {
      "epoch": 0.3939393939393939,
      "grad_norm": 0.27818748354911804,
      "learning_rate": 2.233766233766234e-05,
      "loss": 0.5364,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5366330146789551,
      "step": 130,
      "valid_targets_mean": 11813.2,
      "valid_targets_min": 1946
    },
    {
      "epoch": 0.4090909090909091,
      "grad_norm": 0.2648671269416809,
      "learning_rate": 2.3203463203463207e-05,
      "loss": 0.5305,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5115221738815308,
      "step": 135,
      "valid_targets_mean": 9778.8,
      "valid_targets_min": 1214
    },
    {
      "epoch": 0.42424242424242425,
      "grad_norm": 0.381243497133255,
      "learning_rate": 2.4069264069264074e-05,
      "loss": 0.5251,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5276926159858704,
      "step": 140,
      "valid_targets_mean": 11106.9,
      "valid_targets_min": 343
    },
    {
      "epoch": 0.4393939393939394,
      "grad_norm": 0.2855749726295471,
      "learning_rate": 2.4935064935064935e-05,
      "loss": 0.5226,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5156047344207764,
      "step": 145,
      "valid_targets_mean": 9952.1,
      "valid_targets_min": 1655
    },
    {
      "epoch": 0.45454545454545453,
      "grad_norm": 0.2713521420955658,
      "learning_rate": 2.5800865800865803e-05,
      "loss": 0.5189,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5275256037712097,
      "step": 150,
      "valid_targets_mean": 10284.1,
      "valid_targets_min": 1852
    },
    {
      "epoch": 0.4696969696969697,
      "grad_norm": 0.43924787640571594,
      "learning_rate": 2.6666666666666667e-05,
      "loss": 0.5126,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5050260424613953,
      "step": 155,
      "valid_targets_mean": 9857.3,
      "valid_targets_min": 1508
    },
    {
      "epoch": 0.48484848484848486,
      "grad_norm": 0.3208458721637726,
      "learning_rate": 2.7532467532467534e-05,
      "loss": 0.5205,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5271092653274536,
      "step": 160,
      "valid_targets_mean": 9489.1,
      "valid_targets_min": 164
    },
    {
      "epoch": 0.5,
      "grad_norm": 0.2969353199005127,
      "learning_rate": 2.83982683982684e-05,
      "loss": 0.5078,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5159955620765686,
      "step": 165,
      "valid_targets_mean": 11467.4,
      "valid_targets_min": 1363
    },
    {
      "epoch": 0.5151515151515151,
      "grad_norm": 0.24687372148036957,
      "learning_rate": 2.9264069264069265e-05,
      "loss": 0.504,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5024310946464539,
      "step": 170,
      "valid_targets_mean": 10679.8,
      "valid_targets_min": 996
    },
    {
      "epoch": 0.5303030303030303,
      "grad_norm": 0.22991038858890533,
      "learning_rate": 3.0129870129870133e-05,
      "loss": 0.5106,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4929521381855011,
      "step": 175,
      "valid_targets_mean": 10705.8,
      "valid_targets_min": 2042
    },
    {
      "epoch": 0.5454545454545454,
      "grad_norm": 0.2597404420375824,
      "learning_rate": 3.0995670995671e-05,
      "loss": 0.5122,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5060884356498718,
      "step": 180,
      "valid_targets_mean": 10133.8,
      "valid_targets_min": 1074
    },
    {
      "epoch": 0.5606060606060606,
      "grad_norm": 0.2618630826473236,
      "learning_rate": 3.1861471861471864e-05,
      "loss": 0.5062,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5122888684272766,
      "step": 185,
      "valid_targets_mean": 11173.3,
      "valid_targets_min": 812
    },
    {
      "epoch": 0.5757575757575758,
      "grad_norm": 0.2614921033382416,
      "learning_rate": 3.272727272727273e-05,
      "loss": 0.5063,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5075699687004089,
      "step": 190,
      "valid_targets_mean": 10824.9,
      "valid_targets_min": 1149
    },
    {
      "epoch": 0.5909090909090909,
      "grad_norm": 0.27503690123558044,
      "learning_rate": 3.359307359307359e-05,
      "loss": 0.5051,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5058221817016602,
      "step": 195,
      "valid_targets_mean": 10300.2,
      "valid_targets_min": 2006
    },
    {
      "epoch": 0.6060606060606061,
      "grad_norm": 0.28348851203918457,
      "learning_rate": 3.445887445887446e-05,
      "loss": 0.5005,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.49363622069358826,
      "step": 200,
      "valid_targets_mean": 10661.9,
      "valid_targets_min": 1772
    },
    {
      "epoch": 0.6212121212121212,
      "grad_norm": 0.3084920644760132,
      "learning_rate": 3.532467532467533e-05,
      "loss": 0.4991,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4782664477825165,
      "step": 205,
      "valid_targets_mean": 9761.5,
      "valid_targets_min": 1884
    },
    {
      "epoch": 0.6363636363636364,
      "grad_norm": 0.32024168968200684,
      "learning_rate": 3.6190476190476195e-05,
      "loss": 0.4995,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5088427066802979,
      "step": 210,
      "valid_targets_mean": 10326.9,
      "valid_targets_min": 1318
    },
    {
      "epoch": 0.6515151515151515,
      "grad_norm": 0.29676759243011475,
      "learning_rate": 3.705627705627706e-05,
      "loss": 0.4952,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5000290274620056,
      "step": 215,
      "valid_targets_mean": 11271.6,
      "valid_targets_min": 1957
    },
    {
      "epoch": 0.6666666666666666,
      "grad_norm": 0.28237777948379517,
      "learning_rate": 3.792207792207792e-05,
      "loss": 0.4901,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.480541855096817,
      "step": 220,
      "valid_targets_mean": 10033.5,
      "valid_targets_min": 2683
    },
    {
      "epoch": 0.6818181818181818,
      "grad_norm": 0.2501182556152344,
      "learning_rate": 3.878787878787879e-05,
      "loss": 0.4807,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4864450991153717,
      "step": 225,
      "valid_targets_mean": 9482.9,
      "valid_targets_min": 2446
    },
    {
      "epoch": 0.696969696969697,
      "grad_norm": 0.2756481170654297,
      "learning_rate": 3.965367965367966e-05,
      "loss": 0.4973,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.48787084221839905,
      "step": 230,
      "valid_targets_mean": 11122.2,
      "valid_targets_min": 1090
    },
    {
      "epoch": 0.7121212121212122,
      "grad_norm": 0.32130560278892517,
      "learning_rate": 3.999979449020199e-05,
      "loss": 0.4955,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.484106183052063,
      "step": 235,
      "valid_targets_mean": 9360.9,
      "valid_targets_min": 755
    },
    {
      "epoch": 0.7272727272727273,
      "grad_norm": 0.2478799968957901,
      "learning_rate": 3.99985386122866e-05,
      "loss": 0.4852,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4770452082157135,
      "step": 240,
      "valid_targets_mean": 9947.7,
      "valid_targets_min": 326
    },
    {
      "epoch": 0.7424242424242424,
      "grad_norm": 0.29648807644844055,
      "learning_rate": 3.999614110017182e-05,
      "loss": 0.4804,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4880082309246063,
      "step": 245,
      "valid_targets_mean": 10222.1,
      "valid_targets_min": 1518
    },
    {
      "epoch": 0.7575757575757576,
      "grad_norm": 0.2879868745803833,
      "learning_rate": 3.999260209072175e-05,
      "loss": 0.4862,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.5047332644462585,
      "step": 250,
      "valid_targets_mean": 11265.0,
      "valid_targets_min": 282
    },
    {
      "epoch": 0.7727272727272727,
      "grad_norm": 0.31799957156181335,
      "learning_rate": 3.998792178596384e-05,
      "loss": 0.4812,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4731893539428711,
      "step": 255,
      "valid_targets_mean": 9441.6,
      "valid_targets_min": 1364
    },
    {
      "epoch": 0.7878787878787878,
      "grad_norm": 0.29029300808906555,
      "learning_rate": 3.998210045307744e-05,
      "loss": 0.4739,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4720456302165985,
      "step": 260,
      "valid_targets_mean": 11430.9,
      "valid_targets_min": 1778
    },
    {
      "epoch": 0.803030303030303,
      "grad_norm": 0.2993226647377014,
      "learning_rate": 3.997513842437845e-05,
      "loss": 0.4811,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4912916123867035,
      "step": 265,
      "valid_targets_mean": 11393.1,
      "valid_targets_min": 1756
    },
    {
      "epoch": 0.8181818181818182,
      "grad_norm": 0.28473803400993347,
      "learning_rate": 3.996703609730042e-05,
      "loss": 0.4734,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4677492082118988,
      "step": 270,
      "valid_targets_mean": 10704.3,
      "valid_targets_min": 1454
    },
    {
      "epoch": 0.8333333333333334,
      "grad_norm": 0.2601090371608734,
      "learning_rate": 3.995779393437183e-05,
      "loss": 0.472,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4640270173549652,
      "step": 275,
      "valid_targets_mean": 10002.1,
      "valid_targets_min": 1636
    },
    {
      "epoch": 0.8484848484848485,
      "grad_norm": 0.2986535131931305,
      "learning_rate": 3.99474124631897e-05,
      "loss": 0.481,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4762050211429596,
      "step": 280,
      "valid_targets_mean": 10547.1,
      "valid_targets_min": 1222
    },
    {
      "epoch": 0.8636363636363636,
      "grad_norm": 0.2527305483818054,
      "learning_rate": 3.993589227638944e-05,
      "loss": 0.4739,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.47839856147766113,
      "step": 285,
      "valid_targets_mean": 9557.9,
      "valid_targets_min": 1879
    },
    {
      "epoch": 0.8787878787878788,
      "grad_norm": 0.2590777575969696,
      "learning_rate": 3.992323403161109e-05,
      "loss": 0.4764,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4745398759841919,
      "step": 290,
      "valid_targets_mean": 10366.5,
      "valid_targets_min": 1483
    },
    {
      "epoch": 0.8939393939393939,
      "grad_norm": 0.27018365263938904,
      "learning_rate": 3.9909438451461695e-05,
      "loss": 0.4727,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.45954784750938416,
      "step": 295,
      "valid_targets_mean": 9689.8,
      "valid_targets_min": 322
    },
    {
      "epoch": 0.9090909090909091,
      "grad_norm": 0.2571364641189575,
      "learning_rate": 3.989450632347411e-05,
      "loss": 0.4729,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.46253645420074463,
      "step": 300,
      "valid_targets_mean": 9145.9,
      "valid_targets_min": 1569
    },
    {
      "epoch": 0.9242424242424242,
      "grad_norm": 0.26115307211875916,
      "learning_rate": 3.987843850006201e-05,
      "loss": 0.4746,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.47390684485435486,
      "step": 305,
      "valid_targets_mean": 10386.2,
      "valid_targets_min": 1218
    },
    {
      "epoch": 0.9393939393939394,
      "grad_norm": 0.2708868980407715,
      "learning_rate": 3.9861235898471265e-05,
      "loss": 0.4805,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4682181179523468,
      "step": 310,
      "valid_targets_mean": 9678.5,
      "valid_targets_min": 1660
    },
    {
      "epoch": 0.9545454545454546,
      "grad_norm": 0.2515220046043396,
      "learning_rate": 3.984289950072754e-05,
      "loss": 0.4602,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.44889721274375916,
      "step": 315,
      "valid_targets_mean": 10232.6,
      "valid_targets_min": 1685
    },
    {
      "epoch": 0.9696969696969697,
      "grad_norm": 0.2677609622478485,
      "learning_rate": 3.982343035358026e-05,
      "loss": 0.4722,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4712284505367279,
      "step": 320,
      "valid_targets_mean": 11043.2,
      "valid_targets_min": 1271
    },
    {
      "epoch": 0.9848484848484849,
      "grad_norm": 0.28743913769721985,
      "learning_rate": 3.980282956844284e-05,
      "loss": 0.4763,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4863670766353607,
      "step": 325,
      "valid_targets_mean": 9911.7,
      "valid_targets_min": 1649
    },
    {
      "epoch": 1.0,
      "grad_norm": 0.250603049993515,
      "learning_rate": 3.9781098321329266e-05,
      "loss": 0.4584,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4435311555862427,
      "step": 330,
      "valid_targets_mean": 10388.2,
      "valid_targets_min": 1503
    },
    {
      "epoch": 1.0151515151515151,
      "grad_norm": 0.24998654425144196,
      "learning_rate": 3.975823785278691e-05,
      "loss": 0.4661,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.47927233576774597,
      "step": 335,
      "valid_targets_mean": 11059.3,
      "valid_targets_min": 888
    },
    {
      "epoch": 1.0303030303030303,
      "grad_norm": 0.2713046371936798,
      "learning_rate": 3.973424946782578e-05,
      "loss": 0.4624,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4718210697174072,
      "step": 340,
      "valid_targets_mean": 10540.3,
      "valid_targets_min": 2162
    },
    {
      "epoch": 1.0454545454545454,
      "grad_norm": 0.2729426622390747,
      "learning_rate": 3.970913453584397e-05,
      "loss": 0.4598,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4586052894592285,
      "step": 345,
      "valid_targets_mean": 10517.5,
      "valid_targets_min": 1042
    },
    {
      "epoch": 1.0606060606060606,
      "grad_norm": 0.25700637698173523,
      "learning_rate": 3.9682894490549485e-05,
      "loss": 0.4672,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.47395238280296326,
      "step": 350,
      "valid_targets_mean": 10677.8,
      "valid_targets_min": 1696
    },
    {
      "epoch": 1.0757575757575757,
      "grad_norm": 0.2859768569469452,
      "learning_rate": 3.965553082987846e-05,
      "loss": 0.4596,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4566628932952881,
      "step": 355,
      "valid_targets_mean": 10823.8,
      "valid_targets_min": 755
    },
    {
      "epoch": 1.0909090909090908,
      "grad_norm": 0.26943156123161316,
      "learning_rate": 3.9627045115909556e-05,
      "loss": 0.4594,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4670834541320801,
      "step": 360,
      "valid_targets_mean": 9804.6,
      "valid_targets_min": 1462
    },
    {
      "epoch": 1.106060606060606,
      "grad_norm": 0.29486578702926636,
      "learning_rate": 3.959743897477486e-05,
      "loss": 0.4653,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4612759053707123,
      "step": 365,
      "valid_targets_mean": 9946.5,
      "valid_targets_min": 209
    },
    {
      "epoch": 1.121212121212121,
      "grad_norm": 0.28370463848114014,
      "learning_rate": 3.9566714096567035e-05,
      "loss": 0.4719,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.46476736664772034,
      "step": 370,
      "valid_targets_mean": 9901.6,
      "valid_targets_min": 1649
    },
    {
      "epoch": 1.1363636363636362,
      "grad_norm": 0.2426619678735733,
      "learning_rate": 3.953487223524283e-05,
      "loss": 0.4528,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.44726458191871643,
      "step": 375,
      "valid_targets_mean": 10916.4,
      "valid_targets_min": 2120
    },
    {
      "epoch": 1.1515151515151516,
      "grad_norm": 0.23527388274669647,
      "learning_rate": 3.950191520852294e-05,
      "loss": 0.4569,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4509541988372803,
      "step": 380,
      "valid_targets_mean": 9661.9,
      "valid_targets_min": 1867
    },
    {
      "epoch": 1.1666666666666667,
      "grad_norm": 0.2603897452354431,
      "learning_rate": 3.94678448977883e-05,
      "loss": 0.4563,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.46255016326904297,
      "step": 385,
      "valid_targets_mean": 10519.6,
      "valid_targets_min": 2057
    },
    {
      "epoch": 1.1818181818181819,
      "grad_norm": 0.2214042991399765,
      "learning_rate": 3.9432663247972614e-05,
      "loss": 0.4606,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.45441946387290955,
      "step": 390,
      "valid_targets_mean": 10959.6,
      "valid_targets_min": 2036
    },
    {
      "epoch": 1.196969696969697,
      "grad_norm": 0.24642735719680786,
      "learning_rate": 3.9396372267451356e-05,
      "loss": 0.4606,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4440148174762726,
      "step": 395,
      "valid_targets_mean": 9717.3,
      "valid_targets_min": 1135
    },
    {
      "epoch": 1.2121212121212122,
      "grad_norm": 0.275397926568985,
      "learning_rate": 3.935897402792713e-05,
      "loss": 0.4536,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.45174315571784973,
      "step": 400,
      "valid_targets_mean": 11004.2,
      "valid_targets_min": 1795
    },
    {
      "epoch": 1.2272727272727273,
      "grad_norm": 0.22368259727954865,
      "learning_rate": 3.93204706643114e-05,
      "loss": 0.4555,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4407337009906769,
      "step": 405,
      "valid_targets_mean": 9665.6,
      "valid_targets_min": 1125
    },
    {
      "epoch": 1.2424242424242424,
      "grad_norm": 0.24048179388046265,
      "learning_rate": 3.92808643746026e-05,
      "loss": 0.457,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4605706036090851,
      "step": 410,
      "valid_targets_mean": 10396.8,
      "valid_targets_min": 2126
    },
    {
      "epoch": 1.2575757575757576,
      "grad_norm": 0.2720293700695038,
      "learning_rate": 3.924015741976069e-05,
      "loss": 0.4605,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.47806239128112793,
      "step": 415,
      "valid_targets_mean": 9976.4,
      "valid_targets_min": 1110
    },
    {
      "epoch": 1.2727272727272727,
      "grad_norm": 0.2574106454849243,
      "learning_rate": 3.919835212357805e-05,
      "loss": 0.4597,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4556029736995697,
      "step": 420,
      "valid_targets_mean": 11291.8,
      "valid_targets_min": 2367
    },
    {
      "epoch": 1.2878787878787878,
      "grad_norm": 0.2812650203704834,
      "learning_rate": 3.915545087254685e-05,
      "loss": 0.4498,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.46366044878959656,
      "step": 425,
      "valid_targets_mean": 10460.3,
      "valid_targets_min": 1297
    },
    {
      "epoch": 1.303030303030303,
      "grad_norm": 0.23735706508159637,
      "learning_rate": 3.911145611572282e-05,
      "loss": 0.4458,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.44674479961395264,
      "step": 430,
      "valid_targets_mean": 10087.7,
      "valid_targets_min": 1144
    },
    {
      "epoch": 1.3181818181818181,
      "grad_norm": 0.25153276324272156,
      "learning_rate": 3.906637036458541e-05,
      "loss": 0.451,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.45029929280281067,
      "step": 435,
      "valid_targets_mean": 9777.4,
      "valid_targets_min": 1204
    },
    {
      "epoch": 1.3333333333333333,
      "grad_norm": 0.25070884823799133,
      "learning_rate": 3.902019619289446e-05,
      "loss": 0.4622,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.46781739592552185,
      "step": 440,
      "valid_targets_mean": 10395.3,
      "valid_targets_min": 1728
    },
    {
      "epoch": 1.3484848484848486,
      "grad_norm": 0.24307523667812347,
      "learning_rate": 3.897293623654325e-05,
      "loss": 0.4482,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.44386550784111023,
      "step": 445,
      "valid_targets_mean": 9848.8,
      "valid_targets_min": 1607
    },
    {
      "epoch": 1.3636363636363638,
      "grad_norm": 0.24276429414749146,
      "learning_rate": 3.892459319340802e-05,
      "loss": 0.4548,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43963709473609924,
      "step": 450,
      "valid_targets_mean": 10101.5,
      "valid_targets_min": 1663
    },
    {
      "epoch": 1.378787878787879,
      "grad_norm": 0.249475359916687,
      "learning_rate": 3.8875169823194e-05,
      "loss": 0.4502,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4495883285999298,
      "step": 455,
      "valid_targets_mean": 10685.9,
      "valid_targets_min": 1915
    },
    {
      "epoch": 1.393939393939394,
      "grad_norm": 0.2366575002670288,
      "learning_rate": 3.882466894727784e-05,
      "loss": 0.4546,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4443645477294922,
      "step": 460,
      "valid_targets_mean": 10432.0,
      "valid_targets_min": 1490
    },
    {
      "epoch": 1.4090909090909092,
      "grad_norm": 0.2537292540073395,
      "learning_rate": 3.8773093448546525e-05,
      "loss": 0.4564,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.44205543398857117,
      "step": 465,
      "valid_targets_mean": 9609.5,
      "valid_targets_min": 1488
    },
    {
      "epoch": 1.4242424242424243,
      "grad_norm": 0.2168312817811966,
      "learning_rate": 3.872044627123285e-05,
      "loss": 0.4575,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4481028616428375,
      "step": 470,
      "valid_targets_mean": 11010.4,
      "valid_targets_min": 1958
    },
    {
      "epoch": 1.4393939393939394,
      "grad_norm": 0.27858564257621765,
      "learning_rate": 3.8666730420747336e-05,
      "loss": 0.4485,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4395351707935333,
      "step": 475,
      "valid_targets_mean": 10812.4,
      "valid_targets_min": 1165
    },
    {
      "epoch": 1.4545454545454546,
      "grad_norm": 0.2815878093242645,
      "learning_rate": 3.861194896350664e-05,
      "loss": 0.4512,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4515993595123291,
      "step": 480,
      "valid_targets_mean": 9597.8,
      "valid_targets_min": 1439
    },
    {
      "epoch": 1.4696969696969697,
      "grad_norm": 0.36264467239379883,
      "learning_rate": 3.855610502675851e-05,
      "loss": 0.458,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4578615725040436,
      "step": 485,
      "valid_targets_mean": 11116.2,
      "valid_targets_min": 1083
    },
    {
      "epoch": 1.4848484848484849,
      "grad_norm": 0.336082398891449,
      "learning_rate": 3.8499201798403303e-05,
      "loss": 0.4493,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4488205909729004,
      "step": 490,
      "valid_targets_mean": 10286.6,
      "valid_targets_min": 1727
    },
    {
      "epoch": 1.5,
      "grad_norm": 0.2562655210494995,
      "learning_rate": 3.8441242526811934e-05,
      "loss": 0.4502,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4542854130268097,
      "step": 495,
      "valid_targets_mean": 9711.1,
      "valid_targets_min": 1772
    },
    {
      "epoch": 1.5151515151515151,
      "grad_norm": 0.2148941308259964,
      "learning_rate": 3.8382230520640506e-05,
      "loss": 0.4559,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.45196518301963806,
      "step": 500,
      "valid_targets_mean": 11280.3,
      "valid_targets_min": 1100
    },
    {
      "epoch": 1.5303030303030303,
      "grad_norm": 0.22475746273994446,
      "learning_rate": 3.832216914864139e-05,
      "loss": 0.4544,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4607529938220978,
      "step": 505,
      "valid_targets_mean": 10677.9,
      "valid_targets_min": 1161
    },
    {
      "epoch": 1.5454545454545454,
      "grad_norm": 0.2460007518529892,
      "learning_rate": 3.826106183947095e-05,
      "loss": 0.4406,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43116143345832825,
      "step": 510,
      "valid_targets_mean": 9641.8,
      "valid_targets_min": 349
    },
    {
      "epoch": 1.5606060606060606,
      "grad_norm": 0.2652975916862488,
      "learning_rate": 3.8198912081493735e-05,
      "loss": 0.4462,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4421050250530243,
      "step": 515,
      "valid_targets_mean": 10400.5,
      "valid_targets_min": 1542
    },
    {
      "epoch": 1.5757575757575757,
      "grad_norm": 0.22775985300540924,
      "learning_rate": 3.813572342258349e-05,
      "loss": 0.4492,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4473375380039215,
      "step": 520,
      "valid_targets_mean": 9271.8,
      "valid_targets_min": 751
    },
    {
      "epoch": 1.5909090909090908,
      "grad_norm": 0.216358482837677,
      "learning_rate": 3.807149946992047e-05,
      "loss": 0.4526,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4686870574951172,
      "step": 525,
      "valid_targets_mean": 11210.2,
      "valid_targets_min": 1942
    },
    {
      "epoch": 1.606060606060606,
      "grad_norm": 0.2305910736322403,
      "learning_rate": 3.800624388978561e-05,
      "loss": 0.4373,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4444570541381836,
      "step": 530,
      "valid_targets_mean": 11183.9,
      "valid_targets_min": 366
    },
    {
      "epoch": 1.621212121212121,
      "grad_norm": 0.2128409594297409,
      "learning_rate": 3.793996040735119e-05,
      "loss": 0.4479,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4667322635650635,
      "step": 535,
      "valid_targets_mean": 10397.2,
      "valid_targets_min": 975
    },
    {
      "epoch": 1.6363636363636362,
      "grad_norm": 0.20614521205425262,
      "learning_rate": 3.787265280646825e-05,
      "loss": 0.4499,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43987563252449036,
      "step": 540,
      "valid_targets_mean": 11167.8,
      "valid_targets_min": 891
    },
    {
      "epoch": 1.6515151515151514,
      "grad_norm": 0.28170475363731384,
      "learning_rate": 3.7804324929450465e-05,
      "loss": 0.4433,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4500798285007477,
      "step": 545,
      "valid_targets_mean": 10240.6,
      "valid_targets_min": 442
    },
    {
      "epoch": 1.6666666666666665,
      "grad_norm": 0.22459731996059418,
      "learning_rate": 3.7734980676854916e-05,
      "loss": 0.4356,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4508148729801178,
      "step": 550,
      "valid_targets_mean": 9347.4,
      "valid_targets_min": 1364
    },
    {
      "epoch": 1.6818181818181817,
      "grad_norm": 0.20802566409111023,
      "learning_rate": 3.7664624007259375e-05,
      "loss": 0.4459,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.44437363743782043,
      "step": 555,
      "valid_targets_mean": 10428.0,
      "valid_targets_min": 1455
    },
    {
      "epoch": 1.696969696969697,
      "grad_norm": 0.2239496111869812,
      "learning_rate": 3.759325893703631e-05,
      "loss": 0.4477,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.45915719866752625,
      "step": 560,
      "valid_targets_mean": 10657.2,
      "valid_targets_min": 1250
    },
    {
      "epoch": 1.7121212121212122,
      "grad_norm": 0.26243460178375244,
      "learning_rate": 3.752088954012366e-05,
      "loss": 0.4522,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4665350914001465,
      "step": 565,
      "valid_targets_mean": 10759.9,
      "valid_targets_min": 1218
    },
    {
      "epoch": 1.7272727272727273,
      "grad_norm": 0.23304878175258636,
      "learning_rate": 3.74475199477922e-05,
      "loss": 0.4408,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43713513016700745,
      "step": 570,
      "valid_targets_mean": 10863.8,
      "valid_targets_min": 1753
    },
    {
      "epoch": 1.7424242424242424,
      "grad_norm": 0.2664192318916321,
      "learning_rate": 3.7373154348409775e-05,
      "loss": 0.4345,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4253872334957123,
      "step": 575,
      "valid_targets_mean": 10234.6,
      "valid_targets_min": 1912
    },
    {
      "epoch": 1.7575757575757576,
      "grad_norm": 0.24994930624961853,
      "learning_rate": 3.729779698720215e-05,
      "loss": 0.4484,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4481925070285797,
      "step": 580,
      "valid_targets_mean": 10805.3,
      "valid_targets_min": 1526
    },
    {
      "epoch": 1.7727272727272727,
      "grad_norm": 0.2276647984981537,
      "learning_rate": 3.7221452166010704e-05,
      "loss": 0.4446,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4426651895046234,
      "step": 585,
      "valid_targets_mean": 9971.8,
      "valid_targets_min": 2410
    },
    {
      "epoch": 1.7878787878787878,
      "grad_norm": 0.20459814369678497,
      "learning_rate": 3.7144124243046815e-05,
      "loss": 0.4468,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4608907401561737,
      "step": 590,
      "valid_targets_mean": 10947.9,
      "valid_targets_min": 1074
    },
    {
      "epoch": 1.803030303030303,
      "grad_norm": 0.2181711047887802,
      "learning_rate": 3.7065817632643115e-05,
      "loss": 0.4385,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4517947733402252,
      "step": 595,
      "valid_targets_mean": 10850.7,
      "valid_targets_min": 2683
    },
    {
      "epoch": 1.8181818181818183,
      "grad_norm": 0.23002681136131287,
      "learning_rate": 3.6986536805001466e-05,
      "loss": 0.4459,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4446265697479248,
      "step": 600,
      "valid_targets_mean": 10814.6,
      "valid_targets_min": 1822
    },
    {
      "epoch": 1.8333333333333335,
      "grad_norm": 0.22724074125289917,
      "learning_rate": 3.690628628593777e-05,
      "loss": 0.444,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4290066063404083,
      "step": 605,
      "valid_targets_mean": 10667.6,
      "valid_targets_min": 2435
    },
    {
      "epoch": 1.8484848484848486,
      "grad_norm": 0.22361937165260315,
      "learning_rate": 3.6825070656623626e-05,
      "loss": 0.4392,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4376356601715088,
      "step": 610,
      "valid_targets_mean": 10062.2,
      "valid_targets_min": 477
    },
    {
      "epoch": 1.8636363636363638,
      "grad_norm": 0.23989582061767578,
      "learning_rate": 3.67428945533248e-05,
      "loss": 0.4523,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.47073546051979065,
      "step": 615,
      "valid_targets_mean": 9944.7,
      "valid_targets_min": 2013
    },
    {
      "epoch": 1.878787878787879,
      "grad_norm": 0.250827819108963,
      "learning_rate": 3.6659762667136567e-05,
      "loss": 0.4423,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4399884045124054,
      "step": 620,
      "valid_targets_mean": 10567.1,
      "valid_targets_min": 1374
    },
    {
      "epoch": 1.893939393939394,
      "grad_norm": 0.22220048308372498,
      "learning_rate": 3.65756797437159e-05,
      "loss": 0.4385,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4380852282047272,
      "step": 625,
      "valid_targets_mean": 10540.8,
      "valid_targets_min": 406
    },
    {
      "epoch": 1.9090909090909092,
      "grad_norm": 0.24989663064479828,
      "learning_rate": 3.6490650583010585e-05,
      "loss": 0.4487,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4289959967136383,
      "step": 630,
      "valid_targets_mean": 10362.9,
      "valid_targets_min": 1140
    },
    {
      "epoch": 1.9242424242424243,
      "grad_norm": 0.23470529913902283,
      "learning_rate": 3.640468003898518e-05,
      "loss": 0.4349,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4211912453174591,
      "step": 635,
      "valid_targets_mean": 10182.2,
      "valid_targets_min": 1609
    },
    {
      "epoch": 1.9393939393939394,
      "grad_norm": 0.23092474043369293,
      "learning_rate": 3.631777301934395e-05,
      "loss": 0.4499,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43969210982322693,
      "step": 640,
      "valid_targets_mean": 11027.6,
      "valid_targets_min": 766
    },
    {
      "epoch": 1.9545454545454546,
      "grad_norm": 0.24192656576633453,
      "learning_rate": 3.6229934485250684e-05,
      "loss": 0.4335,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.44624972343444824,
      "step": 645,
      "valid_targets_mean": 9532.8,
      "valid_targets_min": 1717
    },
    {
      "epoch": 1.9696969696969697,
      "grad_norm": 0.27132418751716614,
      "learning_rate": 3.6141169451045526e-05,
      "loss": 0.4414,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.445227712392807,
      "step": 650,
      "valid_targets_mean": 10008.7,
      "valid_targets_min": 1695
    },
    {
      "epoch": 1.9848484848484849,
      "grad_norm": 0.2537304759025574,
      "learning_rate": 3.605148298395865e-05,
      "loss": 0.4341,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42087364196777344,
      "step": 655,
      "valid_targets_mean": 9409.2,
      "valid_targets_min": 2144
    },
    {
      "epoch": 2.0,
      "grad_norm": 0.22206485271453857,
      "learning_rate": 3.5960880203821086e-05,
      "loss": 0.4409,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4437296688556671,
      "step": 660,
      "valid_targets_mean": 10845.7,
      "valid_targets_min": 1506
    },
    {
      "epoch": 2.015151515151515,
      "grad_norm": 0.2415200173854828,
      "learning_rate": 3.5869366282772354e-05,
      "loss": 0.4263,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.427796870470047,
      "step": 665,
      "valid_targets_mean": 10783.5,
      "valid_targets_min": 2372
    },
    {
      "epoch": 2.0303030303030303,
      "grad_norm": 0.2219669669866562,
      "learning_rate": 3.577694644496529e-05,
      "loss": 0.4387,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.44274500012397766,
      "step": 670,
      "valid_targets_mean": 10079.4,
      "valid_targets_min": 597
    },
    {
      "epoch": 2.0454545454545454,
      "grad_norm": 0.2286299616098404,
      "learning_rate": 3.568362596626778e-05,
      "loss": 0.4322,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42993131279945374,
      "step": 675,
      "valid_targets_mean": 10566.7,
      "valid_targets_min": 209
    },
    {
      "epoch": 2.0606060606060606,
      "grad_norm": 0.2039332240819931,
      "learning_rate": 3.558941017396158e-05,
      "loss": 0.4467,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42933186888694763,
      "step": 680,
      "valid_targets_mean": 10816.9,
      "valid_targets_min": 2207
    },
    {
      "epoch": 2.0757575757575757,
      "grad_norm": 0.22693563997745514,
      "learning_rate": 3.549430444643821e-05,
      "loss": 0.4349,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4335421621799469,
      "step": 685,
      "valid_targets_mean": 10083.9,
      "valid_targets_min": 810
    },
    {
      "epoch": 2.090909090909091,
      "grad_norm": 0.21606650948524475,
      "learning_rate": 3.539831421289195e-05,
      "loss": 0.4354,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4249735176563263,
      "step": 690,
      "valid_targets_mean": 9284.2,
      "valid_targets_min": 2048
    },
    {
      "epoch": 2.106060606060606,
      "grad_norm": 0.2451007068157196,
      "learning_rate": 3.530144495300984e-05,
      "loss": 0.4323,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42328715324401855,
      "step": 695,
      "valid_targets_mean": 10665.6,
      "valid_targets_min": 1087
    },
    {
      "epoch": 2.121212121212121,
      "grad_norm": 0.206374391913414,
      "learning_rate": 3.520370219665897e-05,
      "loss": 0.4287,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41378211975097656,
      "step": 700,
      "valid_targets_mean": 9846.3,
      "valid_targets_min": 349
    },
    {
      "epoch": 2.1363636363636362,
      "grad_norm": 0.21436245739459991,
      "learning_rate": 3.51050915235707e-05,
      "loss": 0.437,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41966649889945984,
      "step": 705,
      "valid_targets_mean": 9895.6,
      "valid_targets_min": 1697
    },
    {
      "epoch": 2.1515151515151514,
      "grad_norm": 0.21009962260723114,
      "learning_rate": 3.500561856302221e-05,
      "loss": 0.43,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42929863929748535,
      "step": 710,
      "valid_targets_mean": 10501.0,
      "valid_targets_min": 1895
    },
    {
      "epoch": 2.1666666666666665,
      "grad_norm": 0.19127050042152405,
      "learning_rate": 3.49052889935151e-05,
      "loss": 0.4322,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42854270339012146,
      "step": 715,
      "valid_targets_mean": 11479.1,
      "valid_targets_min": 1830
    },
    {
      "epoch": 2.1818181818181817,
      "grad_norm": 0.22553083300590515,
      "learning_rate": 3.480410854245125e-05,
      "loss": 0.4391,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4565233290195465,
      "step": 720,
      "valid_targets_mean": 11101.6,
      "valid_targets_min": 1640
    },
    {
      "epoch": 2.196969696969697,
      "grad_norm": 0.7163257002830505,
      "learning_rate": 3.470208298580589e-05,
      "loss": 0.4309,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4279252588748932,
      "step": 725,
      "valid_targets_mean": 9545.3,
      "valid_targets_min": 1526
    },
    {
      "epoch": 2.212121212121212,
      "grad_norm": 0.20563149452209473,
      "learning_rate": 3.459921814779781e-05,
      "loss": 0.4321,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4246140420436859,
      "step": 730,
      "valid_targets_mean": 10329.9,
      "valid_targets_min": 1483
    },
    {
      "epoch": 2.227272727272727,
      "grad_norm": 0.20377157628536224,
      "learning_rate": 3.449551990055694e-05,
      "loss": 0.4345,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4424196183681488,
      "step": 735,
      "valid_targets_mean": 11953.7,
      "valid_targets_min": 1887
    },
    {
      "epoch": 2.242424242424242,
      "grad_norm": 0.23602178692817688,
      "learning_rate": 3.43909941637891e-05,
      "loss": 0.4386,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43257784843444824,
      "step": 740,
      "valid_targets_mean": 10169.9,
      "valid_targets_min": 1542
    },
    {
      "epoch": 2.257575757575758,
      "grad_norm": 0.2238876223564148,
      "learning_rate": 3.428564690443807e-05,
      "loss": 0.4368,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42592504620552063,
      "step": 745,
      "valid_targets_mean": 10588.3,
      "valid_targets_min": 1805
    },
    {
      "epoch": 2.2727272727272725,
      "grad_norm": 0.20250310003757477,
      "learning_rate": 3.4179484136345e-05,
      "loss": 0.4399,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4334385395050049,
      "step": 750,
      "valid_targets_mean": 11248.8,
      "valid_targets_min": 1192
    },
    {
      "epoch": 2.287878787878788,
      "grad_norm": 0.2033395618200302,
      "learning_rate": 3.4072511919905055e-05,
      "loss": 0.4293,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42527851462364197,
      "step": 755,
      "valid_targets_mean": 10569.0,
      "valid_targets_min": 417
    },
    {
      "epoch": 2.303030303030303,
      "grad_norm": 0.2330806702375412,
      "learning_rate": 3.396473636172146e-05,
      "loss": 0.4372,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43682384490966797,
      "step": 760,
      "valid_targets_mean": 10564.2,
      "valid_targets_min": 1991
    },
    {
      "epoch": 2.3181818181818183,
      "grad_norm": 0.22088605165481567,
      "learning_rate": 3.385616361425694e-05,
      "loss": 0.4291,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43701648712158203,
      "step": 765,
      "valid_targets_mean": 10605.5,
      "valid_targets_min": 1296
    },
    {
      "epoch": 2.3333333333333335,
      "grad_norm": 0.20546109974384308,
      "learning_rate": 3.374679987548245e-05,
      "loss": 0.4301,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4188653230667114,
      "step": 770,
      "valid_targets_mean": 9971.1,
      "valid_targets_min": 3344
    },
    {
      "epoch": 2.3484848484848486,
      "grad_norm": 0.22689083218574524,
      "learning_rate": 3.363665138852339e-05,
      "loss": 0.4265,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4298488199710846,
      "step": 775,
      "valid_targets_mean": 10632.2,
      "valid_targets_min": 1811
    },
    {
      "epoch": 2.3636363636363638,
      "grad_norm": 0.20038045942783356,
      "learning_rate": 3.3525724441303206e-05,
      "loss": 0.4268,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4342721998691559,
      "step": 780,
      "valid_targets_mean": 11544.0,
      "valid_targets_min": 1656
    },
    {
      "epoch": 2.378787878787879,
      "grad_norm": 0.22986885905265808,
      "learning_rate": 3.341402536618442e-05,
      "loss": 0.4392,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43103542923927307,
      "step": 785,
      "valid_targets_mean": 9285.2,
      "valid_targets_min": 1612
    },
    {
      "epoch": 2.393939393939394,
      "grad_norm": 0.21807241439819336,
      "learning_rate": 3.3301560539607174e-05,
      "loss": 0.4344,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43380093574523926,
      "step": 790,
      "valid_targets_mean": 10724.3,
      "valid_targets_min": 1125
    },
    {
      "epoch": 2.409090909090909,
      "grad_norm": 0.2292519211769104,
      "learning_rate": 3.318833638172519e-05,
      "loss": 0.43,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.45009514689445496,
      "step": 795,
      "valid_targets_mean": 10604.0,
      "valid_targets_min": 1504
    },
    {
      "epoch": 2.4242424242424243,
      "grad_norm": 0.19641590118408203,
      "learning_rate": 3.3074359356039306e-05,
      "loss": 0.4352,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43692293763160706,
      "step": 800,
      "valid_targets_mean": 10434.1,
      "valid_targets_min": 1765
    },
    {
      "epoch": 2.4393939393939394,
      "grad_norm": 0.21825142204761505,
      "learning_rate": 3.295963596902846e-05,
      "loss": 0.4278,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42220747470855713,
      "step": 805,
      "valid_targets_mean": 9952.1,
      "valid_targets_min": 1655
    },
    {
      "epoch": 2.4545454545454546,
      "grad_norm": 0.21460793912410736,
      "learning_rate": 3.284417276977829e-05,
      "loss": 0.4272,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4363503158092499,
      "step": 810,
      "valid_targets_mean": 10284.1,
      "valid_targets_min": 1852
    },
    {
      "epoch": 2.4696969696969697,
      "grad_norm": 0.2176777571439743,
      "learning_rate": 3.2727976349607276e-05,
      "loss": 0.4246,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41815733909606934,
      "step": 815,
      "valid_targets_mean": 9857.3,
      "valid_targets_min": 1508
    },
    {
      "epoch": 2.484848484848485,
      "grad_norm": 0.21343909204006195,
      "learning_rate": 3.261105334169045e-05,
      "loss": 0.4316,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43810808658599854,
      "step": 820,
      "valid_targets_mean": 9489.1,
      "valid_targets_min": 164
    },
    {
      "epoch": 2.5,
      "grad_norm": 0.20652568340301514,
      "learning_rate": 3.249341042068077e-05,
      "loss": 0.4237,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4342770278453827,
      "step": 825,
      "valid_targets_mean": 11467.4,
      "valid_targets_min": 1363
    },
    {
      "epoch": 2.515151515151515,
      "grad_norm": 0.21475526690483093,
      "learning_rate": 3.237505430232803e-05,
      "loss": 0.4224,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4227277338504791,
      "step": 830,
      "valid_targets_mean": 10679.8,
      "valid_targets_min": 996
    },
    {
      "epoch": 2.5303030303030303,
      "grad_norm": 0.21706214547157288,
      "learning_rate": 3.225599174309554e-05,
      "loss": 0.4302,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4138771593570709,
      "step": 835,
      "valid_targets_mean": 10705.8,
      "valid_targets_min": 2042
    },
    {
      "epoch": 2.5454545454545454,
      "grad_norm": 0.19915887713432312,
      "learning_rate": 3.213622953977443e-05,
      "loss": 0.4326,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42662036418914795,
      "step": 840,
      "valid_targets_mean": 10133.8,
      "valid_targets_min": 1074
    },
    {
      "epoch": 2.5606060606060606,
      "grad_norm": 0.21421875059604645,
      "learning_rate": 3.2015774529095595e-05,
      "loss": 0.4293,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4351594150066376,
      "step": 845,
      "valid_targets_mean": 11173.3,
      "valid_targets_min": 812
    },
    {
      "epoch": 2.5757575757575757,
      "grad_norm": 0.19874396920204163,
      "learning_rate": 3.189463358733947e-05,
      "loss": 0.4306,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4339146614074707,
      "step": 850,
      "valid_targets_mean": 10824.9,
      "valid_targets_min": 1149
    },
    {
      "epoch": 2.590909090909091,
      "grad_norm": 0.21312716603279114,
      "learning_rate": 3.1772813629943455e-05,
      "loss": 0.4315,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43114426732063293,
      "step": 855,
      "valid_targets_mean": 10300.2,
      "valid_targets_min": 2006
    },
    {
      "epoch": 2.606060606060606,
      "grad_norm": 0.18753370642662048,
      "learning_rate": 3.1650321611107195e-05,
      "loss": 0.429,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4246825873851776,
      "step": 860,
      "valid_targets_mean": 10661.9,
      "valid_targets_min": 1772
    },
    {
      "epoch": 2.621212121212121,
      "grad_norm": 0.21000739932060242,
      "learning_rate": 3.152716452339549e-05,
      "loss": 0.4289,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4096165597438812,
      "step": 865,
      "valid_targets_mean": 9761.5,
      "valid_targets_min": 1884
    },
    {
      "epoch": 2.6363636363636362,
      "grad_norm": 0.243647038936615,
      "learning_rate": 3.140334939733924e-05,
      "loss": 0.4322,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43968942761421204,
      "step": 870,
      "valid_targets_mean": 10326.9,
      "valid_targets_min": 1318
    },
    {
      "epoch": 2.6515151515151514,
      "grad_norm": 0.2074131965637207,
      "learning_rate": 3.127888330103401e-05,
      "loss": 0.429,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43400755524635315,
      "step": 875,
      "valid_targets_mean": 11271.6,
      "valid_targets_min": 1957
    },
    {
      "epoch": 2.6666666666666665,
      "grad_norm": 0.21518051624298096,
      "learning_rate": 3.115377333973659e-05,
      "loss": 0.4257,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41851577162742615,
      "step": 880,
      "valid_targets_mean": 10033.5,
      "valid_targets_min": 2683
    },
    {
      "epoch": 2.6818181818181817,
      "grad_norm": 0.20207814872264862,
      "learning_rate": 3.1028026655459376e-05,
      "loss": 0.4174,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42288029193878174,
      "step": 885,
      "valid_targets_mean": 9482.9,
      "valid_targets_min": 2446
    },
    {
      "epoch": 2.6969696969696972,
      "grad_norm": 0.21402396261692047,
      "learning_rate": 3.0901650426562634e-05,
      "loss": 0.4339,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4258166551589966,
      "step": 890,
      "valid_targets_mean": 11122.2,
      "valid_targets_min": 1090
    },
    {
      "epoch": 2.712121212121212,
      "grad_norm": 0.22855792939662933,
      "learning_rate": 3.0774651867344765e-05,
      "loss": 0.4331,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4204627573490143,
      "step": 895,
      "valid_targets_mean": 9360.9,
      "valid_targets_min": 755
    },
    {
      "epoch": 2.7272727272727275,
      "grad_norm": 0.20411750674247742,
      "learning_rate": 3.064703822763043e-05,
      "loss": 0.425,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4180368483066559,
      "step": 900,
      "valid_targets_mean": 9947.7,
      "valid_targets_min": 326
    },
    {
      "epoch": 2.742424242424242,
      "grad_norm": 0.20558571815490723,
      "learning_rate": 3.05188167923567e-05,
      "loss": 0.422,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42901313304901123,
      "step": 905,
      "valid_targets_mean": 10222.1,
      "valid_targets_min": 1518
    },
    {
      "epoch": 2.757575757575758,
      "grad_norm": 0.1814568042755127,
      "learning_rate": 3.0389994881157207e-05,
      "loss": 0.4282,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4465599060058594,
      "step": 910,
      "valid_targets_mean": 11265.0,
      "valid_targets_min": 282
    },
    {
      "epoch": 2.7727272727272725,
      "grad_norm": 0.22199641168117523,
      "learning_rate": 3.0260579847944255e-05,
      "loss": 0.4249,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4176905155181885,
      "step": 915,
      "valid_targets_mean": 9441.6,
      "valid_targets_min": 1364
    },
    {
      "epoch": 2.787878787878788,
      "grad_norm": 0.19174708425998688,
      "learning_rate": 3.0130579080489064e-05,
      "loss": 0.4194,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4169464409351349,
      "step": 920,
      "valid_targets_mean": 11430.9,
      "valid_targets_min": 1778
    },
    {
      "epoch": 2.8030303030303028,
      "grad_norm": 0.2250652015209198,
      "learning_rate": 3.0000000000000004e-05,
      "loss": 0.4267,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4370851516723633,
      "step": 925,
      "valid_targets_mean": 11393.1,
      "valid_targets_min": 1756
    },
    {
      "epoch": 2.8181818181818183,
      "grad_norm": 0.23404088616371155,
      "learning_rate": 2.986885006069894e-05,
      "loss": 0.4205,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.414355605840683,
      "step": 930,
      "valid_targets_mean": 10704.3,
      "valid_targets_min": 1454
    },
    {
      "epoch": 2.8333333333333335,
      "grad_norm": 0.21358336508274078,
      "learning_rate": 2.9737136749395743e-05,
      "loss": 0.4203,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4119959771633148,
      "step": 935,
      "valid_targets_mean": 10002.1,
      "valid_targets_min": 1636
    },
    {
      "epoch": 2.8484848484848486,
      "grad_norm": 0.1918504238128662,
      "learning_rate": 2.9604867585060834e-05,
      "loss": 0.4296,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42521974444389343,
      "step": 940,
      "valid_targets_mean": 10547.1,
      "valid_targets_min": 1222
    },
    {
      "epoch": 2.8636363636363638,
      "grad_norm": 0.20002751052379608,
      "learning_rate": 2.9472050118396034e-05,
      "loss": 0.4224,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4263623058795929,
      "step": 945,
      "valid_targets_mean": 9557.9,
      "valid_targets_min": 1879
    },
    {
      "epoch": 2.878787878787879,
      "grad_norm": 0.19519871473312378,
      "learning_rate": 2.933869193140346e-05,
      "loss": 0.4257,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42450690269470215,
      "step": 950,
      "valid_targets_mean": 10366.5,
      "valid_targets_min": 1483
    },
    {
      "epoch": 2.893939393939394,
      "grad_norm": 0.2766372263431549,
      "learning_rate": 2.920480063695272e-05,
      "loss": 0.4232,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41100582480430603,
      "step": 955,
      "valid_targets_mean": 9689.8,
      "valid_targets_min": 322
    },
    {
      "epoch": 2.909090909090909,
      "grad_norm": 0.21853753924369812,
      "learning_rate": 2.9070383878346358e-05,
      "loss": 0.4237,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41399192810058594,
      "step": 960,
      "valid_targets_mean": 9145.9,
      "valid_targets_min": 1569
    },
    {
      "epoch": 2.9242424242424243,
      "grad_norm": 0.19874586164951324,
      "learning_rate": 2.8935449328883478e-05,
      "loss": 0.4259,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4265817105770111,
      "step": 965,
      "valid_targets_mean": 10386.2,
      "valid_targets_min": 1218
    },
    {
      "epoch": 2.9393939393939394,
      "grad_norm": 0.22772009670734406,
      "learning_rate": 2.8800004691421743e-05,
      "loss": 0.4318,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4205353558063507,
      "step": 970,
      "valid_targets_mean": 9678.5,
      "valid_targets_min": 1660
    },
    {
      "epoch": 2.9545454545454546,
      "grad_norm": 0.21222379803657532,
      "learning_rate": 2.866405769793764e-05,
      "loss": 0.4142,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4031359851360321,
      "step": 975,
      "valid_targets_mean": 10232.6,
      "valid_targets_min": 1685
    },
    {
      "epoch": 2.9696969696969697,
      "grad_norm": 0.20302629470825195,
      "learning_rate": 2.8527616109085082e-05,
      "loss": 0.4261,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4246656894683838,
      "step": 980,
      "valid_targets_mean": 11043.2,
      "valid_targets_min": 1271
    },
    {
      "epoch": 2.984848484848485,
      "grad_norm": 0.20401757955551147,
      "learning_rate": 2.8390687713752405e-05,
      "loss": 0.4305,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4396190345287323,
      "step": 985,
      "valid_targets_mean": 9911.7,
      "valid_targets_min": 1649
    },
    {
      "epoch": 3.0,
      "grad_norm": 0.27179184556007385,
      "learning_rate": 2.8253280328617712e-05,
      "loss": 0.4144,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.400782972574234,
      "step": 990,
      "valid_targets_mean": 10388.2,
      "valid_targets_min": 1503
    },
    {
      "epoch": 3.015151515151515,
      "grad_norm": 0.2729860246181488,
      "learning_rate": 2.811540179770268e-05,
      "loss": 0.4247,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42994508147239685,
      "step": 995,
      "valid_targets_mean": 10646.0,
      "valid_targets_min": 1279
    },
    {
      "epoch": 3.0303030303030303,
      "grad_norm": 0.21952061355113983,
      "learning_rate": 2.7977059991924734e-05,
      "loss": 0.421,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4310123026371002,
      "step": 1000,
      "valid_targets_mean": 10762.6,
      "valid_targets_min": 1339
    },
    {
      "epoch": 3.0454545454545454,
      "grad_norm": 0.1868942379951477,
      "learning_rate": 2.7838262808647766e-05,
      "loss": 0.4315,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.429645299911499,
      "step": 1005,
      "valid_targets_mean": 11248.8,
      "valid_targets_min": 2333
    },
    {
      "epoch": 3.0606060606060606,
      "grad_norm": 0.18647056818008423,
      "learning_rate": 2.7699018171231288e-05,
      "loss": 0.4166,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41728949546813965,
      "step": 1010,
      "valid_targets_mean": 10100.0,
      "valid_targets_min": 1898
    },
    {
      "epoch": 3.0757575757575757,
      "grad_norm": 0.17098015546798706,
      "learning_rate": 2.7559334028578135e-05,
      "loss": 0.4212,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42317211627960205,
      "step": 1015,
      "valid_targets_mean": 11445.4,
      "valid_targets_min": 1191
    },
    {
      "epoch": 3.090909090909091,
      "grad_norm": 0.20844846963882446,
      "learning_rate": 2.741921835468068e-05,
      "loss": 0.4239,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42512431740760803,
      "step": 1020,
      "valid_targets_mean": 10569.9,
      "valid_targets_min": 909
    },
    {
      "epoch": 3.106060606060606,
      "grad_norm": 0.20188727974891663,
      "learning_rate": 2.727867914816563e-05,
      "loss": 0.4186,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41753116250038147,
      "step": 1025,
      "valid_targets_mean": 10349.4,
      "valid_targets_min": 1756
    },
    {
      "epoch": 3.121212121212121,
      "grad_norm": 0.18661518394947052,
      "learning_rate": 2.713772443183744e-05,
      "loss": 0.418,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4182397425174713,
      "step": 1030,
      "valid_targets_mean": 9910.7,
      "valid_targets_min": 1645
    },
    {
      "epoch": 3.1363636363636362,
      "grad_norm": 0.20452997088432312,
      "learning_rate": 2.69963622522203e-05,
      "loss": 0.4279,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4188250005245209,
      "step": 1035,
      "valid_targets_mean": 10223.4,
      "valid_targets_min": 1230
    },
    {
      "epoch": 3.1515151515151514,
      "grad_norm": 0.19041919708251953,
      "learning_rate": 2.6854600679098798e-05,
      "loss": 0.4212,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4224275052547455,
      "step": 1040,
      "valid_targets_mean": 9970.6,
      "valid_targets_min": 1756
    },
    {
      "epoch": 3.1666666666666665,
      "grad_norm": 0.19370634853839874,
      "learning_rate": 2.6712447805057254e-05,
      "loss": 0.4169,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40548941493034363,
      "step": 1045,
      "valid_targets_mean": 10416.4,
      "valid_targets_min": 1982
    },
    {
      "epoch": 3.1818181818181817,
      "grad_norm": 0.2554323375225067,
      "learning_rate": 2.656991174501775e-05,
      "loss": 0.4264,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4328724443912506,
      "step": 1050,
      "valid_targets_mean": 10460.4,
      "valid_targets_min": 1156
    },
    {
      "epoch": 3.196969696969697,
      "grad_norm": 0.20309683680534363,
      "learning_rate": 2.642700063577689e-05,
      "loss": 0.4186,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40723714232444763,
      "step": 1055,
      "valid_targets_mean": 10760.2,
      "valid_targets_min": 1389
    },
    {
      "epoch": 3.212121212121212,
      "grad_norm": 0.17737239599227905,
      "learning_rate": 2.6283722635541255e-05,
      "loss": 0.4156,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40439486503601074,
      "step": 1060,
      "valid_targets_mean": 11411.1,
      "valid_targets_min": 2150
    },
    {
      "epoch": 3.227272727272727,
      "grad_norm": 0.18335093557834625,
      "learning_rate": 2.6140085923461756e-05,
      "loss": 0.4228,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4052716791629791,
      "step": 1065,
      "valid_targets_mean": 10802.9,
      "valid_targets_min": 1829
    },
    {
      "epoch": 3.242424242424242,
      "grad_norm": 0.1992240846157074,
      "learning_rate": 2.5996098699166678e-05,
      "loss": 0.42,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4278109073638916,
      "step": 1070,
      "valid_targets_mean": 11340.9,
      "valid_targets_min": 2300
    },
    {
      "epoch": 3.257575757575758,
      "grad_norm": 0.20290203392505646,
      "learning_rate": 2.585176918229359e-05,
      "loss": 0.4261,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4109087884426117,
      "step": 1075,
      "valid_targets_mean": 9812.4,
      "valid_targets_min": 1565
    },
    {
      "epoch": 3.2727272727272725,
      "grad_norm": 0.20068520307540894,
      "learning_rate": 2.5707105612020163e-05,
      "loss": 0.4194,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41564837098121643,
      "step": 1080,
      "valid_targets_mean": 10380.2,
      "valid_targets_min": 1236
    },
    {
      "epoch": 3.287878787878788,
      "grad_norm": 0.22621721029281616,
      "learning_rate": 2.5562116246593765e-05,
      "loss": 0.4352,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.434804767370224,
      "step": 1085,
      "valid_targets_mean": 11949.3,
      "valid_targets_min": 2558
    },
    {
      "epoch": 3.303030303030303,
      "grad_norm": 0.19200244545936584,
      "learning_rate": 2.5416809362860107e-05,
      "loss": 0.4198,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41716840863227844,
      "step": 1090,
      "valid_targets_mean": 10078.6,
      "valid_targets_min": 245
    },
    {
      "epoch": 3.3181818181818183,
      "grad_norm": 0.212563619017601,
      "learning_rate": 2.5271193255790672e-05,
      "loss": 0.4183,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4222122132778168,
      "step": 1095,
      "valid_targets_mean": 10519.9,
      "valid_targets_min": 2548
    },
    {
      "epoch": 3.3333333333333335,
      "grad_norm": 0.20653125643730164,
      "learning_rate": 2.5125276238009254e-05,
      "loss": 0.4173,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4071441888809204,
      "step": 1100,
      "valid_targets_mean": 10728.7,
      "valid_targets_min": 2007
    },
    {
      "epoch": 3.3484848484848486,
      "grad_norm": 0.20140014588832855,
      "learning_rate": 2.4979066639317405e-05,
      "loss": 0.421,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42444849014282227,
      "step": 1105,
      "valid_targets_mean": 11452.4,
      "valid_targets_min": 1647
    },
    {
      "epoch": 3.3636363636363638,
      "grad_norm": 0.1943032592535019,
      "learning_rate": 2.4832572806218892e-05,
      "loss": 0.4122,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4192112386226654,
      "step": 1110,
      "valid_targets_mean": 10823.0,
      "valid_targets_min": 2052
    },
    {
      "epoch": 3.378787878787879,
      "grad_norm": 0.19034847617149353,
      "learning_rate": 2.4685803101443276e-05,
      "loss": 0.4273,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42712780833244324,
      "step": 1115,
      "valid_targets_mean": 9967.3,
      "valid_targets_min": 1519
    },
    {
      "epoch": 3.393939393939394,
      "grad_norm": 0.2004571259021759,
      "learning_rate": 2.4538765903468486e-05,
      "loss": 0.4195,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4275309145450592,
      "step": 1120,
      "valid_targets_mean": 10125.7,
      "valid_targets_min": 2476
    },
    {
      "epoch": 3.409090909090909,
      "grad_norm": 0.19023148715496063,
      "learning_rate": 2.4391469606042533e-05,
      "loss": 0.4121,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4267745018005371,
      "step": 1125,
      "valid_targets_mean": 11666.1,
      "valid_targets_min": 413
    },
    {
      "epoch": 3.4242424242424243,
      "grad_norm": 0.23639102280139923,
      "learning_rate": 2.4243922617704364e-05,
      "loss": 0.4199,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4062090218067169,
      "step": 1130,
      "valid_targets_mean": 11161.1,
      "valid_targets_min": 1048
    },
    {
      "epoch": 3.4393939393939394,
      "grad_norm": 0.19407431781291962,
      "learning_rate": 2.4096133361303816e-05,
      "loss": 0.418,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4056593179702759,
      "step": 1135,
      "valid_targets_mean": 10122.8,
      "valid_targets_min": 1150
    },
    {
      "epoch": 3.4545454545454546,
      "grad_norm": 0.20122820138931274,
      "learning_rate": 2.394811027352083e-05,
      "loss": 0.4208,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41746625304222107,
      "step": 1140,
      "valid_targets_mean": 10889.9,
      "valid_targets_min": 1563
    },
    {
      "epoch": 3.4696969696969697,
      "grad_norm": 0.17768630385398865,
      "learning_rate": 2.3799861804383807e-05,
      "loss": 0.4163,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42235884070396423,
      "step": 1145,
      "valid_targets_mean": 10588.1,
      "valid_targets_min": 449
    },
    {
      "epoch": 3.484848484848485,
      "grad_norm": 0.17705488204956055,
      "learning_rate": 2.365139641678724e-05,
      "loss": 0.4246,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4180980622768402,
      "step": 1150,
      "valid_targets_mean": 10257.1,
      "valid_targets_min": 1782
    },
    {
      "epoch": 3.5,
      "grad_norm": 0.1959957778453827,
      "learning_rate": 2.350272258600861e-05,
      "loss": 0.4118,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4030723571777344,
      "step": 1155,
      "valid_targets_mean": 9848.5,
      "valid_targets_min": 1210
    },
    {
      "epoch": 3.515151515151515,
      "grad_norm": 0.2458437830209732,
      "learning_rate": 2.335384879922456e-05,
      "loss": 0.417,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4224826395511627,
      "step": 1160,
      "valid_targets_mean": 10929.2,
      "valid_targets_min": 1719
    },
    {
      "epoch": 3.5303030303030303,
      "grad_norm": 0.18638093769550323,
      "learning_rate": 2.3204783555026394e-05,
      "loss": 0.4171,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40201738476753235,
      "step": 1165,
      "valid_targets_mean": 9884.4,
      "valid_targets_min": 1480
    },
    {
      "epoch": 3.5454545454545454,
      "grad_norm": 0.32506927847862244,
      "learning_rate": 2.3055535362934945e-05,
      "loss": 0.4107,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3836853504180908,
      "step": 1170,
      "valid_targets_mean": 9067.7,
      "valid_targets_min": 1453
    },
    {
      "epoch": 3.5606060606060606,
      "grad_norm": 0.2119535356760025,
      "learning_rate": 2.2906112742914756e-05,
      "loss": 0.4161,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41747698187828064,
      "step": 1175,
      "valid_targets_mean": 9674.8,
      "valid_targets_min": 787
    },
    {
      "epoch": 3.5757575757575757,
      "grad_norm": 0.18430210649967194,
      "learning_rate": 2.2756524224887776e-05,
      "loss": 0.4155,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4083894193172455,
      "step": 1180,
      "valid_targets_mean": 9974.4,
      "valid_targets_min": 1880
    },
    {
      "epoch": 3.590909090909091,
      "grad_norm": 0.21823269128799438,
      "learning_rate": 2.2606778348246366e-05,
      "loss": 0.4192,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42979535460472107,
      "step": 1185,
      "valid_targets_mean": 10762.7,
      "valid_targets_min": 2336
    },
    {
      "epoch": 3.606060606060606,
      "grad_norm": 0.2095731645822525,
      "learning_rate": 2.245688366136586e-05,
      "loss": 0.4215,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42712774872779846,
      "step": 1190,
      "valid_targets_mean": 10312.4,
      "valid_targets_min": 1558
    },
    {
      "epoch": 3.621212121212121,
      "grad_norm": 0.2071361392736435,
      "learning_rate": 2.2306848721116535e-05,
      "loss": 0.4163,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4190068244934082,
      "step": 1195,
      "valid_targets_mean": 9892.5,
      "valid_targets_min": 2093
    },
    {
      "epoch": 3.6363636363636362,
      "grad_norm": 0.18072761595249176,
      "learning_rate": 2.2156682092375175e-05,
      "loss": 0.4151,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42064690589904785,
      "step": 1200,
      "valid_targets_mean": 10726.5,
      "valid_targets_min": 417
    },
    {
      "epoch": 3.6515151515151514,
      "grad_norm": 0.18834993243217468,
      "learning_rate": 2.2006392347536102e-05,
      "loss": 0.4183,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41616353392601013,
      "step": 1205,
      "valid_targets_mean": 11537.3,
      "valid_targets_min": 1483
    },
    {
      "epoch": 3.6666666666666665,
      "grad_norm": 0.19812461733818054,
      "learning_rate": 2.1855988066021837e-05,
      "loss": 0.415,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4053727686405182,
      "step": 1210,
      "valid_targets_mean": 10439.8,
      "valid_targets_min": 1354
    },
    {
      "epoch": 3.6818181818181817,
      "grad_norm": 0.21153761446475983,
      "learning_rate": 2.1705477833793332e-05,
      "loss": 0.4229,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42289385199546814,
      "step": 1215,
      "valid_targets_mean": 10161.7,
      "valid_targets_min": 1717
    },
    {
      "epoch": 3.6969696969696972,
      "grad_norm": 0.21868152916431427,
      "learning_rate": 2.1554870242859813e-05,
      "loss": 0.4183,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42788711190223694,
      "step": 1220,
      "valid_targets_mean": 10665.2,
      "valid_targets_min": 1074
    },
    {
      "epoch": 3.712121212121212,
      "grad_norm": 0.19481517374515533,
      "learning_rate": 2.140417389078833e-05,
      "loss": 0.4204,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4304725229740143,
      "step": 1225,
      "valid_targets_mean": 11175.5,
      "valid_targets_min": 2383
    },
    {
      "epoch": 3.7272727272727275,
      "grad_norm": 0.20116017758846283,
      "learning_rate": 2.1253397380212935e-05,
      "loss": 0.4251,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42162999510765076,
      "step": 1230,
      "valid_targets_mean": 10527.8,
      "valid_targets_min": 2165
    },
    {
      "epoch": 3.742424242424242,
      "grad_norm": 0.2512374520301819,
      "learning_rate": 2.110254931834359e-05,
      "loss": 0.4156,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.428302526473999,
      "step": 1235,
      "valid_targets_mean": 10738.9,
      "valid_targets_min": 684
    },
    {
      "epoch": 3.757575757575758,
      "grad_norm": 0.23847144842147827,
      "learning_rate": 2.095163831647485e-05,
      "loss": 0.4202,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4156361520290375,
      "step": 1240,
      "valid_targets_mean": 10191.9,
      "valid_targets_min": 1629
    },
    {
      "epoch": 3.7727272727272725,
      "grad_norm": 0.20744843780994415,
      "learning_rate": 2.0800672989494225e-05,
      "loss": 0.4197,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41469821333885193,
      "step": 1245,
      "valid_targets_mean": 9533.6,
      "valid_targets_min": 1230
    },
    {
      "epoch": 3.787878787878788,
      "grad_norm": 0.1912551075220108,
      "learning_rate": 2.0649661955390447e-05,
      "loss": 0.4111,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3924819529056549,
      "step": 1250,
      "valid_targets_mean": 10400.9,
      "valid_targets_min": 1325
    },
    {
      "epoch": 3.8030303030303028,
      "grad_norm": 0.19835317134857178,
      "learning_rate": 2.0498613834761462e-05,
      "loss": 0.4157,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41742661595344543,
      "step": 1255,
      "valid_targets_mean": 10714.8,
      "valid_targets_min": 1652
    },
    {
      "epoch": 3.8181818181818183,
      "grad_norm": 0.1845531016588211,
      "learning_rate": 2.0347537250322338e-05,
      "loss": 0.4214,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4282882511615753,
      "step": 1260,
      "valid_targets_mean": 10514.7,
      "valid_targets_min": 1780
    },
    {
      "epoch": 3.8333333333333335,
      "grad_norm": 0.19716830551624298,
      "learning_rate": 2.0196440826413033e-05,
      "loss": 0.415,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4074436128139496,
      "step": 1265,
      "valid_targets_mean": 10360.2,
      "valid_targets_min": 1629
    },
    {
      "epoch": 3.8484848484848486,
      "grad_norm": 0.20778636634349823,
      "learning_rate": 2.004533318850605e-05,
      "loss": 0.4165,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42352762818336487,
      "step": 1270,
      "valid_targets_mean": 10006.3,
      "valid_targets_min": 1467
    },
    {
      "epoch": 3.8636363636363638,
      "grad_norm": 0.1979368031024933,
      "learning_rate": 1.9894222962714046e-05,
      "loss": 0.4131,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3961637318134308,
      "step": 1275,
      "valid_targets_mean": 11113.1,
      "valid_targets_min": 2815
    },
    {
      "epoch": 3.878787878787879,
      "grad_norm": 0.17837052047252655,
      "learning_rate": 1.9743118775297432e-05,
      "loss": 0.4139,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40569257736206055,
      "step": 1280,
      "valid_targets_mean": 10979.7,
      "valid_targets_min": 330
    },
    {
      "epoch": 3.893939393939394,
      "grad_norm": 0.17641782760620117,
      "learning_rate": 1.95920292521719e-05,
      "loss": 0.421,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41460153460502625,
      "step": 1285,
      "valid_targets_mean": 10661.4,
      "valid_targets_min": 1620
    },
    {
      "epoch": 3.909090909090909,
      "grad_norm": 0.1804283708333969,
      "learning_rate": 1.9440963018416002e-05,
      "loss": 0.4099,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3852957785129547,
      "step": 1290,
      "valid_targets_mean": 10380.7,
      "valid_targets_min": 1640
    },
    {
      "epoch": 3.9242424242424243,
      "grad_norm": 0.22172673046588898,
      "learning_rate": 1.9289928697778826e-05,
      "loss": 0.4111,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4054459035396576,
      "step": 1295,
      "valid_targets_mean": 10142.9,
      "valid_targets_min": 164
    },
    {
      "epoch": 3.9393939393939394,
      "grad_norm": 0.23020152747631073,
      "learning_rate": 1.9138934912187647e-05,
      "loss": 0.4164,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41885724663734436,
      "step": 1300,
      "valid_targets_mean": 9766.0,
      "valid_targets_min": 597
    },
    {
      "epoch": 3.9545454545454546,
      "grad_norm": 0.23106056451797485,
      "learning_rate": 1.898799028125578e-05,
      "loss": 0.4144,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4151775538921356,
      "step": 1305,
      "valid_targets_mean": 11293.2,
      "valid_targets_min": 1135
    },
    {
      "epoch": 3.9696969696969697,
      "grad_norm": 0.2169714868068695,
      "learning_rate": 1.8837103421790486e-05,
      "loss": 0.4135,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4112125635147095,
      "step": 1310,
      "valid_targets_mean": 11151.4,
      "valid_targets_min": 1810
    },
    {
      "epoch": 3.984848484848485,
      "grad_norm": 0.232574000954628,
      "learning_rate": 1.8686282947301108e-05,
      "loss": 0.4134,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4059921205043793,
      "step": 1315,
      "valid_targets_mean": 9344.0,
      "valid_targets_min": 1959
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.23121783137321472,
      "learning_rate": 1.8535537467507326e-05,
      "loss": 0.413,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4057408273220062,
      "step": 1320,
      "valid_targets_mean": 10952.7,
      "valid_targets_min": 1506
    },
    {
      "epoch": 4.015151515151516,
      "grad_norm": 0.20374144613742828,
      "learning_rate": 1.8384875587847698e-05,
      "loss": 0.4089,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4188028872013092,
      "step": 1325,
      "valid_targets_mean": 11590.3,
      "valid_targets_min": 1891
    },
    {
      "epoch": 4.03030303030303,
      "grad_norm": 0.19305098056793213,
      "learning_rate": 1.823430590898839e-05,
      "loss": 0.4102,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41999438405036926,
      "step": 1330,
      "valid_targets_mean": 10782.8,
      "valid_targets_min": 2067
    },
    {
      "epoch": 4.045454545454546,
      "grad_norm": 0.18887659907341003,
      "learning_rate": 1.8083837026332187e-05,
      "loss": 0.4133,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.407478004693985,
      "step": 1335,
      "valid_targets_mean": 11131.2,
      "valid_targets_min": 1432
    },
    {
      "epoch": 4.0606060606060606,
      "grad_norm": 0.19746972620487213,
      "learning_rate": 1.7933477529527862e-05,
      "loss": 0.4041,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39921459555625916,
      "step": 1340,
      "valid_targets_mean": 9728.8,
      "valid_targets_min": 2326
    },
    {
      "epoch": 4.075757575757576,
      "grad_norm": 0.20823736488819122,
      "learning_rate": 1.778323600197978e-05,
      "loss": 0.411,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40398335456848145,
      "step": 1345,
      "valid_targets_mean": 10329.9,
      "valid_targets_min": 1419
    },
    {
      "epoch": 4.090909090909091,
      "grad_norm": 0.1867724061012268,
      "learning_rate": 1.7633121020357945e-05,
      "loss": 0.41,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3993622362613678,
      "step": 1350,
      "valid_targets_mean": 10712.6,
      "valid_targets_min": 2154
    },
    {
      "epoch": 4.106060606060606,
      "grad_norm": 0.19380463659763336,
      "learning_rate": 1.748314115410835e-05,
      "loss": 0.4047,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4130658209323883,
      "step": 1355,
      "valid_targets_mean": 10586.5,
      "valid_targets_min": 1647
    },
    {
      "epoch": 4.121212121212121,
      "grad_norm": 0.21715623140335083,
      "learning_rate": 1.733330496496383e-05,
      "loss": 0.4103,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42087242007255554,
      "step": 1360,
      "valid_targets_mean": 10490.3,
      "valid_targets_min": 1484
    },
    {
      "epoch": 4.136363636363637,
      "grad_norm": 0.18377196788787842,
      "learning_rate": 1.718362100645527e-05,
      "loss": 0.4174,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39659810066223145,
      "step": 1365,
      "valid_targets_mean": 10914.2,
      "valid_targets_min": 684
    },
    {
      "epoch": 4.151515151515151,
      "grad_norm": 0.1994674801826477,
      "learning_rate": 1.7034097823423356e-05,
      "loss": 0.413,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42040011286735535,
      "step": 1370,
      "valid_targets_mean": 10165.2,
      "valid_targets_min": 1426
    },
    {
      "epoch": 4.166666666666667,
      "grad_norm": 0.19943372905254364,
      "learning_rate": 1.6884743951530737e-05,
      "loss": 0.417,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4275229871273041,
      "step": 1375,
      "valid_targets_mean": 10239.5,
      "valid_targets_min": 1042
    },
    {
      "epoch": 4.181818181818182,
      "grad_norm": 0.1934702843427658,
      "learning_rate": 1.6735567916774814e-05,
      "loss": 0.4082,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.423069566488266,
      "step": 1380,
      "valid_targets_mean": 11409.8,
      "valid_targets_min": 1843
    },
    {
      "epoch": 4.196969696969697,
      "grad_norm": 0.17279306054115295,
      "learning_rate": 1.6586578235000975e-05,
      "loss": 0.4154,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.420689195394516,
      "step": 1385,
      "valid_targets_mean": 10533.5,
      "valid_targets_min": 2090
    },
    {
      "epoch": 4.212121212121212,
      "grad_norm": 0.17090025544166565,
      "learning_rate": 1.64377834114165e-05,
      "loss": 0.4118,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4183070659637451,
      "step": 1390,
      "valid_targets_mean": 10416.0,
      "valid_targets_min": 483
    },
    {
      "epoch": 4.2272727272727275,
      "grad_norm": 0.21014520525932312,
      "learning_rate": 1.6289191940105004e-05,
      "loss": 0.4155,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4173717498779297,
      "step": 1395,
      "valid_targets_mean": 10735.2,
      "valid_targets_min": 818
    },
    {
      "epoch": 4.242424242424242,
      "grad_norm": 0.19336336851119995,
      "learning_rate": 1.614081230354158e-05,
      "loss": 0.4142,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41427120566368103,
      "step": 1400,
      "valid_targets_mean": 10855.8,
      "valid_targets_min": 1652
    },
    {
      "epoch": 4.257575757575758,
      "grad_norm": 0.18316613137722015,
      "learning_rate": 1.599265297210852e-05,
      "loss": 0.4138,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41837987303733826,
      "step": 1405,
      "valid_targets_mean": 9897.3,
      "valid_targets_min": 1265
    },
    {
      "epoch": 4.2727272727272725,
      "grad_norm": 0.23567067086696625,
      "learning_rate": 1.5844722403611814e-05,
      "loss": 0.4037,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4032416343688965,
      "step": 1410,
      "valid_targets_mean": 9735.4,
      "valid_targets_min": 1222
    },
    {
      "epoch": 4.287878787878788,
      "grad_norm": 0.18789321184158325,
      "learning_rate": 1.5697029042798334e-05,
      "loss": 0.4196,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41432538628578186,
      "step": 1415,
      "valid_targets_mean": 9544.1,
      "valid_targets_min": 1156
    },
    {
      "epoch": 4.303030303030303,
      "grad_norm": 0.19221526384353638,
      "learning_rate": 1.5549581320873715e-05,
      "loss": 0.4073,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3974875211715698,
      "step": 1420,
      "valid_targets_mean": 9594.2,
      "valid_targets_min": 1542
    },
    {
      "epoch": 4.318181818181818,
      "grad_norm": 0.19721989333629608,
      "learning_rate": 1.5402387655021106e-05,
      "loss": 0.4117,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4071464538574219,
      "step": 1425,
      "valid_targets_mean": 10403.3,
      "valid_targets_min": 1991
    },
    {
      "epoch": 4.333333333333333,
      "grad_norm": 0.17588333785533905,
      "learning_rate": 1.5255456447920622e-05,
      "loss": 0.4065,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41249537467956543,
      "step": 1430,
      "valid_targets_mean": 10742.1,
      "valid_targets_min": 1332
    },
    {
      "epoch": 4.348484848484849,
      "grad_norm": 0.18215610086917877,
      "learning_rate": 1.5108796087269704e-05,
      "loss": 0.4047,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40675318241119385,
      "step": 1435,
      "valid_targets_mean": 10571.2,
      "valid_targets_min": 1038
    },
    {
      "epoch": 4.363636363636363,
      "grad_norm": 0.19428397715091705,
      "learning_rate": 1.4962414945304269e-05,
      "loss": 0.401,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3971422612667084,
      "step": 1440,
      "valid_targets_mean": 9610.1,
      "valid_targets_min": 282
    },
    {
      "epoch": 4.378787878787879,
      "grad_norm": 0.18930122256278992,
      "learning_rate": 1.48163213783208e-05,
      "loss": 0.4118,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4251502454280853,
      "step": 1445,
      "valid_targets_mean": 10501.2,
      "valid_targets_min": 1793
    },
    {
      "epoch": 4.393939393939394,
      "grad_norm": 0.19899514317512512,
      "learning_rate": 1.4670523726199304e-05,
      "loss": 0.4086,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3994264602661133,
      "step": 1450,
      "valid_targets_mean": 9648.5,
      "valid_targets_min": 477
    },
    {
      "epoch": 4.409090909090909,
      "grad_norm": 0.2355262190103531,
      "learning_rate": 1.4525030311927228e-05,
      "loss": 0.4139,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4202699661254883,
      "step": 1455,
      "valid_targets_mean": 10048.0,
      "valid_targets_min": 2274
    },
    {
      "epoch": 4.424242424242424,
      "grad_norm": 0.22981785237789154,
      "learning_rate": 1.4379849441124345e-05,
      "loss": 0.4097,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41982337832450867,
      "step": 1460,
      "valid_targets_mean": 10465.3,
      "valid_targets_min": 2401
    },
    {
      "epoch": 4.4393939393939394,
      "grad_norm": 0.19632099568843842,
      "learning_rate": 1.4234989401568584e-05,
      "loss": 0.4056,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3999885618686676,
      "step": 1465,
      "valid_targets_mean": 9741.8,
      "valid_targets_min": 2272
    },
    {
      "epoch": 4.454545454545454,
      "grad_norm": 0.20335780084133148,
      "learning_rate": 1.4090458462722964e-05,
      "loss": 0.4032,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40087077021598816,
      "step": 1470,
      "valid_targets_mean": 10285.3,
      "valid_targets_min": 1381
    },
    {
      "epoch": 4.46969696969697,
      "grad_norm": 0.19511574506759644,
      "learning_rate": 1.3946264875263485e-05,
      "loss": 0.4085,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40873095393180847,
      "step": 1475,
      "valid_targets_mean": 9055.8,
      "valid_targets_min": 1926
    },
    {
      "epoch": 4.484848484848484,
      "grad_norm": 0.17824706435203552,
      "learning_rate": 1.380241687060815e-05,
      "loss": 0.4077,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3942215144634247,
      "step": 1480,
      "valid_targets_mean": 10574.3,
      "valid_targets_min": 1846
    },
    {
      "epoch": 4.5,
      "grad_norm": 0.180027037858963,
      "learning_rate": 1.3658922660447065e-05,
      "loss": 0.4077,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.403706818819046,
      "step": 1485,
      "valid_targets_mean": 9287.3,
      "valid_targets_min": 1659
    },
    {
      "epoch": 4.515151515151516,
      "grad_norm": 0.1935834437608719,
      "learning_rate": 1.3515790436273664e-05,
      "loss": 0.4023,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38846245408058167,
      "step": 1490,
      "valid_targets_mean": 9509.0,
      "valid_targets_min": 1359
    },
    {
      "epoch": 4.53030303030303,
      "grad_norm": 0.22525715827941895,
      "learning_rate": 1.3373028368917091e-05,
      "loss": 0.4071,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39254388213157654,
      "step": 1495,
      "valid_targets_mean": 10080.8,
      "valid_targets_min": 1546
    },
    {
      "epoch": 4.545454545454545,
      "grad_norm": 0.18843472003936768,
      "learning_rate": 1.3230644608075766e-05,
      "loss": 0.4024,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40034183859825134,
      "step": 1500,
      "valid_targets_mean": 9936.8,
      "valid_targets_min": 1694
    },
    {
      "epoch": 4.5606060606060606,
      "grad_norm": 0.1873057335615158,
      "learning_rate": 1.3088647281852138e-05,
      "loss": 0.4144,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4039515554904938,
      "step": 1505,
      "valid_targets_mean": 10970.2,
      "valid_targets_min": 366
    },
    {
      "epoch": 4.575757575757576,
      "grad_norm": 0.19386930763721466,
      "learning_rate": 1.2947044496288718e-05,
      "loss": 0.4013,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41857337951660156,
      "step": 1510,
      "valid_targets_mean": 11227.6,
      "valid_targets_min": 1384
    },
    {
      "epoch": 4.590909090909091,
      "grad_norm": 0.1858264058828354,
      "learning_rate": 1.2805844334905286e-05,
      "loss": 0.4116,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39640963077545166,
      "step": 1515,
      "valid_targets_mean": 9526.9,
      "valid_targets_min": 373
    },
    {
      "epoch": 4.606060606060606,
      "grad_norm": 0.1939195692539215,
      "learning_rate": 1.266505485823748e-05,
      "loss": 0.4067,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40895378589630127,
      "step": 1520,
      "valid_targets_mean": 11968.8,
      "valid_targets_min": 1948
    },
    {
      "epoch": 4.621212121212121,
      "grad_norm": 0.1939331591129303,
      "learning_rate": 1.2524684103376643e-05,
      "loss": 0.4045,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38713982701301575,
      "step": 1525,
      "valid_targets_mean": 9692.9,
      "valid_targets_min": 562
    },
    {
      "epoch": 4.636363636363637,
      "grad_norm": 0.19022901356220245,
      "learning_rate": 1.238474008351101e-05,
      "loss": 0.4092,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4030486047267914,
      "step": 1530,
      "valid_targets_mean": 10173.5,
      "valid_targets_min": 766
    },
    {
      "epoch": 4.651515151515151,
      "grad_norm": 0.19227585196495056,
      "learning_rate": 1.2245230787468279e-05,
      "loss": 0.4048,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3992983400821686,
      "step": 1535,
      "valid_targets_mean": 9347.5,
      "valid_targets_min": 1735
    },
    {
      "epoch": 4.666666666666667,
      "grad_norm": 0.19001130759716034,
      "learning_rate": 1.210616417925954e-05,
      "loss": 0.4109,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4029116630554199,
      "step": 1540,
      "valid_targets_mean": 10134.9,
      "valid_targets_min": 2170
    },
    {
      "epoch": 4.681818181818182,
      "grad_norm": 0.27013349533081055,
      "learning_rate": 1.1967548197624679e-05,
      "loss": 0.4084,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4037863314151764,
      "step": 1545,
      "valid_targets_mean": 10829.2,
      "valid_targets_min": 1230
    },
    {
      "epoch": 4.696969696969697,
      "grad_norm": 0.19909490644931793,
      "learning_rate": 1.182939075557915e-05,
      "loss": 0.4048,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4084106385707855,
      "step": 1550,
      "valid_targets_mean": 10914.1,
      "valid_targets_min": 1882
    },
    {
      "epoch": 4.712121212121212,
      "grad_norm": 0.20606616139411926,
      "learning_rate": 1.1691699739962275e-05,
      "loss": 0.4117,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4061392843723297,
      "step": 1555,
      "valid_targets_mean": 9403.7,
      "valid_targets_min": 1372
    },
    {
      "epoch": 4.7272727272727275,
      "grad_norm": 0.19522467255592346,
      "learning_rate": 1.1554483010987015e-05,
      "loss": 0.4051,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40971383452415466,
      "step": 1560,
      "valid_targets_mean": 10978.9,
      "valid_targets_min": 1313
    },
    {
      "epoch": 4.742424242424242,
      "grad_norm": 0.1925458461046219,
      "learning_rate": 1.1417748401791261e-05,
      "loss": 0.405,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4002872705459595,
      "step": 1565,
      "valid_targets_mean": 9904.3,
      "valid_targets_min": 1757
    },
    {
      "epoch": 4.757575757575758,
      "grad_norm": 0.19727370142936707,
      "learning_rate": 1.1281503717990675e-05,
      "loss": 0.4085,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4186248779296875,
      "step": 1570,
      "valid_targets_mean": 10988.9,
      "valid_targets_min": 1649
    },
    {
      "epoch": 4.7727272727272725,
      "grad_norm": 0.21095062792301178,
      "learning_rate": 1.1145756737233078e-05,
      "loss": 0.4157,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4183364808559418,
      "step": 1575,
      "valid_targets_mean": 11484.0,
      "valid_targets_min": 1761
    },
    {
      "epoch": 4.787878787878788,
      "grad_norm": 0.19469307363033295,
      "learning_rate": 1.1010515208754487e-05,
      "loss": 0.4193,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4074338972568512,
      "step": 1580,
      "valid_targets_mean": 11323.6,
      "valid_targets_min": 2489
    },
    {
      "epoch": 4.803030303030303,
      "grad_norm": 0.19948354363441467,
      "learning_rate": 1.087578685293674e-05,
      "loss": 0.407,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4115237295627594,
      "step": 1585,
      "valid_targets_mean": 10239.6,
      "valid_targets_min": 1587
    },
    {
      "epoch": 4.818181818181818,
      "grad_norm": 0.1830843836069107,
      "learning_rate": 1.0741579360866752e-05,
      "loss": 0.4149,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.43092045187950134,
      "step": 1590,
      "valid_targets_mean": 11877.0,
      "valid_targets_min": 662
    },
    {
      "epoch": 4.833333333333333,
      "grad_norm": 0.19914160668849945,
      "learning_rate": 1.0607900393897463e-05,
      "loss": 0.4083,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39899182319641113,
      "step": 1595,
      "valid_targets_mean": 10492.9,
      "valid_targets_min": 1610
    },
    {
      "epoch": 4.848484848484849,
      "grad_norm": 0.18953293561935425,
      "learning_rate": 1.0474757583210498e-05,
      "loss": 0.3952,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38818177580833435,
      "step": 1600,
      "valid_targets_mean": 10411.5,
      "valid_targets_min": 1083
    },
    {
      "epoch": 4.863636363636363,
      "grad_norm": 0.22493985295295715,
      "learning_rate": 1.0342158529380544e-05,
      "loss": 0.3987,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4030338227748871,
      "step": 1605,
      "valid_targets_mean": 9557.9,
      "valid_targets_min": 1879
    },
    {
      "epoch": 4.878787878787879,
      "grad_norm": 0.21450503170490265,
      "learning_rate": 1.0210110801941426e-05,
      "loss": 0.4022,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40151020884513855,
      "step": 1610,
      "valid_targets_mean": 10366.5,
      "valid_targets_min": 1483
    },
    {
      "epoch": 4.893939393939394,
      "grad_norm": 0.20370227098464966,
      "learning_rate": 1.0078621938954028e-05,
      "loss": 0.4003,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3890402317047119,
      "step": 1615,
      "valid_targets_mean": 9689.8,
      "valid_targets_min": 322
    },
    {
      "epoch": 4.909090909090909,
      "grad_norm": 0.2002885639667511,
      "learning_rate": 9.947699446575951e-06,
      "loss": 0.4006,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.390987366437912,
      "step": 1620,
      "valid_targets_mean": 9145.9,
      "valid_targets_min": 1569
    },
    {
      "epoch": 4.924242424242424,
      "grad_norm": 0.1956857591867447,
      "learning_rate": 9.817350798633068e-06,
      "loss": 0.4029,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4039744436740875,
      "step": 1625,
      "valid_targets_mean": 10386.2,
      "valid_targets_min": 1218
    },
    {
      "epoch": 4.9393939393939394,
      "grad_norm": 0.19534245133399963,
      "learning_rate": 9.687583436192785e-06,
      "loss": 0.4083,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39688289165496826,
      "step": 1630,
      "valid_targets_mean": 9678.5,
      "valid_targets_min": 1660
    },
    {
      "epoch": 4.954545454545455,
      "grad_norm": 0.2007068395614624,
      "learning_rate": 9.558404767139335e-06,
      "loss": 0.3919,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3812257945537567,
      "step": 1635,
      "valid_targets_mean": 10232.6,
      "valid_targets_min": 1685
    },
    {
      "epoch": 4.96969696969697,
      "grad_norm": 0.18256376683712006,
      "learning_rate": 9.429822165750893e-06,
      "loss": 0.4029,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4019554555416107,
      "step": 1640,
      "valid_targets_mean": 11043.2,
      "valid_targets_min": 1271
    },
    {
      "epoch": 4.984848484848484,
      "grad_norm": 0.1989598423242569,
      "learning_rate": 9.301842972278557e-06,
      "loss": 0.4074,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4155091941356659,
      "step": 1645,
      "valid_targets_mean": 9911.7,
      "valid_targets_min": 1649
    },
    {
      "epoch": 5.0,
      "grad_norm": 0.18358856439590454,
      "learning_rate": 9.174474492527359e-06,
      "loss": 0.3923,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38023051619529724,
      "step": 1650,
      "valid_targets_mean": 10388.2,
      "valid_targets_min": 1503
    },
    {
      "epoch": 5.015151515151516,
      "grad_norm": 0.20470210909843445,
      "learning_rate": 9.047723997439206e-06,
      "loss": 0.4095,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4025757610797882,
      "step": 1655,
      "valid_targets_mean": 9850.1,
      "valid_targets_min": 1074
    },
    {
      "epoch": 5.03030303030303,
      "grad_norm": 0.20453205704689026,
      "learning_rate": 8.921598722677796e-06,
      "loss": 0.4,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3911770284175873,
      "step": 1660,
      "valid_targets_mean": 9092.7,
      "valid_targets_min": 1091
    },
    {
      "epoch": 5.045454545454546,
      "grad_norm": 0.17487764358520508,
      "learning_rate": 8.796105868215592e-06,
      "loss": 0.4058,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40591394901275635,
      "step": 1665,
      "valid_targets_mean": 10583.7,
      "valid_targets_min": 2008
    },
    {
      "epoch": 5.0606060606060606,
      "grad_norm": 0.18619990348815918,
      "learning_rate": 8.671252597922768e-06,
      "loss": 0.4036,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.406268835067749,
      "step": 1670,
      "valid_targets_mean": 9190.0,
      "valid_targets_min": 1868
    },
    {
      "epoch": 5.075757575757576,
      "grad_norm": 0.2434193640947342,
      "learning_rate": 8.547046039158283e-06,
      "loss": 0.4001,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39369162917137146,
      "step": 1675,
      "valid_targets_mean": 10443.4,
      "valid_targets_min": 1296
    },
    {
      "epoch": 5.090909090909091,
      "grad_norm": 0.19456841051578522,
      "learning_rate": 8.423493282362982e-06,
      "loss": 0.3992,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.393932968378067,
      "step": 1680,
      "valid_targets_mean": 11161.7,
      "valid_targets_min": 1942
    },
    {
      "epoch": 5.106060606060606,
      "grad_norm": 0.17499184608459473,
      "learning_rate": 8.300601380654883e-06,
      "loss": 0.4092,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4162634611129761,
      "step": 1685,
      "valid_targets_mean": 11765.9,
      "valid_targets_min": 1414
    },
    {
      "epoch": 5.121212121212121,
      "grad_norm": 0.2031983733177185,
      "learning_rate": 8.178377349426471e-06,
      "loss": 0.4065,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40188470482826233,
      "step": 1690,
      "valid_targets_mean": 10635.6,
      "valid_targets_min": 1299
    },
    {
      "epoch": 5.136363636363637,
      "grad_norm": 0.19300803542137146,
      "learning_rate": 8.056828165944282e-06,
      "loss": 0.4044,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39879536628723145,
      "step": 1695,
      "valid_targets_mean": 9723.2,
      "valid_targets_min": 1488
    },
    {
      "epoch": 5.151515151515151,
      "grad_norm": 0.18970294296741486,
      "learning_rate": 7.93596076895055e-06,
      "loss": 0.3973,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3964475691318512,
      "step": 1700,
      "valid_targets_mean": 10449.8,
      "valid_targets_min": 1410
    },
    {
      "epoch": 5.166666666666667,
      "grad_norm": 0.1875295341014862,
      "learning_rate": 7.815782058267156e-06,
      "loss": 0.3891,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4009249210357666,
      "step": 1705,
      "valid_targets_mean": 11113.9,
      "valid_targets_min": 891
    },
    {
      "epoch": 5.181818181818182,
      "grad_norm": 0.19057181477546692,
      "learning_rate": 7.696298894401697e-06,
      "loss": 0.4054,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4091309607028961,
      "step": 1710,
      "valid_targets_mean": 11027.4,
      "valid_targets_min": 1825
    },
    {
      "epoch": 5.196969696969697,
      "grad_norm": 0.1840340793132782,
      "learning_rate": 7.577518098155869e-06,
      "loss": 0.4079,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40669795870780945,
      "step": 1715,
      "valid_targets_mean": 11521.0,
      "valid_targets_min": 1210
    },
    {
      "epoch": 5.212121212121212,
      "grad_norm": 0.18176555633544922,
      "learning_rate": 7.459446450236087e-06,
      "loss": 0.4088,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40551725029945374,
      "step": 1720,
      "valid_targets_mean": 10480.2,
      "valid_targets_min": 1800
    },
    {
      "epoch": 5.2272727272727275,
      "grad_norm": 0.18051281571388245,
      "learning_rate": 7.3420906908664345e-06,
      "loss": 0.4064,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40220069885253906,
      "step": 1725,
      "valid_targets_mean": 11155.4,
      "valid_targets_min": 1149
    },
    {
      "epoch": 5.242424242424242,
      "grad_norm": 0.20326533913612366,
      "learning_rate": 7.225457519403838e-06,
      "loss": 0.4074,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4038243293762207,
      "step": 1730,
      "valid_targets_mean": 10394.1,
      "valid_targets_min": 1192
    },
    {
      "epoch": 5.257575757575758,
      "grad_norm": 0.1892276406288147,
      "learning_rate": 7.109553593955671e-06,
      "loss": 0.3932,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4165287911891937,
      "step": 1735,
      "valid_targets_mean": 10811.4,
      "valid_targets_min": 1256
    },
    {
      "epoch": 5.2727272727272725,
      "grad_norm": 0.1724826693534851,
      "learning_rate": 6.99438553099965e-06,
      "loss": 0.4053,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3880939483642578,
      "step": 1740,
      "valid_targets_mean": 10665.0,
      "valid_targets_min": 2826
    },
    {
      "epoch": 5.287878787878788,
      "grad_norm": 0.19334332644939423,
      "learning_rate": 6.879959905006135e-06,
      "loss": 0.3941,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38951316475868225,
      "step": 1745,
      "valid_targets_mean": 9358.8,
      "valid_targets_min": 1439
    },
    {
      "epoch": 5.303030303030303,
      "grad_norm": 0.18359558284282684,
      "learning_rate": 6.766283248062817e-06,
      "loss": 0.4061,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4040810167789459,
      "step": 1750,
      "valid_targets_mean": 11006.8,
      "valid_targets_min": 477
    },
    {
      "epoch": 5.318181818181818,
      "grad_norm": 0.18151497840881348,
      "learning_rate": 6.653362049501826e-06,
      "loss": 0.3943,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4058665335178375,
      "step": 1755,
      "valid_targets_mean": 10877.0,
      "valid_targets_min": 1859
    },
    {
      "epoch": 5.333333333333333,
      "grad_norm": 0.17168527841567993,
      "learning_rate": 6.541202755529299e-06,
      "loss": 0.4005,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40380382537841797,
      "step": 1760,
      "valid_targets_mean": 11013.4,
      "valid_targets_min": 343
    },
    {
      "epoch": 5.348484848484849,
      "grad_norm": 0.18425264954566956,
      "learning_rate": 6.429811768857359e-06,
      "loss": 0.4055,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4035809338092804,
      "step": 1765,
      "valid_targets_mean": 11370.3,
      "valid_targets_min": 1230
    },
    {
      "epoch": 5.363636363636363,
      "grad_norm": 0.19658911228179932,
      "learning_rate": 6.319195448338642e-06,
      "loss": 0.3941,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4051226079463959,
      "step": 1770,
      "valid_targets_mean": 10482.6,
      "valid_targets_min": 1519
    },
    {
      "epoch": 5.378787878787879,
      "grad_norm": 0.1981133222579956,
      "learning_rate": 6.2093601086032754e-06,
      "loss": 0.4024,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41745617985725403,
      "step": 1775,
      "valid_targets_mean": 11329.6,
      "valid_targets_min": 1753
    },
    {
      "epoch": 5.393939393939394,
      "grad_norm": 0.20829808712005615,
      "learning_rate": 6.100312019698411e-06,
      "loss": 0.4,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4014301300048828,
      "step": 1780,
      "valid_targets_mean": 10215.2,
      "valid_targets_min": 1453
    },
    {
      "epoch": 5.409090909090909,
      "grad_norm": 0.19712240993976593,
      "learning_rate": 5.9920574067303114e-06,
      "loss": 0.4039,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.395517498254776,
      "step": 1785,
      "valid_targets_mean": 9959.0,
      "valid_targets_min": 1912
    },
    {
      "epoch": 5.424242424242424,
      "grad_norm": 0.18274646997451782,
      "learning_rate": 5.8846024495089425e-06,
      "loss": 0.3933,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3774782121181488,
      "step": 1790,
      "valid_targets_mean": 10087.9,
      "valid_targets_min": 1440
    },
    {
      "epoch": 5.4393939393939394,
      "grad_norm": 0.19544139504432678,
      "learning_rate": 5.777953282195228e-06,
      "loss": 0.3993,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4005298614501953,
      "step": 1795,
      "valid_targets_mean": 9448.6,
      "valid_targets_min": 1569
    },
    {
      "epoch": 5.454545454545454,
      "grad_norm": 0.17123498022556305,
      "learning_rate": 5.672115992950855e-06,
      "loss": 0.398,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38771674036979675,
      "step": 1800,
      "valid_targets_mean": 10957.0,
      "valid_targets_min": 2314
    },
    {
      "epoch": 5.46969696969697,
      "grad_norm": 0.22083625197410583,
      "learning_rate": 5.567096623590758e-06,
      "loss": 0.4078,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4015129804611206,
      "step": 1805,
      "valid_targets_mean": 10334.7,
      "valid_targets_min": 1582
    },
    {
      "epoch": 5.484848484848484,
      "grad_norm": 0.19562523066997528,
      "learning_rate": 5.462901169238175e-06,
      "loss": 0.401,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40464141964912415,
      "step": 1810,
      "valid_targets_mean": 9977.1,
      "valid_targets_min": 1687
    },
    {
      "epoch": 5.5,
      "grad_norm": 0.18308033049106598,
      "learning_rate": 5.359535577982431e-06,
      "loss": 0.4002,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40914642810821533,
      "step": 1815,
      "valid_targets_mean": 11260.5,
      "valid_targets_min": 1942
    },
    {
      "epoch": 5.515151515151516,
      "grad_norm": 0.21989823877811432,
      "learning_rate": 5.257005750539388e-06,
      "loss": 0.4035,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4028076231479645,
      "step": 1820,
      "valid_targets_mean": 9497.4,
      "valid_targets_min": 1612
    },
    {
      "epoch": 5.53030303030303,
      "grad_norm": 0.19132602214813232,
      "learning_rate": 5.155317539914601e-06,
      "loss": 0.4038,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.42036446928977966,
      "step": 1825,
      "valid_targets_mean": 11022.1,
      "valid_targets_min": 1701
    },
    {
      "epoch": 5.545454545454545,
      "grad_norm": 0.17459383606910706,
      "learning_rate": 5.054476751069179e-06,
      "loss": 0.4009,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39751294255256653,
      "step": 1830,
      "valid_targets_mean": 10233.8,
      "valid_targets_min": 1576
    },
    {
      "epoch": 5.5606060606060606,
      "grad_norm": 0.18775145709514618,
      "learning_rate": 4.954489140588412e-06,
      "loss": 0.3976,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40354594588279724,
      "step": 1835,
      "valid_targets_mean": 9377.2,
      "valid_targets_min": 1898
    },
    {
      "epoch": 5.575757575757576,
      "grad_norm": 0.17629046738147736,
      "learning_rate": 4.85536041635315e-06,
      "loss": 0.3981,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4082765579223633,
      "step": 1840,
      "valid_targets_mean": 10685.2,
      "valid_targets_min": 1339
    },
    {
      "epoch": 5.590909090909091,
      "grad_norm": 0.17188449203968048,
      "learning_rate": 4.757096237213976e-06,
      "loss": 0.4048,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3997708857059479,
      "step": 1845,
      "valid_targets_mean": 10888.4,
      "valid_targets_min": 1795
    },
    {
      "epoch": 5.606060606060606,
      "grad_norm": 0.1903722882270813,
      "learning_rate": 4.659702212668151e-06,
      "loss": 0.4109,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4117445945739746,
      "step": 1850,
      "valid_targets_mean": 10827.6,
      "valid_targets_min": 1724
    },
    {
      "epoch": 5.621212121212121,
      "grad_norm": 0.2301056832075119,
      "learning_rate": 4.5631839025393694e-06,
      "loss": 0.3956,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4025770425796509,
      "step": 1855,
      "valid_targets_mean": 10200.8,
      "valid_targets_min": 1891
    },
    {
      "epoch": 5.636363636363637,
      "grad_norm": 0.17692887783050537,
      "learning_rate": 4.467546816660433e-06,
      "loss": 0.3902,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39730289578437805,
      "step": 1860,
      "valid_targets_mean": 10355.4,
      "valid_targets_min": 1919
    },
    {
      "epoch": 5.651515151515151,
      "grad_norm": 0.19610123336315155,
      "learning_rate": 4.372796414558665e-06,
      "loss": 0.4009,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4042603075504303,
      "step": 1865,
      "valid_targets_mean": 10362.4,
      "valid_targets_min": 978
    },
    {
      "epoch": 5.666666666666667,
      "grad_norm": 0.1960899829864502,
      "learning_rate": 4.278938105144255e-06,
      "loss": 0.3978,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39503979682922363,
      "step": 1870,
      "valid_targets_mean": 10548.0,
      "valid_targets_min": 1372
    },
    {
      "epoch": 5.681818181818182,
      "grad_norm": 0.20076261460781097,
      "learning_rate": 4.185977246401509e-06,
      "loss": 0.4034,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38388100266456604,
      "step": 1875,
      "valid_targets_mean": 10947.4,
      "valid_targets_min": 352
    },
    {
      "epoch": 5.696969696969697,
      "grad_norm": 0.17560230195522308,
      "learning_rate": 4.093919145082959e-06,
      "loss": 0.3991,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40157899260520935,
      "step": 1880,
      "valid_targets_mean": 11695.2,
      "valid_targets_min": 562
    },
    {
      "epoch": 5.712121212121212,
      "grad_norm": 0.192967027425766,
      "learning_rate": 4.002769056406453e-06,
      "loss": 0.3968,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4012157618999481,
      "step": 1885,
      "valid_targets_mean": 11320.9,
      "valid_targets_min": 421
    },
    {
      "epoch": 5.7272727272727275,
      "grad_norm": 0.2020249217748642,
      "learning_rate": 3.912532183755115e-06,
      "loss": 0.4029,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4169815480709076,
      "step": 1890,
      "valid_targets_mean": 10135.3,
      "valid_targets_min": 2056
    },
    {
      "epoch": 5.742424242424242,
      "grad_norm": 0.1873956173658371,
      "learning_rate": 3.82321367838034e-06,
      "loss": 0.3993,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40991440415382385,
      "step": 1895,
      "valid_targets_mean": 10930.3,
      "valid_targets_min": 1426
    },
    {
      "epoch": 5.757575757575758,
      "grad_norm": 0.2235776036977768,
      "learning_rate": 3.734818639107709e-06,
      "loss": 0.3942,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.37802746891975403,
      "step": 1900,
      "valid_targets_mean": 9496.6,
      "valid_targets_min": 413
    },
    {
      "epoch": 5.7727272727272725,
      "grad_norm": 0.21321405470371246,
      "learning_rate": 3.647352112045943e-06,
      "loss": 0.402,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3882791996002197,
      "step": 1905,
      "valid_targets_mean": 9789.3,
      "valid_targets_min": 1793
    },
    {
      "epoch": 5.787878787878788,
      "grad_norm": 0.18891166150569916,
      "learning_rate": 3.560819090298808e-06,
      "loss": 0.408,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.390031099319458,
      "step": 1910,
      "valid_targets_mean": 10615.1,
      "valid_targets_min": 1962
    },
    {
      "epoch": 5.803030303030303,
      "grad_norm": 0.19478043913841248,
      "learning_rate": 3.4752245136801065e-06,
      "loss": 0.3988,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40345656871795654,
      "step": 1915,
      "valid_targets_mean": 10180.5,
      "valid_targets_min": 349
    },
    {
      "epoch": 5.818181818181818,
      "grad_norm": 0.19157883524894714,
      "learning_rate": 3.3905732684316626e-06,
      "loss": 0.3985,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38469433784484863,
      "step": 1920,
      "valid_targets_mean": 9206.4,
      "valid_targets_min": 1649
    },
    {
      "epoch": 5.833333333333333,
      "grad_norm": 0.1840132176876068,
      "learning_rate": 3.3068701869444177e-06,
      "loss": 0.4044,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3924840986728668,
      "step": 1925,
      "valid_targets_mean": 10671.4,
      "valid_targets_min": 1161
    },
    {
      "epoch": 5.848484848484849,
      "grad_norm": 0.17908671498298645,
      "learning_rate": 3.2241200474825307e-06,
      "loss": 0.3987,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4008850157260895,
      "step": 1930,
      "valid_targets_mean": 11423.3,
      "valid_targets_min": 1191
    },
    {
      "epoch": 5.863636363636363,
      "grad_norm": 0.21520231664180756,
      "learning_rate": 3.1423275739106353e-06,
      "loss": 0.3967,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39275026321411133,
      "step": 1935,
      "valid_targets_mean": 10805.6,
      "valid_targets_min": 321
    },
    {
      "epoch": 5.878787878787879,
      "grad_norm": 0.18723836541175842,
      "learning_rate": 3.0614974354241547e-06,
      "loss": 0.398,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40686407685279846,
      "step": 1940,
      "valid_targets_mean": 9994.7,
      "valid_targets_min": 1024
    },
    {
      "epoch": 5.893939393939394,
      "grad_norm": 0.1681266576051712,
      "learning_rate": 2.9816342462827806e-06,
      "loss": 0.402,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40341463685035706,
      "step": 1945,
      "valid_targets_mean": 10825.6,
      "valid_targets_min": 1326
    },
    {
      "epoch": 5.909090909090909,
      "grad_norm": 0.182035893201828,
      "learning_rate": 2.9027425655470366e-06,
      "loss": 0.3977,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3859279155731201,
      "step": 1950,
      "valid_targets_mean": 10662.4,
      "valid_targets_min": 810
    },
    {
      "epoch": 5.924242424242424,
      "grad_norm": 0.19270572066307068,
      "learning_rate": 2.824826896818036e-06,
      "loss": 0.403,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41110482811927795,
      "step": 1955,
      "valid_targets_mean": 10985.0,
      "valid_targets_min": 684
    },
    {
      "epoch": 5.9393939393939394,
      "grad_norm": 0.1864854097366333,
      "learning_rate": 2.747891687980384e-06,
      "loss": 0.4087,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41622889041900635,
      "step": 1960,
      "valid_targets_mean": 11076.4,
      "valid_targets_min": 888
    },
    {
      "epoch": 5.954545454545455,
      "grad_norm": 0.17258039116859436,
      "learning_rate": 2.6719413309482843e-06,
      "loss": 0.4139,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41283082962036133,
      "step": 1965,
      "valid_targets_mean": 10695.4,
      "valid_targets_min": 815
    },
    {
      "epoch": 5.96969696969697,
      "grad_norm": 0.17079488933086395,
      "learning_rate": 2.5969801614147838e-06,
      "loss": 0.4002,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39123931527137756,
      "step": 1970,
      "valid_targets_mean": 10217.5,
      "valid_targets_min": 1696
    },
    {
      "epoch": 5.984848484848484,
      "grad_norm": 0.17750884592533112,
      "learning_rate": 2.5230124586043016e-06,
      "loss": 0.3945,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3990228474140167,
      "step": 1975,
      "valid_targets_mean": 10042.9,
      "valid_targets_min": 1354
    },
    {
      "epoch": 6.0,
      "grad_norm": 0.17098312079906464,
      "learning_rate": 2.45004244502834e-06,
      "loss": 0.4014,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41152897477149963,
      "step": 1980,
      "valid_targets_mean": 10263.9,
      "valid_targets_min": 777
    },
    {
      "epoch": 6.015151515151516,
      "grad_norm": 0.181598961353302,
      "learning_rate": 2.3780742862444205e-06,
      "loss": 0.3894,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3900626599788666,
      "step": 1985,
      "valid_targets_mean": 11375.3,
      "valid_targets_min": 2090
    },
    {
      "epoch": 6.03030303030303,
      "grad_norm": 0.25970128178596497,
      "learning_rate": 2.3071120906183064e-06,
      "loss": 0.3937,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39559587836265564,
      "step": 1990,
      "valid_targets_mean": 10437.6,
      "valid_targets_min": 1508
    },
    {
      "epoch": 6.045454545454546,
      "grad_norm": 0.1895422488451004,
      "learning_rate": 2.237159909089468e-06,
      "loss": 0.4006,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39304879307746887,
      "step": 1995,
      "valid_targets_mean": 10778.6,
      "valid_targets_min": 2157
    },
    {
      "epoch": 6.0606060606060606,
      "grad_norm": 0.2517562508583069,
      "learning_rate": 2.168221734939824e-06,
      "loss": 0.3997,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.391062468290329,
      "step": 2000,
      "valid_targets_mean": 10901.2,
      "valid_targets_min": 2141
    },
    {
      "epoch": 6.075757575757576,
      "grad_norm": 0.1763228178024292,
      "learning_rate": 2.1003015035658024e-06,
      "loss": 0.4014,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4006135165691376,
      "step": 2005,
      "valid_targets_mean": 9812.5,
      "valid_targets_min": 1451
    },
    {
      "epoch": 6.090909090909091,
      "grad_norm": 0.20151282846927643,
      "learning_rate": 2.0334030922536606e-06,
      "loss": 0.4001,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38767656683921814,
      "step": 2010,
      "valid_targets_mean": 10602.4,
      "valid_targets_min": 1879
    },
    {
      "epoch": 6.106060606060606,
      "grad_norm": 0.18771740794181824,
      "learning_rate": 1.9675303199581554e-06,
      "loss": 0.3916,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3691408634185791,
      "step": 2015,
      "valid_targets_mean": 9828.1,
      "valid_targets_min": 1867
    },
    {
      "epoch": 6.121212121212121,
      "grad_norm": 0.30418938398361206,
      "learning_rate": 1.90268694708454e-06,
      "loss": 0.387,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3917301595211029,
      "step": 2020,
      "valid_targets_mean": 10257.9,
      "valid_targets_min": 638
    },
    {
      "epoch": 6.136363636363637,
      "grad_norm": 0.2077164500951767,
      "learning_rate": 1.8388766752739017e-06,
      "loss": 0.3952,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.393963485956192,
      "step": 2025,
      "valid_targets_mean": 10017.9,
      "valid_targets_min": 1550
    },
    {
      "epoch": 6.151515151515151,
      "grad_norm": 0.19020657241344452,
      "learning_rate": 1.77610314719183e-06,
      "loss": 0.3891,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.389608234167099,
      "step": 2030,
      "valid_targets_mean": 11069.8,
      "valid_targets_min": 1285
    },
    {
      "epoch": 6.166666666666667,
      "grad_norm": 0.2073080688714981,
      "learning_rate": 1.7143699463204932e-06,
      "loss": 0.3951,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3853774070739746,
      "step": 2035,
      "valid_targets_mean": 9854.9,
      "valid_targets_min": 1069
    },
    {
      "epoch": 6.181818181818182,
      "grad_norm": 0.1797332465648651,
      "learning_rate": 1.6536805967540614e-06,
      "loss": 0.3936,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3885289132595062,
      "step": 2040,
      "valid_targets_mean": 9915.2,
      "valid_targets_min": 1898
    },
    {
      "epoch": 6.196969696969697,
      "grad_norm": 0.17500117421150208,
      "learning_rate": 1.5940385629975353e-06,
      "loss": 0.3893,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4007008969783783,
      "step": 2045,
      "valid_targets_mean": 10949.3,
      "valid_targets_min": 1652
    },
    {
      "epoch": 6.212121212121212,
      "grad_norm": 0.20276956260204315,
      "learning_rate": 1.535447249768971e-06,
      "loss": 0.4022,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3994843661785126,
      "step": 2050,
      "valid_targets_mean": 10397.8,
      "valid_targets_min": 1895
    },
    {
      "epoch": 6.2272727272727275,
      "grad_norm": 0.17442868649959564,
      "learning_rate": 1.4779100018051118e-06,
      "loss": 0.3987,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3956926763057709,
      "step": 2055,
      "valid_targets_mean": 11166.6,
      "valid_targets_min": 1191
    },
    {
      "epoch": 6.242424242424242,
      "grad_norm": 0.17206421494483948,
      "learning_rate": 1.421430103670456e-06,
      "loss": 0.3996,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4041639268398285,
      "step": 2060,
      "valid_targets_mean": 11530.3,
      "valid_targets_min": 1464
    },
    {
      "epoch": 6.257575757575758,
      "grad_norm": 0.18630486726760864,
      "learning_rate": 1.366010779569764e-06,
      "loss": 0.3911,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3783796727657318,
      "step": 2065,
      "valid_targets_mean": 10216.2,
      "valid_targets_min": 449
    },
    {
      "epoch": 6.2727272727272725,
      "grad_norm": 0.21275196969509125,
      "learning_rate": 1.3116551931639899e-06,
      "loss": 0.3921,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.37745991349220276,
      "step": 2070,
      "valid_targets_mean": 9373.4,
      "valid_targets_min": 1886
    },
    {
      "epoch": 6.287878787878788,
      "grad_norm": 0.19589528441429138,
      "learning_rate": 1.258366447389674e-06,
      "loss": 0.3937,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.379587858915329,
      "step": 2075,
      "valid_targets_mean": 9955.4,
      "valid_targets_min": 1440
    },
    {
      "epoch": 6.303030303030303,
      "grad_norm": 0.17381101846694946,
      "learning_rate": 1.2061475842818337e-06,
      "loss": 0.394,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39328789710998535,
      "step": 2080,
      "valid_targets_mean": 10528.4,
      "valid_targets_min": 812
    },
    {
      "epoch": 6.318181818181818,
      "grad_norm": 0.18443268537521362,
      "learning_rate": 1.1550015848002816e-06,
      "loss": 0.3976,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.410473108291626,
      "step": 2085,
      "valid_targets_mean": 10358.1,
      "valid_targets_min": 1467
    },
    {
      "epoch": 6.333333333333333,
      "grad_norm": 0.1822654902935028,
      "learning_rate": 1.1049313686594675e-06,
      "loss": 0.3853,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.37304720282554626,
      "step": 2090,
      "valid_targets_mean": 10051.9,
      "valid_targets_min": 1394
    },
    {
      "epoch": 6.348484848484849,
      "grad_norm": 0.1884860098361969,
      "learning_rate": 1.0559397941618022e-06,
      "loss": 0.3944,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.385576993227005,
      "step": 2095,
      "valid_targets_mean": 10737.3,
      "valid_targets_min": 1255
    },
    {
      "epoch": 6.363636363636363,
      "grad_norm": 0.17882195115089417,
      "learning_rate": 1.008029658034484e-06,
      "loss": 0.3953,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3954743444919586,
      "step": 2100,
      "valid_targets_mean": 10676.4,
      "valid_targets_min": 766
    },
    {
      "epoch": 6.378787878787879,
      "grad_norm": 0.2298562228679657,
      "learning_rate": 9.61203695269859e-07,
      "loss": 0.3919,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3838197886943817,
      "step": 2105,
      "valid_targets_mean": 9619.6,
      "valid_targets_min": 374
    },
    {
      "epoch": 6.393939393939394,
      "grad_norm": 0.2124887853860855,
      "learning_rate": 9.154645789692718e-07,
      "loss": 0.3906,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39249828457832336,
      "step": 2110,
      "valid_targets_mean": 9937.5,
      "valid_targets_min": 1381
    },
    {
      "epoch": 6.409090909090909,
      "grad_norm": 0.1902085840702057,
      "learning_rate": 8.708149201904814e-07,
      "loss": 0.3889,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38680437207221985,
      "step": 2115,
      "valid_targets_mean": 11011.9,
      "valid_targets_min": 2391
    },
    {
      "epoch": 6.424242424242424,
      "grad_norm": 0.18241585791110992,
      "learning_rate": 8.272572677986001e-07,
      "loss": 0.3993,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4119403660297394,
      "step": 2120,
      "valid_targets_mean": 10275.7,
      "valid_targets_min": 1144
    },
    {
      "epoch": 6.4393939393939394,
      "grad_norm": 0.18207530677318573,
      "learning_rate": 7.847941083206057e-07,
      "loss": 0.3972,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39500510692596436,
      "step": 2125,
      "valid_targets_mean": 11102.7,
      "valid_targets_min": 3165
    },
    {
      "epoch": 6.454545454545454,
      "grad_norm": 0.19066762924194336,
      "learning_rate": 7.4342786580337e-07,
      "loss": 0.3896,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.37024879455566406,
      "step": 2130,
      "valid_targets_mean": 9863.9,
      "valid_targets_min": 1078
    },
    {
      "epoch": 6.46969696969697,
      "grad_norm": 0.15881162881851196,
      "learning_rate": 7.031609016753016e-07,
      "loss": 0.3975,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39601197838783264,
      "step": 2135,
      "valid_targets_mean": 10874.5,
      "valid_targets_min": 1870
    },
    {
      "epoch": 6.484848484848484,
      "grad_norm": 0.1725914478302002,
      "learning_rate": 6.639955146115284e-07,
      "loss": 0.4012,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.394885390996933,
      "step": 2140,
      "valid_targets_mean": 11377.2,
      "valid_targets_min": 1135
    },
    {
      "epoch": 6.5,
      "grad_norm": 0.1764625459909439,
      "learning_rate": 6.259339404026876e-07,
      "loss": 0.3939,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3892444372177124,
      "step": 2145,
      "valid_targets_mean": 9995.7,
      "valid_targets_min": 1569
    },
    {
      "epoch": 6.515151515151516,
      "grad_norm": 0.18832962214946747,
      "learning_rate": 5.889783518272785e-07,
      "loss": 0.3933,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39502444863319397,
      "step": 2150,
      "valid_targets_mean": 11390.2,
      "valid_targets_min": 1772
    },
    {
      "epoch": 6.53030303030303,
      "grad_norm": 0.18924881517887115,
      "learning_rate": 5.53130858527644e-07,
      "loss": 0.3896,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3994877338409424,
      "step": 2155,
      "valid_targets_mean": 11837.1,
      "valid_targets_min": 334
    },
    {
      "epoch": 6.545454545454545,
      "grad_norm": 0.1887369155883789,
      "learning_rate": 5.183935068895207e-07,
      "loss": 0.392,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40097928047180176,
      "step": 2160,
      "valid_targets_mean": 10294.0,
      "valid_targets_min": 1022
    },
    {
      "epoch": 6.5606060606060606,
      "grad_norm": 0.17452627420425415,
      "learning_rate": 4.847682799252474e-07,
      "loss": 0.3959,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40073612332344055,
      "step": 2165,
      "valid_targets_mean": 9978.1,
      "valid_targets_min": 1841
    },
    {
      "epoch": 6.575757575757576,
      "grad_norm": 0.19132117927074432,
      "learning_rate": 4.522570971605289e-07,
      "loss": 0.3921,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3851205110549927,
      "step": 2170,
      "valid_targets_mean": 9710.3,
      "valid_targets_min": 1419
    },
    {
      "epoch": 6.590909090909091,
      "grad_norm": 0.19148464500904083,
      "learning_rate": 4.208618145248866e-07,
      "loss": 0.3923,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40141549706459045,
      "step": 2175,
      "valid_targets_mean": 10751.1,
      "valid_targets_min": 373
    },
    {
      "epoch": 6.606060606060606,
      "grad_norm": 0.1952446550130844,
      "learning_rate": 3.9058422424568923e-07,
      "loss": 0.3943,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.40332460403442383,
      "step": 2180,
      "valid_targets_mean": 10758.5,
      "valid_targets_min": 2318
    },
    {
      "epoch": 6.621212121212121,
      "grad_norm": 0.19235672056674957,
      "learning_rate": 3.614260547458659e-07,
      "loss": 0.3958,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.392660528421402,
      "step": 2185,
      "valid_targets_mean": 10613.4,
      "valid_targets_min": 2396
    },
    {
      "epoch": 6.636363636363637,
      "grad_norm": 0.18683838844299316,
      "learning_rate": 3.3338897054521205e-07,
      "loss": 0.3936,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3899310529232025,
      "step": 2190,
      "valid_targets_mean": 10502.3,
      "valid_targets_min": 483
    },
    {
      "epoch": 6.651515151515151,
      "grad_norm": 0.18461862206459045,
      "learning_rate": 3.0647457216538724e-07,
      "loss": 0.3994,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4082428216934204,
      "step": 2195,
      "valid_targets_mean": 11777.3,
      "valid_targets_min": 1735
    },
    {
      "epoch": 6.666666666666667,
      "grad_norm": 0.18447022140026093,
      "learning_rate": 2.8068439603853747e-07,
      "loss": 0.3946,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38763725757598877,
      "step": 2200,
      "valid_targets_mean": 9958.8,
      "valid_targets_min": 1310
    },
    {
      "epoch": 6.681818181818182,
      "grad_norm": 0.2138996422290802,
      "learning_rate": 2.5601991441959407e-07,
      "loss": 0.3929,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3867276608943939,
      "step": 2205,
      "valid_targets_mean": 9999.3,
      "valid_targets_min": 1318
    },
    {
      "epoch": 6.696969696969697,
      "grad_norm": 0.18314942717552185,
      "learning_rate": 2.3248253530222753e-07,
      "loss": 0.3975,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3950660526752472,
      "step": 2210,
      "valid_targets_mean": 11171.1,
      "valid_targets_min": 424
    },
    {
      "epoch": 6.712121212121212,
      "grad_norm": 0.2026413083076477,
      "learning_rate": 2.1007360233846308e-07,
      "loss": 0.394,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38612592220306396,
      "step": 2215,
      "valid_targets_mean": 10497.8,
      "valid_targets_min": 1138
    },
    {
      "epoch": 6.7272727272727275,
      "grad_norm": 0.20692871510982513,
      "learning_rate": 1.8879439476198636e-07,
      "loss": 0.3953,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4036029279232025,
      "step": 2220,
      "valid_targets_mean": 11729.8,
      "valid_targets_min": 2225
    },
    {
      "epoch": 6.742424242424242,
      "grad_norm": 0.16621901094913483,
      "learning_rate": 1.6864612731511298e-07,
      "loss": 0.3981,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4010261595249176,
      "step": 2225,
      "valid_targets_mean": 11597.1,
      "valid_targets_min": 1811
    },
    {
      "epoch": 6.757575757575758,
      "grad_norm": 0.17813605070114136,
      "learning_rate": 1.4962995017944626e-07,
      "loss": 0.3949,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39973902702331543,
      "step": 2230,
      "valid_targets_mean": 10912.5,
      "valid_targets_min": 2173
    },
    {
      "epoch": 6.7727272727272725,
      "grad_norm": 0.18354913592338562,
      "learning_rate": 1.3174694891021188e-07,
      "loss": 0.3925,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39457282423973083,
      "step": 2235,
      "valid_targets_mean": 10132.8,
      "valid_targets_min": 1024
    },
    {
      "epoch": 6.787878787878788,
      "grad_norm": 0.19414886832237244,
      "learning_rate": 1.1499814437429869e-07,
      "loss": 0.3947,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3952203691005707,
      "step": 2240,
      "valid_targets_mean": 10012.4,
      "valid_targets_min": 2021
    },
    {
      "epoch": 6.803030303030303,
      "grad_norm": 0.17182624340057373,
      "learning_rate": 9.938449269197181e-08,
      "loss": 0.389,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.37530478835105896,
      "step": 2245,
      "valid_targets_mean": 10932.0,
      "valid_targets_min": 662
    },
    {
      "epoch": 6.818181818181818,
      "grad_norm": 0.1824752390384674,
      "learning_rate": 8.490688518229651e-08,
      "loss": 0.3971,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3855361044406891,
      "step": 2250,
      "valid_targets_mean": 10957.1,
      "valid_targets_min": 1573
    },
    {
      "epoch": 6.833333333333333,
      "grad_norm": 0.17674382030963898,
      "learning_rate": 7.156614831225428e-08,
      "loss": 0.3919,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.41078484058380127,
      "step": 2255,
      "valid_targets_mean": 11247.8,
      "valid_targets_min": 1218
    },
    {
      "epoch": 6.848484848484849,
      "grad_norm": 0.18867872655391693,
      "learning_rate": 5.936304364956513e-08,
      "loss": 0.392,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.39554348587989807,
      "step": 2260,
      "valid_targets_mean": 10252.4,
      "valid_targets_min": 1888
    },
    {
      "epoch": 6.863636363636363,
      "grad_norm": 0.17342455685138702,
      "learning_rate": 4.829826781921343e-08,
      "loss": 0.3951,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.394491046667099,
      "step": 2265,
      "valid_targets_mean": 10781.4,
      "valid_targets_min": 2212
    },
    {
      "epoch": 6.878787878787879,
      "grad_norm": 0.21512271463871002,
      "learning_rate": 3.837245246367749e-08,
      "loss": 0.3932,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3912753164768219,
      "step": 2270,
      "valid_targets_mean": 10922.4,
      "valid_targets_min": 1110
    },
    {
      "epoch": 6.893939393939394,
      "grad_norm": 0.18096481263637543,
      "learning_rate": 2.958616420687177e-08,
      "loss": 0.3915,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38797757029533386,
      "step": 2275,
      "valid_targets_mean": 10149.2,
      "valid_targets_min": 1286
    },
    {
      "epoch": 6.909090909090909,
      "grad_norm": 0.1930212825536728,
      "learning_rate": 2.1939904621806062e-08,
      "loss": 0.3925,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3880687654018402,
      "step": 2280,
      "valid_targets_mean": 9949.4,
      "valid_targets_min": 2071
    },
    {
      "epoch": 6.924242424242424,
      "grad_norm": 0.19907346367835999,
      "learning_rate": 1.5434110201946184e-08,
      "loss": 0.3923,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.3829110860824585,
      "step": 2285,
      "valid_targets_mean": 10804.1,
      "valid_targets_min": 2758
    },
    {
      "epoch": 6.9393939393939394,
      "grad_norm": 0.23474514484405518,
      "learning_rate": 1.006915233629835e-08,
      "loss": 0.3913,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.4002104699611664,
      "step": 2290,
      "valid_targets_mean": 10622.1,
      "valid_targets_min": 1165
    },
    {
      "epoch": 6.954545454545455,
      "grad_norm": 0.19267421960830688,
      "learning_rate": 5.845337288210573e-09,
      "loss": 0.399,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.384625107049942,
      "step": 2295,
      "valid_targets_mean": 9933.8,
      "valid_targets_min": 1995
    },
    {
      "epoch": 6.96969696969697,
      "grad_norm": 0.18153943121433258,
      "learning_rate": 2.7629061778866597e-09,
      "loss": 0.3914,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38872551918029785,
      "step": 2300,
      "valid_targets_mean": 10496.8,
      "valid_targets_min": 2103
    },
    {
      "epoch": 6.984848484848484,
      "grad_norm": 0.21790501475334167,
      "learning_rate": 8.220349686216545e-10,
      "loss": 0.393,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38566938042640686,
      "step": 2305,
      "valid_targets_mean": 10239.4,
      "valid_targets_min": 562
    },
    {
      "epoch": 7.0,
      "grad_norm": 0.20717570185661316,
      "learning_rate": 2.2834456763209944e-11,
      "loss": 0.3952,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38292744755744934,
      "step": 2310,
      "valid_targets_mean": 9547.0,
      "valid_targets_min": 417
    },
    {
      "epoch": 7.0,
      "loss_nan_ranks": 0,
      "loss_rank_avg": 0.38292744755744934,
      "step": 2310,
      "total_flos": 1.4779110410682368e+16,
      "train_loss": 0.12231839595935046,
      "train_runtime": 28001.9493,
      "train_samples_per_second": 7.899,
      "train_steps_per_second": 0.082,
      "valid_targets_mean": 9547.0,
      "valid_targets_min": 417
    }
  ],
  "logging_steps": 5,
  "max_steps": 2310,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 7,
  "save_steps": 400,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 1.4779110410682368e+16,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}