27977 lines
778 KiB
JSON
27977 lines
778 KiB
JSON
{
|
|
"best_global_step": null,
|
|
"best_metric": null,
|
|
"best_model_checkpoint": null,
|
|
"epoch": 7.0,
|
|
"eval_steps": 500,
|
|
"global_step": 12698,
|
|
"is_hyper_param_search": false,
|
|
"is_local_process_zero": true,
|
|
"is_world_process_zero": true,
|
|
"log_history": [
|
|
{
|
|
"epoch": 0.0027563395810363835,
|
|
"grad_norm": 12.075547478827943,
|
|
"learning_rate": 1.2598425196850394e-07,
|
|
"loss": 0.5455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.25351786613464355,
|
|
"step": 5,
|
|
"valid_targets_mean": 7359.2,
|
|
"valid_targets_min": 5595
|
|
},
|
|
{
|
|
"epoch": 0.005512679162072767,
|
|
"grad_norm": 10.891809397898243,
|
|
"learning_rate": 2.8346456692913386e-07,
|
|
"loss": 0.5804,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19977489113807678,
|
|
"step": 10,
|
|
"valid_targets_mean": 1287.1,
|
|
"valid_targets_min": 420
|
|
},
|
|
{
|
|
"epoch": 0.008269018743109152,
|
|
"grad_norm": 9.98708146065969,
|
|
"learning_rate": 4.409448818897638e-07,
|
|
"loss": 0.5586,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2705141603946686,
|
|
"step": 15,
|
|
"valid_targets_mean": 8512.8,
|
|
"valid_targets_min": 6729
|
|
},
|
|
{
|
|
"epoch": 0.011025358324145534,
|
|
"grad_norm": 9.491596927629763,
|
|
"learning_rate": 5.984251968503937e-07,
|
|
"loss": 0.5104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.23840153217315674,
|
|
"step": 20,
|
|
"valid_targets_mean": 7552.0,
|
|
"valid_targets_min": 6145
|
|
},
|
|
{
|
|
"epoch": 0.013781697905181918,
|
|
"grad_norm": 8.145281285375768,
|
|
"learning_rate": 7.559055118110237e-07,
|
|
"loss": 0.5234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2507321834564209,
|
|
"step": 25,
|
|
"valid_targets_mean": 7301.4,
|
|
"valid_targets_min": 4807
|
|
},
|
|
{
|
|
"epoch": 0.016538037486218304,
|
|
"grad_norm": 6.341176368803666,
|
|
"learning_rate": 9.133858267716536e-07,
|
|
"loss": 0.5296,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2803744971752167,
|
|
"step": 30,
|
|
"valid_targets_mean": 6809.0,
|
|
"valid_targets_min": 5015
|
|
},
|
|
{
|
|
"epoch": 0.019294377067254686,
|
|
"grad_norm": 4.973615112317376,
|
|
"learning_rate": 1.0708661417322836e-06,
|
|
"loss": 0.5221,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2603566646575928,
|
|
"step": 35,
|
|
"valid_targets_mean": 7217.0,
|
|
"valid_targets_min": 5541
|
|
},
|
|
{
|
|
"epoch": 0.022050716648291068,
|
|
"grad_norm": 4.34721740676994,
|
|
"learning_rate": 1.2283464566929135e-06,
|
|
"loss": 0.4927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.23556804656982422,
|
|
"step": 40,
|
|
"valid_targets_mean": 6205.1,
|
|
"valid_targets_min": 5580
|
|
},
|
|
{
|
|
"epoch": 0.024807056229327454,
|
|
"grad_norm": 5.573828440588775,
|
|
"learning_rate": 1.3858267716535435e-06,
|
|
"loss": 0.5001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11648112535476685,
|
|
"step": 45,
|
|
"valid_targets_mean": 598.0,
|
|
"valid_targets_min": 151
|
|
},
|
|
{
|
|
"epoch": 0.027563395810363836,
|
|
"grad_norm": 2.5469565923152278,
|
|
"learning_rate": 1.5433070866141734e-06,
|
|
"loss": 0.4536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2248832881450653,
|
|
"step": 50,
|
|
"valid_targets_mean": 7270.8,
|
|
"valid_targets_min": 6125
|
|
},
|
|
{
|
|
"epoch": 0.03031973539140022,
|
|
"grad_norm": 2.1653257440965903,
|
|
"learning_rate": 1.7007874015748034e-06,
|
|
"loss": 0.4512,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21155396103858948,
|
|
"step": 55,
|
|
"valid_targets_mean": 6108.9,
|
|
"valid_targets_min": 4623
|
|
},
|
|
{
|
|
"epoch": 0.03307607497243661,
|
|
"grad_norm": 1.701370119000033,
|
|
"learning_rate": 1.8582677165354333e-06,
|
|
"loss": 0.4351,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21578280627727509,
|
|
"step": 60,
|
|
"valid_targets_mean": 6571.6,
|
|
"valid_targets_min": 4850
|
|
},
|
|
{
|
|
"epoch": 0.03583241455347299,
|
|
"grad_norm": 1.2782427667676923,
|
|
"learning_rate": 2.015748031496063e-06,
|
|
"loss": 0.4127,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21117806434631348,
|
|
"step": 65,
|
|
"valid_targets_mean": 7984.4,
|
|
"valid_targets_min": 6312
|
|
},
|
|
{
|
|
"epoch": 0.03858875413450937,
|
|
"grad_norm": 1.0840215189567821,
|
|
"learning_rate": 2.173228346456693e-06,
|
|
"loss": 0.4009,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19541636109352112,
|
|
"step": 70,
|
|
"valid_targets_mean": 6273.1,
|
|
"valid_targets_min": 4590
|
|
},
|
|
{
|
|
"epoch": 0.041345093715545754,
|
|
"grad_norm": 0.9840003836907789,
|
|
"learning_rate": 2.330708661417323e-06,
|
|
"loss": 0.4045,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20693877339363098,
|
|
"step": 75,
|
|
"valid_targets_mean": 7860.2,
|
|
"valid_targets_min": 5453
|
|
},
|
|
{
|
|
"epoch": 0.044101433296582136,
|
|
"grad_norm": 0.8281179445285038,
|
|
"learning_rate": 2.488188976377953e-06,
|
|
"loss": 0.4002,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18255400657653809,
|
|
"step": 80,
|
|
"valid_targets_mean": 6988.6,
|
|
"valid_targets_min": 4928
|
|
},
|
|
{
|
|
"epoch": 0.046857772877618525,
|
|
"grad_norm": 0.7096605471413986,
|
|
"learning_rate": 2.645669291338583e-06,
|
|
"loss": 0.3866,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1821506917476654,
|
|
"step": 85,
|
|
"valid_targets_mean": 6495.6,
|
|
"valid_targets_min": 5367
|
|
},
|
|
{
|
|
"epoch": 0.04961411245865491,
|
|
"grad_norm": 0.811462547790769,
|
|
"learning_rate": 2.803149606299213e-06,
|
|
"loss": 0.3849,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19314873218536377,
|
|
"step": 90,
|
|
"valid_targets_mean": 4968.1,
|
|
"valid_targets_min": 615
|
|
},
|
|
{
|
|
"epoch": 0.05237045203969129,
|
|
"grad_norm": 0.6426607267592221,
|
|
"learning_rate": 2.9606299212598427e-06,
|
|
"loss": 0.3962,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1978967785835266,
|
|
"step": 95,
|
|
"valid_targets_mean": 7189.9,
|
|
"valid_targets_min": 5732
|
|
},
|
|
{
|
|
"epoch": 0.05512679162072767,
|
|
"grad_norm": 1.1509744585927528,
|
|
"learning_rate": 3.118110236220473e-06,
|
|
"loss": 0.3826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1987777054309845,
|
|
"step": 100,
|
|
"valid_targets_mean": 7969.4,
|
|
"valid_targets_min": 5870
|
|
},
|
|
{
|
|
"epoch": 0.057883131201764054,
|
|
"grad_norm": 0.5456461030409412,
|
|
"learning_rate": 3.2755905511811026e-06,
|
|
"loss": 0.346,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1557111144065857,
|
|
"step": 105,
|
|
"valid_targets_mean": 7348.0,
|
|
"valid_targets_min": 4983
|
|
},
|
|
{
|
|
"epoch": 0.06063947078280044,
|
|
"grad_norm": 0.5289619260062255,
|
|
"learning_rate": 3.433070866141733e-06,
|
|
"loss": 0.3479,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15491631627082825,
|
|
"step": 110,
|
|
"valid_targets_mean": 6763.5,
|
|
"valid_targets_min": 5251
|
|
},
|
|
{
|
|
"epoch": 0.06339581036383682,
|
|
"grad_norm": 0.5831333276320478,
|
|
"learning_rate": 3.5905511811023625e-06,
|
|
"loss": 0.3632,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20547156035900116,
|
|
"step": 115,
|
|
"valid_targets_mean": 6626.4,
|
|
"valid_targets_min": 4332
|
|
},
|
|
{
|
|
"epoch": 0.06615214994487321,
|
|
"grad_norm": 0.5079379091320437,
|
|
"learning_rate": 3.748031496062993e-06,
|
|
"loss": 0.3494,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1666697859764099,
|
|
"step": 120,
|
|
"valid_targets_mean": 7151.6,
|
|
"valid_targets_min": 4434
|
|
},
|
|
{
|
|
"epoch": 0.0689084895259096,
|
|
"grad_norm": 0.5365621643972469,
|
|
"learning_rate": 3.905511811023622e-06,
|
|
"loss": 0.343,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19373363256454468,
|
|
"step": 125,
|
|
"valid_targets_mean": 7592.0,
|
|
"valid_targets_min": 5705
|
|
},
|
|
{
|
|
"epoch": 0.07166482910694598,
|
|
"grad_norm": 0.45493375224886373,
|
|
"learning_rate": 4.062992125984252e-06,
|
|
"loss": 0.303,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1465461105108261,
|
|
"step": 130,
|
|
"valid_targets_mean": 8727.9,
|
|
"valid_targets_min": 5721
|
|
},
|
|
{
|
|
"epoch": 0.07442116868798236,
|
|
"grad_norm": 0.529452666118903,
|
|
"learning_rate": 4.220472440944882e-06,
|
|
"loss": 0.3016,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16562356054782867,
|
|
"step": 135,
|
|
"valid_targets_mean": 6314.9,
|
|
"valid_targets_min": 4517
|
|
},
|
|
{
|
|
"epoch": 0.07717750826901874,
|
|
"grad_norm": 0.5121376140067504,
|
|
"learning_rate": 4.377952755905512e-06,
|
|
"loss": 0.2895,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1628265529870987,
|
|
"step": 140,
|
|
"valid_targets_mean": 7161.6,
|
|
"valid_targets_min": 4531
|
|
},
|
|
{
|
|
"epoch": 0.07993384785005513,
|
|
"grad_norm": 1.1131050678214447,
|
|
"learning_rate": 4.535433070866142e-06,
|
|
"loss": 0.3527,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13153913617134094,
|
|
"step": 145,
|
|
"valid_targets_mean": 1098.8,
|
|
"valid_targets_min": 160
|
|
},
|
|
{
|
|
"epoch": 0.08269018743109151,
|
|
"grad_norm": 0.4627350910375568,
|
|
"learning_rate": 4.692913385826772e-06,
|
|
"loss": 0.3391,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14615601301193237,
|
|
"step": 150,
|
|
"valid_targets_mean": 6988.4,
|
|
"valid_targets_min": 5392
|
|
},
|
|
{
|
|
"epoch": 0.08544652701212789,
|
|
"grad_norm": 0.49950430514762034,
|
|
"learning_rate": 4.850393700787402e-06,
|
|
"loss": 0.3337,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17031151056289673,
|
|
"step": 155,
|
|
"valid_targets_mean": 7712.9,
|
|
"valid_targets_min": 5297
|
|
},
|
|
{
|
|
"epoch": 0.08820286659316427,
|
|
"grad_norm": 0.5280300649378892,
|
|
"learning_rate": 5.007874015748032e-06,
|
|
"loss": 0.3399,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1665465235710144,
|
|
"step": 160,
|
|
"valid_targets_mean": 6396.9,
|
|
"valid_targets_min": 5536
|
|
},
|
|
{
|
|
"epoch": 0.09095920617420065,
|
|
"grad_norm": 0.5480614214644137,
|
|
"learning_rate": 5.1653543307086615e-06,
|
|
"loss": 0.3317,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17012524604797363,
|
|
"step": 165,
|
|
"valid_targets_mean": 6143.8,
|
|
"valid_targets_min": 3434
|
|
},
|
|
{
|
|
"epoch": 0.09371554575523705,
|
|
"grad_norm": 0.4615911499195326,
|
|
"learning_rate": 5.322834645669291e-06,
|
|
"loss": 0.3151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16160203516483307,
|
|
"step": 170,
|
|
"valid_targets_mean": 7182.1,
|
|
"valid_targets_min": 5349
|
|
},
|
|
{
|
|
"epoch": 0.09647188533627343,
|
|
"grad_norm": 0.488810230106606,
|
|
"learning_rate": 5.480314960629922e-06,
|
|
"loss": 0.3191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16232289373874664,
|
|
"step": 175,
|
|
"valid_targets_mean": 7570.5,
|
|
"valid_targets_min": 5209
|
|
},
|
|
{
|
|
"epoch": 0.09922822491730982,
|
|
"grad_norm": 0.5540162367605067,
|
|
"learning_rate": 5.637795275590552e-06,
|
|
"loss": 0.3105,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17014867067337036,
|
|
"step": 180,
|
|
"valid_targets_mean": 7069.0,
|
|
"valid_targets_min": 5581
|
|
},
|
|
{
|
|
"epoch": 0.1019845644983462,
|
|
"grad_norm": 0.5221412948869323,
|
|
"learning_rate": 5.795275590551181e-06,
|
|
"loss": 0.3156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16441379487514496,
|
|
"step": 185,
|
|
"valid_targets_mean": 6925.6,
|
|
"valid_targets_min": 5504
|
|
},
|
|
{
|
|
"epoch": 0.10474090407938258,
|
|
"grad_norm": 0.5308795314743368,
|
|
"learning_rate": 5.952755905511811e-06,
|
|
"loss": 0.308,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15678948163986206,
|
|
"step": 190,
|
|
"valid_targets_mean": 6716.9,
|
|
"valid_targets_min": 5332
|
|
},
|
|
{
|
|
"epoch": 0.10749724366041896,
|
|
"grad_norm": 0.5311149585849814,
|
|
"learning_rate": 6.110236220472442e-06,
|
|
"loss": 0.3118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15208759903907776,
|
|
"step": 195,
|
|
"valid_targets_mean": 6241.8,
|
|
"valid_targets_min": 5406
|
|
},
|
|
{
|
|
"epoch": 0.11025358324145534,
|
|
"grad_norm": 0.4830411193613476,
|
|
"learning_rate": 6.2677165354330715e-06,
|
|
"loss": 0.3118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1529456526041031,
|
|
"step": 200,
|
|
"valid_targets_mean": 6387.8,
|
|
"valid_targets_min": 4847
|
|
},
|
|
{
|
|
"epoch": 0.11300992282249173,
|
|
"grad_norm": 0.4755751133549269,
|
|
"learning_rate": 6.425196850393701e-06,
|
|
"loss": 0.3082,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1523471474647522,
|
|
"step": 205,
|
|
"valid_targets_mean": 6457.6,
|
|
"valid_targets_min": 4288
|
|
},
|
|
{
|
|
"epoch": 0.11576626240352811,
|
|
"grad_norm": 0.4900378091230735,
|
|
"learning_rate": 6.5826771653543306e-06,
|
|
"loss": 0.3068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17025655508041382,
|
|
"step": 210,
|
|
"valid_targets_mean": 6793.1,
|
|
"valid_targets_min": 5173
|
|
},
|
|
{
|
|
"epoch": 0.1185226019845645,
|
|
"grad_norm": 0.511079954412969,
|
|
"learning_rate": 6.740157480314962e-06,
|
|
"loss": 0.3001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13845129311084747,
|
|
"step": 215,
|
|
"valid_targets_mean": 6697.4,
|
|
"valid_targets_min": 5168
|
|
},
|
|
{
|
|
"epoch": 0.12127894156560089,
|
|
"grad_norm": 0.4754998794775802,
|
|
"learning_rate": 6.897637795275591e-06,
|
|
"loss": 0.297,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14565712213516235,
|
|
"step": 220,
|
|
"valid_targets_mean": 6062.0,
|
|
"valid_targets_min": 5270
|
|
},
|
|
{
|
|
"epoch": 0.12403528114663727,
|
|
"grad_norm": 0.49279474277078555,
|
|
"learning_rate": 7.055118110236221e-06,
|
|
"loss": 0.2998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15451404452323914,
|
|
"step": 225,
|
|
"valid_targets_mean": 7110.1,
|
|
"valid_targets_min": 5253
|
|
},
|
|
{
|
|
"epoch": 0.12679162072767364,
|
|
"grad_norm": 0.42933580391362647,
|
|
"learning_rate": 7.21259842519685e-06,
|
|
"loss": 0.2895,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13672415912151337,
|
|
"step": 230,
|
|
"valid_targets_mean": 7395.9,
|
|
"valid_targets_min": 5248
|
|
},
|
|
{
|
|
"epoch": 0.12954796030871002,
|
|
"grad_norm": 0.4999650136228687,
|
|
"learning_rate": 7.3700787401574816e-06,
|
|
"loss": 0.2965,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17963960766792297,
|
|
"step": 235,
|
|
"valid_targets_mean": 7033.5,
|
|
"valid_targets_min": 5354
|
|
},
|
|
{
|
|
"epoch": 0.13230429988974643,
|
|
"grad_norm": 0.5201921876075898,
|
|
"learning_rate": 7.527559055118111e-06,
|
|
"loss": 0.2964,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15595601499080658,
|
|
"step": 240,
|
|
"valid_targets_mean": 6798.8,
|
|
"valid_targets_min": 4532
|
|
},
|
|
{
|
|
"epoch": 0.1350606394707828,
|
|
"grad_norm": 1.0105660237468501,
|
|
"learning_rate": 7.68503937007874e-06,
|
|
"loss": 0.2746,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09508383274078369,
|
|
"step": 245,
|
|
"valid_targets_mean": 2045.0,
|
|
"valid_targets_min": 164
|
|
},
|
|
{
|
|
"epoch": 0.1378169790518192,
|
|
"grad_norm": 0.44606735524421404,
|
|
"learning_rate": 7.84251968503937e-06,
|
|
"loss": 0.2859,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14478397369384766,
|
|
"step": 250,
|
|
"valid_targets_mean": 6723.1,
|
|
"valid_targets_min": 5164
|
|
},
|
|
{
|
|
"epoch": 0.14057331863285558,
|
|
"grad_norm": 0.5119738303973728,
|
|
"learning_rate": 8.000000000000001e-06,
|
|
"loss": 0.2858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1455782651901245,
|
|
"step": 255,
|
|
"valid_targets_mean": 5954.1,
|
|
"valid_targets_min": 4996
|
|
},
|
|
{
|
|
"epoch": 0.14332965821389196,
|
|
"grad_norm": 0.5049448033059593,
|
|
"learning_rate": 8.157480314960631e-06,
|
|
"loss": 0.2914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1333935707807541,
|
|
"step": 260,
|
|
"valid_targets_mean": 5932.0,
|
|
"valid_targets_min": 5205
|
|
},
|
|
{
|
|
"epoch": 0.14608599779492834,
|
|
"grad_norm": 0.5138161558228334,
|
|
"learning_rate": 8.31496062992126e-06,
|
|
"loss": 0.2851,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1488344669342041,
|
|
"step": 265,
|
|
"valid_targets_mean": 6288.4,
|
|
"valid_targets_min": 5380
|
|
},
|
|
{
|
|
"epoch": 0.14884233737596472,
|
|
"grad_norm": 0.445107896206196,
|
|
"learning_rate": 8.47244094488189e-06,
|
|
"loss": 0.2702,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11576345562934875,
|
|
"step": 270,
|
|
"valid_targets_mean": 7095.9,
|
|
"valid_targets_min": 5048
|
|
},
|
|
{
|
|
"epoch": 0.1515986769570011,
|
|
"grad_norm": 0.488604055995414,
|
|
"learning_rate": 8.629921259842521e-06,
|
|
"loss": 0.2856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14103032648563385,
|
|
"step": 275,
|
|
"valid_targets_mean": 6093.5,
|
|
"valid_targets_min": 4489
|
|
},
|
|
{
|
|
"epoch": 0.1543550165380375,
|
|
"grad_norm": 2.775375872004364,
|
|
"learning_rate": 8.78740157480315e-06,
|
|
"loss": 0.3423,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.24638140201568604,
|
|
"step": 280,
|
|
"valid_targets_mean": 1879.2,
|
|
"valid_targets_min": 1107
|
|
},
|
|
{
|
|
"epoch": 0.15711135611907387,
|
|
"grad_norm": 1.441227619722244,
|
|
"learning_rate": 8.94488188976378e-06,
|
|
"loss": 0.4272,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21375879645347595,
|
|
"step": 285,
|
|
"valid_targets_mean": 1493.0,
|
|
"valid_targets_min": 892
|
|
},
|
|
{
|
|
"epoch": 0.15986769570011025,
|
|
"grad_norm": 1.2195308742958342,
|
|
"learning_rate": 9.10236220472441e-06,
|
|
"loss": 0.3757,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1320396065711975,
|
|
"step": 290,
|
|
"valid_targets_mean": 1350.0,
|
|
"valid_targets_min": 908
|
|
},
|
|
{
|
|
"epoch": 0.16262403528114663,
|
|
"grad_norm": 1.150074913031321,
|
|
"learning_rate": 9.259842519685041e-06,
|
|
"loss": 0.3584,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15151718258857727,
|
|
"step": 295,
|
|
"valid_targets_mean": 1319.6,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 0.16538037486218302,
|
|
"grad_norm": 0.9687913721638708,
|
|
"learning_rate": 9.41732283464567e-06,
|
|
"loss": 0.3481,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16266636550426483,
|
|
"step": 300,
|
|
"valid_targets_mean": 1923.8,
|
|
"valid_targets_min": 875
|
|
},
|
|
{
|
|
"epoch": 0.1681367144432194,
|
|
"grad_norm": 1.0609682985213373,
|
|
"learning_rate": 9.5748031496063e-06,
|
|
"loss": 0.3348,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15054169297218323,
|
|
"step": 305,
|
|
"valid_targets_mean": 1288.6,
|
|
"valid_targets_min": 731
|
|
},
|
|
{
|
|
"epoch": 0.17089305402425578,
|
|
"grad_norm": 0.99179710699151,
|
|
"learning_rate": 9.73228346456693e-06,
|
|
"loss": 0.3395,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13270239531993866,
|
|
"step": 310,
|
|
"valid_targets_mean": 1464.1,
|
|
"valid_targets_min": 652
|
|
},
|
|
{
|
|
"epoch": 0.17364939360529216,
|
|
"grad_norm": 1.1341059924131913,
|
|
"learning_rate": 9.88976377952756e-06,
|
|
"loss": 0.307,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1618819683790207,
|
|
"step": 315,
|
|
"valid_targets_mean": 1542.4,
|
|
"valid_targets_min": 736
|
|
},
|
|
{
|
|
"epoch": 0.17640573318632854,
|
|
"grad_norm": 1.1997885706945806,
|
|
"learning_rate": 1.0047244094488188e-05,
|
|
"loss": 0.3267,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14182543754577637,
|
|
"step": 320,
|
|
"valid_targets_mean": 1312.0,
|
|
"valid_targets_min": 758
|
|
},
|
|
{
|
|
"epoch": 0.17916207276736493,
|
|
"grad_norm": 1.022021883133075,
|
|
"learning_rate": 1.0204724409448821e-05,
|
|
"loss": 0.3045,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11816409975290298,
|
|
"step": 325,
|
|
"valid_targets_mean": 1246.1,
|
|
"valid_targets_min": 752
|
|
},
|
|
{
|
|
"epoch": 0.1819184123484013,
|
|
"grad_norm": 1.0760778887241602,
|
|
"learning_rate": 1.0362204724409451e-05,
|
|
"loss": 0.3018,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19339129328727722,
|
|
"step": 330,
|
|
"valid_targets_mean": 1828.1,
|
|
"valid_targets_min": 926
|
|
},
|
|
{
|
|
"epoch": 0.18467475192943772,
|
|
"grad_norm": 1.0654082032603325,
|
|
"learning_rate": 1.051968503937008e-05,
|
|
"loss": 0.3111,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14686307311058044,
|
|
"step": 335,
|
|
"valid_targets_mean": 1334.8,
|
|
"valid_targets_min": 676
|
|
},
|
|
{
|
|
"epoch": 0.1874310915104741,
|
|
"grad_norm": 1.1297295236868727,
|
|
"learning_rate": 1.067716535433071e-05,
|
|
"loss": 0.2988,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13286492228507996,
|
|
"step": 340,
|
|
"valid_targets_mean": 1240.0,
|
|
"valid_targets_min": 769
|
|
},
|
|
{
|
|
"epoch": 0.19018743109151048,
|
|
"grad_norm": 1.0415065950273228,
|
|
"learning_rate": 1.083464566929134e-05,
|
|
"loss": 0.2791,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13291111588478088,
|
|
"step": 345,
|
|
"valid_targets_mean": 1725.1,
|
|
"valid_targets_min": 982
|
|
},
|
|
{
|
|
"epoch": 0.19294377067254687,
|
|
"grad_norm": 0.9781842383825073,
|
|
"learning_rate": 1.0992125984251969e-05,
|
|
"loss": 0.2883,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1279613971710205,
|
|
"step": 350,
|
|
"valid_targets_mean": 1385.1,
|
|
"valid_targets_min": 1099
|
|
},
|
|
{
|
|
"epoch": 0.19570011025358325,
|
|
"grad_norm": 0.9656503373590308,
|
|
"learning_rate": 1.1149606299212599e-05,
|
|
"loss": 0.2907,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1577109694480896,
|
|
"step": 355,
|
|
"valid_targets_mean": 1588.1,
|
|
"valid_targets_min": 674
|
|
},
|
|
{
|
|
"epoch": 0.19845644983461963,
|
|
"grad_norm": 0.9735447605625639,
|
|
"learning_rate": 1.1307086614173228e-05,
|
|
"loss": 0.289,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17590075731277466,
|
|
"step": 360,
|
|
"valid_targets_mean": 1918.5,
|
|
"valid_targets_min": 973
|
|
},
|
|
{
|
|
"epoch": 0.201212789415656,
|
|
"grad_norm": 1.023617657326464,
|
|
"learning_rate": 1.1464566929133861e-05,
|
|
"loss": 0.2937,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17073982954025269,
|
|
"step": 365,
|
|
"valid_targets_mean": 1943.9,
|
|
"valid_targets_min": 865
|
|
},
|
|
{
|
|
"epoch": 0.2039691289966924,
|
|
"grad_norm": 1.0026599725658034,
|
|
"learning_rate": 1.162204724409449e-05,
|
|
"loss": 0.2913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12909388542175293,
|
|
"step": 370,
|
|
"valid_targets_mean": 1323.0,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 0.20672546857772878,
|
|
"grad_norm": 0.9684110258183802,
|
|
"learning_rate": 1.177952755905512e-05,
|
|
"loss": 0.2933,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15493203699588776,
|
|
"step": 375,
|
|
"valid_targets_mean": 1712.6,
|
|
"valid_targets_min": 708
|
|
},
|
|
{
|
|
"epoch": 0.20948180815876516,
|
|
"grad_norm": 1.2171262527001416,
|
|
"learning_rate": 1.193700787401575e-05,
|
|
"loss": 0.2792,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12298417091369629,
|
|
"step": 380,
|
|
"valid_targets_mean": 1137.1,
|
|
"valid_targets_min": 657
|
|
},
|
|
{
|
|
"epoch": 0.21223814773980154,
|
|
"grad_norm": 1.0030073116623575,
|
|
"learning_rate": 1.2094488188976379e-05,
|
|
"loss": 0.2879,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13702605664730072,
|
|
"step": 385,
|
|
"valid_targets_mean": 1440.4,
|
|
"valid_targets_min": 661
|
|
},
|
|
{
|
|
"epoch": 0.21499448732083792,
|
|
"grad_norm": 0.9439151485330421,
|
|
"learning_rate": 1.2251968503937009e-05,
|
|
"loss": 0.2946,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10151394456624985,
|
|
"step": 390,
|
|
"valid_targets_mean": 1238.1,
|
|
"valid_targets_min": 858
|
|
},
|
|
{
|
|
"epoch": 0.2177508269018743,
|
|
"grad_norm": 1.1937652205842209,
|
|
"learning_rate": 1.2409448818897638e-05,
|
|
"loss": 0.2651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16189324855804443,
|
|
"step": 395,
|
|
"valid_targets_mean": 1564.0,
|
|
"valid_targets_min": 949
|
|
},
|
|
{
|
|
"epoch": 0.2205071664829107,
|
|
"grad_norm": 0.9630529800475339,
|
|
"learning_rate": 1.2566929133858268e-05,
|
|
"loss": 0.2774,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13084812462329865,
|
|
"step": 400,
|
|
"valid_targets_mean": 1533.5,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 0.22326350606394707,
|
|
"grad_norm": 1.013663282485196,
|
|
"learning_rate": 1.2724409448818899e-05,
|
|
"loss": 0.2716,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15152710676193237,
|
|
"step": 405,
|
|
"valid_targets_mean": 1604.9,
|
|
"valid_targets_min": 701
|
|
},
|
|
{
|
|
"epoch": 0.22601984564498345,
|
|
"grad_norm": 1.2747746644200577,
|
|
"learning_rate": 1.288188976377953e-05,
|
|
"loss": 0.2665,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10031342506408691,
|
|
"step": 410,
|
|
"valid_targets_mean": 1186.8,
|
|
"valid_targets_min": 776
|
|
},
|
|
{
|
|
"epoch": 0.22877618522601983,
|
|
"grad_norm": 1.0045241440870982,
|
|
"learning_rate": 1.303937007874016e-05,
|
|
"loss": 0.2818,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14396819472312927,
|
|
"step": 415,
|
|
"valid_targets_mean": 1462.8,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 0.23153252480705622,
|
|
"grad_norm": 1.1221966632471743,
|
|
"learning_rate": 1.3196850393700789e-05,
|
|
"loss": 0.2592,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11405763030052185,
|
|
"step": 420,
|
|
"valid_targets_mean": 1345.4,
|
|
"valid_targets_min": 867
|
|
},
|
|
{
|
|
"epoch": 0.2342888643880926,
|
|
"grad_norm": 1.1024764767248383,
|
|
"learning_rate": 1.3354330708661419e-05,
|
|
"loss": 0.2712,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11062923073768616,
|
|
"step": 425,
|
|
"valid_targets_mean": 1279.9,
|
|
"valid_targets_min": 780
|
|
},
|
|
{
|
|
"epoch": 0.237045203969129,
|
|
"grad_norm": 1.0126668533386174,
|
|
"learning_rate": 1.3511811023622048e-05,
|
|
"loss": 0.2728,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11329410970211029,
|
|
"step": 430,
|
|
"valid_targets_mean": 1299.2,
|
|
"valid_targets_min": 792
|
|
},
|
|
{
|
|
"epoch": 0.2398015435501654,
|
|
"grad_norm": 0.9858509249350107,
|
|
"learning_rate": 1.3669291338582678e-05,
|
|
"loss": 0.269,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11664732545614243,
|
|
"step": 435,
|
|
"valid_targets_mean": 1305.6,
|
|
"valid_targets_min": 805
|
|
},
|
|
{
|
|
"epoch": 0.24255788313120177,
|
|
"grad_norm": 0.9650157455414021,
|
|
"learning_rate": 1.3826771653543307e-05,
|
|
"loss": 0.2669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1122482493519783,
|
|
"step": 440,
|
|
"valid_targets_mean": 1392.9,
|
|
"valid_targets_min": 802
|
|
},
|
|
{
|
|
"epoch": 0.24531422271223816,
|
|
"grad_norm": 1.0767398510619248,
|
|
"learning_rate": 1.3984251968503938e-05,
|
|
"loss": 0.2633,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09951470792293549,
|
|
"step": 445,
|
|
"valid_targets_mean": 1183.9,
|
|
"valid_targets_min": 658
|
|
},
|
|
{
|
|
"epoch": 0.24807056229327454,
|
|
"grad_norm": 0.9353557534438305,
|
|
"learning_rate": 1.4141732283464568e-05,
|
|
"loss": 0.2463,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10740942507982254,
|
|
"step": 450,
|
|
"valid_targets_mean": 1540.5,
|
|
"valid_targets_min": 852
|
|
},
|
|
{
|
|
"epoch": 0.2508269018743109,
|
|
"grad_norm": 0.9578004827876149,
|
|
"learning_rate": 1.4299212598425199e-05,
|
|
"loss": 0.2649,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13750100135803223,
|
|
"step": 455,
|
|
"valid_targets_mean": 1697.2,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 0.2535832414553473,
|
|
"grad_norm": 0.9824159553885203,
|
|
"learning_rate": 1.4456692913385829e-05,
|
|
"loss": 0.2677,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09515532851219177,
|
|
"step": 460,
|
|
"valid_targets_mean": 1251.9,
|
|
"valid_targets_min": 763
|
|
},
|
|
{
|
|
"epoch": 0.25633958103638366,
|
|
"grad_norm": 0.97689002243497,
|
|
"learning_rate": 1.4614173228346458e-05,
|
|
"loss": 0.2614,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1298409104347229,
|
|
"step": 465,
|
|
"valid_targets_mean": 1545.5,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 0.25909592061742004,
|
|
"grad_norm": 1.0422840548566852,
|
|
"learning_rate": 1.4771653543307088e-05,
|
|
"loss": 0.2589,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12738534808158875,
|
|
"step": 470,
|
|
"valid_targets_mean": 1356.6,
|
|
"valid_targets_min": 736
|
|
},
|
|
{
|
|
"epoch": 0.2618522601984565,
|
|
"grad_norm": 0.9754211110993389,
|
|
"learning_rate": 1.4929133858267717e-05,
|
|
"loss": 0.2697,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14909060299396515,
|
|
"step": 475,
|
|
"valid_targets_mean": 1661.6,
|
|
"valid_targets_min": 759
|
|
},
|
|
{
|
|
"epoch": 0.26460859977949286,
|
|
"grad_norm": 1.0416795259799272,
|
|
"learning_rate": 1.5086614173228347e-05,
|
|
"loss": 0.2513,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11041361093521118,
|
|
"step": 480,
|
|
"valid_targets_mean": 1315.0,
|
|
"valid_targets_min": 1045
|
|
},
|
|
{
|
|
"epoch": 0.26736493936052924,
|
|
"grad_norm": 1.5132575842185814,
|
|
"learning_rate": 1.5244094488188978e-05,
|
|
"loss": 0.2713,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14459916949272156,
|
|
"step": 485,
|
|
"valid_targets_mean": 1722.8,
|
|
"valid_targets_min": 847
|
|
},
|
|
{
|
|
"epoch": 0.2701212789415656,
|
|
"grad_norm": 1.229596865988589,
|
|
"learning_rate": 1.5401574803149607e-05,
|
|
"loss": 0.2709,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1392192542552948,
|
|
"step": 490,
|
|
"valid_targets_mean": 1637.4,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 0.272877618522602,
|
|
"grad_norm": 1.018369930018895,
|
|
"learning_rate": 1.555905511811024e-05,
|
|
"loss": 0.2749,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11700350046157837,
|
|
"step": 495,
|
|
"valid_targets_mean": 1473.2,
|
|
"valid_targets_min": 760
|
|
},
|
|
{
|
|
"epoch": 0.2756339581036384,
|
|
"grad_norm": 0.8957757268045856,
|
|
"learning_rate": 1.5716535433070866e-05,
|
|
"loss": 0.2545,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11958117038011551,
|
|
"step": 500,
|
|
"valid_targets_mean": 1519.1,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 0.27839029768467477,
|
|
"grad_norm": 0.9374016494813244,
|
|
"learning_rate": 1.5874015748031498e-05,
|
|
"loss": 0.2538,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12481006979942322,
|
|
"step": 505,
|
|
"valid_targets_mean": 1582.4,
|
|
"valid_targets_min": 982
|
|
},
|
|
{
|
|
"epoch": 0.28114663726571115,
|
|
"grad_norm": 3.1426216470752353,
|
|
"learning_rate": 1.6031496062992126e-05,
|
|
"loss": 0.2653,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12947680056095123,
|
|
"step": 510,
|
|
"valid_targets_mean": 1568.8,
|
|
"valid_targets_min": 815
|
|
},
|
|
{
|
|
"epoch": 0.28390297684674753,
|
|
"grad_norm": 0.9692321238203334,
|
|
"learning_rate": 1.6188976377952757e-05,
|
|
"loss": 0.2546,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12825597822666168,
|
|
"step": 515,
|
|
"valid_targets_mean": 1417.9,
|
|
"valid_targets_min": 720
|
|
},
|
|
{
|
|
"epoch": 0.2866593164277839,
|
|
"grad_norm": 1.021941382458699,
|
|
"learning_rate": 1.6346456692913385e-05,
|
|
"loss": 0.2477,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09396955370903015,
|
|
"step": 520,
|
|
"valid_targets_mean": 1146.1,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 0.2894156560088203,
|
|
"grad_norm": 1.0325986156767561,
|
|
"learning_rate": 1.650393700787402e-05,
|
|
"loss": 0.2562,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1393774449825287,
|
|
"step": 525,
|
|
"valid_targets_mean": 1676.6,
|
|
"valid_targets_min": 733
|
|
},
|
|
{
|
|
"epoch": 0.2921719955898567,
|
|
"grad_norm": 0.9898836513074519,
|
|
"learning_rate": 1.6661417322834647e-05,
|
|
"loss": 0.2593,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1543339192867279,
|
|
"step": 530,
|
|
"valid_targets_mean": 1928.6,
|
|
"valid_targets_min": 743
|
|
},
|
|
{
|
|
"epoch": 0.29492833517089306,
|
|
"grad_norm": 0.9652968639736843,
|
|
"learning_rate": 1.6818897637795278e-05,
|
|
"loss": 0.2561,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14497727155685425,
|
|
"step": 535,
|
|
"valid_targets_mean": 1836.4,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 0.29768467475192945,
|
|
"grad_norm": 0.9203871220770407,
|
|
"learning_rate": 1.6976377952755906e-05,
|
|
"loss": 0.249,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14922620356082916,
|
|
"step": 540,
|
|
"valid_targets_mean": 1631.6,
|
|
"valid_targets_min": 784
|
|
},
|
|
{
|
|
"epoch": 0.30044101433296583,
|
|
"grad_norm": 0.8366961734159202,
|
|
"learning_rate": 1.7133858267716537e-05,
|
|
"loss": 0.2542,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12488574534654617,
|
|
"step": 545,
|
|
"valid_targets_mean": 1962.4,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 0.3031973539140022,
|
|
"grad_norm": 1.049401734887797,
|
|
"learning_rate": 1.7291338582677165e-05,
|
|
"loss": 0.2489,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1140659749507904,
|
|
"step": 550,
|
|
"valid_targets_mean": 1426.4,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 0.3059536934950386,
|
|
"grad_norm": 0.9808829995648646,
|
|
"learning_rate": 1.7448818897637796e-05,
|
|
"loss": 0.2504,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13592153787612915,
|
|
"step": 555,
|
|
"valid_targets_mean": 1690.1,
|
|
"valid_targets_min": 1121
|
|
},
|
|
{
|
|
"epoch": 0.308710033076075,
|
|
"grad_norm": 1.0312967179269659,
|
|
"learning_rate": 1.7606299212598424e-05,
|
|
"loss": 0.2482,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13333091139793396,
|
|
"step": 560,
|
|
"valid_targets_mean": 1519.1,
|
|
"valid_targets_min": 721
|
|
},
|
|
{
|
|
"epoch": 0.31146637265711136,
|
|
"grad_norm": 0.979250505106601,
|
|
"learning_rate": 1.776377952755906e-05,
|
|
"loss": 0.2344,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10655708611011505,
|
|
"step": 565,
|
|
"valid_targets_mean": 1284.8,
|
|
"valid_targets_min": 660
|
|
},
|
|
{
|
|
"epoch": 0.31422271223814774,
|
|
"grad_norm": 1.1425190157218426,
|
|
"learning_rate": 1.7921259842519687e-05,
|
|
"loss": 0.2405,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11416280269622803,
|
|
"step": 570,
|
|
"valid_targets_mean": 1513.5,
|
|
"valid_targets_min": 1009
|
|
},
|
|
{
|
|
"epoch": 0.3169790518191841,
|
|
"grad_norm": 0.9522749729264568,
|
|
"learning_rate": 1.8078740157480318e-05,
|
|
"loss": 0.2515,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13329800963401794,
|
|
"step": 575,
|
|
"valid_targets_mean": 1724.6,
|
|
"valid_targets_min": 1273
|
|
},
|
|
{
|
|
"epoch": 0.3197353914002205,
|
|
"grad_norm": 1.266482727575943,
|
|
"learning_rate": 1.8236220472440946e-05,
|
|
"loss": 0.2379,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1285274773836136,
|
|
"step": 580,
|
|
"valid_targets_mean": 1481.1,
|
|
"valid_targets_min": 806
|
|
},
|
|
{
|
|
"epoch": 0.3224917309812569,
|
|
"grad_norm": 0.9123829825437727,
|
|
"learning_rate": 1.8393700787401577e-05,
|
|
"loss": 0.252,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1404200792312622,
|
|
"step": 585,
|
|
"valid_targets_mean": 1545.4,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 0.32524807056229327,
|
|
"grad_norm": 0.9452884021577623,
|
|
"learning_rate": 1.8551181102362205e-05,
|
|
"loss": 0.2466,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11781813204288483,
|
|
"step": 590,
|
|
"valid_targets_mean": 1649.8,
|
|
"valid_targets_min": 1290
|
|
},
|
|
{
|
|
"epoch": 0.32800441014332965,
|
|
"grad_norm": 1.0797530398147628,
|
|
"learning_rate": 1.8708661417322836e-05,
|
|
"loss": 0.2473,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11188550293445587,
|
|
"step": 595,
|
|
"valid_targets_mean": 1363.4,
|
|
"valid_targets_min": 789
|
|
},
|
|
{
|
|
"epoch": 0.33076074972436603,
|
|
"grad_norm": 0.9742387097574635,
|
|
"learning_rate": 1.8866141732283464e-05,
|
|
"loss": 0.243,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13744091987609863,
|
|
"step": 600,
|
|
"valid_targets_mean": 1756.9,
|
|
"valid_targets_min": 589
|
|
},
|
|
{
|
|
"epoch": 0.3335170893054024,
|
|
"grad_norm": 1.0007130907728197,
|
|
"learning_rate": 1.9023622047244098e-05,
|
|
"loss": 0.2493,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13109591603279114,
|
|
"step": 605,
|
|
"valid_targets_mean": 1581.2,
|
|
"valid_targets_min": 659
|
|
},
|
|
{
|
|
"epoch": 0.3362734288864388,
|
|
"grad_norm": 0.9836978799108544,
|
|
"learning_rate": 1.9181102362204726e-05,
|
|
"loss": 0.2528,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.121638223528862,
|
|
"step": 610,
|
|
"valid_targets_mean": 1332.0,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 0.3390297684674752,
|
|
"grad_norm": 1.0137480335223172,
|
|
"learning_rate": 1.9338582677165357e-05,
|
|
"loss": 0.2493,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07734552025794983,
|
|
"step": 615,
|
|
"valid_targets_mean": 1093.6,
|
|
"valid_targets_min": 624
|
|
},
|
|
{
|
|
"epoch": 0.34178610804851156,
|
|
"grad_norm": 0.9702634228924609,
|
|
"learning_rate": 1.9496062992125985e-05,
|
|
"loss": 0.2623,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09776420891284943,
|
|
"step": 620,
|
|
"valid_targets_mean": 1478.8,
|
|
"valid_targets_min": 883
|
|
},
|
|
{
|
|
"epoch": 0.34454244762954794,
|
|
"grad_norm": 0.8711980109735163,
|
|
"learning_rate": 1.9653543307086616e-05,
|
|
"loss": 0.2536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13349077105522156,
|
|
"step": 625,
|
|
"valid_targets_mean": 1750.6,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 0.3472987872105843,
|
|
"grad_norm": 1.036347932698403,
|
|
"learning_rate": 1.9811023622047244e-05,
|
|
"loss": 0.2558,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14377659559249878,
|
|
"step": 630,
|
|
"valid_targets_mean": 1472.2,
|
|
"valid_targets_min": 520
|
|
},
|
|
{
|
|
"epoch": 0.3500551267916207,
|
|
"grad_norm": 1.1103956320828516,
|
|
"learning_rate": 1.9968503937007875e-05,
|
|
"loss": 0.2415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11092190444469452,
|
|
"step": 635,
|
|
"valid_targets_mean": 1226.2,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 0.3528114663726571,
|
|
"grad_norm": 0.912955910938678,
|
|
"learning_rate": 2.0125984251968503e-05,
|
|
"loss": 0.2383,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13074325025081635,
|
|
"step": 640,
|
|
"valid_targets_mean": 1665.5,
|
|
"valid_targets_min": 1008
|
|
},
|
|
{
|
|
"epoch": 0.35556780595369347,
|
|
"grad_norm": 1.0010063000344573,
|
|
"learning_rate": 2.0283464566929134e-05,
|
|
"loss": 0.2451,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12039319425821304,
|
|
"step": 645,
|
|
"valid_targets_mean": 1405.9,
|
|
"valid_targets_min": 768
|
|
},
|
|
{
|
|
"epoch": 0.35832414553472985,
|
|
"grad_norm": 0.9147869696446752,
|
|
"learning_rate": 2.0440944881889762e-05,
|
|
"loss": 0.2475,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10318435728549957,
|
|
"step": 650,
|
|
"valid_targets_mean": 1294.5,
|
|
"valid_targets_min": 744
|
|
},
|
|
{
|
|
"epoch": 0.36108048511576624,
|
|
"grad_norm": 1.1019785053346651,
|
|
"learning_rate": 2.0598425196850393e-05,
|
|
"loss": 0.2428,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10286827385425568,
|
|
"step": 655,
|
|
"valid_targets_mean": 1116.0,
|
|
"valid_targets_min": 679
|
|
},
|
|
{
|
|
"epoch": 0.3638368246968026,
|
|
"grad_norm": 0.8919284575300424,
|
|
"learning_rate": 2.0755905511811028e-05,
|
|
"loss": 0.2359,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14463713765144348,
|
|
"step": 660,
|
|
"valid_targets_mean": 1898.9,
|
|
"valid_targets_min": 1340
|
|
},
|
|
{
|
|
"epoch": 0.36659316427783906,
|
|
"grad_norm": 0.9701501265595001,
|
|
"learning_rate": 2.0913385826771656e-05,
|
|
"loss": 0.2517,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17601154744625092,
|
|
"step": 665,
|
|
"valid_targets_mean": 1631.6,
|
|
"valid_targets_min": 850
|
|
},
|
|
{
|
|
"epoch": 0.36934950385887544,
|
|
"grad_norm": 0.9901937001000649,
|
|
"learning_rate": 2.1070866141732287e-05,
|
|
"loss": 0.237,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10573569685220718,
|
|
"step": 670,
|
|
"valid_targets_mean": 1451.9,
|
|
"valid_targets_min": 902
|
|
},
|
|
{
|
|
"epoch": 0.3721058434399118,
|
|
"grad_norm": 0.9756766540498443,
|
|
"learning_rate": 2.1228346456692915e-05,
|
|
"loss": 0.2342,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13148336112499237,
|
|
"step": 675,
|
|
"valid_targets_mean": 1667.8,
|
|
"valid_targets_min": 1237
|
|
},
|
|
{
|
|
"epoch": 0.3748621830209482,
|
|
"grad_norm": 0.7044212434733603,
|
|
"learning_rate": 2.1385826771653546e-05,
|
|
"loss": 0.2318,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09703606367111206,
|
|
"step": 680,
|
|
"valid_targets_mean": 1690.6,
|
|
"valid_targets_min": 916
|
|
},
|
|
{
|
|
"epoch": 0.3776185226019846,
|
|
"grad_norm": 1.0033465968147017,
|
|
"learning_rate": 2.1543307086614177e-05,
|
|
"loss": 0.2182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10079711675643921,
|
|
"step": 685,
|
|
"valid_targets_mean": 1573.8,
|
|
"valid_targets_min": 925
|
|
},
|
|
{
|
|
"epoch": 0.38037486218302097,
|
|
"grad_norm": 0.8841758471203265,
|
|
"learning_rate": 2.1700787401574805e-05,
|
|
"loss": 0.2416,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12383813410997391,
|
|
"step": 690,
|
|
"valid_targets_mean": 1649.0,
|
|
"valid_targets_min": 1220
|
|
},
|
|
{
|
|
"epoch": 0.38313120176405735,
|
|
"grad_norm": 0.895200736657787,
|
|
"learning_rate": 2.1858267716535436e-05,
|
|
"loss": 0.2365,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11923413723707199,
|
|
"step": 695,
|
|
"valid_targets_mean": 1560.5,
|
|
"valid_targets_min": 869
|
|
},
|
|
{
|
|
"epoch": 0.38588754134509373,
|
|
"grad_norm": 0.9656032825338556,
|
|
"learning_rate": 2.2015748031496064e-05,
|
|
"loss": 0.2405,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11378061771392822,
|
|
"step": 700,
|
|
"valid_targets_mean": 1559.4,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 0.3886438809261301,
|
|
"grad_norm": 1.0812058433860616,
|
|
"learning_rate": 2.2173228346456695e-05,
|
|
"loss": 0.2439,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10770085453987122,
|
|
"step": 705,
|
|
"valid_targets_mean": 1535.9,
|
|
"valid_targets_min": 1211
|
|
},
|
|
{
|
|
"epoch": 0.3914002205071665,
|
|
"grad_norm": 0.9754684079077902,
|
|
"learning_rate": 2.2330708661417323e-05,
|
|
"loss": 0.2331,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10976245999336243,
|
|
"step": 710,
|
|
"valid_targets_mean": 1264.5,
|
|
"valid_targets_min": 590
|
|
},
|
|
{
|
|
"epoch": 0.3941565600882029,
|
|
"grad_norm": 1.002206594754782,
|
|
"learning_rate": 2.2488188976377954e-05,
|
|
"loss": 0.248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09050289541482925,
|
|
"step": 715,
|
|
"valid_targets_mean": 1228.1,
|
|
"valid_targets_min": 644
|
|
},
|
|
{
|
|
"epoch": 0.39691289966923926,
|
|
"grad_norm": 0.9653029505100796,
|
|
"learning_rate": 2.2645669291338582e-05,
|
|
"loss": 0.3437,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20324426889419556,
|
|
"step": 720,
|
|
"valid_targets_mean": 5088.5,
|
|
"valid_targets_min": 1966
|
|
},
|
|
{
|
|
"epoch": 0.39966923925027564,
|
|
"grad_norm": 0.7136745563133086,
|
|
"learning_rate": 2.2803149606299214e-05,
|
|
"loss": 0.4111,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.22606512904167175,
|
|
"step": 725,
|
|
"valid_targets_mean": 5210.0,
|
|
"valid_targets_min": 1628
|
|
},
|
|
{
|
|
"epoch": 0.402425578831312,
|
|
"grad_norm": 0.940106817222379,
|
|
"learning_rate": 2.296062992125984e-05,
|
|
"loss": 0.404,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2563844919204712,
|
|
"step": 730,
|
|
"valid_targets_mean": 5955.5,
|
|
"valid_targets_min": 3288
|
|
},
|
|
{
|
|
"epoch": 0.4051819184123484,
|
|
"grad_norm": 0.6270625170060502,
|
|
"learning_rate": 2.3118110236220473e-05,
|
|
"loss": 0.3837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16112765669822693,
|
|
"step": 735,
|
|
"valid_targets_mean": 4237.4,
|
|
"valid_targets_min": 2734
|
|
},
|
|
{
|
|
"epoch": 0.4079382579933848,
|
|
"grad_norm": 0.6152741400818889,
|
|
"learning_rate": 2.3275590551181107e-05,
|
|
"loss": 0.3798,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15670859813690186,
|
|
"step": 740,
|
|
"valid_targets_mean": 3888.1,
|
|
"valid_targets_min": 1439
|
|
},
|
|
{
|
|
"epoch": 0.41069459757442117,
|
|
"grad_norm": 0.5597278248256056,
|
|
"learning_rate": 2.3433070866141735e-05,
|
|
"loss": 0.3967,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15280428528785706,
|
|
"step": 745,
|
|
"valid_targets_mean": 4472.4,
|
|
"valid_targets_min": 1512
|
|
},
|
|
{
|
|
"epoch": 0.41345093715545755,
|
|
"grad_norm": 0.589097976392343,
|
|
"learning_rate": 2.3590551181102366e-05,
|
|
"loss": 0.3637,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.201160728931427,
|
|
"step": 750,
|
|
"valid_targets_mean": 5120.1,
|
|
"valid_targets_min": 2394
|
|
},
|
|
{
|
|
"epoch": 0.41620727673649394,
|
|
"grad_norm": 0.5472108289313626,
|
|
"learning_rate": 2.3748031496062994e-05,
|
|
"loss": 0.3746,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15171462297439575,
|
|
"step": 755,
|
|
"valid_targets_mean": 3248.4,
|
|
"valid_targets_min": 1654
|
|
},
|
|
{
|
|
"epoch": 0.4189636163175303,
|
|
"grad_norm": 0.5009310919395555,
|
|
"learning_rate": 2.3905511811023625e-05,
|
|
"loss": 0.3699,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17371916770935059,
|
|
"step": 760,
|
|
"valid_targets_mean": 5733.6,
|
|
"valid_targets_min": 4625
|
|
},
|
|
{
|
|
"epoch": 0.4217199558985667,
|
|
"grad_norm": 0.6314557641365345,
|
|
"learning_rate": 2.4062992125984253e-05,
|
|
"loss": 0.3883,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1840207576751709,
|
|
"step": 765,
|
|
"valid_targets_mean": 4134.8,
|
|
"valid_targets_min": 1544
|
|
},
|
|
{
|
|
"epoch": 0.4244762954796031,
|
|
"grad_norm": 0.5199745549887628,
|
|
"learning_rate": 2.4220472440944884e-05,
|
|
"loss": 0.3687,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17755737900733948,
|
|
"step": 770,
|
|
"valid_targets_mean": 5856.9,
|
|
"valid_targets_min": 950
|
|
},
|
|
{
|
|
"epoch": 0.42723263506063947,
|
|
"grad_norm": 0.5545263570617492,
|
|
"learning_rate": 2.4377952755905516e-05,
|
|
"loss": 0.3694,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19994176924228668,
|
|
"step": 775,
|
|
"valid_targets_mean": 4966.5,
|
|
"valid_targets_min": 1263
|
|
},
|
|
{
|
|
"epoch": 0.42998897464167585,
|
|
"grad_norm": 0.5275397494180947,
|
|
"learning_rate": 2.4535433070866143e-05,
|
|
"loss": 0.3654,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17413097620010376,
|
|
"step": 780,
|
|
"valid_targets_mean": 5353.5,
|
|
"valid_targets_min": 1980
|
|
},
|
|
{
|
|
"epoch": 0.43274531422271223,
|
|
"grad_norm": 0.6151386572669745,
|
|
"learning_rate": 2.4692913385826775e-05,
|
|
"loss": 0.3711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15998968482017517,
|
|
"step": 785,
|
|
"valid_targets_mean": 3883.9,
|
|
"valid_targets_min": 1755
|
|
},
|
|
{
|
|
"epoch": 0.4355016538037486,
|
|
"grad_norm": 0.6935850768067239,
|
|
"learning_rate": 2.4850393700787402e-05,
|
|
"loss": 0.3788,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17765235900878906,
|
|
"step": 790,
|
|
"valid_targets_mean": 3182.1,
|
|
"valid_targets_min": 1451
|
|
},
|
|
{
|
|
"epoch": 0.438257993384785,
|
|
"grad_norm": 0.6257142169924204,
|
|
"learning_rate": 2.5007874015748034e-05,
|
|
"loss": 0.3744,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18989720940589905,
|
|
"step": 795,
|
|
"valid_targets_mean": 4275.6,
|
|
"valid_targets_min": 1858
|
|
},
|
|
{
|
|
"epoch": 0.4410143329658214,
|
|
"grad_norm": 0.6273796431822124,
|
|
"learning_rate": 2.516535433070866e-05,
|
|
"loss": 0.3704,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19122779369354248,
|
|
"step": 800,
|
|
"valid_targets_mean": 4332.5,
|
|
"valid_targets_min": 2339
|
|
},
|
|
{
|
|
"epoch": 0.44377067254685776,
|
|
"grad_norm": 0.6099524507798427,
|
|
"learning_rate": 2.5322834645669293e-05,
|
|
"loss": 0.3817,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19384393095970154,
|
|
"step": 805,
|
|
"valid_targets_mean": 3915.5,
|
|
"valid_targets_min": 2766
|
|
},
|
|
{
|
|
"epoch": 0.44652701212789414,
|
|
"grad_norm": 0.6948751597682489,
|
|
"learning_rate": 2.548031496062992e-05,
|
|
"loss": 0.387,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20750582218170166,
|
|
"step": 810,
|
|
"valid_targets_mean": 3645.6,
|
|
"valid_targets_min": 1205
|
|
},
|
|
{
|
|
"epoch": 0.4492833517089305,
|
|
"grad_norm": 0.5636526389704282,
|
|
"learning_rate": 2.563779527559055e-05,
|
|
"loss": 0.3729,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15280693769454956,
|
|
"step": 815,
|
|
"valid_targets_mean": 3829.5,
|
|
"valid_targets_min": 2000
|
|
},
|
|
{
|
|
"epoch": 0.4520396912899669,
|
|
"grad_norm": 0.6329330164837811,
|
|
"learning_rate": 2.5795275590551186e-05,
|
|
"loss": 0.3777,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18266147375106812,
|
|
"step": 820,
|
|
"valid_targets_mean": 3772.6,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 0.4547960308710033,
|
|
"grad_norm": 0.60374305961393,
|
|
"learning_rate": 2.5952755905511814e-05,
|
|
"loss": 0.3571,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2090250849723816,
|
|
"step": 825,
|
|
"valid_targets_mean": 4300.1,
|
|
"valid_targets_min": 2297
|
|
},
|
|
{
|
|
"epoch": 0.45755237045203967,
|
|
"grad_norm": 0.6241781392542569,
|
|
"learning_rate": 2.6110236220472445e-05,
|
|
"loss": 0.3521,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19918176531791687,
|
|
"step": 830,
|
|
"valid_targets_mean": 4547.6,
|
|
"valid_targets_min": 2045
|
|
},
|
|
{
|
|
"epoch": 0.46030871003307605,
|
|
"grad_norm": 0.6545456915390673,
|
|
"learning_rate": 2.6267716535433073e-05,
|
|
"loss": 0.356,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2449052631855011,
|
|
"step": 835,
|
|
"valid_targets_mean": 4654.5,
|
|
"valid_targets_min": 3370
|
|
},
|
|
{
|
|
"epoch": 0.46306504961411243,
|
|
"grad_norm": 0.6224734961527387,
|
|
"learning_rate": 2.6425196850393704e-05,
|
|
"loss": 0.3681,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17789021134376526,
|
|
"step": 840,
|
|
"valid_targets_mean": 3721.4,
|
|
"valid_targets_min": 1743
|
|
},
|
|
{
|
|
"epoch": 0.4658213891951488,
|
|
"grad_norm": 0.6101007030559096,
|
|
"learning_rate": 2.6582677165354332e-05,
|
|
"loss": 0.3689,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1693587601184845,
|
|
"step": 845,
|
|
"valid_targets_mean": 3319.9,
|
|
"valid_targets_min": 1569
|
|
},
|
|
{
|
|
"epoch": 0.4685777287761852,
|
|
"grad_norm": 0.5778946527703491,
|
|
"learning_rate": 2.6740157480314963e-05,
|
|
"loss": 0.3504,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17173618078231812,
|
|
"step": 850,
|
|
"valid_targets_mean": 4595.6,
|
|
"valid_targets_min": 1537
|
|
},
|
|
{
|
|
"epoch": 0.47133406835722164,
|
|
"grad_norm": 0.5626214839049044,
|
|
"learning_rate": 2.689763779527559e-05,
|
|
"loss": 0.3514,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1453714668750763,
|
|
"step": 855,
|
|
"valid_targets_mean": 4320.4,
|
|
"valid_targets_min": 1934
|
|
},
|
|
{
|
|
"epoch": 0.474090407938258,
|
|
"grad_norm": 0.6256849943100633,
|
|
"learning_rate": 2.7055118110236222e-05,
|
|
"loss": 0.3607,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15242540836334229,
|
|
"step": 860,
|
|
"valid_targets_mean": 3487.9,
|
|
"valid_targets_min": 2556
|
|
},
|
|
{
|
|
"epoch": 0.4768467475192944,
|
|
"grad_norm": 0.6037494235011468,
|
|
"learning_rate": 2.7212598425196854e-05,
|
|
"loss": 0.3587,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16787093877792358,
|
|
"step": 865,
|
|
"valid_targets_mean": 4068.8,
|
|
"valid_targets_min": 2648
|
|
},
|
|
{
|
|
"epoch": 0.4796030871003308,
|
|
"grad_norm": 0.5971338817629256,
|
|
"learning_rate": 2.737007874015748e-05,
|
|
"loss": 0.3481,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16925574839115143,
|
|
"step": 870,
|
|
"valid_targets_mean": 4032.1,
|
|
"valid_targets_min": 1349
|
|
},
|
|
{
|
|
"epoch": 0.48235942668136716,
|
|
"grad_norm": 0.6505109914583836,
|
|
"learning_rate": 2.7527559055118113e-05,
|
|
"loss": 0.353,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1337091475725174,
|
|
"step": 875,
|
|
"valid_targets_mean": 2994.6,
|
|
"valid_targets_min": 1566
|
|
},
|
|
{
|
|
"epoch": 0.48511576626240355,
|
|
"grad_norm": 0.716702439020288,
|
|
"learning_rate": 2.768503937007874e-05,
|
|
"loss": 0.3634,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14347165822982788,
|
|
"step": 880,
|
|
"valid_targets_mean": 2621.4,
|
|
"valid_targets_min": 1499
|
|
},
|
|
{
|
|
"epoch": 0.48787210584343993,
|
|
"grad_norm": 0.6216700206602679,
|
|
"learning_rate": 2.7842519685039372e-05,
|
|
"loss": 0.3422,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15769225358963013,
|
|
"step": 885,
|
|
"valid_targets_mean": 3408.1,
|
|
"valid_targets_min": 2030
|
|
},
|
|
{
|
|
"epoch": 0.4906284454244763,
|
|
"grad_norm": 0.7455336132785676,
|
|
"learning_rate": 2.8e-05,
|
|
"loss": 0.3685,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2071845531463623,
|
|
"step": 890,
|
|
"valid_targets_mean": 2845.1,
|
|
"valid_targets_min": 1346
|
|
},
|
|
{
|
|
"epoch": 0.4933847850055127,
|
|
"grad_norm": 0.8013081494066265,
|
|
"learning_rate": 2.815748031496063e-05,
|
|
"loss": 0.3593,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19865600764751434,
|
|
"step": 895,
|
|
"valid_targets_mean": 2974.1,
|
|
"valid_targets_min": 1821
|
|
},
|
|
{
|
|
"epoch": 0.4961411245865491,
|
|
"grad_norm": 0.6288299107638378,
|
|
"learning_rate": 2.8314960629921265e-05,
|
|
"loss": 0.353,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1870521903038025,
|
|
"step": 900,
|
|
"valid_targets_mean": 3943.6,
|
|
"valid_targets_min": 2525
|
|
},
|
|
{
|
|
"epoch": 0.49889746416758546,
|
|
"grad_norm": 0.6916368657675858,
|
|
"learning_rate": 2.8472440944881893e-05,
|
|
"loss": 0.3472,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16577574610710144,
|
|
"step": 905,
|
|
"valid_targets_mean": 3039.4,
|
|
"valid_targets_min": 2070
|
|
},
|
|
{
|
|
"epoch": 0.5016538037486218,
|
|
"grad_norm": 0.699768360392489,
|
|
"learning_rate": 2.8629921259842524e-05,
|
|
"loss": 0.3481,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15919555723667145,
|
|
"step": 910,
|
|
"valid_targets_mean": 2970.5,
|
|
"valid_targets_min": 1152
|
|
},
|
|
{
|
|
"epoch": 0.5044101433296582,
|
|
"grad_norm": 0.6475930225779623,
|
|
"learning_rate": 2.8787401574803152e-05,
|
|
"loss": 0.3446,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15473319590091705,
|
|
"step": 915,
|
|
"valid_targets_mean": 3201.1,
|
|
"valid_targets_min": 1850
|
|
},
|
|
{
|
|
"epoch": 0.5071664829106945,
|
|
"grad_norm": 0.6435583752225621,
|
|
"learning_rate": 2.8944881889763783e-05,
|
|
"loss": 0.3555,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1666574776172638,
|
|
"step": 920,
|
|
"valid_targets_mean": 3636.6,
|
|
"valid_targets_min": 2196
|
|
},
|
|
{
|
|
"epoch": 0.5099228224917309,
|
|
"grad_norm": 0.6107755406083548,
|
|
"learning_rate": 2.910236220472441e-05,
|
|
"loss": 0.3535,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14116963744163513,
|
|
"step": 925,
|
|
"valid_targets_mean": 3762.9,
|
|
"valid_targets_min": 1390
|
|
},
|
|
{
|
|
"epoch": 0.5126791620727673,
|
|
"grad_norm": 0.5792609842458837,
|
|
"learning_rate": 2.9259842519685043e-05,
|
|
"loss": 0.349,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1683681309223175,
|
|
"step": 930,
|
|
"valid_targets_mean": 4341.6,
|
|
"valid_targets_min": 1844
|
|
},
|
|
{
|
|
"epoch": 0.5154355016538037,
|
|
"grad_norm": 0.7647537740460914,
|
|
"learning_rate": 2.941732283464567e-05,
|
|
"loss": 0.3464,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1972532421350479,
|
|
"step": 935,
|
|
"valid_targets_mean": 4335.2,
|
|
"valid_targets_min": 1733
|
|
},
|
|
{
|
|
"epoch": 0.5181918412348401,
|
|
"grad_norm": 0.6093904312433199,
|
|
"learning_rate": 2.95748031496063e-05,
|
|
"loss": 0.3418,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.191727876663208,
|
|
"step": 940,
|
|
"valid_targets_mean": 4243.1,
|
|
"valid_targets_min": 2065
|
|
},
|
|
{
|
|
"epoch": 0.5209481808158766,
|
|
"grad_norm": 0.5815924860097553,
|
|
"learning_rate": 2.9732283464566933e-05,
|
|
"loss": 0.3469,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16618087887763977,
|
|
"step": 945,
|
|
"valid_targets_mean": 4171.9,
|
|
"valid_targets_min": 2348
|
|
},
|
|
{
|
|
"epoch": 0.523704520396913,
|
|
"grad_norm": 0.564107060656102,
|
|
"learning_rate": 2.988976377952756e-05,
|
|
"loss": 0.3398,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18566842377185822,
|
|
"step": 950,
|
|
"valid_targets_mean": 4153.1,
|
|
"valid_targets_min": 1603
|
|
},
|
|
{
|
|
"epoch": 0.5264608599779493,
|
|
"grad_norm": 0.7171076200840043,
|
|
"learning_rate": 3.0047244094488192e-05,
|
|
"loss": 0.3581,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18485206365585327,
|
|
"step": 955,
|
|
"valid_targets_mean": 3267.9,
|
|
"valid_targets_min": 1377
|
|
},
|
|
{
|
|
"epoch": 0.5292171995589857,
|
|
"grad_norm": 0.6410217222662009,
|
|
"learning_rate": 3.020472440944882e-05,
|
|
"loss": 0.3578,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15395814180374146,
|
|
"step": 960,
|
|
"valid_targets_mean": 3298.8,
|
|
"valid_targets_min": 1902
|
|
},
|
|
{
|
|
"epoch": 0.5319735391400221,
|
|
"grad_norm": 0.6693956946083943,
|
|
"learning_rate": 3.036220472440945e-05,
|
|
"loss": 0.3442,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1517816185951233,
|
|
"step": 965,
|
|
"valid_targets_mean": 3316.8,
|
|
"valid_targets_min": 653
|
|
},
|
|
{
|
|
"epoch": 0.5347298787210585,
|
|
"grad_norm": 0.6171129373012616,
|
|
"learning_rate": 3.051968503937008e-05,
|
|
"loss": 0.3461,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1666797548532486,
|
|
"step": 970,
|
|
"valid_targets_mean": 3900.8,
|
|
"valid_targets_min": 2593
|
|
},
|
|
{
|
|
"epoch": 0.5374862183020949,
|
|
"grad_norm": 0.5926348193407744,
|
|
"learning_rate": 3.067716535433071e-05,
|
|
"loss": 0.3496,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1847071349620819,
|
|
"step": 975,
|
|
"valid_targets_mean": 4461.6,
|
|
"valid_targets_min": 1736
|
|
},
|
|
{
|
|
"epoch": 0.5402425578831312,
|
|
"grad_norm": 0.6314373721385794,
|
|
"learning_rate": 3.0834645669291345e-05,
|
|
"loss": 0.3452,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15181858837604523,
|
|
"step": 980,
|
|
"valid_targets_mean": 3218.8,
|
|
"valid_targets_min": 1137
|
|
},
|
|
{
|
|
"epoch": 0.5429988974641676,
|
|
"grad_norm": 0.6289395736029709,
|
|
"learning_rate": 3.099212598425197e-05,
|
|
"loss": 0.3244,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1689888834953308,
|
|
"step": 985,
|
|
"valid_targets_mean": 3535.5,
|
|
"valid_targets_min": 2500
|
|
},
|
|
{
|
|
"epoch": 0.545755237045204,
|
|
"grad_norm": 0.5858419132990518,
|
|
"learning_rate": 3.11496062992126e-05,
|
|
"loss": 0.3387,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17873020470142365,
|
|
"step": 990,
|
|
"valid_targets_mean": 4358.2,
|
|
"valid_targets_min": 1297
|
|
},
|
|
{
|
|
"epoch": 0.5485115766262404,
|
|
"grad_norm": 0.5895538445175481,
|
|
"learning_rate": 3.1307086614173235e-05,
|
|
"loss": 0.3199,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17114953696727753,
|
|
"step": 995,
|
|
"valid_targets_mean": 4368.6,
|
|
"valid_targets_min": 1101
|
|
},
|
|
{
|
|
"epoch": 0.5512679162072768,
|
|
"grad_norm": 0.6038337510510758,
|
|
"learning_rate": 3.146456692913386e-05,
|
|
"loss": 0.3618,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1916867196559906,
|
|
"step": 1000,
|
|
"valid_targets_mean": 3886.8,
|
|
"valid_targets_min": 2078
|
|
},
|
|
{
|
|
"epoch": 0.5540242557883132,
|
|
"grad_norm": 0.6783770860512053,
|
|
"learning_rate": 3.162204724409449e-05,
|
|
"loss": 0.3345,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13468030095100403,
|
|
"step": 1005,
|
|
"valid_targets_mean": 3047.1,
|
|
"valid_targets_min": 1739
|
|
},
|
|
{
|
|
"epoch": 0.5567805953693495,
|
|
"grad_norm": 0.626091666273822,
|
|
"learning_rate": 3.177952755905512e-05,
|
|
"loss": 0.3446,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18287551403045654,
|
|
"step": 1010,
|
|
"valid_targets_mean": 3684.5,
|
|
"valid_targets_min": 1233
|
|
},
|
|
{
|
|
"epoch": 0.5595369349503859,
|
|
"grad_norm": 0.6195581402861309,
|
|
"learning_rate": 3.193700787401575e-05,
|
|
"loss": 0.3473,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17982032895088196,
|
|
"step": 1015,
|
|
"valid_targets_mean": 3783.5,
|
|
"valid_targets_min": 1815
|
|
},
|
|
{
|
|
"epoch": 0.5622932745314223,
|
|
"grad_norm": 0.5967993749069237,
|
|
"learning_rate": 3.209448818897638e-05,
|
|
"loss": 0.3509,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17668241262435913,
|
|
"step": 1020,
|
|
"valid_targets_mean": 4099.9,
|
|
"valid_targets_min": 1649
|
|
},
|
|
{
|
|
"epoch": 0.5650496141124587,
|
|
"grad_norm": 0.703782546254294,
|
|
"learning_rate": 3.225196850393701e-05,
|
|
"loss": 0.3431,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18843522667884827,
|
|
"step": 1025,
|
|
"valid_targets_mean": 3281.1,
|
|
"valid_targets_min": 1366
|
|
},
|
|
{
|
|
"epoch": 0.5678059536934951,
|
|
"grad_norm": 0.5835931350793472,
|
|
"learning_rate": 3.240944881889764e-05,
|
|
"loss": 0.3272,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17816738784313202,
|
|
"step": 1030,
|
|
"valid_targets_mean": 4390.2,
|
|
"valid_targets_min": 2716
|
|
},
|
|
{
|
|
"epoch": 0.5705622932745315,
|
|
"grad_norm": 0.7758427071293258,
|
|
"learning_rate": 3.256692913385827e-05,
|
|
"loss": 0.3326,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17936623096466064,
|
|
"step": 1035,
|
|
"valid_targets_mean": 4261.9,
|
|
"valid_targets_min": 1955
|
|
},
|
|
{
|
|
"epoch": 0.5733186328555678,
|
|
"grad_norm": 0.692504545393788,
|
|
"learning_rate": 3.27244094488189e-05,
|
|
"loss": 0.337,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18274366855621338,
|
|
"step": 1040,
|
|
"valid_targets_mean": 3709.5,
|
|
"valid_targets_min": 1561
|
|
},
|
|
{
|
|
"epoch": 0.5760749724366042,
|
|
"grad_norm": 0.6022653582683537,
|
|
"learning_rate": 3.2881889763779527e-05,
|
|
"loss": 0.3296,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1696901172399521,
|
|
"step": 1045,
|
|
"valid_targets_mean": 3744.2,
|
|
"valid_targets_min": 1442
|
|
},
|
|
{
|
|
"epoch": 0.5788313120176406,
|
|
"grad_norm": 0.6423045908640236,
|
|
"learning_rate": 3.303937007874016e-05,
|
|
"loss": 0.3168,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14521223306655884,
|
|
"step": 1050,
|
|
"valid_targets_mean": 3034.4,
|
|
"valid_targets_min": 1501
|
|
},
|
|
{
|
|
"epoch": 0.581587651598677,
|
|
"grad_norm": 0.6451322722341951,
|
|
"learning_rate": 3.319685039370079e-05,
|
|
"loss": 0.3256,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16142389178276062,
|
|
"step": 1055,
|
|
"valid_targets_mean": 3149.1,
|
|
"valid_targets_min": 857
|
|
},
|
|
{
|
|
"epoch": 0.5843439911797134,
|
|
"grad_norm": 0.80169755299376,
|
|
"learning_rate": 3.3354330708661424e-05,
|
|
"loss": 0.338,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18803809583187103,
|
|
"step": 1060,
|
|
"valid_targets_mean": 4058.2,
|
|
"valid_targets_min": 1285
|
|
},
|
|
{
|
|
"epoch": 0.5871003307607497,
|
|
"grad_norm": 0.6160871530185313,
|
|
"learning_rate": 3.351181102362205e-05,
|
|
"loss": 0.3596,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16094212234020233,
|
|
"step": 1065,
|
|
"valid_targets_mean": 3909.1,
|
|
"valid_targets_min": 1530
|
|
},
|
|
{
|
|
"epoch": 0.5898566703417861,
|
|
"grad_norm": 0.6501385918054338,
|
|
"learning_rate": 3.366929133858268e-05,
|
|
"loss": 0.3318,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1822768896818161,
|
|
"step": 1070,
|
|
"valid_targets_mean": 4228.5,
|
|
"valid_targets_min": 2964
|
|
},
|
|
{
|
|
"epoch": 0.5926130099228225,
|
|
"grad_norm": 0.5994254124642484,
|
|
"learning_rate": 3.3826771653543314e-05,
|
|
"loss": 0.3463,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18721994757652283,
|
|
"step": 1075,
|
|
"valid_targets_mean": 4405.6,
|
|
"valid_targets_min": 2010
|
|
},
|
|
{
|
|
"epoch": 0.5953693495038589,
|
|
"grad_norm": 0.7151198708434924,
|
|
"learning_rate": 3.398425196850394e-05,
|
|
"loss": 0.3252,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1714073121547699,
|
|
"step": 1080,
|
|
"valid_targets_mean": 3541.2,
|
|
"valid_targets_min": 1731
|
|
},
|
|
{
|
|
"epoch": 0.5981256890848953,
|
|
"grad_norm": 0.5524918605306506,
|
|
"learning_rate": 3.414173228346457e-05,
|
|
"loss": 0.3368,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14528022706508636,
|
|
"step": 1085,
|
|
"valid_targets_mean": 4123.8,
|
|
"valid_targets_min": 1911
|
|
},
|
|
{
|
|
"epoch": 0.6008820286659317,
|
|
"grad_norm": 0.633476214112512,
|
|
"learning_rate": 3.42992125984252e-05,
|
|
"loss": 0.3295,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14453305304050446,
|
|
"step": 1090,
|
|
"valid_targets_mean": 3818.8,
|
|
"valid_targets_min": 2408
|
|
},
|
|
{
|
|
"epoch": 0.603638368246968,
|
|
"grad_norm": 0.5812149537876431,
|
|
"learning_rate": 3.445669291338583e-05,
|
|
"loss": 0.3154,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1805756390094757,
|
|
"step": 1095,
|
|
"valid_targets_mean": 4187.2,
|
|
"valid_targets_min": 1927
|
|
},
|
|
{
|
|
"epoch": 0.6063947078280044,
|
|
"grad_norm": 0.6609995555188877,
|
|
"learning_rate": 3.461417322834646e-05,
|
|
"loss": 0.3451,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1914011538028717,
|
|
"step": 1100,
|
|
"valid_targets_mean": 3147.5,
|
|
"valid_targets_min": 1643
|
|
},
|
|
{
|
|
"epoch": 0.6091510474090408,
|
|
"grad_norm": 0.73012671511645,
|
|
"learning_rate": 3.477165354330709e-05,
|
|
"loss": 0.3276,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14493557810783386,
|
|
"step": 1105,
|
|
"valid_targets_mean": 3082.9,
|
|
"valid_targets_min": 1459
|
|
},
|
|
{
|
|
"epoch": 0.6119073869900772,
|
|
"grad_norm": 0.9397961001916695,
|
|
"learning_rate": 3.4929133858267715e-05,
|
|
"loss": 0.3435,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1640443503856659,
|
|
"step": 1110,
|
|
"valid_targets_mean": 3235.0,
|
|
"valid_targets_min": 1425
|
|
},
|
|
{
|
|
"epoch": 0.6146637265711136,
|
|
"grad_norm": 0.6618822152756154,
|
|
"learning_rate": 3.508661417322835e-05,
|
|
"loss": 0.343,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14405249059200287,
|
|
"step": 1115,
|
|
"valid_targets_mean": 2764.4,
|
|
"valid_targets_min": 537
|
|
},
|
|
{
|
|
"epoch": 0.61742006615215,
|
|
"grad_norm": 0.5728827976754084,
|
|
"learning_rate": 3.524409448818898e-05,
|
|
"loss": 0.3364,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17296361923217773,
|
|
"step": 1120,
|
|
"valid_targets_mean": 4477.2,
|
|
"valid_targets_min": 1645
|
|
},
|
|
{
|
|
"epoch": 0.6201764057331863,
|
|
"grad_norm": 0.6660302333653578,
|
|
"learning_rate": 3.5401574803149606e-05,
|
|
"loss": 0.3679,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19500517845153809,
|
|
"step": 1125,
|
|
"valid_targets_mean": 3753.4,
|
|
"valid_targets_min": 2165
|
|
},
|
|
{
|
|
"epoch": 0.6229327453142227,
|
|
"grad_norm": 0.6518277779189289,
|
|
"learning_rate": 3.555905511811024e-05,
|
|
"loss": 0.3341,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14726774394512177,
|
|
"step": 1130,
|
|
"valid_targets_mean": 3160.9,
|
|
"valid_targets_min": 1623
|
|
},
|
|
{
|
|
"epoch": 0.6256890848952591,
|
|
"grad_norm": 0.5840464429398144,
|
|
"learning_rate": 3.571653543307087e-05,
|
|
"loss": 0.3433,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13763117790222168,
|
|
"step": 1135,
|
|
"valid_targets_mean": 3442.0,
|
|
"valid_targets_min": 1996
|
|
},
|
|
{
|
|
"epoch": 0.6284454244762955,
|
|
"grad_norm": 0.6437278871150419,
|
|
"learning_rate": 3.58740157480315e-05,
|
|
"loss": 0.347,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20802661776542664,
|
|
"step": 1140,
|
|
"valid_targets_mean": 4399.0,
|
|
"valid_targets_min": 2938
|
|
},
|
|
{
|
|
"epoch": 0.6312017640573319,
|
|
"grad_norm": 0.5430115972970458,
|
|
"learning_rate": 3.603149606299213e-05,
|
|
"loss": 0.3428,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18838849663734436,
|
|
"step": 1145,
|
|
"valid_targets_mean": 5350.9,
|
|
"valid_targets_min": 2644
|
|
},
|
|
{
|
|
"epoch": 0.6339581036383682,
|
|
"grad_norm": 0.5916790175996349,
|
|
"learning_rate": 3.618897637795276e-05,
|
|
"loss": 0.3431,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1996316909790039,
|
|
"step": 1150,
|
|
"valid_targets_mean": 4307.1,
|
|
"valid_targets_min": 1377
|
|
},
|
|
{
|
|
"epoch": 0.6367144432194046,
|
|
"grad_norm": 0.7970014743281822,
|
|
"learning_rate": 3.634645669291339e-05,
|
|
"loss": 0.3416,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1353786289691925,
|
|
"step": 1155,
|
|
"valid_targets_mean": 3389.5,
|
|
"valid_targets_min": 636
|
|
},
|
|
{
|
|
"epoch": 0.639470782800441,
|
|
"grad_norm": 0.6304661745572893,
|
|
"learning_rate": 3.650393700787402e-05,
|
|
"loss": 0.342,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1490386426448822,
|
|
"step": 1160,
|
|
"valid_targets_mean": 3251.0,
|
|
"valid_targets_min": 718
|
|
},
|
|
{
|
|
"epoch": 0.6422271223814774,
|
|
"grad_norm": 0.5114795282718989,
|
|
"learning_rate": 3.666141732283465e-05,
|
|
"loss": 0.3392,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14698046445846558,
|
|
"step": 1165,
|
|
"valid_targets_mean": 4718.9,
|
|
"valid_targets_min": 3205
|
|
},
|
|
{
|
|
"epoch": 0.6449834619625138,
|
|
"grad_norm": 0.728216751307702,
|
|
"learning_rate": 3.6818897637795276e-05,
|
|
"loss": 0.3307,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17887794971466064,
|
|
"step": 1170,
|
|
"valid_targets_mean": 3768.5,
|
|
"valid_targets_min": 1369
|
|
},
|
|
{
|
|
"epoch": 0.6477398015435502,
|
|
"grad_norm": 0.647102012641455,
|
|
"learning_rate": 3.697637795275591e-05,
|
|
"loss": 0.3169,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1897294670343399,
|
|
"step": 1175,
|
|
"valid_targets_mean": 4258.1,
|
|
"valid_targets_min": 2762
|
|
},
|
|
{
|
|
"epoch": 0.6504961411245865,
|
|
"grad_norm": 0.6026936345631803,
|
|
"learning_rate": 3.713385826771654e-05,
|
|
"loss": 0.3323,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12606516480445862,
|
|
"step": 1180,
|
|
"valid_targets_mean": 2535.5,
|
|
"valid_targets_min": 529
|
|
},
|
|
{
|
|
"epoch": 0.6532524807056229,
|
|
"grad_norm": 0.5876378190861556,
|
|
"learning_rate": 3.729133858267717e-05,
|
|
"loss": 0.3256,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16346250474452972,
|
|
"step": 1185,
|
|
"valid_targets_mean": 3895.6,
|
|
"valid_targets_min": 1832
|
|
},
|
|
{
|
|
"epoch": 0.6560088202866593,
|
|
"grad_norm": 0.6092078740490913,
|
|
"learning_rate": 3.7448818897637795e-05,
|
|
"loss": 0.3419,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1504027545452118,
|
|
"step": 1190,
|
|
"valid_targets_mean": 3470.2,
|
|
"valid_targets_min": 1726
|
|
},
|
|
{
|
|
"epoch": 0.6587651598676957,
|
|
"grad_norm": 0.7093029211924505,
|
|
"learning_rate": 3.760629921259843e-05,
|
|
"loss": 0.336,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19063466787338257,
|
|
"step": 1195,
|
|
"valid_targets_mean": 3707.2,
|
|
"valid_targets_min": 1361
|
|
},
|
|
{
|
|
"epoch": 0.6615214994487321,
|
|
"grad_norm": 0.5704547865659726,
|
|
"learning_rate": 3.776377952755906e-05,
|
|
"loss": 0.3306,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17092472314834595,
|
|
"step": 1200,
|
|
"valid_targets_mean": 3686.8,
|
|
"valid_targets_min": 851
|
|
},
|
|
{
|
|
"epoch": 0.6642778390297684,
|
|
"grad_norm": 0.7005594038455405,
|
|
"learning_rate": 3.7921259842519685e-05,
|
|
"loss": 0.3368,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16275273263454437,
|
|
"step": 1205,
|
|
"valid_targets_mean": 2817.6,
|
|
"valid_targets_min": 1190
|
|
},
|
|
{
|
|
"epoch": 0.6670341786108048,
|
|
"grad_norm": 0.5914026651518667,
|
|
"learning_rate": 3.807874015748032e-05,
|
|
"loss": 0.3309,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19426694512367249,
|
|
"step": 1210,
|
|
"valid_targets_mean": 4741.0,
|
|
"valid_targets_min": 1394
|
|
},
|
|
{
|
|
"epoch": 0.6697905181918412,
|
|
"grad_norm": 0.5668159468765804,
|
|
"learning_rate": 3.823622047244095e-05,
|
|
"loss": 0.3205,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12422381341457367,
|
|
"step": 1215,
|
|
"valid_targets_mean": 3478.5,
|
|
"valid_targets_min": 1972
|
|
},
|
|
{
|
|
"epoch": 0.6725468577728776,
|
|
"grad_norm": 0.588777323260611,
|
|
"learning_rate": 3.839370078740158e-05,
|
|
"loss": 0.3133,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17612385749816895,
|
|
"step": 1220,
|
|
"valid_targets_mean": 4379.2,
|
|
"valid_targets_min": 2550
|
|
},
|
|
{
|
|
"epoch": 0.675303197353914,
|
|
"grad_norm": 0.4809971589847626,
|
|
"learning_rate": 3.855118110236221e-05,
|
|
"loss": 0.3183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15179793536663055,
|
|
"step": 1225,
|
|
"valid_targets_mean": 6398.4,
|
|
"valid_targets_min": 2588
|
|
},
|
|
{
|
|
"epoch": 0.6780595369349504,
|
|
"grad_norm": 0.5651619183957602,
|
|
"learning_rate": 3.870866141732284e-05,
|
|
"loss": 0.3277,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17175175249576569,
|
|
"step": 1230,
|
|
"valid_targets_mean": 4008.5,
|
|
"valid_targets_min": 2867
|
|
},
|
|
{
|
|
"epoch": 0.6808158765159867,
|
|
"grad_norm": 0.6431771370207091,
|
|
"learning_rate": 3.886614173228347e-05,
|
|
"loss": 0.3153,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1401413530111313,
|
|
"step": 1235,
|
|
"valid_targets_mean": 2879.9,
|
|
"valid_targets_min": 992
|
|
},
|
|
{
|
|
"epoch": 0.6835722160970231,
|
|
"grad_norm": 0.6352702727505442,
|
|
"learning_rate": 3.90236220472441e-05,
|
|
"loss": 0.3186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15455858409404755,
|
|
"step": 1240,
|
|
"valid_targets_mean": 3798.8,
|
|
"valid_targets_min": 1573
|
|
},
|
|
{
|
|
"epoch": 0.6863285556780595,
|
|
"grad_norm": 0.6331316955077211,
|
|
"learning_rate": 3.918110236220473e-05,
|
|
"loss": 0.3235,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18990033864974976,
|
|
"step": 1245,
|
|
"valid_targets_mean": 3714.9,
|
|
"valid_targets_min": 1357
|
|
},
|
|
{
|
|
"epoch": 0.6890848952590959,
|
|
"grad_norm": 0.5682877716626864,
|
|
"learning_rate": 3.9338582677165356e-05,
|
|
"loss": 0.3171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15082508325576782,
|
|
"step": 1250,
|
|
"valid_targets_mean": 3661.1,
|
|
"valid_targets_min": 2409
|
|
},
|
|
{
|
|
"epoch": 0.6918412348401323,
|
|
"grad_norm": 0.5640542544247416,
|
|
"learning_rate": 3.949606299212599e-05,
|
|
"loss": 0.3438,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18801799416542053,
|
|
"step": 1255,
|
|
"valid_targets_mean": 4322.4,
|
|
"valid_targets_min": 1452
|
|
},
|
|
{
|
|
"epoch": 0.6945975744211687,
|
|
"grad_norm": 0.6035912496659972,
|
|
"learning_rate": 3.965354330708662e-05,
|
|
"loss": 0.3328,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17514300346374512,
|
|
"step": 1260,
|
|
"valid_targets_mean": 3613.9,
|
|
"valid_targets_min": 1708
|
|
},
|
|
{
|
|
"epoch": 0.697353914002205,
|
|
"grad_norm": 0.6816005697313289,
|
|
"learning_rate": 3.9811023622047246e-05,
|
|
"loss": 0.3325,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17177540063858032,
|
|
"step": 1265,
|
|
"valid_targets_mean": 3097.6,
|
|
"valid_targets_min": 1272
|
|
},
|
|
{
|
|
"epoch": 0.7001102535832414,
|
|
"grad_norm": 0.7388136697541232,
|
|
"learning_rate": 3.9968503937007874e-05,
|
|
"loss": 0.3149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1763586699962616,
|
|
"step": 1270,
|
|
"valid_targets_mean": 3143.1,
|
|
"valid_targets_min": 1858
|
|
},
|
|
{
|
|
"epoch": 0.7028665931642778,
|
|
"grad_norm": 0.6280268328542205,
|
|
"learning_rate": 3.9999987908526714e-05,
|
|
"loss": 0.3261,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12840086221694946,
|
|
"step": 1275,
|
|
"valid_targets_mean": 2977.6,
|
|
"valid_targets_min": 627
|
|
},
|
|
{
|
|
"epoch": 0.7056229327453142,
|
|
"grad_norm": 0.6657722420837537,
|
|
"learning_rate": 3.999993878694153e-05,
|
|
"loss": 0.3233,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1944282352924347,
|
|
"step": 1280,
|
|
"valid_targets_mean": 2598.8,
|
|
"valid_targets_min": 1036
|
|
},
|
|
{
|
|
"epoch": 0.7083792723263506,
|
|
"grad_norm": 0.573413902517194,
|
|
"learning_rate": 3.99998518796201e-05,
|
|
"loss": 0.3327,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1541140079498291,
|
|
"step": 1285,
|
|
"valid_targets_mean": 4311.2,
|
|
"valid_targets_min": 2524
|
|
},
|
|
{
|
|
"epoch": 0.7111356119073869,
|
|
"grad_norm": 0.6661500608027006,
|
|
"learning_rate": 3.9999727186726634e-05,
|
|
"loss": 0.3272,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.188973069190979,
|
|
"step": 1290,
|
|
"valid_targets_mean": 3706.4,
|
|
"valid_targets_min": 809
|
|
},
|
|
{
|
|
"epoch": 0.7138919514884233,
|
|
"grad_norm": 0.5889367365490404,
|
|
"learning_rate": 3.9999564708496694e-05,
|
|
"loss": 0.3246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1422090232372284,
|
|
"step": 1295,
|
|
"valid_targets_mean": 3445.4,
|
|
"valid_targets_min": 1778
|
|
},
|
|
{
|
|
"epoch": 0.7166482910694597,
|
|
"grad_norm": 0.5951283693828998,
|
|
"learning_rate": 3.9999364445237255e-05,
|
|
"loss": 0.3329,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18106815218925476,
|
|
"step": 1300,
|
|
"valid_targets_mean": 4350.0,
|
|
"valid_targets_min": 2181
|
|
},
|
|
{
|
|
"epoch": 0.7194046306504961,
|
|
"grad_norm": 0.7691318290376224,
|
|
"learning_rate": 3.999912639732668e-05,
|
|
"loss": 0.2913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08929453790187836,
|
|
"step": 1305,
|
|
"valid_targets_mean": 2416.5,
|
|
"valid_targets_min": 872
|
|
},
|
|
{
|
|
"epoch": 0.7221609702315325,
|
|
"grad_norm": 0.6546465037418128,
|
|
"learning_rate": 3.99988505652147e-05,
|
|
"loss": 0.2011,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07087012380361557,
|
|
"step": 1310,
|
|
"valid_targets_mean": 1543.9,
|
|
"valid_targets_min": 665
|
|
},
|
|
{
|
|
"epoch": 0.7249173098125689,
|
|
"grad_norm": 0.5031460848693546,
|
|
"learning_rate": 3.999853694942246e-05,
|
|
"loss": 0.1537,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07913365215063095,
|
|
"step": 1315,
|
|
"valid_targets_mean": 3129.8,
|
|
"valid_targets_min": 2254
|
|
},
|
|
{
|
|
"epoch": 0.7276736493936052,
|
|
"grad_norm": 0.5124059721612367,
|
|
"learning_rate": 3.999818555054245e-05,
|
|
"loss": 0.148,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07036252319812775,
|
|
"step": 1320,
|
|
"valid_targets_mean": 3026.8,
|
|
"valid_targets_min": 706
|
|
},
|
|
{
|
|
"epoch": 0.7304299889746417,
|
|
"grad_norm": 0.4653086654408342,
|
|
"learning_rate": 3.999779636923858e-05,
|
|
"loss": 0.1625,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04697272181510925,
|
|
"step": 1325,
|
|
"valid_targets_mean": 3207.8,
|
|
"valid_targets_min": 1216
|
|
},
|
|
{
|
|
"epoch": 0.7331863285556781,
|
|
"grad_norm": 0.5563131198294855,
|
|
"learning_rate": 3.9997369406246115e-05,
|
|
"loss": 0.1472,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08922342956066132,
|
|
"step": 1330,
|
|
"valid_targets_mean": 3540.0,
|
|
"valid_targets_min": 2551
|
|
},
|
|
{
|
|
"epoch": 0.7359426681367145,
|
|
"grad_norm": 0.859172636471933,
|
|
"learning_rate": 3.9996904662371736e-05,
|
|
"loss": 0.2028,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09161075204610825,
|
|
"step": 1335,
|
|
"valid_targets_mean": 960.9,
|
|
"valid_targets_min": 555
|
|
},
|
|
{
|
|
"epoch": 0.7386990077177509,
|
|
"grad_norm": 0.44400913147407683,
|
|
"learning_rate": 3.999640213849345e-05,
|
|
"loss": 0.1891,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06400476396083832,
|
|
"step": 1340,
|
|
"valid_targets_mean": 3465.4,
|
|
"valid_targets_min": 1008
|
|
},
|
|
{
|
|
"epoch": 0.7414553472987873,
|
|
"grad_norm": 0.5365293675457038,
|
|
"learning_rate": 3.99958618355607e-05,
|
|
"loss": 0.1605,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08003666996955872,
|
|
"step": 1345,
|
|
"valid_targets_mean": 3522.1,
|
|
"valid_targets_min": 3039
|
|
},
|
|
{
|
|
"epoch": 0.7442116868798236,
|
|
"grad_norm": 0.5896054962791752,
|
|
"learning_rate": 3.9995283754594256e-05,
|
|
"loss": 0.202,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11615628004074097,
|
|
"step": 1350,
|
|
"valid_targets_mean": 3356.1,
|
|
"valid_targets_min": 1938
|
|
},
|
|
{
|
|
"epoch": 0.74696802646086,
|
|
"grad_norm": 0.3582104720218796,
|
|
"learning_rate": 3.999466789668629e-05,
|
|
"loss": 0.1218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06885728985071182,
|
|
"step": 1355,
|
|
"valid_targets_mean": 3973.6,
|
|
"valid_targets_min": 924
|
|
},
|
|
{
|
|
"epoch": 0.7497243660418964,
|
|
"grad_norm": 0.5748775608151325,
|
|
"learning_rate": 3.999401426300034e-05,
|
|
"loss": 0.1423,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0702914297580719,
|
|
"step": 1360,
|
|
"valid_targets_mean": 2262.8,
|
|
"valid_targets_min": 444
|
|
},
|
|
{
|
|
"epoch": 0.7524807056229328,
|
|
"grad_norm": 0.4523812603360419,
|
|
"learning_rate": 3.99933228547713e-05,
|
|
"loss": 0.1385,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06659045815467834,
|
|
"step": 1365,
|
|
"valid_targets_mean": 3298.9,
|
|
"valid_targets_min": 749
|
|
},
|
|
{
|
|
"epoch": 0.7552370452039692,
|
|
"grad_norm": 0.6215784091433992,
|
|
"learning_rate": 3.999259367330546e-05,
|
|
"loss": 0.1701,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08409664779901505,
|
|
"step": 1370,
|
|
"valid_targets_mean": 1400.5,
|
|
"valid_targets_min": 601
|
|
},
|
|
{
|
|
"epoch": 0.7579933847850056,
|
|
"grad_norm": 0.49899515133899536,
|
|
"learning_rate": 3.999182671998044e-05,
|
|
"loss": 0.1592,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07465004920959473,
|
|
"step": 1375,
|
|
"valid_targets_mean": 3387.0,
|
|
"valid_targets_min": 1852
|
|
},
|
|
{
|
|
"epoch": 0.7607497243660419,
|
|
"grad_norm": 0.4838266361307885,
|
|
"learning_rate": 3.9991021996245256e-05,
|
|
"loss": 0.1432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07216821610927582,
|
|
"step": 1380,
|
|
"valid_targets_mean": 2827.2,
|
|
"valid_targets_min": 571
|
|
},
|
|
{
|
|
"epoch": 0.7635060639470783,
|
|
"grad_norm": 0.5507340257152413,
|
|
"learning_rate": 3.9990179503620255e-05,
|
|
"loss": 0.1368,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06556040048599243,
|
|
"step": 1385,
|
|
"valid_targets_mean": 2584.8,
|
|
"valid_targets_min": 520
|
|
},
|
|
{
|
|
"epoch": 0.7662624035281147,
|
|
"grad_norm": 0.6623761715000506,
|
|
"learning_rate": 3.9989299243697156e-05,
|
|
"loss": 0.1338,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09241770952939987,
|
|
"step": 1390,
|
|
"valid_targets_mean": 3477.0,
|
|
"valid_targets_min": 2179
|
|
},
|
|
{
|
|
"epoch": 0.7690187431091511,
|
|
"grad_norm": 0.49395148699056135,
|
|
"learning_rate": 3.9988381218139024e-05,
|
|
"loss": 0.1455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0599089041352272,
|
|
"step": 1395,
|
|
"valid_targets_mean": 3043.8,
|
|
"valid_targets_min": 807
|
|
},
|
|
{
|
|
"epoch": 0.7717750826901875,
|
|
"grad_norm": 0.48168129215053757,
|
|
"learning_rate": 3.998742542868028e-05,
|
|
"loss": 0.1541,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.062112756073474884,
|
|
"step": 1400,
|
|
"valid_targets_mean": 1830.4,
|
|
"valid_targets_min": 594
|
|
},
|
|
{
|
|
"epoch": 0.7745314222712238,
|
|
"grad_norm": 0.5453863250176623,
|
|
"learning_rate": 3.998643187712669e-05,
|
|
"loss": 0.1579,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.075777068734169,
|
|
"step": 1405,
|
|
"valid_targets_mean": 2210.0,
|
|
"valid_targets_min": 714
|
|
},
|
|
{
|
|
"epoch": 0.7772877618522602,
|
|
"grad_norm": 0.6363888408191681,
|
|
"learning_rate": 3.9985400565355365e-05,
|
|
"loss": 0.176,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1529623419046402,
|
|
"step": 1410,
|
|
"valid_targets_mean": 3163.6,
|
|
"valid_targets_min": 1290
|
|
},
|
|
{
|
|
"epoch": 0.7800441014332966,
|
|
"grad_norm": 0.4902322121658304,
|
|
"learning_rate": 3.998433149531475e-05,
|
|
"loss": 0.1631,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06318273395299911,
|
|
"step": 1415,
|
|
"valid_targets_mean": 2710.6,
|
|
"valid_targets_min": 782
|
|
},
|
|
{
|
|
"epoch": 0.782800441014333,
|
|
"grad_norm": 0.7250187732784953,
|
|
"learning_rate": 3.998322466902464e-05,
|
|
"loss": 0.1378,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10529972612857819,
|
|
"step": 1420,
|
|
"valid_targets_mean": 1954.5,
|
|
"valid_targets_min": 950
|
|
},
|
|
{
|
|
"epoch": 0.7855567805953694,
|
|
"grad_norm": 0.47129604593351004,
|
|
"learning_rate": 3.998208008857614e-05,
|
|
"loss": 0.1855,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07639536261558533,
|
|
"step": 1425,
|
|
"valid_targets_mean": 3065.0,
|
|
"valid_targets_min": 1296
|
|
},
|
|
{
|
|
"epoch": 0.7883131201764058,
|
|
"grad_norm": 0.49568756791840474,
|
|
"learning_rate": 3.9980897756131716e-05,
|
|
"loss": 0.1645,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.053294893354177475,
|
|
"step": 1430,
|
|
"valid_targets_mean": 3146.6,
|
|
"valid_targets_min": 789
|
|
},
|
|
{
|
|
"epoch": 0.7910694597574421,
|
|
"grad_norm": 0.6232404994532379,
|
|
"learning_rate": 3.9979677673925125e-05,
|
|
"loss": 0.1462,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08636526763439178,
|
|
"step": 1435,
|
|
"valid_targets_mean": 2340.4,
|
|
"valid_targets_min": 713
|
|
},
|
|
{
|
|
"epoch": 0.7938257993384785,
|
|
"grad_norm": 0.5985143072117078,
|
|
"learning_rate": 3.997841984426147e-05,
|
|
"loss": 0.3636,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15173931419849396,
|
|
"step": 1440,
|
|
"valid_targets_mean": 2638.0,
|
|
"valid_targets_min": 980
|
|
},
|
|
{
|
|
"epoch": 0.7965821389195149,
|
|
"grad_norm": 0.7172136727656446,
|
|
"learning_rate": 3.997712426951715e-05,
|
|
"loss": 0.1674,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1251123547554016,
|
|
"step": 1445,
|
|
"valid_targets_mean": 1792.1,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 0.7993384785005513,
|
|
"grad_norm": 0.5933310485610418,
|
|
"learning_rate": 3.997579095213988e-05,
|
|
"loss": 0.1455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059026338160037994,
|
|
"step": 1450,
|
|
"valid_targets_mean": 1459.5,
|
|
"valid_targets_min": 670
|
|
},
|
|
{
|
|
"epoch": 0.8020948180815877,
|
|
"grad_norm": 0.5541454152527887,
|
|
"learning_rate": 3.997441989464871e-05,
|
|
"loss": 0.2542,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08276881277561188,
|
|
"step": 1455,
|
|
"valid_targets_mean": 2546.0,
|
|
"valid_targets_min": 663
|
|
},
|
|
{
|
|
"epoch": 0.804851157662624,
|
|
"grad_norm": 0.30548261423906253,
|
|
"learning_rate": 3.997301109963393e-05,
|
|
"loss": 0.1249,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03921825438737869,
|
|
"step": 1460,
|
|
"valid_targets_mean": 5186.2,
|
|
"valid_targets_min": 3522
|
|
},
|
|
{
|
|
"epoch": 0.8076074972436604,
|
|
"grad_norm": 0.4628954539760422,
|
|
"learning_rate": 3.997156456975721e-05,
|
|
"loss": 0.1417,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0826883465051651,
|
|
"step": 1465,
|
|
"valid_targets_mean": 4087.4,
|
|
"valid_targets_min": 2327
|
|
},
|
|
{
|
|
"epoch": 0.8103638368246968,
|
|
"grad_norm": 0.42852274084816355,
|
|
"learning_rate": 3.997008030775144e-05,
|
|
"loss": 0.1399,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0741652101278305,
|
|
"step": 1470,
|
|
"valid_targets_mean": 3420.1,
|
|
"valid_targets_min": 1218
|
|
},
|
|
{
|
|
"epoch": 0.8131201764057332,
|
|
"grad_norm": 0.43232688498222793,
|
|
"learning_rate": 3.9968558316420834e-05,
|
|
"loss": 0.1326,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07875731587409973,
|
|
"step": 1475,
|
|
"valid_targets_mean": 4261.5,
|
|
"valid_targets_min": 3261
|
|
},
|
|
{
|
|
"epoch": 0.8158765159867696,
|
|
"grad_norm": 0.45169827687932035,
|
|
"learning_rate": 3.996699859864088e-05,
|
|
"loss": 0.1306,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06318311393260956,
|
|
"step": 1480,
|
|
"valid_targets_mean": 2142.2,
|
|
"valid_targets_min": 719
|
|
},
|
|
{
|
|
"epoch": 0.818632855567806,
|
|
"grad_norm": 0.4510827353474178,
|
|
"learning_rate": 3.996540115735833e-05,
|
|
"loss": 0.1291,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07042223960161209,
|
|
"step": 1485,
|
|
"valid_targets_mean": 3431.4,
|
|
"valid_targets_min": 2224
|
|
},
|
|
{
|
|
"epoch": 0.8213891951488423,
|
|
"grad_norm": 0.3724439203006909,
|
|
"learning_rate": 3.996376599559124e-05,
|
|
"loss": 0.123,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04451250284910202,
|
|
"step": 1490,
|
|
"valid_targets_mean": 3946.9,
|
|
"valid_targets_min": 3217
|
|
},
|
|
{
|
|
"epoch": 0.8241455347298787,
|
|
"grad_norm": 0.5013968759573806,
|
|
"learning_rate": 3.996209311642889e-05,
|
|
"loss": 0.1312,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06530219316482544,
|
|
"step": 1495,
|
|
"valid_targets_mean": 2442.1,
|
|
"valid_targets_min": 735
|
|
},
|
|
{
|
|
"epoch": 0.8269018743109151,
|
|
"grad_norm": 0.4719620167274468,
|
|
"learning_rate": 3.996038252303184e-05,
|
|
"loss": 0.1798,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.062212176620960236,
|
|
"step": 1500,
|
|
"valid_targets_mean": 1551.5,
|
|
"valid_targets_min": 603
|
|
},
|
|
{
|
|
"epoch": 0.8296582138919515,
|
|
"grad_norm": 0.5699569876623435,
|
|
"learning_rate": 3.995863421863191e-05,
|
|
"loss": 0.1103,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04505588486790657,
|
|
"step": 1505,
|
|
"valid_targets_mean": 1912.9,
|
|
"valid_targets_min": 582
|
|
},
|
|
{
|
|
"epoch": 0.8324145534729879,
|
|
"grad_norm": 0.485307068362443,
|
|
"learning_rate": 3.995684820653216e-05,
|
|
"loss": 0.1331,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07175744324922562,
|
|
"step": 1510,
|
|
"valid_targets_mean": 2670.5,
|
|
"valid_targets_min": 579
|
|
},
|
|
{
|
|
"epoch": 0.8351708930540243,
|
|
"grad_norm": 0.4109560643380955,
|
|
"learning_rate": 3.9955024490106875e-05,
|
|
"loss": 0.1172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07521124929189682,
|
|
"step": 1515,
|
|
"valid_targets_mean": 3709.5,
|
|
"valid_targets_min": 3248
|
|
},
|
|
{
|
|
"epoch": 0.8379272326350606,
|
|
"grad_norm": 0.6724232971653719,
|
|
"learning_rate": 3.9953163072801595e-05,
|
|
"loss": 0.1828,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08695890009403229,
|
|
"step": 1520,
|
|
"valid_targets_mean": 3222.9,
|
|
"valid_targets_min": 994
|
|
},
|
|
{
|
|
"epoch": 0.840683572216097,
|
|
"grad_norm": 0.5441851171734943,
|
|
"learning_rate": 3.995126395813309e-05,
|
|
"loss": 0.148,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10447517037391663,
|
|
"step": 1525,
|
|
"valid_targets_mean": 2046.2,
|
|
"valid_targets_min": 692
|
|
},
|
|
{
|
|
"epoch": 0.8434399117971334,
|
|
"grad_norm": 0.4321235182979236,
|
|
"learning_rate": 3.994932714968934e-05,
|
|
"loss": 0.1611,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07051797211170197,
|
|
"step": 1530,
|
|
"valid_targets_mean": 2707.2,
|
|
"valid_targets_min": 632
|
|
},
|
|
{
|
|
"epoch": 0.8461962513781698,
|
|
"grad_norm": 0.48898183739507495,
|
|
"learning_rate": 3.994735265112954e-05,
|
|
"loss": 0.142,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0575670450925827,
|
|
"step": 1535,
|
|
"valid_targets_mean": 1847.8,
|
|
"valid_targets_min": 812
|
|
},
|
|
{
|
|
"epoch": 0.8489525909592062,
|
|
"grad_norm": 0.5024632125478398,
|
|
"learning_rate": 3.9945340466184086e-05,
|
|
"loss": 0.2051,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20655810832977295,
|
|
"step": 1540,
|
|
"valid_targets_mean": 3254.0,
|
|
"valid_targets_min": 1177
|
|
},
|
|
{
|
|
"epoch": 0.8517089305402425,
|
|
"grad_norm": 0.44611263140796,
|
|
"learning_rate": 3.994329059865461e-05,
|
|
"loss": 0.1408,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06651468575000763,
|
|
"step": 1545,
|
|
"valid_targets_mean": 3402.8,
|
|
"valid_targets_min": 2331
|
|
},
|
|
{
|
|
"epoch": 0.8544652701212789,
|
|
"grad_norm": 0.6951183095947749,
|
|
"learning_rate": 3.994120305241389e-05,
|
|
"loss": 0.1705,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0899491161108017,
|
|
"step": 1550,
|
|
"valid_targets_mean": 1632.0,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 0.8572216097023153,
|
|
"grad_norm": 0.5454734340937385,
|
|
"learning_rate": 3.993907783140591e-05,
|
|
"loss": 0.134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04196039214730263,
|
|
"step": 1555,
|
|
"valid_targets_mean": 1552.2,
|
|
"valid_targets_min": 446
|
|
},
|
|
{
|
|
"epoch": 0.8599779492833517,
|
|
"grad_norm": 0.4371815445798495,
|
|
"learning_rate": 3.993691493964585e-05,
|
|
"loss": 0.1408,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043956734240055084,
|
|
"step": 1560,
|
|
"valid_targets_mean": 2990.9,
|
|
"valid_targets_min": 655
|
|
},
|
|
{
|
|
"epoch": 0.8627342888643881,
|
|
"grad_norm": 0.7364949321163858,
|
|
"learning_rate": 3.993471438122003e-05,
|
|
"loss": 0.1398,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06241379678249359,
|
|
"step": 1565,
|
|
"valid_targets_mean": 1293.9,
|
|
"valid_targets_min": 574
|
|
},
|
|
{
|
|
"epoch": 0.8654906284454245,
|
|
"grad_norm": 0.38380273671439724,
|
|
"learning_rate": 3.993247616028596e-05,
|
|
"loss": 0.1176,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06743113696575165,
|
|
"step": 1570,
|
|
"valid_targets_mean": 4529.1,
|
|
"valid_targets_min": 2872
|
|
},
|
|
{
|
|
"epoch": 0.8682469680264608,
|
|
"grad_norm": 0.36523247961234206,
|
|
"learning_rate": 3.993020028107229e-05,
|
|
"loss": 0.1141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0670817643404007,
|
|
"step": 1575,
|
|
"valid_targets_mean": 3615.2,
|
|
"valid_targets_min": 2242
|
|
},
|
|
{
|
|
"epoch": 0.8710033076074972,
|
|
"grad_norm": 0.4675771225645347,
|
|
"learning_rate": 3.9927886747878826e-05,
|
|
"loss": 0.127,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05731653422117233,
|
|
"step": 1580,
|
|
"valid_targets_mean": 2017.4,
|
|
"valid_targets_min": 651
|
|
},
|
|
{
|
|
"epoch": 0.8737596471885336,
|
|
"grad_norm": 0.42085126827400404,
|
|
"learning_rate": 3.99255355650765e-05,
|
|
"loss": 0.1448,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07115451991558075,
|
|
"step": 1585,
|
|
"valid_targets_mean": 3131.8,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 0.87651598676957,
|
|
"grad_norm": 0.452641951466591,
|
|
"learning_rate": 3.992314673710739e-05,
|
|
"loss": 0.142,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08104560524225235,
|
|
"step": 1590,
|
|
"valid_targets_mean": 3651.6,
|
|
"valid_targets_min": 3006
|
|
},
|
|
{
|
|
"epoch": 0.8792723263506064,
|
|
"grad_norm": 0.33781851102393595,
|
|
"learning_rate": 3.992072026848469e-05,
|
|
"loss": 0.1295,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04587774723768234,
|
|
"step": 1595,
|
|
"valid_targets_mean": 3210.8,
|
|
"valid_targets_min": 2353
|
|
},
|
|
{
|
|
"epoch": 0.8820286659316428,
|
|
"grad_norm": 0.3765972887952825,
|
|
"learning_rate": 3.991825616379271e-05,
|
|
"loss": 0.1212,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05509277433156967,
|
|
"step": 1600,
|
|
"valid_targets_mean": 1954.4,
|
|
"valid_targets_min": 665
|
|
},
|
|
{
|
|
"epoch": 0.8847850055126791,
|
|
"grad_norm": 0.45071063124502425,
|
|
"learning_rate": 3.9915754427686867e-05,
|
|
"loss": 0.149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08051882684230804,
|
|
"step": 1605,
|
|
"valid_targets_mean": 3512.6,
|
|
"valid_targets_min": 1776
|
|
},
|
|
{
|
|
"epoch": 0.8875413450937155,
|
|
"grad_norm": 0.281592724172698,
|
|
"learning_rate": 3.9913215064893664e-05,
|
|
"loss": 0.1105,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05110562592744827,
|
|
"step": 1610,
|
|
"valid_targets_mean": 4054.6,
|
|
"valid_targets_min": 3675
|
|
},
|
|
{
|
|
"epoch": 0.8902976846747519,
|
|
"grad_norm": 0.41045788797443794,
|
|
"learning_rate": 3.9910638080210714e-05,
|
|
"loss": 0.1375,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05886303633451462,
|
|
"step": 1615,
|
|
"valid_targets_mean": 3238.4,
|
|
"valid_targets_min": 591
|
|
},
|
|
{
|
|
"epoch": 0.8930540242557883,
|
|
"grad_norm": 0.4446571427125136,
|
|
"learning_rate": 3.990802347850668e-05,
|
|
"loss": 0.1407,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05325305834412575,
|
|
"step": 1620,
|
|
"valid_targets_mean": 2646.6,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 0.8958103638368247,
|
|
"grad_norm": 0.6301836244103415,
|
|
"learning_rate": 3.990537126472132e-05,
|
|
"loss": 0.1477,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12450476735830307,
|
|
"step": 1625,
|
|
"valid_targets_mean": 1379.1,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 0.898566703417861,
|
|
"grad_norm": 0.5889931182356986,
|
|
"learning_rate": 3.990268144386543e-05,
|
|
"loss": 0.1395,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07146436721086502,
|
|
"step": 1630,
|
|
"valid_targets_mean": 2145.4,
|
|
"valid_targets_min": 693
|
|
},
|
|
{
|
|
"epoch": 0.9013230429988974,
|
|
"grad_norm": 0.7218555878289816,
|
|
"learning_rate": 3.989995402102089e-05,
|
|
"loss": 0.1406,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09036028385162354,
|
|
"step": 1635,
|
|
"valid_targets_mean": 1268.9,
|
|
"valid_targets_min": 732
|
|
},
|
|
{
|
|
"epoch": 0.9040793825799338,
|
|
"grad_norm": 0.407745518634487,
|
|
"learning_rate": 3.989718900134058e-05,
|
|
"loss": 0.1317,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04444864019751549,
|
|
"step": 1640,
|
|
"valid_targets_mean": 2524.1,
|
|
"valid_targets_min": 799
|
|
},
|
|
{
|
|
"epoch": 0.9068357221609702,
|
|
"grad_norm": 0.4087268978933364,
|
|
"learning_rate": 3.9894386390048444e-05,
|
|
"loss": 0.1553,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04996391758322716,
|
|
"step": 1645,
|
|
"valid_targets_mean": 2966.2,
|
|
"valid_targets_min": 1012
|
|
},
|
|
{
|
|
"epoch": 0.9095920617420066,
|
|
"grad_norm": 0.3953371471334868,
|
|
"learning_rate": 3.989154619243943e-05,
|
|
"loss": 0.1304,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06219456344842911,
|
|
"step": 1650,
|
|
"valid_targets_mean": 3617.0,
|
|
"valid_targets_min": 1622
|
|
},
|
|
{
|
|
"epoch": 0.912348401323043,
|
|
"grad_norm": 0.4235865416244623,
|
|
"learning_rate": 3.988866841387949e-05,
|
|
"loss": 0.1182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06617286056280136,
|
|
"step": 1655,
|
|
"valid_targets_mean": 3110.0,
|
|
"valid_targets_min": 1001
|
|
},
|
|
{
|
|
"epoch": 0.9151047409040793,
|
|
"grad_norm": 0.4629060357826618,
|
|
"learning_rate": 3.988575305980561e-05,
|
|
"loss": 0.1435,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06073075160384178,
|
|
"step": 1660,
|
|
"valid_targets_mean": 2685.8,
|
|
"valid_targets_min": 897
|
|
},
|
|
{
|
|
"epoch": 0.9178610804851157,
|
|
"grad_norm": 0.3470966021012663,
|
|
"learning_rate": 3.988280013572573e-05,
|
|
"loss": 0.14,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056100573390722275,
|
|
"step": 1665,
|
|
"valid_targets_mean": 3777.6,
|
|
"valid_targets_min": 3040
|
|
},
|
|
{
|
|
"epoch": 0.9206174200661521,
|
|
"grad_norm": 0.5753004182560467,
|
|
"learning_rate": 3.9879809647218805e-05,
|
|
"loss": 0.1145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08498317748308182,
|
|
"step": 1670,
|
|
"valid_targets_mean": 2964.5,
|
|
"valid_targets_min": 561
|
|
},
|
|
{
|
|
"epoch": 0.9233737596471885,
|
|
"grad_norm": 0.3718025330418027,
|
|
"learning_rate": 3.9876781599934726e-05,
|
|
"loss": 0.1265,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07072757929563522,
|
|
"step": 1675,
|
|
"valid_targets_mean": 3642.1,
|
|
"valid_targets_min": 2542
|
|
},
|
|
{
|
|
"epoch": 0.9261300992282249,
|
|
"grad_norm": 0.3308619471796676,
|
|
"learning_rate": 3.9873715999594374e-05,
|
|
"loss": 0.1134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05305442214012146,
|
|
"step": 1680,
|
|
"valid_targets_mean": 4175.0,
|
|
"valid_targets_min": 2983
|
|
},
|
|
{
|
|
"epoch": 0.9288864388092613,
|
|
"grad_norm": 0.4067555326776259,
|
|
"learning_rate": 3.987061285198955e-05,
|
|
"loss": 0.1051,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04848351329565048,
|
|
"step": 1685,
|
|
"valid_targets_mean": 2350.6,
|
|
"valid_targets_min": 615
|
|
},
|
|
{
|
|
"epoch": 0.9316427783902976,
|
|
"grad_norm": 0.34018141423253334,
|
|
"learning_rate": 3.9867472162983024e-05,
|
|
"loss": 0.1342,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04568084329366684,
|
|
"step": 1690,
|
|
"valid_targets_mean": 3647.6,
|
|
"valid_targets_min": 2936
|
|
},
|
|
{
|
|
"epoch": 0.934399117971334,
|
|
"grad_norm": 0.36242431431802835,
|
|
"learning_rate": 3.9864293938508465e-05,
|
|
"loss": 0.1241,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09017286449670792,
|
|
"step": 1695,
|
|
"valid_targets_mean": 3372.6,
|
|
"valid_targets_min": 1532
|
|
},
|
|
{
|
|
"epoch": 0.9371554575523704,
|
|
"grad_norm": 0.40279817500278875,
|
|
"learning_rate": 3.986107818457048e-05,
|
|
"loss": 0.1451,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039335522800683975,
|
|
"step": 1700,
|
|
"valid_targets_mean": 2305.8,
|
|
"valid_targets_min": 506
|
|
},
|
|
{
|
|
"epoch": 0.9399117971334069,
|
|
"grad_norm": 0.33392411653894755,
|
|
"learning_rate": 3.985782490724455e-05,
|
|
"loss": 0.1322,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04187878221273422,
|
|
"step": 1705,
|
|
"valid_targets_mean": 2746.1,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 0.9426681367144433,
|
|
"grad_norm": 0.4322442346306531,
|
|
"learning_rate": 3.985453411267709e-05,
|
|
"loss": 0.1566,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06901313364505768,
|
|
"step": 1710,
|
|
"valid_targets_mean": 2745.8,
|
|
"valid_targets_min": 733
|
|
},
|
|
{
|
|
"epoch": 0.9454244762954797,
|
|
"grad_norm": 0.34484748093868567,
|
|
"learning_rate": 3.985120580708536e-05,
|
|
"loss": 0.114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057368770241737366,
|
|
"step": 1715,
|
|
"valid_targets_mean": 3070.9,
|
|
"valid_targets_min": 489
|
|
},
|
|
{
|
|
"epoch": 0.948180815876516,
|
|
"grad_norm": 0.5534773063668448,
|
|
"learning_rate": 3.98478399967575e-05,
|
|
"loss": 0.1409,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06249939650297165,
|
|
"step": 1720,
|
|
"valid_targets_mean": 1427.6,
|
|
"valid_targets_min": 693
|
|
},
|
|
{
|
|
"epoch": 0.9509371554575524,
|
|
"grad_norm": 0.3767944003374936,
|
|
"learning_rate": 3.9844436688052526e-05,
|
|
"loss": 0.121,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04625030979514122,
|
|
"step": 1725,
|
|
"valid_targets_mean": 3525.9,
|
|
"valid_targets_min": 918
|
|
},
|
|
{
|
|
"epoch": 0.9536934950385888,
|
|
"grad_norm": 0.679623013516448,
|
|
"learning_rate": 3.9840995887400274e-05,
|
|
"loss": 0.1772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0719040036201477,
|
|
"step": 1730,
|
|
"valid_targets_mean": 1479.6,
|
|
"valid_targets_min": 528
|
|
},
|
|
{
|
|
"epoch": 0.9564498346196252,
|
|
"grad_norm": 0.5064331103889584,
|
|
"learning_rate": 3.9837517601301424e-05,
|
|
"loss": 0.1661,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09957806766033173,
|
|
"step": 1735,
|
|
"valid_targets_mean": 2086.1,
|
|
"valid_targets_min": 572
|
|
},
|
|
{
|
|
"epoch": 0.9592061742006616,
|
|
"grad_norm": 0.4542511190639091,
|
|
"learning_rate": 3.983400183632747e-05,
|
|
"loss": 0.1269,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06869851052761078,
|
|
"step": 1740,
|
|
"valid_targets_mean": 3714.4,
|
|
"valid_targets_min": 2773
|
|
},
|
|
{
|
|
"epoch": 0.961962513781698,
|
|
"grad_norm": 0.3856971291796678,
|
|
"learning_rate": 3.983044859912073e-05,
|
|
"loss": 0.1252,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0590018704533577,
|
|
"step": 1745,
|
|
"valid_targets_mean": 3416.0,
|
|
"valid_targets_min": 2516
|
|
},
|
|
{
|
|
"epoch": 0.9647188533627343,
|
|
"grad_norm": 0.37189735746964914,
|
|
"learning_rate": 3.9826857896394306e-05,
|
|
"loss": 0.124,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04796933755278587,
|
|
"step": 1750,
|
|
"valid_targets_mean": 2494.9,
|
|
"valid_targets_min": 965
|
|
},
|
|
{
|
|
"epoch": 0.9674751929437707,
|
|
"grad_norm": 0.8567401444591685,
|
|
"learning_rate": 3.982322973493209e-05,
|
|
"loss": 0.1183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05484459549188614,
|
|
"step": 1755,
|
|
"valid_targets_mean": 1522.5,
|
|
"valid_targets_min": 725
|
|
},
|
|
{
|
|
"epoch": 0.9702315325248071,
|
|
"grad_norm": 0.7481339512192541,
|
|
"learning_rate": 3.9819564121588736e-05,
|
|
"loss": 0.1537,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10105057060718536,
|
|
"step": 1760,
|
|
"valid_targets_mean": 1261.0,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 0.9729878721058435,
|
|
"grad_norm": 0.3702279535716411,
|
|
"learning_rate": 3.981586106328966e-05,
|
|
"loss": 0.1123,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04988981783390045,
|
|
"step": 1765,
|
|
"valid_targets_mean": 3052.4,
|
|
"valid_targets_min": 889
|
|
},
|
|
{
|
|
"epoch": 0.9757442116868799,
|
|
"grad_norm": 0.34464284873957685,
|
|
"learning_rate": 3.981212056703102e-05,
|
|
"loss": 0.1183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05826319381594658,
|
|
"step": 1770,
|
|
"valid_targets_mean": 3543.8,
|
|
"valid_targets_min": 2396
|
|
},
|
|
{
|
|
"epoch": 0.9785005512679162,
|
|
"grad_norm": 0.3874834463251156,
|
|
"learning_rate": 3.980834263987971e-05,
|
|
"loss": 0.1237,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06502720713615417,
|
|
"step": 1775,
|
|
"valid_targets_mean": 3317.0,
|
|
"valid_targets_min": 981
|
|
},
|
|
{
|
|
"epoch": 0.9812568908489526,
|
|
"grad_norm": 0.3397133391497705,
|
|
"learning_rate": 3.980452728897334e-05,
|
|
"loss": 0.1319,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.058445245027542114,
|
|
"step": 1780,
|
|
"valid_targets_mean": 3465.1,
|
|
"valid_targets_min": 1071
|
|
},
|
|
{
|
|
"epoch": 0.984013230429989,
|
|
"grad_norm": 0.30705620104786974,
|
|
"learning_rate": 3.9800674521520234e-05,
|
|
"loss": 0.1094,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0629330649971962,
|
|
"step": 1785,
|
|
"valid_targets_mean": 4144.0,
|
|
"valid_targets_min": 757
|
|
},
|
|
{
|
|
"epoch": 0.9867695700110254,
|
|
"grad_norm": 0.38410819478609937,
|
|
"learning_rate": 3.9796784344799385e-05,
|
|
"loss": 0.12,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047083545476198196,
|
|
"step": 1790,
|
|
"valid_targets_mean": 1548.0,
|
|
"valid_targets_min": 504
|
|
},
|
|
{
|
|
"epoch": 0.9895259095920618,
|
|
"grad_norm": 0.5565304289747816,
|
|
"learning_rate": 3.979285676616048e-05,
|
|
"loss": 0.1388,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11265057325363159,
|
|
"step": 1795,
|
|
"valid_targets_mean": 2696.8,
|
|
"valid_targets_min": 775
|
|
},
|
|
{
|
|
"epoch": 0.9922822491730982,
|
|
"grad_norm": 0.4052992544151912,
|
|
"learning_rate": 3.978889179302385e-05,
|
|
"loss": 0.2097,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05838824808597565,
|
|
"step": 1800,
|
|
"valid_targets_mean": 3880.9,
|
|
"valid_targets_min": 2338
|
|
},
|
|
{
|
|
"epoch": 0.9950385887541345,
|
|
"grad_norm": 0.4224249322678252,
|
|
"learning_rate": 3.9784889432880513e-05,
|
|
"loss": 0.1206,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07626640796661377,
|
|
"step": 1805,
|
|
"valid_targets_mean": 3655.5,
|
|
"valid_targets_min": 2463
|
|
},
|
|
{
|
|
"epoch": 0.9977949283351709,
|
|
"grad_norm": 0.39536853771645086,
|
|
"learning_rate": 3.9780849693292094e-05,
|
|
"loss": 0.1191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05672920495271683,
|
|
"step": 1810,
|
|
"valid_targets_mean": 3319.2,
|
|
"valid_targets_min": 2237
|
|
},
|
|
{
|
|
"epoch": 1.0005512679162072,
|
|
"grad_norm": 0.5391699051741186,
|
|
"learning_rate": 3.977677258189082e-05,
|
|
"loss": 0.1928,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13977041840553284,
|
|
"step": 1815,
|
|
"valid_targets_mean": 8186.1,
|
|
"valid_targets_min": 6131
|
|
},
|
|
{
|
|
"epoch": 1.0033076074972436,
|
|
"grad_norm": 0.4429434546579974,
|
|
"learning_rate": 3.977265810637957e-05,
|
|
"loss": 0.2823,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14157527685165405,
|
|
"step": 1820,
|
|
"valid_targets_mean": 8564.9,
|
|
"valid_targets_min": 5974
|
|
},
|
|
{
|
|
"epoch": 1.00606394707828,
|
|
"grad_norm": 0.5328821993462085,
|
|
"learning_rate": 3.9768506274531795e-05,
|
|
"loss": 0.2759,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1231263130903244,
|
|
"step": 1825,
|
|
"valid_targets_mean": 3567.1,
|
|
"valid_targets_min": 287
|
|
},
|
|
{
|
|
"epoch": 1.0088202866593163,
|
|
"grad_norm": 0.40045178073964055,
|
|
"learning_rate": 3.97643170941915e-05,
|
|
"loss": 0.2642,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13034161925315857,
|
|
"step": 1830,
|
|
"valid_targets_mean": 6991.4,
|
|
"valid_targets_min": 4953
|
|
},
|
|
{
|
|
"epoch": 1.0115766262403527,
|
|
"grad_norm": 0.4038436496286876,
|
|
"learning_rate": 3.9760090573273286e-05,
|
|
"loss": 0.2425,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12636935710906982,
|
|
"step": 1835,
|
|
"valid_targets_mean": 6929.9,
|
|
"valid_targets_min": 1449
|
|
},
|
|
{
|
|
"epoch": 1.014332965821389,
|
|
"grad_norm": 0.3820745745461578,
|
|
"learning_rate": 3.975582671976229e-05,
|
|
"loss": 0.2492,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11822760105133057,
|
|
"step": 1840,
|
|
"valid_targets_mean": 7335.5,
|
|
"valid_targets_min": 4961
|
|
},
|
|
{
|
|
"epoch": 1.0170893054024255,
|
|
"grad_norm": 0.44481177360388846,
|
|
"learning_rate": 3.9751525541714164e-05,
|
|
"loss": 0.2649,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15675359964370728,
|
|
"step": 1845,
|
|
"valid_targets_mean": 7028.9,
|
|
"valid_targets_min": 5952
|
|
},
|
|
{
|
|
"epoch": 1.0198456449834619,
|
|
"grad_norm": 0.5731778258177014,
|
|
"learning_rate": 3.974718704725511e-05,
|
|
"loss": 0.2616,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14083576202392578,
|
|
"step": 1850,
|
|
"valid_targets_mean": 7634.2,
|
|
"valid_targets_min": 5143
|
|
},
|
|
{
|
|
"epoch": 1.0226019845644982,
|
|
"grad_norm": 0.38429349234475857,
|
|
"learning_rate": 3.9742811244581806e-05,
|
|
"loss": 0.2496,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12699735164642334,
|
|
"step": 1855,
|
|
"valid_targets_mean": 7079.8,
|
|
"valid_targets_min": 5316
|
|
},
|
|
{
|
|
"epoch": 1.0253583241455346,
|
|
"grad_norm": 0.4516162039984377,
|
|
"learning_rate": 3.973839814196142e-05,
|
|
"loss": 0.2467,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11604298651218414,
|
|
"step": 1860,
|
|
"valid_targets_mean": 4218.1,
|
|
"valid_targets_min": 151
|
|
},
|
|
{
|
|
"epoch": 1.028114663726571,
|
|
"grad_norm": 0.38841955535142336,
|
|
"learning_rate": 3.973394774773159e-05,
|
|
"loss": 0.2413,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1270502358675003,
|
|
"step": 1865,
|
|
"valid_targets_mean": 7171.6,
|
|
"valid_targets_min": 4244
|
|
},
|
|
{
|
|
"epoch": 1.0308710033076074,
|
|
"grad_norm": 0.4216035638713189,
|
|
"learning_rate": 3.972946007030043e-05,
|
|
"loss": 0.2435,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11914072185754776,
|
|
"step": 1870,
|
|
"valid_targets_mean": 7214.5,
|
|
"valid_targets_min": 5799
|
|
},
|
|
{
|
|
"epoch": 1.0336273428886438,
|
|
"grad_norm": 0.39495303374536844,
|
|
"learning_rate": 3.972493511814646e-05,
|
|
"loss": 0.2393,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11811669170856476,
|
|
"step": 1875,
|
|
"valid_targets_mean": 6923.1,
|
|
"valid_targets_min": 3844
|
|
},
|
|
{
|
|
"epoch": 1.0363836824696802,
|
|
"grad_norm": 0.4105671090945868,
|
|
"learning_rate": 3.972037289981865e-05,
|
|
"loss": 0.2371,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12813234329223633,
|
|
"step": 1880,
|
|
"valid_targets_mean": 6961.0,
|
|
"valid_targets_min": 5664
|
|
},
|
|
{
|
|
"epoch": 1.0391400220507165,
|
|
"grad_norm": 0.40074737118762155,
|
|
"learning_rate": 3.971577342393635e-05,
|
|
"loss": 0.2295,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12755727767944336,
|
|
"step": 1885,
|
|
"valid_targets_mean": 6933.6,
|
|
"valid_targets_min": 4915
|
|
},
|
|
{
|
|
"epoch": 1.041896361631753,
|
|
"grad_norm": 0.36855107336545717,
|
|
"learning_rate": 3.971113669918933e-05,
|
|
"loss": 0.2343,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.111843042075634,
|
|
"step": 1890,
|
|
"valid_targets_mean": 6407.2,
|
|
"valid_targets_min": 4966
|
|
},
|
|
{
|
|
"epoch": 1.0446527012127893,
|
|
"grad_norm": 0.38409204260266694,
|
|
"learning_rate": 3.9706462734337714e-05,
|
|
"loss": 0.2372,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11797580122947693,
|
|
"step": 1895,
|
|
"valid_targets_mean": 7489.0,
|
|
"valid_targets_min": 5451
|
|
},
|
|
{
|
|
"epoch": 1.0474090407938257,
|
|
"grad_norm": 0.3681827912396921,
|
|
"learning_rate": 3.970175153821199e-05,
|
|
"loss": 0.2288,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10866495966911316,
|
|
"step": 1900,
|
|
"valid_targets_mean": 7392.1,
|
|
"valid_targets_min": 5195
|
|
},
|
|
{
|
|
"epoch": 1.0501653803748623,
|
|
"grad_norm": 0.42383931143830844,
|
|
"learning_rate": 3.9697003119712994e-05,
|
|
"loss": 0.249,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15298643708229065,
|
|
"step": 1905,
|
|
"valid_targets_mean": 7050.0,
|
|
"valid_targets_min": 3733
|
|
},
|
|
{
|
|
"epoch": 1.0529217199558987,
|
|
"grad_norm": 0.44675277648990114,
|
|
"learning_rate": 3.969221748781186e-05,
|
|
"loss": 0.2602,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13354279100894928,
|
|
"step": 1910,
|
|
"valid_targets_mean": 6038.4,
|
|
"valid_targets_min": 5143
|
|
},
|
|
{
|
|
"epoch": 1.055678059536935,
|
|
"grad_norm": 0.3981813570788902,
|
|
"learning_rate": 3.9687394651550055e-05,
|
|
"loss": 0.2187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11283302307128906,
|
|
"step": 1915,
|
|
"valid_targets_mean": 7356.9,
|
|
"valid_targets_min": 5765
|
|
},
|
|
{
|
|
"epoch": 1.0584343991179714,
|
|
"grad_norm": 0.37541392508849375,
|
|
"learning_rate": 3.968253462003934e-05,
|
|
"loss": 0.2229,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1066504716873169,
|
|
"step": 1920,
|
|
"valid_targets_mean": 7668.2,
|
|
"valid_targets_min": 4640
|
|
},
|
|
{
|
|
"epoch": 1.0611907386990078,
|
|
"grad_norm": 0.6725515725973464,
|
|
"learning_rate": 3.967763740246172e-05,
|
|
"loss": 0.2294,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12626959383487701,
|
|
"step": 1925,
|
|
"valid_targets_mean": 7774.5,
|
|
"valid_targets_min": 5483
|
|
},
|
|
{
|
|
"epoch": 1.0639470782800442,
|
|
"grad_norm": 0.3762662737775176,
|
|
"learning_rate": 3.967270300806948e-05,
|
|
"loss": 0.2415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11693324148654938,
|
|
"step": 1930,
|
|
"valid_targets_mean": 7311.8,
|
|
"valid_targets_min": 6424
|
|
},
|
|
{
|
|
"epoch": 1.0667034178610806,
|
|
"grad_norm": 0.40058665657313214,
|
|
"learning_rate": 3.9667731446185126e-05,
|
|
"loss": 0.2318,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10364173352718353,
|
|
"step": 1935,
|
|
"valid_targets_mean": 6788.4,
|
|
"valid_targets_min": 4371
|
|
},
|
|
{
|
|
"epoch": 1.069459757442117,
|
|
"grad_norm": 0.38564487869391756,
|
|
"learning_rate": 3.966272272620141e-05,
|
|
"loss": 0.2312,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10268288850784302,
|
|
"step": 1940,
|
|
"valid_targets_mean": 7830.8,
|
|
"valid_targets_min": 5371
|
|
},
|
|
{
|
|
"epoch": 1.0722160970231533,
|
|
"grad_norm": 0.33541649336395335,
|
|
"learning_rate": 3.965767685758126e-05,
|
|
"loss": 0.202,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09285418689250946,
|
|
"step": 1945,
|
|
"valid_targets_mean": 8286.8,
|
|
"valid_targets_min": 4818
|
|
},
|
|
{
|
|
"epoch": 1.0749724366041897,
|
|
"grad_norm": 0.32906340798106043,
|
|
"learning_rate": 3.965259384985779e-05,
|
|
"loss": 0.2044,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08702705800533295,
|
|
"step": 1950,
|
|
"valid_targets_mean": 7339.1,
|
|
"valid_targets_min": 5995
|
|
},
|
|
{
|
|
"epoch": 1.0777287761852261,
|
|
"grad_norm": 0.40905337737414466,
|
|
"learning_rate": 3.96474737126343e-05,
|
|
"loss": 0.2139,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11576525121927261,
|
|
"step": 1955,
|
|
"valid_targets_mean": 6405.4,
|
|
"valid_targets_min": 4206
|
|
},
|
|
{
|
|
"epoch": 1.0804851157662625,
|
|
"grad_norm": 1.7247639593979691,
|
|
"learning_rate": 3.964231645558422e-05,
|
|
"loss": 0.2368,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09979398548603058,
|
|
"step": 1960,
|
|
"valid_targets_mean": 246.4,
|
|
"valid_targets_min": 136
|
|
},
|
|
{
|
|
"epoch": 1.0832414553472989,
|
|
"grad_norm": 0.4057176959401024,
|
|
"learning_rate": 3.9637122088451125e-05,
|
|
"loss": 0.241,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11096513271331787,
|
|
"step": 1965,
|
|
"valid_targets_mean": 6778.4,
|
|
"valid_targets_min": 5133
|
|
},
|
|
{
|
|
"epoch": 1.0859977949283353,
|
|
"grad_norm": 0.4788985312489239,
|
|
"learning_rate": 3.963189062104869e-05,
|
|
"loss": 0.2455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1283767819404602,
|
|
"step": 1970,
|
|
"valid_targets_mean": 6963.0,
|
|
"valid_targets_min": 5072
|
|
},
|
|
{
|
|
"epoch": 1.0887541345093716,
|
|
"grad_norm": 0.4044479444843954,
|
|
"learning_rate": 3.962662206326069e-05,
|
|
"loss": 0.2438,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13194270431995392,
|
|
"step": 1975,
|
|
"valid_targets_mean": 6662.5,
|
|
"valid_targets_min": 5036
|
|
},
|
|
{
|
|
"epoch": 1.091510474090408,
|
|
"grad_norm": 0.41227911767434583,
|
|
"learning_rate": 3.962131642504097e-05,
|
|
"loss": 0.2369,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14123860001564026,
|
|
"step": 1980,
|
|
"valid_targets_mean": 7389.0,
|
|
"valid_targets_min": 5546
|
|
},
|
|
{
|
|
"epoch": 1.0942668136714444,
|
|
"grad_norm": 0.3846369049442408,
|
|
"learning_rate": 3.9615973716413434e-05,
|
|
"loss": 0.226,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11498785018920898,
|
|
"step": 1985,
|
|
"valid_targets_mean": 7507.0,
|
|
"valid_targets_min": 5051
|
|
},
|
|
{
|
|
"epoch": 1.0970231532524808,
|
|
"grad_norm": 0.36727354006124485,
|
|
"learning_rate": 3.961059394747202e-05,
|
|
"loss": 0.2346,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1061241552233696,
|
|
"step": 1990,
|
|
"valid_targets_mean": 7192.6,
|
|
"valid_targets_min": 5809
|
|
},
|
|
{
|
|
"epoch": 1.0997794928335172,
|
|
"grad_norm": 0.43746035822859525,
|
|
"learning_rate": 3.9605177128380686e-05,
|
|
"loss": 0.2282,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11287423968315125,
|
|
"step": 1995,
|
|
"valid_targets_mean": 7255.8,
|
|
"valid_targets_min": 5119
|
|
},
|
|
{
|
|
"epoch": 1.1025358324145536,
|
|
"grad_norm": 0.38095977347770466,
|
|
"learning_rate": 3.9599723269373393e-05,
|
|
"loss": 0.2391,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1104314774274826,
|
|
"step": 2000,
|
|
"valid_targets_mean": 6710.9,
|
|
"valid_targets_min": 4029
|
|
},
|
|
{
|
|
"epoch": 1.10529217199559,
|
|
"grad_norm": 0.418337316793876,
|
|
"learning_rate": 3.959423238075407e-05,
|
|
"loss": 0.2296,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12884628772735596,
|
|
"step": 2005,
|
|
"valid_targets_mean": 6299.6,
|
|
"valid_targets_min": 5464
|
|
},
|
|
{
|
|
"epoch": 1.1080485115766263,
|
|
"grad_norm": 0.4253904059983968,
|
|
"learning_rate": 3.958870447289662e-05,
|
|
"loss": 0.2343,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11437705159187317,
|
|
"step": 2010,
|
|
"valid_targets_mean": 6239.8,
|
|
"valid_targets_min": 4663
|
|
},
|
|
{
|
|
"epoch": 1.1108048511576627,
|
|
"grad_norm": 0.4397397487968829,
|
|
"learning_rate": 3.958313955624487e-05,
|
|
"loss": 0.2349,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12122131139039993,
|
|
"step": 2015,
|
|
"valid_targets_mean": 5759.1,
|
|
"valid_targets_min": 4989
|
|
},
|
|
{
|
|
"epoch": 1.113561190738699,
|
|
"grad_norm": 0.5062657925191542,
|
|
"learning_rate": 3.957753764131258e-05,
|
|
"loss": 0.2263,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09378443658351898,
|
|
"step": 2020,
|
|
"valid_targets_mean": 3368.0,
|
|
"valid_targets_min": 2264
|
|
},
|
|
{
|
|
"epoch": 1.1163175303197355,
|
|
"grad_norm": 0.4400220455110204,
|
|
"learning_rate": 3.9571898738683394e-05,
|
|
"loss": 0.2277,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11328338831663132,
|
|
"step": 2025,
|
|
"valid_targets_mean": 6208.5,
|
|
"valid_targets_min": 5061
|
|
},
|
|
{
|
|
"epoch": 1.1190738699007718,
|
|
"grad_norm": 0.40398983565890384,
|
|
"learning_rate": 3.9566222859010875e-05,
|
|
"loss": 0.2274,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11828620731830597,
|
|
"step": 2030,
|
|
"valid_targets_mean": 6299.4,
|
|
"valid_targets_min": 4566
|
|
},
|
|
{
|
|
"epoch": 1.1218302094818082,
|
|
"grad_norm": 0.39478971283641073,
|
|
"learning_rate": 3.95605100130184e-05,
|
|
"loss": 0.2263,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12121611833572388,
|
|
"step": 2035,
|
|
"valid_targets_mean": 7560.9,
|
|
"valid_targets_min": 4931
|
|
},
|
|
{
|
|
"epoch": 1.1245865490628446,
|
|
"grad_norm": 0.4030780461018089,
|
|
"learning_rate": 3.9554760211499214e-05,
|
|
"loss": 0.227,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09764682501554489,
|
|
"step": 2040,
|
|
"valid_targets_mean": 5546.4,
|
|
"valid_targets_min": 4426
|
|
},
|
|
{
|
|
"epoch": 1.127342888643881,
|
|
"grad_norm": 0.4253177407177079,
|
|
"learning_rate": 3.9548973465316374e-05,
|
|
"loss": 0.2256,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10873885452747345,
|
|
"step": 2045,
|
|
"valid_targets_mean": 5400.6,
|
|
"valid_targets_min": 4401
|
|
},
|
|
{
|
|
"epoch": 1.1300992282249174,
|
|
"grad_norm": 0.4168690587947074,
|
|
"learning_rate": 3.9543149785402733e-05,
|
|
"loss": 0.2297,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11768406629562378,
|
|
"step": 2050,
|
|
"valid_targets_mean": 6330.9,
|
|
"valid_targets_min": 5178
|
|
},
|
|
{
|
|
"epoch": 1.1328555678059538,
|
|
"grad_norm": 0.38943320294407135,
|
|
"learning_rate": 3.9537289182760926e-05,
|
|
"loss": 0.2305,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10606523603200912,
|
|
"step": 2055,
|
|
"valid_targets_mean": 5793.9,
|
|
"valid_targets_min": 5097
|
|
},
|
|
{
|
|
"epoch": 1.1356119073869901,
|
|
"grad_norm": 0.825438699399564,
|
|
"learning_rate": 3.953139166846335e-05,
|
|
"loss": 0.2046,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16150851547718048,
|
|
"step": 2060,
|
|
"valid_targets_mean": 1627.8,
|
|
"valid_targets_min": 136
|
|
},
|
|
{
|
|
"epoch": 1.1383682469680265,
|
|
"grad_norm": 0.561224944484612,
|
|
"learning_rate": 3.952545725365215e-05,
|
|
"loss": 0.2255,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11346971988677979,
|
|
"step": 2065,
|
|
"valid_targets_mean": 6442.6,
|
|
"valid_targets_min": 4719
|
|
},
|
|
{
|
|
"epoch": 1.141124586549063,
|
|
"grad_norm": 0.39824062062272925,
|
|
"learning_rate": 3.951948594953914e-05,
|
|
"loss": 0.222,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12175564467906952,
|
|
"step": 2070,
|
|
"valid_targets_mean": 7150.1,
|
|
"valid_targets_min": 5326
|
|
},
|
|
{
|
|
"epoch": 1.1438809261300993,
|
|
"grad_norm": 0.4242053644112044,
|
|
"learning_rate": 3.951347776740588e-05,
|
|
"loss": 0.2285,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11423341929912567,
|
|
"step": 2075,
|
|
"valid_targets_mean": 7101.2,
|
|
"valid_targets_min": 4056
|
|
},
|
|
{
|
|
"epoch": 1.1466372657111357,
|
|
"grad_norm": 0.4289522277088913,
|
|
"learning_rate": 3.9507432718603584e-05,
|
|
"loss": 0.2201,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11060268431901932,
|
|
"step": 2080,
|
|
"valid_targets_mean": 6992.9,
|
|
"valid_targets_min": 4903
|
|
},
|
|
{
|
|
"epoch": 1.149393605292172,
|
|
"grad_norm": 0.40321655528621014,
|
|
"learning_rate": 3.9501350814553114e-05,
|
|
"loss": 0.2149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11475573480129242,
|
|
"step": 2085,
|
|
"valid_targets_mean": 5869.9,
|
|
"valid_targets_min": 4747
|
|
},
|
|
{
|
|
"epoch": 1.1521499448732084,
|
|
"grad_norm": 0.40175083435061454,
|
|
"learning_rate": 3.9495232066744965e-05,
|
|
"loss": 0.2252,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11527985334396362,
|
|
"step": 2090,
|
|
"valid_targets_mean": 6258.1,
|
|
"valid_targets_min": 4588
|
|
},
|
|
{
|
|
"epoch": 1.1549062844542448,
|
|
"grad_norm": 1.1990498220939767,
|
|
"learning_rate": 3.9489076486739245e-05,
|
|
"loss": 0.258,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10560241341590881,
|
|
"step": 2095,
|
|
"valid_targets_mean": 1159.1,
|
|
"valid_targets_min": 787
|
|
},
|
|
{
|
|
"epoch": 1.1576626240352812,
|
|
"grad_norm": 0.863837441062856,
|
|
"learning_rate": 3.948288408616565e-05,
|
|
"loss": 0.2553,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12062355875968933,
|
|
"step": 2100,
|
|
"valid_targets_mean": 1495.9,
|
|
"valid_targets_min": 956
|
|
},
|
|
{
|
|
"epoch": 1.1604189636163176,
|
|
"grad_norm": 0.8640986205233645,
|
|
"learning_rate": 3.947665487672342e-05,
|
|
"loss": 0.2404,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11295288801193237,
|
|
"step": 2105,
|
|
"valid_targets_mean": 1366.0,
|
|
"valid_targets_min": 795
|
|
},
|
|
{
|
|
"epoch": 1.163175303197354,
|
|
"grad_norm": 0.8951633283095337,
|
|
"learning_rate": 3.9470388870181376e-05,
|
|
"loss": 0.2399,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13539494574069977,
|
|
"step": 2110,
|
|
"valid_targets_mean": 1429.9,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 1.1659316427783903,
|
|
"grad_norm": 1.018554227871174,
|
|
"learning_rate": 3.9464086078377824e-05,
|
|
"loss": 0.2366,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11292294412851334,
|
|
"step": 2115,
|
|
"valid_targets_mean": 1574.1,
|
|
"valid_targets_min": 637
|
|
},
|
|
{
|
|
"epoch": 1.1686879823594267,
|
|
"grad_norm": 0.7244229799968689,
|
|
"learning_rate": 3.945774651322058e-05,
|
|
"loss": 0.2349,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11890330910682678,
|
|
"step": 2120,
|
|
"valid_targets_mean": 1737.8,
|
|
"valid_targets_min": 723
|
|
},
|
|
{
|
|
"epoch": 1.171444321940463,
|
|
"grad_norm": 0.8021547101345716,
|
|
"learning_rate": 3.945137018668695e-05,
|
|
"loss": 0.2421,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10122287273406982,
|
|
"step": 2125,
|
|
"valid_targets_mean": 1306.0,
|
|
"valid_targets_min": 760
|
|
},
|
|
{
|
|
"epoch": 1.1742006615214995,
|
|
"grad_norm": 0.8764497453102833,
|
|
"learning_rate": 3.9444957110823665e-05,
|
|
"loss": 0.2241,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1343308538198471,
|
|
"step": 2130,
|
|
"valid_targets_mean": 1363.9,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 1.1769570011025359,
|
|
"grad_norm": 0.7907792765403534,
|
|
"learning_rate": 3.943850729774692e-05,
|
|
"loss": 0.2432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1148378923535347,
|
|
"step": 2135,
|
|
"valid_targets_mean": 1488.0,
|
|
"valid_targets_min": 866
|
|
},
|
|
{
|
|
"epoch": 1.1797133406835723,
|
|
"grad_norm": 0.7243215532477045,
|
|
"learning_rate": 3.943202075964229e-05,
|
|
"loss": 0.2308,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12787649035453796,
|
|
"step": 2140,
|
|
"valid_targets_mean": 1986.4,
|
|
"valid_targets_min": 788
|
|
},
|
|
{
|
|
"epoch": 1.1824696802646086,
|
|
"grad_norm": 0.7877738992439743,
|
|
"learning_rate": 3.942549750876473e-05,
|
|
"loss": 0.2245,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11777092516422272,
|
|
"step": 2145,
|
|
"valid_targets_mean": 1518.4,
|
|
"valid_targets_min": 895
|
|
},
|
|
{
|
|
"epoch": 1.185226019845645,
|
|
"grad_norm": 0.7768965716910806,
|
|
"learning_rate": 3.941893755743859e-05,
|
|
"loss": 0.2374,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11837087571620941,
|
|
"step": 2150,
|
|
"valid_targets_mean": 1784.0,
|
|
"valid_targets_min": 1005
|
|
},
|
|
{
|
|
"epoch": 1.1879823594266814,
|
|
"grad_norm": 0.7926017237940828,
|
|
"learning_rate": 3.9412340918057525e-05,
|
|
"loss": 0.2269,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09464045614004135,
|
|
"step": 2155,
|
|
"valid_targets_mean": 1389.1,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 1.1907386990077178,
|
|
"grad_norm": 0.9261976823239991,
|
|
"learning_rate": 3.9405707603084525e-05,
|
|
"loss": 0.2163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08825404942035675,
|
|
"step": 2160,
|
|
"valid_targets_mean": 1125.1,
|
|
"valid_targets_min": 658
|
|
},
|
|
{
|
|
"epoch": 1.1934950385887542,
|
|
"grad_norm": 0.9566215356885488,
|
|
"learning_rate": 3.939903762505186e-05,
|
|
"loss": 0.2153,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10395936667919159,
|
|
"step": 2165,
|
|
"valid_targets_mean": 1460.8,
|
|
"valid_targets_min": 639
|
|
},
|
|
{
|
|
"epoch": 1.1962513781697905,
|
|
"grad_norm": 0.8347408288853052,
|
|
"learning_rate": 3.9392330996561064e-05,
|
|
"loss": 0.225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07904361933469772,
|
|
"step": 2170,
|
|
"valid_targets_mean": 1127.0,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 1.199007717750827,
|
|
"grad_norm": 0.850596486112364,
|
|
"learning_rate": 3.9385587730282936e-05,
|
|
"loss": 0.2347,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10767305642366409,
|
|
"step": 2175,
|
|
"valid_targets_mean": 1318.9,
|
|
"valid_targets_min": 629
|
|
},
|
|
{
|
|
"epoch": 1.2017640573318633,
|
|
"grad_norm": 0.8322703339122393,
|
|
"learning_rate": 3.937880783895746e-05,
|
|
"loss": 0.2344,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13187266886234283,
|
|
"step": 2180,
|
|
"valid_targets_mean": 1605.4,
|
|
"valid_targets_min": 622
|
|
},
|
|
{
|
|
"epoch": 1.2045203969128997,
|
|
"grad_norm": 0.7023489969446237,
|
|
"learning_rate": 3.9371991335393846e-05,
|
|
"loss": 0.2276,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08459800481796265,
|
|
"step": 2185,
|
|
"valid_targets_mean": 1396.2,
|
|
"valid_targets_min": 939
|
|
},
|
|
{
|
|
"epoch": 1.207276736493936,
|
|
"grad_norm": 0.7111782733516041,
|
|
"learning_rate": 3.936513823247047e-05,
|
|
"loss": 0.2272,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10252651572227478,
|
|
"step": 2190,
|
|
"valid_targets_mean": 1333.2,
|
|
"valid_targets_min": 775
|
|
},
|
|
{
|
|
"epoch": 1.2100330760749725,
|
|
"grad_norm": 1.0188517619372066,
|
|
"learning_rate": 3.935824854313483e-05,
|
|
"loss": 0.225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1316230446100235,
|
|
"step": 2195,
|
|
"valid_targets_mean": 1520.2,
|
|
"valid_targets_min": 620
|
|
},
|
|
{
|
|
"epoch": 1.2127894156560088,
|
|
"grad_norm": 1.2410525860213264,
|
|
"learning_rate": 3.935132228040357e-05,
|
|
"loss": 0.2298,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12410677969455719,
|
|
"step": 2200,
|
|
"valid_targets_mean": 1436.1,
|
|
"valid_targets_min": 690
|
|
},
|
|
{
|
|
"epoch": 1.2155457552370452,
|
|
"grad_norm": 0.8644256214436915,
|
|
"learning_rate": 3.934435945736244e-05,
|
|
"loss": 0.2254,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06392615288496017,
|
|
"step": 2205,
|
|
"valid_targets_mean": 846.4,
|
|
"valid_targets_min": 610
|
|
},
|
|
{
|
|
"epoch": 1.2183020948180816,
|
|
"grad_norm": 0.8438891418597277,
|
|
"learning_rate": 3.933736008716624e-05,
|
|
"loss": 0.2183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11855478584766388,
|
|
"step": 2210,
|
|
"valid_targets_mean": 1519.8,
|
|
"valid_targets_min": 792
|
|
},
|
|
{
|
|
"epoch": 1.221058434399118,
|
|
"grad_norm": 0.7211711670003869,
|
|
"learning_rate": 3.933032418303883e-05,
|
|
"loss": 0.2176,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11930051445960999,
|
|
"step": 2215,
|
|
"valid_targets_mean": 1771.8,
|
|
"valid_targets_min": 561
|
|
},
|
|
{
|
|
"epoch": 1.2238147739801544,
|
|
"grad_norm": 0.6978861320978671,
|
|
"learning_rate": 3.932325175827309e-05,
|
|
"loss": 0.2181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12333820015192032,
|
|
"step": 2220,
|
|
"valid_targets_mean": 1734.2,
|
|
"valid_targets_min": 1336
|
|
},
|
|
{
|
|
"epoch": 1.2265711135611908,
|
|
"grad_norm": 0.7934913073934879,
|
|
"learning_rate": 3.9316142826230905e-05,
|
|
"loss": 0.219,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13259224593639374,
|
|
"step": 2225,
|
|
"valid_targets_mean": 1868.2,
|
|
"valid_targets_min": 1103
|
|
},
|
|
{
|
|
"epoch": 1.2293274531422271,
|
|
"grad_norm": 0.791912543129638,
|
|
"learning_rate": 3.9308997400343114e-05,
|
|
"loss": 0.2281,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10461173951625824,
|
|
"step": 2230,
|
|
"valid_targets_mean": 1414.5,
|
|
"valid_targets_min": 677
|
|
},
|
|
{
|
|
"epoch": 1.2320837927232635,
|
|
"grad_norm": 0.7377503620364259,
|
|
"learning_rate": 3.930181549410954e-05,
|
|
"loss": 0.2122,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1498279571533203,
|
|
"step": 2235,
|
|
"valid_targets_mean": 1962.0,
|
|
"valid_targets_min": 978
|
|
},
|
|
{
|
|
"epoch": 1.2348401323043,
|
|
"grad_norm": 0.7349646336124053,
|
|
"learning_rate": 3.9294597121098884e-05,
|
|
"loss": 0.2171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10808219015598297,
|
|
"step": 2240,
|
|
"valid_targets_mean": 1562.5,
|
|
"valid_targets_min": 935
|
|
},
|
|
{
|
|
"epoch": 1.2375964718853363,
|
|
"grad_norm": 1.2150503361466123,
|
|
"learning_rate": 3.928734229494878e-05,
|
|
"loss": 0.2218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11851302534341812,
|
|
"step": 2245,
|
|
"valid_targets_mean": 1639.5,
|
|
"valid_targets_min": 803
|
|
},
|
|
{
|
|
"epoch": 1.2403528114663727,
|
|
"grad_norm": 0.7506887914389411,
|
|
"learning_rate": 3.928005102936571e-05,
|
|
"loss": 0.2234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10622420907020569,
|
|
"step": 2250,
|
|
"valid_targets_mean": 1568.8,
|
|
"valid_targets_min": 1015
|
|
},
|
|
{
|
|
"epoch": 1.243109151047409,
|
|
"grad_norm": 0.7952263161833848,
|
|
"learning_rate": 3.9272723338125e-05,
|
|
"loss": 0.2138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09869219362735748,
|
|
"step": 2255,
|
|
"valid_targets_mean": 1119.9,
|
|
"valid_targets_min": 881
|
|
},
|
|
{
|
|
"epoch": 1.2458654906284454,
|
|
"grad_norm": 0.7107939930905063,
|
|
"learning_rate": 3.926535923507081e-05,
|
|
"loss": 0.2108,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11727292835712433,
|
|
"step": 2260,
|
|
"valid_targets_mean": 1812.2,
|
|
"valid_targets_min": 1167
|
|
},
|
|
{
|
|
"epoch": 1.2486218302094818,
|
|
"grad_norm": 0.8257018456697217,
|
|
"learning_rate": 3.92579587341161e-05,
|
|
"loss": 0.2049,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10165152698755264,
|
|
"step": 2265,
|
|
"valid_targets_mean": 1413.4,
|
|
"valid_targets_min": 676
|
|
},
|
|
{
|
|
"epoch": 1.2513781697905182,
|
|
"grad_norm": 0.8626248822778579,
|
|
"learning_rate": 3.9250521849242555e-05,
|
|
"loss": 0.2195,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08355280011892319,
|
|
"step": 2270,
|
|
"valid_targets_mean": 1048.5,
|
|
"valid_targets_min": 675
|
|
},
|
|
{
|
|
"epoch": 1.2541345093715546,
|
|
"grad_norm": 0.723018143571444,
|
|
"learning_rate": 3.9243048594500647e-05,
|
|
"loss": 0.2149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09599076211452484,
|
|
"step": 2275,
|
|
"valid_targets_mean": 1424.6,
|
|
"valid_targets_min": 793
|
|
},
|
|
{
|
|
"epoch": 1.256890848952591,
|
|
"grad_norm": 0.7544125591670842,
|
|
"learning_rate": 3.923553898400953e-05,
|
|
"loss": 0.2157,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11298521608114243,
|
|
"step": 2280,
|
|
"valid_targets_mean": 1482.8,
|
|
"valid_targets_min": 741
|
|
},
|
|
{
|
|
"epoch": 1.2596471885336273,
|
|
"grad_norm": 0.7683246850469455,
|
|
"learning_rate": 3.922799303195707e-05,
|
|
"loss": 0.2174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1288297176361084,
|
|
"step": 2285,
|
|
"valid_targets_mean": 1727.5,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 1.2624035281146637,
|
|
"grad_norm": 0.924446561709709,
|
|
"learning_rate": 3.922041075259975e-05,
|
|
"loss": 0.2269,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1282481700181961,
|
|
"step": 2290,
|
|
"valid_targets_mean": 1706.5,
|
|
"valid_targets_min": 1154
|
|
},
|
|
{
|
|
"epoch": 1.2651598676957,
|
|
"grad_norm": 0.7295437113701345,
|
|
"learning_rate": 3.921279216026275e-05,
|
|
"loss": 0.2049,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14163079857826233,
|
|
"step": 2295,
|
|
"valid_targets_mean": 1717.0,
|
|
"valid_targets_min": 858
|
|
},
|
|
{
|
|
"epoch": 1.2679162072767365,
|
|
"grad_norm": 0.8698013467403112,
|
|
"learning_rate": 3.920513726933979e-05,
|
|
"loss": 0.2276,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1358645260334015,
|
|
"step": 2300,
|
|
"valid_targets_mean": 1628.0,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 1.2706725468577729,
|
|
"grad_norm": 0.793361775147602,
|
|
"learning_rate": 3.919744609429321e-05,
|
|
"loss": 0.2179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.102980837225914,
|
|
"step": 2305,
|
|
"valid_targets_mean": 1371.6,
|
|
"valid_targets_min": 1053
|
|
},
|
|
{
|
|
"epoch": 1.2734288864388092,
|
|
"grad_norm": 0.8623279632621945,
|
|
"learning_rate": 3.918971864965389e-05,
|
|
"loss": 0.2332,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11874578893184662,
|
|
"step": 2310,
|
|
"valid_targets_mean": 1571.4,
|
|
"valid_targets_min": 617
|
|
},
|
|
{
|
|
"epoch": 1.2761852260198456,
|
|
"grad_norm": 0.7344609964695565,
|
|
"learning_rate": 3.9181954950021236e-05,
|
|
"loss": 0.2087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12706609070301056,
|
|
"step": 2315,
|
|
"valid_targets_mean": 1611.0,
|
|
"valid_targets_min": 626
|
|
},
|
|
{
|
|
"epoch": 1.278941565600882,
|
|
"grad_norm": 0.6731590161382996,
|
|
"learning_rate": 3.917415501006315e-05,
|
|
"loss": 0.2156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11158636212348938,
|
|
"step": 2320,
|
|
"valid_targets_mean": 1875.5,
|
|
"valid_targets_min": 1166
|
|
},
|
|
{
|
|
"epoch": 1.2816979051819184,
|
|
"grad_norm": 0.7734608989746682,
|
|
"learning_rate": 3.9166318844516e-05,
|
|
"loss": 0.2182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1138259544968605,
|
|
"step": 2325,
|
|
"valid_targets_mean": 1362.9,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 1.2844542447629548,
|
|
"grad_norm": 0.8564823564775201,
|
|
"learning_rate": 3.915844646818459e-05,
|
|
"loss": 0.2056,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08460142463445663,
|
|
"step": 2330,
|
|
"valid_targets_mean": 1171.2,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 1.2872105843439912,
|
|
"grad_norm": 0.7327706705093682,
|
|
"learning_rate": 3.915053789594214e-05,
|
|
"loss": 0.21,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10057958960533142,
|
|
"step": 2335,
|
|
"valid_targets_mean": 1548.8,
|
|
"valid_targets_min": 502
|
|
},
|
|
{
|
|
"epoch": 1.2899669239250275,
|
|
"grad_norm": 0.6860219789847906,
|
|
"learning_rate": 3.914259314273027e-05,
|
|
"loss": 0.2152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12251529842615128,
|
|
"step": 2340,
|
|
"valid_targets_mean": 1877.4,
|
|
"valid_targets_min": 1330
|
|
},
|
|
{
|
|
"epoch": 1.292723263506064,
|
|
"grad_norm": 0.6975759342744783,
|
|
"learning_rate": 3.913461222355893e-05,
|
|
"loss": 0.2152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07897192239761353,
|
|
"step": 2345,
|
|
"valid_targets_mean": 1301.9,
|
|
"valid_targets_min": 659
|
|
},
|
|
{
|
|
"epoch": 1.2954796030871003,
|
|
"grad_norm": 0.764417356058545,
|
|
"learning_rate": 3.9126595153506424e-05,
|
|
"loss": 0.219,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08148185908794403,
|
|
"step": 2350,
|
|
"valid_targets_mean": 1325.8,
|
|
"valid_targets_min": 955
|
|
},
|
|
{
|
|
"epoch": 1.2982359426681367,
|
|
"grad_norm": 0.6984240531107728,
|
|
"learning_rate": 3.9118541947719334e-05,
|
|
"loss": 0.2073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08307754993438721,
|
|
"step": 2355,
|
|
"valid_targets_mean": 1461.5,
|
|
"valid_targets_min": 962
|
|
},
|
|
{
|
|
"epoch": 1.300992282249173,
|
|
"grad_norm": 0.7816061851789948,
|
|
"learning_rate": 3.911045262141252e-05,
|
|
"loss": 0.2143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10532218962907791,
|
|
"step": 2360,
|
|
"valid_targets_mean": 1423.1,
|
|
"valid_targets_min": 792
|
|
},
|
|
{
|
|
"epoch": 1.3037486218302095,
|
|
"grad_norm": 0.7519708343218594,
|
|
"learning_rate": 3.910232718986909e-05,
|
|
"loss": 0.2138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13661381602287292,
|
|
"step": 2365,
|
|
"valid_targets_mean": 1629.1,
|
|
"valid_targets_min": 916
|
|
},
|
|
{
|
|
"epoch": 1.3065049614112458,
|
|
"grad_norm": 0.7553857960253325,
|
|
"learning_rate": 3.909416566844036e-05,
|
|
"loss": 0.2075,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08479543030261993,
|
|
"step": 2370,
|
|
"valid_targets_mean": 1243.4,
|
|
"valid_targets_min": 871
|
|
},
|
|
{
|
|
"epoch": 1.3092613009922822,
|
|
"grad_norm": 0.7096032332146929,
|
|
"learning_rate": 3.908596807254585e-05,
|
|
"loss": 0.2152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11088590323925018,
|
|
"step": 2375,
|
|
"valid_targets_mean": 1896.4,
|
|
"valid_targets_min": 1216
|
|
},
|
|
{
|
|
"epoch": 1.3120176405733186,
|
|
"grad_norm": 0.8322490628353374,
|
|
"learning_rate": 3.9077734417673186e-05,
|
|
"loss": 0.1935,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10097792744636536,
|
|
"step": 2380,
|
|
"valid_targets_mean": 1387.0,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 1.314773980154355,
|
|
"grad_norm": 0.8122940429267342,
|
|
"learning_rate": 3.906946471937817e-05,
|
|
"loss": 0.2077,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11288172006607056,
|
|
"step": 2385,
|
|
"valid_targets_mean": 1290.0,
|
|
"valid_targets_min": 602
|
|
},
|
|
{
|
|
"epoch": 1.3175303197353914,
|
|
"grad_norm": 0.6848779285989502,
|
|
"learning_rate": 3.906115899328468e-05,
|
|
"loss": 0.2129,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06991253793239594,
|
|
"step": 2390,
|
|
"valid_targets_mean": 1410.0,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 1.3202866593164277,
|
|
"grad_norm": 0.7903401879274046,
|
|
"learning_rate": 3.905281725508466e-05,
|
|
"loss": 0.2005,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09129732847213745,
|
|
"step": 2395,
|
|
"valid_targets_mean": 1057.5,
|
|
"valid_targets_min": 722
|
|
},
|
|
{
|
|
"epoch": 1.3230429988974641,
|
|
"grad_norm": 0.7457498612807043,
|
|
"learning_rate": 3.90444395205381e-05,
|
|
"loss": 0.2135,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11177748441696167,
|
|
"step": 2400,
|
|
"valid_targets_mean": 1625.4,
|
|
"valid_targets_min": 1323
|
|
},
|
|
{
|
|
"epoch": 1.3257993384785005,
|
|
"grad_norm": 0.693098984608495,
|
|
"learning_rate": 3.9036025805472976e-05,
|
|
"loss": 0.2085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09320472180843353,
|
|
"step": 2405,
|
|
"valid_targets_mean": 1408.6,
|
|
"valid_targets_min": 901
|
|
},
|
|
{
|
|
"epoch": 1.328555678059537,
|
|
"grad_norm": 0.7262117094237536,
|
|
"learning_rate": 3.902757612578528e-05,
|
|
"loss": 0.2143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09907413274049759,
|
|
"step": 2410,
|
|
"valid_targets_mean": 1553.9,
|
|
"valid_targets_min": 802
|
|
},
|
|
{
|
|
"epoch": 1.3313120176405733,
|
|
"grad_norm": 0.7335497317167226,
|
|
"learning_rate": 3.9019090497438916e-05,
|
|
"loss": 0.2118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1077362596988678,
|
|
"step": 2415,
|
|
"valid_targets_mean": 1436.0,
|
|
"valid_targets_min": 1078
|
|
},
|
|
{
|
|
"epoch": 1.3340683572216097,
|
|
"grad_norm": 0.7135564086396033,
|
|
"learning_rate": 3.901056893646572e-05,
|
|
"loss": 0.2079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10013362020254135,
|
|
"step": 2420,
|
|
"valid_targets_mean": 1443.0,
|
|
"valid_targets_min": 1258
|
|
},
|
|
{
|
|
"epoch": 1.336824696802646,
|
|
"grad_norm": 0.7855791194741032,
|
|
"learning_rate": 3.900201145896541e-05,
|
|
"loss": 0.2194,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10709736496210098,
|
|
"step": 2425,
|
|
"valid_targets_mean": 1482.1,
|
|
"valid_targets_min": 843
|
|
},
|
|
{
|
|
"epoch": 1.3395810363836824,
|
|
"grad_norm": 0.7580670007310235,
|
|
"learning_rate": 3.899341808110558e-05,
|
|
"loss": 0.2151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09351199865341187,
|
|
"step": 2430,
|
|
"valid_targets_mean": 1415.9,
|
|
"valid_targets_min": 666
|
|
},
|
|
{
|
|
"epoch": 1.3423373759647188,
|
|
"grad_norm": 0.7282167214807848,
|
|
"learning_rate": 3.898478881912162e-05,
|
|
"loss": 0.2187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12833508849143982,
|
|
"step": 2435,
|
|
"valid_targets_mean": 1815.6,
|
|
"valid_targets_min": 1039
|
|
},
|
|
{
|
|
"epoch": 1.3450937155457552,
|
|
"grad_norm": 0.7531562555670924,
|
|
"learning_rate": 3.897612368931674e-05,
|
|
"loss": 0.2151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09941570460796356,
|
|
"step": 2440,
|
|
"valid_targets_mean": 1323.5,
|
|
"valid_targets_min": 775
|
|
},
|
|
{
|
|
"epoch": 1.3478500551267916,
|
|
"grad_norm": 0.725650973636521,
|
|
"learning_rate": 3.896742270806189e-05,
|
|
"loss": 0.2152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10683807730674744,
|
|
"step": 2445,
|
|
"valid_targets_mean": 1817.0,
|
|
"valid_targets_min": 1280
|
|
},
|
|
{
|
|
"epoch": 1.350606394707828,
|
|
"grad_norm": 0.7893252076163679,
|
|
"learning_rate": 3.895868589179579e-05,
|
|
"loss": 0.2031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07101763784885406,
|
|
"step": 2450,
|
|
"valid_targets_mean": 922.0,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 1.3533627342888643,
|
|
"grad_norm": 0.6756139981603474,
|
|
"learning_rate": 3.894991325702483e-05,
|
|
"loss": 0.2125,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09092555195093155,
|
|
"step": 2455,
|
|
"valid_targets_mean": 1475.8,
|
|
"valid_targets_min": 655
|
|
},
|
|
{
|
|
"epoch": 1.3561190738699007,
|
|
"grad_norm": 0.6811041334419775,
|
|
"learning_rate": 3.894110482032309e-05,
|
|
"loss": 0.2118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08369474112987518,
|
|
"step": 2460,
|
|
"valid_targets_mean": 1410.9,
|
|
"valid_targets_min": 694
|
|
},
|
|
{
|
|
"epoch": 1.358875413450937,
|
|
"grad_norm": 0.6984323058419214,
|
|
"learning_rate": 3.893226059833227e-05,
|
|
"loss": 0.2067,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10465934872627258,
|
|
"step": 2465,
|
|
"valid_targets_mean": 1651.0,
|
|
"valid_targets_min": 814
|
|
},
|
|
{
|
|
"epoch": 1.3616317530319735,
|
|
"grad_norm": 0.6787259317223129,
|
|
"learning_rate": 3.892338060776172e-05,
|
|
"loss": 0.2036,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0964624434709549,
|
|
"step": 2470,
|
|
"valid_targets_mean": 1547.8,
|
|
"valid_targets_min": 849
|
|
},
|
|
{
|
|
"epoch": 1.3643880926130099,
|
|
"grad_norm": 0.748917016671887,
|
|
"learning_rate": 3.891446486538831e-05,
|
|
"loss": 0.1997,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10137563198804855,
|
|
"step": 2475,
|
|
"valid_targets_mean": 1349.2,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 1.3671444321940462,
|
|
"grad_norm": 0.7247570704449319,
|
|
"learning_rate": 3.890551338805651e-05,
|
|
"loss": 0.2188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09162464737892151,
|
|
"step": 2480,
|
|
"valid_targets_mean": 1415.6,
|
|
"valid_targets_min": 929
|
|
},
|
|
{
|
|
"epoch": 1.3699007717750826,
|
|
"grad_norm": 0.731799059895408,
|
|
"learning_rate": 3.8896526192678265e-05,
|
|
"loss": 0.2075,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09941557049751282,
|
|
"step": 2485,
|
|
"valid_targets_mean": 1389.9,
|
|
"valid_targets_min": 826
|
|
},
|
|
{
|
|
"epoch": 1.372657111356119,
|
|
"grad_norm": 0.7274715144402939,
|
|
"learning_rate": 3.8887503296233034e-05,
|
|
"loss": 0.1996,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09162037074565887,
|
|
"step": 2490,
|
|
"valid_targets_mean": 1378.1,
|
|
"valid_targets_min": 908
|
|
},
|
|
{
|
|
"epoch": 1.3754134509371554,
|
|
"grad_norm": 0.690109544983653,
|
|
"learning_rate": 3.8878444715767686e-05,
|
|
"loss": 0.1991,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1137581318616867,
|
|
"step": 2495,
|
|
"valid_targets_mean": 1676.0,
|
|
"valid_targets_min": 1090
|
|
},
|
|
{
|
|
"epoch": 1.3781697905181918,
|
|
"grad_norm": 0.7896862906404408,
|
|
"learning_rate": 3.886935046839654e-05,
|
|
"loss": 0.1918,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09316205978393555,
|
|
"step": 2500,
|
|
"valid_targets_mean": 1115.2,
|
|
"valid_targets_min": 719
|
|
},
|
|
{
|
|
"epoch": 1.3809261300992282,
|
|
"grad_norm": 0.8504997157166785,
|
|
"learning_rate": 3.886022057130129e-05,
|
|
"loss": 0.2086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10884758830070496,
|
|
"step": 2505,
|
|
"valid_targets_mean": 1533.0,
|
|
"valid_targets_min": 609
|
|
},
|
|
{
|
|
"epoch": 1.3836824696802645,
|
|
"grad_norm": 0.7497043146670845,
|
|
"learning_rate": 3.885105504173099e-05,
|
|
"loss": 0.2029,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11243531107902527,
|
|
"step": 2510,
|
|
"valid_targets_mean": 1564.6,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 1.386438809261301,
|
|
"grad_norm": 0.6715401001809181,
|
|
"learning_rate": 3.8841853897002e-05,
|
|
"loss": 0.2125,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1002802774310112,
|
|
"step": 2515,
|
|
"valid_targets_mean": 1533.1,
|
|
"valid_targets_min": 873
|
|
},
|
|
{
|
|
"epoch": 1.3891951488423373,
|
|
"grad_norm": 0.6652500622489002,
|
|
"learning_rate": 3.8832617154497974e-05,
|
|
"loss": 0.2031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10130589455366135,
|
|
"step": 2520,
|
|
"valid_targets_mean": 1465.5,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 1.3919514884233737,
|
|
"grad_norm": 0.6793761321944495,
|
|
"learning_rate": 3.882334483166983e-05,
|
|
"loss": 0.2067,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08963647484779358,
|
|
"step": 2525,
|
|
"valid_targets_mean": 1249.2,
|
|
"valid_targets_min": 815
|
|
},
|
|
{
|
|
"epoch": 1.39470782800441,
|
|
"grad_norm": 1.004313643262738,
|
|
"learning_rate": 3.881403694603567e-05,
|
|
"loss": 0.2121,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10261203348636627,
|
|
"step": 2530,
|
|
"valid_targets_mean": 1477.4,
|
|
"valid_targets_min": 1179
|
|
},
|
|
{
|
|
"epoch": 1.3974641675854464,
|
|
"grad_norm": 0.5247292522599006,
|
|
"learning_rate": 3.880469351518086e-05,
|
|
"loss": 0.3036,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14284640550613403,
|
|
"step": 2535,
|
|
"valid_targets_mean": 4555.2,
|
|
"valid_targets_min": 700
|
|
},
|
|
{
|
|
"epoch": 1.4002205071664828,
|
|
"grad_norm": 0.5667743900921924,
|
|
"learning_rate": 3.8795314556757835e-05,
|
|
"loss": 0.3277,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18902640044689178,
|
|
"step": 2540,
|
|
"valid_targets_mean": 5515.5,
|
|
"valid_targets_min": 1004
|
|
},
|
|
{
|
|
"epoch": 1.4029768467475192,
|
|
"grad_norm": 0.46315243516155186,
|
|
"learning_rate": 3.878590008848621e-05,
|
|
"loss": 0.3143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12766319513320923,
|
|
"step": 2545,
|
|
"valid_targets_mean": 4310.8,
|
|
"valid_targets_min": 2049
|
|
},
|
|
{
|
|
"epoch": 1.4057331863285556,
|
|
"grad_norm": 0.5086101496882789,
|
|
"learning_rate": 3.877645012815267e-05,
|
|
"loss": 0.3138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12383601069450378,
|
|
"step": 2550,
|
|
"valid_targets_mean": 3495.1,
|
|
"valid_targets_min": 1559
|
|
},
|
|
{
|
|
"epoch": 1.4084895259095922,
|
|
"grad_norm": 0.5200050138200063,
|
|
"learning_rate": 3.8766964693610945e-05,
|
|
"loss": 0.3099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19253048300743103,
|
|
"step": 2555,
|
|
"valid_targets_mean": 5011.6,
|
|
"valid_targets_min": 1969
|
|
},
|
|
{
|
|
"epoch": 1.4112458654906286,
|
|
"grad_norm": 0.4605329678645425,
|
|
"learning_rate": 3.875744380278181e-05,
|
|
"loss": 0.3204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15327517688274384,
|
|
"step": 2560,
|
|
"valid_targets_mean": 6144.0,
|
|
"valid_targets_min": 1649
|
|
},
|
|
{
|
|
"epoch": 1.414002205071665,
|
|
"grad_norm": 0.42357493154224807,
|
|
"learning_rate": 3.8747887473653004e-05,
|
|
"loss": 0.3012,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16235151886940002,
|
|
"step": 2565,
|
|
"valid_targets_mean": 7660.9,
|
|
"valid_targets_min": 2225
|
|
},
|
|
{
|
|
"epoch": 1.4167585446527013,
|
|
"grad_norm": 0.510522270867259,
|
|
"learning_rate": 3.8738295724279226e-05,
|
|
"loss": 0.3092,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10453785955905914,
|
|
"step": 2570,
|
|
"valid_targets_mean": 3117.5,
|
|
"valid_targets_min": 466
|
|
},
|
|
{
|
|
"epoch": 1.4195148842337377,
|
|
"grad_norm": 0.5546393121261465,
|
|
"learning_rate": 3.8728668572782103e-05,
|
|
"loss": 0.3145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17512211203575134,
|
|
"step": 2575,
|
|
"valid_targets_mean": 6088.8,
|
|
"valid_targets_min": 1747
|
|
},
|
|
{
|
|
"epoch": 1.422271223814774,
|
|
"grad_norm": 0.5839502780630056,
|
|
"learning_rate": 3.871900603735015e-05,
|
|
"loss": 0.3151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14682859182357788,
|
|
"step": 2580,
|
|
"valid_targets_mean": 4500.0,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 1.4250275633958105,
|
|
"grad_norm": 0.5864374887853592,
|
|
"learning_rate": 3.870930813623871e-05,
|
|
"loss": 0.3117,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16442151367664337,
|
|
"step": 2585,
|
|
"valid_targets_mean": 3932.4,
|
|
"valid_targets_min": 1221
|
|
},
|
|
{
|
|
"epoch": 1.4277839029768469,
|
|
"grad_norm": 0.47940524059941014,
|
|
"learning_rate": 3.8699574887769953e-05,
|
|
"loss": 0.3079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1591748148202896,
|
|
"step": 2590,
|
|
"valid_targets_mean": 5551.0,
|
|
"valid_targets_min": 1550
|
|
},
|
|
{
|
|
"epoch": 1.4305402425578833,
|
|
"grad_norm": 0.5217135834643721,
|
|
"learning_rate": 3.8689806310332855e-05,
|
|
"loss": 0.3136,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1708628535270691,
|
|
"step": 2595,
|
|
"valid_targets_mean": 4318.0,
|
|
"valid_targets_min": 1546
|
|
},
|
|
{
|
|
"epoch": 1.4332965821389196,
|
|
"grad_norm": 0.5857354613565579,
|
|
"learning_rate": 3.868000242238309e-05,
|
|
"loss": 0.3065,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17321570217609406,
|
|
"step": 2600,
|
|
"valid_targets_mean": 4605.4,
|
|
"valid_targets_min": 1956
|
|
},
|
|
{
|
|
"epoch": 1.436052921719956,
|
|
"grad_norm": 0.5359482144815007,
|
|
"learning_rate": 3.867016324244308e-05,
|
|
"loss": 0.3182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12835712730884552,
|
|
"step": 2605,
|
|
"valid_targets_mean": 3389.5,
|
|
"valid_targets_min": 1538
|
|
},
|
|
{
|
|
"epoch": 1.4388092613009924,
|
|
"grad_norm": 0.6211103306463823,
|
|
"learning_rate": 3.8660288789101925e-05,
|
|
"loss": 0.3172,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16628815233707428,
|
|
"step": 2610,
|
|
"valid_targets_mean": 3408.9,
|
|
"valid_targets_min": 1639
|
|
},
|
|
{
|
|
"epoch": 1.4415656008820288,
|
|
"grad_norm": 0.6038331162343423,
|
|
"learning_rate": 3.865037908101535e-05,
|
|
"loss": 0.313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15019799768924713,
|
|
"step": 2615,
|
|
"valid_targets_mean": 3183.5,
|
|
"valid_targets_min": 935
|
|
},
|
|
{
|
|
"epoch": 1.4443219404630652,
|
|
"grad_norm": 0.5866829991038175,
|
|
"learning_rate": 3.864043413690569e-05,
|
|
"loss": 0.3235,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16500423848628998,
|
|
"step": 2620,
|
|
"valid_targets_mean": 3576.2,
|
|
"valid_targets_min": 1266
|
|
},
|
|
{
|
|
"epoch": 1.4470782800441016,
|
|
"grad_norm": 0.6225661852848867,
|
|
"learning_rate": 3.8630453975561856e-05,
|
|
"loss": 0.3217,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12444554269313812,
|
|
"step": 2625,
|
|
"valid_targets_mean": 2092.0,
|
|
"valid_targets_min": 1170
|
|
},
|
|
{
|
|
"epoch": 1.449834619625138,
|
|
"grad_norm": 0.5249607258607903,
|
|
"learning_rate": 3.8620438615839286e-05,
|
|
"loss": 0.3152,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14304453134536743,
|
|
"step": 2630,
|
|
"valid_targets_mean": 4129.1,
|
|
"valid_targets_min": 1668
|
|
},
|
|
{
|
|
"epoch": 1.4525909592061743,
|
|
"grad_norm": 0.5832437110570117,
|
|
"learning_rate": 3.8610388076659944e-05,
|
|
"loss": 0.3168,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15339717268943787,
|
|
"step": 2635,
|
|
"valid_targets_mean": 3367.0,
|
|
"valid_targets_min": 1329
|
|
},
|
|
{
|
|
"epoch": 1.4553472987872107,
|
|
"grad_norm": 0.6180816667354885,
|
|
"learning_rate": 3.8600302377012226e-05,
|
|
"loss": 0.3056,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18806253373622894,
|
|
"step": 2640,
|
|
"valid_targets_mean": 3943.2,
|
|
"valid_targets_min": 2108
|
|
},
|
|
{
|
|
"epoch": 1.458103638368247,
|
|
"grad_norm": 0.5021729228587204,
|
|
"learning_rate": 3.8590181535950966e-05,
|
|
"loss": 0.2859,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12701448798179626,
|
|
"step": 2645,
|
|
"valid_targets_mean": 4311.9,
|
|
"valid_targets_min": 2214
|
|
},
|
|
{
|
|
"epoch": 1.4608599779492835,
|
|
"grad_norm": 3.078676929508479,
|
|
"learning_rate": 3.858002557259741e-05,
|
|
"loss": 0.3128,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15151967108249664,
|
|
"step": 2650,
|
|
"valid_targets_mean": 3457.4,
|
|
"valid_targets_min": 1561
|
|
},
|
|
{
|
|
"epoch": 1.4636163175303198,
|
|
"grad_norm": 0.7346797372641912,
|
|
"learning_rate": 3.856983450613912e-05,
|
|
"loss": 0.3116,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15501129627227783,
|
|
"step": 2655,
|
|
"valid_targets_mean": 3474.9,
|
|
"valid_targets_min": 771
|
|
},
|
|
{
|
|
"epoch": 1.4663726571113562,
|
|
"grad_norm": 0.5190618013138691,
|
|
"learning_rate": 3.855960835583003e-05,
|
|
"loss": 0.3186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14920921623706818,
|
|
"step": 2660,
|
|
"valid_targets_mean": 4008.4,
|
|
"valid_targets_min": 1770
|
|
},
|
|
{
|
|
"epoch": 1.4691289966923926,
|
|
"grad_norm": 0.608435130838089,
|
|
"learning_rate": 3.854934714099031e-05,
|
|
"loss": 0.2972,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12759672105312347,
|
|
"step": 2665,
|
|
"valid_targets_mean": 2430.8,
|
|
"valid_targets_min": 994
|
|
},
|
|
{
|
|
"epoch": 1.471885336273429,
|
|
"grad_norm": 0.5701099904818253,
|
|
"learning_rate": 3.8539050881006404e-05,
|
|
"loss": 0.2965,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14958828687667847,
|
|
"step": 2670,
|
|
"valid_targets_mean": 3961.1,
|
|
"valid_targets_min": 1738
|
|
},
|
|
{
|
|
"epoch": 1.4746416758544654,
|
|
"grad_norm": 0.590643679793242,
|
|
"learning_rate": 3.852871959533096e-05,
|
|
"loss": 0.3132,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1502712070941925,
|
|
"step": 2675,
|
|
"valid_targets_mean": 3741.0,
|
|
"valid_targets_min": 862
|
|
},
|
|
{
|
|
"epoch": 1.4773980154355018,
|
|
"grad_norm": 0.7155303436273717,
|
|
"learning_rate": 3.8518353303482796e-05,
|
|
"loss": 0.3039,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1428161859512329,
|
|
"step": 2680,
|
|
"valid_targets_mean": 3093.5,
|
|
"valid_targets_min": 1408
|
|
},
|
|
{
|
|
"epoch": 1.4801543550165381,
|
|
"grad_norm": 0.5078335986518911,
|
|
"learning_rate": 3.850795202504688e-05,
|
|
"loss": 0.2921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1329430341720581,
|
|
"step": 2685,
|
|
"valid_targets_mean": 4835.2,
|
|
"valid_targets_min": 2110
|
|
},
|
|
{
|
|
"epoch": 1.4829106945975745,
|
|
"grad_norm": 0.5858496276010408,
|
|
"learning_rate": 3.8497515779674254e-05,
|
|
"loss": 0.31,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1456671953201294,
|
|
"step": 2690,
|
|
"valid_targets_mean": 3404.2,
|
|
"valid_targets_min": 1317
|
|
},
|
|
{
|
|
"epoch": 1.485667034178611,
|
|
"grad_norm": 0.556786955640969,
|
|
"learning_rate": 3.848704458708205e-05,
|
|
"loss": 0.3074,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15538150072097778,
|
|
"step": 2695,
|
|
"valid_targets_mean": 4223.4,
|
|
"valid_targets_min": 1863
|
|
},
|
|
{
|
|
"epoch": 1.4884233737596473,
|
|
"grad_norm": 0.5935414868515891,
|
|
"learning_rate": 3.847653846705342e-05,
|
|
"loss": 0.3,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17564596235752106,
|
|
"step": 2700,
|
|
"valid_targets_mean": 3774.9,
|
|
"valid_targets_min": 2594
|
|
},
|
|
{
|
|
"epoch": 1.4911797133406837,
|
|
"grad_norm": 0.5620570404062514,
|
|
"learning_rate": 3.846599743943749e-05,
|
|
"loss": 0.3104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13488194346427917,
|
|
"step": 2705,
|
|
"valid_targets_mean": 3062.8,
|
|
"valid_targets_min": 1443
|
|
},
|
|
{
|
|
"epoch": 1.49393605292172,
|
|
"grad_norm": 0.5602319967064161,
|
|
"learning_rate": 3.845542152414934e-05,
|
|
"loss": 0.31,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14981862902641296,
|
|
"step": 2710,
|
|
"valid_targets_mean": 3855.1,
|
|
"valid_targets_min": 1268
|
|
},
|
|
{
|
|
"epoch": 1.4966923925027564,
|
|
"grad_norm": 0.48894027716196065,
|
|
"learning_rate": 3.844481074116998e-05,
|
|
"loss": 0.3042,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15919747948646545,
|
|
"step": 2715,
|
|
"valid_targets_mean": 5580.1,
|
|
"valid_targets_min": 2023
|
|
},
|
|
{
|
|
"epoch": 1.4994487320837928,
|
|
"grad_norm": 0.5620282113827801,
|
|
"learning_rate": 3.8434165110546287e-05,
|
|
"loss": 0.2961,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17603543400764465,
|
|
"step": 2720,
|
|
"valid_targets_mean": 4074.2,
|
|
"valid_targets_min": 1922
|
|
},
|
|
{
|
|
"epoch": 1.5022050716648292,
|
|
"grad_norm": 0.5797605592364853,
|
|
"learning_rate": 3.842348465239096e-05,
|
|
"loss": 0.3048,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15549787878990173,
|
|
"step": 2725,
|
|
"valid_targets_mean": 3672.6,
|
|
"valid_targets_min": 2089
|
|
},
|
|
{
|
|
"epoch": 1.5049614112458656,
|
|
"grad_norm": 0.6294390688131681,
|
|
"learning_rate": 3.841276938688251e-05,
|
|
"loss": 0.3072,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20381753146648407,
|
|
"step": 2730,
|
|
"valid_targets_mean": 4324.4,
|
|
"valid_targets_min": 1363
|
|
},
|
|
{
|
|
"epoch": 1.507717750826902,
|
|
"grad_norm": 0.5955664930562039,
|
|
"learning_rate": 3.8402019334265216e-05,
|
|
"loss": 0.2976,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16087070107460022,
|
|
"step": 2735,
|
|
"valid_targets_mean": 3350.5,
|
|
"valid_targets_min": 1087
|
|
},
|
|
{
|
|
"epoch": 1.5104740904079383,
|
|
"grad_norm": 0.6032150957578843,
|
|
"learning_rate": 3.839123451484907e-05,
|
|
"loss": 0.3077,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.18486779928207397,
|
|
"step": 2740,
|
|
"valid_targets_mean": 4624.4,
|
|
"valid_targets_min": 2476
|
|
},
|
|
{
|
|
"epoch": 1.5132304299889747,
|
|
"grad_norm": 0.5422235996563903,
|
|
"learning_rate": 3.838041494900975e-05,
|
|
"loss": 0.3014,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17244887351989746,
|
|
"step": 2745,
|
|
"valid_targets_mean": 4545.9,
|
|
"valid_targets_min": 2311
|
|
},
|
|
{
|
|
"epoch": 1.515986769570011,
|
|
"grad_norm": 0.9041835503447105,
|
|
"learning_rate": 3.836956065718859e-05,
|
|
"loss": 0.2947,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14451348781585693,
|
|
"step": 2750,
|
|
"valid_targets_mean": 3913.2,
|
|
"valid_targets_min": 2010
|
|
},
|
|
{
|
|
"epoch": 1.5187431091510475,
|
|
"grad_norm": 0.5845183952326188,
|
|
"learning_rate": 3.8358671659892507e-05,
|
|
"loss": 0.2885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10664557665586472,
|
|
"step": 2755,
|
|
"valid_targets_mean": 2885.6,
|
|
"valid_targets_min": 1294
|
|
},
|
|
{
|
|
"epoch": 1.5214994487320839,
|
|
"grad_norm": 0.623860843417031,
|
|
"learning_rate": 3.8347747977694014e-05,
|
|
"loss": 0.3103,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17998671531677246,
|
|
"step": 2760,
|
|
"valid_targets_mean": 3632.0,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 1.5242557883131203,
|
|
"grad_norm": 0.6345712595670284,
|
|
"learning_rate": 3.8336789631231136e-05,
|
|
"loss": 0.2957,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12867692112922668,
|
|
"step": 2765,
|
|
"valid_targets_mean": 2715.8,
|
|
"valid_targets_min": 1582
|
|
},
|
|
{
|
|
"epoch": 1.5270121278941566,
|
|
"grad_norm": 0.6446908014602838,
|
|
"learning_rate": 3.832579664120741e-05,
|
|
"loss": 0.3038,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.172246053814888,
|
|
"step": 2770,
|
|
"valid_targets_mean": 3353.0,
|
|
"valid_targets_min": 2403
|
|
},
|
|
{
|
|
"epoch": 1.529768467475193,
|
|
"grad_norm": 0.5698721508646416,
|
|
"learning_rate": 3.83147690283918e-05,
|
|
"loss": 0.3097,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16058051586151123,
|
|
"step": 2775,
|
|
"valid_targets_mean": 4056.6,
|
|
"valid_targets_min": 2288
|
|
},
|
|
{
|
|
"epoch": 1.5325248070562294,
|
|
"grad_norm": 0.5527802210507292,
|
|
"learning_rate": 3.830370681361869e-05,
|
|
"loss": 0.3042,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19106632471084595,
|
|
"step": 2780,
|
|
"valid_targets_mean": 5186.2,
|
|
"valid_targets_min": 2187
|
|
},
|
|
{
|
|
"epoch": 1.5352811466372658,
|
|
"grad_norm": 0.6081507903403066,
|
|
"learning_rate": 3.829261001778785e-05,
|
|
"loss": 0.2994,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19244495034217834,
|
|
"step": 2785,
|
|
"valid_targets_mean": 4289.0,
|
|
"valid_targets_min": 2606
|
|
},
|
|
{
|
|
"epoch": 1.5380374862183022,
|
|
"grad_norm": 0.566606367644529,
|
|
"learning_rate": 3.8281478661864365e-05,
|
|
"loss": 0.3126,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21449314057826996,
|
|
"step": 2790,
|
|
"valid_targets_mean": 4530.1,
|
|
"valid_targets_min": 2009
|
|
},
|
|
{
|
|
"epoch": 1.5407938257993385,
|
|
"grad_norm": 0.522494060993465,
|
|
"learning_rate": 3.8270312766878624e-05,
|
|
"loss": 0.2872,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12325183302164078,
|
|
"step": 2795,
|
|
"valid_targets_mean": 3603.1,
|
|
"valid_targets_min": 1902
|
|
},
|
|
{
|
|
"epoch": 1.543550165380375,
|
|
"grad_norm": 0.5648684579804323,
|
|
"learning_rate": 3.825911235392628e-05,
|
|
"loss": 0.2872,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16402718424797058,
|
|
"step": 2800,
|
|
"valid_targets_mean": 3794.1,
|
|
"valid_targets_min": 2261
|
|
},
|
|
{
|
|
"epoch": 1.5463065049614113,
|
|
"grad_norm": 0.5633462191610494,
|
|
"learning_rate": 3.824787744416818e-05,
|
|
"loss": 0.2874,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13553448021411896,
|
|
"step": 2805,
|
|
"valid_targets_mean": 4446.4,
|
|
"valid_targets_min": 902
|
|
},
|
|
{
|
|
"epoch": 1.5490628445424477,
|
|
"grad_norm": 0.6109969660415735,
|
|
"learning_rate": 3.823660805883036e-05,
|
|
"loss": 0.2969,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16743594408035278,
|
|
"step": 2810,
|
|
"valid_targets_mean": 3782.0,
|
|
"valid_targets_min": 1218
|
|
},
|
|
{
|
|
"epoch": 1.551819184123484,
|
|
"grad_norm": 0.657084098559412,
|
|
"learning_rate": 3.822530421920399e-05,
|
|
"loss": 0.3011,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14217126369476318,
|
|
"step": 2815,
|
|
"valid_targets_mean": 4039.6,
|
|
"valid_targets_min": 1658
|
|
},
|
|
{
|
|
"epoch": 1.5545755237045205,
|
|
"grad_norm": 0.6072867313927341,
|
|
"learning_rate": 3.8213965946645324e-05,
|
|
"loss": 0.2976,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14695695042610168,
|
|
"step": 2820,
|
|
"valid_targets_mean": 3492.2,
|
|
"valid_targets_min": 1421
|
|
},
|
|
{
|
|
"epoch": 1.5573318632855568,
|
|
"grad_norm": 0.5191114416270458,
|
|
"learning_rate": 3.8202593262575674e-05,
|
|
"loss": 0.2944,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11723275482654572,
|
|
"step": 2825,
|
|
"valid_targets_mean": 3909.8,
|
|
"valid_targets_min": 2315
|
|
},
|
|
{
|
|
"epoch": 1.5600882028665932,
|
|
"grad_norm": 0.5552534282815287,
|
|
"learning_rate": 3.819118618848137e-05,
|
|
"loss": 0.3058,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1425410807132721,
|
|
"step": 2830,
|
|
"valid_targets_mean": 3644.5,
|
|
"valid_targets_min": 1187
|
|
},
|
|
{
|
|
"epoch": 1.5628445424476296,
|
|
"grad_norm": 0.5403691531553851,
|
|
"learning_rate": 3.8179744745913724e-05,
|
|
"loss": 0.3035,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1270756721496582,
|
|
"step": 2835,
|
|
"valid_targets_mean": 3410.8,
|
|
"valid_targets_min": 616
|
|
},
|
|
{
|
|
"epoch": 1.565600882028666,
|
|
"grad_norm": 0.6666354832350139,
|
|
"learning_rate": 3.816826895648895e-05,
|
|
"loss": 0.2994,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13888883590698242,
|
|
"step": 2840,
|
|
"valid_targets_mean": 2662.8,
|
|
"valid_targets_min": 1363
|
|
},
|
|
{
|
|
"epoch": 1.5683572216097024,
|
|
"grad_norm": 0.606852599372683,
|
|
"learning_rate": 3.815675884188818e-05,
|
|
"loss": 0.2894,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15082953870296478,
|
|
"step": 2845,
|
|
"valid_targets_mean": 3497.0,
|
|
"valid_targets_min": 1393
|
|
},
|
|
{
|
|
"epoch": 1.5711135611907387,
|
|
"grad_norm": 0.5056916427936858,
|
|
"learning_rate": 3.81452144238574e-05,
|
|
"loss": 0.2839,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14923208951950073,
|
|
"step": 2850,
|
|
"valid_targets_mean": 4659.9,
|
|
"valid_targets_min": 1414
|
|
},
|
|
{
|
|
"epoch": 1.5738699007717751,
|
|
"grad_norm": 0.5646552602958356,
|
|
"learning_rate": 3.813363572420738e-05,
|
|
"loss": 0.2912,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1571073830127716,
|
|
"step": 2855,
|
|
"valid_targets_mean": 3795.8,
|
|
"valid_targets_min": 620
|
|
},
|
|
{
|
|
"epoch": 1.5766262403528115,
|
|
"grad_norm": 0.5605125067315445,
|
|
"learning_rate": 3.8122022764813674e-05,
|
|
"loss": 0.2814,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11232224106788635,
|
|
"step": 2860,
|
|
"valid_targets_mean": 3543.5,
|
|
"valid_targets_min": 1286
|
|
},
|
|
{
|
|
"epoch": 1.579382579933848,
|
|
"grad_norm": 0.6550439415233656,
|
|
"learning_rate": 3.811037556761656e-05,
|
|
"loss": 0.2826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11929789185523987,
|
|
"step": 2865,
|
|
"valid_targets_mean": 2120.8,
|
|
"valid_targets_min": 641
|
|
},
|
|
{
|
|
"epoch": 1.5821389195148843,
|
|
"grad_norm": 0.6844143734025898,
|
|
"learning_rate": 3.8098694154621004e-05,
|
|
"loss": 0.284,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15691444277763367,
|
|
"step": 2870,
|
|
"valid_targets_mean": 3445.9,
|
|
"valid_targets_min": 2096
|
|
},
|
|
{
|
|
"epoch": 1.5848952590959207,
|
|
"grad_norm": 0.595604531880007,
|
|
"learning_rate": 3.808697854789661e-05,
|
|
"loss": 0.2961,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.20834125578403473,
|
|
"step": 2875,
|
|
"valid_targets_mean": 4586.4,
|
|
"valid_targets_min": 2783
|
|
},
|
|
{
|
|
"epoch": 1.587651598676957,
|
|
"grad_norm": 0.6168088037025864,
|
|
"learning_rate": 3.8075228769577586e-05,
|
|
"loss": 0.3092,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1272270679473877,
|
|
"step": 2880,
|
|
"valid_targets_mean": 3270.9,
|
|
"valid_targets_min": 1502
|
|
},
|
|
{
|
|
"epoch": 1.5904079382579934,
|
|
"grad_norm": 0.7017273082496397,
|
|
"learning_rate": 3.806344484186271e-05,
|
|
"loss": 0.2926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1860450655221939,
|
|
"step": 2885,
|
|
"valid_targets_mean": 3216.4,
|
|
"valid_targets_min": 986
|
|
},
|
|
{
|
|
"epoch": 1.5931642778390298,
|
|
"grad_norm": 0.5394010521497201,
|
|
"learning_rate": 3.805162678701526e-05,
|
|
"loss": 0.2935,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1287517100572586,
|
|
"step": 2890,
|
|
"valid_targets_mean": 3652.0,
|
|
"valid_targets_min": 1605
|
|
},
|
|
{
|
|
"epoch": 1.5959206174200662,
|
|
"grad_norm": 0.5385163530091687,
|
|
"learning_rate": 3.8039774627363e-05,
|
|
"loss": 0.2882,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15816931426525116,
|
|
"step": 2895,
|
|
"valid_targets_mean": 4645.4,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 1.5986769570011026,
|
|
"grad_norm": 0.4492595948700624,
|
|
"learning_rate": 3.802788838529813e-05,
|
|
"loss": 0.2827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12778984010219574,
|
|
"step": 2900,
|
|
"valid_targets_mean": 5431.8,
|
|
"valid_targets_min": 1762
|
|
},
|
|
{
|
|
"epoch": 1.601433296582139,
|
|
"grad_norm": 0.4815474856472237,
|
|
"learning_rate": 3.801596808327724e-05,
|
|
"loss": 0.2939,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13336020708084106,
|
|
"step": 2905,
|
|
"valid_targets_mean": 4498.6,
|
|
"valid_targets_min": 2379
|
|
},
|
|
{
|
|
"epoch": 1.6041896361631753,
|
|
"grad_norm": 0.7001336256097759,
|
|
"learning_rate": 3.800401374382127e-05,
|
|
"loss": 0.2816,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12304878234863281,
|
|
"step": 2910,
|
|
"valid_targets_mean": 2672.5,
|
|
"valid_targets_min": 1531
|
|
},
|
|
{
|
|
"epoch": 1.6069459757442117,
|
|
"grad_norm": 0.5288121019734653,
|
|
"learning_rate": 3.799202538951546e-05,
|
|
"loss": 0.2925,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10649672150611877,
|
|
"step": 2915,
|
|
"valid_targets_mean": 3118.6,
|
|
"valid_targets_min": 1471
|
|
},
|
|
{
|
|
"epoch": 1.609702315325248,
|
|
"grad_norm": 0.6207995285704719,
|
|
"learning_rate": 3.798000304300932e-05,
|
|
"loss": 0.2885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15830457210540771,
|
|
"step": 2920,
|
|
"valid_targets_mean": 3860.9,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 1.6124586549062845,
|
|
"grad_norm": 0.587505709706613,
|
|
"learning_rate": 3.796794672701658e-05,
|
|
"loss": 0.2978,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13932064175605774,
|
|
"step": 2925,
|
|
"valid_targets_mean": 3230.8,
|
|
"valid_targets_min": 1667
|
|
},
|
|
{
|
|
"epoch": 1.6152149944873209,
|
|
"grad_norm": 0.6046989769645399,
|
|
"learning_rate": 3.795585646431516e-05,
|
|
"loss": 0.3022,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14174184203147888,
|
|
"step": 2930,
|
|
"valid_targets_mean": 3201.0,
|
|
"valid_targets_min": 1316
|
|
},
|
|
{
|
|
"epoch": 1.6179713340683572,
|
|
"grad_norm": 0.7223830328238721,
|
|
"learning_rate": 3.794373227774708e-05,
|
|
"loss": 0.2993,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17771393060684204,
|
|
"step": 2935,
|
|
"valid_targets_mean": 2652.2,
|
|
"valid_targets_min": 1001
|
|
},
|
|
{
|
|
"epoch": 1.6207276736493936,
|
|
"grad_norm": 0.6666729704255393,
|
|
"learning_rate": 3.7931574190218504e-05,
|
|
"loss": 0.3151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11714617908000946,
|
|
"step": 2940,
|
|
"valid_targets_mean": 2528.5,
|
|
"valid_targets_min": 843
|
|
},
|
|
{
|
|
"epoch": 1.62348401323043,
|
|
"grad_norm": 0.5228552011176496,
|
|
"learning_rate": 3.791938222469959e-05,
|
|
"loss": 0.2863,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15076830983161926,
|
|
"step": 2945,
|
|
"valid_targets_mean": 4447.8,
|
|
"valid_targets_min": 2600
|
|
},
|
|
{
|
|
"epoch": 1.6262403528114664,
|
|
"grad_norm": 0.5669941521212518,
|
|
"learning_rate": 3.790715640422455e-05,
|
|
"loss": 0.3062,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13325557112693787,
|
|
"step": 2950,
|
|
"valid_targets_mean": 3453.4,
|
|
"valid_targets_min": 1241
|
|
},
|
|
{
|
|
"epoch": 1.6289966923925028,
|
|
"grad_norm": 0.625004814774253,
|
|
"learning_rate": 3.789489675189152e-05,
|
|
"loss": 0.3099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.19623157382011414,
|
|
"step": 2955,
|
|
"valid_targets_mean": 3746.0,
|
|
"valid_targets_min": 990
|
|
},
|
|
{
|
|
"epoch": 1.6317530319735392,
|
|
"grad_norm": 0.5951102262431854,
|
|
"learning_rate": 3.788260329086258e-05,
|
|
"loss": 0.2915,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1430695652961731,
|
|
"step": 2960,
|
|
"valid_targets_mean": 3809.4,
|
|
"valid_targets_min": 1131
|
|
},
|
|
{
|
|
"epoch": 1.6345093715545755,
|
|
"grad_norm": 0.6248059904749236,
|
|
"learning_rate": 3.787027604436367e-05,
|
|
"loss": 0.2956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17055535316467285,
|
|
"step": 2965,
|
|
"valid_targets_mean": 4040.2,
|
|
"valid_targets_min": 2248
|
|
},
|
|
{
|
|
"epoch": 1.637265711135612,
|
|
"grad_norm": 0.5548686553471324,
|
|
"learning_rate": 3.785791503568457e-05,
|
|
"loss": 0.2985,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1215679869055748,
|
|
"step": 2970,
|
|
"valid_targets_mean": 3809.2,
|
|
"valid_targets_min": 1909
|
|
},
|
|
{
|
|
"epoch": 1.6400220507166483,
|
|
"grad_norm": 0.5726640929081117,
|
|
"learning_rate": 3.784552028817884e-05,
|
|
"loss": 0.3028,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1445842832326889,
|
|
"step": 2975,
|
|
"valid_targets_mean": 3358.8,
|
|
"valid_targets_min": 2760
|
|
},
|
|
{
|
|
"epoch": 1.6427783902976847,
|
|
"grad_norm": 0.7060860053285646,
|
|
"learning_rate": 3.7833091825263785e-05,
|
|
"loss": 0.3009,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1519981324672699,
|
|
"step": 2980,
|
|
"valid_targets_mean": 2591.8,
|
|
"valid_targets_min": 616
|
|
},
|
|
{
|
|
"epoch": 1.645534729878721,
|
|
"grad_norm": 0.5218260611217675,
|
|
"learning_rate": 3.7820629670420406e-05,
|
|
"loss": 0.2768,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13153928518295288,
|
|
"step": 2985,
|
|
"valid_targets_mean": 3763.9,
|
|
"valid_targets_min": 3138
|
|
},
|
|
{
|
|
"epoch": 1.6482910694597575,
|
|
"grad_norm": 0.591571822423482,
|
|
"learning_rate": 3.780813384719336e-05,
|
|
"loss": 0.2837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14383168518543243,
|
|
"step": 2990,
|
|
"valid_targets_mean": 3772.1,
|
|
"valid_targets_min": 1843
|
|
},
|
|
{
|
|
"epoch": 1.6510474090407938,
|
|
"grad_norm": 0.5301815939034309,
|
|
"learning_rate": 3.779560437919091e-05,
|
|
"loss": 0.2904,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13876739144325256,
|
|
"step": 2995,
|
|
"valid_targets_mean": 3888.1,
|
|
"valid_targets_min": 2322
|
|
},
|
|
{
|
|
"epoch": 1.6538037486218302,
|
|
"grad_norm": 0.6729632529074006,
|
|
"learning_rate": 3.7783041290084904e-05,
|
|
"loss": 0.2935,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14491796493530273,
|
|
"step": 3000,
|
|
"valid_targets_mean": 2851.5,
|
|
"valid_targets_min": 1064
|
|
},
|
|
{
|
|
"epoch": 1.6565600882028666,
|
|
"grad_norm": 0.5136645857633937,
|
|
"learning_rate": 3.777044460361067e-05,
|
|
"loss": 0.2831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1273820698261261,
|
|
"step": 3005,
|
|
"valid_targets_mean": 3490.5,
|
|
"valid_targets_min": 1710
|
|
},
|
|
{
|
|
"epoch": 1.659316427783903,
|
|
"grad_norm": 0.6519671522369229,
|
|
"learning_rate": 3.775781434356705e-05,
|
|
"loss": 0.2989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14385581016540527,
|
|
"step": 3010,
|
|
"valid_targets_mean": 3912.6,
|
|
"valid_targets_min": 981
|
|
},
|
|
{
|
|
"epoch": 1.6620727673649394,
|
|
"grad_norm": 0.5857565186716999,
|
|
"learning_rate": 3.77451505338163e-05,
|
|
"loss": 0.2851,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1561882495880127,
|
|
"step": 3015,
|
|
"valid_targets_mean": 4673.2,
|
|
"valid_targets_min": 3322
|
|
},
|
|
{
|
|
"epoch": 1.6648291069459757,
|
|
"grad_norm": 0.6986766713134465,
|
|
"learning_rate": 3.773245319828405e-05,
|
|
"loss": 0.2948,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15250897407531738,
|
|
"step": 3020,
|
|
"valid_targets_mean": 3132.5,
|
|
"valid_targets_min": 1350
|
|
},
|
|
{
|
|
"epoch": 1.6675854465270121,
|
|
"grad_norm": 0.5343705069259208,
|
|
"learning_rate": 3.771972236095929e-05,
|
|
"loss": 0.2856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1229620948433876,
|
|
"step": 3025,
|
|
"valid_targets_mean": 3979.8,
|
|
"valid_targets_min": 2268
|
|
},
|
|
{
|
|
"epoch": 1.6703417861080485,
|
|
"grad_norm": 0.9774810647078945,
|
|
"learning_rate": 3.77069580458943e-05,
|
|
"loss": 0.2786,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1208256408572197,
|
|
"step": 3030,
|
|
"valid_targets_mean": 3196.8,
|
|
"valid_targets_min": 2098
|
|
},
|
|
{
|
|
"epoch": 1.673098125689085,
|
|
"grad_norm": 0.5416977052248985,
|
|
"learning_rate": 3.76941602772046e-05,
|
|
"loss": 0.2707,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12059114128351212,
|
|
"step": 3035,
|
|
"valid_targets_mean": 3781.6,
|
|
"valid_targets_min": 1686
|
|
},
|
|
{
|
|
"epoch": 1.6758544652701213,
|
|
"grad_norm": 0.5805730473297891,
|
|
"learning_rate": 3.768132907906893e-05,
|
|
"loss": 0.2794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15451817214488983,
|
|
"step": 3040,
|
|
"valid_targets_mean": 3496.2,
|
|
"valid_targets_min": 2243
|
|
},
|
|
{
|
|
"epoch": 1.6786108048511577,
|
|
"grad_norm": 0.5631485136939947,
|
|
"learning_rate": 3.766846447572916e-05,
|
|
"loss": 0.2817,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13542982935905457,
|
|
"step": 3045,
|
|
"valid_targets_mean": 3654.4,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 1.681367144432194,
|
|
"grad_norm": 0.5484598473739655,
|
|
"learning_rate": 3.765556649149031e-05,
|
|
"loss": 0.2768,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17363861203193665,
|
|
"step": 3050,
|
|
"valid_targets_mean": 4804.6,
|
|
"valid_targets_min": 1866
|
|
},
|
|
{
|
|
"epoch": 1.6841234840132304,
|
|
"grad_norm": 0.5667094829557431,
|
|
"learning_rate": 3.7642635150720426e-05,
|
|
"loss": 0.2731,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14888909459114075,
|
|
"step": 3055,
|
|
"valid_targets_mean": 3809.4,
|
|
"valid_targets_min": 872
|
|
},
|
|
{
|
|
"epoch": 1.6868798235942668,
|
|
"grad_norm": 0.6406593299871054,
|
|
"learning_rate": 3.7629670477850616e-05,
|
|
"loss": 0.279,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12725286185741425,
|
|
"step": 3060,
|
|
"valid_targets_mean": 3245.6,
|
|
"valid_targets_min": 1694
|
|
},
|
|
{
|
|
"epoch": 1.6896361631753032,
|
|
"grad_norm": 0.7416120728805293,
|
|
"learning_rate": 3.761667249737491e-05,
|
|
"loss": 0.2795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1141820102930069,
|
|
"step": 3065,
|
|
"valid_targets_mean": 2389.2,
|
|
"valid_targets_min": 897
|
|
},
|
|
{
|
|
"epoch": 1.6923925027563396,
|
|
"grad_norm": 0.6041151503981338,
|
|
"learning_rate": 3.760364123385033e-05,
|
|
"loss": 0.2996,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13265225291252136,
|
|
"step": 3070,
|
|
"valid_targets_mean": 3098.4,
|
|
"valid_targets_min": 1695
|
|
},
|
|
{
|
|
"epoch": 1.695148842337376,
|
|
"grad_norm": 0.6341655522886623,
|
|
"learning_rate": 3.759057671189673e-05,
|
|
"loss": 0.2858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13373008370399475,
|
|
"step": 3075,
|
|
"valid_targets_mean": 3205.5,
|
|
"valid_targets_min": 1710
|
|
},
|
|
{
|
|
"epoch": 1.6979051819184123,
|
|
"grad_norm": 0.5823320768439156,
|
|
"learning_rate": 3.757747895619681e-05,
|
|
"loss": 0.2819,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1254759281873703,
|
|
"step": 3080,
|
|
"valid_targets_mean": 3492.4,
|
|
"valid_targets_min": 992
|
|
},
|
|
{
|
|
"epoch": 1.7006615214994487,
|
|
"grad_norm": 0.552434665249138,
|
|
"learning_rate": 3.756434799149607e-05,
|
|
"loss": 0.2749,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14672215282917023,
|
|
"step": 3085,
|
|
"valid_targets_mean": 4175.6,
|
|
"valid_targets_min": 1673
|
|
},
|
|
{
|
|
"epoch": 1.703417861080485,
|
|
"grad_norm": 0.5369246874987358,
|
|
"learning_rate": 3.7551183842602735e-05,
|
|
"loss": 0.2807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12637808918952942,
|
|
"step": 3090,
|
|
"valid_targets_mean": 3773.9,
|
|
"valid_targets_min": 2251
|
|
},
|
|
{
|
|
"epoch": 1.7061742006615215,
|
|
"grad_norm": 0.5528521135533061,
|
|
"learning_rate": 3.7537986534387754e-05,
|
|
"loss": 0.2883,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1411515772342682,
|
|
"step": 3095,
|
|
"valid_targets_mean": 4170.9,
|
|
"valid_targets_min": 2888
|
|
},
|
|
{
|
|
"epoch": 1.7089305402425579,
|
|
"grad_norm": 0.5280204727761292,
|
|
"learning_rate": 3.752475609178468e-05,
|
|
"loss": 0.2886,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14853471517562866,
|
|
"step": 3100,
|
|
"valid_targets_mean": 4478.4,
|
|
"valid_targets_min": 2960
|
|
},
|
|
{
|
|
"epoch": 1.7116868798235942,
|
|
"grad_norm": 0.5458158687112293,
|
|
"learning_rate": 3.75114925397897e-05,
|
|
"loss": 0.2832,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11362916231155396,
|
|
"step": 3105,
|
|
"valid_targets_mean": 3730.8,
|
|
"valid_targets_min": 1815
|
|
},
|
|
{
|
|
"epoch": 1.7144432194046306,
|
|
"grad_norm": 0.5258489467697137,
|
|
"learning_rate": 3.7498195903461556e-05,
|
|
"loss": 0.2854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16439346969127655,
|
|
"step": 3110,
|
|
"valid_targets_mean": 4718.6,
|
|
"valid_targets_min": 1725
|
|
},
|
|
{
|
|
"epoch": 1.717199558985667,
|
|
"grad_norm": 0.5163088890763808,
|
|
"learning_rate": 3.748486620792147e-05,
|
|
"loss": 0.2795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12315109372138977,
|
|
"step": 3115,
|
|
"valid_targets_mean": 3794.4,
|
|
"valid_targets_min": 1951
|
|
},
|
|
{
|
|
"epoch": 1.7199558985667034,
|
|
"grad_norm": 0.5339145759980849,
|
|
"learning_rate": 3.747150347835314e-05,
|
|
"loss": 0.2248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09877531230449677,
|
|
"step": 3120,
|
|
"valid_targets_mean": 2261.1,
|
|
"valid_targets_min": 331
|
|
},
|
|
{
|
|
"epoch": 1.7227122381477398,
|
|
"grad_norm": 0.3881075348581907,
|
|
"learning_rate": 3.7458107740002676e-05,
|
|
"loss": 0.1335,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06161794438958168,
|
|
"step": 3125,
|
|
"valid_targets_mean": 3604.4,
|
|
"valid_targets_min": 708
|
|
},
|
|
{
|
|
"epoch": 1.7254685777287762,
|
|
"grad_norm": 0.4750440862339531,
|
|
"learning_rate": 3.744467901817854e-05,
|
|
"loss": 0.1166,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06372582167387009,
|
|
"step": 3130,
|
|
"valid_targets_mean": 2933.4,
|
|
"valid_targets_min": 729
|
|
},
|
|
{
|
|
"epoch": 1.7282249173098125,
|
|
"grad_norm": 0.44702826431290754,
|
|
"learning_rate": 3.743121733825152e-05,
|
|
"loss": 0.1201,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09848805516958237,
|
|
"step": 3135,
|
|
"valid_targets_mean": 2718.9,
|
|
"valid_targets_min": 921
|
|
},
|
|
{
|
|
"epoch": 1.730981256890849,
|
|
"grad_norm": 0.42997582888385405,
|
|
"learning_rate": 3.741772272565468e-05,
|
|
"loss": 0.1085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04737251624464989,
|
|
"step": 3140,
|
|
"valid_targets_mean": 2827.9,
|
|
"valid_targets_min": 813
|
|
},
|
|
{
|
|
"epoch": 1.7337375964718853,
|
|
"grad_norm": 0.44026972963129035,
|
|
"learning_rate": 3.7404195205883274e-05,
|
|
"loss": 0.1197,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.061276860535144806,
|
|
"step": 3145,
|
|
"valid_targets_mean": 2900.5,
|
|
"valid_targets_min": 854
|
|
},
|
|
{
|
|
"epoch": 1.7364939360529217,
|
|
"grad_norm": 0.7702107663014082,
|
|
"learning_rate": 3.7390634804494757e-05,
|
|
"loss": 0.1676,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06564486026763916,
|
|
"step": 3150,
|
|
"valid_targets_mean": 1034.4,
|
|
"valid_targets_min": 536
|
|
},
|
|
{
|
|
"epoch": 1.739250275633958,
|
|
"grad_norm": 0.39887241364165865,
|
|
"learning_rate": 3.737704154710868e-05,
|
|
"loss": 0.1297,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05171442776918411,
|
|
"step": 3155,
|
|
"valid_targets_mean": 2992.8,
|
|
"valid_targets_min": 915
|
|
},
|
|
{
|
|
"epoch": 1.7420066152149944,
|
|
"grad_norm": 0.39795106750423354,
|
|
"learning_rate": 3.73634154594067e-05,
|
|
"loss": 0.1253,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06273576617240906,
|
|
"step": 3160,
|
|
"valid_targets_mean": 3356.0,
|
|
"valid_targets_min": 2281
|
|
},
|
|
{
|
|
"epoch": 1.7447629547960308,
|
|
"grad_norm": 0.34445866777428863,
|
|
"learning_rate": 3.734975656713248e-05,
|
|
"loss": 0.1593,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042710188776254654,
|
|
"step": 3165,
|
|
"valid_targets_mean": 3776.1,
|
|
"valid_targets_min": 958
|
|
},
|
|
{
|
|
"epoch": 1.7475192943770672,
|
|
"grad_norm": 0.490519632867232,
|
|
"learning_rate": 3.733606489609166e-05,
|
|
"loss": 0.0971,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06546507775783539,
|
|
"step": 3170,
|
|
"valid_targets_mean": 4237.5,
|
|
"valid_targets_min": 2082
|
|
},
|
|
{
|
|
"epoch": 1.7502756339581036,
|
|
"grad_norm": 0.35789985839512845,
|
|
"learning_rate": 3.732234047215181e-05,
|
|
"loss": 0.111,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05938592553138733,
|
|
"step": 3175,
|
|
"valid_targets_mean": 4378.9,
|
|
"valid_targets_min": 3337
|
|
},
|
|
{
|
|
"epoch": 1.75303197353914,
|
|
"grad_norm": 0.33816283071650105,
|
|
"learning_rate": 3.730858332124239e-05,
|
|
"loss": 0.11,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05424990504980087,
|
|
"step": 3180,
|
|
"valid_targets_mean": 3934.4,
|
|
"valid_targets_min": 515
|
|
},
|
|
{
|
|
"epoch": 1.7557883131201764,
|
|
"grad_norm": 0.5512988372682727,
|
|
"learning_rate": 3.729479346935468e-05,
|
|
"loss": 0.1634,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06325031816959381,
|
|
"step": 3185,
|
|
"valid_targets_mean": 1878.6,
|
|
"valid_targets_min": 546
|
|
},
|
|
{
|
|
"epoch": 1.7585446527012127,
|
|
"grad_norm": 0.37186111019494583,
|
|
"learning_rate": 3.728097094254174e-05,
|
|
"loss": 0.1043,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04918120056390762,
|
|
"step": 3190,
|
|
"valid_targets_mean": 3432.2,
|
|
"valid_targets_min": 2715
|
|
},
|
|
{
|
|
"epoch": 1.7613009922822491,
|
|
"grad_norm": 0.4770183244929656,
|
|
"learning_rate": 3.726711576691838e-05,
|
|
"loss": 0.1196,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047410741448402405,
|
|
"step": 3195,
|
|
"valid_targets_mean": 1612.8,
|
|
"valid_targets_min": 515
|
|
},
|
|
{
|
|
"epoch": 1.7640573318632855,
|
|
"grad_norm": 0.3719765186234999,
|
|
"learning_rate": 3.725322796866106e-05,
|
|
"loss": 0.1025,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031192610040307045,
|
|
"step": 3200,
|
|
"valid_targets_mean": 3322.8,
|
|
"valid_targets_min": 840
|
|
},
|
|
{
|
|
"epoch": 1.7668136714443219,
|
|
"grad_norm": 0.44077928239574515,
|
|
"learning_rate": 3.7239307574007916e-05,
|
|
"loss": 0.1161,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057165950536727905,
|
|
"step": 3205,
|
|
"valid_targets_mean": 2764.0,
|
|
"valid_targets_min": 574
|
|
},
|
|
{
|
|
"epoch": 1.7695700110253583,
|
|
"grad_norm": 0.4912760105644369,
|
|
"learning_rate": 3.722535460925864e-05,
|
|
"loss": 0.1214,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06553533673286438,
|
|
"step": 3210,
|
|
"valid_targets_mean": 2077.9,
|
|
"valid_targets_min": 847
|
|
},
|
|
{
|
|
"epoch": 1.7723263506063947,
|
|
"grad_norm": 0.42572462655376986,
|
|
"learning_rate": 3.721136910077446e-05,
|
|
"loss": 0.1293,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07079510390758514,
|
|
"step": 3215,
|
|
"valid_targets_mean": 2857.1,
|
|
"valid_targets_min": 770
|
|
},
|
|
{
|
|
"epoch": 1.775082690187431,
|
|
"grad_norm": 0.4020851260492457,
|
|
"learning_rate": 3.71973510749781e-05,
|
|
"loss": 0.1268,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05171942338347435,
|
|
"step": 3220,
|
|
"valid_targets_mean": 1736.5,
|
|
"valid_targets_min": 611
|
|
},
|
|
{
|
|
"epoch": 1.7778390297684674,
|
|
"grad_norm": 0.5419630267999331,
|
|
"learning_rate": 3.7183300558353704e-05,
|
|
"loss": 0.1732,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1274569034576416,
|
|
"step": 3225,
|
|
"valid_targets_mean": 3779.8,
|
|
"valid_targets_min": 2812
|
|
},
|
|
{
|
|
"epoch": 1.7805953693495038,
|
|
"grad_norm": 0.3228801699540472,
|
|
"learning_rate": 3.716921757744682e-05,
|
|
"loss": 0.1073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045997750014066696,
|
|
"step": 3230,
|
|
"valid_targets_mean": 4061.4,
|
|
"valid_targets_min": 3722
|
|
},
|
|
{
|
|
"epoch": 1.7833517089305402,
|
|
"grad_norm": 0.47299990957316146,
|
|
"learning_rate": 3.715510215886431e-05,
|
|
"loss": 0.1174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05145843327045441,
|
|
"step": 3235,
|
|
"valid_targets_mean": 1199.0,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 1.7861080485115766,
|
|
"grad_norm": 0.37754071925959404,
|
|
"learning_rate": 3.714095432927433e-05,
|
|
"loss": 0.1562,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05151469260454178,
|
|
"step": 3240,
|
|
"valid_targets_mean": 3407.8,
|
|
"valid_targets_min": 1803
|
|
},
|
|
{
|
|
"epoch": 1.788864388092613,
|
|
"grad_norm": 0.3632759567176424,
|
|
"learning_rate": 3.712677411540627e-05,
|
|
"loss": 0.1374,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057262301445007324,
|
|
"step": 3245,
|
|
"valid_targets_mean": 3789.4,
|
|
"valid_targets_min": 3167
|
|
},
|
|
{
|
|
"epoch": 1.7916207276736493,
|
|
"grad_norm": 0.7326147673126931,
|
|
"learning_rate": 3.711256154405071e-05,
|
|
"loss": 0.1513,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1422143280506134,
|
|
"step": 3250,
|
|
"valid_targets_mean": 1682.0,
|
|
"valid_targets_min": 915
|
|
},
|
|
{
|
|
"epoch": 1.7943770672546857,
|
|
"grad_norm": 0.4298218846828181,
|
|
"learning_rate": 3.709831664205935e-05,
|
|
"loss": 0.305,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06525042653083801,
|
|
"step": 3255,
|
|
"valid_targets_mean": 3663.0,
|
|
"valid_targets_min": 2876
|
|
},
|
|
{
|
|
"epoch": 1.797133406835722,
|
|
"grad_norm": 0.42473767748790403,
|
|
"learning_rate": 3.708403943634499e-05,
|
|
"loss": 0.1302,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05557072162628174,
|
|
"step": 3260,
|
|
"valid_targets_mean": 3140.6,
|
|
"valid_targets_min": 828
|
|
},
|
|
{
|
|
"epoch": 1.7998897464167585,
|
|
"grad_norm": 0.5263055896576978,
|
|
"learning_rate": 3.706972995388143e-05,
|
|
"loss": 0.1279,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1038624495267868,
|
|
"step": 3265,
|
|
"valid_targets_mean": 3329.5,
|
|
"valid_targets_min": 1794
|
|
},
|
|
{
|
|
"epoch": 1.8026460859977949,
|
|
"grad_norm": 0.4512062244849988,
|
|
"learning_rate": 3.705538822170348e-05,
|
|
"loss": 0.2226,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06986123323440552,
|
|
"step": 3270,
|
|
"valid_targets_mean": 3535.5,
|
|
"valid_targets_min": 2771
|
|
},
|
|
{
|
|
"epoch": 1.8054024255788312,
|
|
"grad_norm": 0.27598586360807054,
|
|
"learning_rate": 3.704101426690686e-05,
|
|
"loss": 0.0937,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04326866567134857,
|
|
"step": 3275,
|
|
"valid_targets_mean": 4594.6,
|
|
"valid_targets_min": 2655
|
|
},
|
|
{
|
|
"epoch": 1.8081587651598676,
|
|
"grad_norm": 0.33184094413933124,
|
|
"learning_rate": 3.702660811664819e-05,
|
|
"loss": 0.1257,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02866976521909237,
|
|
"step": 3280,
|
|
"valid_targets_mean": 2743.9,
|
|
"valid_targets_min": 557
|
|
},
|
|
{
|
|
"epoch": 1.810915104740904,
|
|
"grad_norm": 0.38266588457867284,
|
|
"learning_rate": 3.70121697981449e-05,
|
|
"loss": 0.1243,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06082747131586075,
|
|
"step": 3285,
|
|
"valid_targets_mean": 4249.6,
|
|
"valid_targets_min": 1606
|
|
},
|
|
{
|
|
"epoch": 1.8136714443219404,
|
|
"grad_norm": 0.37425380062750546,
|
|
"learning_rate": 3.6997699338675184e-05,
|
|
"loss": 0.108,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.048022154718637466,
|
|
"step": 3290,
|
|
"valid_targets_mean": 3570.9,
|
|
"valid_targets_min": 2163
|
|
},
|
|
{
|
|
"epoch": 1.8164277839029768,
|
|
"grad_norm": 0.4052441520761462,
|
|
"learning_rate": 3.698319676557799e-05,
|
|
"loss": 0.107,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0578053817152977,
|
|
"step": 3295,
|
|
"valid_targets_mean": 3057.5,
|
|
"valid_targets_min": 533
|
|
},
|
|
{
|
|
"epoch": 1.8191841234840131,
|
|
"grad_norm": 0.4781912022159935,
|
|
"learning_rate": 3.696866210625291e-05,
|
|
"loss": 0.1083,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05583494156599045,
|
|
"step": 3300,
|
|
"valid_targets_mean": 3622.0,
|
|
"valid_targets_min": 2249
|
|
},
|
|
{
|
|
"epoch": 1.8219404630650495,
|
|
"grad_norm": 0.3623886934987538,
|
|
"learning_rate": 3.695409538816018e-05,
|
|
"loss": 0.1032,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04643484205007553,
|
|
"step": 3305,
|
|
"valid_targets_mean": 2801.5,
|
|
"valid_targets_min": 817
|
|
},
|
|
{
|
|
"epoch": 1.824696802646086,
|
|
"grad_norm": 0.582785571023652,
|
|
"learning_rate": 3.693949663882058e-05,
|
|
"loss": 0.1577,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1477086842060089,
|
|
"step": 3310,
|
|
"valid_targets_mean": 2015.1,
|
|
"valid_targets_min": 859
|
|
},
|
|
{
|
|
"epoch": 1.8274531422271223,
|
|
"grad_norm": 0.35947197437297973,
|
|
"learning_rate": 3.692486588581543e-05,
|
|
"loss": 0.1032,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05858401954174042,
|
|
"step": 3315,
|
|
"valid_targets_mean": 3601.5,
|
|
"valid_targets_min": 2709
|
|
},
|
|
{
|
|
"epoch": 1.8302094818081587,
|
|
"grad_norm": 0.6804209198056856,
|
|
"learning_rate": 3.6910203156786496e-05,
|
|
"loss": 0.1021,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07768724113702774,
|
|
"step": 3320,
|
|
"valid_targets_mean": 1605.5,
|
|
"valid_targets_min": 781
|
|
},
|
|
{
|
|
"epoch": 1.832965821389195,
|
|
"grad_norm": 0.3341185199076039,
|
|
"learning_rate": 3.6895508479435964e-05,
|
|
"loss": 0.0984,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042303942143917084,
|
|
"step": 3325,
|
|
"valid_targets_mean": 3703.2,
|
|
"valid_targets_min": 2343
|
|
},
|
|
{
|
|
"epoch": 1.8357221609702314,
|
|
"grad_norm": 0.40469774965758526,
|
|
"learning_rate": 3.688078188152639e-05,
|
|
"loss": 0.1002,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04548351839184761,
|
|
"step": 3330,
|
|
"valid_targets_mean": 1990.2,
|
|
"valid_targets_min": 545
|
|
},
|
|
{
|
|
"epoch": 1.8384785005512678,
|
|
"grad_norm": 0.3656525932930328,
|
|
"learning_rate": 3.6866023390880605e-05,
|
|
"loss": 0.1604,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05424724891781807,
|
|
"step": 3335,
|
|
"valid_targets_mean": 3322.1,
|
|
"valid_targets_min": 904
|
|
},
|
|
{
|
|
"epoch": 1.8412348401323042,
|
|
"grad_norm": 0.36669007682614035,
|
|
"learning_rate": 3.6851233035381746e-05,
|
|
"loss": 0.133,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09258392453193665,
|
|
"step": 3340,
|
|
"valid_targets_mean": 3514.5,
|
|
"valid_targets_min": 2184
|
|
},
|
|
{
|
|
"epoch": 1.8439911797133406,
|
|
"grad_norm": 0.3304897596974123,
|
|
"learning_rate": 3.683641084297309e-05,
|
|
"loss": 0.1244,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03585117310285568,
|
|
"step": 3345,
|
|
"valid_targets_mean": 1680.6,
|
|
"valid_targets_min": 595
|
|
},
|
|
{
|
|
"epoch": 1.846747519294377,
|
|
"grad_norm": 0.9384493604738458,
|
|
"learning_rate": 3.6821556841658116e-05,
|
|
"loss": 0.1433,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1252661496400833,
|
|
"step": 3350,
|
|
"valid_targets_mean": 1323.4,
|
|
"valid_targets_min": 575
|
|
},
|
|
{
|
|
"epoch": 1.8495038588754134,
|
|
"grad_norm": 0.5576903831753593,
|
|
"learning_rate": 3.680667105950037e-05,
|
|
"loss": 0.1671,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06596562266349792,
|
|
"step": 3355,
|
|
"valid_targets_mean": 1671.9,
|
|
"valid_targets_min": 691
|
|
},
|
|
{
|
|
"epoch": 1.8522601984564497,
|
|
"grad_norm": 0.49365125619174294,
|
|
"learning_rate": 3.6791753524623456e-05,
|
|
"loss": 0.1174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06557995080947876,
|
|
"step": 3360,
|
|
"valid_targets_mean": 1879.5,
|
|
"valid_targets_min": 687
|
|
},
|
|
{
|
|
"epoch": 1.8550165380374861,
|
|
"grad_norm": 0.3823875017106786,
|
|
"learning_rate": 3.6776804265210963e-05,
|
|
"loss": 0.1378,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049259331077337265,
|
|
"step": 3365,
|
|
"valid_targets_mean": 3679.5,
|
|
"valid_targets_min": 2015
|
|
},
|
|
{
|
|
"epoch": 1.8577728776185225,
|
|
"grad_norm": 0.64112370598362,
|
|
"learning_rate": 3.676182330950641e-05,
|
|
"loss": 0.1201,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05985262617468834,
|
|
"step": 3370,
|
|
"valid_targets_mean": 1351.2,
|
|
"valid_targets_min": 520
|
|
},
|
|
{
|
|
"epoch": 1.8605292171995589,
|
|
"grad_norm": 0.29783270959740243,
|
|
"learning_rate": 3.67468106858132e-05,
|
|
"loss": 0.1108,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0397835299372673,
|
|
"step": 3375,
|
|
"valid_targets_mean": 4111.8,
|
|
"valid_targets_min": 466
|
|
},
|
|
{
|
|
"epoch": 1.8632855567805953,
|
|
"grad_norm": 0.26763669292526,
|
|
"learning_rate": 3.673176642249459e-05,
|
|
"loss": 0.1145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0310286246240139,
|
|
"step": 3380,
|
|
"valid_targets_mean": 5544.5,
|
|
"valid_targets_min": 4219
|
|
},
|
|
{
|
|
"epoch": 1.8660418963616316,
|
|
"grad_norm": 0.3144262250132265,
|
|
"learning_rate": 3.671669054797358e-05,
|
|
"loss": 0.1032,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.040369194000959396,
|
|
"step": 3385,
|
|
"valid_targets_mean": 3800.1,
|
|
"valid_targets_min": 456
|
|
},
|
|
{
|
|
"epoch": 1.868798235942668,
|
|
"grad_norm": 0.31493514517800114,
|
|
"learning_rate": 3.670158309073291e-05,
|
|
"loss": 0.1016,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.048036083579063416,
|
|
"step": 3390,
|
|
"valid_targets_mean": 3487.1,
|
|
"valid_targets_min": 976
|
|
},
|
|
{
|
|
"epoch": 1.8715545755237044,
|
|
"grad_norm": 0.37615471477286844,
|
|
"learning_rate": 3.6686444079314995e-05,
|
|
"loss": 0.1128,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06550219655036926,
|
|
"step": 3395,
|
|
"valid_targets_mean": 2431.6,
|
|
"valid_targets_min": 591
|
|
},
|
|
{
|
|
"epoch": 1.8743109151047408,
|
|
"grad_norm": 0.41367045504545585,
|
|
"learning_rate": 3.667127354232185e-05,
|
|
"loss": 0.1218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06726361811161041,
|
|
"step": 3400,
|
|
"valid_targets_mean": 3660.8,
|
|
"valid_targets_min": 2753
|
|
},
|
|
{
|
|
"epoch": 1.8770672546857772,
|
|
"grad_norm": 0.45022929162345987,
|
|
"learning_rate": 3.6656071508415064e-05,
|
|
"loss": 0.1209,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06036003679037094,
|
|
"step": 3405,
|
|
"valid_targets_mean": 3690.9,
|
|
"valid_targets_min": 2892
|
|
},
|
|
{
|
|
"epoch": 1.8798235942668136,
|
|
"grad_norm": 0.3589709233578578,
|
|
"learning_rate": 3.6640838006315715e-05,
|
|
"loss": 0.1098,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04646208509802818,
|
|
"step": 3410,
|
|
"valid_targets_mean": 3447.0,
|
|
"valid_targets_min": 1093
|
|
},
|
|
{
|
|
"epoch": 1.88257993384785,
|
|
"grad_norm": 0.39959418555970533,
|
|
"learning_rate": 3.6625573064804366e-05,
|
|
"loss": 0.1073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06968662142753601,
|
|
"step": 3415,
|
|
"valid_targets_mean": 3798.8,
|
|
"valid_targets_min": 1861
|
|
},
|
|
{
|
|
"epoch": 1.8853362734288863,
|
|
"grad_norm": 0.34542567884524256,
|
|
"learning_rate": 3.661027671272094e-05,
|
|
"loss": 0.1221,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04565516859292984,
|
|
"step": 3420,
|
|
"valid_targets_mean": 3005.0,
|
|
"valid_targets_min": 720
|
|
},
|
|
{
|
|
"epoch": 1.8880926130099227,
|
|
"grad_norm": 0.2856138590115849,
|
|
"learning_rate": 3.659494897896473e-05,
|
|
"loss": 0.0926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04063495621085167,
|
|
"step": 3425,
|
|
"valid_targets_mean": 3797.5,
|
|
"valid_targets_min": 2761
|
|
},
|
|
{
|
|
"epoch": 1.890848952590959,
|
|
"grad_norm": 0.4362649432975086,
|
|
"learning_rate": 3.65795898924943e-05,
|
|
"loss": 0.1234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04981014505028725,
|
|
"step": 3430,
|
|
"valid_targets_mean": 2420.0,
|
|
"valid_targets_min": 588
|
|
},
|
|
{
|
|
"epoch": 1.8936052921719955,
|
|
"grad_norm": 0.36481077444315557,
|
|
"learning_rate": 3.656419948232747e-05,
|
|
"loss": 0.1163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05692782625555992,
|
|
"step": 3435,
|
|
"valid_targets_mean": 3433.5,
|
|
"valid_targets_min": 596
|
|
},
|
|
{
|
|
"epoch": 1.8963616317530319,
|
|
"grad_norm": 0.42640460805306746,
|
|
"learning_rate": 3.654877777754123e-05,
|
|
"loss": 0.1299,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06541642546653748,
|
|
"step": 3440,
|
|
"valid_targets_mean": 3530.0,
|
|
"valid_targets_min": 1636
|
|
},
|
|
{
|
|
"epoch": 1.8991179713340682,
|
|
"grad_norm": 0.39245561322063804,
|
|
"learning_rate": 3.653332480727168e-05,
|
|
"loss": 0.1185,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06083942949771881,
|
|
"step": 3445,
|
|
"valid_targets_mean": 1950.8,
|
|
"valid_targets_min": 454
|
|
},
|
|
{
|
|
"epoch": 1.9018743109151046,
|
|
"grad_norm": 0.7245980218830017,
|
|
"learning_rate": 3.651784060071401e-05,
|
|
"loss": 0.1298,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0647321417927742,
|
|
"step": 3450,
|
|
"valid_targets_mean": 819.5,
|
|
"valid_targets_min": 609
|
|
},
|
|
{
|
|
"epoch": 1.904630650496141,
|
|
"grad_norm": 0.5128065582900938,
|
|
"learning_rate": 3.6502325187122425e-05,
|
|
"loss": 0.1175,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09567753970623016,
|
|
"step": 3455,
|
|
"valid_targets_mean": 2364.8,
|
|
"valid_targets_min": 893
|
|
},
|
|
{
|
|
"epoch": 1.9073869900771774,
|
|
"grad_norm": 0.42149361038206046,
|
|
"learning_rate": 3.648677859581006e-05,
|
|
"loss": 0.1226,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04841095209121704,
|
|
"step": 3460,
|
|
"valid_targets_mean": 3029.5,
|
|
"valid_targets_min": 1140
|
|
},
|
|
{
|
|
"epoch": 1.9101433296582138,
|
|
"grad_norm": 0.41664435830439317,
|
|
"learning_rate": 3.6471200856148984e-05,
|
|
"loss": 0.1153,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0723855048418045,
|
|
"step": 3465,
|
|
"valid_targets_mean": 3029.6,
|
|
"valid_targets_min": 1502
|
|
},
|
|
{
|
|
"epoch": 1.9128996692392501,
|
|
"grad_norm": 0.5912780695164739,
|
|
"learning_rate": 3.6455591997570115e-05,
|
|
"loss": 0.105,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06043563038110733,
|
|
"step": 3470,
|
|
"valid_targets_mean": 1295.8,
|
|
"valid_targets_min": 624
|
|
},
|
|
{
|
|
"epoch": 1.9156560088202865,
|
|
"grad_norm": 0.5440223761607231,
|
|
"learning_rate": 3.643995204956315e-05,
|
|
"loss": 0.1269,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07425834983587265,
|
|
"step": 3475,
|
|
"valid_targets_mean": 1481.6,
|
|
"valid_targets_min": 687
|
|
},
|
|
{
|
|
"epoch": 1.918412348401323,
|
|
"grad_norm": 0.30079732842974444,
|
|
"learning_rate": 3.6424281041676526e-05,
|
|
"loss": 0.1109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03549587354063988,
|
|
"step": 3480,
|
|
"valid_targets_mean": 3963.5,
|
|
"valid_targets_min": 3651
|
|
},
|
|
{
|
|
"epoch": 1.9211686879823593,
|
|
"grad_norm": 0.35952748967172854,
|
|
"learning_rate": 3.6408579003517347e-05,
|
|
"loss": 0.1005,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06538692116737366,
|
|
"step": 3485,
|
|
"valid_targets_mean": 5041.0,
|
|
"valid_targets_min": 3813
|
|
},
|
|
{
|
|
"epoch": 1.9239250275633957,
|
|
"grad_norm": 0.32031918727750436,
|
|
"learning_rate": 3.639284596475138e-05,
|
|
"loss": 0.1104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04800261929631233,
|
|
"step": 3490,
|
|
"valid_targets_mean": 2392.4,
|
|
"valid_targets_min": 757
|
|
},
|
|
{
|
|
"epoch": 1.926681367144432,
|
|
"grad_norm": 0.35946139159634866,
|
|
"learning_rate": 3.637708195510293e-05,
|
|
"loss": 0.0977,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05562688410282135,
|
|
"step": 3495,
|
|
"valid_targets_mean": 3566.1,
|
|
"valid_targets_min": 664
|
|
},
|
|
{
|
|
"epoch": 1.9294377067254684,
|
|
"grad_norm": 0.36091930506798275,
|
|
"learning_rate": 3.636128700435481e-05,
|
|
"loss": 0.0933,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07015244662761688,
|
|
"step": 3500,
|
|
"valid_targets_mean": 4990.0,
|
|
"valid_targets_min": 4088
|
|
},
|
|
{
|
|
"epoch": 1.9321940463065048,
|
|
"grad_norm": 0.3437621885173031,
|
|
"learning_rate": 3.634546114234833e-05,
|
|
"loss": 0.1163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047505348920822144,
|
|
"step": 3505,
|
|
"valid_targets_mean": 3248.4,
|
|
"valid_targets_min": 779
|
|
},
|
|
{
|
|
"epoch": 1.9349503858875412,
|
|
"grad_norm": 0.45243250975881016,
|
|
"learning_rate": 3.632960439898315e-05,
|
|
"loss": 0.1287,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05756726488471031,
|
|
"step": 3510,
|
|
"valid_targets_mean": 3469.2,
|
|
"valid_targets_min": 1084
|
|
},
|
|
{
|
|
"epoch": 1.9377067254685776,
|
|
"grad_norm": 0.5763553300386708,
|
|
"learning_rate": 3.631371680421732e-05,
|
|
"loss": 0.1096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06473484635353088,
|
|
"step": 3515,
|
|
"valid_targets_mean": 3227.9,
|
|
"valid_targets_min": 700
|
|
},
|
|
{
|
|
"epoch": 1.940463065049614,
|
|
"grad_norm": 0.30471419423607765,
|
|
"learning_rate": 3.6297798388067126e-05,
|
|
"loss": 0.1103,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04565959423780441,
|
|
"step": 3520,
|
|
"valid_targets_mean": 2815.9,
|
|
"valid_targets_min": 769
|
|
},
|
|
{
|
|
"epoch": 1.9432194046306503,
|
|
"grad_norm": 0.2901633878930167,
|
|
"learning_rate": 3.628184918060714e-05,
|
|
"loss": 0.1364,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04687845706939697,
|
|
"step": 3525,
|
|
"valid_targets_mean": 3541.8,
|
|
"valid_targets_min": 982
|
|
},
|
|
{
|
|
"epoch": 1.9459757442116867,
|
|
"grad_norm": 0.3783452876217284,
|
|
"learning_rate": 3.626586921197007e-05,
|
|
"loss": 0.1073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06292902678251266,
|
|
"step": 3530,
|
|
"valid_targets_mean": 3189.4,
|
|
"valid_targets_min": 1043
|
|
},
|
|
{
|
|
"epoch": 1.9487320837927231,
|
|
"grad_norm": 0.35550694279639933,
|
|
"learning_rate": 3.624985851234676e-05,
|
|
"loss": 0.1171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05560785531997681,
|
|
"step": 3535,
|
|
"valid_targets_mean": 3380.1,
|
|
"valid_targets_min": 1067
|
|
},
|
|
{
|
|
"epoch": 1.9514884233737595,
|
|
"grad_norm": 0.37699019374355947,
|
|
"learning_rate": 3.6233817111986096e-05,
|
|
"loss": 0.1114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08471566438674927,
|
|
"step": 3540,
|
|
"valid_targets_mean": 3092.6,
|
|
"valid_targets_min": 901
|
|
},
|
|
{
|
|
"epoch": 1.9542447629547959,
|
|
"grad_norm": 0.5681208968805245,
|
|
"learning_rate": 3.621774504119498e-05,
|
|
"loss": 0.1687,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06799792498350143,
|
|
"step": 3545,
|
|
"valid_targets_mean": 1287.2,
|
|
"valid_targets_min": 567
|
|
},
|
|
{
|
|
"epoch": 1.9570011025358323,
|
|
"grad_norm": 0.46478288014248637,
|
|
"learning_rate": 3.620164233033826e-05,
|
|
"loss": 0.1313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0767936259508133,
|
|
"step": 3550,
|
|
"valid_targets_mean": 3460.5,
|
|
"valid_targets_min": 831
|
|
},
|
|
{
|
|
"epoch": 1.9597574421168686,
|
|
"grad_norm": 0.4086240087391217,
|
|
"learning_rate": 3.618550900983867e-05,
|
|
"loss": 0.1078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04509853944182396,
|
|
"step": 3555,
|
|
"valid_targets_mean": 2148.4,
|
|
"valid_targets_min": 632
|
|
},
|
|
{
|
|
"epoch": 1.962513781697905,
|
|
"grad_norm": 0.3973058779884969,
|
|
"learning_rate": 3.616934511017677e-05,
|
|
"loss": 0.1089,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05371444672346115,
|
|
"step": 3560,
|
|
"valid_targets_mean": 2754.4,
|
|
"valid_targets_min": 1583
|
|
},
|
|
{
|
|
"epoch": 1.9652701212789414,
|
|
"grad_norm": 0.5088480009994887,
|
|
"learning_rate": 3.615315066189089e-05,
|
|
"loss": 0.1062,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04610132426023483,
|
|
"step": 3565,
|
|
"valid_targets_mean": 3267.4,
|
|
"valid_targets_min": 796
|
|
},
|
|
{
|
|
"epoch": 1.9680264608599778,
|
|
"grad_norm": 0.32070295483452965,
|
|
"learning_rate": 3.6136925695577085e-05,
|
|
"loss": 0.1012,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04962703585624695,
|
|
"step": 3570,
|
|
"valid_targets_mean": 3208.9,
|
|
"valid_targets_min": 952
|
|
},
|
|
{
|
|
"epoch": 1.9707828004410142,
|
|
"grad_norm": 0.4938026576221601,
|
|
"learning_rate": 3.612067024188907e-05,
|
|
"loss": 0.1396,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06726624071598053,
|
|
"step": 3575,
|
|
"valid_targets_mean": 2115.5,
|
|
"valid_targets_min": 885
|
|
},
|
|
{
|
|
"epoch": 1.9735391400220506,
|
|
"grad_norm": 0.4040825932658148,
|
|
"learning_rate": 3.6104384331538144e-05,
|
|
"loss": 0.0962,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05516896769404411,
|
|
"step": 3580,
|
|
"valid_targets_mean": 3298.5,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 1.976295479603087,
|
|
"grad_norm": 0.4427170595115168,
|
|
"learning_rate": 3.608806799529317e-05,
|
|
"loss": 0.1078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05540916323661804,
|
|
"step": 3585,
|
|
"valid_targets_mean": 2170.4,
|
|
"valid_targets_min": 782
|
|
},
|
|
{
|
|
"epoch": 1.9790518191841233,
|
|
"grad_norm": 0.43389766417091813,
|
|
"learning_rate": 3.607172126398046e-05,
|
|
"loss": 0.1098,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06508636474609375,
|
|
"step": 3590,
|
|
"valid_targets_mean": 3052.4,
|
|
"valid_targets_min": 2563
|
|
},
|
|
{
|
|
"epoch": 1.98180815876516,
|
|
"grad_norm": 0.2870379058742612,
|
|
"learning_rate": 3.60553441684838e-05,
|
|
"loss": 0.109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05260235816240311,
|
|
"step": 3595,
|
|
"valid_targets_mean": 3972.9,
|
|
"valid_targets_min": 2264
|
|
},
|
|
{
|
|
"epoch": 1.9845644983461963,
|
|
"grad_norm": 0.36443529412119263,
|
|
"learning_rate": 3.603893673974429e-05,
|
|
"loss": 0.0956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06760720908641815,
|
|
"step": 3600,
|
|
"valid_targets_mean": 4005.2,
|
|
"valid_targets_min": 2423
|
|
},
|
|
{
|
|
"epoch": 1.9873208379272327,
|
|
"grad_norm": 0.34125652061122536,
|
|
"learning_rate": 3.6022499008760374e-05,
|
|
"loss": 0.1059,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06714841723442078,
|
|
"step": 3605,
|
|
"valid_targets_mean": 4350.4,
|
|
"valid_targets_min": 2545
|
|
},
|
|
{
|
|
"epoch": 1.990077177508269,
|
|
"grad_norm": 0.5141021878966751,
|
|
"learning_rate": 3.600603100658773e-05,
|
|
"loss": 0.1422,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09961037337779999,
|
|
"step": 3610,
|
|
"valid_targets_mean": 2195.4,
|
|
"valid_targets_min": 1523
|
|
},
|
|
{
|
|
"epoch": 1.9928335170893055,
|
|
"grad_norm": 0.30275818367539886,
|
|
"learning_rate": 3.5989532764339254e-05,
|
|
"loss": 0.163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03203801065683365,
|
|
"step": 3615,
|
|
"valid_targets_mean": 3601.6,
|
|
"valid_targets_min": 1092
|
|
},
|
|
{
|
|
"epoch": 1.9955898566703418,
|
|
"grad_norm": 0.31067005020170935,
|
|
"learning_rate": 3.5973004313184923e-05,
|
|
"loss": 0.1083,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.040281422436237335,
|
|
"step": 3620,
|
|
"valid_targets_mean": 3158.4,
|
|
"valid_targets_min": 835
|
|
},
|
|
{
|
|
"epoch": 1.9983461962513782,
|
|
"grad_norm": 0.3168677761280718,
|
|
"learning_rate": 3.5956445684351837e-05,
|
|
"loss": 0.1037,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04685748368501663,
|
|
"step": 3625,
|
|
"valid_targets_mean": 3701.5,
|
|
"valid_targets_min": 2543
|
|
},
|
|
{
|
|
"epoch": 2.0011025358324144,
|
|
"grad_norm": 0.4212588233617524,
|
|
"learning_rate": 3.5939856909124085e-05,
|
|
"loss": 0.1925,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10481789708137512,
|
|
"step": 3630,
|
|
"valid_targets_mean": 8485.9,
|
|
"valid_targets_min": 6756
|
|
},
|
|
{
|
|
"epoch": 2.0038588754134508,
|
|
"grad_norm": 0.41887281787018227,
|
|
"learning_rate": 3.5923238018842715e-05,
|
|
"loss": 0.2355,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11623367667198181,
|
|
"step": 3635,
|
|
"valid_targets_mean": 6639.4,
|
|
"valid_targets_min": 5905
|
|
},
|
|
{
|
|
"epoch": 2.006615214994487,
|
|
"grad_norm": 0.4358461898832399,
|
|
"learning_rate": 3.590658904490568e-05,
|
|
"loss": 0.2232,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10299892723560333,
|
|
"step": 3640,
|
|
"valid_targets_mean": 4087.4,
|
|
"valid_targets_min": 160
|
|
},
|
|
{
|
|
"epoch": 2.0093715545755235,
|
|
"grad_norm": 0.3856839083602044,
|
|
"learning_rate": 3.588991001876775e-05,
|
|
"loss": 0.2196,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10012112557888031,
|
|
"step": 3645,
|
|
"valid_targets_mean": 7917.6,
|
|
"valid_targets_min": 5567
|
|
},
|
|
{
|
|
"epoch": 2.01212789415656,
|
|
"grad_norm": 0.3936310164571927,
|
|
"learning_rate": 3.58732009719405e-05,
|
|
"loss": 0.21,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1125057265162468,
|
|
"step": 3650,
|
|
"valid_targets_mean": 7361.2,
|
|
"valid_targets_min": 5601
|
|
},
|
|
{
|
|
"epoch": 2.0148842337375963,
|
|
"grad_norm": 0.37280111358408574,
|
|
"learning_rate": 3.5856461935992194e-05,
|
|
"loss": 0.2134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10876008123159409,
|
|
"step": 3655,
|
|
"valid_targets_mean": 6850.4,
|
|
"valid_targets_min": 5085
|
|
},
|
|
{
|
|
"epoch": 2.0176405733186327,
|
|
"grad_norm": 0.4106493040988781,
|
|
"learning_rate": 3.583969294254779e-05,
|
|
"loss": 0.2306,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11171063035726547,
|
|
"step": 3660,
|
|
"valid_targets_mean": 6890.8,
|
|
"valid_targets_min": 5057
|
|
},
|
|
{
|
|
"epoch": 2.020396912899669,
|
|
"grad_norm": 0.39387194325762725,
|
|
"learning_rate": 3.58228940232888e-05,
|
|
"loss": 0.2246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09857937693595886,
|
|
"step": 3665,
|
|
"valid_targets_mean": 6346.6,
|
|
"valid_targets_min": 3920
|
|
},
|
|
{
|
|
"epoch": 2.0231532524807054,
|
|
"grad_norm": 0.41439376356362845,
|
|
"learning_rate": 3.580606520995331e-05,
|
|
"loss": 0.2107,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10577057301998138,
|
|
"step": 3670,
|
|
"valid_targets_mean": 6362.6,
|
|
"valid_targets_min": 4343
|
|
},
|
|
{
|
|
"epoch": 2.025909592061742,
|
|
"grad_norm": 0.45805655886947205,
|
|
"learning_rate": 3.578920653433588e-05,
|
|
"loss": 0.2052,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11995893716812134,
|
|
"step": 3675,
|
|
"valid_targets_mean": 5179.8,
|
|
"valid_targets_min": 136
|
|
},
|
|
{
|
|
"epoch": 2.028665931642778,
|
|
"grad_norm": 0.42553832390890756,
|
|
"learning_rate": 3.5772318028287464e-05,
|
|
"loss": 0.2093,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09376346319913864,
|
|
"step": 3680,
|
|
"valid_targets_mean": 7313.2,
|
|
"valid_targets_min": 5219
|
|
},
|
|
{
|
|
"epoch": 2.0314222712238146,
|
|
"grad_norm": 0.34491699481380983,
|
|
"learning_rate": 3.5755399723715396e-05,
|
|
"loss": 0.2031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08826358616352081,
|
|
"step": 3685,
|
|
"valid_targets_mean": 7894.1,
|
|
"valid_targets_min": 5279
|
|
},
|
|
{
|
|
"epoch": 2.034178610804851,
|
|
"grad_norm": 0.3779707831317998,
|
|
"learning_rate": 3.573845165258331e-05,
|
|
"loss": 0.2123,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10727985203266144,
|
|
"step": 3690,
|
|
"valid_targets_mean": 7899.9,
|
|
"valid_targets_min": 5983
|
|
},
|
|
{
|
|
"epoch": 2.0369349503858873,
|
|
"grad_norm": 0.35110152494277064,
|
|
"learning_rate": 3.572147384691107e-05,
|
|
"loss": 0.2018,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09551440924406052,
|
|
"step": 3695,
|
|
"valid_targets_mean": 7612.1,
|
|
"valid_targets_min": 5224
|
|
},
|
|
{
|
|
"epoch": 2.0396912899669237,
|
|
"grad_norm": 0.3578438268926307,
|
|
"learning_rate": 3.5704466338774714e-05,
|
|
"loss": 0.2006,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10048002004623413,
|
|
"step": 3700,
|
|
"valid_targets_mean": 7219.4,
|
|
"valid_targets_min": 5185
|
|
},
|
|
{
|
|
"epoch": 2.04244762954796,
|
|
"grad_norm": 0.38243034080119537,
|
|
"learning_rate": 3.5687429160306405e-05,
|
|
"loss": 0.2086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11481879651546478,
|
|
"step": 3705,
|
|
"valid_targets_mean": 7588.1,
|
|
"valid_targets_min": 5062
|
|
},
|
|
{
|
|
"epoch": 2.0452039691289965,
|
|
"grad_norm": 0.36214178058092517,
|
|
"learning_rate": 3.567036234369435e-05,
|
|
"loss": 0.2023,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09711915254592896,
|
|
"step": 3710,
|
|
"valid_targets_mean": 7702.1,
|
|
"valid_targets_min": 5810
|
|
},
|
|
{
|
|
"epoch": 2.047960308710033,
|
|
"grad_norm": 0.34297102356020653,
|
|
"learning_rate": 3.565326592118277e-05,
|
|
"loss": 0.2044,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10361095517873764,
|
|
"step": 3715,
|
|
"valid_targets_mean": 7926.0,
|
|
"valid_targets_min": 5470
|
|
},
|
|
{
|
|
"epoch": 2.0507166482910693,
|
|
"grad_norm": 0.38370976690611414,
|
|
"learning_rate": 3.563613992507181e-05,
|
|
"loss": 0.2211,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10950224846601486,
|
|
"step": 3720,
|
|
"valid_targets_mean": 7969.2,
|
|
"valid_targets_min": 5422
|
|
},
|
|
{
|
|
"epoch": 2.0534729878721056,
|
|
"grad_norm": 0.47154918456850187,
|
|
"learning_rate": 3.5618984387717475e-05,
|
|
"loss": 0.225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07937896251678467,
|
|
"step": 3725,
|
|
"valid_targets_mean": 2969.4,
|
|
"valid_targets_min": 1886
|
|
},
|
|
{
|
|
"epoch": 2.056229327453142,
|
|
"grad_norm": 0.37479992032712517,
|
|
"learning_rate": 3.5601799341531605e-05,
|
|
"loss": 0.1838,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08824830502271652,
|
|
"step": 3730,
|
|
"valid_targets_mean": 6684.6,
|
|
"valid_targets_min": 4885
|
|
},
|
|
{
|
|
"epoch": 2.0589856670341784,
|
|
"grad_norm": 0.3671608217110651,
|
|
"learning_rate": 3.558458481898178e-05,
|
|
"loss": 0.2015,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11002112925052643,
|
|
"step": 3735,
|
|
"valid_targets_mean": 7565.5,
|
|
"valid_targets_min": 5456
|
|
},
|
|
{
|
|
"epoch": 2.061742006615215,
|
|
"grad_norm": 0.38784491959354267,
|
|
"learning_rate": 3.556734085259127e-05,
|
|
"loss": 0.2041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10963855683803558,
|
|
"step": 3740,
|
|
"valid_targets_mean": 7017.6,
|
|
"valid_targets_min": 5547
|
|
},
|
|
{
|
|
"epoch": 2.064498346196251,
|
|
"grad_norm": 0.3772818192873377,
|
|
"learning_rate": 3.555006747493896e-05,
|
|
"loss": 0.2098,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10144318640232086,
|
|
"step": 3745,
|
|
"valid_targets_mean": 7078.5,
|
|
"valid_targets_min": 4931
|
|
},
|
|
{
|
|
"epoch": 2.0672546857772875,
|
|
"grad_norm": 0.3877922614739497,
|
|
"learning_rate": 3.553276471865934e-05,
|
|
"loss": 0.206,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10357306897640228,
|
|
"step": 3750,
|
|
"valid_targets_mean": 6780.6,
|
|
"valid_targets_min": 5634
|
|
},
|
|
{
|
|
"epoch": 2.070011025358324,
|
|
"grad_norm": 0.35170912232664525,
|
|
"learning_rate": 3.551543261644236e-05,
|
|
"loss": 0.1973,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0956396609544754,
|
|
"step": 3755,
|
|
"valid_targets_mean": 8042.4,
|
|
"valid_targets_min": 5492
|
|
},
|
|
{
|
|
"epoch": 2.0727673649393603,
|
|
"grad_norm": 0.379233182423579,
|
|
"learning_rate": 3.549807120103343e-05,
|
|
"loss": 0.1783,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08451464027166367,
|
|
"step": 3760,
|
|
"valid_targets_mean": 6528.6,
|
|
"valid_targets_min": 3602
|
|
},
|
|
{
|
|
"epoch": 2.075523704520397,
|
|
"grad_norm": 0.3575563871861672,
|
|
"learning_rate": 3.548068050523336e-05,
|
|
"loss": 0.1831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09630825370550156,
|
|
"step": 3765,
|
|
"valid_targets_mean": 6736.2,
|
|
"valid_targets_min": 4945
|
|
},
|
|
{
|
|
"epoch": 2.078280044101433,
|
|
"grad_norm": 0.3983259016756161,
|
|
"learning_rate": 3.5463260561898246e-05,
|
|
"loss": 0.1954,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09854140877723694,
|
|
"step": 3770,
|
|
"valid_targets_mean": 6115.0,
|
|
"valid_targets_min": 4885
|
|
},
|
|
{
|
|
"epoch": 2.08103638368247,
|
|
"grad_norm": 0.5449771917535857,
|
|
"learning_rate": 3.544581140393947e-05,
|
|
"loss": 0.2033,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13940992951393127,
|
|
"step": 3775,
|
|
"valid_targets_mean": 4565.0,
|
|
"valid_targets_min": 214
|
|
},
|
|
{
|
|
"epoch": 2.083792723263506,
|
|
"grad_norm": 0.35705180724211405,
|
|
"learning_rate": 3.54283330643236e-05,
|
|
"loss": 0.2098,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10565194487571716,
|
|
"step": 3780,
|
|
"valid_targets_mean": 7283.1,
|
|
"valid_targets_min": 4916
|
|
},
|
|
{
|
|
"epoch": 2.0865490628445427,
|
|
"grad_norm": 0.43251827022496514,
|
|
"learning_rate": 3.541082557607231e-05,
|
|
"loss": 0.2235,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10721534490585327,
|
|
"step": 3785,
|
|
"valid_targets_mean": 6411.9,
|
|
"valid_targets_min": 4488
|
|
},
|
|
{
|
|
"epoch": 2.0893054024255786,
|
|
"grad_norm": 0.38689060342529924,
|
|
"learning_rate": 3.53932889722624e-05,
|
|
"loss": 0.2154,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11651946604251862,
|
|
"step": 3790,
|
|
"valid_targets_mean": 7681.2,
|
|
"valid_targets_min": 5767
|
|
},
|
|
{
|
|
"epoch": 2.0920617420066154,
|
|
"grad_norm": 0.34835742098125144,
|
|
"learning_rate": 3.5375723286025615e-05,
|
|
"loss": 0.2042,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09989902377128601,
|
|
"step": 3795,
|
|
"valid_targets_mean": 7875.4,
|
|
"valid_targets_min": 5611
|
|
},
|
|
{
|
|
"epoch": 2.0948180815876514,
|
|
"grad_norm": 0.37665009378856784,
|
|
"learning_rate": 3.53581285505487e-05,
|
|
"loss": 0.2055,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11268219351768494,
|
|
"step": 3800,
|
|
"valid_targets_mean": 7596.0,
|
|
"valid_targets_min": 5711
|
|
},
|
|
{
|
|
"epoch": 2.097574421168688,
|
|
"grad_norm": 0.3560667231861221,
|
|
"learning_rate": 3.534050479907324e-05,
|
|
"loss": 0.2098,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11904726177453995,
|
|
"step": 3805,
|
|
"valid_targets_mean": 8694.8,
|
|
"valid_targets_min": 5112
|
|
},
|
|
{
|
|
"epoch": 2.1003307607497246,
|
|
"grad_norm": 0.3991359271299437,
|
|
"learning_rate": 3.532285206489567e-05,
|
|
"loss": 0.2067,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11471976339817047,
|
|
"step": 3810,
|
|
"valid_targets_mean": 6940.6,
|
|
"valid_targets_min": 6063
|
|
},
|
|
{
|
|
"epoch": 2.103087100330761,
|
|
"grad_norm": 0.3444889252742341,
|
|
"learning_rate": 3.5305170381367165e-05,
|
|
"loss": 0.2093,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09647957980632782,
|
|
"step": 3815,
|
|
"valid_targets_mean": 7110.2,
|
|
"valid_targets_min": 4811
|
|
},
|
|
{
|
|
"epoch": 2.1058434399117973,
|
|
"grad_norm": 0.3732330235528218,
|
|
"learning_rate": 3.528745978189361e-05,
|
|
"loss": 0.2078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09664006531238556,
|
|
"step": 3820,
|
|
"valid_targets_mean": 6677.2,
|
|
"valid_targets_min": 5076
|
|
},
|
|
{
|
|
"epoch": 2.1085997794928337,
|
|
"grad_norm": 0.4009330105552859,
|
|
"learning_rate": 3.52697202999355e-05,
|
|
"loss": 0.2117,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10417863726615906,
|
|
"step": 3825,
|
|
"valid_targets_mean": 5899.2,
|
|
"valid_targets_min": 4676
|
|
},
|
|
{
|
|
"epoch": 2.11135611907387,
|
|
"grad_norm": 0.3549218659661393,
|
|
"learning_rate": 3.52519519690079e-05,
|
|
"loss": 0.203,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09577707946300507,
|
|
"step": 3830,
|
|
"valid_targets_mean": 7114.4,
|
|
"valid_targets_min": 5248
|
|
},
|
|
{
|
|
"epoch": 2.1141124586549065,
|
|
"grad_norm": 0.641007265843844,
|
|
"learning_rate": 3.5234154822680394e-05,
|
|
"loss": 0.198,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06841850280761719,
|
|
"step": 3835,
|
|
"valid_targets_mean": 1548.5,
|
|
"valid_targets_min": 587
|
|
},
|
|
{
|
|
"epoch": 2.116868798235943,
|
|
"grad_norm": 0.4724209741921488,
|
|
"learning_rate": 3.5216328894577e-05,
|
|
"loss": 0.2012,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11833006888628006,
|
|
"step": 3840,
|
|
"valid_targets_mean": 7040.5,
|
|
"valid_targets_min": 4733
|
|
},
|
|
{
|
|
"epoch": 2.1196251378169793,
|
|
"grad_norm": 0.373486758383562,
|
|
"learning_rate": 3.5198474218376106e-05,
|
|
"loss": 0.2035,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10229101032018661,
|
|
"step": 3845,
|
|
"valid_targets_mean": 6161.2,
|
|
"valid_targets_min": 5555
|
|
},
|
|
{
|
|
"epoch": 2.1223814773980156,
|
|
"grad_norm": 0.38729171720930483,
|
|
"learning_rate": 3.518059082781043e-05,
|
|
"loss": 0.2059,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10758139193058014,
|
|
"step": 3850,
|
|
"valid_targets_mean": 6580.2,
|
|
"valid_targets_min": 4862
|
|
},
|
|
{
|
|
"epoch": 2.125137816979052,
|
|
"grad_norm": 0.3607142870026607,
|
|
"learning_rate": 3.516267875666692e-05,
|
|
"loss": 0.1988,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09489966928958893,
|
|
"step": 3855,
|
|
"valid_targets_mean": 6575.2,
|
|
"valid_targets_min": 3768
|
|
},
|
|
{
|
|
"epoch": 2.1278941565600884,
|
|
"grad_norm": 0.3843030445248785,
|
|
"learning_rate": 3.514473803878672e-05,
|
|
"loss": 0.2044,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10326842963695526,
|
|
"step": 3860,
|
|
"valid_targets_mean": 7154.8,
|
|
"valid_targets_min": 4858
|
|
},
|
|
{
|
|
"epoch": 2.130650496141125,
|
|
"grad_norm": 0.38836112817156543,
|
|
"learning_rate": 3.51267687080651e-05,
|
|
"loss": 0.2071,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10299030691385269,
|
|
"step": 3865,
|
|
"valid_targets_mean": 6462.0,
|
|
"valid_targets_min": 4583
|
|
},
|
|
{
|
|
"epoch": 2.133406835722161,
|
|
"grad_norm": 0.39200106811552965,
|
|
"learning_rate": 3.5108770798451376e-05,
|
|
"loss": 0.2095,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11022090166807175,
|
|
"step": 3870,
|
|
"valid_targets_mean": 6308.5,
|
|
"valid_targets_min": 4766
|
|
},
|
|
{
|
|
"epoch": 2.1361631753031975,
|
|
"grad_norm": 0.7857048867525476,
|
|
"learning_rate": 3.5090744343948875e-05,
|
|
"loss": 0.1757,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11766932904720306,
|
|
"step": 3875,
|
|
"valid_targets_mean": 4534.4,
|
|
"valid_targets_min": 147
|
|
},
|
|
{
|
|
"epoch": 2.138919514884234,
|
|
"grad_norm": 0.38172483677448826,
|
|
"learning_rate": 3.507268937861484e-05,
|
|
"loss": 0.2053,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11243122816085815,
|
|
"step": 3880,
|
|
"valid_targets_mean": 6778.1,
|
|
"valid_targets_min": 5132
|
|
},
|
|
{
|
|
"epoch": 2.1416758544652703,
|
|
"grad_norm": 0.3838778055931196,
|
|
"learning_rate": 3.505460593656039e-05,
|
|
"loss": 0.1987,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1024523451924324,
|
|
"step": 3885,
|
|
"valid_targets_mean": 6979.8,
|
|
"valid_targets_min": 4942
|
|
},
|
|
{
|
|
"epoch": 2.1444321940463067,
|
|
"grad_norm": 0.4416581858200025,
|
|
"learning_rate": 3.5036494051950414e-05,
|
|
"loss": 0.2061,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09254634380340576,
|
|
"step": 3890,
|
|
"valid_targets_mean": 5762.2,
|
|
"valid_targets_min": 4777
|
|
},
|
|
{
|
|
"epoch": 2.147188533627343,
|
|
"grad_norm": 0.3950302226200239,
|
|
"learning_rate": 3.5018353759003586e-05,
|
|
"loss": 0.1943,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09175986796617508,
|
|
"step": 3895,
|
|
"valid_targets_mean": 6373.9,
|
|
"valid_targets_min": 4810
|
|
},
|
|
{
|
|
"epoch": 2.1499448732083795,
|
|
"grad_norm": 0.38937456586576313,
|
|
"learning_rate": 3.500018509199222e-05,
|
|
"loss": 0.1997,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10262373089790344,
|
|
"step": 3900,
|
|
"valid_targets_mean": 6151.0,
|
|
"valid_targets_min": 5084
|
|
},
|
|
{
|
|
"epoch": 2.152701212789416,
|
|
"grad_norm": 0.36281242677233017,
|
|
"learning_rate": 3.4981988085242243e-05,
|
|
"loss": 0.1998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09190413355827332,
|
|
"step": 3905,
|
|
"valid_targets_mean": 6133.5,
|
|
"valid_targets_min": 5736
|
|
},
|
|
{
|
|
"epoch": 2.1554575523704522,
|
|
"grad_norm": 0.7202031965609307,
|
|
"learning_rate": 3.4963762773133126e-05,
|
|
"loss": 0.2264,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08337302505970001,
|
|
"step": 3910,
|
|
"valid_targets_mean": 1423.2,
|
|
"valid_targets_min": 1061
|
|
},
|
|
{
|
|
"epoch": 2.1582138919514886,
|
|
"grad_norm": 0.8847552732561023,
|
|
"learning_rate": 3.494550919009782e-05,
|
|
"loss": 0.2031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09689651429653168,
|
|
"step": 3915,
|
|
"valid_targets_mean": 1656.5,
|
|
"valid_targets_min": 959
|
|
},
|
|
{
|
|
"epoch": 2.160970231532525,
|
|
"grad_norm": 1.1044139324714701,
|
|
"learning_rate": 3.4927227370622675e-05,
|
|
"loss": 0.1985,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09642351418733597,
|
|
"step": 3920,
|
|
"valid_targets_mean": 1471.4,
|
|
"valid_targets_min": 932
|
|
},
|
|
{
|
|
"epoch": 2.1637265711135614,
|
|
"grad_norm": 0.7568271874688526,
|
|
"learning_rate": 3.49089173492474e-05,
|
|
"loss": 0.1952,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09728054702281952,
|
|
"step": 3925,
|
|
"valid_targets_mean": 1473.5,
|
|
"valid_targets_min": 880
|
|
},
|
|
{
|
|
"epoch": 2.1664829106945978,
|
|
"grad_norm": 0.9885581808660554,
|
|
"learning_rate": 3.4890579160564985e-05,
|
|
"loss": 0.1998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11882907152175903,
|
|
"step": 3930,
|
|
"valid_targets_mean": 1636.2,
|
|
"valid_targets_min": 868
|
|
},
|
|
{
|
|
"epoch": 2.169239250275634,
|
|
"grad_norm": 0.7331110523840421,
|
|
"learning_rate": 3.487221283922164e-05,
|
|
"loss": 0.2001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10501580685377121,
|
|
"step": 3935,
|
|
"valid_targets_mean": 1703.0,
|
|
"valid_targets_min": 1089
|
|
},
|
|
{
|
|
"epoch": 2.1719955898566705,
|
|
"grad_norm": 0.7579381973458933,
|
|
"learning_rate": 3.485381841991671e-05,
|
|
"loss": 0.1911,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07788132131099701,
|
|
"step": 3940,
|
|
"valid_targets_mean": 1154.2,
|
|
"valid_targets_min": 758
|
|
},
|
|
{
|
|
"epoch": 2.174751929437707,
|
|
"grad_norm": 0.7756160060086923,
|
|
"learning_rate": 3.4835395937402636e-05,
|
|
"loss": 0.206,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09442643821239471,
|
|
"step": 3945,
|
|
"valid_targets_mean": 1547.6,
|
|
"valid_targets_min": 1238
|
|
},
|
|
{
|
|
"epoch": 2.1775082690187433,
|
|
"grad_norm": 0.7719238589534311,
|
|
"learning_rate": 3.481694542648488e-05,
|
|
"loss": 0.1901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08923796564340591,
|
|
"step": 3950,
|
|
"valid_targets_mean": 1532.5,
|
|
"valid_targets_min": 732
|
|
},
|
|
{
|
|
"epoch": 2.1802646085997797,
|
|
"grad_norm": 0.7038253896815054,
|
|
"learning_rate": 3.479846692202187e-05,
|
|
"loss": 0.1961,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.060153014957904816,
|
|
"step": 3955,
|
|
"valid_targets_mean": 1246.9,
|
|
"valid_targets_min": 834
|
|
},
|
|
{
|
|
"epoch": 2.183020948180816,
|
|
"grad_norm": 0.7205350836551213,
|
|
"learning_rate": 3.4779960458924876e-05,
|
|
"loss": 0.189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11061939597129822,
|
|
"step": 3960,
|
|
"valid_targets_mean": 1591.6,
|
|
"valid_targets_min": 1040
|
|
},
|
|
{
|
|
"epoch": 2.1857772877618524,
|
|
"grad_norm": 0.7843146346139855,
|
|
"learning_rate": 3.476142607215805e-05,
|
|
"loss": 0.2018,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09544992446899414,
|
|
"step": 3965,
|
|
"valid_targets_mean": 1641.1,
|
|
"valid_targets_min": 796
|
|
},
|
|
{
|
|
"epoch": 2.188533627342889,
|
|
"grad_norm": 0.747953655463711,
|
|
"learning_rate": 3.474286379673826e-05,
|
|
"loss": 0.1885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09579872339963913,
|
|
"step": 3970,
|
|
"valid_targets_mean": 1443.5,
|
|
"valid_targets_min": 569
|
|
},
|
|
{
|
|
"epoch": 2.191289966923925,
|
|
"grad_norm": 0.7305289766748679,
|
|
"learning_rate": 3.472427366773508e-05,
|
|
"loss": 0.1815,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07025370746850967,
|
|
"step": 3975,
|
|
"valid_targets_mean": 1161.8,
|
|
"valid_targets_min": 546
|
|
},
|
|
{
|
|
"epoch": 2.1940463065049616,
|
|
"grad_norm": 0.771689774256239,
|
|
"learning_rate": 3.47056557202707e-05,
|
|
"loss": 0.181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09436297416687012,
|
|
"step": 3980,
|
|
"valid_targets_mean": 1333.5,
|
|
"valid_targets_min": 681
|
|
},
|
|
{
|
|
"epoch": 2.196802646085998,
|
|
"grad_norm": 0.7303274823082866,
|
|
"learning_rate": 3.468700998951987e-05,
|
|
"loss": 0.1983,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09966295957565308,
|
|
"step": 3985,
|
|
"valid_targets_mean": 1627.6,
|
|
"valid_targets_min": 895
|
|
},
|
|
{
|
|
"epoch": 2.1995589856670343,
|
|
"grad_norm": 0.7874419602910386,
|
|
"learning_rate": 3.466833651070983e-05,
|
|
"loss": 0.1965,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1235324814915657,
|
|
"step": 3990,
|
|
"valid_targets_mean": 1801.4,
|
|
"valid_targets_min": 596
|
|
},
|
|
{
|
|
"epoch": 2.2023153252480707,
|
|
"grad_norm": 0.7691046803072038,
|
|
"learning_rate": 3.464963531912024e-05,
|
|
"loss": 0.1993,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08461252599954605,
|
|
"step": 3995,
|
|
"valid_targets_mean": 1286.4,
|
|
"valid_targets_min": 703
|
|
},
|
|
{
|
|
"epoch": 2.205071664829107,
|
|
"grad_norm": 0.7166877665484572,
|
|
"learning_rate": 3.463090645008313e-05,
|
|
"loss": 0.1911,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0612972229719162,
|
|
"step": 4000,
|
|
"valid_targets_mean": 1350.0,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 2.2078280044101435,
|
|
"grad_norm": 0.6785887913706472,
|
|
"learning_rate": 3.461214993898282e-05,
|
|
"loss": 0.1885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08955728262662888,
|
|
"step": 4005,
|
|
"valid_targets_mean": 1457.0,
|
|
"valid_targets_min": 818
|
|
},
|
|
{
|
|
"epoch": 2.21058434399118,
|
|
"grad_norm": 0.6959944336187041,
|
|
"learning_rate": 3.459336582125584e-05,
|
|
"loss": 0.1969,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07366833835840225,
|
|
"step": 4010,
|
|
"valid_targets_mean": 1221.6,
|
|
"valid_targets_min": 683
|
|
},
|
|
{
|
|
"epoch": 2.2133406835722162,
|
|
"grad_norm": 0.7534217251869129,
|
|
"learning_rate": 3.4574554132390884e-05,
|
|
"loss": 0.2026,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0990917831659317,
|
|
"step": 4015,
|
|
"valid_targets_mean": 1278.4,
|
|
"valid_targets_min": 709
|
|
},
|
|
{
|
|
"epoch": 2.2160970231532526,
|
|
"grad_norm": 0.7191084094610598,
|
|
"learning_rate": 3.455571490792873e-05,
|
|
"loss": 0.188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1019989401102066,
|
|
"step": 4020,
|
|
"valid_targets_mean": 1629.5,
|
|
"valid_targets_min": 1187
|
|
},
|
|
{
|
|
"epoch": 2.218853362734289,
|
|
"grad_norm": 0.7032244683283996,
|
|
"learning_rate": 3.45368481834622e-05,
|
|
"loss": 0.1839,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09727083891630173,
|
|
"step": 4025,
|
|
"valid_targets_mean": 1610.2,
|
|
"valid_targets_min": 771
|
|
},
|
|
{
|
|
"epoch": 2.2216097023153254,
|
|
"grad_norm": 0.6374024607067189,
|
|
"learning_rate": 3.451795399463605e-05,
|
|
"loss": 0.1871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09787138551473618,
|
|
"step": 4030,
|
|
"valid_targets_mean": 1840.9,
|
|
"valid_targets_min": 731
|
|
},
|
|
{
|
|
"epoch": 2.224366041896362,
|
|
"grad_norm": 0.7369218723137947,
|
|
"learning_rate": 3.4499032377146926e-05,
|
|
"loss": 0.1858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0969679057598114,
|
|
"step": 4035,
|
|
"valid_targets_mean": 1564.5,
|
|
"valid_targets_min": 1009
|
|
},
|
|
{
|
|
"epoch": 2.227122381477398,
|
|
"grad_norm": 0.7649879174923954,
|
|
"learning_rate": 3.448008336674331e-05,
|
|
"loss": 0.1906,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11178130656480789,
|
|
"step": 4040,
|
|
"valid_targets_mean": 1690.0,
|
|
"valid_targets_min": 802
|
|
},
|
|
{
|
|
"epoch": 2.2298787210584345,
|
|
"grad_norm": 0.7203774840201107,
|
|
"learning_rate": 3.446110699922541e-05,
|
|
"loss": 0.1896,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08316947519779205,
|
|
"step": 4045,
|
|
"valid_targets_mean": 1302.5,
|
|
"valid_targets_min": 781
|
|
},
|
|
{
|
|
"epoch": 2.232635060639471,
|
|
"grad_norm": 0.6830074566001085,
|
|
"learning_rate": 3.4442103310445166e-05,
|
|
"loss": 0.184,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09928393363952637,
|
|
"step": 4050,
|
|
"valid_targets_mean": 1592.2,
|
|
"valid_targets_min": 644
|
|
},
|
|
{
|
|
"epoch": 2.2353914002205073,
|
|
"grad_norm": 0.7018781679000964,
|
|
"learning_rate": 3.442307233630608e-05,
|
|
"loss": 0.1959,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13319098949432373,
|
|
"step": 4055,
|
|
"valid_targets_mean": 1812.5,
|
|
"valid_targets_min": 921
|
|
},
|
|
{
|
|
"epoch": 2.2381477398015437,
|
|
"grad_norm": 0.7342802503528026,
|
|
"learning_rate": 3.4404014112763236e-05,
|
|
"loss": 0.1916,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10575231909751892,
|
|
"step": 4060,
|
|
"valid_targets_mean": 1815.8,
|
|
"valid_targets_min": 795
|
|
},
|
|
{
|
|
"epoch": 2.24090407938258,
|
|
"grad_norm": 0.6841993780833887,
|
|
"learning_rate": 3.43849286758232e-05,
|
|
"loss": 0.1823,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09562225639820099,
|
|
"step": 4065,
|
|
"valid_targets_mean": 1536.6,
|
|
"valid_targets_min": 901
|
|
},
|
|
{
|
|
"epoch": 2.2436604189636165,
|
|
"grad_norm": 0.8138285124310337,
|
|
"learning_rate": 3.436581606154394e-05,
|
|
"loss": 0.1846,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11228308081626892,
|
|
"step": 4070,
|
|
"valid_targets_mean": 1455.1,
|
|
"valid_targets_min": 779
|
|
},
|
|
{
|
|
"epoch": 2.246416758544653,
|
|
"grad_norm": 0.6871591858989954,
|
|
"learning_rate": 3.434667630603478e-05,
|
|
"loss": 0.1796,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09701648354530334,
|
|
"step": 4075,
|
|
"valid_targets_mean": 1577.2,
|
|
"valid_targets_min": 780
|
|
},
|
|
{
|
|
"epoch": 2.249173098125689,
|
|
"grad_norm": 0.7094871278452979,
|
|
"learning_rate": 3.4327509445456325e-05,
|
|
"loss": 0.1824,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07242260873317719,
|
|
"step": 4080,
|
|
"valid_targets_mean": 1337.5,
|
|
"valid_targets_min": 659
|
|
},
|
|
{
|
|
"epoch": 2.2519294377067256,
|
|
"grad_norm": 0.7100778440634945,
|
|
"learning_rate": 3.430831551602038e-05,
|
|
"loss": 0.1848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08265615999698639,
|
|
"step": 4085,
|
|
"valid_targets_mean": 1265.5,
|
|
"valid_targets_min": 723
|
|
},
|
|
{
|
|
"epoch": 2.254685777287762,
|
|
"grad_norm": 0.7671674690052557,
|
|
"learning_rate": 3.4289094553989885e-05,
|
|
"loss": 0.1819,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10191851854324341,
|
|
"step": 4090,
|
|
"valid_targets_mean": 1505.5,
|
|
"valid_targets_min": 965
|
|
},
|
|
{
|
|
"epoch": 2.2574421168687984,
|
|
"grad_norm": 0.7476189897235228,
|
|
"learning_rate": 3.426984659567887e-05,
|
|
"loss": 0.1877,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08090505003929138,
|
|
"step": 4095,
|
|
"valid_targets_mean": 1322.6,
|
|
"valid_targets_min": 772
|
|
},
|
|
{
|
|
"epoch": 2.2601984564498347,
|
|
"grad_norm": 0.7199909774986637,
|
|
"learning_rate": 3.425057167745236e-05,
|
|
"loss": 0.1848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07850509881973267,
|
|
"step": 4100,
|
|
"valid_targets_mean": 1348.5,
|
|
"valid_targets_min": 806
|
|
},
|
|
{
|
|
"epoch": 2.262954796030871,
|
|
"grad_norm": 0.8317378823302307,
|
|
"learning_rate": 3.42312698357263e-05,
|
|
"loss": 0.1927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07383224368095398,
|
|
"step": 4105,
|
|
"valid_targets_mean": 1139.5,
|
|
"valid_targets_min": 781
|
|
},
|
|
{
|
|
"epoch": 2.2657111356119075,
|
|
"grad_norm": 0.654771864389965,
|
|
"learning_rate": 3.421194110696754e-05,
|
|
"loss": 0.1811,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10218144953250885,
|
|
"step": 4110,
|
|
"valid_targets_mean": 1843.2,
|
|
"valid_targets_min": 632
|
|
},
|
|
{
|
|
"epoch": 2.268467475192944,
|
|
"grad_norm": 0.6728139552098787,
|
|
"learning_rate": 3.419258552769369e-05,
|
|
"loss": 0.1955,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09376795589923859,
|
|
"step": 4115,
|
|
"valid_targets_mean": 1594.5,
|
|
"valid_targets_min": 571
|
|
},
|
|
{
|
|
"epoch": 2.2712238147739803,
|
|
"grad_norm": 0.6872084571665308,
|
|
"learning_rate": 3.41732031344731e-05,
|
|
"loss": 0.1885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10285481065511703,
|
|
"step": 4120,
|
|
"valid_targets_mean": 1987.6,
|
|
"valid_targets_min": 1101
|
|
},
|
|
{
|
|
"epoch": 2.2739801543550167,
|
|
"grad_norm": 0.7067896021556421,
|
|
"learning_rate": 3.4153793963924794e-05,
|
|
"loss": 0.1924,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07028794288635254,
|
|
"step": 4125,
|
|
"valid_targets_mean": 1234.9,
|
|
"valid_targets_min": 840
|
|
},
|
|
{
|
|
"epoch": 2.276736493936053,
|
|
"grad_norm": 0.7073346172998418,
|
|
"learning_rate": 3.413435805271836e-05,
|
|
"loss": 0.1809,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08970724046230316,
|
|
"step": 4130,
|
|
"valid_targets_mean": 1463.4,
|
|
"valid_targets_min": 954
|
|
},
|
|
{
|
|
"epoch": 2.2794928335170894,
|
|
"grad_norm": 0.7776019703736581,
|
|
"learning_rate": 3.411489543757394e-05,
|
|
"loss": 0.1891,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09646405279636383,
|
|
"step": 4135,
|
|
"valid_targets_mean": 1297.6,
|
|
"valid_targets_min": 854
|
|
},
|
|
{
|
|
"epoch": 2.282249173098126,
|
|
"grad_norm": 0.6394749922675212,
|
|
"learning_rate": 3.409540615526209e-05,
|
|
"loss": 0.1891,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09843998402357101,
|
|
"step": 4140,
|
|
"valid_targets_mean": 1385.2,
|
|
"valid_targets_min": 807
|
|
},
|
|
{
|
|
"epoch": 2.285005512679162,
|
|
"grad_norm": 0.7002617330180707,
|
|
"learning_rate": 3.407589024260378e-05,
|
|
"loss": 0.1749,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07928720116615295,
|
|
"step": 4145,
|
|
"valid_targets_mean": 1299.0,
|
|
"valid_targets_min": 674
|
|
},
|
|
{
|
|
"epoch": 2.2877618522601986,
|
|
"grad_norm": 0.6984137195473014,
|
|
"learning_rate": 3.405634773647027e-05,
|
|
"loss": 0.1802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07871744781732559,
|
|
"step": 4150,
|
|
"valid_targets_mean": 1215.9,
|
|
"valid_targets_min": 724
|
|
},
|
|
{
|
|
"epoch": 2.290518191841235,
|
|
"grad_norm": 0.8030696066571337,
|
|
"learning_rate": 3.403677867378307e-05,
|
|
"loss": 0.1842,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07708248496055603,
|
|
"step": 4155,
|
|
"valid_targets_mean": 1257.6,
|
|
"valid_targets_min": 725
|
|
},
|
|
{
|
|
"epoch": 2.2932745314222713,
|
|
"grad_norm": 0.6872007464850188,
|
|
"learning_rate": 3.4017183091513884e-05,
|
|
"loss": 0.1896,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10817970335483551,
|
|
"step": 4160,
|
|
"valid_targets_mean": 1608.9,
|
|
"valid_targets_min": 667
|
|
},
|
|
{
|
|
"epoch": 2.2960308710033077,
|
|
"grad_norm": 0.606338900596094,
|
|
"learning_rate": 3.399756102668448e-05,
|
|
"loss": 0.1837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10406054556369781,
|
|
"step": 4165,
|
|
"valid_targets_mean": 1799.0,
|
|
"valid_targets_min": 791
|
|
},
|
|
{
|
|
"epoch": 2.298787210584344,
|
|
"grad_norm": 0.7260325951087828,
|
|
"learning_rate": 3.3977912516366685e-05,
|
|
"loss": 0.1837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09613724052906036,
|
|
"step": 4170,
|
|
"valid_targets_mean": 1622.6,
|
|
"valid_targets_min": 735
|
|
},
|
|
{
|
|
"epoch": 2.3015435501653805,
|
|
"grad_norm": 0.7163994839309861,
|
|
"learning_rate": 3.395823759768229e-05,
|
|
"loss": 0.1869,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1285693347454071,
|
|
"step": 4175,
|
|
"valid_targets_mean": 1817.1,
|
|
"valid_targets_min": 691
|
|
},
|
|
{
|
|
"epoch": 2.304299889746417,
|
|
"grad_norm": 0.6827063007576076,
|
|
"learning_rate": 3.393853630780297e-05,
|
|
"loss": 0.1808,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11781921982765198,
|
|
"step": 4180,
|
|
"valid_targets_mean": 1841.2,
|
|
"valid_targets_min": 1432
|
|
},
|
|
{
|
|
"epoch": 2.3070562293274532,
|
|
"grad_norm": 0.6934625357337632,
|
|
"learning_rate": 3.391880868395022e-05,
|
|
"loss": 0.1817,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1135096549987793,
|
|
"step": 4185,
|
|
"valid_targets_mean": 1844.2,
|
|
"valid_targets_min": 808
|
|
},
|
|
{
|
|
"epoch": 2.3098125689084896,
|
|
"grad_norm": 0.7324848813538705,
|
|
"learning_rate": 3.38990547633953e-05,
|
|
"loss": 0.1807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08447502553462982,
|
|
"step": 4190,
|
|
"valid_targets_mean": 1372.2,
|
|
"valid_targets_min": 789
|
|
},
|
|
{
|
|
"epoch": 2.312568908489526,
|
|
"grad_norm": 0.8607971865269517,
|
|
"learning_rate": 3.3879274583459145e-05,
|
|
"loss": 0.1713,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08694937825202942,
|
|
"step": 4195,
|
|
"valid_targets_mean": 1332.4,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 2.3153252480705624,
|
|
"grad_norm": 0.7319388311634608,
|
|
"learning_rate": 3.3859468181512305e-05,
|
|
"loss": 0.1833,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11193642765283585,
|
|
"step": 4200,
|
|
"valid_targets_mean": 1773.8,
|
|
"valid_targets_min": 824
|
|
},
|
|
{
|
|
"epoch": 2.3180815876515988,
|
|
"grad_norm": 0.7052785157666032,
|
|
"learning_rate": 3.3839635594974864e-05,
|
|
"loss": 0.1816,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08402714133262634,
|
|
"step": 4205,
|
|
"valid_targets_mean": 1595.6,
|
|
"valid_targets_min": 905
|
|
},
|
|
{
|
|
"epoch": 2.320837927232635,
|
|
"grad_norm": 0.7053790218853652,
|
|
"learning_rate": 3.381977686131639e-05,
|
|
"loss": 0.1777,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09687889367341995,
|
|
"step": 4210,
|
|
"valid_targets_mean": 1469.9,
|
|
"valid_targets_min": 1088
|
|
},
|
|
{
|
|
"epoch": 2.3235942668136715,
|
|
"grad_norm": 0.7003132242594865,
|
|
"learning_rate": 3.379989201805582e-05,
|
|
"loss": 0.1834,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09392642974853516,
|
|
"step": 4215,
|
|
"valid_targets_mean": 1572.0,
|
|
"valid_targets_min": 1118
|
|
},
|
|
{
|
|
"epoch": 2.326350606394708,
|
|
"grad_norm": 0.6826956861925462,
|
|
"learning_rate": 3.377998110276147e-05,
|
|
"loss": 0.183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1040303111076355,
|
|
"step": 4220,
|
|
"valid_targets_mean": 1508.0,
|
|
"valid_targets_min": 699
|
|
},
|
|
{
|
|
"epoch": 2.3291069459757443,
|
|
"grad_norm": 0.6878542401166445,
|
|
"learning_rate": 3.376004415305086e-05,
|
|
"loss": 0.1819,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07384923100471497,
|
|
"step": 4225,
|
|
"valid_targets_mean": 1283.0,
|
|
"valid_targets_min": 809
|
|
},
|
|
{
|
|
"epoch": 2.3318632855567807,
|
|
"grad_norm": 0.7734502307180114,
|
|
"learning_rate": 3.374008120659073e-05,
|
|
"loss": 0.1852,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08726200461387634,
|
|
"step": 4230,
|
|
"valid_targets_mean": 1323.5,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 2.334619625137817,
|
|
"grad_norm": 0.7175375609782297,
|
|
"learning_rate": 3.372009230109694e-05,
|
|
"loss": 0.182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10982063412666321,
|
|
"step": 4235,
|
|
"valid_targets_mean": 1537.9,
|
|
"valid_targets_min": 672
|
|
},
|
|
{
|
|
"epoch": 2.3373759647188534,
|
|
"grad_norm": 0.7799144210508859,
|
|
"learning_rate": 3.3700077474334354e-05,
|
|
"loss": 0.1917,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10207492113113403,
|
|
"step": 4240,
|
|
"valid_targets_mean": 1212.2,
|
|
"valid_targets_min": 638
|
|
},
|
|
{
|
|
"epoch": 2.34013230429989,
|
|
"grad_norm": 0.7551319806908273,
|
|
"learning_rate": 3.3680036764116866e-05,
|
|
"loss": 0.1789,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08111283928155899,
|
|
"step": 4245,
|
|
"valid_targets_mean": 1219.4,
|
|
"valid_targets_min": 700
|
|
},
|
|
{
|
|
"epoch": 2.342888643880926,
|
|
"grad_norm": 0.7064546182215252,
|
|
"learning_rate": 3.365997020830722e-05,
|
|
"loss": 0.1972,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09956347942352295,
|
|
"step": 4250,
|
|
"valid_targets_mean": 1845.4,
|
|
"valid_targets_min": 1201
|
|
},
|
|
{
|
|
"epoch": 2.3456449834619626,
|
|
"grad_norm": 0.6893468029410061,
|
|
"learning_rate": 3.363987784481702e-05,
|
|
"loss": 0.1816,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07128924131393433,
|
|
"step": 4255,
|
|
"valid_targets_mean": 1260.8,
|
|
"valid_targets_min": 656
|
|
},
|
|
{
|
|
"epoch": 2.348401323042999,
|
|
"grad_norm": 0.704400221522243,
|
|
"learning_rate": 3.361975971160662e-05,
|
|
"loss": 0.1869,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08587305247783661,
|
|
"step": 4260,
|
|
"valid_targets_mean": 1383.4,
|
|
"valid_targets_min": 669
|
|
},
|
|
{
|
|
"epoch": 2.3511576626240354,
|
|
"grad_norm": 0.6935996163176144,
|
|
"learning_rate": 3.359961584668505e-05,
|
|
"loss": 0.1804,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1116408258676529,
|
|
"step": 4265,
|
|
"valid_targets_mean": 1807.0,
|
|
"valid_targets_min": 1013
|
|
},
|
|
{
|
|
"epoch": 2.3539140022050717,
|
|
"grad_norm": 0.7175996898335646,
|
|
"learning_rate": 3.357944628810998e-05,
|
|
"loss": 0.1785,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07271032780408859,
|
|
"step": 4270,
|
|
"valid_targets_mean": 1112.4,
|
|
"valid_targets_min": 731
|
|
},
|
|
{
|
|
"epoch": 2.356670341786108,
|
|
"grad_norm": 0.7202969647190771,
|
|
"learning_rate": 3.35592510739876e-05,
|
|
"loss": 0.1865,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06504243612289429,
|
|
"step": 4275,
|
|
"valid_targets_mean": 1191.0,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 2.3594266813671445,
|
|
"grad_norm": 0.5948229025057867,
|
|
"learning_rate": 3.3539030242472574e-05,
|
|
"loss": 0.1761,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0913083553314209,
|
|
"step": 4280,
|
|
"valid_targets_mean": 1751.4,
|
|
"valid_targets_min": 904
|
|
},
|
|
{
|
|
"epoch": 2.362183020948181,
|
|
"grad_norm": 0.716172402901676,
|
|
"learning_rate": 3.351878383176797e-05,
|
|
"loss": 0.1783,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10536319017410278,
|
|
"step": 4285,
|
|
"valid_targets_mean": 1550.0,
|
|
"valid_targets_min": 594
|
|
},
|
|
{
|
|
"epoch": 2.3649393605292173,
|
|
"grad_norm": 0.7070441338335257,
|
|
"learning_rate": 3.34985118801252e-05,
|
|
"loss": 0.174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07276949286460876,
|
|
"step": 4290,
|
|
"valid_targets_mean": 1211.1,
|
|
"valid_targets_min": 686
|
|
},
|
|
{
|
|
"epoch": 2.3676957001102537,
|
|
"grad_norm": 0.6628328974240931,
|
|
"learning_rate": 3.347821442584389e-05,
|
|
"loss": 0.1908,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10310603678226471,
|
|
"step": 4295,
|
|
"valid_targets_mean": 1945.8,
|
|
"valid_targets_min": 1197
|
|
},
|
|
{
|
|
"epoch": 2.37045203969129,
|
|
"grad_norm": 0.6516016697500501,
|
|
"learning_rate": 3.3457891507271876e-05,
|
|
"loss": 0.1788,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09196461737155914,
|
|
"step": 4300,
|
|
"valid_targets_mean": 1752.8,
|
|
"valid_targets_min": 1211
|
|
},
|
|
{
|
|
"epoch": 2.3732083792723264,
|
|
"grad_norm": 0.64520895757867,
|
|
"learning_rate": 3.34375431628051e-05,
|
|
"loss": 0.1776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09301188588142395,
|
|
"step": 4305,
|
|
"valid_targets_mean": 1704.5,
|
|
"valid_targets_min": 1121
|
|
},
|
|
{
|
|
"epoch": 2.375964718853363,
|
|
"grad_norm": 0.7142211125820697,
|
|
"learning_rate": 3.341716943088754e-05,
|
|
"loss": 0.173,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09237653017044067,
|
|
"step": 4310,
|
|
"valid_targets_mean": 1484.0,
|
|
"valid_targets_min": 849
|
|
},
|
|
{
|
|
"epoch": 2.378721058434399,
|
|
"grad_norm": 0.6428424278862529,
|
|
"learning_rate": 3.3396770350011134e-05,
|
|
"loss": 0.1648,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07822799682617188,
|
|
"step": 4315,
|
|
"valid_targets_mean": 1565.6,
|
|
"valid_targets_min": 1313
|
|
},
|
|
{
|
|
"epoch": 2.3814773980154356,
|
|
"grad_norm": 0.711034658605113,
|
|
"learning_rate": 3.3376345958715716e-05,
|
|
"loss": 0.1833,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10037054121494293,
|
|
"step": 4320,
|
|
"valid_targets_mean": 1494.0,
|
|
"valid_targets_min": 1039
|
|
},
|
|
{
|
|
"epoch": 2.384233737596472,
|
|
"grad_norm": 0.6626711784087037,
|
|
"learning_rate": 3.335589629558894e-05,
|
|
"loss": 0.1818,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1116810292005539,
|
|
"step": 4325,
|
|
"valid_targets_mean": 1769.0,
|
|
"valid_targets_min": 919
|
|
},
|
|
{
|
|
"epoch": 2.3869900771775083,
|
|
"grad_norm": 0.6311122512223118,
|
|
"learning_rate": 3.33354213992662e-05,
|
|
"loss": 0.1783,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09387040138244629,
|
|
"step": 4330,
|
|
"valid_targets_mean": 1541.0,
|
|
"valid_targets_min": 846
|
|
},
|
|
{
|
|
"epoch": 2.3897464167585447,
|
|
"grad_norm": 0.6888366594354899,
|
|
"learning_rate": 3.3314921308430554e-05,
|
|
"loss": 0.1799,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09652800858020782,
|
|
"step": 4335,
|
|
"valid_targets_mean": 1763.1,
|
|
"valid_targets_min": 736
|
|
},
|
|
{
|
|
"epoch": 2.392502756339581,
|
|
"grad_norm": 0.690823090987391,
|
|
"learning_rate": 3.32943960618127e-05,
|
|
"loss": 0.182,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09987470507621765,
|
|
"step": 4340,
|
|
"valid_targets_mean": 1535.2,
|
|
"valid_targets_min": 593
|
|
},
|
|
{
|
|
"epoch": 2.3952590959206175,
|
|
"grad_norm": 0.5945662502074668,
|
|
"learning_rate": 3.32738456981908e-05,
|
|
"loss": 0.1857,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09591243416070938,
|
|
"step": 4345,
|
|
"valid_targets_mean": 2219.4,
|
|
"valid_targets_min": 1389
|
|
},
|
|
{
|
|
"epoch": 2.398015435501654,
|
|
"grad_norm": 0.6060810282499857,
|
|
"learning_rate": 3.325327025639054e-05,
|
|
"loss": 0.2918,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1524161398410797,
|
|
"step": 4350,
|
|
"valid_targets_mean": 3689.0,
|
|
"valid_targets_min": 770
|
|
},
|
|
{
|
|
"epoch": 2.4007717750826902,
|
|
"grad_norm": 0.6082448187662206,
|
|
"learning_rate": 3.3232669775284926e-05,
|
|
"loss": 0.2815,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14881551265716553,
|
|
"step": 4355,
|
|
"valid_targets_mean": 5227.8,
|
|
"valid_targets_min": 1798
|
|
},
|
|
{
|
|
"epoch": 2.4035281146637266,
|
|
"grad_norm": 0.4675464280366617,
|
|
"learning_rate": 3.321204429379429e-05,
|
|
"loss": 0.2725,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14512908458709717,
|
|
"step": 4360,
|
|
"valid_targets_mean": 5876.1,
|
|
"valid_targets_min": 3005
|
|
},
|
|
{
|
|
"epoch": 2.406284454244763,
|
|
"grad_norm": 0.46185135244843234,
|
|
"learning_rate": 3.3191393850886225e-05,
|
|
"loss": 0.2696,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13734814524650574,
|
|
"step": 4365,
|
|
"valid_targets_mean": 6019.1,
|
|
"valid_targets_min": 1679
|
|
},
|
|
{
|
|
"epoch": 2.4090407938257994,
|
|
"grad_norm": 0.5348588994531697,
|
|
"learning_rate": 3.317071848557544e-05,
|
|
"loss": 0.278,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11896532773971558,
|
|
"step": 4370,
|
|
"valid_targets_mean": 3295.9,
|
|
"valid_targets_min": 1373
|
|
},
|
|
{
|
|
"epoch": 2.4117971334068358,
|
|
"grad_norm": 0.5905424479990715,
|
|
"learning_rate": 3.315001823692376e-05,
|
|
"loss": 0.2803,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1370130479335785,
|
|
"step": 4375,
|
|
"valid_targets_mean": 3851.8,
|
|
"valid_targets_min": 1950
|
|
},
|
|
{
|
|
"epoch": 2.414553472987872,
|
|
"grad_norm": 0.4762539178113444,
|
|
"learning_rate": 3.3129293144039995e-05,
|
|
"loss": 0.2662,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13153642416000366,
|
|
"step": 4380,
|
|
"valid_targets_mean": 5243.2,
|
|
"valid_targets_min": 2040
|
|
},
|
|
{
|
|
"epoch": 2.4173098125689085,
|
|
"grad_norm": 0.47043084867500007,
|
|
"learning_rate": 3.310854324607993e-05,
|
|
"loss": 0.2711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11843539774417877,
|
|
"step": 4385,
|
|
"valid_targets_mean": 4928.8,
|
|
"valid_targets_min": 1830
|
|
},
|
|
{
|
|
"epoch": 2.420066152149945,
|
|
"grad_norm": 0.500048439405165,
|
|
"learning_rate": 3.308776858224619e-05,
|
|
"loss": 0.2827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1450158655643463,
|
|
"step": 4390,
|
|
"valid_targets_mean": 4039.6,
|
|
"valid_targets_min": 2071
|
|
},
|
|
{
|
|
"epoch": 2.4228224917309813,
|
|
"grad_norm": 0.48087801008610814,
|
|
"learning_rate": 3.3066969191788184e-05,
|
|
"loss": 0.2769,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12331129610538483,
|
|
"step": 4395,
|
|
"valid_targets_mean": 4077.1,
|
|
"valid_targets_min": 2021
|
|
},
|
|
{
|
|
"epoch": 2.4255788313120177,
|
|
"grad_norm": 0.49054437469965273,
|
|
"learning_rate": 3.304614511400205e-05,
|
|
"loss": 0.2806,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1505330204963684,
|
|
"step": 4400,
|
|
"valid_targets_mean": 4889.5,
|
|
"valid_targets_min": 1498
|
|
},
|
|
{
|
|
"epoch": 2.428335170893054,
|
|
"grad_norm": 0.4495668390980617,
|
|
"learning_rate": 3.3025296388230564e-05,
|
|
"loss": 0.2684,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1413159966468811,
|
|
"step": 4405,
|
|
"valid_targets_mean": 4812.9,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 2.4310915104740904,
|
|
"grad_norm": 0.505754152621846,
|
|
"learning_rate": 3.300442305386308e-05,
|
|
"loss": 0.2752,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13405068218708038,
|
|
"step": 4410,
|
|
"valid_targets_mean": 4664.6,
|
|
"valid_targets_min": 1934
|
|
},
|
|
{
|
|
"epoch": 2.433847850055127,
|
|
"grad_norm": 0.5534215885194713,
|
|
"learning_rate": 3.2983525150335414e-05,
|
|
"loss": 0.2726,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16056929528713226,
|
|
"step": 4415,
|
|
"valid_targets_mean": 4002.8,
|
|
"valid_targets_min": 1813
|
|
},
|
|
{
|
|
"epoch": 2.436604189636163,
|
|
"grad_norm": 0.584886934625671,
|
|
"learning_rate": 3.2962602717129834e-05,
|
|
"loss": 0.2827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16010001301765442,
|
|
"step": 4420,
|
|
"valid_targets_mean": 3818.5,
|
|
"valid_targets_min": 1587
|
|
},
|
|
{
|
|
"epoch": 2.4393605292171996,
|
|
"grad_norm": 0.5106226436701461,
|
|
"learning_rate": 3.294165579377493e-05,
|
|
"loss": 0.2722,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12410934269428253,
|
|
"step": 4425,
|
|
"valid_targets_mean": 4037.5,
|
|
"valid_targets_min": 2658
|
|
},
|
|
{
|
|
"epoch": 2.442116868798236,
|
|
"grad_norm": 0.5506137872583496,
|
|
"learning_rate": 3.292068441984558e-05,
|
|
"loss": 0.2756,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13618406653404236,
|
|
"step": 4430,
|
|
"valid_targets_mean": 3588.2,
|
|
"valid_targets_min": 1267
|
|
},
|
|
{
|
|
"epoch": 2.4448732083792724,
|
|
"grad_norm": 0.5953928316190624,
|
|
"learning_rate": 3.2899688634962846e-05,
|
|
"loss": 0.2871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13676932454109192,
|
|
"step": 4435,
|
|
"valid_targets_mean": 3418.8,
|
|
"valid_targets_min": 1692
|
|
},
|
|
{
|
|
"epoch": 2.4476295479603087,
|
|
"grad_norm": 0.5687353164076167,
|
|
"learning_rate": 3.287866847879389e-05,
|
|
"loss": 0.2832,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1336851269006729,
|
|
"step": 4440,
|
|
"valid_targets_mean": 3515.8,
|
|
"valid_targets_min": 1125
|
|
},
|
|
{
|
|
"epoch": 2.450385887541345,
|
|
"grad_norm": 0.5721182512728366,
|
|
"learning_rate": 3.285762399105195e-05,
|
|
"loss": 0.2707,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13906799256801605,
|
|
"step": 4445,
|
|
"valid_targets_mean": 3526.9,
|
|
"valid_targets_min": 1437
|
|
},
|
|
{
|
|
"epoch": 2.4531422271223815,
|
|
"grad_norm": 0.5480685644026277,
|
|
"learning_rate": 3.283655521149623e-05,
|
|
"loss": 0.2805,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13203464448451996,
|
|
"step": 4450,
|
|
"valid_targets_mean": 3837.6,
|
|
"valid_targets_min": 1266
|
|
},
|
|
{
|
|
"epoch": 2.455898566703418,
|
|
"grad_norm": 0.519538844041324,
|
|
"learning_rate": 3.281546217993181e-05,
|
|
"loss": 0.2668,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11915551126003265,
|
|
"step": 4455,
|
|
"valid_targets_mean": 3666.1,
|
|
"valid_targets_min": 1025
|
|
},
|
|
{
|
|
"epoch": 2.4586549062844543,
|
|
"grad_norm": 0.6190553757852973,
|
|
"learning_rate": 3.27943449362096e-05,
|
|
"loss": 0.2554,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12359665334224701,
|
|
"step": 4460,
|
|
"valid_targets_mean": 2861.6,
|
|
"valid_targets_min": 1403
|
|
},
|
|
{
|
|
"epoch": 2.4614112458654906,
|
|
"grad_norm": 0.589075486518532,
|
|
"learning_rate": 3.277320352022624e-05,
|
|
"loss": 0.2785,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15374290943145752,
|
|
"step": 4465,
|
|
"valid_targets_mean": 3876.2,
|
|
"valid_targets_min": 1586
|
|
},
|
|
{
|
|
"epoch": 2.464167585446527,
|
|
"grad_norm": 0.5952165574463202,
|
|
"learning_rate": 3.275203797192406e-05,
|
|
"loss": 0.2786,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1218034029006958,
|
|
"step": 4470,
|
|
"valid_targets_mean": 2824.6,
|
|
"valid_targets_min": 1274
|
|
},
|
|
{
|
|
"epoch": 2.4669239250275634,
|
|
"grad_norm": 0.457084394026027,
|
|
"learning_rate": 3.2730848331290984e-05,
|
|
"loss": 0.2671,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11772876977920532,
|
|
"step": 4475,
|
|
"valid_targets_mean": 4224.9,
|
|
"valid_targets_min": 1631
|
|
},
|
|
{
|
|
"epoch": 2.4696802646086,
|
|
"grad_norm": 0.5106255372634201,
|
|
"learning_rate": 3.270963463836043e-05,
|
|
"loss": 0.2693,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16350074112415314,
|
|
"step": 4480,
|
|
"valid_targets_mean": 5064.2,
|
|
"valid_targets_min": 3082
|
|
},
|
|
{
|
|
"epoch": 2.472436604189636,
|
|
"grad_norm": 0.5592886672495634,
|
|
"learning_rate": 3.268839693321128e-05,
|
|
"loss": 0.2598,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16275565326213837,
|
|
"step": 4485,
|
|
"valid_targets_mean": 3725.6,
|
|
"valid_targets_min": 1801
|
|
},
|
|
{
|
|
"epoch": 2.4751929437706726,
|
|
"grad_norm": 0.603126267997095,
|
|
"learning_rate": 3.266713525596776e-05,
|
|
"loss": 0.2744,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15631866455078125,
|
|
"step": 4490,
|
|
"valid_targets_mean": 3485.0,
|
|
"valid_targets_min": 1680
|
|
},
|
|
{
|
|
"epoch": 2.477949283351709,
|
|
"grad_norm": 0.5272710390828589,
|
|
"learning_rate": 3.2645849646799406e-05,
|
|
"loss": 0.261,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12210951745510101,
|
|
"step": 4495,
|
|
"valid_targets_mean": 3653.2,
|
|
"valid_targets_min": 1590
|
|
},
|
|
{
|
|
"epoch": 2.4807056229327453,
|
|
"grad_norm": 0.526678863464346,
|
|
"learning_rate": 3.262454014592097e-05,
|
|
"loss": 0.2666,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14927881956100464,
|
|
"step": 4500,
|
|
"valid_targets_mean": 4299.8,
|
|
"valid_targets_min": 2251
|
|
},
|
|
{
|
|
"epoch": 2.4834619625137817,
|
|
"grad_norm": 0.6253533543719939,
|
|
"learning_rate": 3.260320679359232e-05,
|
|
"loss": 0.2634,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1374313235282898,
|
|
"step": 4505,
|
|
"valid_targets_mean": 3100.4,
|
|
"valid_targets_min": 1456
|
|
},
|
|
{
|
|
"epoch": 2.486218302094818,
|
|
"grad_norm": 0.5849042479272506,
|
|
"learning_rate": 3.258184963011842e-05,
|
|
"loss": 0.2692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1283252239227295,
|
|
"step": 4510,
|
|
"valid_targets_mean": 3659.6,
|
|
"valid_targets_min": 1160
|
|
},
|
|
{
|
|
"epoch": 2.4889746416758545,
|
|
"grad_norm": 0.6027198319742977,
|
|
"learning_rate": 3.2560468695849174e-05,
|
|
"loss": 0.2647,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1237245425581932,
|
|
"step": 4515,
|
|
"valid_targets_mean": 2829.0,
|
|
"valid_targets_min": 1479
|
|
},
|
|
{
|
|
"epoch": 2.491730981256891,
|
|
"grad_norm": 0.5764632429524174,
|
|
"learning_rate": 3.253906403117945e-05,
|
|
"loss": 0.2752,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1405836045742035,
|
|
"step": 4520,
|
|
"valid_targets_mean": 3789.6,
|
|
"valid_targets_min": 1228
|
|
},
|
|
{
|
|
"epoch": 2.4944873208379272,
|
|
"grad_norm": 0.5563984891705012,
|
|
"learning_rate": 3.25176356765489e-05,
|
|
"loss": 0.2717,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12148723006248474,
|
|
"step": 4525,
|
|
"valid_targets_mean": 3551.9,
|
|
"valid_targets_min": 1268
|
|
},
|
|
{
|
|
"epoch": 2.4972436604189636,
|
|
"grad_norm": 0.549562653312022,
|
|
"learning_rate": 3.249618367244197e-05,
|
|
"loss": 0.2642,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1202792227268219,
|
|
"step": 4530,
|
|
"valid_targets_mean": 3473.5,
|
|
"valid_targets_min": 1355
|
|
},
|
|
{
|
|
"epoch": 2.5,
|
|
"grad_norm": 0.5975071320690827,
|
|
"learning_rate": 3.2474708059387775e-05,
|
|
"loss": 0.2648,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13821253180503845,
|
|
"step": 4535,
|
|
"valid_targets_mean": 2919.6,
|
|
"valid_targets_min": 922
|
|
},
|
|
{
|
|
"epoch": 2.5027563395810364,
|
|
"grad_norm": 0.5632577184943831,
|
|
"learning_rate": 3.245320887796001e-05,
|
|
"loss": 0.2707,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13934670388698578,
|
|
"step": 4540,
|
|
"valid_targets_mean": 4023.8,
|
|
"valid_targets_min": 1671
|
|
},
|
|
{
|
|
"epoch": 2.5055126791620728,
|
|
"grad_norm": 0.5789046905053755,
|
|
"learning_rate": 3.2431686168776944e-05,
|
|
"loss": 0.2665,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13213962316513062,
|
|
"step": 4545,
|
|
"valid_targets_mean": 3168.5,
|
|
"valid_targets_min": 1639
|
|
},
|
|
{
|
|
"epoch": 2.508269018743109,
|
|
"grad_norm": 0.5099347302872038,
|
|
"learning_rate": 3.241013997250126e-05,
|
|
"loss": 0.259,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1415427178144455,
|
|
"step": 4550,
|
|
"valid_targets_mean": 4252.4,
|
|
"valid_targets_min": 1242
|
|
},
|
|
{
|
|
"epoch": 2.5110253583241455,
|
|
"grad_norm": 0.5522632815057266,
|
|
"learning_rate": 3.2388570329840025e-05,
|
|
"loss": 0.2741,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1261557787656784,
|
|
"step": 4555,
|
|
"valid_targets_mean": 4060.6,
|
|
"valid_targets_min": 729
|
|
},
|
|
{
|
|
"epoch": 2.513781697905182,
|
|
"grad_norm": 0.611673062928991,
|
|
"learning_rate": 3.236697728154461e-05,
|
|
"loss": 0.2631,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11874519288539886,
|
|
"step": 4560,
|
|
"valid_targets_mean": 2941.5,
|
|
"valid_targets_min": 975
|
|
},
|
|
{
|
|
"epoch": 2.5165380374862183,
|
|
"grad_norm": 0.5838833192142227,
|
|
"learning_rate": 3.234536086841061e-05,
|
|
"loss": 0.2635,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14937260746955872,
|
|
"step": 4565,
|
|
"valid_targets_mean": 5109.2,
|
|
"valid_targets_min": 2085
|
|
},
|
|
{
|
|
"epoch": 2.5192943770672547,
|
|
"grad_norm": 0.5512299095915336,
|
|
"learning_rate": 3.232372113127774e-05,
|
|
"loss": 0.2517,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11611886322498322,
|
|
"step": 4570,
|
|
"valid_targets_mean": 3116.8,
|
|
"valid_targets_min": 1649
|
|
},
|
|
{
|
|
"epoch": 2.522050716648291,
|
|
"grad_norm": 0.5448186543026713,
|
|
"learning_rate": 3.230205811102979e-05,
|
|
"loss": 0.2735,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13560880720615387,
|
|
"step": 4575,
|
|
"valid_targets_mean": 4074.8,
|
|
"valid_targets_min": 1427
|
|
},
|
|
{
|
|
"epoch": 2.5248070562293274,
|
|
"grad_norm": 0.534437224698198,
|
|
"learning_rate": 3.228037184859456e-05,
|
|
"loss": 0.2559,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12374728918075562,
|
|
"step": 4580,
|
|
"valid_targets_mean": 3861.9,
|
|
"valid_targets_min": 1261
|
|
},
|
|
{
|
|
"epoch": 2.527563395810364,
|
|
"grad_norm": 0.5767029549110786,
|
|
"learning_rate": 3.2258662384943745e-05,
|
|
"loss": 0.2756,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1584603488445282,
|
|
"step": 4585,
|
|
"valid_targets_mean": 4072.1,
|
|
"valid_targets_min": 2725
|
|
},
|
|
{
|
|
"epoch": 2.5303197353914,
|
|
"grad_norm": 0.5434878768971206,
|
|
"learning_rate": 3.2236929761092865e-05,
|
|
"loss": 0.2642,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1119743138551712,
|
|
"step": 4590,
|
|
"valid_targets_mean": 3464.9,
|
|
"valid_targets_min": 1217
|
|
},
|
|
{
|
|
"epoch": 2.5330760749724366,
|
|
"grad_norm": 0.5970326998402175,
|
|
"learning_rate": 3.221517401810121e-05,
|
|
"loss": 0.2756,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1449601948261261,
|
|
"step": 4595,
|
|
"valid_targets_mean": 4082.0,
|
|
"valid_targets_min": 2057
|
|
},
|
|
{
|
|
"epoch": 2.535832414553473,
|
|
"grad_norm": 0.5878789554245125,
|
|
"learning_rate": 3.219339519707173e-05,
|
|
"loss": 0.2554,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12189297378063202,
|
|
"step": 4600,
|
|
"valid_targets_mean": 2729.9,
|
|
"valid_targets_min": 915
|
|
},
|
|
{
|
|
"epoch": 2.5385887541345094,
|
|
"grad_norm": 0.5465305540384525,
|
|
"learning_rate": 3.217159333915101e-05,
|
|
"loss": 0.2771,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11429809033870697,
|
|
"step": 4605,
|
|
"valid_targets_mean": 3687.0,
|
|
"valid_targets_min": 1296
|
|
},
|
|
{
|
|
"epoch": 2.5413450937155457,
|
|
"grad_norm": 0.5774697918243201,
|
|
"learning_rate": 3.214976848552913e-05,
|
|
"loss": 0.2564,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16515560448169708,
|
|
"step": 4610,
|
|
"valid_targets_mean": 4240.2,
|
|
"valid_targets_min": 1499
|
|
},
|
|
{
|
|
"epoch": 2.544101433296582,
|
|
"grad_norm": 0.5560713624594463,
|
|
"learning_rate": 3.212792067743963e-05,
|
|
"loss": 0.2505,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1467883437871933,
|
|
"step": 4615,
|
|
"valid_targets_mean": 4828.5,
|
|
"valid_targets_min": 2791
|
|
},
|
|
{
|
|
"epoch": 2.5468577728776185,
|
|
"grad_norm": 0.5885622736828782,
|
|
"learning_rate": 3.21060499561594e-05,
|
|
"loss": 0.2521,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1357017457485199,
|
|
"step": 4620,
|
|
"valid_targets_mean": 3344.4,
|
|
"valid_targets_min": 877
|
|
},
|
|
{
|
|
"epoch": 2.549614112458655,
|
|
"grad_norm": 0.5182897605745844,
|
|
"learning_rate": 3.208415636300864e-05,
|
|
"loss": 0.2717,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1192808523774147,
|
|
"step": 4625,
|
|
"valid_targets_mean": 3746.5,
|
|
"valid_targets_min": 2654
|
|
},
|
|
{
|
|
"epoch": 2.5523704520396913,
|
|
"grad_norm": 0.5276749356369358,
|
|
"learning_rate": 3.2062239939350756e-05,
|
|
"loss": 0.2622,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1242235004901886,
|
|
"step": 4630,
|
|
"valid_targets_mean": 3960.0,
|
|
"valid_targets_min": 1084
|
|
},
|
|
{
|
|
"epoch": 2.5551267916207276,
|
|
"grad_norm": 0.5418517526294038,
|
|
"learning_rate": 3.204030072659229e-05,
|
|
"loss": 0.2657,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14493471384048462,
|
|
"step": 4635,
|
|
"valid_targets_mean": 5284.0,
|
|
"valid_targets_min": 1565
|
|
},
|
|
{
|
|
"epoch": 2.557883131201764,
|
|
"grad_norm": 0.5868346221194314,
|
|
"learning_rate": 3.201833876618281e-05,
|
|
"loss": 0.2654,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15977177023887634,
|
|
"step": 4640,
|
|
"valid_targets_mean": 4159.5,
|
|
"valid_targets_min": 2492
|
|
},
|
|
{
|
|
"epoch": 2.5606394707828004,
|
|
"grad_norm": 0.5529897695229956,
|
|
"learning_rate": 3.199635409961491e-05,
|
|
"loss": 0.2716,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16260987520217896,
|
|
"step": 4645,
|
|
"valid_targets_mean": 4593.1,
|
|
"valid_targets_min": 1546
|
|
},
|
|
{
|
|
"epoch": 2.563395810363837,
|
|
"grad_norm": 0.5681004659729655,
|
|
"learning_rate": 3.197434676842404e-05,
|
|
"loss": 0.2658,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12388027459383011,
|
|
"step": 4650,
|
|
"valid_targets_mean": 3282.5,
|
|
"valid_targets_min": 1350
|
|
},
|
|
{
|
|
"epoch": 2.566152149944873,
|
|
"grad_norm": 0.49436033993167217,
|
|
"learning_rate": 3.19523168141885e-05,
|
|
"loss": 0.2576,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1363145112991333,
|
|
"step": 4655,
|
|
"valid_targets_mean": 5273.9,
|
|
"valid_targets_min": 2339
|
|
},
|
|
{
|
|
"epoch": 2.5689084895259096,
|
|
"grad_norm": 0.5473249402573735,
|
|
"learning_rate": 3.193026427852932e-05,
|
|
"loss": 0.2544,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12649378180503845,
|
|
"step": 4660,
|
|
"valid_targets_mean": 4047.0,
|
|
"valid_targets_min": 2008
|
|
},
|
|
{
|
|
"epoch": 2.571664829106946,
|
|
"grad_norm": 0.5619898696826447,
|
|
"learning_rate": 3.190818920311018e-05,
|
|
"loss": 0.2528,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15290437638759613,
|
|
"step": 4665,
|
|
"valid_targets_mean": 4831.5,
|
|
"valid_targets_min": 1728
|
|
},
|
|
{
|
|
"epoch": 2.5744211686879823,
|
|
"grad_norm": 0.6122682506629524,
|
|
"learning_rate": 3.1886091629637376e-05,
|
|
"loss": 0.2553,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1224728524684906,
|
|
"step": 4670,
|
|
"valid_targets_mean": 3047.1,
|
|
"valid_targets_min": 1457
|
|
},
|
|
{
|
|
"epoch": 2.5771775082690187,
|
|
"grad_norm": 0.5884348410151253,
|
|
"learning_rate": 3.186397159985967e-05,
|
|
"loss": 0.2456,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.138324573636055,
|
|
"step": 4675,
|
|
"valid_targets_mean": 4615.9,
|
|
"valid_targets_min": 1280
|
|
},
|
|
{
|
|
"epoch": 2.579933847850055,
|
|
"grad_norm": 0.5494103454446826,
|
|
"learning_rate": 3.1841829155568284e-05,
|
|
"loss": 0.2452,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10916861891746521,
|
|
"step": 4680,
|
|
"valid_targets_mean": 3367.5,
|
|
"valid_targets_min": 1420
|
|
},
|
|
{
|
|
"epoch": 2.5826901874310915,
|
|
"grad_norm": 0.5748191435392036,
|
|
"learning_rate": 3.181966433859676e-05,
|
|
"loss": 0.2538,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15969008207321167,
|
|
"step": 4685,
|
|
"valid_targets_mean": 4477.8,
|
|
"valid_targets_min": 1763
|
|
},
|
|
{
|
|
"epoch": 2.585446527012128,
|
|
"grad_norm": 0.6624266718173617,
|
|
"learning_rate": 3.179747719082094e-05,
|
|
"loss": 0.2706,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1552642285823822,
|
|
"step": 4690,
|
|
"valid_targets_mean": 3284.8,
|
|
"valid_targets_min": 1175
|
|
},
|
|
{
|
|
"epoch": 2.5882028665931642,
|
|
"grad_norm": 0.48852172194169297,
|
|
"learning_rate": 3.1775267754158824e-05,
|
|
"loss": 0.2627,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1174711361527443,
|
|
"step": 4695,
|
|
"valid_targets_mean": 4651.1,
|
|
"valid_targets_min": 2544
|
|
},
|
|
{
|
|
"epoch": 2.5909592061742006,
|
|
"grad_norm": 0.6805815220755985,
|
|
"learning_rate": 3.1753036070570554e-05,
|
|
"loss": 0.2626,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16039074957370758,
|
|
"step": 4700,
|
|
"valid_targets_mean": 3827.2,
|
|
"valid_targets_min": 1242
|
|
},
|
|
{
|
|
"epoch": 2.593715545755237,
|
|
"grad_norm": 0.5281262423044831,
|
|
"learning_rate": 3.1730782182058255e-05,
|
|
"loss": 0.2558,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.129986971616745,
|
|
"step": 4705,
|
|
"valid_targets_mean": 4398.1,
|
|
"valid_targets_min": 1851
|
|
},
|
|
{
|
|
"epoch": 2.5964718853362734,
|
|
"grad_norm": 0.5467471602528055,
|
|
"learning_rate": 3.170850613066607e-05,
|
|
"loss": 0.255,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.143656387925148,
|
|
"step": 4710,
|
|
"valid_targets_mean": 3830.4,
|
|
"valid_targets_min": 1949
|
|
},
|
|
{
|
|
"epoch": 2.5992282249173098,
|
|
"grad_norm": 0.5023088114488351,
|
|
"learning_rate": 3.168620795847994e-05,
|
|
"loss": 0.2481,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1077934131026268,
|
|
"step": 4715,
|
|
"valid_targets_mean": 3330.5,
|
|
"valid_targets_min": 1247
|
|
},
|
|
{
|
|
"epoch": 2.601984564498346,
|
|
"grad_norm": 0.48528689588553625,
|
|
"learning_rate": 3.1663887707627666e-05,
|
|
"loss": 0.2548,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11805665493011475,
|
|
"step": 4720,
|
|
"valid_targets_mean": 4298.8,
|
|
"valid_targets_min": 1143
|
|
},
|
|
{
|
|
"epoch": 2.6047409040793825,
|
|
"grad_norm": 0.555162458448305,
|
|
"learning_rate": 3.164154542027872e-05,
|
|
"loss": 0.2552,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12123659253120422,
|
|
"step": 4725,
|
|
"valid_targets_mean": 3224.0,
|
|
"valid_targets_min": 1424
|
|
},
|
|
{
|
|
"epoch": 2.607497243660419,
|
|
"grad_norm": 0.547601941814992,
|
|
"learning_rate": 3.161918113864423e-05,
|
|
"loss": 0.2551,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14641651511192322,
|
|
"step": 4730,
|
|
"valid_targets_mean": 4420.4,
|
|
"valid_targets_min": 1706
|
|
},
|
|
{
|
|
"epoch": 2.6102535832414553,
|
|
"grad_norm": 0.6758742005846232,
|
|
"learning_rate": 3.159679490497686e-05,
|
|
"loss": 0.2509,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08945570886135101,
|
|
"step": 4735,
|
|
"valid_targets_mean": 2927.4,
|
|
"valid_targets_min": 1299
|
|
},
|
|
{
|
|
"epoch": 2.6130099228224917,
|
|
"grad_norm": 0.5928386902365377,
|
|
"learning_rate": 3.157438676157075e-05,
|
|
"loss": 0.2729,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11042401194572449,
|
|
"step": 4740,
|
|
"valid_targets_mean": 3006.4,
|
|
"valid_targets_min": 1961
|
|
},
|
|
{
|
|
"epoch": 2.615766262403528,
|
|
"grad_norm": 0.536198783776861,
|
|
"learning_rate": 3.1551956750761444e-05,
|
|
"loss": 0.2643,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15939512848854065,
|
|
"step": 4745,
|
|
"valid_targets_mean": 4881.2,
|
|
"valid_targets_min": 1053
|
|
},
|
|
{
|
|
"epoch": 2.6185226019845644,
|
|
"grad_norm": 0.569498442551076,
|
|
"learning_rate": 3.15295049149258e-05,
|
|
"loss": 0.2656,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14446908235549927,
|
|
"step": 4750,
|
|
"valid_targets_mean": 3521.1,
|
|
"valid_targets_min": 1609
|
|
},
|
|
{
|
|
"epoch": 2.621278941565601,
|
|
"grad_norm": 0.549336909715908,
|
|
"learning_rate": 3.15070312964819e-05,
|
|
"loss": 0.2719,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1096280962228775,
|
|
"step": 4755,
|
|
"valid_targets_mean": 3525.4,
|
|
"valid_targets_min": 720
|
|
},
|
|
{
|
|
"epoch": 2.624035281146637,
|
|
"grad_norm": 0.5478506496631415,
|
|
"learning_rate": 3.148453593788899e-05,
|
|
"loss": 0.2594,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13057929277420044,
|
|
"step": 4760,
|
|
"valid_targets_mean": 3893.9,
|
|
"valid_targets_min": 1623
|
|
},
|
|
{
|
|
"epoch": 2.6267916207276736,
|
|
"grad_norm": 0.5339890629644338,
|
|
"learning_rate": 3.146201888164738e-05,
|
|
"loss": 0.2692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1291407346725464,
|
|
"step": 4765,
|
|
"valid_targets_mean": 4168.8,
|
|
"valid_targets_min": 1380
|
|
},
|
|
{
|
|
"epoch": 2.62954796030871,
|
|
"grad_norm": 0.6168444703763231,
|
|
"learning_rate": 3.143948017029838e-05,
|
|
"loss": 0.2729,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1310538351535797,
|
|
"step": 4770,
|
|
"valid_targets_mean": 3230.8,
|
|
"valid_targets_min": 995
|
|
},
|
|
{
|
|
"epoch": 2.6323042998897463,
|
|
"grad_norm": 0.5386223611000888,
|
|
"learning_rate": 3.141691984642421e-05,
|
|
"loss": 0.2575,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09724453091621399,
|
|
"step": 4775,
|
|
"valid_targets_mean": 2951.9,
|
|
"valid_targets_min": 1186
|
|
},
|
|
{
|
|
"epoch": 2.6350606394707827,
|
|
"grad_norm": 0.5557605337049826,
|
|
"learning_rate": 3.139433795264791e-05,
|
|
"loss": 0.2622,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15454351902008057,
|
|
"step": 4780,
|
|
"valid_targets_mean": 3578.2,
|
|
"valid_targets_min": 1194
|
|
},
|
|
{
|
|
"epoch": 2.637816979051819,
|
|
"grad_norm": 0.6153083085196538,
|
|
"learning_rate": 3.137173453163332e-05,
|
|
"loss": 0.2635,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12959834933280945,
|
|
"step": 4785,
|
|
"valid_targets_mean": 2691.1,
|
|
"valid_targets_min": 884
|
|
},
|
|
{
|
|
"epoch": 2.6405733186328555,
|
|
"grad_norm": 0.5662174640494001,
|
|
"learning_rate": 3.134910962608487e-05,
|
|
"loss": 0.2656,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1241687536239624,
|
|
"step": 4790,
|
|
"valid_targets_mean": 3355.0,
|
|
"valid_targets_min": 813
|
|
},
|
|
{
|
|
"epoch": 2.643329658213892,
|
|
"grad_norm": 0.5607899763778619,
|
|
"learning_rate": 3.132646327874766e-05,
|
|
"loss": 0.264,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11414383351802826,
|
|
"step": 4795,
|
|
"valid_targets_mean": 3605.6,
|
|
"valid_targets_min": 2165
|
|
},
|
|
{
|
|
"epoch": 2.6460859977949283,
|
|
"grad_norm": 0.48190331769308214,
|
|
"learning_rate": 3.130379553240726e-05,
|
|
"loss": 0.2442,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14459747076034546,
|
|
"step": 4800,
|
|
"valid_targets_mean": 5464.9,
|
|
"valid_targets_min": 3020
|
|
},
|
|
{
|
|
"epoch": 2.6488423373759646,
|
|
"grad_norm": 0.6468866298642583,
|
|
"learning_rate": 3.1281106429889655e-05,
|
|
"loss": 0.2579,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12449154257774353,
|
|
"step": 4805,
|
|
"valid_targets_mean": 3293.8,
|
|
"valid_targets_min": 1434
|
|
},
|
|
{
|
|
"epoch": 2.651598676957001,
|
|
"grad_norm": 0.6094053154279868,
|
|
"learning_rate": 3.125839601406123e-05,
|
|
"loss": 0.2492,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12536941468715668,
|
|
"step": 4810,
|
|
"valid_targets_mean": 3257.8,
|
|
"valid_targets_min": 1071
|
|
},
|
|
{
|
|
"epoch": 2.6543550165380374,
|
|
"grad_norm": 0.5712572263970925,
|
|
"learning_rate": 3.123566432782858e-05,
|
|
"loss": 0.259,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10148803889751434,
|
|
"step": 4815,
|
|
"valid_targets_mean": 3140.2,
|
|
"valid_targets_min": 978
|
|
},
|
|
{
|
|
"epoch": 2.657111356119074,
|
|
"grad_norm": 0.604952751529664,
|
|
"learning_rate": 3.121291141413853e-05,
|
|
"loss": 0.2556,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1431146264076233,
|
|
"step": 4820,
|
|
"valid_targets_mean": 4568.9,
|
|
"valid_targets_min": 2363
|
|
},
|
|
{
|
|
"epoch": 2.65986769570011,
|
|
"grad_norm": 0.5377599516503931,
|
|
"learning_rate": 3.1190137315977976e-05,
|
|
"loss": 0.2649,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13327714800834656,
|
|
"step": 4825,
|
|
"valid_targets_mean": 4992.5,
|
|
"valid_targets_min": 3135
|
|
},
|
|
{
|
|
"epoch": 2.6626240352811466,
|
|
"grad_norm": 0.6119262254182417,
|
|
"learning_rate": 3.116734207637386e-05,
|
|
"loss": 0.2521,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15674933791160583,
|
|
"step": 4830,
|
|
"valid_targets_mean": 4028.0,
|
|
"valid_targets_min": 2136
|
|
},
|
|
{
|
|
"epoch": 2.665380374862183,
|
|
"grad_norm": 0.5611207556589975,
|
|
"learning_rate": 3.114452573839306e-05,
|
|
"loss": 0.256,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1284465491771698,
|
|
"step": 4835,
|
|
"valid_targets_mean": 3603.0,
|
|
"valid_targets_min": 2009
|
|
},
|
|
{
|
|
"epoch": 2.6681367144432193,
|
|
"grad_norm": 0.6171337393789802,
|
|
"learning_rate": 3.112168834514231e-05,
|
|
"loss": 0.258,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14494982361793518,
|
|
"step": 4840,
|
|
"valid_targets_mean": 3787.1,
|
|
"valid_targets_min": 1082
|
|
},
|
|
{
|
|
"epoch": 2.6708930540242557,
|
|
"grad_norm": 0.5468485046969078,
|
|
"learning_rate": 3.1098829939768134e-05,
|
|
"loss": 0.248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12874338030815125,
|
|
"step": 4845,
|
|
"valid_targets_mean": 3678.5,
|
|
"valid_targets_min": 1614
|
|
},
|
|
{
|
|
"epoch": 2.673649393605292,
|
|
"grad_norm": 0.5768706031407824,
|
|
"learning_rate": 3.107595056545675e-05,
|
|
"loss": 0.2415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1188037246465683,
|
|
"step": 4850,
|
|
"valid_targets_mean": 3374.5,
|
|
"valid_targets_min": 1190
|
|
},
|
|
{
|
|
"epoch": 2.6764057331863285,
|
|
"grad_norm": 0.5192595082000558,
|
|
"learning_rate": 3.1053050265433975e-05,
|
|
"loss": 0.2489,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13302139937877655,
|
|
"step": 4855,
|
|
"valid_targets_mean": 4725.9,
|
|
"valid_targets_min": 2081
|
|
},
|
|
{
|
|
"epoch": 2.679162072767365,
|
|
"grad_norm": 0.5643689010657639,
|
|
"learning_rate": 3.103012908296519e-05,
|
|
"loss": 0.2492,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15316656231880188,
|
|
"step": 4860,
|
|
"valid_targets_mean": 4228.9,
|
|
"valid_targets_min": 2782
|
|
},
|
|
{
|
|
"epoch": 2.6819184123484012,
|
|
"grad_norm": 0.5319588903673653,
|
|
"learning_rate": 3.100718706135521e-05,
|
|
"loss": 0.236,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10256774723529816,
|
|
"step": 4865,
|
|
"valid_targets_mean": 4030.4,
|
|
"valid_targets_min": 1244
|
|
},
|
|
{
|
|
"epoch": 2.6846747519294376,
|
|
"grad_norm": 0.5407009916326783,
|
|
"learning_rate": 3.098422424394824e-05,
|
|
"loss": 0.2406,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10410472750663757,
|
|
"step": 4870,
|
|
"valid_targets_mean": 3429.5,
|
|
"valid_targets_min": 1483
|
|
},
|
|
{
|
|
"epoch": 2.687431091510474,
|
|
"grad_norm": 0.5494062433076609,
|
|
"learning_rate": 3.096124067412774e-05,
|
|
"loss": 0.2507,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13365547358989716,
|
|
"step": 4875,
|
|
"valid_targets_mean": 4149.5,
|
|
"valid_targets_min": 1502
|
|
},
|
|
{
|
|
"epoch": 2.6901874310915104,
|
|
"grad_norm": 0.515736792895505,
|
|
"learning_rate": 3.093823639531641e-05,
|
|
"loss": 0.246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0966002494096756,
|
|
"step": 4880,
|
|
"valid_targets_mean": 3779.1,
|
|
"valid_targets_min": 1522
|
|
},
|
|
{
|
|
"epoch": 2.6929437706725468,
|
|
"grad_norm": 0.6204347160197397,
|
|
"learning_rate": 3.091521145097606e-05,
|
|
"loss": 0.2725,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14043346047401428,
|
|
"step": 4885,
|
|
"valid_targets_mean": 3240.9,
|
|
"valid_targets_min": 1312
|
|
},
|
|
{
|
|
"epoch": 2.695700110253583,
|
|
"grad_norm": 0.6111064265339511,
|
|
"learning_rate": 3.0892165884607545e-05,
|
|
"loss": 0.2451,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13614505529403687,
|
|
"step": 4890,
|
|
"valid_targets_mean": 5018.9,
|
|
"valid_targets_min": 2731
|
|
},
|
|
{
|
|
"epoch": 2.6984564498346195,
|
|
"grad_norm": 0.5756767165989757,
|
|
"learning_rate": 3.086909973975069e-05,
|
|
"loss": 0.246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1301257312297821,
|
|
"step": 4895,
|
|
"valid_targets_mean": 4021.8,
|
|
"valid_targets_min": 1748
|
|
},
|
|
{
|
|
"epoch": 2.701212789415656,
|
|
"grad_norm": 0.5649775223407608,
|
|
"learning_rate": 3.084601305998419e-05,
|
|
"loss": 0.2448,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10901471972465515,
|
|
"step": 4900,
|
|
"valid_targets_mean": 3414.4,
|
|
"valid_targets_min": 2151
|
|
},
|
|
{
|
|
"epoch": 2.7039691289966923,
|
|
"grad_norm": 0.6590650191489065,
|
|
"learning_rate": 3.082290588892553e-05,
|
|
"loss": 0.2531,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1241842657327652,
|
|
"step": 4905,
|
|
"valid_targets_mean": 3216.1,
|
|
"valid_targets_min": 2041
|
|
},
|
|
{
|
|
"epoch": 2.7067254685777287,
|
|
"grad_norm": 0.5908389923339759,
|
|
"learning_rate": 3.079977827023092e-05,
|
|
"loss": 0.2549,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10504922270774841,
|
|
"step": 4910,
|
|
"valid_targets_mean": 2650.9,
|
|
"valid_targets_min": 1233
|
|
},
|
|
{
|
|
"epoch": 2.709481808158765,
|
|
"grad_norm": 0.5326638753476255,
|
|
"learning_rate": 3.077663024759521e-05,
|
|
"loss": 0.2534,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11452393233776093,
|
|
"step": 4915,
|
|
"valid_targets_mean": 3692.6,
|
|
"valid_targets_min": 1382
|
|
},
|
|
{
|
|
"epoch": 2.7122381477398014,
|
|
"grad_norm": 0.5674713293788852,
|
|
"learning_rate": 3.075346186475178e-05,
|
|
"loss": 0.2525,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11910741776227951,
|
|
"step": 4920,
|
|
"valid_targets_mean": 3760.1,
|
|
"valid_targets_min": 1800
|
|
},
|
|
{
|
|
"epoch": 2.714994487320838,
|
|
"grad_norm": 0.629060355735984,
|
|
"learning_rate": 3.0730273165472486e-05,
|
|
"loss": 0.252,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11711858212947845,
|
|
"step": 4925,
|
|
"valid_targets_mean": 3108.4,
|
|
"valid_targets_min": 2324
|
|
},
|
|
{
|
|
"epoch": 2.717750826901874,
|
|
"grad_norm": 0.6362370704485214,
|
|
"learning_rate": 3.070706419356756e-05,
|
|
"loss": 0.2492,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12467217445373535,
|
|
"step": 4930,
|
|
"valid_targets_mean": 3552.0,
|
|
"valid_targets_min": 1211
|
|
},
|
|
{
|
|
"epoch": 2.7205071664829106,
|
|
"grad_norm": 0.484617419374427,
|
|
"learning_rate": 3.0683834992885535e-05,
|
|
"loss": 0.1753,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07266230881214142,
|
|
"step": 4935,
|
|
"valid_targets_mean": 2719.9,
|
|
"valid_targets_min": 540
|
|
},
|
|
{
|
|
"epoch": 2.723263506063947,
|
|
"grad_norm": 0.36340033482200335,
|
|
"learning_rate": 3.066058560731319e-05,
|
|
"loss": 0.1051,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04052020609378815,
|
|
"step": 4940,
|
|
"valid_targets_mean": 3632.6,
|
|
"valid_targets_min": 2726
|
|
},
|
|
{
|
|
"epoch": 2.7260198456449833,
|
|
"grad_norm": 0.3844706426070914,
|
|
"learning_rate": 3.0637316080775396e-05,
|
|
"loss": 0.0996,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03910723328590393,
|
|
"step": 4945,
|
|
"valid_targets_mean": 3739.5,
|
|
"valid_targets_min": 2923
|
|
},
|
|
{
|
|
"epoch": 2.7287761852260197,
|
|
"grad_norm": 0.3978349019971758,
|
|
"learning_rate": 3.0614026457235114e-05,
|
|
"loss": 0.1105,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0680786669254303,
|
|
"step": 4950,
|
|
"valid_targets_mean": 3755.0,
|
|
"valid_targets_min": 2940
|
|
},
|
|
{
|
|
"epoch": 2.731532524807056,
|
|
"grad_norm": 0.41834367276524304,
|
|
"learning_rate": 3.059071678069324e-05,
|
|
"loss": 0.0955,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0548199787735939,
|
|
"step": 4955,
|
|
"valid_targets_mean": 2684.9,
|
|
"valid_targets_min": 557
|
|
},
|
|
{
|
|
"epoch": 2.7342888643880925,
|
|
"grad_norm": 0.38274555687844236,
|
|
"learning_rate": 3.0567387095188586e-05,
|
|
"loss": 0.1026,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049349404871463776,
|
|
"step": 4960,
|
|
"valid_targets_mean": 2843.5,
|
|
"valid_targets_min": 997
|
|
},
|
|
{
|
|
"epoch": 2.737045203969129,
|
|
"grad_norm": 0.6684986233211074,
|
|
"learning_rate": 3.054403744479774e-05,
|
|
"loss": 0.1626,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09795743227005005,
|
|
"step": 4965,
|
|
"valid_targets_mean": 1788.0,
|
|
"valid_targets_min": 920
|
|
},
|
|
{
|
|
"epoch": 2.7398015435501653,
|
|
"grad_norm": 0.3769134875178669,
|
|
"learning_rate": 3.0520667873635044e-05,
|
|
"loss": 0.1006,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044134922325611115,
|
|
"step": 4970,
|
|
"valid_targets_mean": 3279.5,
|
|
"valid_targets_min": 1032
|
|
},
|
|
{
|
|
"epoch": 2.7425578831312016,
|
|
"grad_norm": 0.5267375228143274,
|
|
"learning_rate": 3.0497278425852445e-05,
|
|
"loss": 0.1242,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1012573316693306,
|
|
"step": 4975,
|
|
"valid_targets_mean": 2088.2,
|
|
"valid_targets_min": 878
|
|
},
|
|
{
|
|
"epoch": 2.745314222712238,
|
|
"grad_norm": 0.22560318021537637,
|
|
"learning_rate": 3.0473869145639453e-05,
|
|
"loss": 0.1209,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03094373643398285,
|
|
"step": 4980,
|
|
"valid_targets_mean": 5159.8,
|
|
"valid_targets_min": 3479
|
|
},
|
|
{
|
|
"epoch": 2.7480705622932744,
|
|
"grad_norm": 0.32871994652241737,
|
|
"learning_rate": 3.0450440077223037e-05,
|
|
"loss": 0.0929,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031137222424149513,
|
|
"step": 4985,
|
|
"valid_targets_mean": 3110.0,
|
|
"valid_targets_min": 760
|
|
},
|
|
{
|
|
"epoch": 2.750826901874311,
|
|
"grad_norm": 0.3460274320735208,
|
|
"learning_rate": 3.0426991264867586e-05,
|
|
"loss": 0.0983,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041986722499132156,
|
|
"step": 4990,
|
|
"valid_targets_mean": 3693.5,
|
|
"valid_targets_min": 768
|
|
},
|
|
{
|
|
"epoch": 2.753583241455347,
|
|
"grad_norm": 0.44365402500803014,
|
|
"learning_rate": 3.0403522752874754e-05,
|
|
"loss": 0.0984,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05000906437635422,
|
|
"step": 4995,
|
|
"valid_targets_mean": 2497.2,
|
|
"valid_targets_min": 489
|
|
},
|
|
{
|
|
"epoch": 2.7563395810363835,
|
|
"grad_norm": 0.3005265483160692,
|
|
"learning_rate": 3.0380034585583425e-05,
|
|
"loss": 0.1414,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0402534194290638,
|
|
"step": 5000,
|
|
"valid_targets_mean": 4180.9,
|
|
"valid_targets_min": 1952
|
|
},
|
|
{
|
|
"epoch": 2.75909592061742,
|
|
"grad_norm": 0.2828095033046206,
|
|
"learning_rate": 3.0356526807369624e-05,
|
|
"loss": 0.0923,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.040765151381492615,
|
|
"step": 5005,
|
|
"valid_targets_mean": 3900.0,
|
|
"valid_targets_min": 3681
|
|
},
|
|
{
|
|
"epoch": 2.7618522601984563,
|
|
"grad_norm": 0.39548418107712213,
|
|
"learning_rate": 3.0332999462646414e-05,
|
|
"loss": 0.1079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05122110992670059,
|
|
"step": 5010,
|
|
"valid_targets_mean": 2939.8,
|
|
"valid_targets_min": 836
|
|
},
|
|
{
|
|
"epoch": 2.7646085997794927,
|
|
"grad_norm": 0.33673619595360027,
|
|
"learning_rate": 3.030945259586385e-05,
|
|
"loss": 0.0863,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04271479696035385,
|
|
"step": 5015,
|
|
"valid_targets_mean": 3101.9,
|
|
"valid_targets_min": 858
|
|
},
|
|
{
|
|
"epoch": 2.767364939360529,
|
|
"grad_norm": 0.5614426253462308,
|
|
"learning_rate": 3.0285886251508842e-05,
|
|
"loss": 0.1066,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054506585001945496,
|
|
"step": 5020,
|
|
"valid_targets_mean": 1645.6,
|
|
"valid_targets_min": 512
|
|
},
|
|
{
|
|
"epoch": 2.7701212789415655,
|
|
"grad_norm": 0.451923883880089,
|
|
"learning_rate": 3.026230047410511e-05,
|
|
"loss": 0.1097,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044937796890735626,
|
|
"step": 5025,
|
|
"valid_targets_mean": 2134.5,
|
|
"valid_targets_min": 586
|
|
},
|
|
{
|
|
"epoch": 2.772877618522602,
|
|
"grad_norm": 0.39242261781894094,
|
|
"learning_rate": 3.0238695308213086e-05,
|
|
"loss": 0.113,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05756910890340805,
|
|
"step": 5030,
|
|
"valid_targets_mean": 3461.8,
|
|
"valid_targets_min": 2907
|
|
},
|
|
{
|
|
"epoch": 2.775633958103638,
|
|
"grad_norm": 0.401946683443002,
|
|
"learning_rate": 3.0215070798429837e-05,
|
|
"loss": 0.1157,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07013443857431412,
|
|
"step": 5035,
|
|
"valid_targets_mean": 3035.8,
|
|
"valid_targets_min": 1335
|
|
},
|
|
{
|
|
"epoch": 2.7783902976846746,
|
|
"grad_norm": 0.3776853577022536,
|
|
"learning_rate": 3.0191426989388985e-05,
|
|
"loss": 0.1479,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05581440031528473,
|
|
"step": 5040,
|
|
"valid_targets_mean": 3608.2,
|
|
"valid_targets_min": 2411
|
|
},
|
|
{
|
|
"epoch": 2.781146637265711,
|
|
"grad_norm": 0.36612802245621856,
|
|
"learning_rate": 3.0167763925760592e-05,
|
|
"loss": 0.094,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05209590494632721,
|
|
"step": 5045,
|
|
"valid_targets_mean": 3239.6,
|
|
"valid_targets_min": 665
|
|
},
|
|
{
|
|
"epoch": 2.7839029768467474,
|
|
"grad_norm": 0.6452629242891981,
|
|
"learning_rate": 3.0144081652251107e-05,
|
|
"loss": 0.1151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07512808591127396,
|
|
"step": 5050,
|
|
"valid_targets_mean": 1610.0,
|
|
"valid_targets_min": 698
|
|
},
|
|
{
|
|
"epoch": 2.7866593164277838,
|
|
"grad_norm": 0.5811784776793605,
|
|
"learning_rate": 3.012038021360328e-05,
|
|
"loss": 0.1359,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07257714867591858,
|
|
"step": 5055,
|
|
"valid_targets_mean": 1275.4,
|
|
"valid_targets_min": 173
|
|
},
|
|
{
|
|
"epoch": 2.78941565600882,
|
|
"grad_norm": 0.3654484812715468,
|
|
"learning_rate": 3.0096659654596064e-05,
|
|
"loss": 0.1156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05093298852443695,
|
|
"step": 5060,
|
|
"valid_targets_mean": 3774.6,
|
|
"valid_targets_min": 2819
|
|
},
|
|
{
|
|
"epoch": 2.7921719955898565,
|
|
"grad_norm": 0.4682017072379619,
|
|
"learning_rate": 3.0072920020044533e-05,
|
|
"loss": 0.1627,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14800626039505005,
|
|
"step": 5065,
|
|
"valid_targets_mean": 3735.9,
|
|
"valid_targets_min": 3094
|
|
},
|
|
{
|
|
"epoch": 2.794928335170893,
|
|
"grad_norm": 0.39881180206456995,
|
|
"learning_rate": 3.00491613547998e-05,
|
|
"loss": 0.2548,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04761160537600517,
|
|
"step": 5070,
|
|
"valid_targets_mean": 3513.1,
|
|
"valid_targets_min": 1924
|
|
},
|
|
{
|
|
"epoch": 2.7976846747519293,
|
|
"grad_norm": 0.499012108121896,
|
|
"learning_rate": 3.002538370374894e-05,
|
|
"loss": 0.1185,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06343312561511993,
|
|
"step": 5075,
|
|
"valid_targets_mean": 1947.0,
|
|
"valid_targets_min": 815
|
|
},
|
|
{
|
|
"epoch": 2.8004410143329657,
|
|
"grad_norm": 0.42926051453139163,
|
|
"learning_rate": 3.000158711181489e-05,
|
|
"loss": 0.1124,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07281750440597534,
|
|
"step": 5080,
|
|
"valid_targets_mean": 2904.6,
|
|
"valid_targets_min": 959
|
|
},
|
|
{
|
|
"epoch": 2.803197353914002,
|
|
"grad_norm": 0.3418633762717022,
|
|
"learning_rate": 2.997777162395638e-05,
|
|
"loss": 0.1981,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04314558953046799,
|
|
"step": 5085,
|
|
"valid_targets_mean": 4504.0,
|
|
"valid_targets_min": 2253
|
|
},
|
|
{
|
|
"epoch": 2.8059536934950384,
|
|
"grad_norm": 0.3511804434923005,
|
|
"learning_rate": 2.9953937285167827e-05,
|
|
"loss": 0.0953,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0781187191605568,
|
|
"step": 5090,
|
|
"valid_targets_mean": 4589.0,
|
|
"valid_targets_min": 3860
|
|
},
|
|
{
|
|
"epoch": 2.808710033076075,
|
|
"grad_norm": 0.39648432429000097,
|
|
"learning_rate": 2.993008414047928e-05,
|
|
"loss": 0.1036,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056735169142484665,
|
|
"step": 5095,
|
|
"valid_targets_mean": 3729.4,
|
|
"valid_targets_min": 510
|
|
},
|
|
{
|
|
"epoch": 2.811466372657111,
|
|
"grad_norm": 0.3206254634215307,
|
|
"learning_rate": 2.9906212234956302e-05,
|
|
"loss": 0.1062,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038246236741542816,
|
|
"step": 5100,
|
|
"valid_targets_mean": 4107.1,
|
|
"valid_targets_min": 3247
|
|
},
|
|
{
|
|
"epoch": 2.8142227122381476,
|
|
"grad_norm": 0.3303819279046746,
|
|
"learning_rate": 2.9882321613699905e-05,
|
|
"loss": 0.0951,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04678600654006004,
|
|
"step": 5105,
|
|
"valid_targets_mean": 4856.0,
|
|
"valid_targets_min": 3351
|
|
},
|
|
{
|
|
"epoch": 2.8169790518191844,
|
|
"grad_norm": 0.36710495589327674,
|
|
"learning_rate": 2.9858412321846467e-05,
|
|
"loss": 0.0958,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05426638945937157,
|
|
"step": 5110,
|
|
"valid_targets_mean": 3239.2,
|
|
"valid_targets_min": 1866
|
|
},
|
|
{
|
|
"epoch": 2.8197353914002203,
|
|
"grad_norm": 0.4459394143461181,
|
|
"learning_rate": 2.983448440456764e-05,
|
|
"loss": 0.1001,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057131651788949966,
|
|
"step": 5115,
|
|
"valid_targets_mean": 2683.0,
|
|
"valid_targets_min": 2126
|
|
},
|
|
{
|
|
"epoch": 2.822491730981257,
|
|
"grad_norm": 0.4037021159892385,
|
|
"learning_rate": 2.9810537907070256e-05,
|
|
"loss": 0.0846,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03479522466659546,
|
|
"step": 5120,
|
|
"valid_targets_mean": 2194.0,
|
|
"valid_targets_min": 594
|
|
},
|
|
{
|
|
"epoch": 2.825248070562293,
|
|
"grad_norm": 0.5707432738418996,
|
|
"learning_rate": 2.9786572874596254e-05,
|
|
"loss": 0.1556,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07945086807012558,
|
|
"step": 5125,
|
|
"valid_targets_mean": 2108.0,
|
|
"valid_targets_min": 703
|
|
},
|
|
{
|
|
"epoch": 2.82800441014333,
|
|
"grad_norm": 0.36061410471145355,
|
|
"learning_rate": 2.976258935242259e-05,
|
|
"loss": 0.0775,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03625667840242386,
|
|
"step": 5130,
|
|
"valid_targets_mean": 2886.8,
|
|
"valid_targets_min": 469
|
|
},
|
|
{
|
|
"epoch": 2.830760749724366,
|
|
"grad_norm": 0.4167817990911004,
|
|
"learning_rate": 2.9738587385861164e-05,
|
|
"loss": 0.0926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04535949230194092,
|
|
"step": 5135,
|
|
"valid_targets_mean": 1476.8,
|
|
"valid_targets_min": 613
|
|
},
|
|
{
|
|
"epoch": 2.8335170893054027,
|
|
"grad_norm": 0.3272637552576123,
|
|
"learning_rate": 2.9714567020258712e-05,
|
|
"loss": 0.0838,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04203474149107933,
|
|
"step": 5140,
|
|
"valid_targets_mean": 4232.9,
|
|
"valid_targets_min": 3120
|
|
},
|
|
{
|
|
"epoch": 2.8362734288864386,
|
|
"grad_norm": 0.40643917658457357,
|
|
"learning_rate": 2.9690528300996725e-05,
|
|
"loss": 0.1041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10844376683235168,
|
|
"step": 5145,
|
|
"valid_targets_mean": 2670.0,
|
|
"valid_targets_min": 790
|
|
},
|
|
{
|
|
"epoch": 2.8390297684674755,
|
|
"grad_norm": 0.33205106432354825,
|
|
"learning_rate": 2.9666471273491382e-05,
|
|
"loss": 0.1273,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04660022258758545,
|
|
"step": 5150,
|
|
"valid_targets_mean": 3183.9,
|
|
"valid_targets_min": 724
|
|
},
|
|
{
|
|
"epoch": 2.8417861080485114,
|
|
"grad_norm": 0.5801973370970757,
|
|
"learning_rate": 2.9642395983193452e-05,
|
|
"loss": 0.1297,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051767781376838684,
|
|
"step": 5155,
|
|
"valid_targets_mean": 1333.8,
|
|
"valid_targets_min": 685
|
|
},
|
|
{
|
|
"epoch": 2.844542447629548,
|
|
"grad_norm": 0.6311876963306967,
|
|
"learning_rate": 2.961830247558819e-05,
|
|
"loss": 0.1127,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08383244276046753,
|
|
"step": 5160,
|
|
"valid_targets_mean": 1587.6,
|
|
"valid_targets_min": 660
|
|
},
|
|
{
|
|
"epoch": 2.847298787210584,
|
|
"grad_norm": 0.46338754856694236,
|
|
"learning_rate": 2.9594190796195295e-05,
|
|
"loss": 0.125,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05817720293998718,
|
|
"step": 5165,
|
|
"valid_targets_mean": 2035.5,
|
|
"valid_targets_min": 827
|
|
},
|
|
{
|
|
"epoch": 2.850055126791621,
|
|
"grad_norm": 0.3473892682341388,
|
|
"learning_rate": 2.957006099056878e-05,
|
|
"loss": 0.1396,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05253785848617554,
|
|
"step": 5170,
|
|
"valid_targets_mean": 3783.5,
|
|
"valid_targets_min": 2431
|
|
},
|
|
{
|
|
"epoch": 2.852811466372657,
|
|
"grad_norm": 0.3634166889255431,
|
|
"learning_rate": 2.9545913104296898e-05,
|
|
"loss": 0.1051,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04488586261868477,
|
|
"step": 5175,
|
|
"valid_targets_mean": 3359.5,
|
|
"valid_targets_min": 1344
|
|
},
|
|
{
|
|
"epoch": 2.8555678059536937,
|
|
"grad_norm": 0.4021918478544107,
|
|
"learning_rate": 2.9521747183002086e-05,
|
|
"loss": 0.1231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03999020904302597,
|
|
"step": 5180,
|
|
"valid_targets_mean": 2813.1,
|
|
"valid_targets_min": 751
|
|
},
|
|
{
|
|
"epoch": 2.8583241455347297,
|
|
"grad_norm": 0.421893271489034,
|
|
"learning_rate": 2.9497563272340834e-05,
|
|
"loss": 0.108,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04774616286158562,
|
|
"step": 5185,
|
|
"valid_targets_mean": 1752.9,
|
|
"valid_targets_min": 482
|
|
},
|
|
{
|
|
"epoch": 2.8610804851157665,
|
|
"grad_norm": 0.43134934300121197,
|
|
"learning_rate": 2.947336141800363e-05,
|
|
"loss": 0.0959,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04139981418848038,
|
|
"step": 5190,
|
|
"valid_targets_mean": 3314.1,
|
|
"valid_targets_min": 805
|
|
},
|
|
{
|
|
"epoch": 2.8638368246968025,
|
|
"grad_norm": 0.25865470151829983,
|
|
"learning_rate": 2.9449141665714846e-05,
|
|
"loss": 0.0956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037146370857954025,
|
|
"step": 5195,
|
|
"valid_targets_mean": 5540.6,
|
|
"valid_targets_min": 5089
|
|
},
|
|
{
|
|
"epoch": 2.8665931642778393,
|
|
"grad_norm": 0.4177903034539172,
|
|
"learning_rate": 2.9424904061232687e-05,
|
|
"loss": 0.0967,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05946866795420647,
|
|
"step": 5200,
|
|
"valid_targets_mean": 3626.1,
|
|
"valid_targets_min": 879
|
|
},
|
|
{
|
|
"epoch": 2.869349503858875,
|
|
"grad_norm": 0.3436122143974236,
|
|
"learning_rate": 2.9400648650349094e-05,
|
|
"loss": 0.0913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04458882659673691,
|
|
"step": 5205,
|
|
"valid_targets_mean": 3559.1,
|
|
"valid_targets_min": 934
|
|
},
|
|
{
|
|
"epoch": 2.872105843439912,
|
|
"grad_norm": 0.4133030841805535,
|
|
"learning_rate": 2.9376375478889613e-05,
|
|
"loss": 0.1078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0834861546754837,
|
|
"step": 5210,
|
|
"valid_targets_mean": 2775.2,
|
|
"valid_targets_min": 804
|
|
},
|
|
{
|
|
"epoch": 2.874862183020948,
|
|
"grad_norm": 0.4148029246507272,
|
|
"learning_rate": 2.9352084592713378e-05,
|
|
"loss": 0.1031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05819116160273552,
|
|
"step": 5215,
|
|
"valid_targets_mean": 3218.0,
|
|
"valid_targets_min": 1803
|
|
},
|
|
{
|
|
"epoch": 2.877618522601985,
|
|
"grad_norm": 0.3922955915629398,
|
|
"learning_rate": 2.9327776037712978e-05,
|
|
"loss": 0.1101,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06251122802495956,
|
|
"step": 5220,
|
|
"valid_targets_mean": 3551.5,
|
|
"valid_targets_min": 2193
|
|
},
|
|
{
|
|
"epoch": 2.8803748621830207,
|
|
"grad_norm": 0.3348310002113072,
|
|
"learning_rate": 2.9303449859814385e-05,
|
|
"loss": 0.0916,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042417652904987335,
|
|
"step": 5225,
|
|
"valid_targets_mean": 3896.2,
|
|
"valid_targets_min": 2937
|
|
},
|
|
{
|
|
"epoch": 2.8831312017640576,
|
|
"grad_norm": 0.4220257901737759,
|
|
"learning_rate": 2.927910610497687e-05,
|
|
"loss": 0.0993,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056269384920597076,
|
|
"step": 5230,
|
|
"valid_targets_mean": 2332.9,
|
|
"valid_targets_min": 880
|
|
},
|
|
{
|
|
"epoch": 2.8858875413450935,
|
|
"grad_norm": 0.331726030026401,
|
|
"learning_rate": 2.925474481919291e-05,
|
|
"loss": 0.104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045178383588790894,
|
|
"step": 5235,
|
|
"valid_targets_mean": 3415.6,
|
|
"valid_targets_min": 2300
|
|
},
|
|
{
|
|
"epoch": 2.8886438809261303,
|
|
"grad_norm": 0.37962554103409885,
|
|
"learning_rate": 2.9230366048488097e-05,
|
|
"loss": 0.0837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043932266533374786,
|
|
"step": 5240,
|
|
"valid_targets_mean": 3456.6,
|
|
"valid_targets_min": 1028
|
|
},
|
|
{
|
|
"epoch": 2.8914002205071663,
|
|
"grad_norm": 0.33575189666443794,
|
|
"learning_rate": 2.9205969838921065e-05,
|
|
"loss": 0.1074,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046517737209796906,
|
|
"step": 5245,
|
|
"valid_targets_mean": 2957.1,
|
|
"valid_targets_min": 528
|
|
},
|
|
{
|
|
"epoch": 2.894156560088203,
|
|
"grad_norm": 0.31405049031239946,
|
|
"learning_rate": 2.9181556236583385e-05,
|
|
"loss": 0.0998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042906031012535095,
|
|
"step": 5250,
|
|
"valid_targets_mean": 3214.2,
|
|
"valid_targets_min": 997
|
|
},
|
|
{
|
|
"epoch": 2.896912899669239,
|
|
"grad_norm": 0.455623179635456,
|
|
"learning_rate": 2.9157125287599508e-05,
|
|
"loss": 0.1218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03529912978410721,
|
|
"step": 5255,
|
|
"valid_targets_mean": 1525.6,
|
|
"valid_targets_min": 594
|
|
},
|
|
{
|
|
"epoch": 2.899669239250276,
|
|
"grad_norm": 0.4094077554438432,
|
|
"learning_rate": 2.9132677038126637e-05,
|
|
"loss": 0.1009,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03489510342478752,
|
|
"step": 5260,
|
|
"valid_targets_mean": 1607.8,
|
|
"valid_targets_min": 662
|
|
},
|
|
{
|
|
"epoch": 2.902425578831312,
|
|
"grad_norm": 0.5643122919486534,
|
|
"learning_rate": 2.9108211534354672e-05,
|
|
"loss": 0.1184,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05163051187992096,
|
|
"step": 5265,
|
|
"valid_targets_mean": 1200.2,
|
|
"valid_targets_min": 849
|
|
},
|
|
{
|
|
"epoch": 2.9051819184123486,
|
|
"grad_norm": 0.6714198225106331,
|
|
"learning_rate": 2.9083728822506103e-05,
|
|
"loss": 0.1092,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07662265002727509,
|
|
"step": 5270,
|
|
"valid_targets_mean": 1718.0,
|
|
"valid_targets_min": 892
|
|
},
|
|
{
|
|
"epoch": 2.9079382579933846,
|
|
"grad_norm": 0.49918059729559333,
|
|
"learning_rate": 2.9059228948835934e-05,
|
|
"loss": 0.1068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057096295058727264,
|
|
"step": 5275,
|
|
"valid_targets_mean": 2259.9,
|
|
"valid_targets_min": 735
|
|
},
|
|
{
|
|
"epoch": 2.9106945975744214,
|
|
"grad_norm": 0.31587873566061625,
|
|
"learning_rate": 2.9034711959631607e-05,
|
|
"loss": 0.0958,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03585599735379219,
|
|
"step": 5280,
|
|
"valid_targets_mean": 3359.8,
|
|
"valid_targets_min": 2814
|
|
},
|
|
{
|
|
"epoch": 2.9134509371554573,
|
|
"grad_norm": 0.35126871168674884,
|
|
"learning_rate": 2.901017790121288e-05,
|
|
"loss": 0.0944,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042547404766082764,
|
|
"step": 5285,
|
|
"valid_targets_mean": 3794.9,
|
|
"valid_targets_min": 2873
|
|
},
|
|
{
|
|
"epoch": 2.916207276736494,
|
|
"grad_norm": 0.5139436225755849,
|
|
"learning_rate": 2.898562681993177e-05,
|
|
"loss": 0.125,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05583256855607033,
|
|
"step": 5290,
|
|
"valid_targets_mean": 1490.1,
|
|
"valid_targets_min": 443
|
|
},
|
|
{
|
|
"epoch": 2.91896361631753,
|
|
"grad_norm": 0.37783500303476836,
|
|
"learning_rate": 2.8961058762172448e-05,
|
|
"loss": 0.0882,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04075084626674652,
|
|
"step": 5295,
|
|
"valid_targets_mean": 2663.0,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 2.921719955898567,
|
|
"grad_norm": 0.4098014043788719,
|
|
"learning_rate": 2.8936473774351167e-05,
|
|
"loss": 0.0964,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06916703283786774,
|
|
"step": 5300,
|
|
"valid_targets_mean": 3080.0,
|
|
"valid_targets_min": 795
|
|
},
|
|
{
|
|
"epoch": 2.924476295479603,
|
|
"grad_norm": 0.35986730343189643,
|
|
"learning_rate": 2.891187190291617e-05,
|
|
"loss": 0.0958,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04997006803750992,
|
|
"step": 5305,
|
|
"valid_targets_mean": 4051.4,
|
|
"valid_targets_min": 2461
|
|
},
|
|
{
|
|
"epoch": 2.9272326350606397,
|
|
"grad_norm": 0.2890546153765392,
|
|
"learning_rate": 2.888725319434758e-05,
|
|
"loss": 0.0826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03368262201547623,
|
|
"step": 5310,
|
|
"valid_targets_mean": 3571.1,
|
|
"valid_targets_min": 856
|
|
},
|
|
{
|
|
"epoch": 2.9299889746416756,
|
|
"grad_norm": 0.38966239020306725,
|
|
"learning_rate": 2.886261769515734e-05,
|
|
"loss": 0.0913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0612705759704113,
|
|
"step": 5315,
|
|
"valid_targets_mean": 3969.9,
|
|
"valid_targets_min": 3010
|
|
},
|
|
{
|
|
"epoch": 2.9327453142227125,
|
|
"grad_norm": 0.3251461445601138,
|
|
"learning_rate": 2.883796545188912e-05,
|
|
"loss": 0.0989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.060133226215839386,
|
|
"step": 5320,
|
|
"valid_targets_mean": 3696.4,
|
|
"valid_targets_min": 2055
|
|
},
|
|
{
|
|
"epoch": 2.9355016538037484,
|
|
"grad_norm": 0.40295866502201705,
|
|
"learning_rate": 2.8813296511118234e-05,
|
|
"loss": 0.1195,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.058293890208005905,
|
|
"step": 5325,
|
|
"valid_targets_mean": 3658.6,
|
|
"valid_targets_min": 2646
|
|
},
|
|
{
|
|
"epoch": 2.938257993384785,
|
|
"grad_norm": 0.40465714910192857,
|
|
"learning_rate": 2.8788610919451504e-05,
|
|
"loss": 0.1151,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12296296656131744,
|
|
"step": 5330,
|
|
"valid_targets_mean": 3360.6,
|
|
"valid_targets_min": 953
|
|
},
|
|
{
|
|
"epoch": 2.941014332965821,
|
|
"grad_norm": 0.42970725492986217,
|
|
"learning_rate": 2.8763908723527256e-05,
|
|
"loss": 0.0834,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04473197087645531,
|
|
"step": 5335,
|
|
"valid_targets_mean": 1481.0,
|
|
"valid_targets_min": 811
|
|
},
|
|
{
|
|
"epoch": 2.943770672546858,
|
|
"grad_norm": 0.3761757839956298,
|
|
"learning_rate": 2.8739189970015152e-05,
|
|
"loss": 0.1202,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.048818476498126984,
|
|
"step": 5340,
|
|
"valid_targets_mean": 3004.9,
|
|
"valid_targets_min": 719
|
|
},
|
|
{
|
|
"epoch": 2.946527012127894,
|
|
"grad_norm": 0.3691460696801811,
|
|
"learning_rate": 2.8714454705616157e-05,
|
|
"loss": 0.0952,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05576256290078163,
|
|
"step": 5345,
|
|
"valid_targets_mean": 3818.9,
|
|
"valid_targets_min": 3374
|
|
},
|
|
{
|
|
"epoch": 2.9492833517089307,
|
|
"grad_norm": 0.3576692724246759,
|
|
"learning_rate": 2.868970297706242e-05,
|
|
"loss": 0.1022,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044196829199790955,
|
|
"step": 5350,
|
|
"valid_targets_mean": 3588.9,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 2.9520396912899667,
|
|
"grad_norm": 0.5408520783477055,
|
|
"learning_rate": 2.8664934831117205e-05,
|
|
"loss": 0.1136,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10145282745361328,
|
|
"step": 5355,
|
|
"valid_targets_mean": 2562.2,
|
|
"valid_targets_min": 837
|
|
},
|
|
{
|
|
"epoch": 2.9547960308710035,
|
|
"grad_norm": 0.4050597118916316,
|
|
"learning_rate": 2.864015031457478e-05,
|
|
"loss": 0.1445,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09561051428318024,
|
|
"step": 5360,
|
|
"valid_targets_mean": 2838.9,
|
|
"valid_targets_min": 1043
|
|
},
|
|
{
|
|
"epoch": 2.9575523704520394,
|
|
"grad_norm": 0.41880705001519125,
|
|
"learning_rate": 2.8615349474260345e-05,
|
|
"loss": 0.1086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.040141619741916656,
|
|
"step": 5365,
|
|
"valid_targets_mean": 2451.2,
|
|
"valid_targets_min": 659
|
|
},
|
|
{
|
|
"epoch": 2.9603087100330763,
|
|
"grad_norm": 0.4565300607992464,
|
|
"learning_rate": 2.8590532357029955e-05,
|
|
"loss": 0.099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05463661253452301,
|
|
"step": 5370,
|
|
"valid_targets_mean": 1652.0,
|
|
"valid_targets_min": 713
|
|
},
|
|
{
|
|
"epoch": 2.963065049614112,
|
|
"grad_norm": 0.3891108556072015,
|
|
"learning_rate": 2.8565699009770393e-05,
|
|
"loss": 0.0956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05098123475909233,
|
|
"step": 5375,
|
|
"valid_targets_mean": 2270.1,
|
|
"valid_targets_min": 539
|
|
},
|
|
{
|
|
"epoch": 2.965821389195149,
|
|
"grad_norm": 0.3971695367227798,
|
|
"learning_rate": 2.8540849479399134e-05,
|
|
"loss": 0.0955,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05267088860273361,
|
|
"step": 5380,
|
|
"valid_targets_mean": 3042.4,
|
|
"valid_targets_min": 1014
|
|
},
|
|
{
|
|
"epoch": 2.968577728776185,
|
|
"grad_norm": 0.3679980376634279,
|
|
"learning_rate": 2.8515983812864197e-05,
|
|
"loss": 0.0861,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04018281027674675,
|
|
"step": 5385,
|
|
"valid_targets_mean": 1789.0,
|
|
"valid_targets_min": 631
|
|
},
|
|
{
|
|
"epoch": 2.971334068357222,
|
|
"grad_norm": 0.3546894261163911,
|
|
"learning_rate": 2.849110205714411e-05,
|
|
"loss": 0.1201,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03981973975896835,
|
|
"step": 5390,
|
|
"valid_targets_mean": 2367.8,
|
|
"valid_targets_min": 790
|
|
},
|
|
{
|
|
"epoch": 2.9740904079382577,
|
|
"grad_norm": 0.3561670713886429,
|
|
"learning_rate": 2.8466204259247786e-05,
|
|
"loss": 0.0858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03286431357264519,
|
|
"step": 5395,
|
|
"valid_targets_mean": 1381.9,
|
|
"valid_targets_min": 700
|
|
},
|
|
{
|
|
"epoch": 2.9768467475192946,
|
|
"grad_norm": 0.36033913580407506,
|
|
"learning_rate": 2.8441290466214453e-05,
|
|
"loss": 0.0964,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035906389355659485,
|
|
"step": 5400,
|
|
"valid_targets_mean": 2650.1,
|
|
"valid_targets_min": 527
|
|
},
|
|
{
|
|
"epoch": 2.9796030871003305,
|
|
"grad_norm": 0.3838597774437746,
|
|
"learning_rate": 2.841636072511355e-05,
|
|
"loss": 0.1028,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05610041692852974,
|
|
"step": 5405,
|
|
"valid_targets_mean": 3526.8,
|
|
"valid_targets_min": 2880
|
|
},
|
|
{
|
|
"epoch": 2.9823594266813673,
|
|
"grad_norm": 0.2956362544670108,
|
|
"learning_rate": 2.839141508304466e-05,
|
|
"loss": 0.094,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0408497229218483,
|
|
"step": 5410,
|
|
"valid_targets_mean": 3915.8,
|
|
"valid_targets_min": 513
|
|
},
|
|
{
|
|
"epoch": 2.9851157662624033,
|
|
"grad_norm": 0.29419682858869745,
|
|
"learning_rate": 2.8366453587137393e-05,
|
|
"loss": 0.0856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03885408490896225,
|
|
"step": 5415,
|
|
"valid_targets_mean": 3510.1,
|
|
"valid_targets_min": 1025
|
|
},
|
|
{
|
|
"epoch": 2.98787210584344,
|
|
"grad_norm": 0.2867009689790259,
|
|
"learning_rate": 2.8341476284551324e-05,
|
|
"loss": 0.0903,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046296630054712296,
|
|
"step": 5420,
|
|
"valid_targets_mean": 3487.0,
|
|
"valid_targets_min": 601
|
|
},
|
|
{
|
|
"epoch": 2.990628445424476,
|
|
"grad_norm": 0.5005574737234342,
|
|
"learning_rate": 2.8316483222475883e-05,
|
|
"loss": 0.1673,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.16088847815990448,
|
|
"step": 5425,
|
|
"valid_targets_mean": 2794.5,
|
|
"valid_targets_min": 1898
|
|
},
|
|
{
|
|
"epoch": 2.993384785005513,
|
|
"grad_norm": 0.3551897950496882,
|
|
"learning_rate": 2.829147444813028e-05,
|
|
"loss": 0.1153,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04540003091096878,
|
|
"step": 5430,
|
|
"valid_targets_mean": 2930.8,
|
|
"valid_targets_min": 1228
|
|
},
|
|
{
|
|
"epoch": 2.996141124586549,
|
|
"grad_norm": 0.38842927297160135,
|
|
"learning_rate": 2.82664500087634e-05,
|
|
"loss": 0.1007,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051071420311927795,
|
|
"step": 5435,
|
|
"valid_targets_mean": 2879.8,
|
|
"valid_targets_min": 713
|
|
},
|
|
{
|
|
"epoch": 2.9988974641675856,
|
|
"grad_norm": 0.3754810951896936,
|
|
"learning_rate": 2.8241409951653732e-05,
|
|
"loss": 0.0878,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04938063025474548,
|
|
"step": 5440,
|
|
"valid_targets_mean": 2256.6,
|
|
"valid_targets_min": 1193
|
|
},
|
|
{
|
|
"epoch": 3.001653803748622,
|
|
"grad_norm": 0.5345421324863573,
|
|
"learning_rate": 2.8216354324109284e-05,
|
|
"loss": 0.1996,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10521845519542694,
|
|
"step": 5445,
|
|
"valid_targets_mean": 5409.0,
|
|
"valid_targets_min": 3500
|
|
},
|
|
{
|
|
"epoch": 3.0044101433296584,
|
|
"grad_norm": 0.3723728554498553,
|
|
"learning_rate": 2.819128317346745e-05,
|
|
"loss": 0.2096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10074079781770706,
|
|
"step": 5450,
|
|
"valid_targets_mean": 7623.9,
|
|
"valid_targets_min": 6322
|
|
},
|
|
{
|
|
"epoch": 3.0071664829106948,
|
|
"grad_norm": 0.3511131071220079,
|
|
"learning_rate": 2.8166196547094986e-05,
|
|
"loss": 0.2002,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11535365134477615,
|
|
"step": 5455,
|
|
"valid_targets_mean": 7311.9,
|
|
"valid_targets_min": 5722
|
|
},
|
|
{
|
|
"epoch": 3.009922822491731,
|
|
"grad_norm": 0.37679923509245017,
|
|
"learning_rate": 2.8141094492387853e-05,
|
|
"loss": 0.1949,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08716094493865967,
|
|
"step": 5460,
|
|
"valid_targets_mean": 7635.8,
|
|
"valid_targets_min": 5634
|
|
},
|
|
{
|
|
"epoch": 3.0126791620727675,
|
|
"grad_norm": 0.3665972166321447,
|
|
"learning_rate": 2.811597705677118e-05,
|
|
"loss": 0.1954,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1011427715420723,
|
|
"step": 5465,
|
|
"valid_targets_mean": 7440.2,
|
|
"valid_targets_min": 5614
|
|
},
|
|
{
|
|
"epoch": 3.015435501653804,
|
|
"grad_norm": 0.3836304388678673,
|
|
"learning_rate": 2.8090844287699165e-05,
|
|
"loss": 0.1956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11480289697647095,
|
|
"step": 5470,
|
|
"valid_targets_mean": 7015.1,
|
|
"valid_targets_min": 5539
|
|
},
|
|
{
|
|
"epoch": 3.0181918412348403,
|
|
"grad_norm": 0.3737007968520039,
|
|
"learning_rate": 2.806569623265495e-05,
|
|
"loss": 0.2114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11037489771842957,
|
|
"step": 5475,
|
|
"valid_targets_mean": 7799.6,
|
|
"valid_targets_min": 5373
|
|
},
|
|
{
|
|
"epoch": 3.0209481808158767,
|
|
"grad_norm": 0.37062097027438295,
|
|
"learning_rate": 2.8040532939150577e-05,
|
|
"loss": 0.2002,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10399997979402542,
|
|
"step": 5480,
|
|
"valid_targets_mean": 6996.5,
|
|
"valid_targets_min": 5640
|
|
},
|
|
{
|
|
"epoch": 3.023704520396913,
|
|
"grad_norm": 0.3419865630015972,
|
|
"learning_rate": 2.8015354454726863e-05,
|
|
"loss": 0.1912,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0938965380191803,
|
|
"step": 5485,
|
|
"valid_targets_mean": 7040.8,
|
|
"valid_targets_min": 5071
|
|
},
|
|
{
|
|
"epoch": 3.0264608599779494,
|
|
"grad_norm": 0.3701113946869666,
|
|
"learning_rate": 2.7990160826953336e-05,
|
|
"loss": 0.1802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09258505702018738,
|
|
"step": 5490,
|
|
"valid_targets_mean": 6681.4,
|
|
"valid_targets_min": 5437
|
|
},
|
|
{
|
|
"epoch": 3.029217199558986,
|
|
"grad_norm": 0.35841632613948493,
|
|
"learning_rate": 2.796495210342813e-05,
|
|
"loss": 0.1949,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09194891154766083,
|
|
"step": 5495,
|
|
"valid_targets_mean": 7258.4,
|
|
"valid_targets_min": 5481
|
|
},
|
|
{
|
|
"epoch": 3.031973539140022,
|
|
"grad_norm": 0.3771078234011868,
|
|
"learning_rate": 2.7939728331777892e-05,
|
|
"loss": 0.1847,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09178681671619415,
|
|
"step": 5500,
|
|
"valid_targets_mean": 6993.5,
|
|
"valid_targets_min": 5666
|
|
},
|
|
{
|
|
"epoch": 3.0347298787210586,
|
|
"grad_norm": 0.3632833659930408,
|
|
"learning_rate": 2.7914489559657726e-05,
|
|
"loss": 0.1941,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1003008708357811,
|
|
"step": 5505,
|
|
"valid_targets_mean": 6706.2,
|
|
"valid_targets_min": 5518
|
|
},
|
|
{
|
|
"epoch": 3.037486218302095,
|
|
"grad_norm": 0.3671150905630996,
|
|
"learning_rate": 2.788923583475104e-05,
|
|
"loss": 0.1803,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08863619714975357,
|
|
"step": 5510,
|
|
"valid_targets_mean": 6317.2,
|
|
"valid_targets_min": 5034
|
|
},
|
|
{
|
|
"epoch": 3.0402425578831314,
|
|
"grad_norm": 0.3833830202083012,
|
|
"learning_rate": 2.786396720476951e-05,
|
|
"loss": 0.1845,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09462518990039825,
|
|
"step": 5515,
|
|
"valid_targets_mean": 6752.5,
|
|
"valid_targets_min": 5586
|
|
},
|
|
{
|
|
"epoch": 3.0429988974641677,
|
|
"grad_norm": 0.38601857474048024,
|
|
"learning_rate": 2.7838683717452988e-05,
|
|
"loss": 0.1926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1082678884267807,
|
|
"step": 5520,
|
|
"valid_targets_mean": 7536.1,
|
|
"valid_targets_min": 5824
|
|
},
|
|
{
|
|
"epoch": 3.045755237045204,
|
|
"grad_norm": 0.36615106864908703,
|
|
"learning_rate": 2.7813385420569364e-05,
|
|
"loss": 0.1838,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08990936726331711,
|
|
"step": 5525,
|
|
"valid_targets_mean": 7168.0,
|
|
"valid_targets_min": 4761
|
|
},
|
|
{
|
|
"epoch": 3.0485115766262405,
|
|
"grad_norm": 0.3764343765533296,
|
|
"learning_rate": 2.778807236191452e-05,
|
|
"loss": 0.1894,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09619498252868652,
|
|
"step": 5530,
|
|
"valid_targets_mean": 6651.6,
|
|
"valid_targets_min": 5318
|
|
},
|
|
{
|
|
"epoch": 3.051267916207277,
|
|
"grad_norm": 0.37639901398421144,
|
|
"learning_rate": 2.7762744589312238e-05,
|
|
"loss": 0.203,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09997943043708801,
|
|
"step": 5535,
|
|
"valid_targets_mean": 6741.1,
|
|
"valid_targets_min": 5735
|
|
},
|
|
{
|
|
"epoch": 3.0540242557883133,
|
|
"grad_norm": 0.7184669207033333,
|
|
"learning_rate": 2.7737402150614098e-05,
|
|
"loss": 0.1992,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0734001100063324,
|
|
"step": 5540,
|
|
"valid_targets_mean": 1413.6,
|
|
"valid_targets_min": 287
|
|
},
|
|
{
|
|
"epoch": 3.0567805953693497,
|
|
"grad_norm": 0.4716672075846354,
|
|
"learning_rate": 2.7712045093699372e-05,
|
|
"loss": 0.1659,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09860876202583313,
|
|
"step": 5545,
|
|
"valid_targets_mean": 6754.0,
|
|
"valid_targets_min": 5024
|
|
},
|
|
{
|
|
"epoch": 3.059536934950386,
|
|
"grad_norm": 0.3687557952209285,
|
|
"learning_rate": 2.768667346647496e-05,
|
|
"loss": 0.1829,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09052016586065292,
|
|
"step": 5550,
|
|
"valid_targets_mean": 6746.0,
|
|
"valid_targets_min": 3856
|
|
},
|
|
{
|
|
"epoch": 3.0622932745314224,
|
|
"grad_norm": 0.47667576603532547,
|
|
"learning_rate": 2.766128731687529e-05,
|
|
"loss": 0.1884,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09741353988647461,
|
|
"step": 5555,
|
|
"valid_targets_mean": 7015.4,
|
|
"valid_targets_min": 6101
|
|
},
|
|
{
|
|
"epoch": 3.065049614112459,
|
|
"grad_norm": 0.449052909861335,
|
|
"learning_rate": 2.763588669286223e-05,
|
|
"loss": 0.1911,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10042963922023773,
|
|
"step": 5560,
|
|
"valid_targets_mean": 6858.5,
|
|
"valid_targets_min": 5844
|
|
},
|
|
{
|
|
"epoch": 3.067805953693495,
|
|
"grad_norm": 0.3435701092085784,
|
|
"learning_rate": 2.761047164242499e-05,
|
|
"loss": 0.1865,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08629290014505386,
|
|
"step": 5565,
|
|
"valid_targets_mean": 7529.2,
|
|
"valid_targets_min": 6113
|
|
},
|
|
{
|
|
"epoch": 3.0705622932745316,
|
|
"grad_norm": 0.31262931770570923,
|
|
"learning_rate": 2.758504221358003e-05,
|
|
"loss": 0.1754,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08457375317811966,
|
|
"step": 5570,
|
|
"valid_targets_mean": 8891.4,
|
|
"valid_targets_min": 4760
|
|
},
|
|
{
|
|
"epoch": 3.073318632855568,
|
|
"grad_norm": 0.30433355717783583,
|
|
"learning_rate": 2.7559598454370995e-05,
|
|
"loss": 0.1629,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07489997893571854,
|
|
"step": 5575,
|
|
"valid_targets_mean": 7368.4,
|
|
"valid_targets_min": 5090
|
|
},
|
|
{
|
|
"epoch": 3.0760749724366043,
|
|
"grad_norm": 0.3129318050147033,
|
|
"learning_rate": 2.753414041286858e-05,
|
|
"loss": 0.1669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08243705332279205,
|
|
"step": 5580,
|
|
"valid_targets_mean": 8453.1,
|
|
"valid_targets_min": 4010
|
|
},
|
|
{
|
|
"epoch": 3.0788313120176407,
|
|
"grad_norm": 0.3706830509457027,
|
|
"learning_rate": 2.7508668137170473e-05,
|
|
"loss": 0.1855,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09332611411809921,
|
|
"step": 5585,
|
|
"valid_targets_mean": 6708.9,
|
|
"valid_targets_min": 4625
|
|
},
|
|
{
|
|
"epoch": 3.081587651598677,
|
|
"grad_norm": 0.40546012111778607,
|
|
"learning_rate": 2.748318167540126e-05,
|
|
"loss": 0.1773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0961044579744339,
|
|
"step": 5590,
|
|
"valid_targets_mean": 7523.5,
|
|
"valid_targets_min": 5429
|
|
},
|
|
{
|
|
"epoch": 3.0843439911797135,
|
|
"grad_norm": 0.41098814846925885,
|
|
"learning_rate": 2.7457681075712326e-05,
|
|
"loss": 0.1967,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11096566915512085,
|
|
"step": 5595,
|
|
"valid_targets_mean": 7059.1,
|
|
"valid_targets_min": 4847
|
|
},
|
|
{
|
|
"epoch": 3.08710033076075,
|
|
"grad_norm": 0.4326773073224714,
|
|
"learning_rate": 2.7432166386281768e-05,
|
|
"loss": 0.1998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08876191079616547,
|
|
"step": 5600,
|
|
"valid_targets_mean": 6242.9,
|
|
"valid_targets_min": 4585
|
|
},
|
|
{
|
|
"epoch": 3.0898566703417862,
|
|
"grad_norm": 0.356159539496199,
|
|
"learning_rate": 2.7406637655314288e-05,
|
|
"loss": 0.1982,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0887247771024704,
|
|
"step": 5605,
|
|
"valid_targets_mean": 6787.5,
|
|
"valid_targets_min": 5493
|
|
},
|
|
{
|
|
"epoch": 3.0926130099228226,
|
|
"grad_norm": 0.4686461006808657,
|
|
"learning_rate": 2.7381094931041134e-05,
|
|
"loss": 0.1846,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0938146635890007,
|
|
"step": 5610,
|
|
"valid_targets_mean": 7336.0,
|
|
"valid_targets_min": 4968
|
|
},
|
|
{
|
|
"epoch": 3.095369349503859,
|
|
"grad_norm": 0.3885233796003193,
|
|
"learning_rate": 2.7355538261719994e-05,
|
|
"loss": 0.1903,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08492517471313477,
|
|
"step": 5615,
|
|
"valid_targets_mean": 6181.1,
|
|
"valid_targets_min": 3964
|
|
},
|
|
{
|
|
"epoch": 3.0981256890848954,
|
|
"grad_norm": 0.34420893791692747,
|
|
"learning_rate": 2.7329967695634885e-05,
|
|
"loss": 0.1912,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0935019850730896,
|
|
"step": 5620,
|
|
"valid_targets_mean": 8367.6,
|
|
"valid_targets_min": 6099
|
|
},
|
|
{
|
|
"epoch": 3.1008820286659318,
|
|
"grad_norm": 0.35996413668301924,
|
|
"learning_rate": 2.7304383281096096e-05,
|
|
"loss": 0.1901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09571622312068939,
|
|
"step": 5625,
|
|
"valid_targets_mean": 7629.2,
|
|
"valid_targets_min": 4042
|
|
},
|
|
{
|
|
"epoch": 3.103638368246968,
|
|
"grad_norm": 0.3469173124669671,
|
|
"learning_rate": 2.7278785066440058e-05,
|
|
"loss": 0.1923,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09375506639480591,
|
|
"step": 5630,
|
|
"valid_targets_mean": 7061.6,
|
|
"valid_targets_min": 5185
|
|
},
|
|
{
|
|
"epoch": 3.1063947078280045,
|
|
"grad_norm": 0.35263443809134365,
|
|
"learning_rate": 2.7253173100029297e-05,
|
|
"loss": 0.1921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09024305641651154,
|
|
"step": 5635,
|
|
"valid_targets_mean": 6969.5,
|
|
"valid_targets_min": 5329
|
|
},
|
|
{
|
|
"epoch": 3.109151047409041,
|
|
"grad_norm": 0.3633610421975031,
|
|
"learning_rate": 2.7227547430252315e-05,
|
|
"loss": 0.1931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09405043721199036,
|
|
"step": 5640,
|
|
"valid_targets_mean": 6327.2,
|
|
"valid_targets_min": 5204
|
|
},
|
|
{
|
|
"epoch": 3.1119073869900773,
|
|
"grad_norm": 0.42303095444387007,
|
|
"learning_rate": 2.7201908105523496e-05,
|
|
"loss": 0.1897,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10259703546762466,
|
|
"step": 5645,
|
|
"valid_targets_mean": 5908.1,
|
|
"valid_targets_min": 4666
|
|
},
|
|
{
|
|
"epoch": 3.1146637265711137,
|
|
"grad_norm": 1.3636784014690222,
|
|
"learning_rate": 2.717625517428303e-05,
|
|
"loss": 0.1625,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03642156720161438,
|
|
"step": 5650,
|
|
"valid_targets_mean": 207.1,
|
|
"valid_targets_min": 133
|
|
},
|
|
{
|
|
"epoch": 3.11742006615215,
|
|
"grad_norm": 0.4465692042172449,
|
|
"learning_rate": 2.7150588684996803e-05,
|
|
"loss": 0.193,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09883013367652893,
|
|
"step": 5655,
|
|
"valid_targets_mean": 5976.6,
|
|
"valid_targets_min": 4924
|
|
},
|
|
{
|
|
"epoch": 3.1201764057331864,
|
|
"grad_norm": 0.36632543071613327,
|
|
"learning_rate": 2.712490868615633e-05,
|
|
"loss": 0.1842,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08314917981624603,
|
|
"step": 5660,
|
|
"valid_targets_mean": 6010.1,
|
|
"valid_targets_min": 4785
|
|
},
|
|
{
|
|
"epoch": 3.122932745314223,
|
|
"grad_norm": 0.3740453463868335,
|
|
"learning_rate": 2.7099215226278635e-05,
|
|
"loss": 0.1914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09292158484458923,
|
|
"step": 5665,
|
|
"valid_targets_mean": 6578.8,
|
|
"valid_targets_min": 5070
|
|
},
|
|
{
|
|
"epoch": 3.125689084895259,
|
|
"grad_norm": 0.3921550247125368,
|
|
"learning_rate": 2.707350835390619e-05,
|
|
"loss": 0.1833,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09590593725442886,
|
|
"step": 5670,
|
|
"valid_targets_mean": 6638.1,
|
|
"valid_targets_min": 4837
|
|
},
|
|
{
|
|
"epoch": 3.1284454244762956,
|
|
"grad_norm": 0.3606308846247167,
|
|
"learning_rate": 2.7047788117606794e-05,
|
|
"loss": 0.1836,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0834389179944992,
|
|
"step": 5675,
|
|
"valid_targets_mean": 6058.5,
|
|
"valid_targets_min": 4914
|
|
},
|
|
{
|
|
"epoch": 3.131201764057332,
|
|
"grad_norm": 0.39954895354752207,
|
|
"learning_rate": 2.7022054565973498e-05,
|
|
"loss": 0.1934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09039755165576935,
|
|
"step": 5680,
|
|
"valid_targets_mean": 6196.4,
|
|
"valid_targets_min": 4824
|
|
},
|
|
{
|
|
"epoch": 3.1339581036383684,
|
|
"grad_norm": 0.38931937267214356,
|
|
"learning_rate": 2.6996307747624514e-05,
|
|
"loss": 0.1921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09171608090400696,
|
|
"step": 5685,
|
|
"valid_targets_mean": 5347.9,
|
|
"valid_targets_min": 4122
|
|
},
|
|
{
|
|
"epoch": 3.1367144432194047,
|
|
"grad_norm": 0.4678828890825313,
|
|
"learning_rate": 2.6970547711203114e-05,
|
|
"loss": 0.1578,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09446283429861069,
|
|
"step": 5690,
|
|
"valid_targets_mean": 6277.8,
|
|
"valid_targets_min": 5273
|
|
},
|
|
{
|
|
"epoch": 3.139470782800441,
|
|
"grad_norm": 0.4312188498482889,
|
|
"learning_rate": 2.6944774505377545e-05,
|
|
"loss": 0.1847,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08975284546613693,
|
|
"step": 5695,
|
|
"valid_targets_mean": 6190.9,
|
|
"valid_targets_min": 4605
|
|
},
|
|
{
|
|
"epoch": 3.1422271223814775,
|
|
"grad_norm": 0.3985914616508528,
|
|
"learning_rate": 2.6918988178840932e-05,
|
|
"loss": 0.1877,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10301844775676727,
|
|
"step": 5700,
|
|
"valid_targets_mean": 6797.0,
|
|
"valid_targets_min": 5636
|
|
},
|
|
{
|
|
"epoch": 3.144983461962514,
|
|
"grad_norm": 0.37258322063112864,
|
|
"learning_rate": 2.6893188780311182e-05,
|
|
"loss": 0.188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0984998494386673,
|
|
"step": 5705,
|
|
"valid_targets_mean": 6405.5,
|
|
"valid_targets_min": 4972
|
|
},
|
|
{
|
|
"epoch": 3.1477398015435503,
|
|
"grad_norm": 0.41298579001824337,
|
|
"learning_rate": 2.686737635853093e-05,
|
|
"loss": 0.1794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08721563965082169,
|
|
"step": 5710,
|
|
"valid_targets_mean": 6174.4,
|
|
"valid_targets_min": 5309
|
|
},
|
|
{
|
|
"epoch": 3.1504961411245866,
|
|
"grad_norm": 0.38705079443852347,
|
|
"learning_rate": 2.6841550962267376e-05,
|
|
"loss": 0.1844,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09890256077051163,
|
|
"step": 5715,
|
|
"valid_targets_mean": 6717.0,
|
|
"valid_targets_min": 5331
|
|
},
|
|
{
|
|
"epoch": 3.153252480705623,
|
|
"grad_norm": 0.4040792732493771,
|
|
"learning_rate": 2.681571264031225e-05,
|
|
"loss": 0.1831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08728047460317612,
|
|
"step": 5720,
|
|
"valid_targets_mean": 5186.0,
|
|
"valid_targets_min": 790
|
|
},
|
|
{
|
|
"epoch": 3.1560088202866594,
|
|
"grad_norm": 0.7037629344738677,
|
|
"learning_rate": 2.678986144148171e-05,
|
|
"loss": 0.1996,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07662524282932281,
|
|
"step": 5725,
|
|
"valid_targets_mean": 1275.9,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 3.158765159867696,
|
|
"grad_norm": 0.7530060270829984,
|
|
"learning_rate": 2.6763997414616226e-05,
|
|
"loss": 0.178,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09695053845643997,
|
|
"step": 5730,
|
|
"valid_targets_mean": 1629.0,
|
|
"valid_targets_min": 660
|
|
},
|
|
{
|
|
"epoch": 3.161521499448732,
|
|
"grad_norm": 0.7433723619258658,
|
|
"learning_rate": 2.673812060858052e-05,
|
|
"loss": 0.1723,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08200408518314362,
|
|
"step": 5735,
|
|
"valid_targets_mean": 1415.2,
|
|
"valid_targets_min": 825
|
|
},
|
|
{
|
|
"epoch": 3.1642778390297686,
|
|
"grad_norm": 0.7186736300323973,
|
|
"learning_rate": 2.6712231072263453e-05,
|
|
"loss": 0.1692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07920856773853302,
|
|
"step": 5740,
|
|
"valid_targets_mean": 1395.1,
|
|
"valid_targets_min": 741
|
|
},
|
|
{
|
|
"epoch": 3.167034178610805,
|
|
"grad_norm": 0.7234238631988296,
|
|
"learning_rate": 2.6686328854577923e-05,
|
|
"loss": 0.1821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06963801383972168,
|
|
"step": 5745,
|
|
"valid_targets_mean": 1304.8,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 3.1697905181918413,
|
|
"grad_norm": 0.6804717159784225,
|
|
"learning_rate": 2.6660414004460806e-05,
|
|
"loss": 0.1755,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10395242273807526,
|
|
"step": 5750,
|
|
"valid_targets_mean": 1723.5,
|
|
"valid_targets_min": 763
|
|
},
|
|
{
|
|
"epoch": 3.1725468577728777,
|
|
"grad_norm": 0.7135483794202085,
|
|
"learning_rate": 2.663448657087283e-05,
|
|
"loss": 0.1623,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08328169584274292,
|
|
"step": 5755,
|
|
"valid_targets_mean": 1645.1,
|
|
"valid_targets_min": 813
|
|
},
|
|
{
|
|
"epoch": 3.175303197353914,
|
|
"grad_norm": 0.7598872151755668,
|
|
"learning_rate": 2.6608546602798512e-05,
|
|
"loss": 0.1801,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06543715298175812,
|
|
"step": 5760,
|
|
"valid_targets_mean": 1187.5,
|
|
"valid_targets_min": 607
|
|
},
|
|
{
|
|
"epoch": 3.1780595369349505,
|
|
"grad_norm": 0.7198710115205064,
|
|
"learning_rate": 2.6582594149246035e-05,
|
|
"loss": 0.166,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09502364695072174,
|
|
"step": 5765,
|
|
"valid_targets_mean": 1872.4,
|
|
"valid_targets_min": 816
|
|
},
|
|
{
|
|
"epoch": 3.180815876515987,
|
|
"grad_norm": 0.6630535937810428,
|
|
"learning_rate": 2.6556629259247175e-05,
|
|
"loss": 0.1698,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.079322949051857,
|
|
"step": 5770,
|
|
"valid_targets_mean": 1494.2,
|
|
"valid_targets_min": 836
|
|
},
|
|
{
|
|
"epoch": 3.1835722160970232,
|
|
"grad_norm": 0.7755273664856227,
|
|
"learning_rate": 2.6530651981857217e-05,
|
|
"loss": 0.1701,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1135888546705246,
|
|
"step": 5775,
|
|
"valid_targets_mean": 1655.4,
|
|
"valid_targets_min": 590
|
|
},
|
|
{
|
|
"epoch": 3.1863285556780596,
|
|
"grad_norm": 0.678070612371881,
|
|
"learning_rate": 2.6504662366154826e-05,
|
|
"loss": 0.1696,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08234953880310059,
|
|
"step": 5780,
|
|
"valid_targets_mean": 1532.5,
|
|
"valid_targets_min": 702
|
|
},
|
|
{
|
|
"epoch": 3.189084895259096,
|
|
"grad_norm": 0.6932169696813631,
|
|
"learning_rate": 2.6478660461242e-05,
|
|
"loss": 0.1606,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.062241796404123306,
|
|
"step": 5785,
|
|
"valid_targets_mean": 1223.5,
|
|
"valid_targets_min": 721
|
|
},
|
|
{
|
|
"epoch": 3.1918412348401324,
|
|
"grad_norm": 0.7058424055536338,
|
|
"learning_rate": 2.6452646316243943e-05,
|
|
"loss": 0.1595,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07513256371021271,
|
|
"step": 5790,
|
|
"valid_targets_mean": 1361.0,
|
|
"valid_targets_min": 640
|
|
},
|
|
{
|
|
"epoch": 3.1945975744211688,
|
|
"grad_norm": 0.7877371960345096,
|
|
"learning_rate": 2.6426619980308988e-05,
|
|
"loss": 0.1601,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08057187497615814,
|
|
"step": 5795,
|
|
"valid_targets_mean": 1353.9,
|
|
"valid_targets_min": 516
|
|
},
|
|
{
|
|
"epoch": 3.197353914002205,
|
|
"grad_norm": 0.7566200644661817,
|
|
"learning_rate": 2.640058150260849e-05,
|
|
"loss": 0.1734,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09616967290639877,
|
|
"step": 5800,
|
|
"valid_targets_mean": 1585.9,
|
|
"valid_targets_min": 727
|
|
},
|
|
{
|
|
"epoch": 3.2001102535832415,
|
|
"grad_norm": 0.7017170384841674,
|
|
"learning_rate": 2.6374530932336763e-05,
|
|
"loss": 0.1715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10249229520559311,
|
|
"step": 5805,
|
|
"valid_targets_mean": 1783.9,
|
|
"valid_targets_min": 819
|
|
},
|
|
{
|
|
"epoch": 3.202866593164278,
|
|
"grad_norm": 0.8111907703351192,
|
|
"learning_rate": 2.6348468318710952e-05,
|
|
"loss": 0.1757,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09271591156721115,
|
|
"step": 5810,
|
|
"valid_targets_mean": 1466.4,
|
|
"valid_targets_min": 815
|
|
},
|
|
{
|
|
"epoch": 3.2056229327453143,
|
|
"grad_norm": 0.7889700414305706,
|
|
"learning_rate": 2.6322393710970963e-05,
|
|
"loss": 0.1625,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07034605741500854,
|
|
"step": 5815,
|
|
"valid_targets_mean": 1274.6,
|
|
"valid_targets_min": 623
|
|
},
|
|
{
|
|
"epoch": 3.2083792723263507,
|
|
"grad_norm": 0.6754893187703334,
|
|
"learning_rate": 2.6296307158379356e-05,
|
|
"loss": 0.1679,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10925255715847015,
|
|
"step": 5820,
|
|
"valid_targets_mean": 2305.0,
|
|
"valid_targets_min": 789
|
|
},
|
|
{
|
|
"epoch": 3.211135611907387,
|
|
"grad_norm": 0.7402110033602038,
|
|
"learning_rate": 2.6270208710221253e-05,
|
|
"loss": 0.1739,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11757739633321762,
|
|
"step": 5825,
|
|
"valid_targets_mean": 2222.4,
|
|
"valid_targets_min": 1330
|
|
},
|
|
{
|
|
"epoch": 3.2138919514884234,
|
|
"grad_norm": 0.6566285200279769,
|
|
"learning_rate": 2.6244098415804272e-05,
|
|
"loss": 0.167,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09257020056247711,
|
|
"step": 5830,
|
|
"valid_targets_mean": 1723.1,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 3.21664829106946,
|
|
"grad_norm": 0.7327820229386447,
|
|
"learning_rate": 2.62179763244584e-05,
|
|
"loss": 0.16,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06514466553926468,
|
|
"step": 5835,
|
|
"valid_targets_mean": 1219.0,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 3.219404630650496,
|
|
"grad_norm": 0.7648472924378685,
|
|
"learning_rate": 2.61918424855359e-05,
|
|
"loss": 0.1618,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09546822309494019,
|
|
"step": 5840,
|
|
"valid_targets_mean": 1515.6,
|
|
"valid_targets_min": 996
|
|
},
|
|
{
|
|
"epoch": 3.2221609702315326,
|
|
"grad_norm": 0.7188859719820919,
|
|
"learning_rate": 2.616569694841125e-05,
|
|
"loss": 0.162,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08607321977615356,
|
|
"step": 5845,
|
|
"valid_targets_mean": 1686.1,
|
|
"valid_targets_min": 626
|
|
},
|
|
{
|
|
"epoch": 3.224917309812569,
|
|
"grad_norm": 0.783043055471916,
|
|
"learning_rate": 2.6139539762481007e-05,
|
|
"loss": 0.1651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08706812560558319,
|
|
"step": 5850,
|
|
"valid_targets_mean": 1624.0,
|
|
"valid_targets_min": 1072
|
|
},
|
|
{
|
|
"epoch": 3.2276736493936053,
|
|
"grad_norm": 0.6726418973266549,
|
|
"learning_rate": 2.6113370977163762e-05,
|
|
"loss": 0.1653,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08502975851297379,
|
|
"step": 5855,
|
|
"valid_targets_mean": 1836.2,
|
|
"valid_targets_min": 1071
|
|
},
|
|
{
|
|
"epoch": 3.2304299889746417,
|
|
"grad_norm": 0.7777891780765533,
|
|
"learning_rate": 2.6087190641900005e-05,
|
|
"loss": 0.168,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0926479697227478,
|
|
"step": 5860,
|
|
"valid_targets_mean": 1841.2,
|
|
"valid_targets_min": 1313
|
|
},
|
|
{
|
|
"epoch": 3.233186328555678,
|
|
"grad_norm": 0.7164378388492749,
|
|
"learning_rate": 2.6060998806152045e-05,
|
|
"loss": 0.1612,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09424290060997009,
|
|
"step": 5865,
|
|
"valid_targets_mean": 1720.5,
|
|
"valid_targets_min": 956
|
|
},
|
|
{
|
|
"epoch": 3.2359426681367145,
|
|
"grad_norm": 0.7951640369495439,
|
|
"learning_rate": 2.6034795519403927e-05,
|
|
"loss": 0.1649,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06589596718549728,
|
|
"step": 5870,
|
|
"valid_targets_mean": 1111.9,
|
|
"valid_targets_min": 642
|
|
},
|
|
{
|
|
"epoch": 3.238699007717751,
|
|
"grad_norm": 1.5605364114912275,
|
|
"learning_rate": 2.600858083116132e-05,
|
|
"loss": 0.1701,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08355779200792313,
|
|
"step": 5875,
|
|
"valid_targets_mean": 1541.8,
|
|
"valid_targets_min": 1110
|
|
},
|
|
{
|
|
"epoch": 3.2414553472987873,
|
|
"grad_norm": 0.6874183525225576,
|
|
"learning_rate": 2.598235479095145e-05,
|
|
"loss": 0.1569,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07314138114452362,
|
|
"step": 5880,
|
|
"valid_targets_mean": 1445.0,
|
|
"valid_targets_min": 704
|
|
},
|
|
{
|
|
"epoch": 3.2442116868798236,
|
|
"grad_norm": 0.7101517696962755,
|
|
"learning_rate": 2.5956117448322972e-05,
|
|
"loss": 0.1629,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08403250575065613,
|
|
"step": 5885,
|
|
"valid_targets_mean": 1532.5,
|
|
"valid_targets_min": 1284
|
|
},
|
|
{
|
|
"epoch": 3.24696802646086,
|
|
"grad_norm": 0.6980265827872365,
|
|
"learning_rate": 2.592986885284591e-05,
|
|
"loss": 0.1507,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07715944945812225,
|
|
"step": 5890,
|
|
"valid_targets_mean": 1316.0,
|
|
"valid_targets_min": 752
|
|
},
|
|
{
|
|
"epoch": 3.2497243660418964,
|
|
"grad_norm": 0.7845131237865528,
|
|
"learning_rate": 2.590360905411154e-05,
|
|
"loss": 0.1619,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07715801149606705,
|
|
"step": 5895,
|
|
"valid_targets_mean": 1474.1,
|
|
"valid_targets_min": 816
|
|
},
|
|
{
|
|
"epoch": 3.252480705622933,
|
|
"grad_norm": 0.6777224756659469,
|
|
"learning_rate": 2.5877338101732293e-05,
|
|
"loss": 0.16,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06594108045101166,
|
|
"step": 5900,
|
|
"valid_targets_mean": 1290.6,
|
|
"valid_targets_min": 972
|
|
},
|
|
{
|
|
"epoch": 3.255237045203969,
|
|
"grad_norm": 0.7257120848052234,
|
|
"learning_rate": 2.5851056045341704e-05,
|
|
"loss": 0.1625,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08703707158565521,
|
|
"step": 5905,
|
|
"valid_targets_mean": 1523.5,
|
|
"valid_targets_min": 655
|
|
},
|
|
{
|
|
"epoch": 3.2579933847850056,
|
|
"grad_norm": 0.8704048984795587,
|
|
"learning_rate": 2.5824762934594256e-05,
|
|
"loss": 0.1617,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09733042120933533,
|
|
"step": 5910,
|
|
"valid_targets_mean": 1672.4,
|
|
"valid_targets_min": 879
|
|
},
|
|
{
|
|
"epoch": 3.260749724366042,
|
|
"grad_norm": 0.7514053920946749,
|
|
"learning_rate": 2.5798458819165335e-05,
|
|
"loss": 0.1642,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0953604057431221,
|
|
"step": 5915,
|
|
"valid_targets_mean": 1604.0,
|
|
"valid_targets_min": 1049
|
|
},
|
|
{
|
|
"epoch": 3.2635060639470783,
|
|
"grad_norm": 0.7284375627541742,
|
|
"learning_rate": 2.57721437487511e-05,
|
|
"loss": 0.16,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07512304186820984,
|
|
"step": 5920,
|
|
"valid_targets_mean": 1444.1,
|
|
"valid_targets_min": 679
|
|
},
|
|
{
|
|
"epoch": 3.2662624035281147,
|
|
"grad_norm": 0.6612520687208383,
|
|
"learning_rate": 2.5745817773068428e-05,
|
|
"loss": 0.1594,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07188956439495087,
|
|
"step": 5925,
|
|
"valid_targets_mean": 1340.0,
|
|
"valid_targets_min": 702
|
|
},
|
|
{
|
|
"epoch": 3.269018743109151,
|
|
"grad_norm": 0.7452631269131852,
|
|
"learning_rate": 2.571948094185479e-05,
|
|
"loss": 0.1691,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0772596001625061,
|
|
"step": 5930,
|
|
"valid_targets_mean": 1464.9,
|
|
"valid_targets_min": 933
|
|
},
|
|
{
|
|
"epoch": 3.2717750826901875,
|
|
"grad_norm": 0.8020399812776332,
|
|
"learning_rate": 2.5693133304868164e-05,
|
|
"loss": 0.1733,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11898557841777802,
|
|
"step": 5935,
|
|
"valid_targets_mean": 1824.2,
|
|
"valid_targets_min": 733
|
|
},
|
|
{
|
|
"epoch": 3.274531422271224,
|
|
"grad_norm": 0.6902949594577399,
|
|
"learning_rate": 2.5666774911886948e-05,
|
|
"loss": 0.1585,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10325020551681519,
|
|
"step": 5940,
|
|
"valid_targets_mean": 2003.6,
|
|
"valid_targets_min": 1433
|
|
},
|
|
{
|
|
"epoch": 3.2772877618522602,
|
|
"grad_norm": 0.703576516382865,
|
|
"learning_rate": 2.5640405812709847e-05,
|
|
"loss": 0.1583,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06290339678525925,
|
|
"step": 5945,
|
|
"valid_targets_mean": 1400.9,
|
|
"valid_targets_min": 549
|
|
},
|
|
{
|
|
"epoch": 3.2800441014332966,
|
|
"grad_norm": 0.7888292463783313,
|
|
"learning_rate": 2.5614026057155822e-05,
|
|
"loss": 0.165,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10196740925312042,
|
|
"step": 5950,
|
|
"valid_targets_mean": 1812.0,
|
|
"valid_targets_min": 930
|
|
},
|
|
{
|
|
"epoch": 3.282800441014333,
|
|
"grad_norm": 0.7084562182668427,
|
|
"learning_rate": 2.558763569506394e-05,
|
|
"loss": 0.159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07969842851161957,
|
|
"step": 5955,
|
|
"valid_targets_mean": 1594.1,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 3.2855567805953694,
|
|
"grad_norm": 0.6752390503934562,
|
|
"learning_rate": 2.556123477629332e-05,
|
|
"loss": 0.1536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08806701749563217,
|
|
"step": 5960,
|
|
"valid_targets_mean": 1767.1,
|
|
"valid_targets_min": 1152
|
|
},
|
|
{
|
|
"epoch": 3.2883131201764058,
|
|
"grad_norm": 0.7226042699560578,
|
|
"learning_rate": 2.5534823350723022e-05,
|
|
"loss": 0.1517,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07272815704345703,
|
|
"step": 5965,
|
|
"valid_targets_mean": 1226.6,
|
|
"valid_targets_min": 755
|
|
},
|
|
{
|
|
"epoch": 3.291069459757442,
|
|
"grad_norm": 0.779308170450985,
|
|
"learning_rate": 2.550840146825195e-05,
|
|
"loss": 0.1668,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09173888713121414,
|
|
"step": 5970,
|
|
"valid_targets_mean": 1532.1,
|
|
"valid_targets_min": 853
|
|
},
|
|
{
|
|
"epoch": 3.2938257993384785,
|
|
"grad_norm": 0.6888456112298955,
|
|
"learning_rate": 2.548196917879878e-05,
|
|
"loss": 0.1604,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08668754994869232,
|
|
"step": 5975,
|
|
"valid_targets_mean": 1733.0,
|
|
"valid_targets_min": 1455
|
|
},
|
|
{
|
|
"epoch": 3.296582138919515,
|
|
"grad_norm": 0.6705726392476253,
|
|
"learning_rate": 2.5455526532301838e-05,
|
|
"loss": 0.1567,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07058720290660858,
|
|
"step": 5980,
|
|
"valid_targets_mean": 1470.8,
|
|
"valid_targets_min": 940
|
|
},
|
|
{
|
|
"epoch": 3.2993384785005513,
|
|
"grad_norm": 0.685909685753295,
|
|
"learning_rate": 2.5429073578719018e-05,
|
|
"loss": 0.1573,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07246449589729309,
|
|
"step": 5985,
|
|
"valid_targets_mean": 1477.2,
|
|
"valid_targets_min": 1224
|
|
},
|
|
{
|
|
"epoch": 3.3020948180815877,
|
|
"grad_norm": 0.6757125766577314,
|
|
"learning_rate": 2.540261036802769e-05,
|
|
"loss": 0.1605,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06590719521045685,
|
|
"step": 5990,
|
|
"valid_targets_mean": 1271.5,
|
|
"valid_targets_min": 738
|
|
},
|
|
{
|
|
"epoch": 3.304851157662624,
|
|
"grad_norm": 0.665315383514823,
|
|
"learning_rate": 2.53761369502246e-05,
|
|
"loss": 0.1607,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09540765732526779,
|
|
"step": 5995,
|
|
"valid_targets_mean": 2017.0,
|
|
"valid_targets_min": 877
|
|
},
|
|
{
|
|
"epoch": 3.3076074972436604,
|
|
"grad_norm": 0.7628088968155343,
|
|
"learning_rate": 2.5349653375325788e-05,
|
|
"loss": 0.1587,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08156589418649673,
|
|
"step": 6000,
|
|
"valid_targets_mean": 1284.6,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 3.310363836824697,
|
|
"grad_norm": 0.7204950856895939,
|
|
"learning_rate": 2.5323159693366472e-05,
|
|
"loss": 0.156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05616691708564758,
|
|
"step": 6005,
|
|
"valid_targets_mean": 1100.9,
|
|
"valid_targets_min": 813
|
|
},
|
|
{
|
|
"epoch": 3.313120176405733,
|
|
"grad_norm": 0.7592815109298671,
|
|
"learning_rate": 2.529665595440096e-05,
|
|
"loss": 0.1506,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07726922631263733,
|
|
"step": 6010,
|
|
"valid_targets_mean": 1388.8,
|
|
"valid_targets_min": 816
|
|
},
|
|
{
|
|
"epoch": 3.3158765159867696,
|
|
"grad_norm": 0.7601899698687988,
|
|
"learning_rate": 2.5270142208502586e-05,
|
|
"loss": 0.1586,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07069392502307892,
|
|
"step": 6015,
|
|
"valid_targets_mean": 1387.8,
|
|
"valid_targets_min": 660
|
|
},
|
|
{
|
|
"epoch": 3.318632855567806,
|
|
"grad_norm": 0.7719461183220124,
|
|
"learning_rate": 2.5243618505763562e-05,
|
|
"loss": 0.1599,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07524304836988449,
|
|
"step": 6020,
|
|
"valid_targets_mean": 1437.5,
|
|
"valid_targets_min": 727
|
|
},
|
|
{
|
|
"epoch": 3.3213891951488423,
|
|
"grad_norm": 0.7262686864401563,
|
|
"learning_rate": 2.5217084896294926e-05,
|
|
"loss": 0.1546,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07308384776115417,
|
|
"step": 6025,
|
|
"valid_targets_mean": 1407.8,
|
|
"valid_targets_min": 691
|
|
},
|
|
{
|
|
"epoch": 3.3241455347298787,
|
|
"grad_norm": 0.7448257518304007,
|
|
"learning_rate": 2.5190541430226432e-05,
|
|
"loss": 0.1595,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08132465928792953,
|
|
"step": 6030,
|
|
"valid_targets_mean": 1319.6,
|
|
"valid_targets_min": 839
|
|
},
|
|
{
|
|
"epoch": 3.326901874310915,
|
|
"grad_norm": 0.7645225512561269,
|
|
"learning_rate": 2.516398815770645e-05,
|
|
"loss": 0.16,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08001318573951721,
|
|
"step": 6035,
|
|
"valid_targets_mean": 1483.9,
|
|
"valid_targets_min": 949
|
|
},
|
|
{
|
|
"epoch": 3.3296582138919515,
|
|
"grad_norm": 0.6927221461779683,
|
|
"learning_rate": 2.513742512890189e-05,
|
|
"loss": 0.1556,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08853611350059509,
|
|
"step": 6040,
|
|
"valid_targets_mean": 1929.4,
|
|
"valid_targets_min": 877
|
|
},
|
|
{
|
|
"epoch": 3.332414553472988,
|
|
"grad_norm": 0.7845756917752043,
|
|
"learning_rate": 2.511085239399807e-05,
|
|
"loss": 0.1602,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08278684318065643,
|
|
"step": 6045,
|
|
"valid_targets_mean": 1496.9,
|
|
"valid_targets_min": 835
|
|
},
|
|
{
|
|
"epoch": 3.3351708930540243,
|
|
"grad_norm": 0.7621394481704554,
|
|
"learning_rate": 2.5084270003198672e-05,
|
|
"loss": 0.159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08645252883434296,
|
|
"step": 6050,
|
|
"valid_targets_mean": 1529.4,
|
|
"valid_targets_min": 1009
|
|
},
|
|
{
|
|
"epoch": 3.3379272326350606,
|
|
"grad_norm": 0.7620803483861601,
|
|
"learning_rate": 2.5057678006725613e-05,
|
|
"loss": 0.162,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07133709639310837,
|
|
"step": 6055,
|
|
"valid_targets_mean": 1341.5,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 3.340683572216097,
|
|
"grad_norm": 0.8046936095021902,
|
|
"learning_rate": 2.503107645481895e-05,
|
|
"loss": 0.1585,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08323918282985687,
|
|
"step": 6060,
|
|
"valid_targets_mean": 1537.8,
|
|
"valid_targets_min": 726
|
|
},
|
|
{
|
|
"epoch": 3.3434399117971334,
|
|
"grad_norm": 0.6659078190884129,
|
|
"learning_rate": 2.5004465397736794e-05,
|
|
"loss": 0.1689,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07124647498130798,
|
|
"step": 6065,
|
|
"valid_targets_mean": 1535.6,
|
|
"valid_targets_min": 1215
|
|
},
|
|
{
|
|
"epoch": 3.34619625137817,
|
|
"grad_norm": 0.6572783349938557,
|
|
"learning_rate": 2.4977844885755228e-05,
|
|
"loss": 0.1587,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07897505164146423,
|
|
"step": 6070,
|
|
"valid_targets_mean": 1626.0,
|
|
"valid_targets_min": 912
|
|
},
|
|
{
|
|
"epoch": 3.348952590959206,
|
|
"grad_norm": 0.6974227469322377,
|
|
"learning_rate": 2.4951214969168177e-05,
|
|
"loss": 0.1596,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07202719151973724,
|
|
"step": 6075,
|
|
"valid_targets_mean": 1346.5,
|
|
"valid_targets_min": 645
|
|
},
|
|
{
|
|
"epoch": 3.3517089305402425,
|
|
"grad_norm": 0.754359893227626,
|
|
"learning_rate": 2.492457569828736e-05,
|
|
"loss": 0.156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08316866308450699,
|
|
"step": 6080,
|
|
"valid_targets_mean": 1379.2,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 3.354465270121279,
|
|
"grad_norm": 0.6997188512291971,
|
|
"learning_rate": 2.4897927123442135e-05,
|
|
"loss": 0.1528,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0747290700674057,
|
|
"step": 6085,
|
|
"valid_targets_mean": 1572.8,
|
|
"valid_targets_min": 844
|
|
},
|
|
{
|
|
"epoch": 3.3572216097023153,
|
|
"grad_norm": 0.7833276553779296,
|
|
"learning_rate": 2.4871269294979473e-05,
|
|
"loss": 0.1634,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0762481689453125,
|
|
"step": 6090,
|
|
"valid_targets_mean": 1180.0,
|
|
"valid_targets_min": 690
|
|
},
|
|
{
|
|
"epoch": 3.3599779492833517,
|
|
"grad_norm": 0.7374274485578655,
|
|
"learning_rate": 2.4844602263263805e-05,
|
|
"loss": 0.1509,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07163633406162262,
|
|
"step": 6095,
|
|
"valid_targets_mean": 1370.9,
|
|
"valid_targets_min": 820
|
|
},
|
|
{
|
|
"epoch": 3.362734288864388,
|
|
"grad_norm": 0.7070767348408084,
|
|
"learning_rate": 2.481792607867696e-05,
|
|
"loss": 0.1521,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0685414969921112,
|
|
"step": 6100,
|
|
"valid_targets_mean": 1423.6,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 3.3654906284454245,
|
|
"grad_norm": 0.7083711465813802,
|
|
"learning_rate": 2.479124079161805e-05,
|
|
"loss": 0.1525,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0720287337899208,
|
|
"step": 6105,
|
|
"valid_targets_mean": 1416.4,
|
|
"valid_targets_min": 780
|
|
},
|
|
{
|
|
"epoch": 3.368246968026461,
|
|
"grad_norm": 0.7846320707578657,
|
|
"learning_rate": 2.4764546452503394e-05,
|
|
"loss": 0.1679,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08491784334182739,
|
|
"step": 6110,
|
|
"valid_targets_mean": 1251.2,
|
|
"valid_targets_min": 546
|
|
},
|
|
{
|
|
"epoch": 3.3710033076074972,
|
|
"grad_norm": 0.6976360561727827,
|
|
"learning_rate": 2.4737843111766407e-05,
|
|
"loss": 0.1522,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08084502071142197,
|
|
"step": 6115,
|
|
"valid_targets_mean": 1406.4,
|
|
"valid_targets_min": 693
|
|
},
|
|
{
|
|
"epoch": 3.3737596471885336,
|
|
"grad_norm": 0.7192730520151608,
|
|
"learning_rate": 2.471113081985752e-05,
|
|
"loss": 0.1597,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09040600061416626,
|
|
"step": 6120,
|
|
"valid_targets_mean": 1968.0,
|
|
"valid_targets_min": 1029
|
|
},
|
|
{
|
|
"epoch": 3.37651598676957,
|
|
"grad_norm": 0.6366363849327463,
|
|
"learning_rate": 2.4684409627244064e-05,
|
|
"loss": 0.1449,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06264091283082962,
|
|
"step": 6125,
|
|
"valid_targets_mean": 1574.9,
|
|
"valid_targets_min": 679
|
|
},
|
|
{
|
|
"epoch": 3.3792723263506064,
|
|
"grad_norm": 0.7952282339421747,
|
|
"learning_rate": 2.4657679584410187e-05,
|
|
"loss": 0.1489,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07549738883972168,
|
|
"step": 6130,
|
|
"valid_targets_mean": 1420.1,
|
|
"valid_targets_min": 986
|
|
},
|
|
{
|
|
"epoch": 3.3820286659316428,
|
|
"grad_norm": 0.7308122730264502,
|
|
"learning_rate": 2.4630940741856762e-05,
|
|
"loss": 0.156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0581502839922905,
|
|
"step": 6135,
|
|
"valid_targets_mean": 1295.1,
|
|
"valid_targets_min": 638
|
|
},
|
|
{
|
|
"epoch": 3.384785005512679,
|
|
"grad_norm": 0.7133550239520771,
|
|
"learning_rate": 2.4604193150101287e-05,
|
|
"loss": 0.1554,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06828457862138748,
|
|
"step": 6140,
|
|
"valid_targets_mean": 1453.1,
|
|
"valid_targets_min": 720
|
|
},
|
|
{
|
|
"epoch": 3.3875413450937155,
|
|
"grad_norm": 0.7443190111146352,
|
|
"learning_rate": 2.4577436859677795e-05,
|
|
"loss": 0.1552,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0740957260131836,
|
|
"step": 6145,
|
|
"valid_targets_mean": 1325.5,
|
|
"valid_targets_min": 873
|
|
},
|
|
{
|
|
"epoch": 3.390297684674752,
|
|
"grad_norm": 0.7585141385815627,
|
|
"learning_rate": 2.455067192113674e-05,
|
|
"loss": 0.1566,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07758934795856476,
|
|
"step": 6150,
|
|
"valid_targets_mean": 1553.0,
|
|
"valid_targets_min": 685
|
|
},
|
|
{
|
|
"epoch": 3.3930540242557883,
|
|
"grad_norm": 0.7383372588524464,
|
|
"learning_rate": 2.4523898385044924e-05,
|
|
"loss": 0.1593,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08464233577251434,
|
|
"step": 6155,
|
|
"valid_targets_mean": 1387.9,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 3.3958103638368247,
|
|
"grad_norm": 0.8225932430652274,
|
|
"learning_rate": 2.449711630198539e-05,
|
|
"loss": 0.1701,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14323914051055908,
|
|
"step": 6160,
|
|
"valid_targets_mean": 2335.2,
|
|
"valid_targets_min": 915
|
|
},
|
|
{
|
|
"epoch": 3.398566703417861,
|
|
"grad_norm": 0.5815496349556788,
|
|
"learning_rate": 2.447032572255733e-05,
|
|
"loss": 0.2728,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12792423367500305,
|
|
"step": 6165,
|
|
"valid_targets_mean": 5145.2,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 3.4013230429988974,
|
|
"grad_norm": 0.4527711025741082,
|
|
"learning_rate": 2.4443526697375987e-05,
|
|
"loss": 0.2514,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12657275795936584,
|
|
"step": 6170,
|
|
"valid_targets_mean": 5444.0,
|
|
"valid_targets_min": 1842
|
|
},
|
|
{
|
|
"epoch": 3.404079382579934,
|
|
"grad_norm": 0.49166131951796893,
|
|
"learning_rate": 2.441671927707256e-05,
|
|
"loss": 0.2424,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0933639258146286,
|
|
"step": 6175,
|
|
"valid_targets_mean": 4881.1,
|
|
"valid_targets_min": 1319
|
|
},
|
|
{
|
|
"epoch": 3.40683572216097,
|
|
"grad_norm": 0.4430129654366527,
|
|
"learning_rate": 2.4389903512294115e-05,
|
|
"loss": 0.2438,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12261107563972473,
|
|
"step": 6180,
|
|
"valid_targets_mean": 5722.4,
|
|
"valid_targets_min": 2365
|
|
},
|
|
{
|
|
"epoch": 3.4095920617420066,
|
|
"grad_norm": 0.5282950814834866,
|
|
"learning_rate": 2.4363079453703467e-05,
|
|
"loss": 0.2536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11502567678689957,
|
|
"step": 6185,
|
|
"valid_targets_mean": 4214.8,
|
|
"valid_targets_min": 2074
|
|
},
|
|
{
|
|
"epoch": 3.412348401323043,
|
|
"grad_norm": 0.5287170876839649,
|
|
"learning_rate": 2.4336247151979128e-05,
|
|
"loss": 0.2477,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12764747440814972,
|
|
"step": 6190,
|
|
"valid_targets_mean": 5447.8,
|
|
"valid_targets_min": 1991
|
|
},
|
|
{
|
|
"epoch": 3.4151047409040793,
|
|
"grad_norm": 0.5449218280056792,
|
|
"learning_rate": 2.4309406657815146e-05,
|
|
"loss": 0.2415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10804009437561035,
|
|
"step": 6195,
|
|
"valid_targets_mean": 3284.6,
|
|
"valid_targets_min": 1747
|
|
},
|
|
{
|
|
"epoch": 3.4178610804851157,
|
|
"grad_norm": 0.4944975415646534,
|
|
"learning_rate": 2.4282558021921086e-05,
|
|
"loss": 0.2447,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10888747870922089,
|
|
"step": 6200,
|
|
"valid_targets_mean": 4710.8,
|
|
"valid_targets_min": 1165
|
|
},
|
|
{
|
|
"epoch": 3.420617420066152,
|
|
"grad_norm": 0.5042434182701653,
|
|
"learning_rate": 2.425570129502187e-05,
|
|
"loss": 0.2543,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10952840745449066,
|
|
"step": 6205,
|
|
"valid_targets_mean": 4457.4,
|
|
"valid_targets_min": 1239
|
|
},
|
|
{
|
|
"epoch": 3.4233737596471885,
|
|
"grad_norm": 0.49424506700189763,
|
|
"learning_rate": 2.4228836527857705e-05,
|
|
"loss": 0.248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13219739496707916,
|
|
"step": 6210,
|
|
"valid_targets_mean": 4907.9,
|
|
"valid_targets_min": 2744
|
|
},
|
|
{
|
|
"epoch": 3.426130099228225,
|
|
"grad_norm": 0.44456662731307445,
|
|
"learning_rate": 2.4201963771184008e-05,
|
|
"loss": 0.2424,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11132525652647018,
|
|
"step": 6215,
|
|
"valid_targets_mean": 5654.8,
|
|
"valid_targets_min": 2138
|
|
},
|
|
{
|
|
"epoch": 3.4288864388092613,
|
|
"grad_norm": 0.517378652332287,
|
|
"learning_rate": 2.4175083075771283e-05,
|
|
"loss": 0.2432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11823809146881104,
|
|
"step": 6220,
|
|
"valid_targets_mean": 5088.0,
|
|
"valid_targets_min": 2584
|
|
},
|
|
{
|
|
"epoch": 3.4316427783902976,
|
|
"grad_norm": 0.48526798859423875,
|
|
"learning_rate": 2.4148194492405013e-05,
|
|
"loss": 0.2491,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12625981867313385,
|
|
"step": 6225,
|
|
"valid_targets_mean": 5116.5,
|
|
"valid_targets_min": 1365
|
|
},
|
|
{
|
|
"epoch": 3.434399117971334,
|
|
"grad_norm": 0.5447290615866072,
|
|
"learning_rate": 2.4121298071885613e-05,
|
|
"loss": 0.2424,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11234936118125916,
|
|
"step": 6230,
|
|
"valid_targets_mean": 3449.4,
|
|
"valid_targets_min": 1717
|
|
},
|
|
{
|
|
"epoch": 3.4371554575523704,
|
|
"grad_norm": 0.5882270819301996,
|
|
"learning_rate": 2.409439386502828e-05,
|
|
"loss": 0.2528,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11351832747459412,
|
|
"step": 6235,
|
|
"valid_targets_mean": 3208.2,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 3.439911797133407,
|
|
"grad_norm": 0.5770140578754144,
|
|
"learning_rate": 2.4067481922662945e-05,
|
|
"loss": 0.2337,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12162694334983826,
|
|
"step": 6240,
|
|
"valid_targets_mean": 3638.1,
|
|
"valid_targets_min": 1538
|
|
},
|
|
{
|
|
"epoch": 3.442668136714443,
|
|
"grad_norm": 0.5495050999089264,
|
|
"learning_rate": 2.4040562295634137e-05,
|
|
"loss": 0.2439,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10051672160625458,
|
|
"step": 6245,
|
|
"valid_targets_mean": 3239.1,
|
|
"valid_targets_min": 1449
|
|
},
|
|
{
|
|
"epoch": 3.4454244762954795,
|
|
"grad_norm": 0.5620620775855073,
|
|
"learning_rate": 2.4013635034800897e-05,
|
|
"loss": 0.2542,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12960955500602722,
|
|
"step": 6250,
|
|
"valid_targets_mean": 4332.6,
|
|
"valid_targets_min": 2546
|
|
},
|
|
{
|
|
"epoch": 3.448180815876516,
|
|
"grad_norm": 0.5932655242750333,
|
|
"learning_rate": 2.398670019103672e-05,
|
|
"loss": 0.2526,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10642994195222855,
|
|
"step": 6255,
|
|
"valid_targets_mean": 3518.0,
|
|
"valid_targets_min": 1313
|
|
},
|
|
{
|
|
"epoch": 3.4509371554575523,
|
|
"grad_norm": 0.5575731692804224,
|
|
"learning_rate": 2.395975781522938e-05,
|
|
"loss": 0.2449,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1418878436088562,
|
|
"step": 6260,
|
|
"valid_targets_mean": 4586.4,
|
|
"valid_targets_min": 1915
|
|
},
|
|
{
|
|
"epoch": 3.4536934950385887,
|
|
"grad_norm": 0.5405104472698241,
|
|
"learning_rate": 2.3932807958280938e-05,
|
|
"loss": 0.232,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10514342784881592,
|
|
"step": 6265,
|
|
"valid_targets_mean": 4111.0,
|
|
"valid_targets_min": 1998
|
|
},
|
|
{
|
|
"epoch": 3.456449834619625,
|
|
"grad_norm": 0.6285484091757434,
|
|
"learning_rate": 2.390585067110754e-05,
|
|
"loss": 0.24,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09783228486776352,
|
|
"step": 6270,
|
|
"valid_targets_mean": 3531.4,
|
|
"valid_targets_min": 2369
|
|
},
|
|
{
|
|
"epoch": 3.4592061742006615,
|
|
"grad_norm": 0.6146706670380021,
|
|
"learning_rate": 2.38788860046394e-05,
|
|
"loss": 0.2231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1293877214193344,
|
|
"step": 6275,
|
|
"valid_targets_mean": 3523.4,
|
|
"valid_targets_min": 2176
|
|
},
|
|
{
|
|
"epoch": 3.461962513781698,
|
|
"grad_norm": 0.6172235003576945,
|
|
"learning_rate": 2.3851914009820645e-05,
|
|
"loss": 0.2519,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14595340192317963,
|
|
"step": 6280,
|
|
"valid_targets_mean": 4304.4,
|
|
"valid_targets_min": 3027
|
|
},
|
|
{
|
|
"epoch": 3.464718853362734,
|
|
"grad_norm": 0.5441066265027635,
|
|
"learning_rate": 2.3824934737609283e-05,
|
|
"loss": 0.2384,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09061658382415771,
|
|
"step": 6285,
|
|
"valid_targets_mean": 3186.2,
|
|
"valid_targets_min": 1475
|
|
},
|
|
{
|
|
"epoch": 3.4674751929437706,
|
|
"grad_norm": 0.5465377515237156,
|
|
"learning_rate": 2.3797948238977044e-05,
|
|
"loss": 0.2376,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1014673039317131,
|
|
"step": 6290,
|
|
"valid_targets_mean": 3240.8,
|
|
"valid_targets_min": 646
|
|
},
|
|
{
|
|
"epoch": 3.470231532524807,
|
|
"grad_norm": 0.5555477784253697,
|
|
"learning_rate": 2.3770954564909326e-05,
|
|
"loss": 0.2418,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12894469499588013,
|
|
"step": 6295,
|
|
"valid_targets_mean": 3814.9,
|
|
"valid_targets_min": 1229
|
|
},
|
|
{
|
|
"epoch": 3.4729878721058434,
|
|
"grad_norm": 0.5742628054970059,
|
|
"learning_rate": 2.3743953766405067e-05,
|
|
"loss": 0.2251,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10887102782726288,
|
|
"step": 6300,
|
|
"valid_targets_mean": 3464.8,
|
|
"valid_targets_min": 2826
|
|
},
|
|
{
|
|
"epoch": 3.4757442116868797,
|
|
"grad_norm": 0.5962623004402987,
|
|
"learning_rate": 2.3716945894476692e-05,
|
|
"loss": 0.2405,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11787180602550507,
|
|
"step": 6305,
|
|
"valid_targets_mean": 3609.2,
|
|
"valid_targets_min": 1781
|
|
},
|
|
{
|
|
"epoch": 3.478500551267916,
|
|
"grad_norm": 0.6102011607782009,
|
|
"learning_rate": 2.3689931000149962e-05,
|
|
"loss": 0.2348,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11969827115535736,
|
|
"step": 6310,
|
|
"valid_targets_mean": 3260.4,
|
|
"valid_targets_min": 1108
|
|
},
|
|
{
|
|
"epoch": 3.4812568908489525,
|
|
"grad_norm": 0.6017122457889318,
|
|
"learning_rate": 2.3662909134463917e-05,
|
|
"loss": 0.2338,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08846394717693329,
|
|
"step": 6315,
|
|
"valid_targets_mean": 2584.4,
|
|
"valid_targets_min": 1910
|
|
},
|
|
{
|
|
"epoch": 3.484013230429989,
|
|
"grad_norm": 0.5852234971390976,
|
|
"learning_rate": 2.363588034847077e-05,
|
|
"loss": 0.2302,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12393698841333389,
|
|
"step": 6320,
|
|
"valid_targets_mean": 3782.1,
|
|
"valid_targets_min": 1826
|
|
},
|
|
{
|
|
"epoch": 3.4867695700110253,
|
|
"grad_norm": 0.5799168628509384,
|
|
"learning_rate": 2.3608844693235807e-05,
|
|
"loss": 0.2374,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11487237364053726,
|
|
"step": 6325,
|
|
"valid_targets_mean": 4205.8,
|
|
"valid_targets_min": 1587
|
|
},
|
|
{
|
|
"epoch": 3.4895259095920617,
|
|
"grad_norm": 0.6078556288462733,
|
|
"learning_rate": 2.358180221983729e-05,
|
|
"loss": 0.233,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1130157858133316,
|
|
"step": 6330,
|
|
"valid_targets_mean": 3286.8,
|
|
"valid_targets_min": 1326
|
|
},
|
|
{
|
|
"epoch": 3.492282249173098,
|
|
"grad_norm": 0.6304666694172644,
|
|
"learning_rate": 2.3554752979366367e-05,
|
|
"loss": 0.2377,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09004136174917221,
|
|
"step": 6335,
|
|
"valid_targets_mean": 2832.0,
|
|
"valid_targets_min": 1240
|
|
},
|
|
{
|
|
"epoch": 3.4950385887541344,
|
|
"grad_norm": 0.5939784539038993,
|
|
"learning_rate": 2.352769702292696e-05,
|
|
"loss": 0.2444,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10852304100990295,
|
|
"step": 6340,
|
|
"valid_targets_mean": 3817.9,
|
|
"valid_targets_min": 1722
|
|
},
|
|
{
|
|
"epoch": 3.497794928335171,
|
|
"grad_norm": 0.5986642591121064,
|
|
"learning_rate": 2.3500634401635695e-05,
|
|
"loss": 0.2313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12920263409614563,
|
|
"step": 6345,
|
|
"valid_targets_mean": 4388.9,
|
|
"valid_targets_min": 1960
|
|
},
|
|
{
|
|
"epoch": 3.500551267916207,
|
|
"grad_norm": 0.5939490098274539,
|
|
"learning_rate": 2.3473565166621778e-05,
|
|
"loss": 0.2334,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10748495161533356,
|
|
"step": 6350,
|
|
"valid_targets_mean": 3341.1,
|
|
"valid_targets_min": 1183
|
|
},
|
|
{
|
|
"epoch": 3.5033076074972436,
|
|
"grad_norm": 0.5616617310737125,
|
|
"learning_rate": 2.3446489369026912e-05,
|
|
"loss": 0.2362,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11116499453783035,
|
|
"step": 6355,
|
|
"valid_targets_mean": 4164.9,
|
|
"valid_targets_min": 2488
|
|
},
|
|
{
|
|
"epoch": 3.50606394707828,
|
|
"grad_norm": 0.5822043567777454,
|
|
"learning_rate": 2.3419407060005207e-05,
|
|
"loss": 0.2421,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12118351459503174,
|
|
"step": 6360,
|
|
"valid_targets_mean": 3456.2,
|
|
"valid_targets_min": 2495
|
|
},
|
|
{
|
|
"epoch": 3.5088202866593163,
|
|
"grad_norm": 0.6450334503145673,
|
|
"learning_rate": 2.339231829072307e-05,
|
|
"loss": 0.2306,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14776748418807983,
|
|
"step": 6365,
|
|
"valid_targets_mean": 3753.0,
|
|
"valid_targets_min": 1644
|
|
},
|
|
{
|
|
"epoch": 3.5115766262403527,
|
|
"grad_norm": 0.6000955606376773,
|
|
"learning_rate": 2.33652231123591e-05,
|
|
"loss": 0.2367,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11825516819953918,
|
|
"step": 6370,
|
|
"valid_targets_mean": 3786.4,
|
|
"valid_targets_min": 1579
|
|
},
|
|
{
|
|
"epoch": 3.514332965821389,
|
|
"grad_norm": 0.5889797874581091,
|
|
"learning_rate": 2.3338121576104028e-05,
|
|
"loss": 0.2337,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1244833916425705,
|
|
"step": 6375,
|
|
"valid_targets_mean": 3710.0,
|
|
"valid_targets_min": 1574
|
|
},
|
|
{
|
|
"epoch": 3.5170893054024255,
|
|
"grad_norm": 0.5251695497357054,
|
|
"learning_rate": 2.3311013733160585e-05,
|
|
"loss": 0.2264,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1059429943561554,
|
|
"step": 6380,
|
|
"valid_targets_mean": 4509.0,
|
|
"valid_targets_min": 2092
|
|
},
|
|
{
|
|
"epoch": 3.519845644983462,
|
|
"grad_norm": 0.5771742318417672,
|
|
"learning_rate": 2.3283899634743414e-05,
|
|
"loss": 0.2326,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10951802134513855,
|
|
"step": 6385,
|
|
"valid_targets_mean": 3671.8,
|
|
"valid_targets_min": 1672
|
|
},
|
|
{
|
|
"epoch": 3.5226019845644982,
|
|
"grad_norm": 0.5351123123884047,
|
|
"learning_rate": 2.3256779332078986e-05,
|
|
"loss": 0.2379,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13419727981090546,
|
|
"step": 6390,
|
|
"valid_targets_mean": 4776.5,
|
|
"valid_targets_min": 2483
|
|
},
|
|
{
|
|
"epoch": 3.5253583241455346,
|
|
"grad_norm": 0.5945321016766635,
|
|
"learning_rate": 2.322965287640548e-05,
|
|
"loss": 0.2311,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10337097942829132,
|
|
"step": 6395,
|
|
"valid_targets_mean": 3441.5,
|
|
"valid_targets_min": 741
|
|
},
|
|
{
|
|
"epoch": 3.528114663726571,
|
|
"grad_norm": 0.5581309912396306,
|
|
"learning_rate": 2.3202520318972708e-05,
|
|
"loss": 0.2422,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12688183784484863,
|
|
"step": 6400,
|
|
"valid_targets_mean": 4108.2,
|
|
"valid_targets_min": 1756
|
|
},
|
|
{
|
|
"epoch": 3.5308710033076074,
|
|
"grad_norm": 0.5695767839328957,
|
|
"learning_rate": 2.317538171104202e-05,
|
|
"loss": 0.2315,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11882161349058151,
|
|
"step": 6405,
|
|
"valid_targets_mean": 4258.1,
|
|
"valid_targets_min": 1504
|
|
},
|
|
{
|
|
"epoch": 3.5336273428886438,
|
|
"grad_norm": 0.5617972611638663,
|
|
"learning_rate": 2.314823710388618e-05,
|
|
"loss": 0.2421,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12450900673866272,
|
|
"step": 6410,
|
|
"valid_targets_mean": 3893.6,
|
|
"valid_targets_min": 1168
|
|
},
|
|
{
|
|
"epoch": 3.53638368246968,
|
|
"grad_norm": 0.592800038113691,
|
|
"learning_rate": 2.312108654878929e-05,
|
|
"loss": 0.2293,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11428281664848328,
|
|
"step": 6415,
|
|
"valid_targets_mean": 2834.5,
|
|
"valid_targets_min": 1953
|
|
},
|
|
{
|
|
"epoch": 3.5391400220507165,
|
|
"grad_norm": 0.5753603336714727,
|
|
"learning_rate": 2.3093930097046697e-05,
|
|
"loss": 0.2415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11373285204172134,
|
|
"step": 6420,
|
|
"valid_targets_mean": 3374.2,
|
|
"valid_targets_min": 1213
|
|
},
|
|
{
|
|
"epoch": 3.541896361631753,
|
|
"grad_norm": 0.5007177129114649,
|
|
"learning_rate": 2.3066767799964885e-05,
|
|
"loss": 0.2234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09671217948198318,
|
|
"step": 6425,
|
|
"valid_targets_mean": 4139.9,
|
|
"valid_targets_min": 2749
|
|
},
|
|
{
|
|
"epoch": 3.5446527012127893,
|
|
"grad_norm": 0.6095629790019452,
|
|
"learning_rate": 2.303959970886138e-05,
|
|
"loss": 0.2249,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11914513260126114,
|
|
"step": 6430,
|
|
"valid_targets_mean": 4366.9,
|
|
"valid_targets_min": 2525
|
|
},
|
|
{
|
|
"epoch": 3.5474090407938257,
|
|
"grad_norm": 0.5894601764010948,
|
|
"learning_rate": 2.3012425875064657e-05,
|
|
"loss": 0.2265,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.134830504655838,
|
|
"step": 6435,
|
|
"valid_targets_mean": 4376.4,
|
|
"valid_targets_min": 462
|
|
},
|
|
{
|
|
"epoch": 3.550165380374862,
|
|
"grad_norm": 0.5524726306805005,
|
|
"learning_rate": 2.298524634991403e-05,
|
|
"loss": 0.2403,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11727046966552734,
|
|
"step": 6440,
|
|
"valid_targets_mean": 3758.4,
|
|
"valid_targets_min": 1001
|
|
},
|
|
{
|
|
"epoch": 3.5529217199558984,
|
|
"grad_norm": 0.562559547302879,
|
|
"learning_rate": 2.2958061184759586e-05,
|
|
"loss": 0.2316,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1261035054922104,
|
|
"step": 6445,
|
|
"valid_targets_mean": 3720.1,
|
|
"valid_targets_min": 1622
|
|
},
|
|
{
|
|
"epoch": 3.555678059536935,
|
|
"grad_norm": 0.5755513944348452,
|
|
"learning_rate": 2.2930870430962055e-05,
|
|
"loss": 0.2373,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11407224833965302,
|
|
"step": 6450,
|
|
"valid_targets_mean": 3742.9,
|
|
"valid_targets_min": 1963
|
|
},
|
|
{
|
|
"epoch": 3.558434399117971,
|
|
"grad_norm": 0.5883197472373739,
|
|
"learning_rate": 2.2903674139892717e-05,
|
|
"loss": 0.2329,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11099185794591904,
|
|
"step": 6455,
|
|
"valid_targets_mean": 3838.6,
|
|
"valid_targets_min": 1456
|
|
},
|
|
{
|
|
"epoch": 3.5611907386990076,
|
|
"grad_norm": 0.5570435021398833,
|
|
"learning_rate": 2.2876472362933333e-05,
|
|
"loss": 0.2425,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11305543780326843,
|
|
"step": 6460,
|
|
"valid_targets_mean": 4055.6,
|
|
"valid_targets_min": 1171
|
|
},
|
|
{
|
|
"epoch": 3.563947078280044,
|
|
"grad_norm": 0.5408844468230256,
|
|
"learning_rate": 2.2849265151476016e-05,
|
|
"loss": 0.2363,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.126931831240654,
|
|
"step": 6465,
|
|
"valid_targets_mean": 4132.1,
|
|
"valid_targets_min": 1814
|
|
},
|
|
{
|
|
"epoch": 3.5667034178610804,
|
|
"grad_norm": 0.5660331688767481,
|
|
"learning_rate": 2.2822052556923145e-05,
|
|
"loss": 0.2254,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10984844714403152,
|
|
"step": 6470,
|
|
"valid_targets_mean": 3922.2,
|
|
"valid_targets_min": 1529
|
|
},
|
|
{
|
|
"epoch": 3.5694597574421167,
|
|
"grad_norm": 0.6326048345276221,
|
|
"learning_rate": 2.279483463068728e-05,
|
|
"loss": 0.229,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15491682291030884,
|
|
"step": 6475,
|
|
"valid_targets_mean": 5364.6,
|
|
"valid_targets_min": 1813
|
|
},
|
|
{
|
|
"epoch": 3.572216097023153,
|
|
"grad_norm": 0.661730755332824,
|
|
"learning_rate": 2.2767611424191043e-05,
|
|
"loss": 0.2215,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1261109560728073,
|
|
"step": 6480,
|
|
"valid_targets_mean": 2870.6,
|
|
"valid_targets_min": 1071
|
|
},
|
|
{
|
|
"epoch": 3.5749724366041895,
|
|
"grad_norm": 0.5600630355727813,
|
|
"learning_rate": 2.274038298886704e-05,
|
|
"loss": 0.2233,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11618770658969879,
|
|
"step": 6485,
|
|
"valid_targets_mean": 3716.1,
|
|
"valid_targets_min": 2416
|
|
},
|
|
{
|
|
"epoch": 3.577728776185226,
|
|
"grad_norm": 0.6001017336048586,
|
|
"learning_rate": 2.271314937615775e-05,
|
|
"loss": 0.2146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12243659049272537,
|
|
"step": 6490,
|
|
"valid_targets_mean": 4061.5,
|
|
"valid_targets_min": 1077
|
|
},
|
|
{
|
|
"epoch": 3.5804851157662623,
|
|
"grad_norm": 0.5852133564324503,
|
|
"learning_rate": 2.2685910637515427e-05,
|
|
"loss": 0.2126,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09722046554088593,
|
|
"step": 6495,
|
|
"valid_targets_mean": 3204.9,
|
|
"valid_targets_min": 1143
|
|
},
|
|
{
|
|
"epoch": 3.5832414553472987,
|
|
"grad_norm": 0.5705242616143488,
|
|
"learning_rate": 2.2658666824402037e-05,
|
|
"loss": 0.2235,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09298162907361984,
|
|
"step": 6500,
|
|
"valid_targets_mean": 3009.2,
|
|
"valid_targets_min": 1852
|
|
},
|
|
{
|
|
"epoch": 3.585997794928335,
|
|
"grad_norm": 0.6947001413792235,
|
|
"learning_rate": 2.26314179882891e-05,
|
|
"loss": 0.2493,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11058565974235535,
|
|
"step": 6505,
|
|
"valid_targets_mean": 2987.4,
|
|
"valid_targets_min": 1408
|
|
},
|
|
{
|
|
"epoch": 3.5887541345093714,
|
|
"grad_norm": 0.6159310751710421,
|
|
"learning_rate": 2.2604164180657648e-05,
|
|
"loss": 0.2251,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1000012531876564,
|
|
"step": 6510,
|
|
"valid_targets_mean": 2556.1,
|
|
"valid_targets_min": 1563
|
|
},
|
|
{
|
|
"epoch": 3.591510474090408,
|
|
"grad_norm": 0.5941227363737334,
|
|
"learning_rate": 2.2576905452998093e-05,
|
|
"loss": 0.2352,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10665936022996902,
|
|
"step": 6515,
|
|
"valid_targets_mean": 3792.4,
|
|
"valid_targets_min": 2318
|
|
},
|
|
{
|
|
"epoch": 3.594266813671444,
|
|
"grad_norm": 0.5571640567760464,
|
|
"learning_rate": 2.2549641856810144e-05,
|
|
"loss": 0.2214,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09101991355419159,
|
|
"step": 6520,
|
|
"valid_targets_mean": 2644.8,
|
|
"valid_targets_min": 1251
|
|
},
|
|
{
|
|
"epoch": 3.5970231532524806,
|
|
"grad_norm": 0.5782948249265432,
|
|
"learning_rate": 2.252237344360272e-05,
|
|
"loss": 0.2265,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1310802698135376,
|
|
"step": 6525,
|
|
"valid_targets_mean": 4273.8,
|
|
"valid_targets_min": 2248
|
|
},
|
|
{
|
|
"epoch": 3.599779492833517,
|
|
"grad_norm": 0.5944509254410688,
|
|
"learning_rate": 2.249510026489383e-05,
|
|
"loss": 0.2239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.103484146296978,
|
|
"step": 6530,
|
|
"valid_targets_mean": 3358.5,
|
|
"valid_targets_min": 1272
|
|
},
|
|
{
|
|
"epoch": 3.6025358324145533,
|
|
"grad_norm": 0.5604715051373428,
|
|
"learning_rate": 2.2467822372210487e-05,
|
|
"loss": 0.2237,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10655966401100159,
|
|
"step": 6535,
|
|
"valid_targets_mean": 3453.5,
|
|
"valid_targets_min": 1445
|
|
},
|
|
{
|
|
"epoch": 3.6052921719955897,
|
|
"grad_norm": 0.5904847839315162,
|
|
"learning_rate": 2.244053981708861e-05,
|
|
"loss": 0.2272,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14432498812675476,
|
|
"step": 6540,
|
|
"valid_targets_mean": 4757.5,
|
|
"valid_targets_min": 3700
|
|
},
|
|
{
|
|
"epoch": 3.608048511576626,
|
|
"grad_norm": 0.5801171305046676,
|
|
"learning_rate": 2.241325265107293e-05,
|
|
"loss": 0.2254,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12514647841453552,
|
|
"step": 6545,
|
|
"valid_targets_mean": 4462.0,
|
|
"valid_targets_min": 2708
|
|
},
|
|
{
|
|
"epoch": 3.6108048511576625,
|
|
"grad_norm": 0.5853053055582371,
|
|
"learning_rate": 2.238596092571689e-05,
|
|
"loss": 0.2216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11838936805725098,
|
|
"step": 6550,
|
|
"valid_targets_mean": 3549.4,
|
|
"valid_targets_min": 1505
|
|
},
|
|
{
|
|
"epoch": 3.613561190738699,
|
|
"grad_norm": 0.573076043245387,
|
|
"learning_rate": 2.2358664692582542e-05,
|
|
"loss": 0.2393,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11420096457004547,
|
|
"step": 6555,
|
|
"valid_targets_mean": 4404.0,
|
|
"valid_targets_min": 3497
|
|
},
|
|
{
|
|
"epoch": 3.6163175303197352,
|
|
"grad_norm": 0.5535420909346925,
|
|
"learning_rate": 2.2331364003240456e-05,
|
|
"loss": 0.2338,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11688980460166931,
|
|
"step": 6560,
|
|
"valid_targets_mean": 4091.5,
|
|
"valid_targets_min": 2368
|
|
},
|
|
{
|
|
"epoch": 3.6190738699007716,
|
|
"grad_norm": 0.5663832073840436,
|
|
"learning_rate": 2.230405890926962e-05,
|
|
"loss": 0.2383,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14589238166809082,
|
|
"step": 6565,
|
|
"valid_targets_mean": 4878.1,
|
|
"valid_targets_min": 3436
|
|
},
|
|
{
|
|
"epoch": 3.621830209481808,
|
|
"grad_norm": 0.6100729799613839,
|
|
"learning_rate": 2.2276749462257355e-05,
|
|
"loss": 0.2373,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1151973307132721,
|
|
"step": 6570,
|
|
"valid_targets_mean": 3702.0,
|
|
"valid_targets_min": 1134
|
|
},
|
|
{
|
|
"epoch": 3.6245865490628444,
|
|
"grad_norm": 0.6893642593720102,
|
|
"learning_rate": 2.2249435713799188e-05,
|
|
"loss": 0.2339,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1206585094332695,
|
|
"step": 6575,
|
|
"valid_targets_mean": 4411.8,
|
|
"valid_targets_min": 1443
|
|
},
|
|
{
|
|
"epoch": 3.6273428886438808,
|
|
"grad_norm": 0.610846959478792,
|
|
"learning_rate": 2.2222117715498788e-05,
|
|
"loss": 0.2377,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11065837740898132,
|
|
"step": 6580,
|
|
"valid_targets_mean": 3434.1,
|
|
"valid_targets_min": 1396
|
|
},
|
|
{
|
|
"epoch": 3.630099228224917,
|
|
"grad_norm": 0.5746393040036764,
|
|
"learning_rate": 2.2194795518967847e-05,
|
|
"loss": 0.2429,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09952611476182938,
|
|
"step": 6585,
|
|
"valid_targets_mean": 3704.8,
|
|
"valid_targets_min": 1428
|
|
},
|
|
{
|
|
"epoch": 3.6328555678059535,
|
|
"grad_norm": 0.5564038133351629,
|
|
"learning_rate": 2.2167469175825974e-05,
|
|
"loss": 0.2243,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09552530944347382,
|
|
"step": 6590,
|
|
"valid_targets_mean": 3643.9,
|
|
"valid_targets_min": 613
|
|
},
|
|
{
|
|
"epoch": 3.63561190738699,
|
|
"grad_norm": 0.5993590998109386,
|
|
"learning_rate": 2.214013873770065e-05,
|
|
"loss": 0.2298,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10809554904699326,
|
|
"step": 6595,
|
|
"valid_targets_mean": 3467.8,
|
|
"valid_targets_min": 2040
|
|
},
|
|
{
|
|
"epoch": 3.6383682469680263,
|
|
"grad_norm": 0.643675548444291,
|
|
"learning_rate": 2.211280425622705e-05,
|
|
"loss": 0.2393,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1150679886341095,
|
|
"step": 6600,
|
|
"valid_targets_mean": 3434.5,
|
|
"valid_targets_min": 1331
|
|
},
|
|
{
|
|
"epoch": 3.6411245865490627,
|
|
"grad_norm": 0.5620685538709803,
|
|
"learning_rate": 2.2085465783048022e-05,
|
|
"loss": 0.2277,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11839953064918518,
|
|
"step": 6605,
|
|
"valid_targets_mean": 4514.4,
|
|
"valid_targets_min": 1320
|
|
},
|
|
{
|
|
"epoch": 3.643880926130099,
|
|
"grad_norm": 0.5344708995635962,
|
|
"learning_rate": 2.2058123369813932e-05,
|
|
"loss": 0.2314,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09781419485807419,
|
|
"step": 6610,
|
|
"valid_targets_mean": 3989.2,
|
|
"valid_targets_min": 2752
|
|
},
|
|
{
|
|
"epoch": 3.646637265711136,
|
|
"grad_norm": 0.5863535931725102,
|
|
"learning_rate": 2.203077706818259e-05,
|
|
"loss": 0.2189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10421283543109894,
|
|
"step": 6615,
|
|
"valid_targets_mean": 3245.4,
|
|
"valid_targets_min": 1989
|
|
},
|
|
{
|
|
"epoch": 3.649393605292172,
|
|
"grad_norm": 0.5932942573077412,
|
|
"learning_rate": 2.2003426929819184e-05,
|
|
"loss": 0.2324,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13112995028495789,
|
|
"step": 6620,
|
|
"valid_targets_mean": 3923.4,
|
|
"valid_targets_min": 1074
|
|
},
|
|
{
|
|
"epoch": 3.6521499448732087,
|
|
"grad_norm": 0.5433107827579012,
|
|
"learning_rate": 2.1976073006396113e-05,
|
|
"loss": 0.2117,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1081799641251564,
|
|
"step": 6625,
|
|
"valid_targets_mean": 4182.0,
|
|
"valid_targets_min": 1684
|
|
},
|
|
{
|
|
"epoch": 3.6549062844542446,
|
|
"grad_norm": 0.5561581283675747,
|
|
"learning_rate": 2.1948715349592944e-05,
|
|
"loss": 0.2334,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11031346768140793,
|
|
"step": 6630,
|
|
"valid_targets_mean": 4215.0,
|
|
"valid_targets_min": 2179
|
|
},
|
|
{
|
|
"epoch": 3.6576626240352814,
|
|
"grad_norm": 0.5491635711724802,
|
|
"learning_rate": 2.1921354011096302e-05,
|
|
"loss": 0.2266,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09039963781833649,
|
|
"step": 6635,
|
|
"valid_targets_mean": 3483.4,
|
|
"valid_targets_min": 2023
|
|
},
|
|
{
|
|
"epoch": 3.6604189636163174,
|
|
"grad_norm": 0.5839804271132429,
|
|
"learning_rate": 2.1893989042599744e-05,
|
|
"loss": 0.2336,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11396078020334244,
|
|
"step": 6640,
|
|
"valid_targets_mean": 3810.9,
|
|
"valid_targets_min": 1776
|
|
},
|
|
{
|
|
"epoch": 3.663175303197354,
|
|
"grad_norm": 0.5921243800531575,
|
|
"learning_rate": 2.1866620495803722e-05,
|
|
"loss": 0.2194,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09831385314464569,
|
|
"step": 6645,
|
|
"valid_targets_mean": 3623.2,
|
|
"valid_targets_min": 1964
|
|
},
|
|
{
|
|
"epoch": 3.66593164277839,
|
|
"grad_norm": 0.6178094944790615,
|
|
"learning_rate": 2.183924842241542e-05,
|
|
"loss": 0.234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14169856905937195,
|
|
"step": 6650,
|
|
"valid_targets_mean": 5048.9,
|
|
"valid_targets_min": 2580
|
|
},
|
|
{
|
|
"epoch": 3.668687982359427,
|
|
"grad_norm": 0.5694024200824386,
|
|
"learning_rate": 2.1811872874148688e-05,
|
|
"loss": 0.2258,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11276976019144058,
|
|
"step": 6655,
|
|
"valid_targets_mean": 4726.5,
|
|
"valid_targets_min": 2757
|
|
},
|
|
{
|
|
"epoch": 3.671444321940463,
|
|
"grad_norm": 0.6143196024545351,
|
|
"learning_rate": 2.178449390272396e-05,
|
|
"loss": 0.2218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10178348422050476,
|
|
"step": 6660,
|
|
"valid_targets_mean": 3046.2,
|
|
"valid_targets_min": 1033
|
|
},
|
|
{
|
|
"epoch": 3.6742006615214997,
|
|
"grad_norm": 0.564578648764109,
|
|
"learning_rate": 2.175711155986811e-05,
|
|
"loss": 0.2161,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13365858793258667,
|
|
"step": 6665,
|
|
"valid_targets_mean": 4375.8,
|
|
"valid_targets_min": 1877
|
|
},
|
|
{
|
|
"epoch": 3.6769570011025356,
|
|
"grad_norm": 0.5632339610649862,
|
|
"learning_rate": 2.172972589731441e-05,
|
|
"loss": 0.2183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10958202183246613,
|
|
"step": 6670,
|
|
"valid_targets_mean": 3824.2,
|
|
"valid_targets_min": 2480
|
|
},
|
|
{
|
|
"epoch": 3.6797133406835725,
|
|
"grad_norm": 0.5775008238134777,
|
|
"learning_rate": 2.170233696680238e-05,
|
|
"loss": 0.2162,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10177493095397949,
|
|
"step": 6675,
|
|
"valid_targets_mean": 3920.5,
|
|
"valid_targets_min": 1357
|
|
},
|
|
{
|
|
"epoch": 3.6824696802646084,
|
|
"grad_norm": 0.5531349479236645,
|
|
"learning_rate": 2.167494482007772e-05,
|
|
"loss": 0.2111,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11939731985330582,
|
|
"step": 6680,
|
|
"valid_targets_mean": 4533.2,
|
|
"valid_targets_min": 1495
|
|
},
|
|
{
|
|
"epoch": 3.6852260198456452,
|
|
"grad_norm": 0.6289113094556231,
|
|
"learning_rate": 2.164754950889221e-05,
|
|
"loss": 0.2131,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12104576826095581,
|
|
"step": 6685,
|
|
"valid_targets_mean": 3572.5,
|
|
"valid_targets_min": 2053
|
|
},
|
|
{
|
|
"epoch": 3.687982359426681,
|
|
"grad_norm": 0.5747091363794755,
|
|
"learning_rate": 2.1620151085003625e-05,
|
|
"loss": 0.2178,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09782974421977997,
|
|
"step": 6690,
|
|
"valid_targets_mean": 3716.4,
|
|
"valid_targets_min": 1149
|
|
},
|
|
{
|
|
"epoch": 3.690738699007718,
|
|
"grad_norm": 0.6645967328304315,
|
|
"learning_rate": 2.1592749600175594e-05,
|
|
"loss": 0.224,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09480296075344086,
|
|
"step": 6695,
|
|
"valid_targets_mean": 3558.0,
|
|
"valid_targets_min": 1597
|
|
},
|
|
{
|
|
"epoch": 3.693495038588754,
|
|
"grad_norm": 0.5790936617253304,
|
|
"learning_rate": 2.1565345106177533e-05,
|
|
"loss": 0.238,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10597217082977295,
|
|
"step": 6700,
|
|
"valid_targets_mean": 4039.0,
|
|
"valid_targets_min": 2266
|
|
},
|
|
{
|
|
"epoch": 3.6962513781697908,
|
|
"grad_norm": 0.6347162774551712,
|
|
"learning_rate": 2.1537937654784556e-05,
|
|
"loss": 0.2142,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10220508277416229,
|
|
"step": 6705,
|
|
"valid_targets_mean": 3054.6,
|
|
"valid_targets_min": 1671
|
|
},
|
|
{
|
|
"epoch": 3.6990077177508267,
|
|
"grad_norm": 0.6229222261641609,
|
|
"learning_rate": 2.1510527297777357e-05,
|
|
"loss": 0.2179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10816439986228943,
|
|
"step": 6710,
|
|
"valid_targets_mean": 3339.2,
|
|
"valid_targets_min": 1971
|
|
},
|
|
{
|
|
"epoch": 3.7017640573318635,
|
|
"grad_norm": 0.5952127250472791,
|
|
"learning_rate": 2.148311408694212e-05,
|
|
"loss": 0.2159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08948597311973572,
|
|
"step": 6715,
|
|
"valid_targets_mean": 2995.4,
|
|
"valid_targets_min": 1664
|
|
},
|
|
{
|
|
"epoch": 3.7045203969128995,
|
|
"grad_norm": 0.5071708765109028,
|
|
"learning_rate": 2.1455698074070422e-05,
|
|
"loss": 0.2246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09137539565563202,
|
|
"step": 6720,
|
|
"valid_targets_mean": 4727.0,
|
|
"valid_targets_min": 3214
|
|
},
|
|
{
|
|
"epoch": 3.7072767364939363,
|
|
"grad_norm": 0.5774437962626661,
|
|
"learning_rate": 2.1428279310959136e-05,
|
|
"loss": 0.2308,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12156601250171661,
|
|
"step": 6725,
|
|
"valid_targets_mean": 3910.0,
|
|
"valid_targets_min": 2416
|
|
},
|
|
{
|
|
"epoch": 3.7100330760749722,
|
|
"grad_norm": 0.5719476320152961,
|
|
"learning_rate": 2.1400857849410324e-05,
|
|
"loss": 0.2251,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12000076472759247,
|
|
"step": 6730,
|
|
"valid_targets_mean": 4422.2,
|
|
"valid_targets_min": 2100
|
|
},
|
|
{
|
|
"epoch": 3.712789415656009,
|
|
"grad_norm": 0.5841212467716652,
|
|
"learning_rate": 2.1373433741231144e-05,
|
|
"loss": 0.2243,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11770911514759064,
|
|
"step": 6735,
|
|
"valid_targets_mean": 3553.8,
|
|
"valid_targets_min": 1363
|
|
},
|
|
{
|
|
"epoch": 3.715545755237045,
|
|
"grad_norm": 0.5917765857135034,
|
|
"learning_rate": 2.1346007038233774e-05,
|
|
"loss": 0.2262,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10699576139450073,
|
|
"step": 6740,
|
|
"valid_targets_mean": 3909.9,
|
|
"valid_targets_min": 1572
|
|
},
|
|
{
|
|
"epoch": 3.718302094818082,
|
|
"grad_norm": 0.61730039626189,
|
|
"learning_rate": 2.131857779223528e-05,
|
|
"loss": 0.2273,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1324215531349182,
|
|
"step": 6745,
|
|
"valid_targets_mean": 4026.5,
|
|
"valid_targets_min": 875
|
|
},
|
|
{
|
|
"epoch": 3.7210584343991178,
|
|
"grad_norm": 0.4135958043495425,
|
|
"learning_rate": 2.129114605505752e-05,
|
|
"loss": 0.1265,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0492122657597065,
|
|
"step": 6750,
|
|
"valid_targets_mean": 2741.9,
|
|
"valid_targets_min": 455
|
|
},
|
|
{
|
|
"epoch": 3.7238147739801546,
|
|
"grad_norm": 0.3854434640657211,
|
|
"learning_rate": 2.126371187852708e-05,
|
|
"loss": 0.0892,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04234106466174126,
|
|
"step": 6755,
|
|
"valid_targets_mean": 3359.0,
|
|
"valid_targets_min": 2797
|
|
},
|
|
{
|
|
"epoch": 3.7265711135611905,
|
|
"grad_norm": 0.3667024533832895,
|
|
"learning_rate": 2.1236275314475144e-05,
|
|
"loss": 0.0852,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04928459972143173,
|
|
"step": 6760,
|
|
"valid_targets_mean": 3332.0,
|
|
"valid_targets_min": 2059
|
|
},
|
|
{
|
|
"epoch": 3.7293274531422274,
|
|
"grad_norm": 0.47517264275278537,
|
|
"learning_rate": 2.120883641473742e-05,
|
|
"loss": 0.1054,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0505400076508522,
|
|
"step": 6765,
|
|
"valid_targets_mean": 3402.9,
|
|
"valid_targets_min": 2417
|
|
},
|
|
{
|
|
"epoch": 3.7320837927232633,
|
|
"grad_norm": 0.4911086522043579,
|
|
"learning_rate": 2.1181395231154004e-05,
|
|
"loss": 0.0853,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04971815645694733,
|
|
"step": 6770,
|
|
"valid_targets_mean": 2768.4,
|
|
"valid_targets_min": 1192
|
|
},
|
|
{
|
|
"epoch": 3.7348401323043,
|
|
"grad_norm": 0.6749115665932446,
|
|
"learning_rate": 2.1153951815569326e-05,
|
|
"loss": 0.0989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06025379151105881,
|
|
"step": 6775,
|
|
"valid_targets_mean": 1055.5,
|
|
"valid_targets_min": 803
|
|
},
|
|
{
|
|
"epoch": 3.737596471885336,
|
|
"grad_norm": 0.45345273612381143,
|
|
"learning_rate": 2.112650621983203e-05,
|
|
"loss": 0.1398,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05708978325128555,
|
|
"step": 6780,
|
|
"valid_targets_mean": 3310.8,
|
|
"valid_targets_min": 1943
|
|
},
|
|
{
|
|
"epoch": 3.740352811466373,
|
|
"grad_norm": 0.5579874889887487,
|
|
"learning_rate": 2.1099058495794874e-05,
|
|
"loss": 0.0925,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.061443109065294266,
|
|
"step": 6785,
|
|
"valid_targets_mean": 3068.4,
|
|
"valid_targets_min": 777
|
|
},
|
|
{
|
|
"epoch": 3.743109151047409,
|
|
"grad_norm": 0.32788958952594904,
|
|
"learning_rate": 2.1071608695314644e-05,
|
|
"loss": 0.1057,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03548944368958473,
|
|
"step": 6790,
|
|
"valid_targets_mean": 3493.8,
|
|
"valid_targets_min": 794
|
|
},
|
|
{
|
|
"epoch": 3.7458654906284456,
|
|
"grad_norm": 0.2987017312356992,
|
|
"learning_rate": 2.104415687025204e-05,
|
|
"loss": 0.1091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041137054562568665,
|
|
"step": 6795,
|
|
"valid_targets_mean": 4785.0,
|
|
"valid_targets_min": 2792
|
|
},
|
|
{
|
|
"epoch": 3.7486218302094816,
|
|
"grad_norm": 0.32257409252624825,
|
|
"learning_rate": 2.1016703072471593e-05,
|
|
"loss": 0.0835,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028763340786099434,
|
|
"step": 6800,
|
|
"valid_targets_mean": 2406.2,
|
|
"valid_targets_min": 571
|
|
},
|
|
{
|
|
"epoch": 3.7513781697905184,
|
|
"grad_norm": 0.3555388890722737,
|
|
"learning_rate": 2.0989247353841557e-05,
|
|
"loss": 0.0907,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049447283148765564,
|
|
"step": 6805,
|
|
"valid_targets_mean": 3945.8,
|
|
"valid_targets_min": 2633
|
|
},
|
|
{
|
|
"epoch": 3.7541345093715544,
|
|
"grad_norm": 0.5374256206005917,
|
|
"learning_rate": 2.0961789766233824e-05,
|
|
"loss": 0.0926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06006057560443878,
|
|
"step": 6810,
|
|
"valid_targets_mean": 1659.1,
|
|
"valid_targets_min": 590
|
|
},
|
|
{
|
|
"epoch": 3.756890848952591,
|
|
"grad_norm": 0.3920112094874407,
|
|
"learning_rate": 2.0934330361523812e-05,
|
|
"loss": 0.1232,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0389576330780983,
|
|
"step": 6815,
|
|
"valid_targets_mean": 3209.0,
|
|
"valid_targets_min": 1346
|
|
},
|
|
{
|
|
"epoch": 3.759647188533627,
|
|
"grad_norm": 0.4989553916852877,
|
|
"learning_rate": 2.0906869191590365e-05,
|
|
"loss": 0.0875,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05927523225545883,
|
|
"step": 6820,
|
|
"valid_targets_mean": 2070.0,
|
|
"valid_targets_min": 813
|
|
},
|
|
{
|
|
"epoch": 3.762403528114664,
|
|
"grad_norm": 0.3514893743378231,
|
|
"learning_rate": 2.0879406308315668e-05,
|
|
"loss": 0.0849,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034346580505371094,
|
|
"step": 6825,
|
|
"valid_targets_mean": 1865.4,
|
|
"valid_targets_min": 905
|
|
},
|
|
{
|
|
"epoch": 3.7651598676957,
|
|
"grad_norm": 0.34639684826491884,
|
|
"learning_rate": 2.0851941763585147e-05,
|
|
"loss": 0.078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04195670783519745,
|
|
"step": 6830,
|
|
"valid_targets_mean": 3711.9,
|
|
"valid_targets_min": 2833
|
|
},
|
|
{
|
|
"epoch": 3.7679162072767367,
|
|
"grad_norm": 0.4517818054025793,
|
|
"learning_rate": 2.082447560928736e-05,
|
|
"loss": 0.0983,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046112917363643646,
|
|
"step": 6835,
|
|
"valid_targets_mean": 2754.4,
|
|
"valid_targets_min": 468
|
|
},
|
|
{
|
|
"epoch": 3.7706725468577726,
|
|
"grad_norm": 0.3950533314899727,
|
|
"learning_rate": 2.079700789731391e-05,
|
|
"loss": 0.0976,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056785400956869125,
|
|
"step": 6840,
|
|
"valid_targets_mean": 3056.5,
|
|
"valid_targets_min": 802
|
|
},
|
|
{
|
|
"epoch": 3.7734288864388095,
|
|
"grad_norm": 0.4155411016626426,
|
|
"learning_rate": 2.0769538679559343e-05,
|
|
"loss": 0.1012,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04642801731824875,
|
|
"step": 6845,
|
|
"valid_targets_mean": 2917.2,
|
|
"valid_targets_min": 471
|
|
},
|
|
{
|
|
"epoch": 3.7761852260198454,
|
|
"grad_norm": 0.43339887312212766,
|
|
"learning_rate": 2.074206800792105e-05,
|
|
"loss": 0.0991,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043613798916339874,
|
|
"step": 6850,
|
|
"valid_targets_mean": 3326.2,
|
|
"valid_targets_min": 636
|
|
},
|
|
{
|
|
"epoch": 3.7789415656008822,
|
|
"grad_norm": 0.35991685748750607,
|
|
"learning_rate": 2.071459593429916e-05,
|
|
"loss": 0.1329,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03546847775578499,
|
|
"step": 6855,
|
|
"valid_targets_mean": 3525.8,
|
|
"valid_targets_min": 898
|
|
},
|
|
{
|
|
"epoch": 3.781697905181918,
|
|
"grad_norm": 0.41522757397874177,
|
|
"learning_rate": 2.0687122510596462e-05,
|
|
"loss": 0.0847,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045235127210617065,
|
|
"step": 6860,
|
|
"valid_targets_mean": 2996.2,
|
|
"valid_targets_min": 1829
|
|
},
|
|
{
|
|
"epoch": 3.784454244762955,
|
|
"grad_norm": 0.7101860548240275,
|
|
"learning_rate": 2.06596477887183e-05,
|
|
"loss": 0.1159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08586481213569641,
|
|
"step": 6865,
|
|
"valid_targets_mean": 1320.4,
|
|
"valid_targets_min": 752
|
|
},
|
|
{
|
|
"epoch": 3.787210584343991,
|
|
"grad_norm": 0.5855070458014543,
|
|
"learning_rate": 2.0632171820572454e-05,
|
|
"loss": 0.1187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08742330223321915,
|
|
"step": 6870,
|
|
"valid_targets_mean": 2208.9,
|
|
"valid_targets_min": 596
|
|
},
|
|
{
|
|
"epoch": 3.7899669239250278,
|
|
"grad_norm": 0.373048455662043,
|
|
"learning_rate": 2.0604694658069065e-05,
|
|
"loss": 0.0904,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03316938132047653,
|
|
"step": 6875,
|
|
"valid_targets_mean": 3024.1,
|
|
"valid_targets_min": 890
|
|
},
|
|
{
|
|
"epoch": 3.7927232635060637,
|
|
"grad_norm": 0.5855841371428925,
|
|
"learning_rate": 2.057721635312054e-05,
|
|
"loss": 0.203,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.17197391390800476,
|
|
"step": 6880,
|
|
"valid_targets_mean": 2456.2,
|
|
"valid_targets_min": 875
|
|
},
|
|
{
|
|
"epoch": 3.7954796030871005,
|
|
"grad_norm": 0.44244231586220517,
|
|
"learning_rate": 2.054973695764143e-05,
|
|
"loss": 0.1876,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041602082550525665,
|
|
"step": 6885,
|
|
"valid_targets_mean": 2688.0,
|
|
"valid_targets_min": 752
|
|
},
|
|
{
|
|
"epoch": 3.7982359426681365,
|
|
"grad_norm": 0.3673282823080746,
|
|
"learning_rate": 2.0522256523548362e-05,
|
|
"loss": 0.1038,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.040638431906700134,
|
|
"step": 6890,
|
|
"valid_targets_mean": 3327.1,
|
|
"valid_targets_min": 1960
|
|
},
|
|
{
|
|
"epoch": 3.8009922822491733,
|
|
"grad_norm": 0.7302305577245911,
|
|
"learning_rate": 2.049477510275991e-05,
|
|
"loss": 0.1639,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.21566346287727356,
|
|
"step": 6895,
|
|
"valid_targets_mean": 2729.6,
|
|
"valid_targets_min": 1188
|
|
},
|
|
{
|
|
"epoch": 3.8037486218302092,
|
|
"grad_norm": 0.30540267753911565,
|
|
"learning_rate": 2.0467292747196516e-05,
|
|
"loss": 0.1167,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03313196450471878,
|
|
"step": 6900,
|
|
"valid_targets_mean": 4566.2,
|
|
"valid_targets_min": 1693
|
|
},
|
|
{
|
|
"epoch": 3.806504961411246,
|
|
"grad_norm": 0.342808273944347,
|
|
"learning_rate": 2.04398095087804e-05,
|
|
"loss": 0.0892,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03234358876943588,
|
|
"step": 6905,
|
|
"valid_targets_mean": 3414.8,
|
|
"valid_targets_min": 785
|
|
},
|
|
{
|
|
"epoch": 3.809261300992282,
|
|
"grad_norm": 0.4006466406426037,
|
|
"learning_rate": 2.041232543943543e-05,
|
|
"loss": 0.0982,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05204131454229355,
|
|
"step": 6910,
|
|
"valid_targets_mean": 3699.1,
|
|
"valid_targets_min": 992
|
|
},
|
|
{
|
|
"epoch": 3.812017640573319,
|
|
"grad_norm": 0.36306701211978415,
|
|
"learning_rate": 2.038484059108707e-05,
|
|
"loss": 0.0872,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033380672335624695,
|
|
"step": 6915,
|
|
"valid_targets_mean": 2728.6,
|
|
"valid_targets_min": 485
|
|
},
|
|
{
|
|
"epoch": 3.8147739801543548,
|
|
"grad_norm": 0.4577404801522828,
|
|
"learning_rate": 2.0357355015662228e-05,
|
|
"loss": 0.089,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05728590488433838,
|
|
"step": 6920,
|
|
"valid_targets_mean": 3903.5,
|
|
"valid_targets_min": 2429
|
|
},
|
|
{
|
|
"epoch": 3.8175303197353916,
|
|
"grad_norm": 0.5231985973711474,
|
|
"learning_rate": 2.0329868765089197e-05,
|
|
"loss": 0.0854,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03369203954935074,
|
|
"step": 6925,
|
|
"valid_targets_mean": 1569.0,
|
|
"valid_targets_min": 723
|
|
},
|
|
{
|
|
"epoch": 3.8202866593164275,
|
|
"grad_norm": 0.41973956170633603,
|
|
"learning_rate": 2.0302381891297553e-05,
|
|
"loss": 0.0872,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03964103385806084,
|
|
"step": 6930,
|
|
"valid_targets_mean": 3352.5,
|
|
"valid_targets_min": 1763
|
|
},
|
|
{
|
|
"epoch": 3.8230429988974644,
|
|
"grad_norm": 0.3714278031600684,
|
|
"learning_rate": 2.0274894446218042e-05,
|
|
"loss": 0.0773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04716416075825691,
|
|
"step": 6935,
|
|
"valid_targets_mean": 3818.8,
|
|
"valid_targets_min": 3073
|
|
},
|
|
{
|
|
"epoch": 3.8257993384785003,
|
|
"grad_norm": 0.33654306353984187,
|
|
"learning_rate": 2.024740648178249e-05,
|
|
"loss": 0.1381,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03179306536912918,
|
|
"step": 6940,
|
|
"valid_targets_mean": 3618.4,
|
|
"valid_targets_min": 2915
|
|
},
|
|
{
|
|
"epoch": 3.828555678059537,
|
|
"grad_norm": 0.35567396228841763,
|
|
"learning_rate": 2.0219918049923707e-05,
|
|
"loss": 0.0689,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03320031613111496,
|
|
"step": 6945,
|
|
"valid_targets_mean": 3032.0,
|
|
"valid_targets_min": 718
|
|
},
|
|
{
|
|
"epoch": 3.831312017640573,
|
|
"grad_norm": 0.30073989701225184,
|
|
"learning_rate": 2.0192429202575372e-05,
|
|
"loss": 0.0794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.025534216314554214,
|
|
"step": 6950,
|
|
"valid_targets_mean": 3527.1,
|
|
"valid_targets_min": 2531
|
|
},
|
|
{
|
|
"epoch": 3.83406835722161,
|
|
"grad_norm": 0.38920310845635087,
|
|
"learning_rate": 2.0164939991671976e-05,
|
|
"loss": 0.0796,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03747660294175148,
|
|
"step": 6955,
|
|
"valid_targets_mean": 3218.5,
|
|
"valid_targets_min": 944
|
|
},
|
|
{
|
|
"epoch": 3.836824696802646,
|
|
"grad_norm": 0.4759346024497798,
|
|
"learning_rate": 2.0137450469148674e-05,
|
|
"loss": 0.0952,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03965476155281067,
|
|
"step": 6960,
|
|
"valid_targets_mean": 1023.1,
|
|
"valid_targets_min": 487
|
|
},
|
|
{
|
|
"epoch": 3.8395810363836826,
|
|
"grad_norm": 0.3606623805415833,
|
|
"learning_rate": 2.0109960686941223e-05,
|
|
"loss": 0.1102,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046469129621982574,
|
|
"step": 6965,
|
|
"valid_targets_mean": 3657.1,
|
|
"valid_targets_min": 2788
|
|
},
|
|
{
|
|
"epoch": 3.8423373759647186,
|
|
"grad_norm": 0.4833872087022775,
|
|
"learning_rate": 2.0082470696985856e-05,
|
|
"loss": 0.1236,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07054344564676285,
|
|
"step": 6970,
|
|
"valid_targets_mean": 1852.9,
|
|
"valid_targets_min": 795
|
|
},
|
|
{
|
|
"epoch": 3.8450937155457554,
|
|
"grad_norm": 0.4406894556088469,
|
|
"learning_rate": 2.005498055121921e-05,
|
|
"loss": 0.0946,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04457228258252144,
|
|
"step": 6975,
|
|
"valid_targets_mean": 2663.5,
|
|
"valid_targets_min": 521
|
|
},
|
|
{
|
|
"epoch": 3.8478500551267913,
|
|
"grad_norm": 0.45936993650234254,
|
|
"learning_rate": 2.0027490301578217e-05,
|
|
"loss": 0.1099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05745764076709747,
|
|
"step": 6980,
|
|
"valid_targets_mean": 3276.8,
|
|
"valid_targets_min": 707
|
|
},
|
|
{
|
|
"epoch": 3.850606394707828,
|
|
"grad_norm": 0.39180145490861984,
|
|
"learning_rate": 2e-05,
|
|
"loss": 0.123,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03580094873905182,
|
|
"step": 6985,
|
|
"valid_targets_mean": 2947.5,
|
|
"valid_targets_min": 2657
|
|
},
|
|
{
|
|
"epoch": 3.853362734288864,
|
|
"grad_norm": 0.3891009986488211,
|
|
"learning_rate": 1.997250969842179e-05,
|
|
"loss": 0.0943,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03961687535047531,
|
|
"step": 6990,
|
|
"valid_targets_mean": 3084.2,
|
|
"valid_targets_min": 885
|
|
},
|
|
{
|
|
"epoch": 3.856119073869901,
|
|
"grad_norm": 0.5133557530682479,
|
|
"learning_rate": 1.9945019448780798e-05,
|
|
"loss": 0.1115,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03952065482735634,
|
|
"step": 6995,
|
|
"valid_targets_mean": 3782.1,
|
|
"valid_targets_min": 2384
|
|
},
|
|
{
|
|
"epoch": 3.8588754134509373,
|
|
"grad_norm": 0.4248151932704984,
|
|
"learning_rate": 1.9917529303014148e-05,
|
|
"loss": 0.0979,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05171374976634979,
|
|
"step": 7000,
|
|
"valid_targets_mean": 3404.1,
|
|
"valid_targets_min": 1162
|
|
},
|
|
{
|
|
"epoch": 3.8616317530319737,
|
|
"grad_norm": 0.41180223754446615,
|
|
"learning_rate": 1.9890039313058784e-05,
|
|
"loss": 0.0907,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05209331959486008,
|
|
"step": 7005,
|
|
"valid_targets_mean": 5079.5,
|
|
"valid_targets_min": 3681
|
|
},
|
|
{
|
|
"epoch": 3.86438809261301,
|
|
"grad_norm": 0.476824408749734,
|
|
"learning_rate": 1.986254953085133e-05,
|
|
"loss": 0.0821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05138982832431793,
|
|
"step": 7010,
|
|
"valid_targets_mean": 2498.4,
|
|
"valid_targets_min": 592
|
|
},
|
|
{
|
|
"epoch": 3.8671444321940465,
|
|
"grad_norm": 0.27493830558361904,
|
|
"learning_rate": 1.9835060008328028e-05,
|
|
"loss": 0.0781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02843080833554268,
|
|
"step": 7015,
|
|
"valid_targets_mean": 4521.8,
|
|
"valid_targets_min": 3791
|
|
},
|
|
{
|
|
"epoch": 3.869900771775083,
|
|
"grad_norm": 0.3694039926236481,
|
|
"learning_rate": 1.9807570797424634e-05,
|
|
"loss": 0.0858,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037427064031362534,
|
|
"step": 7020,
|
|
"valid_targets_mean": 3725.8,
|
|
"valid_targets_min": 2852
|
|
},
|
|
{
|
|
"epoch": 3.8726571113561192,
|
|
"grad_norm": 0.4547453963093209,
|
|
"learning_rate": 1.97800819500763e-05,
|
|
"loss": 0.1014,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047768231481313705,
|
|
"step": 7025,
|
|
"valid_targets_mean": 1102.9,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 3.8754134509371556,
|
|
"grad_norm": 0.41341542109000223,
|
|
"learning_rate": 1.9752593518217514e-05,
|
|
"loss": 0.0886,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0456625334918499,
|
|
"step": 7030,
|
|
"valid_targets_mean": 3868.4,
|
|
"valid_targets_min": 2434
|
|
},
|
|
{
|
|
"epoch": 3.878169790518192,
|
|
"grad_norm": 0.7008658944255679,
|
|
"learning_rate": 1.972510555378196e-05,
|
|
"loss": 0.1072,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05970633402466774,
|
|
"step": 7035,
|
|
"valid_targets_mean": 1375.4,
|
|
"valid_targets_min": 780
|
|
},
|
|
{
|
|
"epoch": 3.8809261300992284,
|
|
"grad_norm": 0.3283348673179081,
|
|
"learning_rate": 1.969761810870245e-05,
|
|
"loss": 0.0725,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0396023653447628,
|
|
"step": 7040,
|
|
"valid_targets_mean": 3640.8,
|
|
"valid_targets_min": 900
|
|
},
|
|
{
|
|
"epoch": 3.8836824696802648,
|
|
"grad_norm": 0.4147904920009393,
|
|
"learning_rate": 1.967013123491081e-05,
|
|
"loss": 0.0907,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05819077789783478,
|
|
"step": 7045,
|
|
"valid_targets_mean": 2995.4,
|
|
"valid_targets_min": 653
|
|
},
|
|
{
|
|
"epoch": 3.886438809261301,
|
|
"grad_norm": 0.3990868196190449,
|
|
"learning_rate": 1.964264498433778e-05,
|
|
"loss": 0.0915,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0495319664478302,
|
|
"step": 7050,
|
|
"valid_targets_mean": 3711.2,
|
|
"valid_targets_min": 2758
|
|
},
|
|
{
|
|
"epoch": 3.8891951488423375,
|
|
"grad_norm": 0.5137492743982259,
|
|
"learning_rate": 1.9615159408912937e-05,
|
|
"loss": 0.0787,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051678065210580826,
|
|
"step": 7055,
|
|
"valid_targets_mean": 1395.0,
|
|
"valid_targets_min": 623
|
|
},
|
|
{
|
|
"epoch": 3.891951488423374,
|
|
"grad_norm": 0.39868407464960487,
|
|
"learning_rate": 1.9587674560564573e-05,
|
|
"loss": 0.0899,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04213912785053253,
|
|
"step": 7060,
|
|
"valid_targets_mean": 2325.1,
|
|
"valid_targets_min": 822
|
|
},
|
|
{
|
|
"epoch": 3.8947078280044103,
|
|
"grad_norm": 0.44927128360209534,
|
|
"learning_rate": 1.956019049121961e-05,
|
|
"loss": 0.0914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04818014055490494,
|
|
"step": 7065,
|
|
"valid_targets_mean": 2004.5,
|
|
"valid_targets_min": 935
|
|
},
|
|
{
|
|
"epoch": 3.8974641675854467,
|
|
"grad_norm": 0.39071368008641344,
|
|
"learning_rate": 1.9532707252803488e-05,
|
|
"loss": 0.1066,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04595836251974106,
|
|
"step": 7070,
|
|
"valid_targets_mean": 3336.8,
|
|
"valid_targets_min": 1660
|
|
},
|
|
{
|
|
"epoch": 3.900220507166483,
|
|
"grad_norm": 0.38888297096445135,
|
|
"learning_rate": 1.9505224897240097e-05,
|
|
"loss": 0.0877,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035211022943258286,
|
|
"step": 7075,
|
|
"valid_targets_mean": 2459.1,
|
|
"valid_targets_min": 894
|
|
},
|
|
{
|
|
"epoch": 3.9029768467475194,
|
|
"grad_norm": 0.4882239062155527,
|
|
"learning_rate": 1.9477743476451644e-05,
|
|
"loss": 0.1082,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04037652537226677,
|
|
"step": 7080,
|
|
"valid_targets_mean": 2051.8,
|
|
"valid_targets_min": 755
|
|
},
|
|
{
|
|
"epoch": 3.905733186328556,
|
|
"grad_norm": 0.6013288419610895,
|
|
"learning_rate": 1.9450263042358574e-05,
|
|
"loss": 0.103,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07325807958841324,
|
|
"step": 7085,
|
|
"valid_targets_mean": 1594.8,
|
|
"valid_targets_min": 518
|
|
},
|
|
{
|
|
"epoch": 3.908489525909592,
|
|
"grad_norm": 0.4527045463422284,
|
|
"learning_rate": 1.942278364687947e-05,
|
|
"loss": 0.0921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057140737771987915,
|
|
"step": 7090,
|
|
"valid_targets_mean": 3212.0,
|
|
"valid_targets_min": 1026
|
|
},
|
|
{
|
|
"epoch": 3.9112458654906286,
|
|
"grad_norm": 0.2920401368289559,
|
|
"learning_rate": 1.939530534193094e-05,
|
|
"loss": 0.0766,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029383936896920204,
|
|
"step": 7095,
|
|
"valid_targets_mean": 3631.5,
|
|
"valid_targets_min": 2736
|
|
},
|
|
{
|
|
"epoch": 3.914002205071665,
|
|
"grad_norm": 0.3569251892745568,
|
|
"learning_rate": 1.9367828179427553e-05,
|
|
"loss": 0.0857,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0332898311316967,
|
|
"step": 7100,
|
|
"valid_targets_mean": 2821.0,
|
|
"valid_targets_min": 1681
|
|
},
|
|
{
|
|
"epoch": 3.9167585446527013,
|
|
"grad_norm": 0.3967271487758898,
|
|
"learning_rate": 1.9340352211281707e-05,
|
|
"loss": 0.1161,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04434385895729065,
|
|
"step": 7105,
|
|
"valid_targets_mean": 3006.9,
|
|
"valid_targets_min": 929
|
|
},
|
|
{
|
|
"epoch": 3.9195148842337377,
|
|
"grad_norm": 0.3452259029753954,
|
|
"learning_rate": 1.931287748940354e-05,
|
|
"loss": 0.0735,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.025976676493883133,
|
|
"step": 7110,
|
|
"valid_targets_mean": 3051.2,
|
|
"valid_targets_min": 734
|
|
},
|
|
{
|
|
"epoch": 3.922271223814774,
|
|
"grad_norm": 0.3154986274766866,
|
|
"learning_rate": 1.9285404065700847e-05,
|
|
"loss": 0.0886,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.032151345163583755,
|
|
"step": 7115,
|
|
"valid_targets_mean": 5296.9,
|
|
"valid_targets_min": 3274
|
|
},
|
|
{
|
|
"epoch": 3.9250275633958105,
|
|
"grad_norm": 0.28577551771517906,
|
|
"learning_rate": 1.9257931992078956e-05,
|
|
"loss": 0.0846,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03602185100317001,
|
|
"step": 7120,
|
|
"valid_targets_mean": 4039.6,
|
|
"valid_targets_min": 989
|
|
},
|
|
{
|
|
"epoch": 3.927783902976847,
|
|
"grad_norm": 0.2589461288111413,
|
|
"learning_rate": 1.923046132044066e-05,
|
|
"loss": 0.0706,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.025001443922519684,
|
|
"step": 7125,
|
|
"valid_targets_mean": 4373.0,
|
|
"valid_targets_min": 944
|
|
},
|
|
{
|
|
"epoch": 3.9305402425578833,
|
|
"grad_norm": 0.4172897209875139,
|
|
"learning_rate": 1.9202992102686096e-05,
|
|
"loss": 0.0943,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05997566878795624,
|
|
"step": 7130,
|
|
"valid_targets_mean": 4134.4,
|
|
"valid_targets_min": 2713
|
|
},
|
|
{
|
|
"epoch": 3.9332965821389196,
|
|
"grad_norm": 0.32112720166999226,
|
|
"learning_rate": 1.9175524390712645e-05,
|
|
"loss": 0.0794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02970898523926735,
|
|
"step": 7135,
|
|
"valid_targets_mean": 3617.2,
|
|
"valid_targets_min": 423
|
|
},
|
|
{
|
|
"epoch": 3.936052921719956,
|
|
"grad_norm": 0.33901159500681305,
|
|
"learning_rate": 1.914805823641486e-05,
|
|
"loss": 0.109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034686192870140076,
|
|
"step": 7140,
|
|
"valid_targets_mean": 3218.5,
|
|
"valid_targets_min": 528
|
|
},
|
|
{
|
|
"epoch": 3.9388092613009924,
|
|
"grad_norm": 0.35773615404513137,
|
|
"learning_rate": 1.9120593691684335e-05,
|
|
"loss": 0.1094,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.040162838995456696,
|
|
"step": 7145,
|
|
"valid_targets_mean": 2979.4,
|
|
"valid_targets_min": 850
|
|
},
|
|
{
|
|
"epoch": 3.941565600882029,
|
|
"grad_norm": 0.3742005604519729,
|
|
"learning_rate": 1.909313080840964e-05,
|
|
"loss": 0.0687,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038524262607097626,
|
|
"step": 7150,
|
|
"valid_targets_mean": 3030.6,
|
|
"valid_targets_min": 718
|
|
},
|
|
{
|
|
"epoch": 3.944321940463065,
|
|
"grad_norm": 0.35060830510608243,
|
|
"learning_rate": 1.9065669638476195e-05,
|
|
"loss": 0.1096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035829007625579834,
|
|
"step": 7155,
|
|
"valid_targets_mean": 3710.5,
|
|
"valid_targets_min": 1245
|
|
},
|
|
{
|
|
"epoch": 3.9470782800441016,
|
|
"grad_norm": 0.37565803751758764,
|
|
"learning_rate": 1.903821023376618e-05,
|
|
"loss": 0.0837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05069202184677124,
|
|
"step": 7160,
|
|
"valid_targets_mean": 3328.5,
|
|
"valid_targets_min": 1374
|
|
},
|
|
{
|
|
"epoch": 3.949834619625138,
|
|
"grad_norm": 0.46832589760532256,
|
|
"learning_rate": 1.9010752646158447e-05,
|
|
"loss": 0.0928,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034857943654060364,
|
|
"step": 7165,
|
|
"valid_targets_mean": 1037.5,
|
|
"valid_targets_min": 592
|
|
},
|
|
{
|
|
"epoch": 3.9525909592061743,
|
|
"grad_norm": 0.5288498191879749,
|
|
"learning_rate": 1.8983296927528413e-05,
|
|
"loss": 0.1073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.061042726039886475,
|
|
"step": 7170,
|
|
"valid_targets_mean": 1320.5,
|
|
"valid_targets_min": 517
|
|
},
|
|
{
|
|
"epoch": 3.9553472987872107,
|
|
"grad_norm": 0.3919535457912462,
|
|
"learning_rate": 1.8955843129747965e-05,
|
|
"loss": 0.1232,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03698907047510147,
|
|
"step": 7175,
|
|
"valid_targets_mean": 3405.5,
|
|
"valid_targets_min": 863
|
|
},
|
|
{
|
|
"epoch": 3.958103638368247,
|
|
"grad_norm": 0.33053608622468394,
|
|
"learning_rate": 1.8928391304685363e-05,
|
|
"loss": 0.0909,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03195912390947342,
|
|
"step": 7180,
|
|
"valid_targets_mean": 3737.6,
|
|
"valid_targets_min": 2999
|
|
},
|
|
{
|
|
"epoch": 3.9608599779492835,
|
|
"grad_norm": 0.4019820841992861,
|
|
"learning_rate": 1.8900941504205133e-05,
|
|
"loss": 0.0914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04601987451314926,
|
|
"step": 7185,
|
|
"valid_targets_mean": 3086.1,
|
|
"valid_targets_min": 672
|
|
},
|
|
{
|
|
"epoch": 3.96361631753032,
|
|
"grad_norm": 0.4536837771782932,
|
|
"learning_rate": 1.8873493780167972e-05,
|
|
"loss": 0.0861,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03485861420631409,
|
|
"step": 7190,
|
|
"valid_targets_mean": 1813.5,
|
|
"valid_targets_min": 613
|
|
},
|
|
{
|
|
"epoch": 3.9663726571113562,
|
|
"grad_norm": 0.43129310518227815,
|
|
"learning_rate": 1.8846048184430677e-05,
|
|
"loss": 0.0849,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04179493337869644,
|
|
"step": 7195,
|
|
"valid_targets_mean": 2712.2,
|
|
"valid_targets_min": 538
|
|
},
|
|
{
|
|
"epoch": 3.9691289966923926,
|
|
"grad_norm": 0.6961291706922926,
|
|
"learning_rate": 1.8818604768846003e-05,
|
|
"loss": 0.0803,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054252415895462036,
|
|
"step": 7200,
|
|
"valid_targets_mean": 986.4,
|
|
"valid_targets_min": 754
|
|
},
|
|
{
|
|
"epoch": 3.971885336273429,
|
|
"grad_norm": 0.4004055536216383,
|
|
"learning_rate": 1.8791163585262588e-05,
|
|
"loss": 0.0981,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.040289733558893204,
|
|
"step": 7205,
|
|
"valid_targets_mean": 2936.8,
|
|
"valid_targets_min": 1095
|
|
},
|
|
{
|
|
"epoch": 3.9746416758544654,
|
|
"grad_norm": 0.44901378693854715,
|
|
"learning_rate": 1.876372468552486e-05,
|
|
"loss": 0.0787,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044789381325244904,
|
|
"step": 7210,
|
|
"valid_targets_mean": 3193.0,
|
|
"valid_targets_min": 734
|
|
},
|
|
{
|
|
"epoch": 3.9773980154355018,
|
|
"grad_norm": 0.37255345894505926,
|
|
"learning_rate": 1.8736288121472927e-05,
|
|
"loss": 0.0828,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04594263806939125,
|
|
"step": 7215,
|
|
"valid_targets_mean": 3266.2,
|
|
"valid_targets_min": 1000
|
|
},
|
|
{
|
|
"epoch": 3.980154355016538,
|
|
"grad_norm": 0.38965634056536264,
|
|
"learning_rate": 1.8708853944942486e-05,
|
|
"loss": 0.0956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05038759857416153,
|
|
"step": 7220,
|
|
"valid_targets_mean": 4100.1,
|
|
"valid_targets_min": 2762
|
|
},
|
|
{
|
|
"epoch": 3.9829106945975745,
|
|
"grad_norm": 0.2569824439262299,
|
|
"learning_rate": 1.868142220776473e-05,
|
|
"loss": 0.0762,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028724603354930878,
|
|
"step": 7225,
|
|
"valid_targets_mean": 5708.9,
|
|
"valid_targets_min": 3667
|
|
},
|
|
{
|
|
"epoch": 3.985667034178611,
|
|
"grad_norm": 0.4157588796200929,
|
|
"learning_rate": 1.865399296176623e-05,
|
|
"loss": 0.0815,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04685450345277786,
|
|
"step": 7230,
|
|
"valid_targets_mean": 2413.6,
|
|
"valid_targets_min": 671
|
|
},
|
|
{
|
|
"epoch": 3.9884233737596473,
|
|
"grad_norm": 0.3234512420034083,
|
|
"learning_rate": 1.862656625876886e-05,
|
|
"loss": 0.0784,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03301120176911354,
|
|
"step": 7235,
|
|
"valid_targets_mean": 3936.2,
|
|
"valid_targets_min": 790
|
|
},
|
|
{
|
|
"epoch": 3.9911797133406837,
|
|
"grad_norm": 0.49881529309384837,
|
|
"learning_rate": 1.8599142150589682e-05,
|
|
"loss": 0.1792,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08657403290271759,
|
|
"step": 7240,
|
|
"valid_targets_mean": 3082.2,
|
|
"valid_targets_min": 1054
|
|
},
|
|
{
|
|
"epoch": 3.99393605292172,
|
|
"grad_norm": 0.33501498596295887,
|
|
"learning_rate": 1.8571720689040867e-05,
|
|
"loss": 0.079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03774603456258774,
|
|
"step": 7245,
|
|
"valid_targets_mean": 3091.2,
|
|
"valid_targets_min": 980
|
|
},
|
|
{
|
|
"epoch": 3.9966923925027564,
|
|
"grad_norm": 0.39221121882944615,
|
|
"learning_rate": 1.854430192592958e-05,
|
|
"loss": 0.0899,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04876253381371498,
|
|
"step": 7250,
|
|
"valid_targets_mean": 2980.1,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 3.999448732083793,
|
|
"grad_norm": 0.5863250493398163,
|
|
"learning_rate": 1.8516885913057886e-05,
|
|
"loss": 0.1084,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11331163346767426,
|
|
"step": 7255,
|
|
"valid_targets_mean": 2233.0,
|
|
"valid_targets_min": 1272
|
|
},
|
|
{
|
|
"epoch": 4.002205071664829,
|
|
"grad_norm": 0.6078666580151728,
|
|
"learning_rate": 1.848947270222265e-05,
|
|
"loss": 0.1789,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.15226292610168457,
|
|
"step": 7260,
|
|
"valid_targets_mean": 8001.9,
|
|
"valid_targets_min": 6110
|
|
},
|
|
{
|
|
"epoch": 4.004961411245866,
|
|
"grad_norm": 0.41363689168427437,
|
|
"learning_rate": 1.8462062345215447e-05,
|
|
"loss": 0.1943,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09134787321090698,
|
|
"step": 7265,
|
|
"valid_targets_mean": 6637.6,
|
|
"valid_targets_min": 5446
|
|
},
|
|
{
|
|
"epoch": 4.0077177508269015,
|
|
"grad_norm": 0.3926998485762995,
|
|
"learning_rate": 1.8434654893822474e-05,
|
|
"loss": 0.183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09648989140987396,
|
|
"step": 7270,
|
|
"valid_targets_mean": 7098.2,
|
|
"valid_targets_min": 5340
|
|
},
|
|
{
|
|
"epoch": 4.010474090407938,
|
|
"grad_norm": 0.34945593299069794,
|
|
"learning_rate": 1.8407250399824416e-05,
|
|
"loss": 0.178,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08640488237142563,
|
|
"step": 7275,
|
|
"valid_targets_mean": 7605.0,
|
|
"valid_targets_min": 5582
|
|
},
|
|
{
|
|
"epoch": 4.013230429988974,
|
|
"grad_norm": 0.3729974197520134,
|
|
"learning_rate": 1.837984891499638e-05,
|
|
"loss": 0.1815,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08459024131298065,
|
|
"step": 7280,
|
|
"valid_targets_mean": 8599.5,
|
|
"valid_targets_min": 6330
|
|
},
|
|
{
|
|
"epoch": 4.015986769570011,
|
|
"grad_norm": 0.36920547408917076,
|
|
"learning_rate": 1.8352450491107793e-05,
|
|
"loss": 0.1856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0919535905122757,
|
|
"step": 7285,
|
|
"valid_targets_mean": 7354.9,
|
|
"valid_targets_min": 5812
|
|
},
|
|
{
|
|
"epoch": 4.018743109151047,
|
|
"grad_norm": 0.3543378089432998,
|
|
"learning_rate": 1.8325055179922285e-05,
|
|
"loss": 0.194,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.104297935962677,
|
|
"step": 7290,
|
|
"valid_targets_mean": 7989.4,
|
|
"valid_targets_min": 5251
|
|
},
|
|
{
|
|
"epoch": 4.021499448732084,
|
|
"grad_norm": 0.3520471450789033,
|
|
"learning_rate": 1.829766303319763e-05,
|
|
"loss": 0.1828,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07679878175258636,
|
|
"step": 7295,
|
|
"valid_targets_mean": 7022.4,
|
|
"valid_targets_min": 4116
|
|
},
|
|
{
|
|
"epoch": 4.02425578831312,
|
|
"grad_norm": 0.4269809390918409,
|
|
"learning_rate": 1.82702741026856e-05,
|
|
"loss": 0.1764,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07520589232444763,
|
|
"step": 7300,
|
|
"valid_targets_mean": 5345.5,
|
|
"valid_targets_min": 3897
|
|
},
|
|
{
|
|
"epoch": 4.027012127894157,
|
|
"grad_norm": 0.3847273664267357,
|
|
"learning_rate": 1.8242888440131894e-05,
|
|
"loss": 0.1635,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07981960475444794,
|
|
"step": 7305,
|
|
"valid_targets_mean": 6035.9,
|
|
"valid_targets_min": 4843
|
|
},
|
|
{
|
|
"epoch": 4.029768467475193,
|
|
"grad_norm": 0.38270201259985437,
|
|
"learning_rate": 1.8215506097276047e-05,
|
|
"loss": 0.1817,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08584163337945938,
|
|
"step": 7310,
|
|
"valid_targets_mean": 6969.2,
|
|
"valid_targets_min": 5988
|
|
},
|
|
{
|
|
"epoch": 4.032524807056229,
|
|
"grad_norm": 0.3798113248097001,
|
|
"learning_rate": 1.8188127125851315e-05,
|
|
"loss": 0.171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0853894054889679,
|
|
"step": 7315,
|
|
"valid_targets_mean": 6150.4,
|
|
"valid_targets_min": 4637
|
|
},
|
|
{
|
|
"epoch": 4.035281146637265,
|
|
"grad_norm": 0.3571042461891464,
|
|
"learning_rate": 1.8160751577584587e-05,
|
|
"loss": 0.1765,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07836826890707016,
|
|
"step": 7320,
|
|
"valid_targets_mean": 6220.1,
|
|
"valid_targets_min": 4868
|
|
},
|
|
{
|
|
"epoch": 4.038037486218302,
|
|
"grad_norm": 0.354259541234493,
|
|
"learning_rate": 1.8133379504196288e-05,
|
|
"loss": 0.1694,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07710032165050507,
|
|
"step": 7325,
|
|
"valid_targets_mean": 7126.4,
|
|
"valid_targets_min": 5291
|
|
},
|
|
{
|
|
"epoch": 4.040793825799338,
|
|
"grad_norm": 0.36679961563949465,
|
|
"learning_rate": 1.8106010957400263e-05,
|
|
"loss": 0.1721,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08878771960735321,
|
|
"step": 7330,
|
|
"valid_targets_mean": 7184.9,
|
|
"valid_targets_min": 4699
|
|
},
|
|
{
|
|
"epoch": 4.043550165380375,
|
|
"grad_norm": 0.3767369417168399,
|
|
"learning_rate": 1.807864598890371e-05,
|
|
"loss": 0.1785,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08659182488918304,
|
|
"step": 7335,
|
|
"valid_targets_mean": 6518.6,
|
|
"valid_targets_min": 5707
|
|
},
|
|
{
|
|
"epoch": 4.046306504961411,
|
|
"grad_norm": 0.36392322970891733,
|
|
"learning_rate": 1.805128465040706e-05,
|
|
"loss": 0.1702,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0877913162112236,
|
|
"step": 7340,
|
|
"valid_targets_mean": 6694.6,
|
|
"valid_targets_min": 4508
|
|
},
|
|
{
|
|
"epoch": 4.049062844542448,
|
|
"grad_norm": 0.4697192087104221,
|
|
"learning_rate": 1.8023926993603894e-05,
|
|
"loss": 0.1772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09569253027439117,
|
|
"step": 7345,
|
|
"valid_targets_mean": 4555.8,
|
|
"valid_targets_min": 1033
|
|
},
|
|
{
|
|
"epoch": 4.051819184123484,
|
|
"grad_norm": 0.4049992835009321,
|
|
"learning_rate": 1.7996573070180823e-05,
|
|
"loss": 0.1891,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09000088274478912,
|
|
"step": 7350,
|
|
"valid_targets_mean": 6398.8,
|
|
"valid_targets_min": 5380
|
|
},
|
|
{
|
|
"epoch": 4.0545755237045205,
|
|
"grad_norm": 0.6025020814014947,
|
|
"learning_rate": 1.7969222931817412e-05,
|
|
"loss": 0.1685,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02547072246670723,
|
|
"step": 7355,
|
|
"valid_targets_mean": 579.6,
|
|
"valid_targets_min": 157
|
|
},
|
|
{
|
|
"epoch": 4.057331863285556,
|
|
"grad_norm": 0.42285933024556605,
|
|
"learning_rate": 1.7941876630186078e-05,
|
|
"loss": 0.1618,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09041106700897217,
|
|
"step": 7360,
|
|
"valid_targets_mean": 8202.1,
|
|
"valid_targets_min": 5037
|
|
},
|
|
{
|
|
"epoch": 4.060088202866593,
|
|
"grad_norm": 0.3998934316258495,
|
|
"learning_rate": 1.7914534216951988e-05,
|
|
"loss": 0.1731,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0950043797492981,
|
|
"step": 7365,
|
|
"valid_targets_mean": 7208.6,
|
|
"valid_targets_min": 5089
|
|
},
|
|
{
|
|
"epoch": 4.062844542447629,
|
|
"grad_norm": 0.3866440605194138,
|
|
"learning_rate": 1.7887195743772953e-05,
|
|
"loss": 0.1725,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07695619761943817,
|
|
"step": 7370,
|
|
"valid_targets_mean": 6292.2,
|
|
"valid_targets_min": 4835
|
|
},
|
|
{
|
|
"epoch": 4.065600882028666,
|
|
"grad_norm": 0.3714754947528403,
|
|
"learning_rate": 1.785986126229936e-05,
|
|
"loss": 0.179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08767920732498169,
|
|
"step": 7375,
|
|
"valid_targets_mean": 7856.0,
|
|
"valid_targets_min": 5974
|
|
},
|
|
{
|
|
"epoch": 4.068357221609702,
|
|
"grad_norm": 0.35642691199844273,
|
|
"learning_rate": 1.783253082417403e-05,
|
|
"loss": 0.1692,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07744421064853668,
|
|
"step": 7380,
|
|
"valid_targets_mean": 7059.6,
|
|
"valid_targets_min": 4885
|
|
},
|
|
{
|
|
"epoch": 4.071113561190739,
|
|
"grad_norm": 0.4488435336560105,
|
|
"learning_rate": 1.780520448103216e-05,
|
|
"loss": 0.1637,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07951182126998901,
|
|
"step": 7385,
|
|
"valid_targets_mean": 7414.6,
|
|
"valid_targets_min": 4181
|
|
},
|
|
{
|
|
"epoch": 4.073869900771775,
|
|
"grad_norm": 0.35975353470226124,
|
|
"learning_rate": 1.7777882284501215e-05,
|
|
"loss": 0.1508,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08116020262241364,
|
|
"step": 7390,
|
|
"valid_targets_mean": 7876.2,
|
|
"valid_targets_min": 5037
|
|
},
|
|
{
|
|
"epoch": 4.0766262403528115,
|
|
"grad_norm": 0.33027142809176085,
|
|
"learning_rate": 1.7750564286200816e-05,
|
|
"loss": 0.1517,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.071902796626091,
|
|
"step": 7395,
|
|
"valid_targets_mean": 8526.2,
|
|
"valid_targets_min": 6299
|
|
},
|
|
{
|
|
"epoch": 4.0793825799338475,
|
|
"grad_norm": 0.4164459607714372,
|
|
"learning_rate": 1.772325053774265e-05,
|
|
"loss": 0.1777,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08210623264312744,
|
|
"step": 7400,
|
|
"valid_targets_mean": 4955.2,
|
|
"valid_targets_min": 2831
|
|
},
|
|
{
|
|
"epoch": 4.082138919514884,
|
|
"grad_norm": 0.4144751897928814,
|
|
"learning_rate": 1.7695941090730385e-05,
|
|
"loss": 0.1615,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0955483540892601,
|
|
"step": 7405,
|
|
"valid_targets_mean": 7099.0,
|
|
"valid_targets_min": 5303
|
|
},
|
|
{
|
|
"epoch": 4.08489525909592,
|
|
"grad_norm": 0.40730530788650404,
|
|
"learning_rate": 1.7668635996759547e-05,
|
|
"loss": 0.1811,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09250541031360626,
|
|
"step": 7410,
|
|
"valid_targets_mean": 6837.5,
|
|
"valid_targets_min": 4678
|
|
},
|
|
{
|
|
"epoch": 4.087651598676957,
|
|
"grad_norm": 0.39341986441409216,
|
|
"learning_rate": 1.7641335307417465e-05,
|
|
"loss": 0.187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08619027584791183,
|
|
"step": 7415,
|
|
"valid_targets_mean": 6603.6,
|
|
"valid_targets_min": 4604
|
|
},
|
|
{
|
|
"epoch": 4.090407938257993,
|
|
"grad_norm": 0.37161173372605755,
|
|
"learning_rate": 1.7614039074283117e-05,
|
|
"loss": 0.1793,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0809294804930687,
|
|
"step": 7420,
|
|
"valid_targets_mean": 6302.1,
|
|
"valid_targets_min": 5395
|
|
},
|
|
{
|
|
"epoch": 4.09316427783903,
|
|
"grad_norm": 0.38665118536513754,
|
|
"learning_rate": 1.7586747348927075e-05,
|
|
"loss": 0.1747,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08277405053377151,
|
|
"step": 7425,
|
|
"valid_targets_mean": 6784.9,
|
|
"valid_targets_min": 4932
|
|
},
|
|
{
|
|
"epoch": 4.095920617420066,
|
|
"grad_norm": 0.4381286957995197,
|
|
"learning_rate": 1.7559460182911396e-05,
|
|
"loss": 0.1811,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09550921618938446,
|
|
"step": 7430,
|
|
"valid_targets_mean": 6887.4,
|
|
"valid_targets_min": 5304
|
|
},
|
|
{
|
|
"epoch": 4.098676957001103,
|
|
"grad_norm": 0.377381348862524,
|
|
"learning_rate": 1.7532177627789516e-05,
|
|
"loss": 0.1735,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08894434571266174,
|
|
"step": 7435,
|
|
"valid_targets_mean": 6433.5,
|
|
"valid_targets_min": 5161
|
|
},
|
|
{
|
|
"epoch": 4.1014332965821385,
|
|
"grad_norm": 0.38003882407730316,
|
|
"learning_rate": 1.7504899735106173e-05,
|
|
"loss": 0.1773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09020879864692688,
|
|
"step": 7440,
|
|
"valid_targets_mean": 7024.6,
|
|
"valid_targets_min": 4881
|
|
},
|
|
{
|
|
"epoch": 4.104189636163175,
|
|
"grad_norm": 0.43696754562399215,
|
|
"learning_rate": 1.7477626556397283e-05,
|
|
"loss": 0.1792,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09822157025337219,
|
|
"step": 7445,
|
|
"valid_targets_mean": 7438.8,
|
|
"valid_targets_min": 5396
|
|
},
|
|
{
|
|
"epoch": 4.106945975744211,
|
|
"grad_norm": 0.4119549955440186,
|
|
"learning_rate": 1.745035814318986e-05,
|
|
"loss": 0.1799,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09425225108861923,
|
|
"step": 7450,
|
|
"valid_targets_mean": 6787.4,
|
|
"valid_targets_min": 5254
|
|
},
|
|
{
|
|
"epoch": 4.109702315325248,
|
|
"grad_norm": 0.3820297864520318,
|
|
"learning_rate": 1.7423094547001913e-05,
|
|
"loss": 0.1776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09293752908706665,
|
|
"step": 7455,
|
|
"valid_targets_mean": 7077.1,
|
|
"valid_targets_min": 5169
|
|
},
|
|
{
|
|
"epoch": 4.112458654906284,
|
|
"grad_norm": 0.40942316175263216,
|
|
"learning_rate": 1.7395835819342355e-05,
|
|
"loss": 0.1746,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0850178599357605,
|
|
"step": 7460,
|
|
"valid_targets_mean": 5610.6,
|
|
"valid_targets_min": 4806
|
|
},
|
|
{
|
|
"epoch": 4.115214994487321,
|
|
"grad_norm": 0.8240917775892426,
|
|
"learning_rate": 1.7368582011710904e-05,
|
|
"loss": 0.1443,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14348801970481873,
|
|
"step": 7465,
|
|
"valid_targets_mean": 2044.8,
|
|
"valid_targets_min": 133
|
|
},
|
|
{
|
|
"epoch": 4.117971334068357,
|
|
"grad_norm": 0.4377537370811728,
|
|
"learning_rate": 1.734133317559797e-05,
|
|
"loss": 0.1795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08171463012695312,
|
|
"step": 7470,
|
|
"valid_targets_mean": 6655.9,
|
|
"valid_targets_min": 5224
|
|
},
|
|
{
|
|
"epoch": 4.120727673649394,
|
|
"grad_norm": 0.4000910806857422,
|
|
"learning_rate": 1.731408936248458e-05,
|
|
"loss": 0.1712,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09285025298595428,
|
|
"step": 7475,
|
|
"valid_targets_mean": 6184.5,
|
|
"valid_targets_min": 5509
|
|
},
|
|
{
|
|
"epoch": 4.12348401323043,
|
|
"grad_norm": 0.3857972529866829,
|
|
"learning_rate": 1.7286850623842258e-05,
|
|
"loss": 0.1751,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0868956446647644,
|
|
"step": 7480,
|
|
"valid_targets_mean": 6454.5,
|
|
"valid_targets_min": 5499
|
|
},
|
|
{
|
|
"epoch": 4.126240352811466,
|
|
"grad_norm": 0.38140156936863817,
|
|
"learning_rate": 1.7259617011132967e-05,
|
|
"loss": 0.1713,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08643429726362228,
|
|
"step": 7485,
|
|
"valid_targets_mean": 5767.6,
|
|
"valid_targets_min": 3887
|
|
},
|
|
{
|
|
"epoch": 4.128996692392502,
|
|
"grad_norm": 0.4116263177351103,
|
|
"learning_rate": 1.7232388575808964e-05,
|
|
"loss": 0.171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08689355850219727,
|
|
"step": 7490,
|
|
"valid_targets_mean": 6327.5,
|
|
"valid_targets_min": 5163
|
|
},
|
|
{
|
|
"epoch": 4.131753031973539,
|
|
"grad_norm": 0.4163527291649443,
|
|
"learning_rate": 1.7205165369312728e-05,
|
|
"loss": 0.1799,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08879202604293823,
|
|
"step": 7495,
|
|
"valid_targets_mean": 5777.6,
|
|
"valid_targets_min": 4927
|
|
},
|
|
{
|
|
"epoch": 4.134509371554575,
|
|
"grad_norm": 0.5096899628910012,
|
|
"learning_rate": 1.7177947443076862e-05,
|
|
"loss": 0.1738,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05626387149095535,
|
|
"step": 7500,
|
|
"valid_targets_mean": 2577.2,
|
|
"valid_targets_min": 836
|
|
},
|
|
{
|
|
"epoch": 4.137265711135612,
|
|
"grad_norm": 0.4229381138467187,
|
|
"learning_rate": 1.715073484852399e-05,
|
|
"loss": 0.1455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08328432589769363,
|
|
"step": 7505,
|
|
"valid_targets_mean": 5975.9,
|
|
"valid_targets_min": 4737
|
|
},
|
|
{
|
|
"epoch": 4.140022050716648,
|
|
"grad_norm": 0.3967854988512386,
|
|
"learning_rate": 1.7123527637066673e-05,
|
|
"loss": 0.17,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08383268862962723,
|
|
"step": 7510,
|
|
"valid_targets_mean": 6735.0,
|
|
"valid_targets_min": 5014
|
|
},
|
|
{
|
|
"epoch": 4.142778390297685,
|
|
"grad_norm": 0.42283510457089896,
|
|
"learning_rate": 1.7096325860107286e-05,
|
|
"loss": 0.1771,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09102794528007507,
|
|
"step": 7515,
|
|
"valid_targets_mean": 6024.4,
|
|
"valid_targets_min": 5150
|
|
},
|
|
{
|
|
"epoch": 4.145534729878721,
|
|
"grad_norm": 0.3663436989530665,
|
|
"learning_rate": 1.7069129569037955e-05,
|
|
"loss": 0.1713,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08496168255805969,
|
|
"step": 7520,
|
|
"valid_targets_mean": 7069.6,
|
|
"valid_targets_min": 5334
|
|
},
|
|
{
|
|
"epoch": 4.1482910694597575,
|
|
"grad_norm": 0.38702199866888154,
|
|
"learning_rate": 1.704193881524042e-05,
|
|
"loss": 0.168,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08814899623394012,
|
|
"step": 7525,
|
|
"valid_targets_mean": 7081.1,
|
|
"valid_targets_min": 5160
|
|
},
|
|
{
|
|
"epoch": 4.151047409040794,
|
|
"grad_norm": 0.4228256363536986,
|
|
"learning_rate": 1.7014753650085972e-05,
|
|
"loss": 0.1716,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08567777276039124,
|
|
"step": 7530,
|
|
"valid_targets_mean": 5478.6,
|
|
"valid_targets_min": 4764
|
|
},
|
|
{
|
|
"epoch": 4.15380374862183,
|
|
"grad_norm": 0.8517965822789118,
|
|
"learning_rate": 1.6987574124935353e-05,
|
|
"loss": 0.1715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09757290780544281,
|
|
"step": 7535,
|
|
"valid_targets_mean": 1576.5,
|
|
"valid_targets_min": 860
|
|
},
|
|
{
|
|
"epoch": 4.156560088202866,
|
|
"grad_norm": 0.6879472955885122,
|
|
"learning_rate": 1.6960400291138625e-05,
|
|
"loss": 0.1732,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07717949151992798,
|
|
"step": 7540,
|
|
"valid_targets_mean": 1329.1,
|
|
"valid_targets_min": 631
|
|
},
|
|
{
|
|
"epoch": 4.159316427783903,
|
|
"grad_norm": 0.7495307829027372,
|
|
"learning_rate": 1.693323220003512e-05,
|
|
"loss": 0.1583,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08089318871498108,
|
|
"step": 7545,
|
|
"valid_targets_mean": 1597.4,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 4.16207276736494,
|
|
"grad_norm": 0.9365078461485683,
|
|
"learning_rate": 1.690606990295331e-05,
|
|
"loss": 0.1572,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08308449387550354,
|
|
"step": 7550,
|
|
"valid_targets_mean": 1337.1,
|
|
"valid_targets_min": 698
|
|
},
|
|
{
|
|
"epoch": 4.164829106945976,
|
|
"grad_norm": 0.748951041719024,
|
|
"learning_rate": 1.6878913451210715e-05,
|
|
"loss": 0.1462,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08925335854291916,
|
|
"step": 7555,
|
|
"valid_targets_mean": 1785.5,
|
|
"valid_targets_min": 897
|
|
},
|
|
{
|
|
"epoch": 4.167585446527012,
|
|
"grad_norm": 0.7459710651337684,
|
|
"learning_rate": 1.6851762896113827e-05,
|
|
"loss": 0.1566,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08002932369709015,
|
|
"step": 7560,
|
|
"valid_targets_mean": 1428.4,
|
|
"valid_targets_min": 1069
|
|
},
|
|
{
|
|
"epoch": 4.1703417861080485,
|
|
"grad_norm": 0.7440549372292389,
|
|
"learning_rate": 1.682461828895799e-05,
|
|
"loss": 0.1576,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07806475460529327,
|
|
"step": 7565,
|
|
"valid_targets_mean": 1584.1,
|
|
"valid_targets_min": 969
|
|
},
|
|
{
|
|
"epoch": 4.173098125689085,
|
|
"grad_norm": 0.7377262693241642,
|
|
"learning_rate": 1.67974796810273e-05,
|
|
"loss": 0.1388,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0678805410861969,
|
|
"step": 7570,
|
|
"valid_targets_mean": 1267.4,
|
|
"valid_targets_min": 646
|
|
},
|
|
{
|
|
"epoch": 4.175854465270121,
|
|
"grad_norm": 0.7543036213576726,
|
|
"learning_rate": 1.6770347123594527e-05,
|
|
"loss": 0.1609,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08609826862812042,
|
|
"step": 7575,
|
|
"valid_targets_mean": 1671.1,
|
|
"valid_targets_min": 726
|
|
},
|
|
{
|
|
"epoch": 4.178610804851157,
|
|
"grad_norm": 0.7209583554037061,
|
|
"learning_rate": 1.674322066792102e-05,
|
|
"loss": 0.1416,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057751867920160294,
|
|
"step": 7580,
|
|
"valid_targets_mean": 1400.9,
|
|
"valid_targets_min": 766
|
|
},
|
|
{
|
|
"epoch": 4.181367144432194,
|
|
"grad_norm": 0.7760754839249657,
|
|
"learning_rate": 1.6716100365256593e-05,
|
|
"loss": 0.1509,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06666992604732513,
|
|
"step": 7585,
|
|
"valid_targets_mean": 1376.0,
|
|
"valid_targets_min": 570
|
|
},
|
|
{
|
|
"epoch": 4.184123484013231,
|
|
"grad_norm": 0.736647936680778,
|
|
"learning_rate": 1.668898626683942e-05,
|
|
"loss": 0.1497,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08582430332899094,
|
|
"step": 7590,
|
|
"valid_targets_mean": 1573.8,
|
|
"valid_targets_min": 631
|
|
},
|
|
{
|
|
"epoch": 4.186879823594267,
|
|
"grad_norm": 0.7537541466015941,
|
|
"learning_rate": 1.6661878423895975e-05,
|
|
"loss": 0.1475,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07259409129619598,
|
|
"step": 7595,
|
|
"valid_targets_mean": 1473.2,
|
|
"valid_targets_min": 918
|
|
},
|
|
{
|
|
"epoch": 4.189636163175303,
|
|
"grad_norm": 0.800611881497601,
|
|
"learning_rate": 1.6634776887640902e-05,
|
|
"loss": 0.1388,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06648429483175278,
|
|
"step": 7600,
|
|
"valid_targets_mean": 1395.4,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 4.19239250275634,
|
|
"grad_norm": 0.7289699167532113,
|
|
"learning_rate": 1.6607681709276938e-05,
|
|
"loss": 0.1397,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0654144436120987,
|
|
"step": 7605,
|
|
"valid_targets_mean": 1582.6,
|
|
"valid_targets_min": 869
|
|
},
|
|
{
|
|
"epoch": 4.195148842337376,
|
|
"grad_norm": 0.7471516620956183,
|
|
"learning_rate": 1.6580592939994796e-05,
|
|
"loss": 0.1415,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10127180814743042,
|
|
"step": 7610,
|
|
"valid_targets_mean": 2001.6,
|
|
"valid_targets_min": 1041
|
|
},
|
|
{
|
|
"epoch": 4.197905181918412,
|
|
"grad_norm": 0.7487329153363892,
|
|
"learning_rate": 1.655351063097309e-05,
|
|
"loss": 0.147,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059166040271520615,
|
|
"step": 7615,
|
|
"valid_targets_mean": 1204.6,
|
|
"valid_targets_min": 869
|
|
},
|
|
{
|
|
"epoch": 4.200661521499449,
|
|
"grad_norm": 0.8009711881644724,
|
|
"learning_rate": 1.6526434833378232e-05,
|
|
"loss": 0.1523,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08162828534841537,
|
|
"step": 7620,
|
|
"valid_targets_mean": 1531.1,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 4.203417861080485,
|
|
"grad_norm": 0.6844988342812384,
|
|
"learning_rate": 1.649936559836431e-05,
|
|
"loss": 0.1508,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07326049357652664,
|
|
"step": 7625,
|
|
"valid_targets_mean": 1637.4,
|
|
"valid_targets_min": 905
|
|
},
|
|
{
|
|
"epoch": 4.206174200661522,
|
|
"grad_norm": 0.7934701778566795,
|
|
"learning_rate": 1.6472302977073044e-05,
|
|
"loss": 0.1458,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07691974192857742,
|
|
"step": 7630,
|
|
"valid_targets_mean": 1564.5,
|
|
"valid_targets_min": 813
|
|
},
|
|
{
|
|
"epoch": 4.208930540242558,
|
|
"grad_norm": 0.7059762559648443,
|
|
"learning_rate": 1.644524702063364e-05,
|
|
"loss": 0.1476,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06738974153995514,
|
|
"step": 7635,
|
|
"valid_targets_mean": 1715.5,
|
|
"valid_targets_min": 831
|
|
},
|
|
{
|
|
"epoch": 4.211686879823595,
|
|
"grad_norm": 0.6653905598205795,
|
|
"learning_rate": 1.6418197780162718e-05,
|
|
"loss": 0.1461,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.058646310120821,
|
|
"step": 7640,
|
|
"valid_targets_mean": 1447.2,
|
|
"valid_targets_min": 890
|
|
},
|
|
{
|
|
"epoch": 4.214443219404631,
|
|
"grad_norm": 0.8517085606956124,
|
|
"learning_rate": 1.63911553067642e-05,
|
|
"loss": 0.1441,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08643659204244614,
|
|
"step": 7645,
|
|
"valid_targets_mean": 1580.9,
|
|
"valid_targets_min": 733
|
|
},
|
|
{
|
|
"epoch": 4.2171995589856675,
|
|
"grad_norm": 0.7176976052400766,
|
|
"learning_rate": 1.6364119651529232e-05,
|
|
"loss": 0.1356,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06479166448116302,
|
|
"step": 7650,
|
|
"valid_targets_mean": 1490.0,
|
|
"valid_targets_min": 734
|
|
},
|
|
{
|
|
"epoch": 4.219955898566703,
|
|
"grad_norm": 0.751068190210644,
|
|
"learning_rate": 1.633709086553609e-05,
|
|
"loss": 0.143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07757444679737091,
|
|
"step": 7655,
|
|
"valid_targets_mean": 1729.4,
|
|
"valid_targets_min": 871
|
|
},
|
|
{
|
|
"epoch": 4.22271223814774,
|
|
"grad_norm": 0.782108677239449,
|
|
"learning_rate": 1.6310068999850045e-05,
|
|
"loss": 0.1374,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07181262969970703,
|
|
"step": 7660,
|
|
"valid_targets_mean": 1426.6,
|
|
"valid_targets_min": 804
|
|
},
|
|
{
|
|
"epoch": 4.225468577728776,
|
|
"grad_norm": 0.6993162769936405,
|
|
"learning_rate": 1.6283054105523315e-05,
|
|
"loss": 0.1479,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08327316492795944,
|
|
"step": 7665,
|
|
"valid_targets_mean": 2051.4,
|
|
"valid_targets_min": 1249
|
|
},
|
|
{
|
|
"epoch": 4.228224917309813,
|
|
"grad_norm": 0.7351042518351655,
|
|
"learning_rate": 1.6256046233594937e-05,
|
|
"loss": 0.1495,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08415644615888596,
|
|
"step": 7670,
|
|
"valid_targets_mean": 2055.9,
|
|
"valid_targets_min": 1260
|
|
},
|
|
{
|
|
"epoch": 4.230981256890849,
|
|
"grad_norm": 0.7182022029222869,
|
|
"learning_rate": 1.622904543509068e-05,
|
|
"loss": 0.1408,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05368346348404884,
|
|
"step": 7675,
|
|
"valid_targets_mean": 1033.2,
|
|
"valid_targets_min": 611
|
|
},
|
|
{
|
|
"epoch": 4.233737596471886,
|
|
"grad_norm": 0.6963073558133369,
|
|
"learning_rate": 1.620205176102296e-05,
|
|
"loss": 0.141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06982263177633286,
|
|
"step": 7680,
|
|
"valid_targets_mean": 1668.9,
|
|
"valid_targets_min": 767
|
|
},
|
|
{
|
|
"epoch": 4.236493936052922,
|
|
"grad_norm": 0.9067488282087945,
|
|
"learning_rate": 1.6175065262390724e-05,
|
|
"loss": 0.1419,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06291791796684265,
|
|
"step": 7685,
|
|
"valid_targets_mean": 1245.8,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 4.2392502756339585,
|
|
"grad_norm": 0.749282093525826,
|
|
"learning_rate": 1.614808599017936e-05,
|
|
"loss": 0.1517,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05362901836633682,
|
|
"step": 7690,
|
|
"valid_targets_mean": 1112.0,
|
|
"valid_targets_min": 799
|
|
},
|
|
{
|
|
"epoch": 4.2420066152149944,
|
|
"grad_norm": 0.6985698021966887,
|
|
"learning_rate": 1.612111399536061e-05,
|
|
"loss": 0.1382,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07821168750524521,
|
|
"step": 7695,
|
|
"valid_targets_mean": 1774.2,
|
|
"valid_targets_min": 932
|
|
},
|
|
{
|
|
"epoch": 4.244762954796031,
|
|
"grad_norm": 0.8135625500064216,
|
|
"learning_rate": 1.609414932889246e-05,
|
|
"loss": 0.1394,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06301310658454895,
|
|
"step": 7700,
|
|
"valid_targets_mean": 1109.9,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 4.247519294377067,
|
|
"grad_norm": 0.7669684781099689,
|
|
"learning_rate": 1.6067192041719065e-05,
|
|
"loss": 0.1277,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07107910513877869,
|
|
"step": 7705,
|
|
"valid_targets_mean": 1260.5,
|
|
"valid_targets_min": 813
|
|
},
|
|
{
|
|
"epoch": 4.250275633958104,
|
|
"grad_norm": 0.7747301351427225,
|
|
"learning_rate": 1.604024218477062e-05,
|
|
"loss": 0.1443,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09000284224748611,
|
|
"step": 7710,
|
|
"valid_targets_mean": 1644.2,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 4.25303197353914,
|
|
"grad_norm": 0.7497325324848468,
|
|
"learning_rate": 1.601329980896329e-05,
|
|
"loss": 0.1369,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055581364780664444,
|
|
"step": 7715,
|
|
"valid_targets_mean": 1228.2,
|
|
"valid_targets_min": 738
|
|
},
|
|
{
|
|
"epoch": 4.255788313120177,
|
|
"grad_norm": 0.9110737589212531,
|
|
"learning_rate": 1.59863649651991e-05,
|
|
"loss": 0.1395,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08607631176710129,
|
|
"step": 7720,
|
|
"valid_targets_mean": 1677.0,
|
|
"valid_targets_min": 830
|
|
},
|
|
{
|
|
"epoch": 4.258544652701213,
|
|
"grad_norm": 0.7541273332427687,
|
|
"learning_rate": 1.5959437704365866e-05,
|
|
"loss": 0.1394,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06868889927864075,
|
|
"step": 7725,
|
|
"valid_targets_mean": 1565.6,
|
|
"valid_targets_min": 1010
|
|
},
|
|
{
|
|
"epoch": 4.26130099228225,
|
|
"grad_norm": 0.722415499072246,
|
|
"learning_rate": 1.5932518077337055e-05,
|
|
"loss": 0.1387,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06205160170793533,
|
|
"step": 7730,
|
|
"valid_targets_mean": 1272.6,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 4.2640573318632855,
|
|
"grad_norm": 0.6897599708682252,
|
|
"learning_rate": 1.590560613497172e-05,
|
|
"loss": 0.1379,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04627011716365814,
|
|
"step": 7735,
|
|
"valid_targets_mean": 1143.8,
|
|
"valid_targets_min": 732
|
|
},
|
|
{
|
|
"epoch": 4.266813671444322,
|
|
"grad_norm": 0.8246147869961412,
|
|
"learning_rate": 1.587870192811439e-05,
|
|
"loss": 0.1402,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08678305894136429,
|
|
"step": 7740,
|
|
"valid_targets_mean": 1598.8,
|
|
"valid_targets_min": 951
|
|
},
|
|
{
|
|
"epoch": 4.269570011025358,
|
|
"grad_norm": 0.8064923223558,
|
|
"learning_rate": 1.5851805507594987e-05,
|
|
"loss": 0.1472,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07297685742378235,
|
|
"step": 7745,
|
|
"valid_targets_mean": 1776.8,
|
|
"valid_targets_min": 1006
|
|
},
|
|
{
|
|
"epoch": 4.272326350606395,
|
|
"grad_norm": 0.7842003975019868,
|
|
"learning_rate": 1.582491692422872e-05,
|
|
"loss": 0.1498,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08495980501174927,
|
|
"step": 7750,
|
|
"valid_targets_mean": 1722.5,
|
|
"valid_targets_min": 1140
|
|
},
|
|
{
|
|
"epoch": 4.275082690187431,
|
|
"grad_norm": 0.7738300032032556,
|
|
"learning_rate": 1.5798036228815988e-05,
|
|
"loss": 0.1364,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0637158751487732,
|
|
"step": 7755,
|
|
"valid_targets_mean": 1478.6,
|
|
"valid_targets_min": 697
|
|
},
|
|
{
|
|
"epoch": 4.277839029768468,
|
|
"grad_norm": 0.7676019024372392,
|
|
"learning_rate": 1.57711634721423e-05,
|
|
"loss": 0.1323,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.060051992535591125,
|
|
"step": 7760,
|
|
"valid_targets_mean": 1194.1,
|
|
"valid_targets_min": 676
|
|
},
|
|
{
|
|
"epoch": 4.280595369349504,
|
|
"grad_norm": 0.8118003628398803,
|
|
"learning_rate": 1.5744298704978135e-05,
|
|
"loss": 0.145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07083000242710114,
|
|
"step": 7765,
|
|
"valid_targets_mean": 1360.2,
|
|
"valid_targets_min": 686
|
|
},
|
|
{
|
|
"epoch": 4.283351708930541,
|
|
"grad_norm": 0.7835175476799578,
|
|
"learning_rate": 1.5717441978078914e-05,
|
|
"loss": 0.1374,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05070601403713226,
|
|
"step": 7770,
|
|
"valid_targets_mean": 1140.0,
|
|
"valid_targets_min": 834
|
|
},
|
|
{
|
|
"epoch": 4.286108048511577,
|
|
"grad_norm": 0.7615527068019017,
|
|
"learning_rate": 1.5690593342184854e-05,
|
|
"loss": 0.1304,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05928089842200279,
|
|
"step": 7775,
|
|
"valid_targets_mean": 1312.4,
|
|
"valid_targets_min": 828
|
|
},
|
|
{
|
|
"epoch": 4.288864388092613,
|
|
"grad_norm": 0.7228773360267736,
|
|
"learning_rate": 1.5663752848020875e-05,
|
|
"loss": 0.1317,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0728033185005188,
|
|
"step": 7780,
|
|
"valid_targets_mean": 1733.5,
|
|
"valid_targets_min": 1135
|
|
},
|
|
{
|
|
"epoch": 4.291620727673649,
|
|
"grad_norm": 0.7256053458908281,
|
|
"learning_rate": 1.5636920546296533e-05,
|
|
"loss": 0.1404,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06257686018943787,
|
|
"step": 7785,
|
|
"valid_targets_mean": 1584.9,
|
|
"valid_targets_min": 1172
|
|
},
|
|
{
|
|
"epoch": 4.294377067254686,
|
|
"grad_norm": 0.783004815133053,
|
|
"learning_rate": 1.561009648770589e-05,
|
|
"loss": 0.1403,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07007262110710144,
|
|
"step": 7790,
|
|
"valid_targets_mean": 1538.5,
|
|
"valid_targets_min": 553
|
|
},
|
|
{
|
|
"epoch": 4.297133406835722,
|
|
"grad_norm": 0.7379208696510058,
|
|
"learning_rate": 1.5583280722927437e-05,
|
|
"loss": 0.1344,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06629291921854019,
|
|
"step": 7795,
|
|
"valid_targets_mean": 1425.4,
|
|
"valid_targets_min": 842
|
|
},
|
|
{
|
|
"epoch": 4.299889746416759,
|
|
"grad_norm": 0.6696283032954135,
|
|
"learning_rate": 1.5556473302624017e-05,
|
|
"loss": 0.1334,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06403248012065887,
|
|
"step": 7800,
|
|
"valid_targets_mean": 1617.8,
|
|
"valid_targets_min": 848
|
|
},
|
|
{
|
|
"epoch": 4.302646085997795,
|
|
"grad_norm": 0.7762576532619638,
|
|
"learning_rate": 1.5529674277442674e-05,
|
|
"loss": 0.1411,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.062323540449142456,
|
|
"step": 7805,
|
|
"valid_targets_mean": 1424.4,
|
|
"valid_targets_min": 609
|
|
},
|
|
{
|
|
"epoch": 4.305402425578832,
|
|
"grad_norm": 0.8176593253392116,
|
|
"learning_rate": 1.5502883698014614e-05,
|
|
"loss": 0.1398,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05869539827108383,
|
|
"step": 7810,
|
|
"valid_targets_mean": 1278.6,
|
|
"valid_targets_min": 781
|
|
},
|
|
{
|
|
"epoch": 4.308158765159868,
|
|
"grad_norm": 0.7858321388753888,
|
|
"learning_rate": 1.547610161495508e-05,
|
|
"loss": 0.1371,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07790792733430862,
|
|
"step": 7815,
|
|
"valid_targets_mean": 1452.1,
|
|
"valid_targets_min": 925
|
|
},
|
|
{
|
|
"epoch": 4.3109151047409044,
|
|
"grad_norm": 0.6958760876283344,
|
|
"learning_rate": 1.5449328078863265e-05,
|
|
"loss": 0.1322,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05794057995080948,
|
|
"step": 7820,
|
|
"valid_targets_mean": 1393.9,
|
|
"valid_targets_min": 818
|
|
},
|
|
{
|
|
"epoch": 4.31367144432194,
|
|
"grad_norm": 0.750843630523466,
|
|
"learning_rate": 1.5422563140322208e-05,
|
|
"loss": 0.1277,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06363102048635483,
|
|
"step": 7825,
|
|
"valid_targets_mean": 1459.5,
|
|
"valid_targets_min": 669
|
|
},
|
|
{
|
|
"epoch": 4.316427783902977,
|
|
"grad_norm": 0.7500019271517154,
|
|
"learning_rate": 1.5395806849898713e-05,
|
|
"loss": 0.1396,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0658683180809021,
|
|
"step": 7830,
|
|
"valid_targets_mean": 1498.4,
|
|
"valid_targets_min": 645
|
|
},
|
|
{
|
|
"epoch": 4.319184123484013,
|
|
"grad_norm": 1.2004827619462493,
|
|
"learning_rate": 1.536905925814324e-05,
|
|
"loss": 0.138,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05503794550895691,
|
|
"step": 7835,
|
|
"valid_targets_mean": 1272.6,
|
|
"valid_targets_min": 790
|
|
},
|
|
{
|
|
"epoch": 4.32194046306505,
|
|
"grad_norm": 0.8145938312685909,
|
|
"learning_rate": 1.5342320415589816e-05,
|
|
"loss": 0.1344,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06137322634458542,
|
|
"step": 7840,
|
|
"valid_targets_mean": 1287.6,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 4.324696802646086,
|
|
"grad_norm": 0.7568918234899581,
|
|
"learning_rate": 1.531559037275594e-05,
|
|
"loss": 0.1376,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05559319257736206,
|
|
"step": 7845,
|
|
"valid_targets_mean": 1404.8,
|
|
"valid_targets_min": 891
|
|
},
|
|
{
|
|
"epoch": 4.327453142227123,
|
|
"grad_norm": 0.7758020022751839,
|
|
"learning_rate": 1.5288869180142482e-05,
|
|
"loss": 0.1418,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0776045173406601,
|
|
"step": 7850,
|
|
"valid_targets_mean": 1576.4,
|
|
"valid_targets_min": 913
|
|
},
|
|
{
|
|
"epoch": 4.330209481808159,
|
|
"grad_norm": 0.7448727763536691,
|
|
"learning_rate": 1.5262156888233596e-05,
|
|
"loss": 0.1357,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06284817308187485,
|
|
"step": 7855,
|
|
"valid_targets_mean": 1642.9,
|
|
"valid_targets_min": 635
|
|
},
|
|
{
|
|
"epoch": 4.3329658213891955,
|
|
"grad_norm": 0.7795230321459533,
|
|
"learning_rate": 1.523545354749661e-05,
|
|
"loss": 0.1353,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06956443190574646,
|
|
"step": 7860,
|
|
"valid_targets_mean": 1517.1,
|
|
"valid_targets_min": 638
|
|
},
|
|
{
|
|
"epoch": 4.335722160970231,
|
|
"grad_norm": 0.8218498749240536,
|
|
"learning_rate": 1.5208759208381952e-05,
|
|
"loss": 0.1389,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07637335360050201,
|
|
"step": 7865,
|
|
"valid_targets_mean": 1624.5,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 4.338478500551268,
|
|
"grad_norm": 0.8094392053545977,
|
|
"learning_rate": 1.5182073921323044e-05,
|
|
"loss": 0.1361,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06179517135024071,
|
|
"step": 7870,
|
|
"valid_targets_mean": 1159.5,
|
|
"valid_targets_min": 728
|
|
},
|
|
{
|
|
"epoch": 4.341234840132304,
|
|
"grad_norm": 0.8594914571568866,
|
|
"learning_rate": 1.5155397736736197e-05,
|
|
"loss": 0.1414,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08105720579624176,
|
|
"step": 7875,
|
|
"valid_targets_mean": 1609.1,
|
|
"valid_targets_min": 999
|
|
},
|
|
{
|
|
"epoch": 4.343991179713341,
|
|
"grad_norm": 0.748811590854895,
|
|
"learning_rate": 1.5128730705020528e-05,
|
|
"loss": 0.1433,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06870662420988083,
|
|
"step": 7880,
|
|
"valid_targets_mean": 1561.2,
|
|
"valid_targets_min": 1169
|
|
},
|
|
{
|
|
"epoch": 4.346747519294377,
|
|
"grad_norm": 0.8288540263710195,
|
|
"learning_rate": 1.5102072876557867e-05,
|
|
"loss": 0.1386,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0836237445473671,
|
|
"step": 7885,
|
|
"valid_targets_mean": 1603.1,
|
|
"valid_targets_min": 888
|
|
},
|
|
{
|
|
"epoch": 4.349503858875414,
|
|
"grad_norm": 0.8239508146155772,
|
|
"learning_rate": 1.5075424301712645e-05,
|
|
"loss": 0.1421,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06797022372484207,
|
|
"step": 7890,
|
|
"valid_targets_mean": 1420.6,
|
|
"valid_targets_min": 800
|
|
},
|
|
{
|
|
"epoch": 4.35226019845645,
|
|
"grad_norm": 0.7461971556699174,
|
|
"learning_rate": 1.5048785030831821e-05,
|
|
"loss": 0.1285,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06522107869386673,
|
|
"step": 7895,
|
|
"valid_targets_mean": 1457.4,
|
|
"valid_targets_min": 562
|
|
},
|
|
{
|
|
"epoch": 4.355016538037487,
|
|
"grad_norm": 0.7472270873447806,
|
|
"learning_rate": 1.5022155114244773e-05,
|
|
"loss": 0.1313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06649501621723175,
|
|
"step": 7900,
|
|
"valid_targets_mean": 1742.0,
|
|
"valid_targets_min": 649
|
|
},
|
|
{
|
|
"epoch": 4.3577728776185225,
|
|
"grad_norm": 0.7757791171076373,
|
|
"learning_rate": 1.499553460226321e-05,
|
|
"loss": 0.1382,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054501619189977646,
|
|
"step": 7905,
|
|
"valid_targets_mean": 1157.9,
|
|
"valid_targets_min": 713
|
|
},
|
|
{
|
|
"epoch": 4.360529217199559,
|
|
"grad_norm": 0.7830077279578161,
|
|
"learning_rate": 1.4968923545181055e-05,
|
|
"loss": 0.1321,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06682521104812622,
|
|
"step": 7910,
|
|
"valid_targets_mean": 1420.5,
|
|
"valid_targets_min": 743
|
|
},
|
|
{
|
|
"epoch": 4.363285556780595,
|
|
"grad_norm": 0.7228656337489826,
|
|
"learning_rate": 1.4942321993274387e-05,
|
|
"loss": 0.1284,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06008667126297951,
|
|
"step": 7915,
|
|
"valid_targets_mean": 1386.0,
|
|
"valid_targets_min": 1215
|
|
},
|
|
{
|
|
"epoch": 4.366041896361632,
|
|
"grad_norm": 0.8035411953091355,
|
|
"learning_rate": 1.4915729996801326e-05,
|
|
"loss": 0.1361,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08040818572044373,
|
|
"step": 7920,
|
|
"valid_targets_mean": 1717.0,
|
|
"valid_targets_min": 789
|
|
},
|
|
{
|
|
"epoch": 4.368798235942668,
|
|
"grad_norm": 0.7499046844942393,
|
|
"learning_rate": 1.4889147606001932e-05,
|
|
"loss": 0.1453,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06252499669790268,
|
|
"step": 7925,
|
|
"valid_targets_mean": 1362.4,
|
|
"valid_targets_min": 845
|
|
},
|
|
{
|
|
"epoch": 4.371554575523705,
|
|
"grad_norm": 0.7665367957401568,
|
|
"learning_rate": 1.4862574871098118e-05,
|
|
"loss": 0.1316,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07652462273836136,
|
|
"step": 7930,
|
|
"valid_targets_mean": 1765.2,
|
|
"valid_targets_min": 978
|
|
},
|
|
{
|
|
"epoch": 4.374310915104741,
|
|
"grad_norm": 0.7283339327842994,
|
|
"learning_rate": 1.4836011842293554e-05,
|
|
"loss": 0.1402,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07647274434566498,
|
|
"step": 7935,
|
|
"valid_targets_mean": 1773.0,
|
|
"valid_targets_min": 1092
|
|
},
|
|
{
|
|
"epoch": 4.377067254685778,
|
|
"grad_norm": 0.7200450594943597,
|
|
"learning_rate": 1.4809458569773569e-05,
|
|
"loss": 0.1197,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06358221173286438,
|
|
"step": 7940,
|
|
"valid_targets_mean": 1504.5,
|
|
"valid_targets_min": 593
|
|
},
|
|
{
|
|
"epoch": 4.379823594266814,
|
|
"grad_norm": 0.8201892499944329,
|
|
"learning_rate": 1.4782915103705076e-05,
|
|
"loss": 0.134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06480821222066879,
|
|
"step": 7945,
|
|
"valid_targets_mean": 1326.5,
|
|
"valid_targets_min": 867
|
|
},
|
|
{
|
|
"epoch": 4.38257993384785,
|
|
"grad_norm": 0.7844376952874864,
|
|
"learning_rate": 1.4756381494236443e-05,
|
|
"loss": 0.1324,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07742743194103241,
|
|
"step": 7950,
|
|
"valid_targets_mean": 1499.4,
|
|
"valid_targets_min": 913
|
|
},
|
|
{
|
|
"epoch": 4.385336273428886,
|
|
"grad_norm": 0.7154202935302402,
|
|
"learning_rate": 1.4729857791497417e-05,
|
|
"loss": 0.1347,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07443495094776154,
|
|
"step": 7955,
|
|
"valid_targets_mean": 1965.1,
|
|
"valid_targets_min": 1075
|
|
},
|
|
{
|
|
"epoch": 4.388092613009923,
|
|
"grad_norm": 0.7462003259308799,
|
|
"learning_rate": 1.4703344045599037e-05,
|
|
"loss": 0.134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07034140825271606,
|
|
"step": 7960,
|
|
"valid_targets_mean": 1512.6,
|
|
"valid_targets_min": 655
|
|
},
|
|
{
|
|
"epoch": 4.390848952590959,
|
|
"grad_norm": 0.7824103983302281,
|
|
"learning_rate": 1.467684030663353e-05,
|
|
"loss": 0.1401,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06018107384443283,
|
|
"step": 7965,
|
|
"valid_targets_mean": 1357.0,
|
|
"valid_targets_min": 1014
|
|
},
|
|
{
|
|
"epoch": 4.393605292171996,
|
|
"grad_norm": 0.7415715341604429,
|
|
"learning_rate": 1.4650346624674212e-05,
|
|
"loss": 0.1379,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07600206136703491,
|
|
"step": 7970,
|
|
"valid_targets_mean": 1618.8,
|
|
"valid_targets_min": 1039
|
|
},
|
|
{
|
|
"epoch": 4.396361631753032,
|
|
"grad_norm": 1.1213263186400895,
|
|
"learning_rate": 1.46238630497754e-05,
|
|
"loss": 0.1745,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1794266402721405,
|
|
"step": 7975,
|
|
"valid_targets_mean": 5379.1,
|
|
"valid_targets_min": 2504
|
|
},
|
|
{
|
|
"epoch": 4.399117971334069,
|
|
"grad_norm": 0.704263229807152,
|
|
"learning_rate": 1.4597389631972314e-05,
|
|
"loss": 0.2446,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12897881865501404,
|
|
"step": 7980,
|
|
"valid_targets_mean": 5547.8,
|
|
"valid_targets_min": 531
|
|
},
|
|
{
|
|
"epoch": 4.401874310915105,
|
|
"grad_norm": 0.5510021105239377,
|
|
"learning_rate": 1.4570926421280982e-05,
|
|
"loss": 0.2326,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12001478672027588,
|
|
"step": 7985,
|
|
"valid_targets_mean": 4651.6,
|
|
"valid_targets_min": 1751
|
|
},
|
|
{
|
|
"epoch": 4.404630650496141,
|
|
"grad_norm": 0.5879841380819364,
|
|
"learning_rate": 1.4544473467698165e-05,
|
|
"loss": 0.228,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12472338229417801,
|
|
"step": 7990,
|
|
"valid_targets_mean": 4022.9,
|
|
"valid_targets_min": 1199
|
|
},
|
|
{
|
|
"epoch": 4.407386990077177,
|
|
"grad_norm": 0.5025636496189418,
|
|
"learning_rate": 1.4518030821201223e-05,
|
|
"loss": 0.2163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10731356590986252,
|
|
"step": 7995,
|
|
"valid_targets_mean": 4543.9,
|
|
"valid_targets_min": 2254
|
|
},
|
|
{
|
|
"epoch": 4.410143329658214,
|
|
"grad_norm": 0.5376164773445452,
|
|
"learning_rate": 1.4491598531748053e-05,
|
|
"loss": 0.2358,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14368705451488495,
|
|
"step": 8000,
|
|
"valid_targets_mean": 5558.8,
|
|
"valid_targets_min": 2167
|
|
},
|
|
{
|
|
"epoch": 4.41289966923925,
|
|
"grad_norm": 0.4834771699921208,
|
|
"learning_rate": 1.4465176649276984e-05,
|
|
"loss": 0.2221,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12175144255161285,
|
|
"step": 8005,
|
|
"valid_targets_mean": 7014.4,
|
|
"valid_targets_min": 4490
|
|
},
|
|
{
|
|
"epoch": 4.415656008820287,
|
|
"grad_norm": 0.492330252279134,
|
|
"learning_rate": 1.443876522370668e-05,
|
|
"loss": 0.2248,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12364785373210907,
|
|
"step": 8010,
|
|
"valid_targets_mean": 5679.8,
|
|
"valid_targets_min": 1853
|
|
},
|
|
{
|
|
"epoch": 4.418412348401323,
|
|
"grad_norm": 0.5964018825652971,
|
|
"learning_rate": 1.441236430493606e-05,
|
|
"loss": 0.2208,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1205216497182846,
|
|
"step": 8015,
|
|
"valid_targets_mean": 4010.8,
|
|
"valid_targets_min": 1023
|
|
},
|
|
{
|
|
"epoch": 4.42116868798236,
|
|
"grad_norm": 0.47484659288461506,
|
|
"learning_rate": 1.4385973942844178e-05,
|
|
"loss": 0.2319,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10770849138498306,
|
|
"step": 8020,
|
|
"valid_targets_mean": 4978.2,
|
|
"valid_targets_min": 2570
|
|
},
|
|
{
|
|
"epoch": 4.423925027563396,
|
|
"grad_norm": 0.5947975357083642,
|
|
"learning_rate": 1.4359594187290151e-05,
|
|
"loss": 0.2292,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.125885009765625,
|
|
"step": 8025,
|
|
"valid_targets_mean": 3882.9,
|
|
"valid_targets_min": 801
|
|
},
|
|
{
|
|
"epoch": 4.4266813671444325,
|
|
"grad_norm": 0.44497229141309924,
|
|
"learning_rate": 1.4333225088113058e-05,
|
|
"loss": 0.2154,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11455892026424408,
|
|
"step": 8030,
|
|
"valid_targets_mean": 5685.5,
|
|
"valid_targets_min": 1593
|
|
},
|
|
{
|
|
"epoch": 4.429437706725468,
|
|
"grad_norm": 0.5592395817539539,
|
|
"learning_rate": 1.4306866695131837e-05,
|
|
"loss": 0.2283,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11288316547870636,
|
|
"step": 8035,
|
|
"valid_targets_mean": 3502.1,
|
|
"valid_targets_min": 1650
|
|
},
|
|
{
|
|
"epoch": 4.432194046306505,
|
|
"grad_norm": 0.5467807219621297,
|
|
"learning_rate": 1.4280519058145212e-05,
|
|
"loss": 0.2216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10988324135541916,
|
|
"step": 8040,
|
|
"valid_targets_mean": 4330.5,
|
|
"valid_targets_min": 1382
|
|
},
|
|
{
|
|
"epoch": 4.434950385887541,
|
|
"grad_norm": 0.5749189758886772,
|
|
"learning_rate": 1.4254182226931574e-05,
|
|
"loss": 0.2174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11767753958702087,
|
|
"step": 8045,
|
|
"valid_targets_mean": 3924.1,
|
|
"valid_targets_min": 2545
|
|
},
|
|
{
|
|
"epoch": 4.437706725468578,
|
|
"grad_norm": 0.5051229876771122,
|
|
"learning_rate": 1.4227856251248904e-05,
|
|
"loss": 0.2202,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09680397808551788,
|
|
"step": 8050,
|
|
"valid_targets_mean": 4964.9,
|
|
"valid_targets_min": 2050
|
|
},
|
|
{
|
|
"epoch": 4.440463065049614,
|
|
"grad_norm": 0.6213240295768452,
|
|
"learning_rate": 1.4201541180834673e-05,
|
|
"loss": 0.2157,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10763208568096161,
|
|
"step": 8055,
|
|
"valid_targets_mean": 4022.4,
|
|
"valid_targets_min": 1210
|
|
},
|
|
{
|
|
"epoch": 4.443219404630651,
|
|
"grad_norm": 0.5871766661692827,
|
|
"learning_rate": 1.4175237065405745e-05,
|
|
"loss": 0.2214,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12686407566070557,
|
|
"step": 8060,
|
|
"valid_targets_mean": 4088.6,
|
|
"valid_targets_min": 1062
|
|
},
|
|
{
|
|
"epoch": 4.445975744211687,
|
|
"grad_norm": 0.6398111245581221,
|
|
"learning_rate": 1.4148943954658299e-05,
|
|
"loss": 0.23,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1367061734199524,
|
|
"step": 8065,
|
|
"valid_targets_mean": 3754.2,
|
|
"valid_targets_min": 1678
|
|
},
|
|
{
|
|
"epoch": 4.448732083792724,
|
|
"grad_norm": 0.6029289384747221,
|
|
"learning_rate": 1.4122661898267706e-05,
|
|
"loss": 0.223,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11766141653060913,
|
|
"step": 8070,
|
|
"valid_targets_mean": 3659.9,
|
|
"valid_targets_min": 670
|
|
},
|
|
{
|
|
"epoch": 4.4514884233737595,
|
|
"grad_norm": 0.6160875104148361,
|
|
"learning_rate": 1.4096390945888467e-05,
|
|
"loss": 0.2193,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10212425142526627,
|
|
"step": 8075,
|
|
"valid_targets_mean": 3318.5,
|
|
"valid_targets_min": 1862
|
|
},
|
|
{
|
|
"epoch": 4.454244762954796,
|
|
"grad_norm": 0.6061609070983313,
|
|
"learning_rate": 1.407013114715409e-05,
|
|
"loss": 0.2048,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11563520133495331,
|
|
"step": 8080,
|
|
"valid_targets_mean": 3744.6,
|
|
"valid_targets_min": 1528
|
|
},
|
|
{
|
|
"epoch": 4.457001102535832,
|
|
"grad_norm": 0.5800558542531107,
|
|
"learning_rate": 1.4043882551677028e-05,
|
|
"loss": 0.2126,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0923055112361908,
|
|
"step": 8085,
|
|
"valid_targets_mean": 3768.0,
|
|
"valid_targets_min": 1050
|
|
},
|
|
{
|
|
"epoch": 4.459757442116869,
|
|
"grad_norm": 0.5934229112162975,
|
|
"learning_rate": 1.4017645209048554e-05,
|
|
"loss": 0.202,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11180223524570465,
|
|
"step": 8090,
|
|
"valid_targets_mean": 3672.9,
|
|
"valid_targets_min": 1345
|
|
},
|
|
{
|
|
"epoch": 4.462513781697905,
|
|
"grad_norm": 0.570592972371985,
|
|
"learning_rate": 1.3991419168838683e-05,
|
|
"loss": 0.2252,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10706991702318192,
|
|
"step": 8095,
|
|
"valid_targets_mean": 4454.9,
|
|
"valid_targets_min": 2227
|
|
},
|
|
{
|
|
"epoch": 4.465270121278942,
|
|
"grad_norm": 0.61215883810512,
|
|
"learning_rate": 1.396520448059608e-05,
|
|
"loss": 0.2217,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10485170781612396,
|
|
"step": 8100,
|
|
"valid_targets_mean": 3875.6,
|
|
"valid_targets_min": 2120
|
|
},
|
|
{
|
|
"epoch": 4.468026460859978,
|
|
"grad_norm": 0.6331495159824281,
|
|
"learning_rate": 1.3939001193847955e-05,
|
|
"loss": 0.2071,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10095158219337463,
|
|
"step": 8105,
|
|
"valid_targets_mean": 3256.0,
|
|
"valid_targets_min": 1599
|
|
},
|
|
{
|
|
"epoch": 4.470782800441015,
|
|
"grad_norm": 0.5968996088759305,
|
|
"learning_rate": 1.3912809358099995e-05,
|
|
"loss": 0.2122,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08421427756547928,
|
|
"step": 8110,
|
|
"valid_targets_mean": 2624.0,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 4.4735391400220506,
|
|
"grad_norm": 0.6281069681604908,
|
|
"learning_rate": 1.3886629022836238e-05,
|
|
"loss": 0.2091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13390058279037476,
|
|
"step": 8115,
|
|
"valid_targets_mean": 3453.2,
|
|
"valid_targets_min": 1262
|
|
},
|
|
{
|
|
"epoch": 4.476295479603087,
|
|
"grad_norm": 0.5643036954594765,
|
|
"learning_rate": 1.3860460237518993e-05,
|
|
"loss": 0.2072,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09387161582708359,
|
|
"step": 8120,
|
|
"valid_targets_mean": 2598.9,
|
|
"valid_targets_min": 925
|
|
},
|
|
{
|
|
"epoch": 4.479051819184123,
|
|
"grad_norm": 0.6366818608968566,
|
|
"learning_rate": 1.3834303051588757e-05,
|
|
"loss": 0.215,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10937078297138214,
|
|
"step": 8125,
|
|
"valid_targets_mean": 3400.9,
|
|
"valid_targets_min": 1193
|
|
},
|
|
{
|
|
"epoch": 4.48180815876516,
|
|
"grad_norm": 0.6156750867173805,
|
|
"learning_rate": 1.3808157514464102e-05,
|
|
"loss": 0.2054,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11005501449108124,
|
|
"step": 8130,
|
|
"valid_targets_mean": 3601.6,
|
|
"valid_targets_min": 2016
|
|
},
|
|
{
|
|
"epoch": 4.484564498346196,
|
|
"grad_norm": 0.6015715897153072,
|
|
"learning_rate": 1.3782023675541606e-05,
|
|
"loss": 0.2104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1296803206205368,
|
|
"step": 8135,
|
|
"valid_targets_mean": 4653.8,
|
|
"valid_targets_min": 2905
|
|
},
|
|
{
|
|
"epoch": 4.487320837927233,
|
|
"grad_norm": 0.6543702946107185,
|
|
"learning_rate": 1.3755901584195728e-05,
|
|
"loss": 0.2085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09186176210641861,
|
|
"step": 8140,
|
|
"valid_targets_mean": 3208.2,
|
|
"valid_targets_min": 1108
|
|
},
|
|
{
|
|
"epoch": 4.490077177508269,
|
|
"grad_norm": 0.6088099986069908,
|
|
"learning_rate": 1.372979128977875e-05,
|
|
"loss": 0.2086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10948807746171951,
|
|
"step": 8145,
|
|
"valid_targets_mean": 3715.0,
|
|
"valid_targets_min": 512
|
|
},
|
|
{
|
|
"epoch": 4.492833517089306,
|
|
"grad_norm": 0.5991495510742394,
|
|
"learning_rate": 1.3703692841620653e-05,
|
|
"loss": 0.206,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11291998624801636,
|
|
"step": 8150,
|
|
"valid_targets_mean": 4154.9,
|
|
"valid_targets_min": 2552
|
|
},
|
|
{
|
|
"epoch": 4.495589856670342,
|
|
"grad_norm": 0.604169422893374,
|
|
"learning_rate": 1.367760628902904e-05,
|
|
"loss": 0.2193,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10151398181915283,
|
|
"step": 8155,
|
|
"valid_targets_mean": 3286.0,
|
|
"valid_targets_min": 1671
|
|
},
|
|
{
|
|
"epoch": 4.498346196251378,
|
|
"grad_norm": 0.5981081302647263,
|
|
"learning_rate": 1.3651531681289048e-05,
|
|
"loss": 0.2108,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09351824223995209,
|
|
"step": 8160,
|
|
"valid_targets_mean": 3456.5,
|
|
"valid_targets_min": 1858
|
|
},
|
|
{
|
|
"epoch": 4.501102535832414,
|
|
"grad_norm": 0.5681089224519368,
|
|
"learning_rate": 1.3625469067663237e-05,
|
|
"loss": 0.2113,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11134525388479233,
|
|
"step": 8165,
|
|
"valid_targets_mean": 5265.8,
|
|
"valid_targets_min": 2489
|
|
},
|
|
{
|
|
"epoch": 4.503858875413451,
|
|
"grad_norm": 0.616406823856619,
|
|
"learning_rate": 1.3599418497391512e-05,
|
|
"loss": 0.2084,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12367556989192963,
|
|
"step": 8170,
|
|
"valid_targets_mean": 3907.9,
|
|
"valid_targets_min": 1944
|
|
},
|
|
{
|
|
"epoch": 4.506615214994487,
|
|
"grad_norm": 0.5752148835088711,
|
|
"learning_rate": 1.3573380019691019e-05,
|
|
"loss": 0.2136,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10542194545269012,
|
|
"step": 8175,
|
|
"valid_targets_mean": 4699.1,
|
|
"valid_targets_min": 1907
|
|
},
|
|
{
|
|
"epoch": 4.509371554575524,
|
|
"grad_norm": 0.5927155836709318,
|
|
"learning_rate": 1.3547353683756056e-05,
|
|
"loss": 0.2084,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10827003419399261,
|
|
"step": 8180,
|
|
"valid_targets_mean": 3441.1,
|
|
"valid_targets_min": 1761
|
|
},
|
|
{
|
|
"epoch": 4.51212789415656,
|
|
"grad_norm": 0.5988904164073714,
|
|
"learning_rate": 1.3521339538758e-05,
|
|
"loss": 0.2077,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09567597508430481,
|
|
"step": 8185,
|
|
"valid_targets_mean": 4256.9,
|
|
"valid_targets_min": 2436
|
|
},
|
|
{
|
|
"epoch": 4.514884233737597,
|
|
"grad_norm": 0.6051279999058408,
|
|
"learning_rate": 1.3495337633845175e-05,
|
|
"loss": 0.21,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10057590156793594,
|
|
"step": 8190,
|
|
"valid_targets_mean": 3822.2,
|
|
"valid_targets_min": 2475
|
|
},
|
|
{
|
|
"epoch": 4.517640573318633,
|
|
"grad_norm": 0.6138970472374256,
|
|
"learning_rate": 1.3469348018142786e-05,
|
|
"loss": 0.2096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11191126704216003,
|
|
"step": 8195,
|
|
"valid_targets_mean": 3811.5,
|
|
"valid_targets_min": 1832
|
|
},
|
|
{
|
|
"epoch": 4.5203969128996695,
|
|
"grad_norm": 0.6430743362753338,
|
|
"learning_rate": 1.3443370740752823e-05,
|
|
"loss": 0.2054,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08961810916662216,
|
|
"step": 8200,
|
|
"valid_targets_mean": 2731.0,
|
|
"valid_targets_min": 1179
|
|
},
|
|
{
|
|
"epoch": 4.523153252480705,
|
|
"grad_norm": 0.5419478996098308,
|
|
"learning_rate": 1.3417405850753968e-05,
|
|
"loss": 0.2087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1037023663520813,
|
|
"step": 8205,
|
|
"valid_targets_mean": 4628.8,
|
|
"valid_targets_min": 1859
|
|
},
|
|
{
|
|
"epoch": 4.525909592061742,
|
|
"grad_norm": 0.5799036217155045,
|
|
"learning_rate": 1.3391453397201493e-05,
|
|
"loss": 0.2088,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1174817755818367,
|
|
"step": 8210,
|
|
"valid_targets_mean": 4101.1,
|
|
"valid_targets_min": 2488
|
|
},
|
|
{
|
|
"epoch": 4.528665931642778,
|
|
"grad_norm": 0.6365928837108723,
|
|
"learning_rate": 1.3365513429127172e-05,
|
|
"loss": 0.2208,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09851337969303131,
|
|
"step": 8215,
|
|
"valid_targets_mean": 3183.1,
|
|
"valid_targets_min": 986
|
|
},
|
|
{
|
|
"epoch": 4.531422271223815,
|
|
"grad_norm": 0.637468934853173,
|
|
"learning_rate": 1.33395859955392e-05,
|
|
"loss": 0.2061,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10665715485811234,
|
|
"step": 8220,
|
|
"valid_targets_mean": 3136.1,
|
|
"valid_targets_min": 1361
|
|
},
|
|
{
|
|
"epoch": 4.534178610804851,
|
|
"grad_norm": 0.5711813500649909,
|
|
"learning_rate": 1.3313671145422077e-05,
|
|
"loss": 0.2099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0966787338256836,
|
|
"step": 8225,
|
|
"valid_targets_mean": 4275.5,
|
|
"valid_targets_min": 2250
|
|
},
|
|
{
|
|
"epoch": 4.536934950385888,
|
|
"grad_norm": 0.7283571878786649,
|
|
"learning_rate": 1.328776892773655e-05,
|
|
"loss": 0.2139,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10378735512495041,
|
|
"step": 8230,
|
|
"valid_targets_mean": 3123.0,
|
|
"valid_targets_min": 1436
|
|
},
|
|
{
|
|
"epoch": 4.539691289966924,
|
|
"grad_norm": 0.5726056921082888,
|
|
"learning_rate": 1.3261879391419478e-05,
|
|
"loss": 0.2088,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0908041000366211,
|
|
"step": 8235,
|
|
"valid_targets_mean": 3708.4,
|
|
"valid_targets_min": 2191
|
|
},
|
|
{
|
|
"epoch": 4.5424476295479606,
|
|
"grad_norm": 0.617000263732811,
|
|
"learning_rate": 1.3236002585383774e-05,
|
|
"loss": 0.2028,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10455415397882462,
|
|
"step": 8240,
|
|
"valid_targets_mean": 4332.6,
|
|
"valid_targets_min": 1271
|
|
},
|
|
{
|
|
"epoch": 4.5452039691289965,
|
|
"grad_norm": 0.5984074663196004,
|
|
"learning_rate": 1.3210138558518295e-05,
|
|
"loss": 0.2052,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1064390242099762,
|
|
"step": 8245,
|
|
"valid_targets_mean": 3416.9,
|
|
"valid_targets_min": 1318
|
|
},
|
|
{
|
|
"epoch": 4.547960308710033,
|
|
"grad_norm": 0.5281222845701417,
|
|
"learning_rate": 1.3184287359687751e-05,
|
|
"loss": 0.1936,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09423847496509552,
|
|
"step": 8250,
|
|
"valid_targets_mean": 3720.1,
|
|
"valid_targets_min": 1192
|
|
},
|
|
{
|
|
"epoch": 4.550716648291069,
|
|
"grad_norm": 0.6157455596258554,
|
|
"learning_rate": 1.3158449037732628e-05,
|
|
"loss": 0.2241,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0916229635477066,
|
|
"step": 8255,
|
|
"valid_targets_mean": 2578.4,
|
|
"valid_targets_min": 1451
|
|
},
|
|
{
|
|
"epoch": 4.553472987872106,
|
|
"grad_norm": 0.5268815200098639,
|
|
"learning_rate": 1.3132623641469074e-05,
|
|
"loss": 0.2018,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0854460597038269,
|
|
"step": 8260,
|
|
"valid_targets_mean": 5275.6,
|
|
"valid_targets_min": 3136
|
|
},
|
|
{
|
|
"epoch": 4.556229327453142,
|
|
"grad_norm": 0.6169049146108829,
|
|
"learning_rate": 1.3106811219688815e-05,
|
|
"loss": 0.2166,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09028385579586029,
|
|
"step": 8265,
|
|
"valid_targets_mean": 3110.8,
|
|
"valid_targets_min": 1254
|
|
},
|
|
{
|
|
"epoch": 4.558985667034179,
|
|
"grad_norm": 0.6231229245208012,
|
|
"learning_rate": 1.3081011821159075e-05,
|
|
"loss": 0.2139,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13293887674808502,
|
|
"step": 8270,
|
|
"valid_targets_mean": 4470.8,
|
|
"valid_targets_min": 1607
|
|
},
|
|
{
|
|
"epoch": 4.561742006615215,
|
|
"grad_norm": 0.6163712296504777,
|
|
"learning_rate": 1.3055225494622459e-05,
|
|
"loss": 0.2181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.106694296002388,
|
|
"step": 8275,
|
|
"valid_targets_mean": 3765.4,
|
|
"valid_targets_min": 1609
|
|
},
|
|
{
|
|
"epoch": 4.564498346196252,
|
|
"grad_norm": 0.6012911815078984,
|
|
"learning_rate": 1.3029452288796886e-05,
|
|
"loss": 0.2081,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1174580305814743,
|
|
"step": 8280,
|
|
"valid_targets_mean": 4025.1,
|
|
"valid_targets_min": 2192
|
|
},
|
|
{
|
|
"epoch": 4.5672546857772875,
|
|
"grad_norm": 0.6220742620441545,
|
|
"learning_rate": 1.3003692252375486e-05,
|
|
"loss": 0.2044,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12679538130760193,
|
|
"step": 8285,
|
|
"valid_targets_mean": 4012.1,
|
|
"valid_targets_min": 2574
|
|
},
|
|
{
|
|
"epoch": 4.570011025358324,
|
|
"grad_norm": 0.6266258893640075,
|
|
"learning_rate": 1.2977945434026504e-05,
|
|
"loss": 0.2002,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10828138887882233,
|
|
"step": 8290,
|
|
"valid_targets_mean": 3416.1,
|
|
"valid_targets_min": 2310
|
|
},
|
|
{
|
|
"epoch": 4.57276736493936,
|
|
"grad_norm": 0.6120341408604333,
|
|
"learning_rate": 1.2952211882393211e-05,
|
|
"loss": 0.1993,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10508313030004501,
|
|
"step": 8295,
|
|
"valid_targets_mean": 3984.4,
|
|
"valid_targets_min": 2336
|
|
},
|
|
{
|
|
"epoch": 4.575523704520397,
|
|
"grad_norm": 0.6380600070068723,
|
|
"learning_rate": 1.292649164609381e-05,
|
|
"loss": 0.193,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08658162504434586,
|
|
"step": 8300,
|
|
"valid_targets_mean": 2859.2,
|
|
"valid_targets_min": 1024
|
|
},
|
|
{
|
|
"epoch": 4.578280044101433,
|
|
"grad_norm": 0.6482760289281719,
|
|
"learning_rate": 1.2900784773721365e-05,
|
|
"loss": 0.1965,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10769455879926682,
|
|
"step": 8305,
|
|
"valid_targets_mean": 4017.2,
|
|
"valid_targets_min": 1039
|
|
},
|
|
{
|
|
"epoch": 4.58103638368247,
|
|
"grad_norm": 0.5933483790818476,
|
|
"learning_rate": 1.2875091313843674e-05,
|
|
"loss": 0.1839,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0936768651008606,
|
|
"step": 8310,
|
|
"valid_targets_mean": 3422.9,
|
|
"valid_targets_min": 1273
|
|
},
|
|
{
|
|
"epoch": 4.583792723263506,
|
|
"grad_norm": 0.6319059926195504,
|
|
"learning_rate": 1.2849411315003199e-05,
|
|
"loss": 0.2013,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10939992219209671,
|
|
"step": 8315,
|
|
"valid_targets_mean": 4161.4,
|
|
"valid_targets_min": 913
|
|
},
|
|
{
|
|
"epoch": 4.586549062844543,
|
|
"grad_norm": 0.6267384925461394,
|
|
"learning_rate": 1.2823744825716974e-05,
|
|
"loss": 0.2249,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08714116364717484,
|
|
"step": 8320,
|
|
"valid_targets_mean": 3919.9,
|
|
"valid_targets_min": 2090
|
|
},
|
|
{
|
|
"epoch": 4.589305402425579,
|
|
"grad_norm": 0.6294080533315118,
|
|
"learning_rate": 1.2798091894476503e-05,
|
|
"loss": 0.2007,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11950120329856873,
|
|
"step": 8325,
|
|
"valid_targets_mean": 3876.9,
|
|
"valid_targets_min": 1682
|
|
},
|
|
{
|
|
"epoch": 4.592061742006615,
|
|
"grad_norm": 0.5768623870569041,
|
|
"learning_rate": 1.2772452569747685e-05,
|
|
"loss": 0.2071,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10465048998594284,
|
|
"step": 8330,
|
|
"valid_targets_mean": 4459.4,
|
|
"valid_targets_min": 2246
|
|
},
|
|
{
|
|
"epoch": 4.594818081587651,
|
|
"grad_norm": 0.5965225451634111,
|
|
"learning_rate": 1.2746826899970706e-05,
|
|
"loss": 0.1998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09709534049034119,
|
|
"step": 8335,
|
|
"valid_targets_mean": 3922.6,
|
|
"valid_targets_min": 1782
|
|
},
|
|
{
|
|
"epoch": 4.597574421168688,
|
|
"grad_norm": 0.6215850780326455,
|
|
"learning_rate": 1.2721214933559947e-05,
|
|
"loss": 0.2065,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10847589373588562,
|
|
"step": 8340,
|
|
"valid_targets_mean": 4168.2,
|
|
"valid_targets_min": 1583
|
|
},
|
|
{
|
|
"epoch": 4.600330760749724,
|
|
"grad_norm": 0.6643485152364759,
|
|
"learning_rate": 1.269561671890391e-05,
|
|
"loss": 0.2047,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11964936554431915,
|
|
"step": 8345,
|
|
"valid_targets_mean": 3739.5,
|
|
"valid_targets_min": 2032
|
|
},
|
|
{
|
|
"epoch": 4.603087100330761,
|
|
"grad_norm": 0.5568027020480141,
|
|
"learning_rate": 1.2670032304365116e-05,
|
|
"loss": 0.1918,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09720966964960098,
|
|
"step": 8350,
|
|
"valid_targets_mean": 3939.4,
|
|
"valid_targets_min": 1616
|
|
},
|
|
{
|
|
"epoch": 4.605843439911797,
|
|
"grad_norm": 0.6267359475808878,
|
|
"learning_rate": 1.264446173828001e-05,
|
|
"loss": 0.2078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08924372494220734,
|
|
"step": 8355,
|
|
"valid_targets_mean": 3527.8,
|
|
"valid_targets_min": 2149
|
|
},
|
|
{
|
|
"epoch": 4.608599779492834,
|
|
"grad_norm": 0.6256599142818978,
|
|
"learning_rate": 1.2618905068958867e-05,
|
|
"loss": 0.1967,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10957935452461243,
|
|
"step": 8360,
|
|
"valid_targets_mean": 3629.2,
|
|
"valid_targets_min": 1743
|
|
},
|
|
{
|
|
"epoch": 4.61135611907387,
|
|
"grad_norm": 0.6886745867596549,
|
|
"learning_rate": 1.2593362344685717e-05,
|
|
"loss": 0.2038,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12486985325813293,
|
|
"step": 8365,
|
|
"valid_targets_mean": 3774.0,
|
|
"valid_targets_min": 2117
|
|
},
|
|
{
|
|
"epoch": 4.6141124586549065,
|
|
"grad_norm": 0.6791129900330095,
|
|
"learning_rate": 1.2567833613718237e-05,
|
|
"loss": 0.2158,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11445054411888123,
|
|
"step": 8370,
|
|
"valid_targets_mean": 3750.4,
|
|
"valid_targets_min": 1829
|
|
},
|
|
{
|
|
"epoch": 4.616868798235942,
|
|
"grad_norm": 0.6530636261998571,
|
|
"learning_rate": 1.2542318924287678e-05,
|
|
"loss": 0.2069,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11921622604131699,
|
|
"step": 8375,
|
|
"valid_targets_mean": 4077.9,
|
|
"valid_targets_min": 1570
|
|
},
|
|
{
|
|
"epoch": 4.619625137816979,
|
|
"grad_norm": 0.6813207783258821,
|
|
"learning_rate": 1.2516818324598742e-05,
|
|
"loss": 0.2136,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11030630767345428,
|
|
"step": 8380,
|
|
"valid_targets_mean": 3366.9,
|
|
"valid_targets_min": 1484
|
|
},
|
|
{
|
|
"epoch": 4.622381477398015,
|
|
"grad_norm": 0.6259027814055962,
|
|
"learning_rate": 1.2491331862829532e-05,
|
|
"loss": 0.2058,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08385582268238068,
|
|
"step": 8385,
|
|
"valid_targets_mean": 3818.2,
|
|
"valid_targets_min": 1967
|
|
},
|
|
{
|
|
"epoch": 4.625137816979052,
|
|
"grad_norm": 0.6286665648266219,
|
|
"learning_rate": 1.2465859587131428e-05,
|
|
"loss": 0.2188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10009288787841797,
|
|
"step": 8390,
|
|
"valid_targets_mean": 3949.2,
|
|
"valid_targets_min": 2480
|
|
},
|
|
{
|
|
"epoch": 4.627894156560088,
|
|
"grad_norm": 0.620138821751272,
|
|
"learning_rate": 1.2440401545629009e-05,
|
|
"loss": 0.2091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09128134697675705,
|
|
"step": 8395,
|
|
"valid_targets_mean": 3551.9,
|
|
"valid_targets_min": 1672
|
|
},
|
|
{
|
|
"epoch": 4.630650496141125,
|
|
"grad_norm": 0.627040258968036,
|
|
"learning_rate": 1.2414957786419967e-05,
|
|
"loss": 0.2144,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09909596294164658,
|
|
"step": 8400,
|
|
"valid_targets_mean": 3459.5,
|
|
"valid_targets_min": 2428
|
|
},
|
|
{
|
|
"epoch": 4.633406835722161,
|
|
"grad_norm": 0.6564546676584087,
|
|
"learning_rate": 1.2389528357575013e-05,
|
|
"loss": 0.2085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10428648442029953,
|
|
"step": 8405,
|
|
"valid_targets_mean": 3590.5,
|
|
"valid_targets_min": 1965
|
|
},
|
|
{
|
|
"epoch": 4.6361631753031975,
|
|
"grad_norm": 0.6247360380875417,
|
|
"learning_rate": 1.2364113307137773e-05,
|
|
"loss": 0.2041,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09185223281383514,
|
|
"step": 8410,
|
|
"valid_targets_mean": 3412.5,
|
|
"valid_targets_min": 1004
|
|
},
|
|
{
|
|
"epoch": 4.6389195148842335,
|
|
"grad_norm": 0.6454365979861197,
|
|
"learning_rate": 1.2338712683124717e-05,
|
|
"loss": 0.2088,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08531850576400757,
|
|
"step": 8415,
|
|
"valid_targets_mean": 3219.1,
|
|
"valid_targets_min": 1200
|
|
},
|
|
{
|
|
"epoch": 4.64167585446527,
|
|
"grad_norm": 0.7500049571207106,
|
|
"learning_rate": 1.231332653352505e-05,
|
|
"loss": 0.2087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11344636976718903,
|
|
"step": 8420,
|
|
"valid_targets_mean": 4130.9,
|
|
"valid_targets_min": 1256
|
|
},
|
|
{
|
|
"epoch": 4.644432194046306,
|
|
"grad_norm": 0.6306922233636642,
|
|
"learning_rate": 1.2287954906300638e-05,
|
|
"loss": 0.204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10219918936491013,
|
|
"step": 8425,
|
|
"valid_targets_mean": 3062.2,
|
|
"valid_targets_min": 1355
|
|
},
|
|
{
|
|
"epoch": 4.647188533627343,
|
|
"grad_norm": 0.5477671423425797,
|
|
"learning_rate": 1.2262597849385908e-05,
|
|
"loss": 0.1916,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08368903398513794,
|
|
"step": 8430,
|
|
"valid_targets_mean": 4574.8,
|
|
"valid_targets_min": 1409
|
|
},
|
|
{
|
|
"epoch": 4.649944873208379,
|
|
"grad_norm": 0.6268346554009789,
|
|
"learning_rate": 1.2237255410687766e-05,
|
|
"loss": 0.2068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09109315276145935,
|
|
"step": 8435,
|
|
"valid_targets_mean": 3433.1,
|
|
"valid_targets_min": 1389
|
|
},
|
|
{
|
|
"epoch": 4.652701212789416,
|
|
"grad_norm": 0.6196115441010493,
|
|
"learning_rate": 1.2211927638085487e-05,
|
|
"loss": 0.1947,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08230583369731903,
|
|
"step": 8440,
|
|
"valid_targets_mean": 3033.2,
|
|
"valid_targets_min": 788
|
|
},
|
|
{
|
|
"epoch": 4.655457552370452,
|
|
"grad_norm": 0.6059350193229949,
|
|
"learning_rate": 1.218661457943065e-05,
|
|
"loss": 0.2098,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10016530752182007,
|
|
"step": 8445,
|
|
"valid_targets_mean": 3689.4,
|
|
"valid_targets_min": 1900
|
|
},
|
|
{
|
|
"epoch": 4.658213891951489,
|
|
"grad_norm": 0.6767841741877972,
|
|
"learning_rate": 1.2161316282547024e-05,
|
|
"loss": 0.2052,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09977364540100098,
|
|
"step": 8450,
|
|
"valid_targets_mean": 2805.4,
|
|
"valid_targets_min": 1457
|
|
},
|
|
{
|
|
"epoch": 4.6609702315325245,
|
|
"grad_norm": 0.5654171615547187,
|
|
"learning_rate": 1.2136032795230492e-05,
|
|
"loss": 0.205,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11652840673923492,
|
|
"step": 8455,
|
|
"valid_targets_mean": 5405.9,
|
|
"valid_targets_min": 2893
|
|
},
|
|
{
|
|
"epoch": 4.663726571113561,
|
|
"grad_norm": 0.6968846992447764,
|
|
"learning_rate": 1.211076416524897e-05,
|
|
"loss": 0.203,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11752282083034515,
|
|
"step": 8460,
|
|
"valid_targets_mean": 3204.4,
|
|
"valid_targets_min": 1760
|
|
},
|
|
{
|
|
"epoch": 4.666482910694597,
|
|
"grad_norm": 0.7134207612121916,
|
|
"learning_rate": 1.2085510440342282e-05,
|
|
"loss": 0.2029,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0932270735502243,
|
|
"step": 8465,
|
|
"valid_targets_mean": 4082.4,
|
|
"valid_targets_min": 1929
|
|
},
|
|
{
|
|
"epoch": 4.669239250275634,
|
|
"grad_norm": 0.5210744352148385,
|
|
"learning_rate": 1.2060271668222115e-05,
|
|
"loss": 0.1999,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08791635185480118,
|
|
"step": 8470,
|
|
"valid_targets_mean": 5744.4,
|
|
"valid_targets_min": 1234
|
|
},
|
|
{
|
|
"epoch": 4.67199558985667,
|
|
"grad_norm": 0.5373974347952534,
|
|
"learning_rate": 1.2035047896571884e-05,
|
|
"loss": 0.1983,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09555687755346298,
|
|
"step": 8475,
|
|
"valid_targets_mean": 5335.2,
|
|
"valid_targets_min": 1739
|
|
},
|
|
{
|
|
"epoch": 4.674751929437707,
|
|
"grad_norm": 0.5864785629748774,
|
|
"learning_rate": 1.2009839173046673e-05,
|
|
"loss": 0.1967,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10215507447719574,
|
|
"step": 8480,
|
|
"valid_targets_mean": 3754.6,
|
|
"valid_targets_min": 855
|
|
},
|
|
{
|
|
"epoch": 4.677508269018743,
|
|
"grad_norm": 0.5802857558586093,
|
|
"learning_rate": 1.1984645545273145e-05,
|
|
"loss": 0.1974,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10472627729177475,
|
|
"step": 8485,
|
|
"valid_targets_mean": 3716.0,
|
|
"valid_targets_min": 1060
|
|
},
|
|
{
|
|
"epoch": 4.68026460859978,
|
|
"grad_norm": 0.5773967042527147,
|
|
"learning_rate": 1.1959467060849432e-05,
|
|
"loss": 0.1924,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09129063785076141,
|
|
"step": 8490,
|
|
"valid_targets_mean": 3903.5,
|
|
"valid_targets_min": 974
|
|
},
|
|
{
|
|
"epoch": 4.683020948180816,
|
|
"grad_norm": 0.6461443987269605,
|
|
"learning_rate": 1.1934303767345057e-05,
|
|
"loss": 0.1895,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09532202780246735,
|
|
"step": 8495,
|
|
"valid_targets_mean": 3001.8,
|
|
"valid_targets_min": 1437
|
|
},
|
|
{
|
|
"epoch": 4.685777287761852,
|
|
"grad_norm": 0.6549145611722482,
|
|
"learning_rate": 1.1909155712300844e-05,
|
|
"loss": 0.1912,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11181293427944183,
|
|
"step": 8500,
|
|
"valid_targets_mean": 4849.8,
|
|
"valid_targets_min": 3610
|
|
},
|
|
{
|
|
"epoch": 4.688533627342888,
|
|
"grad_norm": 0.6355067780533544,
|
|
"learning_rate": 1.1884022943228824e-05,
|
|
"loss": 0.1936,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10432341694831848,
|
|
"step": 8505,
|
|
"valid_targets_mean": 4029.4,
|
|
"valid_targets_min": 1515
|
|
},
|
|
{
|
|
"epoch": 4.691289966923925,
|
|
"grad_norm": 0.7138285481884677,
|
|
"learning_rate": 1.1858905507612156e-05,
|
|
"loss": 0.2057,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10473667830228806,
|
|
"step": 8510,
|
|
"valid_targets_mean": 2771.5,
|
|
"valid_targets_min": 1347
|
|
},
|
|
{
|
|
"epoch": 4.694046306504961,
|
|
"grad_norm": 0.627499721173437,
|
|
"learning_rate": 1.1833803452905025e-05,
|
|
"loss": 0.2059,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10486799478530884,
|
|
"step": 8515,
|
|
"valid_targets_mean": 3195.0,
|
|
"valid_targets_min": 1904
|
|
},
|
|
{
|
|
"epoch": 4.696802646085998,
|
|
"grad_norm": 0.6495484946992698,
|
|
"learning_rate": 1.1808716826532556e-05,
|
|
"loss": 0.192,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10146039724349976,
|
|
"step": 8520,
|
|
"valid_targets_mean": 3454.2,
|
|
"valid_targets_min": 2070
|
|
},
|
|
{
|
|
"epoch": 4.699558985667034,
|
|
"grad_norm": 0.633321627665926,
|
|
"learning_rate": 1.1783645675890725e-05,
|
|
"loss": 0.1897,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08290262520313263,
|
|
"step": 8525,
|
|
"valid_targets_mean": 3065.8,
|
|
"valid_targets_min": 1019
|
|
},
|
|
{
|
|
"epoch": 4.702315325248071,
|
|
"grad_norm": 0.6948000846926881,
|
|
"learning_rate": 1.175859004834627e-05,
|
|
"loss": 0.198,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09420207887887955,
|
|
"step": 8530,
|
|
"valid_targets_mean": 3082.1,
|
|
"valid_targets_min": 653
|
|
},
|
|
{
|
|
"epoch": 4.705071664829107,
|
|
"grad_norm": 0.6148716439278388,
|
|
"learning_rate": 1.1733549991236608e-05,
|
|
"loss": 0.1998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10894359648227692,
|
|
"step": 8535,
|
|
"valid_targets_mean": 4172.1,
|
|
"valid_targets_min": 1604
|
|
},
|
|
{
|
|
"epoch": 4.7078280044101435,
|
|
"grad_norm": 0.6870952931196683,
|
|
"learning_rate": 1.1708525551869729e-05,
|
|
"loss": 0.2085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10620259493589401,
|
|
"step": 8540,
|
|
"valid_targets_mean": 2731.9,
|
|
"valid_targets_min": 815
|
|
},
|
|
{
|
|
"epoch": 4.710584343991179,
|
|
"grad_norm": 0.6211050181527924,
|
|
"learning_rate": 1.1683516777524127e-05,
|
|
"loss": 0.1999,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08935414254665375,
|
|
"step": 8545,
|
|
"valid_targets_mean": 3379.6,
|
|
"valid_targets_min": 1874
|
|
},
|
|
{
|
|
"epoch": 4.713340683572216,
|
|
"grad_norm": 0.6032244357255013,
|
|
"learning_rate": 1.1658523715448683e-05,
|
|
"loss": 0.2008,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10111667960882187,
|
|
"step": 8550,
|
|
"valid_targets_mean": 3883.6,
|
|
"valid_targets_min": 1787
|
|
},
|
|
{
|
|
"epoch": 4.716097023153252,
|
|
"grad_norm": 0.6114987825911102,
|
|
"learning_rate": 1.1633546412862612e-05,
|
|
"loss": 0.2063,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0870811939239502,
|
|
"step": 8555,
|
|
"valid_targets_mean": 3233.4,
|
|
"valid_targets_min": 1214
|
|
},
|
|
{
|
|
"epoch": 4.718853362734289,
|
|
"grad_norm": 0.6383127688633033,
|
|
"learning_rate": 1.1608584916955348e-05,
|
|
"loss": 0.1908,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07421775907278061,
|
|
"step": 8560,
|
|
"valid_targets_mean": 1866.6,
|
|
"valid_targets_min": 631
|
|
},
|
|
{
|
|
"epoch": 4.721609702315325,
|
|
"grad_norm": 0.39638803697068736,
|
|
"learning_rate": 1.1583639274886461e-05,
|
|
"loss": 0.1014,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03152263164520264,
|
|
"step": 8565,
|
|
"valid_targets_mean": 2847.2,
|
|
"valid_targets_min": 829
|
|
},
|
|
{
|
|
"epoch": 4.724366041896362,
|
|
"grad_norm": 0.3752784991518403,
|
|
"learning_rate": 1.155870953378556e-05,
|
|
"loss": 0.0803,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03577549010515213,
|
|
"step": 8570,
|
|
"valid_targets_mean": 3424.2,
|
|
"valid_targets_min": 2780
|
|
},
|
|
{
|
|
"epoch": 4.727122381477398,
|
|
"grad_norm": 0.38053521639452637,
|
|
"learning_rate": 1.1533795740752228e-05,
|
|
"loss": 0.0781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04130037873983383,
|
|
"step": 8575,
|
|
"valid_targets_mean": 3527.0,
|
|
"valid_targets_min": 2121
|
|
},
|
|
{
|
|
"epoch": 4.7298787210584345,
|
|
"grad_norm": 0.400430712548472,
|
|
"learning_rate": 1.1508897942855902e-05,
|
|
"loss": 0.0948,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035319894552230835,
|
|
"step": 8580,
|
|
"valid_targets_mean": 2447.4,
|
|
"valid_targets_min": 523
|
|
},
|
|
{
|
|
"epoch": 4.7326350606394705,
|
|
"grad_norm": 0.40273298795849327,
|
|
"learning_rate": 1.1484016187135815e-05,
|
|
"loss": 0.0782,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04246003180742264,
|
|
"step": 8585,
|
|
"valid_targets_mean": 2761.6,
|
|
"valid_targets_min": 885
|
|
},
|
|
{
|
|
"epoch": 4.735391400220507,
|
|
"grad_norm": 1.0719742110181378,
|
|
"learning_rate": 1.1459150520600873e-05,
|
|
"loss": 0.1037,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08043426275253296,
|
|
"step": 8590,
|
|
"valid_targets_mean": 1135.5,
|
|
"valid_targets_min": 556
|
|
},
|
|
{
|
|
"epoch": 4.738147739801543,
|
|
"grad_norm": 0.43739208368429006,
|
|
"learning_rate": 1.143430099022961e-05,
|
|
"loss": 0.116,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031099801883101463,
|
|
"step": 8595,
|
|
"valid_targets_mean": 2674.5,
|
|
"valid_targets_min": 643
|
|
},
|
|
{
|
|
"epoch": 4.74090407938258,
|
|
"grad_norm": 0.46381278008641624,
|
|
"learning_rate": 1.1409467642970053e-05,
|
|
"loss": 0.0865,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03925149887800217,
|
|
"step": 8600,
|
|
"valid_targets_mean": 2996.8,
|
|
"valid_targets_min": 942
|
|
},
|
|
{
|
|
"epoch": 4.743660418963616,
|
|
"grad_norm": 0.5309216912795555,
|
|
"learning_rate": 1.1384650525739662e-05,
|
|
"loss": 0.1103,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09209037572145462,
|
|
"step": 8605,
|
|
"valid_targets_mean": 1892.5,
|
|
"valid_targets_min": 807
|
|
},
|
|
{
|
|
"epoch": 4.746416758544653,
|
|
"grad_norm": 0.30708375191092246,
|
|
"learning_rate": 1.1359849685425232e-05,
|
|
"loss": 0.0819,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04382132738828659,
|
|
"step": 8610,
|
|
"valid_targets_mean": 5422.9,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 4.749173098125689,
|
|
"grad_norm": 0.36832255842739386,
|
|
"learning_rate": 1.1335065168882807e-05,
|
|
"loss": 0.0773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0418543815612793,
|
|
"step": 8615,
|
|
"valid_targets_mean": 2580.2,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 4.751929437706726,
|
|
"grad_norm": 0.3105105110691598,
|
|
"learning_rate": 1.1310297022937589e-05,
|
|
"loss": 0.0792,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037011466920375824,
|
|
"step": 8620,
|
|
"valid_targets_mean": 3741.8,
|
|
"valid_targets_min": 1091
|
|
},
|
|
{
|
|
"epoch": 4.7546857772877615,
|
|
"grad_norm": 0.47318543355609133,
|
|
"learning_rate": 1.1285545294383855e-05,
|
|
"loss": 0.0942,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046977050602436066,
|
|
"step": 8625,
|
|
"valid_targets_mean": 2870.5,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 4.757442116868798,
|
|
"grad_norm": 0.3509818284609481,
|
|
"learning_rate": 1.1260810029984861e-05,
|
|
"loss": 0.1006,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03285171091556549,
|
|
"step": 8630,
|
|
"valid_targets_mean": 3128.6,
|
|
"valid_targets_min": 550
|
|
},
|
|
{
|
|
"epoch": 4.760198456449834,
|
|
"grad_norm": 0.45330114246288405,
|
|
"learning_rate": 1.1236091276472751e-05,
|
|
"loss": 0.0804,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042145825922489166,
|
|
"step": 8635,
|
|
"valid_targets_mean": 2410.5,
|
|
"valid_targets_min": 563
|
|
},
|
|
{
|
|
"epoch": 4.762954796030871,
|
|
"grad_norm": 0.3074452864614632,
|
|
"learning_rate": 1.1211389080548499e-05,
|
|
"loss": 0.072,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03248672932386398,
|
|
"step": 8640,
|
|
"valid_targets_mean": 4067.1,
|
|
"valid_targets_min": 3326
|
|
},
|
|
{
|
|
"epoch": 4.765711135611907,
|
|
"grad_norm": 0.4757636905148467,
|
|
"learning_rate": 1.1186703488881773e-05,
|
|
"loss": 0.0776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05567679926753044,
|
|
"step": 8645,
|
|
"valid_targets_mean": 1677.4,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 4.768467475192944,
|
|
"grad_norm": 0.569381064638921,
|
|
"learning_rate": 1.116203454811088e-05,
|
|
"loss": 0.0908,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0527566596865654,
|
|
"step": 8650,
|
|
"valid_targets_mean": 2418.4,
|
|
"valid_targets_min": 692
|
|
},
|
|
{
|
|
"epoch": 4.77122381477398,
|
|
"grad_norm": 0.5091981756338738,
|
|
"learning_rate": 1.1137382304842665e-05,
|
|
"loss": 0.0868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05491676181554794,
|
|
"step": 8655,
|
|
"valid_targets_mean": 3002.8,
|
|
"valid_targets_min": 1914
|
|
},
|
|
{
|
|
"epoch": 4.773980154355017,
|
|
"grad_norm": 0.41872102716361553,
|
|
"learning_rate": 1.1112746805652432e-05,
|
|
"loss": 0.0862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02869214303791523,
|
|
"step": 8660,
|
|
"valid_targets_mean": 2430.4,
|
|
"valid_targets_min": 797
|
|
},
|
|
{
|
|
"epoch": 4.776736493936053,
|
|
"grad_norm": 0.4949848841636304,
|
|
"learning_rate": 1.1088128097083841e-05,
|
|
"loss": 0.0945,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057574834674596786,
|
|
"step": 8665,
|
|
"valid_targets_mean": 2037.0,
|
|
"valid_targets_min": 952
|
|
},
|
|
{
|
|
"epoch": 4.779492833517089,
|
|
"grad_norm": 0.3713952457269785,
|
|
"learning_rate": 1.106352622564884e-05,
|
|
"loss": 0.1145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04069802910089493,
|
|
"step": 8670,
|
|
"valid_targets_mean": 2504.5,
|
|
"valid_targets_min": 638
|
|
},
|
|
{
|
|
"epoch": 4.782249173098125,
|
|
"grad_norm": 0.39312885662929314,
|
|
"learning_rate": 1.1038941237827565e-05,
|
|
"loss": 0.076,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04487878829240799,
|
|
"step": 8675,
|
|
"valid_targets_mean": 2675.4,
|
|
"valid_targets_min": 1251
|
|
},
|
|
{
|
|
"epoch": 4.785005512679162,
|
|
"grad_norm": 0.5847943181198441,
|
|
"learning_rate": 1.1014373180068246e-05,
|
|
"loss": 0.1161,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08597130328416824,
|
|
"step": 8680,
|
|
"valid_targets_mean": 2135.6,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 4.787761852260198,
|
|
"grad_norm": 0.46172902846894137,
|
|
"learning_rate": 1.0989822098787126e-05,
|
|
"loss": 0.0961,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0563945509493351,
|
|
"step": 8685,
|
|
"valid_targets_mean": 2731.6,
|
|
"valid_targets_min": 607
|
|
},
|
|
{
|
|
"epoch": 4.790518191841235,
|
|
"grad_norm": 0.49734853890039443,
|
|
"learning_rate": 1.09652880403684e-05,
|
|
"loss": 0.0824,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05869923532009125,
|
|
"step": 8690,
|
|
"valid_targets_mean": 2722.5,
|
|
"valid_targets_min": 888
|
|
},
|
|
{
|
|
"epoch": 4.793274531422271,
|
|
"grad_norm": 0.6591071624643435,
|
|
"learning_rate": 1.0940771051164072e-05,
|
|
"loss": 0.2431,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.2542988061904907,
|
|
"step": 8695,
|
|
"valid_targets_mean": 2824.9,
|
|
"valid_targets_min": 1166
|
|
},
|
|
{
|
|
"epoch": 4.796030871003308,
|
|
"grad_norm": 0.4587676801525207,
|
|
"learning_rate": 1.0916271177493907e-05,
|
|
"loss": 0.1196,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039642058312892914,
|
|
"step": 8700,
|
|
"valid_targets_mean": 2113.9,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 4.798787210584344,
|
|
"grad_norm": 0.39235591093558847,
|
|
"learning_rate": 1.0891788465645338e-05,
|
|
"loss": 0.0956,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05093827843666077,
|
|
"step": 8705,
|
|
"valid_targets_mean": 3255.1,
|
|
"valid_targets_min": 612
|
|
},
|
|
{
|
|
"epoch": 4.8015435501653805,
|
|
"grad_norm": 0.5393248102272744,
|
|
"learning_rate": 1.0867322961873373e-05,
|
|
"loss": 0.1743,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07771757245063782,
|
|
"step": 8710,
|
|
"valid_targets_mean": 2770.4,
|
|
"valid_targets_min": 681
|
|
},
|
|
{
|
|
"epoch": 4.804299889746416,
|
|
"grad_norm": 0.4155130080449806,
|
|
"learning_rate": 1.0842874712400499e-05,
|
|
"loss": 0.0826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03190998733043671,
|
|
"step": 8715,
|
|
"valid_targets_mean": 2307.2,
|
|
"valid_targets_min": 825
|
|
},
|
|
{
|
|
"epoch": 4.807056229327453,
|
|
"grad_norm": 0.41980204984158603,
|
|
"learning_rate": 1.0818443763416622e-05,
|
|
"loss": 0.0807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045763175934553146,
|
|
"step": 8720,
|
|
"valid_targets_mean": 5000.9,
|
|
"valid_targets_min": 2660
|
|
},
|
|
{
|
|
"epoch": 4.809812568908489,
|
|
"grad_norm": 0.3402405318219772,
|
|
"learning_rate": 1.0794030161078948e-05,
|
|
"loss": 0.0871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03102729469537735,
|
|
"step": 8725,
|
|
"valid_targets_mean": 3396.2,
|
|
"valid_targets_min": 991
|
|
},
|
|
{
|
|
"epoch": 4.812568908489526,
|
|
"grad_norm": 0.41642920487982527,
|
|
"learning_rate": 1.0769633951511915e-05,
|
|
"loss": 0.0781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04984761029481888,
|
|
"step": 8730,
|
|
"valid_targets_mean": 3447.6,
|
|
"valid_targets_min": 868
|
|
},
|
|
{
|
|
"epoch": 4.815325248070562,
|
|
"grad_norm": 0.44525386561892577,
|
|
"learning_rate": 1.0745255180807094e-05,
|
|
"loss": 0.08,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04330720007419586,
|
|
"step": 8735,
|
|
"valid_targets_mean": 3714.5,
|
|
"valid_targets_min": 668
|
|
},
|
|
{
|
|
"epoch": 4.818081587651599,
|
|
"grad_norm": 0.35136187816734343,
|
|
"learning_rate": 1.0720893895023134e-05,
|
|
"loss": 0.0727,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03281503543257713,
|
|
"step": 8740,
|
|
"valid_targets_mean": 3203.4,
|
|
"valid_targets_min": 1761
|
|
},
|
|
{
|
|
"epoch": 4.820837927232635,
|
|
"grad_norm": 0.37339564325388935,
|
|
"learning_rate": 1.069655014018562e-05,
|
|
"loss": 0.0802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03664547950029373,
|
|
"step": 8745,
|
|
"valid_targets_mean": 3626.9,
|
|
"valid_targets_min": 2530
|
|
},
|
|
{
|
|
"epoch": 4.8235942668136715,
|
|
"grad_norm": 0.4300424111776043,
|
|
"learning_rate": 1.067222396228703e-05,
|
|
"loss": 0.07,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05024304240942001,
|
|
"step": 8750,
|
|
"valid_targets_mean": 3387.9,
|
|
"valid_targets_min": 1233
|
|
},
|
|
{
|
|
"epoch": 4.8263506063947075,
|
|
"grad_norm": 0.29951479814329085,
|
|
"learning_rate": 1.0647915407286632e-05,
|
|
"loss": 0.1188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.024494362995028496,
|
|
"step": 8755,
|
|
"valid_targets_mean": 3297.2,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 4.829106945975744,
|
|
"grad_norm": 0.3934678493078216,
|
|
"learning_rate": 1.0623624521110397e-05,
|
|
"loss": 0.0664,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033943675458431244,
|
|
"step": 8760,
|
|
"valid_targets_mean": 3229.9,
|
|
"valid_targets_min": 585
|
|
},
|
|
{
|
|
"epoch": 4.83186328555678,
|
|
"grad_norm": 0.38024955759215473,
|
|
"learning_rate": 1.059935134965092e-05,
|
|
"loss": 0.0715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03824543207883835,
|
|
"step": 8765,
|
|
"valid_targets_mean": 3873.1,
|
|
"valid_targets_min": 3327
|
|
},
|
|
{
|
|
"epoch": 4.834619625137817,
|
|
"grad_norm": 0.42249177183868675,
|
|
"learning_rate": 1.057509593876732e-05,
|
|
"loss": 0.0693,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.021149223670363426,
|
|
"step": 8770,
|
|
"valid_targets_mean": 1698.1,
|
|
"valid_targets_min": 523
|
|
},
|
|
{
|
|
"epoch": 4.837375964718853,
|
|
"grad_norm": 0.6874971886271674,
|
|
"learning_rate": 1.0550858334285165e-05,
|
|
"loss": 0.111,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09206512570381165,
|
|
"step": 8775,
|
|
"valid_targets_mean": 1589.6,
|
|
"valid_targets_min": 194
|
|
},
|
|
{
|
|
"epoch": 4.84013230429989,
|
|
"grad_norm": 0.37222577880475094,
|
|
"learning_rate": 1.0526638581996378e-05,
|
|
"loss": 0.0761,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04536258056759834,
|
|
"step": 8780,
|
|
"valid_targets_mean": 3791.4,
|
|
"valid_targets_min": 3033
|
|
},
|
|
{
|
|
"epoch": 4.842888643880926,
|
|
"grad_norm": 0.48372147669692217,
|
|
"learning_rate": 1.0502436727659169e-05,
|
|
"loss": 0.1148,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05402200296521187,
|
|
"step": 8785,
|
|
"valid_targets_mean": 2261.0,
|
|
"valid_targets_min": 959
|
|
},
|
|
{
|
|
"epoch": 4.845644983461963,
|
|
"grad_norm": 0.3960955878716764,
|
|
"learning_rate": 1.0478252816997918e-05,
|
|
"loss": 0.0802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0381321981549263,
|
|
"step": 8790,
|
|
"valid_targets_mean": 3329.0,
|
|
"valid_targets_min": 1025
|
|
},
|
|
{
|
|
"epoch": 4.8484013230429985,
|
|
"grad_norm": 0.6256186863586691,
|
|
"learning_rate": 1.0454086895703108e-05,
|
|
"loss": 0.1061,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06390616297721863,
|
|
"step": 8795,
|
|
"valid_targets_mean": 1613.5,
|
|
"valid_targets_min": 912
|
|
},
|
|
{
|
|
"epoch": 4.851157662624035,
|
|
"grad_norm": 0.5585964945060659,
|
|
"learning_rate": 1.0429939009431229e-05,
|
|
"loss": 0.1069,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047930020838975906,
|
|
"step": 8800,
|
|
"valid_targets_mean": 2147.6,
|
|
"valid_targets_min": 1035
|
|
},
|
|
{
|
|
"epoch": 4.853914002205071,
|
|
"grad_norm": 0.6067662965338925,
|
|
"learning_rate": 1.0405809203804713e-05,
|
|
"loss": 0.1029,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08391904085874557,
|
|
"step": 8805,
|
|
"valid_targets_mean": 1769.1,
|
|
"valid_targets_min": 1038
|
|
},
|
|
{
|
|
"epoch": 4.856670341786108,
|
|
"grad_norm": 0.4638813752233067,
|
|
"learning_rate": 1.0381697524411817e-05,
|
|
"loss": 0.0796,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03301233798265457,
|
|
"step": 8810,
|
|
"valid_targets_mean": 2832.5,
|
|
"valid_targets_min": 703
|
|
},
|
|
{
|
|
"epoch": 4.859426681367144,
|
|
"grad_norm": 0.36529263312196186,
|
|
"learning_rate": 1.0357604016806562e-05,
|
|
"loss": 0.0849,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0375593900680542,
|
|
"step": 8815,
|
|
"valid_targets_mean": 3462.8,
|
|
"valid_targets_min": 2841
|
|
},
|
|
{
|
|
"epoch": 4.862183020948181,
|
|
"grad_norm": 0.3991582173104919,
|
|
"learning_rate": 1.0333528726508628e-05,
|
|
"loss": 0.0844,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030580680817365646,
|
|
"step": 8820,
|
|
"valid_targets_mean": 2462.2,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 4.864939360529217,
|
|
"grad_norm": 0.3612592014515738,
|
|
"learning_rate": 1.0309471699003288e-05,
|
|
"loss": 0.0724,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03632736951112747,
|
|
"step": 8825,
|
|
"valid_targets_mean": 4582.6,
|
|
"valid_targets_min": 1048
|
|
},
|
|
{
|
|
"epoch": 4.867695700110254,
|
|
"grad_norm": 0.33466029377652595,
|
|
"learning_rate": 1.0285432979741295e-05,
|
|
"loss": 0.0679,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0361139252781868,
|
|
"step": 8830,
|
|
"valid_targets_mean": 4001.6,
|
|
"valid_targets_min": 2533
|
|
},
|
|
{
|
|
"epoch": 4.87045203969129,
|
|
"grad_norm": 0.32516406713875945,
|
|
"learning_rate": 1.0261412614138839e-05,
|
|
"loss": 0.0786,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.032514214515686035,
|
|
"step": 8835,
|
|
"valid_targets_mean": 3944.2,
|
|
"valid_targets_min": 1850
|
|
},
|
|
{
|
|
"epoch": 4.873208379272326,
|
|
"grad_norm": 0.48977801407192506,
|
|
"learning_rate": 1.0237410647577415e-05,
|
|
"loss": 0.0905,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03447722643613815,
|
|
"step": 8840,
|
|
"valid_targets_mean": 3158.4,
|
|
"valid_targets_min": 889
|
|
},
|
|
{
|
|
"epoch": 4.875964718853362,
|
|
"grad_norm": 0.4522682882646341,
|
|
"learning_rate": 1.0213427125403756e-05,
|
|
"loss": 0.0868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04081519693136215,
|
|
"step": 8845,
|
|
"valid_targets_mean": 2407.2,
|
|
"valid_targets_min": 749
|
|
},
|
|
{
|
|
"epoch": 4.878721058434399,
|
|
"grad_norm": 0.3081356612593599,
|
|
"learning_rate": 1.0189462092929756e-05,
|
|
"loss": 0.0885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03015086054801941,
|
|
"step": 8850,
|
|
"valid_targets_mean": 4191.4,
|
|
"valid_targets_min": 3925
|
|
},
|
|
{
|
|
"epoch": 4.881477398015435,
|
|
"grad_norm": 0.44120271909107134,
|
|
"learning_rate": 1.016551559543237e-05,
|
|
"loss": 0.0718,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04279952496290207,
|
|
"step": 8855,
|
|
"valid_targets_mean": 3605.1,
|
|
"valid_targets_min": 1567
|
|
},
|
|
{
|
|
"epoch": 4.884233737596472,
|
|
"grad_norm": 0.4732580164527041,
|
|
"learning_rate": 1.0141587678153542e-05,
|
|
"loss": 0.0833,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035236798226833344,
|
|
"step": 8860,
|
|
"valid_targets_mean": 2819.0,
|
|
"valid_targets_min": 902
|
|
},
|
|
{
|
|
"epoch": 4.886990077177508,
|
|
"grad_norm": 0.3466115828837437,
|
|
"learning_rate": 1.0117678386300107e-05,
|
|
"loss": 0.0745,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02841678261756897,
|
|
"step": 8865,
|
|
"valid_targets_mean": 3424.2,
|
|
"valid_targets_min": 974
|
|
},
|
|
{
|
|
"epoch": 4.889746416758545,
|
|
"grad_norm": 0.5985630712925906,
|
|
"learning_rate": 1.0093787765043715e-05,
|
|
"loss": 0.0807,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04562564939260483,
|
|
"step": 8870,
|
|
"valid_targets_mean": 1211.8,
|
|
"valid_targets_min": 569
|
|
},
|
|
{
|
|
"epoch": 4.892502756339581,
|
|
"grad_norm": 0.5408893267658812,
|
|
"learning_rate": 1.0069915859520726e-05,
|
|
"loss": 0.0771,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05622847005724907,
|
|
"step": 8875,
|
|
"valid_targets_mean": 2852.6,
|
|
"valid_targets_min": 1015
|
|
},
|
|
{
|
|
"epoch": 4.8952590959206175,
|
|
"grad_norm": 0.5114949934129079,
|
|
"learning_rate": 1.0046062714832176e-05,
|
|
"loss": 0.0803,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04796206206083298,
|
|
"step": 8880,
|
|
"valid_targets_mean": 1880.9,
|
|
"valid_targets_min": 677
|
|
},
|
|
{
|
|
"epoch": 4.898015435501653,
|
|
"grad_norm": 0.4121514777904532,
|
|
"learning_rate": 1.0022228376043626e-05,
|
|
"loss": 0.0936,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04005277156829834,
|
|
"step": 8885,
|
|
"valid_targets_mean": 2526.8,
|
|
"valid_targets_min": 1010
|
|
},
|
|
{
|
|
"epoch": 4.90077177508269,
|
|
"grad_norm": 0.5901749236959384,
|
|
"learning_rate": 9.998412888185117e-06,
|
|
"loss": 0.0776,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04498175531625748,
|
|
"step": 8890,
|
|
"valid_targets_mean": 1515.2,
|
|
"valid_targets_min": 700
|
|
},
|
|
{
|
|
"epoch": 4.903528114663727,
|
|
"grad_norm": 0.3782203957861455,
|
|
"learning_rate": 9.974616296251067e-06,
|
|
"loss": 0.0887,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.027473434805870056,
|
|
"step": 8895,
|
|
"valid_targets_mean": 2491.1,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 4.906284454244763,
|
|
"grad_norm": 0.5041741579658275,
|
|
"learning_rate": 9.950838645200209e-06,
|
|
"loss": 0.0989,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03366616368293762,
|
|
"step": 8900,
|
|
"valid_targets_mean": 3964.8,
|
|
"valid_targets_min": 3561
|
|
},
|
|
{
|
|
"epoch": 4.909040793825799,
|
|
"grad_norm": 0.42760575044277344,
|
|
"learning_rate": 9.927079979955479e-06,
|
|
"loss": 0.0803,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033416710793972015,
|
|
"step": 8905,
|
|
"valid_targets_mean": 2793.2,
|
|
"valid_targets_min": 809
|
|
},
|
|
{
|
|
"epoch": 4.911797133406836,
|
|
"grad_norm": 0.44800758349656516,
|
|
"learning_rate": 9.903340345403947e-06,
|
|
"loss": 0.0696,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043616458773612976,
|
|
"step": 8910,
|
|
"valid_targets_mean": 3057.9,
|
|
"valid_targets_min": 868
|
|
},
|
|
{
|
|
"epoch": 4.914553472987873,
|
|
"grad_norm": 0.48587160128324836,
|
|
"learning_rate": 9.87961978639673e-06,
|
|
"loss": 0.0855,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07806199789047241,
|
|
"step": 8915,
|
|
"valid_targets_mean": 2935.0,
|
|
"valid_targets_min": 1414
|
|
},
|
|
{
|
|
"epoch": 4.9173098125689085,
|
|
"grad_norm": 0.4025640631766033,
|
|
"learning_rate": 9.855918347748905e-06,
|
|
"loss": 0.0937,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03206022083759308,
|
|
"step": 8920,
|
|
"valid_targets_mean": 3051.6,
|
|
"valid_targets_min": 1103
|
|
},
|
|
{
|
|
"epoch": 4.9200661521499445,
|
|
"grad_norm": 0.3812081183089747,
|
|
"learning_rate": 9.832236074239415e-06,
|
|
"loss": 0.0629,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02963365614414215,
|
|
"step": 8925,
|
|
"valid_targets_mean": 2638.6,
|
|
"valid_targets_min": 680
|
|
},
|
|
{
|
|
"epoch": 4.922822491730981,
|
|
"grad_norm": 0.4479737534920497,
|
|
"learning_rate": 9.80857301061102e-06,
|
|
"loss": 0.0835,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037431083619594574,
|
|
"step": 8930,
|
|
"valid_targets_mean": 2300.4,
|
|
"valid_targets_min": 667
|
|
},
|
|
{
|
|
"epoch": 4.925578831312018,
|
|
"grad_norm": 0.38209225268612085,
|
|
"learning_rate": 9.784929201570166e-06,
|
|
"loss": 0.0738,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035658933222293854,
|
|
"step": 8935,
|
|
"valid_targets_mean": 4678.8,
|
|
"valid_targets_min": 3793
|
|
},
|
|
{
|
|
"epoch": 4.928335170893054,
|
|
"grad_norm": 0.3672354844118744,
|
|
"learning_rate": 9.761304691786921e-06,
|
|
"loss": 0.0638,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.032537903636693954,
|
|
"step": 8940,
|
|
"valid_targets_mean": 3685.1,
|
|
"valid_targets_min": 1236
|
|
},
|
|
{
|
|
"epoch": 4.93109151047409,
|
|
"grad_norm": 0.3826400199931243,
|
|
"learning_rate": 9.737699525894902e-06,
|
|
"loss": 0.0864,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03741665184497833,
|
|
"step": 8945,
|
|
"valid_targets_mean": 4091.2,
|
|
"valid_targets_min": 3314
|
|
},
|
|
{
|
|
"epoch": 4.933847850055127,
|
|
"grad_norm": 0.5039158253741157,
|
|
"learning_rate": 9.714113748491167e-06,
|
|
"loss": 0.0726,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043564386665821075,
|
|
"step": 8950,
|
|
"valid_targets_mean": 1614.5,
|
|
"valid_targets_min": 676
|
|
},
|
|
{
|
|
"epoch": 4.936604189636164,
|
|
"grad_norm": 0.43430392808886786,
|
|
"learning_rate": 9.69054740413616e-06,
|
|
"loss": 0.1004,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038885489106178284,
|
|
"step": 8955,
|
|
"valid_targets_mean": 2698.2,
|
|
"valid_targets_min": 561
|
|
},
|
|
{
|
|
"epoch": 4.9393605292172,
|
|
"grad_norm": 0.2680007947563871,
|
|
"learning_rate": 9.667000537353593e-06,
|
|
"loss": 0.0885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.018218006938695908,
|
|
"step": 8960,
|
|
"valid_targets_mean": 3602.9,
|
|
"valid_targets_min": 924
|
|
},
|
|
{
|
|
"epoch": 4.9421168687982355,
|
|
"grad_norm": 0.5870808982739218,
|
|
"learning_rate": 9.643473192630391e-06,
|
|
"loss": 0.0927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08240225911140442,
|
|
"step": 8965,
|
|
"valid_targets_mean": 3488.1,
|
|
"valid_targets_min": 3039
|
|
},
|
|
{
|
|
"epoch": 4.944873208379272,
|
|
"grad_norm": 0.38547362459663964,
|
|
"learning_rate": 9.61996541441659e-06,
|
|
"loss": 0.073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03705599159002304,
|
|
"step": 8970,
|
|
"valid_targets_mean": 3788.9,
|
|
"valid_targets_min": 2613
|
|
},
|
|
{
|
|
"epoch": 4.947629547960309,
|
|
"grad_norm": 0.7033422942005519,
|
|
"learning_rate": 9.596477247125253e-06,
|
|
"loss": 0.0813,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039514631032943726,
|
|
"step": 8975,
|
|
"valid_targets_mean": 865.4,
|
|
"valid_targets_min": 690
|
|
},
|
|
{
|
|
"epoch": 4.950385887541345,
|
|
"grad_norm": 0.3954259885324664,
|
|
"learning_rate": 9.57300873513242e-06,
|
|
"loss": 0.0754,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04018941521644592,
|
|
"step": 8980,
|
|
"valid_targets_mean": 3928.9,
|
|
"valid_targets_min": 3584
|
|
},
|
|
{
|
|
"epoch": 4.953142227122381,
|
|
"grad_norm": 0.8703492264058255,
|
|
"learning_rate": 9.549559922776968e-06,
|
|
"loss": 0.1043,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05118796229362488,
|
|
"step": 8985,
|
|
"valid_targets_mean": 1228.5,
|
|
"valid_targets_min": 540
|
|
},
|
|
{
|
|
"epoch": 4.955898566703418,
|
|
"grad_norm": 0.4786715853883184,
|
|
"learning_rate": 9.52613085436056e-06,
|
|
"loss": 0.1078,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05500349402427673,
|
|
"step": 8990,
|
|
"valid_targets_mean": 2029.6,
|
|
"valid_targets_min": 500
|
|
},
|
|
{
|
|
"epoch": 4.958654906284455,
|
|
"grad_norm": 0.5075106015141674,
|
|
"learning_rate": 9.502721574147567e-06,
|
|
"loss": 0.0813,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051122747361660004,
|
|
"step": 8995,
|
|
"valid_targets_mean": 2198.2,
|
|
"valid_targets_min": 622
|
|
},
|
|
{
|
|
"epoch": 4.961411245865491,
|
|
"grad_norm": 0.4365217322291381,
|
|
"learning_rate": 9.479332126364966e-06,
|
|
"loss": 0.0781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03304377198219299,
|
|
"step": 9000,
|
|
"valid_targets_mean": 3576.6,
|
|
"valid_targets_min": 2812
|
|
},
|
|
{
|
|
"epoch": 4.964167585446527,
|
|
"grad_norm": 0.44010757624753133,
|
|
"learning_rate": 9.455962555202267e-06,
|
|
"loss": 0.0779,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04642971605062485,
|
|
"step": 9005,
|
|
"valid_targets_mean": 3560.0,
|
|
"valid_targets_min": 2977
|
|
},
|
|
{
|
|
"epoch": 4.966923925027563,
|
|
"grad_norm": 0.3862791064514472,
|
|
"learning_rate": 9.432612904811429e-06,
|
|
"loss": 0.0721,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034055061638355255,
|
|
"step": 9010,
|
|
"valid_targets_mean": 2623.5,
|
|
"valid_targets_min": 852
|
|
},
|
|
{
|
|
"epoch": 4.9696802646086,
|
|
"grad_norm": 0.9355868049379382,
|
|
"learning_rate": 9.409283219306773e-06,
|
|
"loss": 0.079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05062797665596008,
|
|
"step": 9015,
|
|
"valid_targets_mean": 830.8,
|
|
"valid_targets_min": 603
|
|
},
|
|
{
|
|
"epoch": 4.972436604189636,
|
|
"grad_norm": 0.42962937439990273,
|
|
"learning_rate": 9.385973542764895e-06,
|
|
"loss": 0.0799,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029777351766824722,
|
|
"step": 9020,
|
|
"valid_targets_mean": 2720.6,
|
|
"valid_targets_min": 747
|
|
},
|
|
{
|
|
"epoch": 4.975192943770672,
|
|
"grad_norm": 0.5868434020241226,
|
|
"learning_rate": 9.362683919224606e-06,
|
|
"loss": 0.0724,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04924938827753067,
|
|
"step": 9025,
|
|
"valid_targets_mean": 2839.5,
|
|
"valid_targets_min": 859
|
|
},
|
|
{
|
|
"epoch": 4.977949283351709,
|
|
"grad_norm": 0.4735690680547003,
|
|
"learning_rate": 9.339414392686816e-06,
|
|
"loss": 0.0727,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045615263283252716,
|
|
"step": 9030,
|
|
"valid_targets_mean": 3792.5,
|
|
"valid_targets_min": 2890
|
|
},
|
|
{
|
|
"epoch": 4.980705622932746,
|
|
"grad_norm": 0.3582261774967449,
|
|
"learning_rate": 9.316165007114469e-06,
|
|
"loss": 0.0814,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029542695730924606,
|
|
"step": 9035,
|
|
"valid_targets_mean": 3551.2,
|
|
"valid_targets_min": 482
|
|
},
|
|
{
|
|
"epoch": 4.983461962513782,
|
|
"grad_norm": 0.29452108831665696,
|
|
"learning_rate": 9.29293580643245e-06,
|
|
"loss": 0.0687,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033004581928253174,
|
|
"step": 9040,
|
|
"valid_targets_mean": 4891.9,
|
|
"valid_targets_min": 2535
|
|
},
|
|
{
|
|
"epoch": 4.986218302094818,
|
|
"grad_norm": 0.3414991674533301,
|
|
"learning_rate": 9.269726834527524e-06,
|
|
"loss": 0.0744,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038802631199359894,
|
|
"step": 9045,
|
|
"valid_targets_mean": 4404.5,
|
|
"valid_targets_min": 3353
|
|
},
|
|
{
|
|
"epoch": 4.9889746416758545,
|
|
"grad_norm": 0.4529317978189164,
|
|
"learning_rate": 9.246538135248227e-06,
|
|
"loss": 0.0757,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04952320456504822,
|
|
"step": 9050,
|
|
"valid_targets_mean": 2717.0,
|
|
"valid_targets_min": 867
|
|
},
|
|
{
|
|
"epoch": 4.991730981256891,
|
|
"grad_norm": 0.3720625013259873,
|
|
"learning_rate": 9.223369752404794e-06,
|
|
"loss": 0.1613,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03378450125455856,
|
|
"step": 9055,
|
|
"valid_targets_mean": 2442.9,
|
|
"valid_targets_min": 1009
|
|
},
|
|
{
|
|
"epoch": 4.994487320837927,
|
|
"grad_norm": 0.37346489607655126,
|
|
"learning_rate": 9.200221729769084e-06,
|
|
"loss": 0.0737,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05096830427646637,
|
|
"step": 9060,
|
|
"valid_targets_mean": 2349.2,
|
|
"valid_targets_min": 1144
|
|
},
|
|
{
|
|
"epoch": 4.997243660418963,
|
|
"grad_norm": 0.36429975358000666,
|
|
"learning_rate": 9.17709411107448e-06,
|
|
"loss": 0.0748,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028960363939404488,
|
|
"step": 9065,
|
|
"valid_targets_mean": 3562.4,
|
|
"valid_targets_min": 2942
|
|
},
|
|
{
|
|
"epoch": 5.0,
|
|
"grad_norm": 0.4288389641789344,
|
|
"learning_rate": 9.153986940015817e-06,
|
|
"loss": 0.1056,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045981429517269135,
|
|
"step": 9070,
|
|
"valid_targets_mean": 4458.2,
|
|
"valid_targets_min": 1268
|
|
},
|
|
{
|
|
"epoch": 5.002756339581037,
|
|
"grad_norm": 0.7108610295102074,
|
|
"learning_rate": 9.130900260249315e-06,
|
|
"loss": 0.1904,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09192080795764923,
|
|
"step": 9075,
|
|
"valid_targets_mean": 7359.2,
|
|
"valid_targets_min": 5595
|
|
},
|
|
{
|
|
"epoch": 5.005512679162073,
|
|
"grad_norm": 0.6800903281655885,
|
|
"learning_rate": 9.10783411539246e-06,
|
|
"loss": 0.1821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04845072329044342,
|
|
"step": 9080,
|
|
"valid_targets_mean": 1287.1,
|
|
"valid_targets_min": 420
|
|
},
|
|
{
|
|
"epoch": 5.00826901874311,
|
|
"grad_norm": 0.39839657435756076,
|
|
"learning_rate": 9.084788549023951e-06,
|
|
"loss": 0.1731,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09145841747522354,
|
|
"step": 9085,
|
|
"valid_targets_mean": 8512.8,
|
|
"valid_targets_min": 6729
|
|
},
|
|
{
|
|
"epoch": 5.0110253583241455,
|
|
"grad_norm": 0.38079009893041005,
|
|
"learning_rate": 9.061763604683603e-06,
|
|
"loss": 0.1695,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0795409083366394,
|
|
"step": 9090,
|
|
"valid_targets_mean": 7552.0,
|
|
"valid_targets_min": 6145
|
|
},
|
|
{
|
|
"epoch": 5.013781697905182,
|
|
"grad_norm": 0.3816383138378206,
|
|
"learning_rate": 9.03875932587227e-06,
|
|
"loss": 0.1709,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07757729291915894,
|
|
"step": 9095,
|
|
"valid_targets_mean": 7301.4,
|
|
"valid_targets_min": 4807
|
|
},
|
|
{
|
|
"epoch": 5.016538037486218,
|
|
"grad_norm": 0.4087945504503888,
|
|
"learning_rate": 9.015775756051774e-06,
|
|
"loss": 0.179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09269396960735321,
|
|
"step": 9100,
|
|
"valid_targets_mean": 6809.0,
|
|
"valid_targets_min": 5015
|
|
},
|
|
{
|
|
"epoch": 5.019294377067255,
|
|
"grad_norm": 0.40336278542390835,
|
|
"learning_rate": 8.992812938644798e-06,
|
|
"loss": 0.1834,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09667226672172546,
|
|
"step": 9105,
|
|
"valid_targets_mean": 7217.0,
|
|
"valid_targets_min": 5541
|
|
},
|
|
{
|
|
"epoch": 5.022050716648291,
|
|
"grad_norm": 0.38557748723107665,
|
|
"learning_rate": 8.969870917034821e-06,
|
|
"loss": 0.1685,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07625013589859009,
|
|
"step": 9110,
|
|
"valid_targets_mean": 6205.1,
|
|
"valid_targets_min": 5580
|
|
},
|
|
{
|
|
"epoch": 5.024807056229328,
|
|
"grad_norm": 0.6482093056587003,
|
|
"learning_rate": 8.946949734566037e-06,
|
|
"loss": 0.1623,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.023855891078710556,
|
|
"step": 9115,
|
|
"valid_targets_mean": 598.0,
|
|
"valid_targets_min": 151
|
|
},
|
|
{
|
|
"epoch": 5.027563395810364,
|
|
"grad_norm": 0.4089926558113124,
|
|
"learning_rate": 8.92404943454326e-06,
|
|
"loss": 0.1576,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08300809562206268,
|
|
"step": 9120,
|
|
"valid_targets_mean": 7270.8,
|
|
"valid_targets_min": 6125
|
|
},
|
|
{
|
|
"epoch": 5.030319735391401,
|
|
"grad_norm": 0.3898842727747409,
|
|
"learning_rate": 8.901170060231873e-06,
|
|
"loss": 0.1713,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0806395560503006,
|
|
"step": 9125,
|
|
"valid_targets_mean": 6108.9,
|
|
"valid_targets_min": 4623
|
|
},
|
|
{
|
|
"epoch": 5.033076074972437,
|
|
"grad_norm": 0.5328988216578553,
|
|
"learning_rate": 8.878311654857697e-06,
|
|
"loss": 0.1636,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08502927422523499,
|
|
"step": 9130,
|
|
"valid_targets_mean": 6571.6,
|
|
"valid_targets_min": 4850
|
|
},
|
|
{
|
|
"epoch": 5.035832414553473,
|
|
"grad_norm": 0.3660454871366222,
|
|
"learning_rate": 8.855474261606948e-06,
|
|
"loss": 0.1651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08359389752149582,
|
|
"step": 9135,
|
|
"valid_targets_mean": 7984.4,
|
|
"valid_targets_min": 6312
|
|
},
|
|
{
|
|
"epoch": 5.038588754134509,
|
|
"grad_norm": 0.36740599002385466,
|
|
"learning_rate": 8.832657923626149e-06,
|
|
"loss": 0.1572,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07316365838050842,
|
|
"step": 9140,
|
|
"valid_targets_mean": 6273.1,
|
|
"valid_targets_min": 4590
|
|
},
|
|
{
|
|
"epoch": 5.041345093715546,
|
|
"grad_norm": 0.38694395334437204,
|
|
"learning_rate": 8.809862684022032e-06,
|
|
"loss": 0.165,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08720166981220245,
|
|
"step": 9145,
|
|
"valid_targets_mean": 7860.2,
|
|
"valid_targets_min": 5453
|
|
},
|
|
{
|
|
"epoch": 5.044101433296582,
|
|
"grad_norm": 0.5908487085576165,
|
|
"learning_rate": 8.787088585861481e-06,
|
|
"loss": 0.1662,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07537272572517395,
|
|
"step": 9150,
|
|
"valid_targets_mean": 6988.6,
|
|
"valid_targets_min": 4928
|
|
},
|
|
{
|
|
"epoch": 5.046857772877619,
|
|
"grad_norm": 0.4272815200028016,
|
|
"learning_rate": 8.764335672171428e-06,
|
|
"loss": 0.1623,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07856979966163635,
|
|
"step": 9155,
|
|
"valid_targets_mean": 6495.6,
|
|
"valid_targets_min": 5367
|
|
},
|
|
{
|
|
"epoch": 5.049614112458655,
|
|
"grad_norm": 0.4501888973106521,
|
|
"learning_rate": 8.74160398593878e-06,
|
|
"loss": 0.1668,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07981684058904648,
|
|
"step": 9160,
|
|
"valid_targets_mean": 4968.1,
|
|
"valid_targets_min": 615
|
|
},
|
|
{
|
|
"epoch": 5.052370452039692,
|
|
"grad_norm": 0.44328536175864364,
|
|
"learning_rate": 8.718893570110343e-06,
|
|
"loss": 0.1808,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09095379710197449,
|
|
"step": 9165,
|
|
"valid_targets_mean": 7189.9,
|
|
"valid_targets_min": 5732
|
|
},
|
|
{
|
|
"epoch": 5.055126791620728,
|
|
"grad_norm": 0.3905429473057596,
|
|
"learning_rate": 8.696204467592749e-06,
|
|
"loss": 0.1464,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08392584323883057,
|
|
"step": 9170,
|
|
"valid_targets_mean": 7969.4,
|
|
"valid_targets_min": 5870
|
|
},
|
|
{
|
|
"epoch": 5.0578831312017645,
|
|
"grad_norm": 0.38302151508030696,
|
|
"learning_rate": 8.673536721252342e-06,
|
|
"loss": 0.1561,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06684687733650208,
|
|
"step": 9175,
|
|
"valid_targets_mean": 7348.0,
|
|
"valid_targets_min": 4983
|
|
},
|
|
{
|
|
"epoch": 5.0606394707828,
|
|
"grad_norm": 0.38779057753443685,
|
|
"learning_rate": 8.65089037391513e-06,
|
|
"loss": 0.1616,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06842659413814545,
|
|
"step": 9180,
|
|
"valid_targets_mean": 6763.5,
|
|
"valid_targets_min": 5251
|
|
},
|
|
{
|
|
"epoch": 5.063395810363837,
|
|
"grad_norm": 0.41333376271706357,
|
|
"learning_rate": 8.62826546836669e-06,
|
|
"loss": 0.1684,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0930933952331543,
|
|
"step": 9185,
|
|
"valid_targets_mean": 6626.4,
|
|
"valid_targets_min": 4332
|
|
},
|
|
{
|
|
"epoch": 5.066152149944873,
|
|
"grad_norm": 0.3886932728506136,
|
|
"learning_rate": 8.605662047352088e-06,
|
|
"loss": 0.1659,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07610159367322922,
|
|
"step": 9190,
|
|
"valid_targets_mean": 7151.6,
|
|
"valid_targets_min": 4434
|
|
},
|
|
{
|
|
"epoch": 5.06890848952591,
|
|
"grad_norm": 0.4171411443250867,
|
|
"learning_rate": 8.583080153575797e-06,
|
|
"loss": 0.1647,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0977592021226883,
|
|
"step": 9195,
|
|
"valid_targets_mean": 7592.0,
|
|
"valid_targets_min": 5705
|
|
},
|
|
{
|
|
"epoch": 5.071664829106946,
|
|
"grad_norm": 0.3584798907538157,
|
|
"learning_rate": 8.560519829701627e-06,
|
|
"loss": 0.1454,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06810429692268372,
|
|
"step": 9200,
|
|
"valid_targets_mean": 8727.9,
|
|
"valid_targets_min": 5721
|
|
},
|
|
{
|
|
"epoch": 5.074421168687983,
|
|
"grad_norm": 0.40452827839823596,
|
|
"learning_rate": 8.537981118352627e-06,
|
|
"loss": 0.1479,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07612192630767822,
|
|
"step": 9205,
|
|
"valid_targets_mean": 6314.9,
|
|
"valid_targets_min": 4517
|
|
},
|
|
{
|
|
"epoch": 5.077177508269019,
|
|
"grad_norm": 0.3776122930927026,
|
|
"learning_rate": 8.515464062111017e-06,
|
|
"loss": 0.1405,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.080564945936203,
|
|
"step": 9210,
|
|
"valid_targets_mean": 7161.6,
|
|
"valid_targets_min": 4531
|
|
},
|
|
{
|
|
"epoch": 5.0799338478500555,
|
|
"grad_norm": 0.6994957190318662,
|
|
"learning_rate": 8.492968703518099e-06,
|
|
"loss": 0.1663,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05203543230891228,
|
|
"step": 9215,
|
|
"valid_targets_mean": 1098.8,
|
|
"valid_targets_min": 160
|
|
},
|
|
{
|
|
"epoch": 5.0826901874310915,
|
|
"grad_norm": 0.4040731968164092,
|
|
"learning_rate": 8.470495085074201e-06,
|
|
"loss": 0.1517,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07460442185401917,
|
|
"step": 9220,
|
|
"valid_targets_mean": 6988.4,
|
|
"valid_targets_min": 5392
|
|
},
|
|
{
|
|
"epoch": 5.085446527012128,
|
|
"grad_norm": 0.41382516668338837,
|
|
"learning_rate": 8.448043249238557e-06,
|
|
"loss": 0.1709,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08761563897132874,
|
|
"step": 9225,
|
|
"valid_targets_mean": 7712.9,
|
|
"valid_targets_min": 5297
|
|
},
|
|
{
|
|
"epoch": 5.088202866593164,
|
|
"grad_norm": 0.42555636579039535,
|
|
"learning_rate": 8.425613238429256e-06,
|
|
"loss": 0.1755,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08295845985412598,
|
|
"step": 9230,
|
|
"valid_targets_mean": 6396.9,
|
|
"valid_targets_min": 5536
|
|
},
|
|
{
|
|
"epoch": 5.090959206174201,
|
|
"grad_norm": 0.4418650460150576,
|
|
"learning_rate": 8.403205095023148e-06,
|
|
"loss": 0.1697,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08417355269193649,
|
|
"step": 9235,
|
|
"valid_targets_mean": 6143.8,
|
|
"valid_targets_min": 3434
|
|
},
|
|
{
|
|
"epoch": 5.093715545755237,
|
|
"grad_norm": 0.3884973059909086,
|
|
"learning_rate": 8.380818861355775e-06,
|
|
"loss": 0.1666,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08613456040620804,
|
|
"step": 9240,
|
|
"valid_targets_mean": 7182.1,
|
|
"valid_targets_min": 5349
|
|
},
|
|
{
|
|
"epoch": 5.096471885336274,
|
|
"grad_norm": 0.3943007084765131,
|
|
"learning_rate": 8.358454579721283e-06,
|
|
"loss": 0.1695,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08749531209468842,
|
|
"step": 9245,
|
|
"valid_targets_mean": 7570.5,
|
|
"valid_targets_min": 5209
|
|
},
|
|
{
|
|
"epoch": 5.09922822491731,
|
|
"grad_norm": 0.38798616341222064,
|
|
"learning_rate": 8.33611229237234e-06,
|
|
"loss": 0.1646,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08912448585033417,
|
|
"step": 9250,
|
|
"valid_targets_mean": 7069.0,
|
|
"valid_targets_min": 5581
|
|
},
|
|
{
|
|
"epoch": 5.101984564498347,
|
|
"grad_norm": 0.4174151109506085,
|
|
"learning_rate": 8.313792041520066e-06,
|
|
"loss": 0.1691,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0892706960439682,
|
|
"step": 9255,
|
|
"valid_targets_mean": 6925.6,
|
|
"valid_targets_min": 5504
|
|
},
|
|
{
|
|
"epoch": 5.1047409040793825,
|
|
"grad_norm": 0.7453459094397573,
|
|
"learning_rate": 8.29149386933394e-06,
|
|
"loss": 0.1689,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08622249960899353,
|
|
"step": 9260,
|
|
"valid_targets_mean": 6716.9,
|
|
"valid_targets_min": 5332
|
|
},
|
|
{
|
|
"epoch": 5.107497243660419,
|
|
"grad_norm": 0.4286754556105685,
|
|
"learning_rate": 8.269217817941747e-06,
|
|
"loss": 0.1701,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08078384399414062,
|
|
"step": 9265,
|
|
"valid_targets_mean": 6241.8,
|
|
"valid_targets_min": 5406
|
|
},
|
|
{
|
|
"epoch": 5.110253583241455,
|
|
"grad_norm": 0.3862064752740063,
|
|
"learning_rate": 8.246963929429454e-06,
|
|
"loss": 0.1651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07794445753097534,
|
|
"step": 9270,
|
|
"valid_targets_mean": 6387.8,
|
|
"valid_targets_min": 4847
|
|
},
|
|
{
|
|
"epoch": 5.113009922822492,
|
|
"grad_norm": 0.4122538105778726,
|
|
"learning_rate": 8.224732245841176e-06,
|
|
"loss": 0.1628,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07740286737680435,
|
|
"step": 9275,
|
|
"valid_targets_mean": 6457.6,
|
|
"valid_targets_min": 4288
|
|
},
|
|
{
|
|
"epoch": 5.115766262403528,
|
|
"grad_norm": 0.44197828422984053,
|
|
"learning_rate": 8.202522809179065e-06,
|
|
"loss": 0.1362,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09640341997146606,
|
|
"step": 9280,
|
|
"valid_targets_mean": 6793.1,
|
|
"valid_targets_min": 5173
|
|
},
|
|
{
|
|
"epoch": 5.118522601984565,
|
|
"grad_norm": 0.41714712204677284,
|
|
"learning_rate": 8.180335661403242e-06,
|
|
"loss": 0.1647,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07658015936613083,
|
|
"step": 9285,
|
|
"valid_targets_mean": 6697.4,
|
|
"valid_targets_min": 5168
|
|
},
|
|
{
|
|
"epoch": 5.121278941565601,
|
|
"grad_norm": 0.4266115203199952,
|
|
"learning_rate": 8.158170844431725e-06,
|
|
"loss": 0.1618,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07894925773143768,
|
|
"step": 9290,
|
|
"valid_targets_mean": 6062.0,
|
|
"valid_targets_min": 5270
|
|
},
|
|
{
|
|
"epoch": 5.124035281146638,
|
|
"grad_norm": 0.4180433545571716,
|
|
"learning_rate": 8.136028400140337e-06,
|
|
"loss": 0.1661,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08554425835609436,
|
|
"step": 9295,
|
|
"valid_targets_mean": 7110.1,
|
|
"valid_targets_min": 5253
|
|
},
|
|
{
|
|
"epoch": 5.126791620727674,
|
|
"grad_norm": 0.39154267366495543,
|
|
"learning_rate": 8.113908370362633e-06,
|
|
"loss": 0.1598,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07759808003902435,
|
|
"step": 9300,
|
|
"valid_targets_mean": 7395.9,
|
|
"valid_targets_min": 5248
|
|
},
|
|
{
|
|
"epoch": 5.12954796030871,
|
|
"grad_norm": 0.413899895124058,
|
|
"learning_rate": 8.091810796889825e-06,
|
|
"loss": 0.1667,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10593347251415253,
|
|
"step": 9305,
|
|
"valid_targets_mean": 7033.5,
|
|
"valid_targets_min": 5354
|
|
},
|
|
{
|
|
"epoch": 5.132304299889746,
|
|
"grad_norm": 0.41492354674338777,
|
|
"learning_rate": 8.069735721470685e-06,
|
|
"loss": 0.1668,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08595170080661774,
|
|
"step": 9310,
|
|
"valid_targets_mean": 6798.8,
|
|
"valid_targets_min": 4532
|
|
},
|
|
{
|
|
"epoch": 5.135060639470783,
|
|
"grad_norm": 0.5602906608455438,
|
|
"learning_rate": 8.047683185811501e-06,
|
|
"loss": 0.1432,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038363490253686905,
|
|
"step": 9315,
|
|
"valid_targets_mean": 2045.0,
|
|
"valid_targets_min": 164
|
|
},
|
|
{
|
|
"epoch": 5.137816979051819,
|
|
"grad_norm": 0.39459285071178246,
|
|
"learning_rate": 8.025653231575963e-06,
|
|
"loss": 0.1484,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07769788801670074,
|
|
"step": 9320,
|
|
"valid_targets_mean": 6723.1,
|
|
"valid_targets_min": 5164
|
|
},
|
|
{
|
|
"epoch": 5.140573318632856,
|
|
"grad_norm": 0.44154007805832657,
|
|
"learning_rate": 8.003645900385097e-06,
|
|
"loss": 0.1616,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08141817152500153,
|
|
"step": 9325,
|
|
"valid_targets_mean": 5954.1,
|
|
"valid_targets_min": 4996
|
|
},
|
|
{
|
|
"epoch": 5.143329658213892,
|
|
"grad_norm": 0.44465937188830107,
|
|
"learning_rate": 7.981661233817196e-06,
|
|
"loss": 0.1678,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0769452452659607,
|
|
"step": 9330,
|
|
"valid_targets_mean": 5932.0,
|
|
"valid_targets_min": 5205
|
|
},
|
|
{
|
|
"epoch": 5.146085997794929,
|
|
"grad_norm": 0.4218214969036919,
|
|
"learning_rate": 7.959699273407722e-06,
|
|
"loss": 0.1625,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08574438095092773,
|
|
"step": 9335,
|
|
"valid_targets_mean": 6288.4,
|
|
"valid_targets_min": 5380
|
|
},
|
|
{
|
|
"epoch": 5.148842337375965,
|
|
"grad_norm": 0.37233550971818064,
|
|
"learning_rate": 7.937760060649247e-06,
|
|
"loss": 0.1536,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06653133034706116,
|
|
"step": 9340,
|
|
"valid_targets_mean": 7095.9,
|
|
"valid_targets_min": 5048
|
|
},
|
|
{
|
|
"epoch": 5.1515986769570015,
|
|
"grad_norm": 0.4261771690223918,
|
|
"learning_rate": 7.915843636991362e-06,
|
|
"loss": 0.1631,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08012393116950989,
|
|
"step": 9345,
|
|
"valid_targets_mean": 6093.5,
|
|
"valid_targets_min": 4489
|
|
},
|
|
{
|
|
"epoch": 5.154355016538037,
|
|
"grad_norm": 0.8762306508636589,
|
|
"learning_rate": 7.893950043840605e-06,
|
|
"loss": 0.1616,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09650160372257233,
|
|
"step": 9350,
|
|
"valid_targets_mean": 1879.2,
|
|
"valid_targets_min": 1107
|
|
},
|
|
{
|
|
"epoch": 5.157111356119074,
|
|
"grad_norm": 0.7516697803432345,
|
|
"learning_rate": 7.87207932256038e-06,
|
|
"loss": 0.1548,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07895572483539581,
|
|
"step": 9355,
|
|
"valid_targets_mean": 1493.0,
|
|
"valid_targets_min": 892
|
|
},
|
|
{
|
|
"epoch": 5.15986769570011,
|
|
"grad_norm": 0.7459830423161832,
|
|
"learning_rate": 7.85023151447087e-06,
|
|
"loss": 0.1413,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05052900314331055,
|
|
"step": 9360,
|
|
"valid_targets_mean": 1350.0,
|
|
"valid_targets_min": 908
|
|
},
|
|
{
|
|
"epoch": 5.162624035281147,
|
|
"grad_norm": 0.7386761374042871,
|
|
"learning_rate": 7.82840666084899e-06,
|
|
"loss": 0.1386,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.056627340614795685,
|
|
"step": 9365,
|
|
"valid_targets_mean": 1319.6,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 5.165380374862183,
|
|
"grad_norm": 0.7784935524292635,
|
|
"learning_rate": 7.806604802928272e-06,
|
|
"loss": 0.1396,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07376453280448914,
|
|
"step": 9370,
|
|
"valid_targets_mean": 1923.8,
|
|
"valid_targets_min": 875
|
|
},
|
|
{
|
|
"epoch": 5.16813671444322,
|
|
"grad_norm": 0.8116615968609623,
|
|
"learning_rate": 7.7848259818988e-06,
|
|
"loss": 0.1352,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05947571247816086,
|
|
"step": 9375,
|
|
"valid_targets_mean": 1288.6,
|
|
"valid_targets_min": 731
|
|
},
|
|
{
|
|
"epoch": 5.170893054024256,
|
|
"grad_norm": 0.7701545672527379,
|
|
"learning_rate": 7.76307023890714e-06,
|
|
"loss": 0.1431,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0570276714861393,
|
|
"step": 9380,
|
|
"valid_targets_mean": 1464.1,
|
|
"valid_targets_min": 652
|
|
},
|
|
{
|
|
"epoch": 5.1736493936052925,
|
|
"grad_norm": 1.0057029157052861,
|
|
"learning_rate": 7.741337615056259e-06,
|
|
"loss": 0.1231,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07136718928813934,
|
|
"step": 9385,
|
|
"valid_targets_mean": 1542.4,
|
|
"valid_targets_min": 736
|
|
},
|
|
{
|
|
"epoch": 5.1764057331863285,
|
|
"grad_norm": 0.7654085837095604,
|
|
"learning_rate": 7.71962815140544e-06,
|
|
"loss": 0.1426,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059405989944934845,
|
|
"step": 9390,
|
|
"valid_targets_mean": 1312.0,
|
|
"valid_targets_min": 758
|
|
},
|
|
{
|
|
"epoch": 5.179162072767365,
|
|
"grad_norm": 1.2129658992530392,
|
|
"learning_rate": 7.697941888970213e-06,
|
|
"loss": 0.1313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051202528178691864,
|
|
"step": 9395,
|
|
"valid_targets_mean": 1246.1,
|
|
"valid_targets_min": 752
|
|
},
|
|
{
|
|
"epoch": 5.181918412348401,
|
|
"grad_norm": 0.8107538781216953,
|
|
"learning_rate": 7.67627886872227e-06,
|
|
"loss": 0.1335,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08009755611419678,
|
|
"step": 9400,
|
|
"valid_targets_mean": 1828.1,
|
|
"valid_targets_min": 926
|
|
},
|
|
{
|
|
"epoch": 5.184674751929438,
|
|
"grad_norm": 0.8323719738383074,
|
|
"learning_rate": 7.654639131589394e-06,
|
|
"loss": 0.1339,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06427835673093796,
|
|
"step": 9405,
|
|
"valid_targets_mean": 1334.8,
|
|
"valid_targets_min": 676
|
|
},
|
|
{
|
|
"epoch": 5.187431091510474,
|
|
"grad_norm": 0.8020217076424375,
|
|
"learning_rate": 7.633022718455387e-06,
|
|
"loss": 0.1309,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0632963627576828,
|
|
"step": 9410,
|
|
"valid_targets_mean": 1240.0,
|
|
"valid_targets_min": 769
|
|
},
|
|
{
|
|
"epoch": 5.190187431091511,
|
|
"grad_norm": 0.7027545721946764,
|
|
"learning_rate": 7.611429670159976e-06,
|
|
"loss": 0.1184,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05917418748140335,
|
|
"step": 9415,
|
|
"valid_targets_mean": 1725.1,
|
|
"valid_targets_min": 982
|
|
},
|
|
{
|
|
"epoch": 5.192943770672547,
|
|
"grad_norm": 0.7556228917966986,
|
|
"learning_rate": 7.589860027498746e-06,
|
|
"loss": 0.1246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0576883964240551,
|
|
"step": 9420,
|
|
"valid_targets_mean": 1385.1,
|
|
"valid_targets_min": 1099
|
|
},
|
|
{
|
|
"epoch": 5.195700110253584,
|
|
"grad_norm": 0.9015548303436165,
|
|
"learning_rate": 7.568313831223062e-06,
|
|
"loss": 0.1281,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07024513930082321,
|
|
"step": 9425,
|
|
"valid_targets_mean": 1588.1,
|
|
"valid_targets_min": 674
|
|
},
|
|
{
|
|
"epoch": 5.1984564498346195,
|
|
"grad_norm": 0.7731081588186494,
|
|
"learning_rate": 7.546791122039994e-06,
|
|
"loss": 0.1294,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07758387178182602,
|
|
"step": 9430,
|
|
"valid_targets_mean": 1918.5,
|
|
"valid_targets_min": 973
|
|
},
|
|
{
|
|
"epoch": 5.201212789415656,
|
|
"grad_norm": 0.7939597665005103,
|
|
"learning_rate": 7.525291940612236e-06,
|
|
"loss": 0.1336,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07463960349559784,
|
|
"step": 9435,
|
|
"valid_targets_mean": 1943.9,
|
|
"valid_targets_min": 865
|
|
},
|
|
{
|
|
"epoch": 5.203969128996692,
|
|
"grad_norm": 0.7579603716333017,
|
|
"learning_rate": 7.503816327558036e-06,
|
|
"loss": 0.1313,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05638907849788666,
|
|
"step": 9440,
|
|
"valid_targets_mean": 1323.0,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 5.206725468577729,
|
|
"grad_norm": 0.7454528270386546,
|
|
"learning_rate": 7.482364323451106e-06,
|
|
"loss": 0.1321,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06451649963855743,
|
|
"step": 9445,
|
|
"valid_targets_mean": 1712.6,
|
|
"valid_targets_min": 708
|
|
},
|
|
{
|
|
"epoch": 5.209481808158765,
|
|
"grad_norm": 0.8114324171467913,
|
|
"learning_rate": 7.46093596882056e-06,
|
|
"loss": 0.1296,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052536312490701675,
|
|
"step": 9450,
|
|
"valid_targets_mean": 1137.1,
|
|
"valid_targets_min": 657
|
|
},
|
|
{
|
|
"epoch": 5.212238147739802,
|
|
"grad_norm": 0.7807211071121418,
|
|
"learning_rate": 7.439531304150824e-06,
|
|
"loss": 0.1312,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06048858165740967,
|
|
"step": 9455,
|
|
"valid_targets_mean": 1440.4,
|
|
"valid_targets_min": 661
|
|
},
|
|
{
|
|
"epoch": 5.214994487320838,
|
|
"grad_norm": 0.7780596276167248,
|
|
"learning_rate": 7.4181503698815845e-06,
|
|
"loss": 0.1291,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046891503036022186,
|
|
"step": 9460,
|
|
"valid_targets_mean": 1238.1,
|
|
"valid_targets_min": 858
|
|
},
|
|
{
|
|
"epoch": 5.217750826901875,
|
|
"grad_norm": 0.780452437750225,
|
|
"learning_rate": 7.396793206407677e-06,
|
|
"loss": 0.1168,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07133437693119049,
|
|
"step": 9465,
|
|
"valid_targets_mean": 1564.0,
|
|
"valid_targets_min": 949
|
|
},
|
|
{
|
|
"epoch": 5.220507166482911,
|
|
"grad_norm": 0.940115909158042,
|
|
"learning_rate": 7.375459854079034e-06,
|
|
"loss": 0.1251,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06167183071374893,
|
|
"step": 9470,
|
|
"valid_targets_mean": 1533.5,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 5.223263506063947,
|
|
"grad_norm": 0.890439935052888,
|
|
"learning_rate": 7.354150353200596e-06,
|
|
"loss": 0.1273,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0776662528514862,
|
|
"step": 9475,
|
|
"valid_targets_mean": 1604.9,
|
|
"valid_targets_min": 701
|
|
},
|
|
{
|
|
"epoch": 5.226019845644983,
|
|
"grad_norm": 0.8143154588586371,
|
|
"learning_rate": 7.332864744032246e-06,
|
|
"loss": 0.1234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04480643570423126,
|
|
"step": 9480,
|
|
"valid_targets_mean": 1186.8,
|
|
"valid_targets_min": 776
|
|
},
|
|
{
|
|
"epoch": 5.22877618522602,
|
|
"grad_norm": 0.7984432103518015,
|
|
"learning_rate": 7.31160306678873e-06,
|
|
"loss": 0.1357,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06651563197374344,
|
|
"step": 9485,
|
|
"valid_targets_mean": 1462.8,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 5.231532524807056,
|
|
"grad_norm": 0.7595087356398955,
|
|
"learning_rate": 7.290365361639577e-06,
|
|
"loss": 0.1234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0564706064760685,
|
|
"step": 9490,
|
|
"valid_targets_mean": 1345.4,
|
|
"valid_targets_min": 867
|
|
},
|
|
{
|
|
"epoch": 5.234288864388093,
|
|
"grad_norm": 0.8200277928085319,
|
|
"learning_rate": 7.269151668709024e-06,
|
|
"loss": 0.1286,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05494692921638489,
|
|
"step": 9495,
|
|
"valid_targets_mean": 1279.9,
|
|
"valid_targets_min": 780
|
|
},
|
|
{
|
|
"epoch": 5.237045203969129,
|
|
"grad_norm": 0.7745075953674273,
|
|
"learning_rate": 7.247962028075946e-06,
|
|
"loss": 0.1246,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05468755215406418,
|
|
"step": 9500,
|
|
"valid_targets_mean": 1299.2,
|
|
"valid_targets_min": 792
|
|
},
|
|
{
|
|
"epoch": 5.239801543550166,
|
|
"grad_norm": 0.7955489535911753,
|
|
"learning_rate": 7.226796479773763e-06,
|
|
"loss": 0.1292,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05063435435295105,
|
|
"step": 9505,
|
|
"valid_targets_mean": 1305.6,
|
|
"valid_targets_min": 805
|
|
},
|
|
{
|
|
"epoch": 5.242557883131202,
|
|
"grad_norm": 0.756039530326261,
|
|
"learning_rate": 7.205655063790408e-06,
|
|
"loss": 0.1235,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05369751527905464,
|
|
"step": 9510,
|
|
"valid_targets_mean": 1392.9,
|
|
"valid_targets_min": 802
|
|
},
|
|
{
|
|
"epoch": 5.2453142227122385,
|
|
"grad_norm": 0.7422065202095264,
|
|
"learning_rate": 7.184537820068194e-06,
|
|
"loss": 0.123,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046857476234436035,
|
|
"step": 9515,
|
|
"valid_targets_mean": 1183.9,
|
|
"valid_targets_min": 658
|
|
},
|
|
{
|
|
"epoch": 5.248070562293274,
|
|
"grad_norm": 0.7616739942021659,
|
|
"learning_rate": 7.1634447885037705e-06,
|
|
"loss": 0.1131,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05396449193358421,
|
|
"step": 9520,
|
|
"valid_targets_mean": 1540.5,
|
|
"valid_targets_min": 852
|
|
},
|
|
{
|
|
"epoch": 5.250826901874311,
|
|
"grad_norm": 0.9026895566744955,
|
|
"learning_rate": 7.142376008948049e-06,
|
|
"loss": 0.1262,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06392336636781693,
|
|
"step": 9525,
|
|
"valid_targets_mean": 1697.2,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 5.253583241455347,
|
|
"grad_norm": 0.7999564425735289,
|
|
"learning_rate": 7.121331521206114e-06,
|
|
"loss": 0.1239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04812490940093994,
|
|
"step": 9530,
|
|
"valid_targets_mean": 1251.9,
|
|
"valid_targets_min": 763
|
|
},
|
|
{
|
|
"epoch": 5.256339581036384,
|
|
"grad_norm": 0.8324407443149139,
|
|
"learning_rate": 7.1003113650371645e-06,
|
|
"loss": 0.1234,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0668843612074852,
|
|
"step": 9535,
|
|
"valid_targets_mean": 1545.5,
|
|
"valid_targets_min": 874
|
|
},
|
|
{
|
|
"epoch": 5.25909592061742,
|
|
"grad_norm": 0.8111347799295774,
|
|
"learning_rate": 7.079315580154425e-06,
|
|
"loss": 0.1181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.060793712735176086,
|
|
"step": 9540,
|
|
"valid_targets_mean": 1356.6,
|
|
"valid_targets_min": 736
|
|
},
|
|
{
|
|
"epoch": 5.261852260198457,
|
|
"grad_norm": 0.8432217446251274,
|
|
"learning_rate": 7.058344206225074e-06,
|
|
"loss": 0.1245,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0684332549571991,
|
|
"step": 9545,
|
|
"valid_targets_mean": 1661.6,
|
|
"valid_targets_min": 759
|
|
},
|
|
{
|
|
"epoch": 5.264608599779493,
|
|
"grad_norm": 0.8053430421563212,
|
|
"learning_rate": 7.037397282870171e-06,
|
|
"loss": 0.1197,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051424670964479446,
|
|
"step": 9550,
|
|
"valid_targets_mean": 1315.0,
|
|
"valid_targets_min": 1045
|
|
},
|
|
{
|
|
"epoch": 5.2673649393605295,
|
|
"grad_norm": 0.8600797447607577,
|
|
"learning_rate": 7.016474849664592e-06,
|
|
"loss": 0.1273,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07262850552797318,
|
|
"step": 9555,
|
|
"valid_targets_mean": 1722.8,
|
|
"valid_targets_min": 847
|
|
},
|
|
{
|
|
"epoch": 5.2701212789415655,
|
|
"grad_norm": 0.821300899139451,
|
|
"learning_rate": 6.995576946136928e-06,
|
|
"loss": 0.1297,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06632635742425919,
|
|
"step": 9560,
|
|
"valid_targets_mean": 1637.4,
|
|
"valid_targets_min": 906
|
|
},
|
|
{
|
|
"epoch": 5.272877618522602,
|
|
"grad_norm": 0.790885215928923,
|
|
"learning_rate": 6.974703611769438e-06,
|
|
"loss": 0.1306,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05546237528324127,
|
|
"step": 9565,
|
|
"valid_targets_mean": 1473.2,
|
|
"valid_targets_min": 760
|
|
},
|
|
{
|
|
"epoch": 5.275633958103638,
|
|
"grad_norm": 0.7695719605512097,
|
|
"learning_rate": 6.9538548859979574e-06,
|
|
"loss": 0.1185,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055496394634246826,
|
|
"step": 9570,
|
|
"valid_targets_mean": 1519.1,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 5.278390297684675,
|
|
"grad_norm": 0.7541576043307808,
|
|
"learning_rate": 6.933030808211825e-06,
|
|
"loss": 0.1163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05974021181464195,
|
|
"step": 9575,
|
|
"valid_targets_mean": 1582.4,
|
|
"valid_targets_min": 982
|
|
},
|
|
{
|
|
"epoch": 5.281146637265711,
|
|
"grad_norm": 0.7796784670312236,
|
|
"learning_rate": 6.912231417753818e-06,
|
|
"loss": 0.127,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06075458228588104,
|
|
"step": 9580,
|
|
"valid_targets_mean": 1568.8,
|
|
"valid_targets_min": 815
|
|
},
|
|
{
|
|
"epoch": 5.283902976846748,
|
|
"grad_norm": 0.8311063586723451,
|
|
"learning_rate": 6.891456753920076e-06,
|
|
"loss": 0.1181,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05546252429485321,
|
|
"step": 9585,
|
|
"valid_targets_mean": 1417.9,
|
|
"valid_targets_min": 720
|
|
},
|
|
{
|
|
"epoch": 5.286659316427784,
|
|
"grad_norm": 0.7795992924015998,
|
|
"learning_rate": 6.8707068559600095e-06,
|
|
"loss": 0.1144,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04650666564702988,
|
|
"step": 9590,
|
|
"valid_targets_mean": 1146.1,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 5.289415656008821,
|
|
"grad_norm": 0.8224887822148372,
|
|
"learning_rate": 6.849981763076255e-06,
|
|
"loss": 0.1207,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06778716295957565,
|
|
"step": 9595,
|
|
"valid_targets_mean": 1676.6,
|
|
"valid_targets_min": 733
|
|
},
|
|
{
|
|
"epoch": 5.2921719955898565,
|
|
"grad_norm": 0.8178545051145197,
|
|
"learning_rate": 6.829281514424564e-06,
|
|
"loss": 0.1222,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07455785572528839,
|
|
"step": 9600,
|
|
"valid_targets_mean": 1928.6,
|
|
"valid_targets_min": 743
|
|
},
|
|
{
|
|
"epoch": 5.294928335170893,
|
|
"grad_norm": 0.7974385922397952,
|
|
"learning_rate": 6.808606149113779e-06,
|
|
"loss": 0.1216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06749487668275833,
|
|
"step": 9605,
|
|
"valid_targets_mean": 1836.4,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 5.297684674751929,
|
|
"grad_norm": 0.7920503160679432,
|
|
"learning_rate": 6.787955706205709e-06,
|
|
"loss": 0.1157,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06736399233341217,
|
|
"step": 9610,
|
|
"valid_targets_mean": 1631.6,
|
|
"valid_targets_min": 784
|
|
},
|
|
{
|
|
"epoch": 5.300441014332966,
|
|
"grad_norm": 0.8167480949110177,
|
|
"learning_rate": 6.76733022471508e-06,
|
|
"loss": 0.1239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0665399357676506,
|
|
"step": 9615,
|
|
"valid_targets_mean": 1962.4,
|
|
"valid_targets_min": 739
|
|
},
|
|
{
|
|
"epoch": 5.303197353914002,
|
|
"grad_norm": 0.7674639977981773,
|
|
"learning_rate": 6.746729743609466e-06,
|
|
"loss": 0.1163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05641740933060646,
|
|
"step": 9620,
|
|
"valid_targets_mean": 1426.4,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 5.305953693495039,
|
|
"grad_norm": 0.7365556508755293,
|
|
"learning_rate": 6.726154301809198e-06,
|
|
"loss": 0.1239,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06366628408432007,
|
|
"step": 9625,
|
|
"valid_targets_mean": 1690.1,
|
|
"valid_targets_min": 1121
|
|
},
|
|
{
|
|
"epoch": 5.308710033076075,
|
|
"grad_norm": 0.7828969252351737,
|
|
"learning_rate": 6.705603938187308e-06,
|
|
"loss": 0.1188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06271663308143616,
|
|
"step": 9630,
|
|
"valid_targets_mean": 1519.1,
|
|
"valid_targets_min": 721
|
|
},
|
|
{
|
|
"epoch": 5.311466372657112,
|
|
"grad_norm": 0.7732399388271711,
|
|
"learning_rate": 6.685078691569446e-06,
|
|
"loss": 0.1148,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0528080016374588,
|
|
"step": 9635,
|
|
"valid_targets_mean": 1284.8,
|
|
"valid_targets_min": 660
|
|
},
|
|
{
|
|
"epoch": 5.314222712238148,
|
|
"grad_norm": 0.7877180731448664,
|
|
"learning_rate": 6.664578600733811e-06,
|
|
"loss": 0.115,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05886983126401901,
|
|
"step": 9640,
|
|
"valid_targets_mean": 1513.5,
|
|
"valid_targets_min": 1009
|
|
},
|
|
{
|
|
"epoch": 5.316979051819184,
|
|
"grad_norm": 0.8230669234714441,
|
|
"learning_rate": 6.64410370441106e-06,
|
|
"loss": 0.1242,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06956428289413452,
|
|
"step": 9645,
|
|
"valid_targets_mean": 1724.6,
|
|
"valid_targets_min": 1273
|
|
},
|
|
{
|
|
"epoch": 5.31973539140022,
|
|
"grad_norm": 0.7533257706127008,
|
|
"learning_rate": 6.623654041284282e-06,
|
|
"loss": 0.1189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06243418902158737,
|
|
"step": 9650,
|
|
"valid_targets_mean": 1481.1,
|
|
"valid_targets_min": 806
|
|
},
|
|
{
|
|
"epoch": 5.322491730981257,
|
|
"grad_norm": 0.8812686238100234,
|
|
"learning_rate": 6.603229649988867e-06,
|
|
"loss": 0.1189,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0618022084236145,
|
|
"step": 9655,
|
|
"valid_targets_mean": 1545.4,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 5.325248070562293,
|
|
"grad_norm": 0.8095718267903595,
|
|
"learning_rate": 6.5828305691124615e-06,
|
|
"loss": 0.1229,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06148924678564072,
|
|
"step": 9660,
|
|
"valid_targets_mean": 1649.8,
|
|
"valid_targets_min": 1290
|
|
},
|
|
{
|
|
"epoch": 5.32800441014333,
|
|
"grad_norm": 0.7628463225848058,
|
|
"learning_rate": 6.562456837194903e-06,
|
|
"loss": 0.1204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05521581694483757,
|
|
"step": 9665,
|
|
"valid_targets_mean": 1363.4,
|
|
"valid_targets_min": 789
|
|
},
|
|
{
|
|
"epoch": 5.330760749724366,
|
|
"grad_norm": 0.7869410476499461,
|
|
"learning_rate": 6.54210849272813e-06,
|
|
"loss": 0.1198,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06755517423152924,
|
|
"step": 9670,
|
|
"valid_targets_mean": 1756.9,
|
|
"valid_targets_min": 589
|
|
},
|
|
{
|
|
"epoch": 5.333517089305403,
|
|
"grad_norm": 0.9293470032793892,
|
|
"learning_rate": 6.521785574156119e-06,
|
|
"loss": 0.1217,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06974601745605469,
|
|
"step": 9675,
|
|
"valid_targets_mean": 1581.2,
|
|
"valid_targets_min": 659
|
|
},
|
|
{
|
|
"epoch": 5.336273428886439,
|
|
"grad_norm": 0.8667099158879656,
|
|
"learning_rate": 6.501488119874811e-06,
|
|
"loss": 0.1183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05387250334024429,
|
|
"step": 9680,
|
|
"valid_targets_mean": 1332.0,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 5.3390297684674755,
|
|
"grad_norm": 1.189920314605054,
|
|
"learning_rate": 6.481216168232034e-06,
|
|
"loss": 0.1159,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.038235343992710114,
|
|
"step": 9685,
|
|
"valid_targets_mean": 1093.6,
|
|
"valid_targets_min": 624
|
|
},
|
|
{
|
|
"epoch": 5.341786108048511,
|
|
"grad_norm": 0.8740127018416167,
|
|
"learning_rate": 6.460969757527438e-06,
|
|
"loss": 0.1275,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.053202565759420395,
|
|
"step": 9690,
|
|
"valid_targets_mean": 1478.8,
|
|
"valid_targets_min": 883
|
|
},
|
|
{
|
|
"epoch": 5.344542447629548,
|
|
"grad_norm": 0.7905895197463041,
|
|
"learning_rate": 6.440748926012408e-06,
|
|
"loss": 0.1256,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06415604054927826,
|
|
"step": 9695,
|
|
"valid_targets_mean": 1750.6,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 5.347298787210584,
|
|
"grad_norm": 0.8884450448177483,
|
|
"learning_rate": 6.420553711890025e-06,
|
|
"loss": 0.1207,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0660836324095726,
|
|
"step": 9700,
|
|
"valid_targets_mean": 1472.2,
|
|
"valid_targets_min": 520
|
|
},
|
|
{
|
|
"epoch": 5.350055126791621,
|
|
"grad_norm": 0.7994303394989438,
|
|
"learning_rate": 6.400384153314953e-06,
|
|
"loss": 0.1208,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051337212324142456,
|
|
"step": 9705,
|
|
"valid_targets_mean": 1226.2,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 5.352811466372657,
|
|
"grad_norm": 0.8065860758040991,
|
|
"learning_rate": 6.380240288393389e-06,
|
|
"loss": 0.1146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06518785655498505,
|
|
"step": 9710,
|
|
"valid_targets_mean": 1665.5,
|
|
"valid_targets_min": 1008
|
|
},
|
|
{
|
|
"epoch": 5.355567805953694,
|
|
"grad_norm": 0.819252387484025,
|
|
"learning_rate": 6.3601221551829865e-06,
|
|
"loss": 0.1146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05535300076007843,
|
|
"step": 9715,
|
|
"valid_targets_mean": 1405.9,
|
|
"valid_targets_min": 768
|
|
},
|
|
{
|
|
"epoch": 5.35832414553473,
|
|
"grad_norm": 0.9937505933949584,
|
|
"learning_rate": 6.340029791692788e-06,
|
|
"loss": 0.1171,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04867292195558548,
|
|
"step": 9720,
|
|
"valid_targets_mean": 1294.5,
|
|
"valid_targets_min": 744
|
|
},
|
|
{
|
|
"epoch": 5.3610804851157665,
|
|
"grad_norm": 0.8930599839421799,
|
|
"learning_rate": 6.319963235883142e-06,
|
|
"loss": 0.1146,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04545201361179352,
|
|
"step": 9725,
|
|
"valid_targets_mean": 1116.0,
|
|
"valid_targets_min": 679
|
|
},
|
|
{
|
|
"epoch": 5.3638368246968025,
|
|
"grad_norm": 0.7827835241437873,
|
|
"learning_rate": 6.299922525665649e-06,
|
|
"loss": 0.1161,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07382596284151077,
|
|
"step": 9730,
|
|
"valid_targets_mean": 1898.9,
|
|
"valid_targets_min": 1340
|
|
},
|
|
{
|
|
"epoch": 5.366593164277839,
|
|
"grad_norm": 0.8891964944489181,
|
|
"learning_rate": 6.279907698903071e-06,
|
|
"loss": 0.1251,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08447828888893127,
|
|
"step": 9735,
|
|
"valid_targets_mean": 1631.6,
|
|
"valid_targets_min": 850
|
|
},
|
|
{
|
|
"epoch": 5.369349503858875,
|
|
"grad_norm": 0.7430853817963424,
|
|
"learning_rate": 6.259918793409276e-06,
|
|
"loss": 0.1162,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.050583366304636,
|
|
"step": 9740,
|
|
"valid_targets_mean": 1451.9,
|
|
"valid_targets_min": 902
|
|
},
|
|
{
|
|
"epoch": 5.372105843439912,
|
|
"grad_norm": 0.7779810544407563,
|
|
"learning_rate": 6.239955846949144e-06,
|
|
"loss": 0.1174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06361684203147888,
|
|
"step": 9745,
|
|
"valid_targets_mean": 1667.8,
|
|
"valid_targets_min": 1237
|
|
},
|
|
{
|
|
"epoch": 5.374862183020948,
|
|
"grad_norm": 0.6022705992954374,
|
|
"learning_rate": 6.220018897238538e-06,
|
|
"loss": 0.119,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0486748069524765,
|
|
"step": 9750,
|
|
"valid_targets_mean": 1690.6,
|
|
"valid_targets_min": 916
|
|
},
|
|
{
|
|
"epoch": 5.377618522601985,
|
|
"grad_norm": 0.7299956327624164,
|
|
"learning_rate": 6.200107981944181e-06,
|
|
"loss": 0.1079,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05533374845981598,
|
|
"step": 9755,
|
|
"valid_targets_mean": 1573.8,
|
|
"valid_targets_min": 925
|
|
},
|
|
{
|
|
"epoch": 5.380374862183021,
|
|
"grad_norm": 0.770828491642045,
|
|
"learning_rate": 6.180223138683619e-06,
|
|
"loss": 0.1186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06148365139961243,
|
|
"step": 9760,
|
|
"valid_targets_mean": 1649.0,
|
|
"valid_targets_min": 1220
|
|
},
|
|
{
|
|
"epoch": 5.383131201764058,
|
|
"grad_norm": 0.7857513506766097,
|
|
"learning_rate": 6.160364405025141e-06,
|
|
"loss": 0.1153,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05975348502397537,
|
|
"step": 9765,
|
|
"valid_targets_mean": 1560.5,
|
|
"valid_targets_min": 869
|
|
},
|
|
{
|
|
"epoch": 5.3858875413450935,
|
|
"grad_norm": 0.7762902561045943,
|
|
"learning_rate": 6.1405318184877e-06,
|
|
"loss": 0.1186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.060023754835128784,
|
|
"step": 9770,
|
|
"valid_targets_mean": 1559.4,
|
|
"valid_targets_min": 821
|
|
},
|
|
{
|
|
"epoch": 5.38864388092613,
|
|
"grad_norm": 0.8663436757012428,
|
|
"learning_rate": 6.120725416540861e-06,
|
|
"loss": 0.1225,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05719350278377533,
|
|
"step": 9775,
|
|
"valid_targets_mean": 1535.9,
|
|
"valid_targets_min": 1211
|
|
},
|
|
{
|
|
"epoch": 5.391400220507166,
|
|
"grad_norm": 0.787849146130452,
|
|
"learning_rate": 6.1009452366047075e-06,
|
|
"loss": 0.115,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05250891298055649,
|
|
"step": 9780,
|
|
"valid_targets_mean": 1264.5,
|
|
"valid_targets_min": 590
|
|
},
|
|
{
|
|
"epoch": 5.394156560088203,
|
|
"grad_norm": 0.8088060499873851,
|
|
"learning_rate": 6.08119131604979e-06,
|
|
"loss": 0.1186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04883202165365219,
|
|
"step": 9785,
|
|
"valid_targets_mean": 1228.1,
|
|
"valid_targets_min": 644
|
|
},
|
|
{
|
|
"epoch": 5.396912899669239,
|
|
"grad_norm": 1.1328905815584298,
|
|
"learning_rate": 6.061463692197036e-06,
|
|
"loss": 0.1847,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12135627120733261,
|
|
"step": 9790,
|
|
"valid_targets_mean": 5088.5,
|
|
"valid_targets_min": 1966
|
|
},
|
|
{
|
|
"epoch": 5.399669239250276,
|
|
"grad_norm": 0.972275824481555,
|
|
"learning_rate": 6.0417624023177145e-06,
|
|
"loss": 0.2265,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11772017180919647,
|
|
"step": 9795,
|
|
"valid_targets_mean": 5210.0,
|
|
"valid_targets_min": 1628
|
|
},
|
|
{
|
|
"epoch": 5.402425578831312,
|
|
"grad_norm": 0.691106002734909,
|
|
"learning_rate": 6.022087483633319e-06,
|
|
"loss": 0.2263,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.14130942523479462,
|
|
"step": 9800,
|
|
"valid_targets_mean": 5955.5,
|
|
"valid_targets_min": 3288
|
|
},
|
|
{
|
|
"epoch": 5.405181918412349,
|
|
"grad_norm": 0.5921400454029208,
|
|
"learning_rate": 6.0024389733155275e-06,
|
|
"loss": 0.2087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08559156954288483,
|
|
"step": 9805,
|
|
"valid_targets_mean": 4237.4,
|
|
"valid_targets_min": 2734
|
|
},
|
|
{
|
|
"epoch": 5.407938257993385,
|
|
"grad_norm": 0.5917058265959155,
|
|
"learning_rate": 5.982816908486122e-06,
|
|
"loss": 0.21,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08739405870437622,
|
|
"step": 9810,
|
|
"valid_targets_mean": 3888.1,
|
|
"valid_targets_min": 1439
|
|
},
|
|
{
|
|
"epoch": 5.410694597574421,
|
|
"grad_norm": 0.5328192106495365,
|
|
"learning_rate": 5.963221326216931e-06,
|
|
"loss": 0.2205,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08616223931312561,
|
|
"step": 9815,
|
|
"valid_targets_mean": 4472.4,
|
|
"valid_targets_min": 1512
|
|
},
|
|
{
|
|
"epoch": 5.413450937155457,
|
|
"grad_norm": 0.6389355701285906,
|
|
"learning_rate": 5.9436522635297375e-06,
|
|
"loss": 0.2034,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11099259555339813,
|
|
"step": 9820,
|
|
"valid_targets_mean": 5120.1,
|
|
"valid_targets_min": 2394
|
|
},
|
|
{
|
|
"epoch": 5.416207276736494,
|
|
"grad_norm": 0.5504443594995156,
|
|
"learning_rate": 5.924109757396228e-06,
|
|
"loss": 0.2111,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08241067826747894,
|
|
"step": 9825,
|
|
"valid_targets_mean": 3248.4,
|
|
"valid_targets_min": 1654
|
|
},
|
|
{
|
|
"epoch": 5.41896361631753,
|
|
"grad_norm": 0.48340859996590635,
|
|
"learning_rate": 5.904593844737916e-06,
|
|
"loss": 0.2071,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10209549963474274,
|
|
"step": 9830,
|
|
"valid_targets_mean": 5733.6,
|
|
"valid_targets_min": 4625
|
|
},
|
|
{
|
|
"epoch": 5.421719955898567,
|
|
"grad_norm": 0.5929679788113232,
|
|
"learning_rate": 5.885104562426067e-06,
|
|
"loss": 0.2199,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10203132778406143,
|
|
"step": 9835,
|
|
"valid_targets_mean": 4134.8,
|
|
"valid_targets_min": 1544
|
|
},
|
|
{
|
|
"epoch": 5.424476295479603,
|
|
"grad_norm": 0.47520195047207237,
|
|
"learning_rate": 5.865641947281638e-06,
|
|
"loss": 0.2084,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10137131065130234,
|
|
"step": 9840,
|
|
"valid_targets_mean": 5856.9,
|
|
"valid_targets_min": 950
|
|
},
|
|
{
|
|
"epoch": 5.42723263506064,
|
|
"grad_norm": 0.5662750102499035,
|
|
"learning_rate": 5.84620603607521e-06,
|
|
"loss": 0.2099,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11511754989624023,
|
|
"step": 9845,
|
|
"valid_targets_mean": 4966.5,
|
|
"valid_targets_min": 1263
|
|
},
|
|
{
|
|
"epoch": 5.429988974641676,
|
|
"grad_norm": 0.5202309497116531,
|
|
"learning_rate": 5.8267968655269026e-06,
|
|
"loss": 0.2067,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09611360728740692,
|
|
"step": 9850,
|
|
"valid_targets_mean": 5353.5,
|
|
"valid_targets_min": 1980
|
|
},
|
|
{
|
|
"epoch": 5.4327453142227125,
|
|
"grad_norm": 0.5730081821950609,
|
|
"learning_rate": 5.807414472306319e-06,
|
|
"loss": 0.2093,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0887514054775238,
|
|
"step": 9855,
|
|
"valid_targets_mean": 3883.9,
|
|
"valid_targets_min": 1755
|
|
},
|
|
{
|
|
"epoch": 5.435501653803748,
|
|
"grad_norm": 0.6334762620005862,
|
|
"learning_rate": 5.788058893032469e-06,
|
|
"loss": 0.2026,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09398546814918518,
|
|
"step": 9860,
|
|
"valid_targets_mean": 3182.1,
|
|
"valid_targets_min": 1451
|
|
},
|
|
{
|
|
"epoch": 5.438257993384785,
|
|
"grad_norm": 0.5864043305939266,
|
|
"learning_rate": 5.7687301642737035e-06,
|
|
"loss": 0.2027,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09907664358615875,
|
|
"step": 9865,
|
|
"valid_targets_mean": 4275.6,
|
|
"valid_targets_min": 1858
|
|
},
|
|
{
|
|
"epoch": 5.441014332965821,
|
|
"grad_norm": 0.6032554309340857,
|
|
"learning_rate": 5.749428322547652e-06,
|
|
"loss": 0.2019,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10838688910007477,
|
|
"step": 9870,
|
|
"valid_targets_mean": 4332.5,
|
|
"valid_targets_min": 2339
|
|
},
|
|
{
|
|
"epoch": 5.443770672546858,
|
|
"grad_norm": 0.5989472018294463,
|
|
"learning_rate": 5.730153404321137e-06,
|
|
"loss": 0.204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10643668472766876,
|
|
"step": 9875,
|
|
"valid_targets_mean": 3915.5,
|
|
"valid_targets_min": 2766
|
|
},
|
|
{
|
|
"epoch": 5.446527012127894,
|
|
"grad_norm": 0.6744140971640274,
|
|
"learning_rate": 5.710905446010122e-06,
|
|
"loss": 0.2108,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10740696638822556,
|
|
"step": 9880,
|
|
"valid_targets_mean": 3645.6,
|
|
"valid_targets_min": 1205
|
|
},
|
|
{
|
|
"epoch": 5.449283351708931,
|
|
"grad_norm": 0.5747196455997998,
|
|
"learning_rate": 5.691684483979629e-06,
|
|
"loss": 0.2017,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08515532314777374,
|
|
"step": 9885,
|
|
"valid_targets_mean": 3829.5,
|
|
"valid_targets_min": 2000
|
|
},
|
|
{
|
|
"epoch": 5.452039691289967,
|
|
"grad_norm": 0.6033022578015141,
|
|
"learning_rate": 5.672490554543673e-06,
|
|
"loss": 0.2085,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10003925859928131,
|
|
"step": 9890,
|
|
"valid_targets_mean": 3772.6,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 5.4547960308710035,
|
|
"grad_norm": 0.6031249971015517,
|
|
"learning_rate": 5.653323693965218e-06,
|
|
"loss": 0.1891,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11387020349502563,
|
|
"step": 9895,
|
|
"valid_targets_mean": 4300.1,
|
|
"valid_targets_min": 2297
|
|
},
|
|
{
|
|
"epoch": 5.4575523704520394,
|
|
"grad_norm": 0.6050246464531305,
|
|
"learning_rate": 5.634183938456061e-06,
|
|
"loss": 0.1919,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10624270886182785,
|
|
"step": 9900,
|
|
"valid_targets_mean": 4547.6,
|
|
"valid_targets_min": 2045
|
|
},
|
|
{
|
|
"epoch": 5.460308710033076,
|
|
"grad_norm": 0.6892655468135599,
|
|
"learning_rate": 5.615071324176806e-06,
|
|
"loss": 0.1899,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12942224740982056,
|
|
"step": 9905,
|
|
"valid_targets_mean": 4654.5,
|
|
"valid_targets_min": 3370
|
|
},
|
|
{
|
|
"epoch": 5.463065049614112,
|
|
"grad_norm": 0.621558693507702,
|
|
"learning_rate": 5.59598588723677e-06,
|
|
"loss": 0.2077,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09805111587047577,
|
|
"step": 9910,
|
|
"valid_targets_mean": 3721.4,
|
|
"valid_targets_min": 1743
|
|
},
|
|
{
|
|
"epoch": 5.465821389195149,
|
|
"grad_norm": 0.6279753805601004,
|
|
"learning_rate": 5.576927663693929e-06,
|
|
"loss": 0.2029,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0903685986995697,
|
|
"step": 9915,
|
|
"valid_targets_mean": 3319.9,
|
|
"valid_targets_min": 1569
|
|
},
|
|
{
|
|
"epoch": 5.468577728776185,
|
|
"grad_norm": 0.6033280215969534,
|
|
"learning_rate": 5.557896689554843e-06,
|
|
"loss": 0.1943,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10109879076480865,
|
|
"step": 9920,
|
|
"valid_targets_mean": 4595.6,
|
|
"valid_targets_min": 1537
|
|
},
|
|
{
|
|
"epoch": 5.471334068357222,
|
|
"grad_norm": 0.5534344484093475,
|
|
"learning_rate": 5.53889300077459e-06,
|
|
"loss": 0.1926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0870056003332138,
|
|
"step": 9925,
|
|
"valid_targets_mean": 4320.4,
|
|
"valid_targets_min": 1934
|
|
},
|
|
{
|
|
"epoch": 5.474090407938258,
|
|
"grad_norm": 0.594616630937982,
|
|
"learning_rate": 5.519916633256699e-06,
|
|
"loss": 0.1927,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08465346693992615,
|
|
"step": 9930,
|
|
"valid_targets_mean": 3487.9,
|
|
"valid_targets_min": 2556
|
|
},
|
|
{
|
|
"epoch": 5.476846747519295,
|
|
"grad_norm": 0.5998662963563667,
|
|
"learning_rate": 5.500967622853073e-06,
|
|
"loss": 0.1947,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0987597405910492,
|
|
"step": 9935,
|
|
"valid_targets_mean": 4068.8,
|
|
"valid_targets_min": 2648
|
|
},
|
|
{
|
|
"epoch": 5.4796030871003305,
|
|
"grad_norm": 0.5721463126450294,
|
|
"learning_rate": 5.482046005363952e-06,
|
|
"loss": 0.1909,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08903186023235321,
|
|
"step": 9940,
|
|
"valid_targets_mean": 4032.1,
|
|
"valid_targets_min": 1349
|
|
},
|
|
{
|
|
"epoch": 5.482359426681367,
|
|
"grad_norm": 0.6116110417557296,
|
|
"learning_rate": 5.463151816537802e-06,
|
|
"loss": 0.1935,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0779203474521637,
|
|
"step": 9945,
|
|
"valid_targets_mean": 2994.6,
|
|
"valid_targets_min": 1566
|
|
},
|
|
{
|
|
"epoch": 5.485115766262403,
|
|
"grad_norm": 0.6643722568065082,
|
|
"learning_rate": 5.4442850920712704e-06,
|
|
"loss": 0.1951,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07825697213411331,
|
|
"step": 9950,
|
|
"valid_targets_mean": 2621.4,
|
|
"valid_targets_min": 1499
|
|
},
|
|
{
|
|
"epoch": 5.48787210584344,
|
|
"grad_norm": 0.5900863657674383,
|
|
"learning_rate": 5.4254458676091245e-06,
|
|
"loss": 0.1895,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0865669995546341,
|
|
"step": 9955,
|
|
"valid_targets_mean": 3408.1,
|
|
"valid_targets_min": 2030
|
|
},
|
|
{
|
|
"epoch": 5.490628445424476,
|
|
"grad_norm": 0.7220839850898707,
|
|
"learning_rate": 5.406634178744168e-06,
|
|
"loss": 0.1967,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11085161566734314,
|
|
"step": 9960,
|
|
"valid_targets_mean": 2845.1,
|
|
"valid_targets_min": 1346
|
|
},
|
|
{
|
|
"epoch": 5.493384785005513,
|
|
"grad_norm": 0.749756368084999,
|
|
"learning_rate": 5.387850061017186e-06,
|
|
"loss": 0.1934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11232651025056839,
|
|
"step": 9965,
|
|
"valid_targets_mean": 2974.1,
|
|
"valid_targets_min": 1821
|
|
},
|
|
{
|
|
"epoch": 5.496141124586549,
|
|
"grad_norm": 0.658589524089065,
|
|
"learning_rate": 5.369093549916873e-06,
|
|
"loss": 0.1931,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10059788078069687,
|
|
"step": 9970,
|
|
"valid_targets_mean": 3943.6,
|
|
"valid_targets_min": 2525
|
|
},
|
|
{
|
|
"epoch": 5.498897464167586,
|
|
"grad_norm": 0.802644704556509,
|
|
"learning_rate": 5.350364680879769e-06,
|
|
"loss": 0.1961,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09178466349840164,
|
|
"step": 9975,
|
|
"valid_targets_mean": 3039.4,
|
|
"valid_targets_min": 2070
|
|
},
|
|
{
|
|
"epoch": 5.501653803748622,
|
|
"grad_norm": 0.6394058095224769,
|
|
"learning_rate": 5.331663489290184e-06,
|
|
"loss": 0.1935,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08715149015188217,
|
|
"step": 9980,
|
|
"valid_targets_mean": 2970.5,
|
|
"valid_targets_min": 1152
|
|
},
|
|
{
|
|
"epoch": 5.504410143329658,
|
|
"grad_norm": 0.6000355352707261,
|
|
"learning_rate": 5.312990010480137e-06,
|
|
"loss": 0.1944,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08905383944511414,
|
|
"step": 9985,
|
|
"valid_targets_mean": 3201.1,
|
|
"valid_targets_min": 1850
|
|
},
|
|
{
|
|
"epoch": 5.507166482910694,
|
|
"grad_norm": 0.6043948186264636,
|
|
"learning_rate": 5.294344279729307e-06,
|
|
"loss": 0.1941,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08876776695251465,
|
|
"step": 9990,
|
|
"valid_targets_mean": 3636.6,
|
|
"valid_targets_min": 2196
|
|
},
|
|
{
|
|
"epoch": 5.509922822491731,
|
|
"grad_norm": 0.6113851037510525,
|
|
"learning_rate": 5.2757263322649235e-06,
|
|
"loss": 0.1913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07471298426389694,
|
|
"step": 9995,
|
|
"valid_targets_mean": 3762.9,
|
|
"valid_targets_min": 1390
|
|
},
|
|
{
|
|
"epoch": 5.512679162072767,
|
|
"grad_norm": 0.6160533543529592,
|
|
"learning_rate": 5.257136203261741e-06,
|
|
"loss": 0.1979,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10310504585504532,
|
|
"step": 10000,
|
|
"valid_targets_mean": 4341.6,
|
|
"valid_targets_min": 1844
|
|
},
|
|
{
|
|
"epoch": 5.515435501653804,
|
|
"grad_norm": 0.6514496298219753,
|
|
"learning_rate": 5.2385739278419525e-06,
|
|
"loss": 0.1901,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11008217185735703,
|
|
"step": 10005,
|
|
"valid_targets_mean": 4335.2,
|
|
"valid_targets_min": 1733
|
|
},
|
|
{
|
|
"epoch": 5.51819184123484,
|
|
"grad_norm": 0.6217465868877567,
|
|
"learning_rate": 5.220039541075126e-06,
|
|
"loss": 0.1964,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10845685005187988,
|
|
"step": 10010,
|
|
"valid_targets_mean": 4243.1,
|
|
"valid_targets_min": 2065
|
|
},
|
|
{
|
|
"epoch": 5.520948180815877,
|
|
"grad_norm": 0.7482894214374928,
|
|
"learning_rate": 5.201533077978141e-06,
|
|
"loss": 0.1882,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09647095203399658,
|
|
"step": 10015,
|
|
"valid_targets_mean": 4171.9,
|
|
"valid_targets_min": 2348
|
|
},
|
|
{
|
|
"epoch": 5.523704520396913,
|
|
"grad_norm": 0.5909139309915316,
|
|
"learning_rate": 5.18305457351512e-06,
|
|
"loss": 0.191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10751405358314514,
|
|
"step": 10020,
|
|
"valid_targets_mean": 4153.1,
|
|
"valid_targets_min": 1603
|
|
},
|
|
{
|
|
"epoch": 5.5264608599779494,
|
|
"grad_norm": 0.6901040750204975,
|
|
"learning_rate": 5.164604062597369e-06,
|
|
"loss": 0.1976,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10262557864189148,
|
|
"step": 10025,
|
|
"valid_targets_mean": 3267.9,
|
|
"valid_targets_min": 1377
|
|
},
|
|
{
|
|
"epoch": 5.529217199558985,
|
|
"grad_norm": 0.6118009475125896,
|
|
"learning_rate": 5.1461815800832935e-06,
|
|
"loss": 0.1966,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0883086621761322,
|
|
"step": 10030,
|
|
"valid_targets_mean": 3298.8,
|
|
"valid_targets_min": 1902
|
|
},
|
|
{
|
|
"epoch": 5.531973539140022,
|
|
"grad_norm": 0.5967135956869002,
|
|
"learning_rate": 5.127787160778366e-06,
|
|
"loss": 0.1919,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08586828410625458,
|
|
"step": 10035,
|
|
"valid_targets_mean": 3316.8,
|
|
"valid_targets_min": 653
|
|
},
|
|
{
|
|
"epoch": 5.534729878721058,
|
|
"grad_norm": 0.6150440913260063,
|
|
"learning_rate": 5.109420839435018e-06,
|
|
"loss": 0.1965,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09655844420194626,
|
|
"step": 10040,
|
|
"valid_targets_mean": 3900.8,
|
|
"valid_targets_min": 2593
|
|
},
|
|
{
|
|
"epoch": 5.537486218302095,
|
|
"grad_norm": 0.5976269747229416,
|
|
"learning_rate": 5.0910826507526055e-06,
|
|
"loss": 0.1975,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10288942605257034,
|
|
"step": 10045,
|
|
"valid_targets_mean": 4461.6,
|
|
"valid_targets_min": 1736
|
|
},
|
|
{
|
|
"epoch": 5.540242557883131,
|
|
"grad_norm": 0.6279404787182344,
|
|
"learning_rate": 5.072772629377334e-06,
|
|
"loss": 0.1918,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0845770463347435,
|
|
"step": 10050,
|
|
"valid_targets_mean": 3218.8,
|
|
"valid_targets_min": 1137
|
|
},
|
|
{
|
|
"epoch": 5.542998897464168,
|
|
"grad_norm": 0.5885042939080697,
|
|
"learning_rate": 5.054490809902188e-06,
|
|
"loss": 0.1799,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08962637931108475,
|
|
"step": 10055,
|
|
"valid_targets_mean": 3535.5,
|
|
"valid_targets_min": 2500
|
|
},
|
|
{
|
|
"epoch": 5.545755237045204,
|
|
"grad_norm": 0.6014634715472776,
|
|
"learning_rate": 5.03623722686688e-06,
|
|
"loss": 0.1963,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.098706915974617,
|
|
"step": 10060,
|
|
"valid_targets_mean": 4358.2,
|
|
"valid_targets_min": 1297
|
|
},
|
|
{
|
|
"epoch": 5.5485115766262405,
|
|
"grad_norm": 0.6054938183235176,
|
|
"learning_rate": 5.018011914757763e-06,
|
|
"loss": 0.1805,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10142864286899567,
|
|
"step": 10065,
|
|
"valid_targets_mean": 4368.6,
|
|
"valid_targets_min": 1101
|
|
},
|
|
{
|
|
"epoch": 5.551267916207276,
|
|
"grad_norm": 0.6263778010422008,
|
|
"learning_rate": 4.999814908007787e-06,
|
|
"loss": 0.2057,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10816408693790436,
|
|
"step": 10070,
|
|
"valid_targets_mean": 3886.8,
|
|
"valid_targets_min": 2078
|
|
},
|
|
{
|
|
"epoch": 5.554024255788313,
|
|
"grad_norm": 0.6019254386141223,
|
|
"learning_rate": 4.981646240996421e-06,
|
|
"loss": 0.1908,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08021493256092072,
|
|
"step": 10075,
|
|
"valid_targets_mean": 3047.1,
|
|
"valid_targets_min": 1739
|
|
},
|
|
{
|
|
"epoch": 5.556780595369349,
|
|
"grad_norm": 0.6141997144730528,
|
|
"learning_rate": 4.963505948049587e-06,
|
|
"loss": 0.1972,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10275082290172577,
|
|
"step": 10080,
|
|
"valid_targets_mean": 3684.5,
|
|
"valid_targets_min": 1233
|
|
},
|
|
{
|
|
"epoch": 5.559536934950386,
|
|
"grad_norm": 0.6688127358510599,
|
|
"learning_rate": 4.945394063439621e-06,
|
|
"loss": 0.2014,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10375227779150009,
|
|
"step": 10085,
|
|
"valid_targets_mean": 3783.5,
|
|
"valid_targets_min": 1815
|
|
},
|
|
{
|
|
"epoch": 5.562293274531422,
|
|
"grad_norm": 0.5987668841954898,
|
|
"learning_rate": 4.927310621385162e-06,
|
|
"loss": 0.1998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09836991131305695,
|
|
"step": 10090,
|
|
"valid_targets_mean": 4099.9,
|
|
"valid_targets_min": 1649
|
|
},
|
|
{
|
|
"epoch": 5.565049614112459,
|
|
"grad_norm": 0.6574085280393294,
|
|
"learning_rate": 4.909255656051126e-06,
|
|
"loss": 0.1932,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10386388003826141,
|
|
"step": 10095,
|
|
"valid_targets_mean": 3281.1,
|
|
"valid_targets_min": 1366
|
|
},
|
|
{
|
|
"epoch": 5.567805953693495,
|
|
"grad_norm": 0.7602086354724722,
|
|
"learning_rate": 4.891229201548626e-06,
|
|
"loss": 0.1831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09485702216625214,
|
|
"step": 10100,
|
|
"valid_targets_mean": 4390.2,
|
|
"valid_targets_min": 2716
|
|
},
|
|
{
|
|
"epoch": 5.570562293274532,
|
|
"grad_norm": 0.6095414353337042,
|
|
"learning_rate": 4.873231291934905e-06,
|
|
"loss": 0.187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10220303386449814,
|
|
"step": 10105,
|
|
"valid_targets_mean": 4261.9,
|
|
"valid_targets_min": 1955
|
|
},
|
|
{
|
|
"epoch": 5.5733186328555675,
|
|
"grad_norm": 0.7410685454533948,
|
|
"learning_rate": 4.855261961213285e-06,
|
|
"loss": 0.1834,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10195191204547882,
|
|
"step": 10110,
|
|
"valid_targets_mean": 3709.5,
|
|
"valid_targets_min": 1561
|
|
},
|
|
{
|
|
"epoch": 5.576074972436604,
|
|
"grad_norm": 0.6375859092904083,
|
|
"learning_rate": 4.8373212433330884e-06,
|
|
"loss": 0.1812,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09476044774055481,
|
|
"step": 10115,
|
|
"valid_targets_mean": 3744.2,
|
|
"valid_targets_min": 1442
|
|
},
|
|
{
|
|
"epoch": 5.57883131201764,
|
|
"grad_norm": 0.6063940942093722,
|
|
"learning_rate": 4.819409172189578e-06,
|
|
"loss": 0.1732,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07704910635948181,
|
|
"step": 10120,
|
|
"valid_targets_mean": 3034.4,
|
|
"valid_targets_min": 1501
|
|
},
|
|
{
|
|
"epoch": 5.581587651598677,
|
|
"grad_norm": 0.6334481911733573,
|
|
"learning_rate": 4.801525781623899e-06,
|
|
"loss": 0.1715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08486004173755646,
|
|
"step": 10125,
|
|
"valid_targets_mean": 3149.1,
|
|
"valid_targets_min": 857
|
|
},
|
|
{
|
|
"epoch": 5.584343991179713,
|
|
"grad_norm": 0.6860046922750456,
|
|
"learning_rate": 4.7836711054230025e-06,
|
|
"loss": 0.1897,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10492058843374252,
|
|
"step": 10130,
|
|
"valid_targets_mean": 4058.2,
|
|
"valid_targets_min": 1285
|
|
},
|
|
{
|
|
"epoch": 5.58710033076075,
|
|
"grad_norm": 0.7146483030374899,
|
|
"learning_rate": 4.765845177319608e-06,
|
|
"loss": 0.2048,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08972454071044922,
|
|
"step": 10135,
|
|
"valid_targets_mean": 3909.1,
|
|
"valid_targets_min": 1530
|
|
},
|
|
{
|
|
"epoch": 5.589856670341786,
|
|
"grad_norm": 0.6593020123674287,
|
|
"learning_rate": 4.748048030992105e-06,
|
|
"loss": 0.1842,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10301342606544495,
|
|
"step": 10140,
|
|
"valid_targets_mean": 4228.5,
|
|
"valid_targets_min": 2964
|
|
},
|
|
{
|
|
"epoch": 5.592613009922823,
|
|
"grad_norm": 0.6445971526652748,
|
|
"learning_rate": 4.730279700064508e-06,
|
|
"loss": 0.1922,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10523516684770584,
|
|
"step": 10145,
|
|
"valid_targets_mean": 4405.6,
|
|
"valid_targets_min": 2010
|
|
},
|
|
{
|
|
"epoch": 5.595369349503859,
|
|
"grad_norm": 0.7403042902943268,
|
|
"learning_rate": 4.712540218106394e-06,
|
|
"loss": 0.1884,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10373927652835846,
|
|
"step": 10150,
|
|
"valid_targets_mean": 3541.2,
|
|
"valid_targets_min": 1731
|
|
},
|
|
{
|
|
"epoch": 5.598125689084895,
|
|
"grad_norm": 0.6367924843507031,
|
|
"learning_rate": 4.694829618632834e-06,
|
|
"loss": 0.1921,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08654695749282837,
|
|
"step": 10155,
|
|
"valid_targets_mean": 4123.8,
|
|
"valid_targets_min": 1911
|
|
},
|
|
{
|
|
"epoch": 5.600882028665931,
|
|
"grad_norm": 0.627867553328428,
|
|
"learning_rate": 4.677147935104334e-06,
|
|
"loss": 0.1877,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07895289361476898,
|
|
"step": 10160,
|
|
"valid_targets_mean": 3818.8,
|
|
"valid_targets_min": 2408
|
|
},
|
|
{
|
|
"epoch": 5.603638368246968,
|
|
"grad_norm": 0.5876658370680258,
|
|
"learning_rate": 4.659495200926767e-06,
|
|
"loss": 0.1756,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09946301579475403,
|
|
"step": 10165,
|
|
"valid_targets_mean": 4187.2,
|
|
"valid_targets_min": 1927
|
|
},
|
|
{
|
|
"epoch": 5.606394707828004,
|
|
"grad_norm": 0.6788479941163675,
|
|
"learning_rate": 4.6418714494513115e-06,
|
|
"loss": 0.1947,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10173513740301132,
|
|
"step": 10170,
|
|
"valid_targets_mean": 3147.5,
|
|
"valid_targets_min": 1643
|
|
},
|
|
{
|
|
"epoch": 5.609151047409041,
|
|
"grad_norm": 0.6319695855684526,
|
|
"learning_rate": 4.624276713974389e-06,
|
|
"loss": 0.1808,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08137942850589752,
|
|
"step": 10175,
|
|
"valid_targets_mean": 3082.9,
|
|
"valid_targets_min": 1459
|
|
},
|
|
{
|
|
"epoch": 5.611907386990077,
|
|
"grad_norm": 0.6883560656883089,
|
|
"learning_rate": 4.60671102773761e-06,
|
|
"loss": 0.1924,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09578832983970642,
|
|
"step": 10180,
|
|
"valid_targets_mean": 3235.0,
|
|
"valid_targets_min": 1425
|
|
},
|
|
{
|
|
"epoch": 5.614663726571114,
|
|
"grad_norm": 0.6570882902178451,
|
|
"learning_rate": 4.589174423927692e-06,
|
|
"loss": 0.1944,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07798336446285248,
|
|
"step": 10185,
|
|
"valid_targets_mean": 2764.4,
|
|
"valid_targets_min": 537
|
|
},
|
|
{
|
|
"epoch": 5.61742006615215,
|
|
"grad_norm": 0.6047643203555934,
|
|
"learning_rate": 4.571666935676411e-06,
|
|
"loss": 0.1935,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09850729256868362,
|
|
"step": 10190,
|
|
"valid_targets_mean": 4477.2,
|
|
"valid_targets_min": 1645
|
|
},
|
|
{
|
|
"epoch": 5.620176405733186,
|
|
"grad_norm": 0.7086546667443888,
|
|
"learning_rate": 4.554188596060532e-06,
|
|
"loss": 0.2026,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11186853051185608,
|
|
"step": 10195,
|
|
"valid_targets_mean": 3753.4,
|
|
"valid_targets_min": 2165
|
|
},
|
|
{
|
|
"epoch": 5.622932745314222,
|
|
"grad_norm": 0.7009525085878983,
|
|
"learning_rate": 4.536739438101756e-06,
|
|
"loss": 0.1865,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08408283442258835,
|
|
"step": 10200,
|
|
"valid_targets_mean": 3160.9,
|
|
"valid_targets_min": 1623
|
|
},
|
|
{
|
|
"epoch": 5.625689084895259,
|
|
"grad_norm": 0.6137424696550252,
|
|
"learning_rate": 4.519319494766645e-06,
|
|
"loss": 0.2019,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08197230100631714,
|
|
"step": 10205,
|
|
"valid_targets_mean": 3442.0,
|
|
"valid_targets_min": 1996
|
|
},
|
|
{
|
|
"epoch": 5.628445424476295,
|
|
"grad_norm": 0.6849828469654942,
|
|
"learning_rate": 4.501928798966573e-06,
|
|
"loss": 0.1947,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1193176656961441,
|
|
"step": 10210,
|
|
"valid_targets_mean": 4399.0,
|
|
"valid_targets_min": 2938
|
|
},
|
|
{
|
|
"epoch": 5.631201764057332,
|
|
"grad_norm": 0.5755172896053252,
|
|
"learning_rate": 4.484567383557648e-06,
|
|
"loss": 0.1936,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10701175779104233,
|
|
"step": 10215,
|
|
"valid_targets_mean": 5350.9,
|
|
"valid_targets_min": 2644
|
|
},
|
|
{
|
|
"epoch": 5.633958103638368,
|
|
"grad_norm": 0.6260171536707625,
|
|
"learning_rate": 4.467235281340667e-06,
|
|
"loss": 0.1961,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10742195695638657,
|
|
"step": 10220,
|
|
"valid_targets_mean": 4307.1,
|
|
"valid_targets_min": 1377
|
|
},
|
|
{
|
|
"epoch": 5.636714443219405,
|
|
"grad_norm": 0.6549673263758611,
|
|
"learning_rate": 4.449932525061036e-06,
|
|
"loss": 0.1903,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07975347340106964,
|
|
"step": 10225,
|
|
"valid_targets_mean": 3389.5,
|
|
"valid_targets_min": 636
|
|
},
|
|
{
|
|
"epoch": 5.639470782800441,
|
|
"grad_norm": 0.6490995591133779,
|
|
"learning_rate": 4.432659147408736e-06,
|
|
"loss": 0.1906,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08237628638744354,
|
|
"step": 10230,
|
|
"valid_targets_mean": 3251.0,
|
|
"valid_targets_min": 718
|
|
},
|
|
{
|
|
"epoch": 5.6422271223814775,
|
|
"grad_norm": 0.6034390542203207,
|
|
"learning_rate": 4.415415181018223e-06,
|
|
"loss": 0.1944,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08380737900733948,
|
|
"step": 10235,
|
|
"valid_targets_mean": 4718.9,
|
|
"valid_targets_min": 3205
|
|
},
|
|
{
|
|
"epoch": 5.644983461962513,
|
|
"grad_norm": 0.676382642566022,
|
|
"learning_rate": 4.398200658468397e-06,
|
|
"loss": 0.1871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10002248734235764,
|
|
"step": 10240,
|
|
"valid_targets_mean": 3768.5,
|
|
"valid_targets_min": 1369
|
|
},
|
|
{
|
|
"epoch": 5.64773980154355,
|
|
"grad_norm": 0.6772254319406622,
|
|
"learning_rate": 4.381015612282529e-06,
|
|
"loss": 0.1798,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11036770045757294,
|
|
"step": 10245,
|
|
"valid_targets_mean": 4258.1,
|
|
"valid_targets_min": 2762
|
|
},
|
|
{
|
|
"epoch": 5.650496141124586,
|
|
"grad_norm": 0.6674387711843542,
|
|
"learning_rate": 4.363860074928197e-06,
|
|
"loss": 0.187,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07170131057500839,
|
|
"step": 10250,
|
|
"valid_targets_mean": 2535.5,
|
|
"valid_targets_min": 529
|
|
},
|
|
{
|
|
"epoch": 5.653252480705623,
|
|
"grad_norm": 0.6436288283739613,
|
|
"learning_rate": 4.346734078817234e-06,
|
|
"loss": 0.1836,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09613151103258133,
|
|
"step": 10255,
|
|
"valid_targets_mean": 3895.6,
|
|
"valid_targets_min": 1832
|
|
},
|
|
{
|
|
"epoch": 5.656008820286659,
|
|
"grad_norm": 0.6618387632926996,
|
|
"learning_rate": 4.329637656305654e-06,
|
|
"loss": 0.1951,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08921610563993454,
|
|
"step": 10260,
|
|
"valid_targets_mean": 3470.2,
|
|
"valid_targets_min": 1726
|
|
},
|
|
{
|
|
"epoch": 5.658765159867696,
|
|
"grad_norm": 0.6978096121637685,
|
|
"learning_rate": 4.312570839693606e-06,
|
|
"loss": 0.1899,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10754061490297318,
|
|
"step": 10265,
|
|
"valid_targets_mean": 3707.2,
|
|
"valid_targets_min": 1361
|
|
},
|
|
{
|
|
"epoch": 5.661521499448732,
|
|
"grad_norm": 0.6025489149557104,
|
|
"learning_rate": 4.295533661225295e-06,
|
|
"loss": 0.1892,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09673164039850235,
|
|
"step": 10270,
|
|
"valid_targets_mean": 3686.8,
|
|
"valid_targets_min": 851
|
|
},
|
|
{
|
|
"epoch": 5.664277839029769,
|
|
"grad_norm": 0.6630044542534439,
|
|
"learning_rate": 4.2785261530889335e-06,
|
|
"loss": 0.1883,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08534131944179535,
|
|
"step": 10275,
|
|
"valid_targets_mean": 2817.6,
|
|
"valid_targets_min": 1190
|
|
},
|
|
{
|
|
"epoch": 5.6670341786108045,
|
|
"grad_norm": 0.6632047024011952,
|
|
"learning_rate": 4.261548347416691e-06,
|
|
"loss": 0.191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11288687586784363,
|
|
"step": 10280,
|
|
"valid_targets_mean": 4741.0,
|
|
"valid_targets_min": 1394
|
|
},
|
|
{
|
|
"epoch": 5.669790518191841,
|
|
"grad_norm": 0.6929519406558764,
|
|
"learning_rate": 4.244600276284607e-06,
|
|
"loss": 0.1868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07848985493183136,
|
|
"step": 10285,
|
|
"valid_targets_mean": 3478.5,
|
|
"valid_targets_min": 1972
|
|
},
|
|
{
|
|
"epoch": 5.672546857772877,
|
|
"grad_norm": 0.6591803585838358,
|
|
"learning_rate": 4.227681971712543e-06,
|
|
"loss": 0.1781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10063079744577408,
|
|
"step": 10290,
|
|
"valid_targets_mean": 4379.2,
|
|
"valid_targets_min": 2550
|
|
},
|
|
{
|
|
"epoch": 5.675303197353914,
|
|
"grad_norm": 0.5597973349779887,
|
|
"learning_rate": 4.210793465664127e-06,
|
|
"loss": 0.1827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08939698338508606,
|
|
"step": 10295,
|
|
"valid_targets_mean": 6398.4,
|
|
"valid_targets_min": 2588
|
|
},
|
|
{
|
|
"epoch": 5.67805953693495,
|
|
"grad_norm": 0.750220849867065,
|
|
"learning_rate": 4.19393479004669e-06,
|
|
"loss": 0.1827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09528659284114838,
|
|
"step": 10300,
|
|
"valid_targets_mean": 4008.5,
|
|
"valid_targets_min": 2867
|
|
},
|
|
{
|
|
"epoch": 5.680815876515987,
|
|
"grad_norm": 0.6518929248059617,
|
|
"learning_rate": 4.177105976711204e-06,
|
|
"loss": 0.1759,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0778331384062767,
|
|
"step": 10305,
|
|
"valid_targets_mean": 2879.9,
|
|
"valid_targets_min": 992
|
|
},
|
|
{
|
|
"epoch": 5.683572216097023,
|
|
"grad_norm": 0.6199157582567867,
|
|
"learning_rate": 4.160307057452218e-06,
|
|
"loss": 0.1762,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08418752998113632,
|
|
"step": 10310,
|
|
"valid_targets_mean": 3798.8,
|
|
"valid_targets_min": 1573
|
|
},
|
|
{
|
|
"epoch": 5.68632855567806,
|
|
"grad_norm": 0.6664431235228347,
|
|
"learning_rate": 4.143538064007808e-06,
|
|
"loss": 0.1781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1024160385131836,
|
|
"step": 10315,
|
|
"valid_targets_mean": 3714.9,
|
|
"valid_targets_min": 1357
|
|
},
|
|
{
|
|
"epoch": 5.6890848952590956,
|
|
"grad_norm": 0.6018795419204901,
|
|
"learning_rate": 4.126799028059505e-06,
|
|
"loss": 0.1788,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08363612741231918,
|
|
"step": 10320,
|
|
"valid_targets_mean": 3661.1,
|
|
"valid_targets_min": 2409
|
|
},
|
|
{
|
|
"epoch": 5.691841234840132,
|
|
"grad_norm": 0.6521163719212921,
|
|
"learning_rate": 4.110089981232253e-06,
|
|
"loss": 0.1942,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10538972169160843,
|
|
"step": 10325,
|
|
"valid_targets_mean": 4322.4,
|
|
"valid_targets_min": 1452
|
|
},
|
|
{
|
|
"epoch": 5.694597574421168,
|
|
"grad_norm": 0.6468917851518979,
|
|
"learning_rate": 4.093410955094328e-06,
|
|
"loss": 0.1827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08858339488506317,
|
|
"step": 10330,
|
|
"valid_targets_mean": 3613.9,
|
|
"valid_targets_min": 1708
|
|
},
|
|
{
|
|
"epoch": 5.697353914002205,
|
|
"grad_norm": 0.7134722440243644,
|
|
"learning_rate": 4.076761981157289e-06,
|
|
"loss": 0.1824,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09580822288990021,
|
|
"step": 10335,
|
|
"valid_targets_mean": 3097.6,
|
|
"valid_targets_min": 1272
|
|
},
|
|
{
|
|
"epoch": 5.700110253583241,
|
|
"grad_norm": 0.7358830163442455,
|
|
"learning_rate": 4.060143090875921e-06,
|
|
"loss": 0.1749,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09806231409311295,
|
|
"step": 10340,
|
|
"valid_targets_mean": 3143.1,
|
|
"valid_targets_min": 1858
|
|
},
|
|
{
|
|
"epoch": 5.702866593164278,
|
|
"grad_norm": 0.8817895344341007,
|
|
"learning_rate": 4.043554315648166e-06,
|
|
"loss": 0.1885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0877530574798584,
|
|
"step": 10345,
|
|
"valid_targets_mean": 2977.6,
|
|
"valid_targets_min": 627
|
|
},
|
|
{
|
|
"epoch": 5.705622932745314,
|
|
"grad_norm": 0.7389368386272487,
|
|
"learning_rate": 4.026995686815081e-06,
|
|
"loss": 0.1837,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10469523072242737,
|
|
"step": 10350,
|
|
"valid_targets_mean": 2598.8,
|
|
"valid_targets_min": 1036
|
|
},
|
|
{
|
|
"epoch": 5.708379272326351,
|
|
"grad_norm": 0.6430671942957336,
|
|
"learning_rate": 4.010467235660753e-06,
|
|
"loss": 0.1897,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09264027327299118,
|
|
"step": 10355,
|
|
"valid_targets_mean": 4311.2,
|
|
"valid_targets_min": 2524
|
|
},
|
|
{
|
|
"epoch": 5.711135611907387,
|
|
"grad_norm": 0.6432509142190611,
|
|
"learning_rate": 3.9939689934122694e-06,
|
|
"loss": 0.185,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10727041959762573,
|
|
"step": 10360,
|
|
"valid_targets_mean": 3706.4,
|
|
"valid_targets_min": 809
|
|
},
|
|
{
|
|
"epoch": 5.713891951488423,
|
|
"grad_norm": 0.6448450109534186,
|
|
"learning_rate": 3.977500991239631e-06,
|
|
"loss": 0.1852,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0832073912024498,
|
|
"step": 10365,
|
|
"valid_targets_mean": 3445.4,
|
|
"valid_targets_min": 1778
|
|
},
|
|
{
|
|
"epoch": 5.716648291069459,
|
|
"grad_norm": 0.6486002978438167,
|
|
"learning_rate": 3.961063260255713e-06,
|
|
"loss": 0.1903,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10102313756942749,
|
|
"step": 10370,
|
|
"valid_targets_mean": 4350.0,
|
|
"valid_targets_min": 2181
|
|
},
|
|
{
|
|
"epoch": 5.719404630650496,
|
|
"grad_norm": 0.49260370560477407,
|
|
"learning_rate": 3.944655831516204e-06,
|
|
"loss": 0.1561,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.040023479610681534,
|
|
"step": 10375,
|
|
"valid_targets_mean": 2416.5,
|
|
"valid_targets_min": 872
|
|
},
|
|
{
|
|
"epoch": 5.722160970231532,
|
|
"grad_norm": 0.5110183251845707,
|
|
"learning_rate": 3.928278736019538e-06,
|
|
"loss": 0.095,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03305526822805405,
|
|
"step": 10380,
|
|
"valid_targets_mean": 1543.9,
|
|
"valid_targets_min": 665
|
|
},
|
|
{
|
|
"epoch": 5.724917309812569,
|
|
"grad_norm": 0.38330435025000503,
|
|
"learning_rate": 3.911932004706836e-06,
|
|
"loss": 0.0727,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041076187044382095,
|
|
"step": 10385,
|
|
"valid_targets_mean": 3129.8,
|
|
"valid_targets_min": 2254
|
|
},
|
|
{
|
|
"epoch": 5.727673649393605,
|
|
"grad_norm": 0.4057825458545097,
|
|
"learning_rate": 3.895615668461856e-06,
|
|
"loss": 0.0724,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034805674105882645,
|
|
"step": 10390,
|
|
"valid_targets_mean": 3026.8,
|
|
"valid_targets_min": 706
|
|
},
|
|
{
|
|
"epoch": 5.730429988974642,
|
|
"grad_norm": 0.33952842924926496,
|
|
"learning_rate": 3.8793297581109325e-06,
|
|
"loss": 0.0855,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02664371207356453,
|
|
"step": 10395,
|
|
"valid_targets_mean": 3207.8,
|
|
"valid_targets_min": 1216
|
|
},
|
|
{
|
|
"epoch": 5.733186328555679,
|
|
"grad_norm": 0.38470094984455816,
|
|
"learning_rate": 3.863074304422918e-06,
|
|
"loss": 0.0741,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0413883700966835,
|
|
"step": 10400,
|
|
"valid_targets_mean": 3540.0,
|
|
"valid_targets_min": 2551
|
|
},
|
|
{
|
|
"epoch": 5.7359426681367145,
|
|
"grad_norm": 0.8635968638296525,
|
|
"learning_rate": 3.846849338109116e-06,
|
|
"loss": 0.1083,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05321991443634033,
|
|
"step": 10405,
|
|
"valid_targets_mean": 960.9,
|
|
"valid_targets_min": 555
|
|
},
|
|
{
|
|
"epoch": 5.73869900771775,
|
|
"grad_norm": 0.37487083761778484,
|
|
"learning_rate": 3.830654889823238e-06,
|
|
"loss": 0.0998,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03464626520872116,
|
|
"step": 10410,
|
|
"valid_targets_mean": 3465.4,
|
|
"valid_targets_min": 1008
|
|
},
|
|
{
|
|
"epoch": 5.741455347298787,
|
|
"grad_norm": 0.40898274809106727,
|
|
"learning_rate": 3.8144909901613303e-06,
|
|
"loss": 0.0833,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042249955236911774,
|
|
"step": 10415,
|
|
"valid_targets_mean": 3522.1,
|
|
"valid_targets_min": 3039
|
|
},
|
|
{
|
|
"epoch": 5.744211686879824,
|
|
"grad_norm": 0.5645394833814121,
|
|
"learning_rate": 3.7983576696617364e-06,
|
|
"loss": 0.1145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06845027208328247,
|
|
"step": 10420,
|
|
"valid_targets_mean": 3356.1,
|
|
"valid_targets_min": 1938
|
|
},
|
|
{
|
|
"epoch": 5.74696802646086,
|
|
"grad_norm": 0.36567641052698313,
|
|
"learning_rate": 3.7822549588050185e-06,
|
|
"loss": 0.063,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03703555092215538,
|
|
"step": 10425,
|
|
"valid_targets_mean": 3973.6,
|
|
"valid_targets_min": 924
|
|
},
|
|
{
|
|
"epoch": 5.749724366041896,
|
|
"grad_norm": 0.4471986682392373,
|
|
"learning_rate": 3.766182888013905e-06,
|
|
"loss": 0.0739,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03784479573369026,
|
|
"step": 10430,
|
|
"valid_targets_mean": 2262.8,
|
|
"valid_targets_min": 444
|
|
},
|
|
{
|
|
"epoch": 5.752480705622933,
|
|
"grad_norm": 0.3618426483252103,
|
|
"learning_rate": 3.7501414876532425e-06,
|
|
"loss": 0.0737,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03629196435213089,
|
|
"step": 10435,
|
|
"valid_targets_mean": 3298.9,
|
|
"valid_targets_min": 749
|
|
},
|
|
{
|
|
"epoch": 5.75523704520397,
|
|
"grad_norm": 0.527136486305893,
|
|
"learning_rate": 3.7341307880299306e-06,
|
|
"loss": 0.0937,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04867885261774063,
|
|
"step": 10440,
|
|
"valid_targets_mean": 1400.5,
|
|
"valid_targets_min": 601
|
|
},
|
|
{
|
|
"epoch": 5.7579933847850056,
|
|
"grad_norm": 0.4650791264406412,
|
|
"learning_rate": 3.718150819392863e-06,
|
|
"loss": 0.0898,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03798801079392433,
|
|
"step": 10445,
|
|
"valid_targets_mean": 3387.0,
|
|
"valid_targets_min": 1852
|
|
},
|
|
{
|
|
"epoch": 5.7607497243660415,
|
|
"grad_norm": 0.3910132898379467,
|
|
"learning_rate": 3.7022016119328763e-06,
|
|
"loss": 0.0726,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03286594897508621,
|
|
"step": 10450,
|
|
"valid_targets_mean": 2827.2,
|
|
"valid_targets_min": 571
|
|
},
|
|
{
|
|
"epoch": 5.763506063947078,
|
|
"grad_norm": 0.4860656150805912,
|
|
"learning_rate": 3.686283195782692e-06,
|
|
"loss": 0.0708,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03288431465625763,
|
|
"step": 10455,
|
|
"valid_targets_mean": 2584.8,
|
|
"valid_targets_min": 520
|
|
},
|
|
{
|
|
"epoch": 5.766262403528115,
|
|
"grad_norm": 0.5301458510262297,
|
|
"learning_rate": 3.670395601016854e-06,
|
|
"loss": 0.0739,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04883969575166702,
|
|
"step": 10460,
|
|
"valid_targets_mean": 3477.0,
|
|
"valid_targets_min": 2179
|
|
},
|
|
{
|
|
"epoch": 5.769018743109151,
|
|
"grad_norm": 0.3877088097399488,
|
|
"learning_rate": 3.654538857651675e-06,
|
|
"loss": 0.077,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030916331335902214,
|
|
"step": 10465,
|
|
"valid_targets_mean": 3043.8,
|
|
"valid_targets_min": 807
|
|
},
|
|
{
|
|
"epoch": 5.771775082690187,
|
|
"grad_norm": 0.4410674734771966,
|
|
"learning_rate": 3.638712995645188e-06,
|
|
"loss": 0.0836,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03403722122311592,
|
|
"step": 10470,
|
|
"valid_targets_mean": 1830.4,
|
|
"valid_targets_min": 594
|
|
},
|
|
{
|
|
"epoch": 5.774531422271224,
|
|
"grad_norm": 0.5129723696336134,
|
|
"learning_rate": 3.622918044897077e-06,
|
|
"loss": 0.0828,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04141794517636299,
|
|
"step": 10475,
|
|
"valid_targets_mean": 2210.0,
|
|
"valid_targets_min": 714
|
|
},
|
|
{
|
|
"epoch": 5.777287761852261,
|
|
"grad_norm": 0.5331083728308998,
|
|
"learning_rate": 3.6071540352486234e-06,
|
|
"loss": 0.1025,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09520633518695831,
|
|
"step": 10480,
|
|
"valid_targets_mean": 3163.6,
|
|
"valid_targets_min": 1290
|
|
},
|
|
{
|
|
"epoch": 5.780044101433297,
|
|
"grad_norm": 0.47054880116045456,
|
|
"learning_rate": 3.591420996482655e-06,
|
|
"loss": 0.0915,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03334933519363403,
|
|
"step": 10485,
|
|
"valid_targets_mean": 2710.6,
|
|
"valid_targets_min": 782
|
|
},
|
|
{
|
|
"epoch": 5.7828004410143325,
|
|
"grad_norm": 0.6770504698182713,
|
|
"learning_rate": 3.5757189583234817e-06,
|
|
"loss": 0.0742,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0567297488451004,
|
|
"step": 10490,
|
|
"valid_targets_mean": 1954.5,
|
|
"valid_targets_min": 950
|
|
},
|
|
{
|
|
"epoch": 5.785556780595369,
|
|
"grad_norm": 0.42485796243632007,
|
|
"learning_rate": 3.5600479504368533e-06,
|
|
"loss": 0.1031,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04026125371456146,
|
|
"step": 10495,
|
|
"valid_targets_mean": 3065.0,
|
|
"valid_targets_min": 1296
|
|
},
|
|
{
|
|
"epoch": 5.788313120176406,
|
|
"grad_norm": 0.41805083680206595,
|
|
"learning_rate": 3.5444080024298844e-06,
|
|
"loss": 0.0908,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028698088601231575,
|
|
"step": 10500,
|
|
"valid_targets_mean": 3146.6,
|
|
"valid_targets_min": 789
|
|
},
|
|
{
|
|
"epoch": 5.791069459757442,
|
|
"grad_norm": 0.5496355189359563,
|
|
"learning_rate": 3.528799143851016e-06,
|
|
"loss": 0.0797,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04778962582349777,
|
|
"step": 10505,
|
|
"valid_targets_mean": 2340.4,
|
|
"valid_targets_min": 713
|
|
},
|
|
{
|
|
"epoch": 5.793825799338478,
|
|
"grad_norm": 0.5809346207927943,
|
|
"learning_rate": 3.5132214041899482e-06,
|
|
"loss": 0.2552,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10462389886379242,
|
|
"step": 10510,
|
|
"valid_targets_mean": 2638.0,
|
|
"valid_targets_min": 980
|
|
},
|
|
{
|
|
"epoch": 5.796582138919515,
|
|
"grad_norm": 0.5940401314260293,
|
|
"learning_rate": 3.497674812877583e-06,
|
|
"loss": 0.095,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07489131391048431,
|
|
"step": 10515,
|
|
"valid_targets_mean": 1792.1,
|
|
"valid_targets_min": 753
|
|
},
|
|
{
|
|
"epoch": 5.799338478500552,
|
|
"grad_norm": 0.46859125369741433,
|
|
"learning_rate": 3.4821593992859893e-06,
|
|
"loss": 0.0809,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.032329168170690536,
|
|
"step": 10520,
|
|
"valid_targets_mean": 1459.5,
|
|
"valid_targets_min": 670
|
|
},
|
|
{
|
|
"epoch": 5.802094818081588,
|
|
"grad_norm": 0.5409103108117076,
|
|
"learning_rate": 3.4666751927283217e-06,
|
|
"loss": 0.1696,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042465876787900925,
|
|
"step": 10525,
|
|
"valid_targets_mean": 2546.0,
|
|
"valid_targets_min": 663
|
|
},
|
|
{
|
|
"epoch": 5.804851157662624,
|
|
"grad_norm": 0.30906644180443515,
|
|
"learning_rate": 3.4512222224587742e-06,
|
|
"loss": 0.0711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02382844313979149,
|
|
"step": 10530,
|
|
"valid_targets_mean": 5186.2,
|
|
"valid_targets_min": 3522
|
|
},
|
|
{
|
|
"epoch": 5.80760749724366,
|
|
"grad_norm": 0.5097660802115372,
|
|
"learning_rate": 3.4358005176725317e-06,
|
|
"loss": 0.0827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049920547753572464,
|
|
"step": 10535,
|
|
"valid_targets_mean": 4087.4,
|
|
"valid_targets_min": 2327
|
|
},
|
|
{
|
|
"epoch": 5.810363836824697,
|
|
"grad_norm": 0.478780521855255,
|
|
"learning_rate": 3.4204101075057026e-06,
|
|
"loss": 0.0782,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0407244935631752,
|
|
"step": 10540,
|
|
"valid_targets_mean": 3420.1,
|
|
"valid_targets_min": 1218
|
|
},
|
|
{
|
|
"epoch": 5.813120176405733,
|
|
"grad_norm": 0.40802520879755927,
|
|
"learning_rate": 3.4050510210352793e-06,
|
|
"loss": 0.0711,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04154260456562042,
|
|
"step": 10545,
|
|
"valid_targets_mean": 4261.5,
|
|
"valid_targets_min": 3261
|
|
},
|
|
{
|
|
"epoch": 5.815876515986769,
|
|
"grad_norm": 0.43534508374229525,
|
|
"learning_rate": 3.3897232872790675e-06,
|
|
"loss": 0.0727,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034647002816200256,
|
|
"step": 10550,
|
|
"valid_targets_mean": 2142.2,
|
|
"valid_targets_min": 719
|
|
},
|
|
{
|
|
"epoch": 5.818632855567806,
|
|
"grad_norm": 0.4425096839927687,
|
|
"learning_rate": 3.374426935195643e-06,
|
|
"loss": 0.07,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.040748901665210724,
|
|
"step": 10555,
|
|
"valid_targets_mean": 3431.4,
|
|
"valid_targets_min": 2224
|
|
},
|
|
{
|
|
"epoch": 5.821389195148843,
|
|
"grad_norm": 0.3302378432957247,
|
|
"learning_rate": 3.3591619936842832e-06,
|
|
"loss": 0.0688,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.025787679478526115,
|
|
"step": 10560,
|
|
"valid_targets_mean": 3946.9,
|
|
"valid_targets_min": 3217
|
|
},
|
|
{
|
|
"epoch": 5.824145534729879,
|
|
"grad_norm": 0.484860103344773,
|
|
"learning_rate": 3.34392849158494e-06,
|
|
"loss": 0.07,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03400131314992905,
|
|
"step": 10565,
|
|
"valid_targets_mean": 2442.1,
|
|
"valid_targets_min": 735
|
|
},
|
|
{
|
|
"epoch": 5.826901874310915,
|
|
"grad_norm": 0.4811739637911985,
|
|
"learning_rate": 3.328726457678153e-06,
|
|
"loss": 0.1088,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03218796104192734,
|
|
"step": 10570,
|
|
"valid_targets_mean": 1551.5,
|
|
"valid_targets_min": 603
|
|
},
|
|
{
|
|
"epoch": 5.8296582138919515,
|
|
"grad_norm": 0.32965833126067934,
|
|
"learning_rate": 3.3135559206850075e-06,
|
|
"loss": 0.0578,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02289218083024025,
|
|
"step": 10575,
|
|
"valid_targets_mean": 1912.9,
|
|
"valid_targets_min": 582
|
|
},
|
|
{
|
|
"epoch": 5.832414553472988,
|
|
"grad_norm": 0.4348958706094496,
|
|
"learning_rate": 3.2984169092670927e-06,
|
|
"loss": 0.0714,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0390004888176918,
|
|
"step": 10580,
|
|
"valid_targets_mean": 2670.5,
|
|
"valid_targets_min": 579
|
|
},
|
|
{
|
|
"epoch": 5.835170893054024,
|
|
"grad_norm": 0.45619163549081493,
|
|
"learning_rate": 3.2833094520264265e-06,
|
|
"loss": 0.064,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04037974774837494,
|
|
"step": 10585,
|
|
"valid_targets_mean": 3709.5,
|
|
"valid_targets_min": 3248
|
|
},
|
|
{
|
|
"epoch": 5.83792723263506,
|
|
"grad_norm": 0.4531840065950375,
|
|
"learning_rate": 3.2682335775054174e-06,
|
|
"loss": 0.1042,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04504851996898651,
|
|
"step": 10590,
|
|
"valid_targets_mean": 3222.9,
|
|
"valid_targets_min": 994
|
|
},
|
|
{
|
|
"epoch": 5.840683572216097,
|
|
"grad_norm": 0.5756550224249012,
|
|
"learning_rate": 3.253189314186802e-06,
|
|
"loss": 0.0887,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06621728837490082,
|
|
"step": 10595,
|
|
"valid_targets_mean": 2046.2,
|
|
"valid_targets_min": 692
|
|
},
|
|
{
|
|
"epoch": 5.843439911797134,
|
|
"grad_norm": 0.46044398698681,
|
|
"learning_rate": 3.2381766904935997e-06,
|
|
"loss": 0.0893,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03590303659439087,
|
|
"step": 10600,
|
|
"valid_targets_mean": 2707.2,
|
|
"valid_targets_min": 632
|
|
},
|
|
{
|
|
"epoch": 5.84619625137817,
|
|
"grad_norm": 0.45839243098442195,
|
|
"learning_rate": 3.223195734789044e-06,
|
|
"loss": 0.0743,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03310862556099892,
|
|
"step": 10605,
|
|
"valid_targets_mean": 1847.8,
|
|
"valid_targets_min": 812
|
|
},
|
|
{
|
|
"epoch": 5.848952590959206,
|
|
"grad_norm": 0.5843768283020991,
|
|
"learning_rate": 3.208246475376542e-06,
|
|
"loss": 0.119,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1391768753528595,
|
|
"step": 10610,
|
|
"valid_targets_mean": 3254.0,
|
|
"valid_targets_min": 1177
|
|
},
|
|
{
|
|
"epoch": 5.8517089305402425,
|
|
"grad_norm": 0.44782401731420657,
|
|
"learning_rate": 3.193328940499629e-06,
|
|
"loss": 0.0809,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03949049115180969,
|
|
"step": 10615,
|
|
"valid_targets_mean": 3402.8,
|
|
"valid_targets_min": 2331
|
|
},
|
|
{
|
|
"epoch": 5.854465270121279,
|
|
"grad_norm": 0.5992140732962553,
|
|
"learning_rate": 3.178443158341886e-06,
|
|
"loss": 0.0992,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04658026993274689,
|
|
"step": 10620,
|
|
"valid_targets_mean": 1632.0,
|
|
"valid_targets_min": 684
|
|
},
|
|
{
|
|
"epoch": 5.857221609702315,
|
|
"grad_norm": 0.5169503304016846,
|
|
"learning_rate": 3.1635891570269115e-06,
|
|
"loss": 0.0693,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02639373391866684,
|
|
"step": 10625,
|
|
"valid_targets_mean": 1552.2,
|
|
"valid_targets_min": 446
|
|
},
|
|
{
|
|
"epoch": 5.859977949283351,
|
|
"grad_norm": 0.41866865161343947,
|
|
"learning_rate": 3.148766964618262e-06,
|
|
"loss": 0.0773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026528827846050262,
|
|
"step": 10630,
|
|
"valid_targets_mean": 2990.9,
|
|
"valid_targets_min": 655
|
|
},
|
|
{
|
|
"epoch": 5.862734288864388,
|
|
"grad_norm": 0.6246067952468632,
|
|
"learning_rate": 3.1339766091193937e-06,
|
|
"loss": 0.0824,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03567548096179962,
|
|
"step": 10635,
|
|
"valid_targets_mean": 1293.9,
|
|
"valid_targets_min": 574
|
|
},
|
|
{
|
|
"epoch": 5.865490628445425,
|
|
"grad_norm": 0.37506883043504513,
|
|
"learning_rate": 3.119218118473617e-06,
|
|
"loss": 0.0669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03946959972381592,
|
|
"step": 10640,
|
|
"valid_targets_mean": 4529.1,
|
|
"valid_targets_min": 2872
|
|
},
|
|
{
|
|
"epoch": 5.868246968026461,
|
|
"grad_norm": 0.4196304095844324,
|
|
"learning_rate": 3.104491520564039e-06,
|
|
"loss": 0.0629,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037202708423137665,
|
|
"step": 10645,
|
|
"valid_targets_mean": 3615.2,
|
|
"valid_targets_min": 2242
|
|
},
|
|
{
|
|
"epoch": 5.871003307607497,
|
|
"grad_norm": 0.4729628638902189,
|
|
"learning_rate": 3.0897968432135105e-06,
|
|
"loss": 0.0722,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03228700906038284,
|
|
"step": 10650,
|
|
"valid_targets_mean": 2017.4,
|
|
"valid_targets_min": 651
|
|
},
|
|
{
|
|
"epoch": 5.873759647188534,
|
|
"grad_norm": 0.4498390886102797,
|
|
"learning_rate": 3.0751341141845794e-06,
|
|
"loss": 0.0834,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03738843649625778,
|
|
"step": 10655,
|
|
"valid_targets_mean": 3131.8,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 5.87651598676957,
|
|
"grad_norm": 0.5100098085486376,
|
|
"learning_rate": 3.060503361179423e-06,
|
|
"loss": 0.0851,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04666534438729286,
|
|
"step": 10660,
|
|
"valid_targets_mean": 3651.6,
|
|
"valid_targets_min": 3006
|
|
},
|
|
{
|
|
"epoch": 5.879272326350606,
|
|
"grad_norm": 0.32103836741939207,
|
|
"learning_rate": 3.045904611839827e-06,
|
|
"loss": 0.0757,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026558440178632736,
|
|
"step": 10665,
|
|
"valid_targets_mean": 3210.8,
|
|
"valid_targets_min": 2353
|
|
},
|
|
{
|
|
"epoch": 5.882028665931642,
|
|
"grad_norm": 0.358435990222605,
|
|
"learning_rate": 3.031337893747095e-06,
|
|
"loss": 0.0681,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03156706690788269,
|
|
"step": 10670,
|
|
"valid_targets_mean": 1954.4,
|
|
"valid_targets_min": 665
|
|
},
|
|
{
|
|
"epoch": 5.884785005512679,
|
|
"grad_norm": 0.545621117516562,
|
|
"learning_rate": 3.0168032344220165e-06,
|
|
"loss": 0.0841,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04634404182434082,
|
|
"step": 10675,
|
|
"valid_targets_mean": 3512.6,
|
|
"valid_targets_min": 1776
|
|
},
|
|
{
|
|
"epoch": 5.887541345093716,
|
|
"grad_norm": 0.31360884249069826,
|
|
"learning_rate": 3.0023006613248217e-06,
|
|
"loss": 0.0606,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026722118258476257,
|
|
"step": 10680,
|
|
"valid_targets_mean": 4054.6,
|
|
"valid_targets_min": 3675
|
|
},
|
|
{
|
|
"epoch": 5.890297684674752,
|
|
"grad_norm": 0.3978495124716693,
|
|
"learning_rate": 2.9878302018551087e-06,
|
|
"loss": 0.0762,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031191488727927208,
|
|
"step": 10685,
|
|
"valid_targets_mean": 3238.4,
|
|
"valid_targets_min": 591
|
|
},
|
|
{
|
|
"epoch": 5.893054024255788,
|
|
"grad_norm": 0.45701549701383903,
|
|
"learning_rate": 2.9733918833518127e-06,
|
|
"loss": 0.074,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0314142219722271,
|
|
"step": 10690,
|
|
"valid_targets_mean": 2646.6,
|
|
"valid_targets_min": 1011
|
|
},
|
|
{
|
|
"epoch": 5.895810363836825,
|
|
"grad_norm": 0.6547791759688245,
|
|
"learning_rate": 2.9589857330931404e-06,
|
|
"loss": 0.0839,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07440979033708572,
|
|
"step": 10695,
|
|
"valid_targets_mean": 1379.1,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 5.8985667034178615,
|
|
"grad_norm": 0.5711936224455019,
|
|
"learning_rate": 2.9446117782965266e-06,
|
|
"loss": 0.0784,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04029003530740738,
|
|
"step": 10700,
|
|
"valid_targets_mean": 2145.4,
|
|
"valid_targets_min": 693
|
|
},
|
|
{
|
|
"epoch": 5.901323042998897,
|
|
"grad_norm": 0.7047654983684193,
|
|
"learning_rate": 2.9302700461185707e-06,
|
|
"loss": 0.0743,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.050310246646404266,
|
|
"step": 10705,
|
|
"valid_targets_mean": 1268.9,
|
|
"valid_targets_min": 732
|
|
},
|
|
{
|
|
"epoch": 5.904079382579933,
|
|
"grad_norm": 0.38806191006249946,
|
|
"learning_rate": 2.915960563655014e-06,
|
|
"loss": 0.0735,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02515799179673195,
|
|
"step": 10710,
|
|
"valid_targets_mean": 2524.1,
|
|
"valid_targets_min": 799
|
|
},
|
|
{
|
|
"epoch": 5.90683572216097,
|
|
"grad_norm": 0.3996660795815713,
|
|
"learning_rate": 2.9016833579406455e-06,
|
|
"loss": 0.0926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030745767056941986,
|
|
"step": 10715,
|
|
"valid_targets_mean": 2966.2,
|
|
"valid_targets_min": 1012
|
|
},
|
|
{
|
|
"epoch": 5.909592061742007,
|
|
"grad_norm": 0.4331389768915479,
|
|
"learning_rate": 2.8874384559492897e-06,
|
|
"loss": 0.0757,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03625374287366867,
|
|
"step": 10720,
|
|
"valid_targets_mean": 3617.0,
|
|
"valid_targets_min": 1622
|
|
},
|
|
{
|
|
"epoch": 5.912348401323043,
|
|
"grad_norm": 0.4708020380952445,
|
|
"learning_rate": 2.8732258845937313e-06,
|
|
"loss": 0.0644,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03645868971943855,
|
|
"step": 10725,
|
|
"valid_targets_mean": 3110.0,
|
|
"valid_targets_min": 1001
|
|
},
|
|
{
|
|
"epoch": 5.915104740904079,
|
|
"grad_norm": 0.6058901397183789,
|
|
"learning_rate": 2.8590456707256776e-06,
|
|
"loss": 0.0856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035730600357055664,
|
|
"step": 10730,
|
|
"valid_targets_mean": 2685.8,
|
|
"valid_targets_min": 897
|
|
},
|
|
{
|
|
"epoch": 5.917861080485116,
|
|
"grad_norm": 0.44187183157903265,
|
|
"learning_rate": 2.8448978411356987e-06,
|
|
"loss": 0.0795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029674295336008072,
|
|
"step": 10735,
|
|
"valid_targets_mean": 3777.6,
|
|
"valid_targets_min": 3040
|
|
},
|
|
{
|
|
"epoch": 5.9206174200661525,
|
|
"grad_norm": 0.5769625536755545,
|
|
"learning_rate": 2.830782422553189e-06,
|
|
"loss": 0.0636,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049918919801712036,
|
|
"step": 10740,
|
|
"valid_targets_mean": 2964.5,
|
|
"valid_targets_min": 561
|
|
},
|
|
{
|
|
"epoch": 5.9233737596471885,
|
|
"grad_norm": 0.4225421291765312,
|
|
"learning_rate": 2.8166994416463023e-06,
|
|
"loss": 0.0736,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04158448427915573,
|
|
"step": 10745,
|
|
"valid_targets_mean": 3642.1,
|
|
"valid_targets_min": 2542
|
|
},
|
|
{
|
|
"epoch": 5.926130099228224,
|
|
"grad_norm": 0.34182282610434084,
|
|
"learning_rate": 2.802648925021907e-06,
|
|
"loss": 0.0644,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029587846249341965,
|
|
"step": 10750,
|
|
"valid_targets_mean": 4175.0,
|
|
"valid_targets_min": 2983
|
|
},
|
|
{
|
|
"epoch": 5.928886438809261,
|
|
"grad_norm": 0.4030298983239053,
|
|
"learning_rate": 2.788630899225542e-06,
|
|
"loss": 0.0593,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026005571708083153,
|
|
"step": 10755,
|
|
"valid_targets_mean": 2350.6,
|
|
"valid_targets_min": 615
|
|
},
|
|
{
|
|
"epoch": 5.931642778390298,
|
|
"grad_norm": 0.3316188756851577,
|
|
"learning_rate": 2.7746453907413616e-06,
|
|
"loss": 0.0814,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028358865529298782,
|
|
"step": 10760,
|
|
"valid_targets_mean": 3647.6,
|
|
"valid_targets_min": 2936
|
|
},
|
|
{
|
|
"epoch": 5.934399117971334,
|
|
"grad_norm": 0.35175168975748267,
|
|
"learning_rate": 2.760692425992084e-06,
|
|
"loss": 0.0723,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05882719159126282,
|
|
"step": 10765,
|
|
"valid_targets_mean": 3372.6,
|
|
"valid_targets_min": 1532
|
|
},
|
|
{
|
|
"epoch": 5.93715545755237,
|
|
"grad_norm": 0.35484728847109304,
|
|
"learning_rate": 2.7467720313389402e-06,
|
|
"loss": 0.0872,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.021428870037198067,
|
|
"step": 10770,
|
|
"valid_targets_mean": 2305.8,
|
|
"valid_targets_min": 506
|
|
},
|
|
{
|
|
"epoch": 5.939911797133407,
|
|
"grad_norm": 0.3542857495543656,
|
|
"learning_rate": 2.73288423308163e-06,
|
|
"loss": 0.0825,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02307962253689766,
|
|
"step": 10775,
|
|
"valid_targets_mean": 2746.1,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 5.942668136714444,
|
|
"grad_norm": 0.4493354379109839,
|
|
"learning_rate": 2.7190290574582647e-06,
|
|
"loss": 0.0934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039251022040843964,
|
|
"step": 10780,
|
|
"valid_targets_mean": 2745.8,
|
|
"valid_targets_min": 733
|
|
},
|
|
{
|
|
"epoch": 5.9454244762954795,
|
|
"grad_norm": 0.3804663797837991,
|
|
"learning_rate": 2.705206530645328e-06,
|
|
"loss": 0.0627,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03265189751982689,
|
|
"step": 10785,
|
|
"valid_targets_mean": 3070.9,
|
|
"valid_targets_min": 489
|
|
},
|
|
{
|
|
"epoch": 5.948180815876516,
|
|
"grad_norm": 0.5449809006075041,
|
|
"learning_rate": 2.6914166787576167e-06,
|
|
"loss": 0.0791,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03299988433718681,
|
|
"step": 10790,
|
|
"valid_targets_mean": 1427.6,
|
|
"valid_targets_min": 693
|
|
},
|
|
{
|
|
"epoch": 5.950937155457552,
|
|
"grad_norm": 0.391684583688166,
|
|
"learning_rate": 2.6776595278481955e-06,
|
|
"loss": 0.0659,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.027263425290584564,
|
|
"step": 10795,
|
|
"valid_targets_mean": 3525.9,
|
|
"valid_targets_min": 918
|
|
},
|
|
{
|
|
"epoch": 5.953693495038589,
|
|
"grad_norm": 0.6698052431851748,
|
|
"learning_rate": 2.6639351039083437e-06,
|
|
"loss": 0.101,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04039355367422104,
|
|
"step": 10800,
|
|
"valid_targets_mean": 1479.6,
|
|
"valid_targets_min": 528
|
|
},
|
|
{
|
|
"epoch": 5.956449834619625,
|
|
"grad_norm": 0.5055479242519756,
|
|
"learning_rate": 2.650243432867521e-06,
|
|
"loss": 0.1006,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05940599739551544,
|
|
"step": 10805,
|
|
"valid_targets_mean": 2086.1,
|
|
"valid_targets_min": 572
|
|
},
|
|
{
|
|
"epoch": 5.959206174200662,
|
|
"grad_norm": 0.45794485923378564,
|
|
"learning_rate": 2.636584540593299e-06,
|
|
"loss": 0.072,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03799443319439888,
|
|
"step": 10810,
|
|
"valid_targets_mean": 3714.4,
|
|
"valid_targets_min": 2773
|
|
},
|
|
{
|
|
"epoch": 5.961962513781698,
|
|
"grad_norm": 0.49712410654139844,
|
|
"learning_rate": 2.62295845289132e-06,
|
|
"loss": 0.0702,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033639438450336456,
|
|
"step": 10815,
|
|
"valid_targets_mean": 3416.0,
|
|
"valid_targets_min": 2516
|
|
},
|
|
{
|
|
"epoch": 5.964718853362735,
|
|
"grad_norm": 0.4164178167485395,
|
|
"learning_rate": 2.6093651955052492e-06,
|
|
"loss": 0.0694,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02712143585085869,
|
|
"step": 10820,
|
|
"valid_targets_mean": 2494.9,
|
|
"valid_targets_min": 965
|
|
},
|
|
{
|
|
"epoch": 5.967475192943771,
|
|
"grad_norm": 0.44136817568036213,
|
|
"learning_rate": 2.5958047941167296e-06,
|
|
"loss": 0.0652,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028853660449385643,
|
|
"step": 10825,
|
|
"valid_targets_mean": 1522.5,
|
|
"valid_targets_min": 725
|
|
},
|
|
{
|
|
"epoch": 5.970231532524807,
|
|
"grad_norm": 0.9447354054608453,
|
|
"learning_rate": 2.582277274345324e-06,
|
|
"loss": 0.0832,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05597830191254616,
|
|
"step": 10830,
|
|
"valid_targets_mean": 1261.0,
|
|
"valid_targets_min": 633
|
|
},
|
|
{
|
|
"epoch": 5.972987872105843,
|
|
"grad_norm": 0.3719759318893918,
|
|
"learning_rate": 2.568782661748479e-06,
|
|
"loss": 0.0624,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029345398768782616,
|
|
"step": 10835,
|
|
"valid_targets_mean": 3052.4,
|
|
"valid_targets_min": 889
|
|
},
|
|
{
|
|
"epoch": 5.97574421168688,
|
|
"grad_norm": 0.4181025419491743,
|
|
"learning_rate": 2.5553209818214632e-06,
|
|
"loss": 0.0667,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0302704069763422,
|
|
"step": 10840,
|
|
"valid_targets_mean": 3543.8,
|
|
"valid_targets_min": 2396
|
|
},
|
|
{
|
|
"epoch": 5.978500551267916,
|
|
"grad_norm": 0.45783763747084116,
|
|
"learning_rate": 2.541892259997332e-06,
|
|
"loss": 0.0669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0308137908577919,
|
|
"step": 10845,
|
|
"valid_targets_mean": 3317.0,
|
|
"valid_targets_min": 981
|
|
},
|
|
{
|
|
"epoch": 5.981256890848953,
|
|
"grad_norm": 0.38192780445207664,
|
|
"learning_rate": 2.5284965216468617e-06,
|
|
"loss": 0.0775,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034714266657829285,
|
|
"step": 10850,
|
|
"valid_targets_mean": 3465.1,
|
|
"valid_targets_min": 1071
|
|
},
|
|
{
|
|
"epoch": 5.984013230429989,
|
|
"grad_norm": 0.3766852407518425,
|
|
"learning_rate": 2.5151337920785323e-06,
|
|
"loss": 0.0638,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03589920327067375,
|
|
"step": 10855,
|
|
"valid_targets_mean": 4144.0,
|
|
"valid_targets_min": 757
|
|
},
|
|
{
|
|
"epoch": 5.986769570011026,
|
|
"grad_norm": 0.38851053769215615,
|
|
"learning_rate": 2.501804096538447e-06,
|
|
"loss": 0.0672,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029379256069660187,
|
|
"step": 10860,
|
|
"valid_targets_mean": 1548.0,
|
|
"valid_targets_min": 504
|
|
},
|
|
{
|
|
"epoch": 5.989525909592062,
|
|
"grad_norm": 0.481536283756934,
|
|
"learning_rate": 2.4885074602102987e-06,
|
|
"loss": 0.0851,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0747687891125679,
|
|
"step": 10865,
|
|
"valid_targets_mean": 2696.8,
|
|
"valid_targets_min": 775
|
|
},
|
|
{
|
|
"epoch": 5.9922822491730985,
|
|
"grad_norm": 0.402022307999217,
|
|
"learning_rate": 2.4752439082153235e-06,
|
|
"loss": 0.141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035179510712623596,
|
|
"step": 10870,
|
|
"valid_targets_mean": 3880.9,
|
|
"valid_targets_min": 2338
|
|
},
|
|
{
|
|
"epoch": 5.995038588754134,
|
|
"grad_norm": 0.46169106398102855,
|
|
"learning_rate": 2.462013465612254e-06,
|
|
"loss": 0.0682,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037871092557907104,
|
|
"step": 10875,
|
|
"valid_targets_mean": 3655.5,
|
|
"valid_targets_min": 2463
|
|
},
|
|
{
|
|
"epoch": 5.997794928335171,
|
|
"grad_norm": 0.3901519072494221,
|
|
"learning_rate": 2.4488161573972667e-06,
|
|
"loss": 0.0669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03265278786420822,
|
|
"step": 10880,
|
|
"valid_targets_mean": 3319.2,
|
|
"valid_targets_min": 2237
|
|
},
|
|
{
|
|
"epoch": 6.000551267916207,
|
|
"grad_norm": 0.8139552711112659,
|
|
"learning_rate": 2.435652008503939e-06,
|
|
"loss": 0.1217,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0961044505238533,
|
|
"step": 10885,
|
|
"valid_targets_mean": 8186.1,
|
|
"valid_targets_min": 6131
|
|
},
|
|
{
|
|
"epoch": 6.003307607497244,
|
|
"grad_norm": 0.8216833835847769,
|
|
"learning_rate": 2.422521043803199e-06,
|
|
"loss": 0.1902,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09788983315229416,
|
|
"step": 10890,
|
|
"valid_targets_mean": 8564.9,
|
|
"valid_targets_min": 5974
|
|
},
|
|
{
|
|
"epoch": 6.00606394707828,
|
|
"grad_norm": 0.753340566383063,
|
|
"learning_rate": 2.4094232881032787e-06,
|
|
"loss": 0.1727,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07500340044498444,
|
|
"step": 10895,
|
|
"valid_targets_mean": 3567.1,
|
|
"valid_targets_min": 287
|
|
},
|
|
{
|
|
"epoch": 6.008820286659317,
|
|
"grad_norm": 0.5740704507294192,
|
|
"learning_rate": 2.3963587661496713e-06,
|
|
"loss": 0.1738,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08845479041337967,
|
|
"step": 10900,
|
|
"valid_targets_mean": 6991.4,
|
|
"valid_targets_min": 4953
|
|
},
|
|
{
|
|
"epoch": 6.011576626240353,
|
|
"grad_norm": 0.6027464018123886,
|
|
"learning_rate": 2.383327502625088e-06,
|
|
"loss": 0.1642,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08496417105197906,
|
|
"step": 10905,
|
|
"valid_targets_mean": 6929.9,
|
|
"valid_targets_min": 1449
|
|
},
|
|
{
|
|
"epoch": 6.0143329658213895,
|
|
"grad_norm": 0.4356382360487854,
|
|
"learning_rate": 2.3703295221493926e-06,
|
|
"loss": 0.1679,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0794033408164978,
|
|
"step": 10910,
|
|
"valid_targets_mean": 7335.5,
|
|
"valid_targets_min": 4961
|
|
},
|
|
{
|
|
"epoch": 6.0170893054024255,
|
|
"grad_norm": 0.5027292448814547,
|
|
"learning_rate": 2.3573648492795754e-06,
|
|
"loss": 0.1816,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10678911209106445,
|
|
"step": 10915,
|
|
"valid_targets_mean": 7028.9,
|
|
"valid_targets_min": 5952
|
|
},
|
|
{
|
|
"epoch": 6.019845644983462,
|
|
"grad_norm": 0.43543526231902435,
|
|
"learning_rate": 2.3444335085096957e-06,
|
|
"loss": 0.1764,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09667710214853287,
|
|
"step": 10920,
|
|
"valid_targets_mean": 7634.2,
|
|
"valid_targets_min": 5143
|
|
},
|
|
{
|
|
"epoch": 6.022601984564498,
|
|
"grad_norm": 0.43508565809444527,
|
|
"learning_rate": 2.33153552427084e-06,
|
|
"loss": 0.1636,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08359694480895996,
|
|
"step": 10925,
|
|
"valid_targets_mean": 7079.8,
|
|
"valid_targets_min": 5316
|
|
},
|
|
{
|
|
"epoch": 6.025358324145535,
|
|
"grad_norm": 0.5006901975015738,
|
|
"learning_rate": 2.318670920931074e-06,
|
|
"loss": 0.1512,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06941013038158417,
|
|
"step": 10930,
|
|
"valid_targets_mean": 4218.1,
|
|
"valid_targets_min": 151
|
|
},
|
|
{
|
|
"epoch": 6.028114663726571,
|
|
"grad_norm": 0.4219790161021891,
|
|
"learning_rate": 2.3058397227954e-06,
|
|
"loss": 0.1589,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08445869386196136,
|
|
"step": 10935,
|
|
"valid_targets_mean": 7171.6,
|
|
"valid_targets_min": 4244
|
|
},
|
|
{
|
|
"epoch": 6.030871003307608,
|
|
"grad_norm": 0.3990957315732198,
|
|
"learning_rate": 2.2930419541057015e-06,
|
|
"loss": 0.1648,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07940733432769775,
|
|
"step": 10940,
|
|
"valid_targets_mean": 7214.5,
|
|
"valid_targets_min": 5799
|
|
},
|
|
{
|
|
"epoch": 6.033627342888644,
|
|
"grad_norm": 0.40926412313904686,
|
|
"learning_rate": 2.2802776390407087e-06,
|
|
"loss": 0.1615,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08152562379837036,
|
|
"step": 10945,
|
|
"valid_targets_mean": 6923.1,
|
|
"valid_targets_min": 3844
|
|
},
|
|
{
|
|
"epoch": 6.036383682469681,
|
|
"grad_norm": 0.3911568395084352,
|
|
"learning_rate": 2.2675468017159518e-06,
|
|
"loss": 0.1603,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08510257303714752,
|
|
"step": 10950,
|
|
"valid_targets_mean": 6961.0,
|
|
"valid_targets_min": 5664
|
|
},
|
|
{
|
|
"epoch": 6.0391400220507165,
|
|
"grad_norm": 0.4175588457734751,
|
|
"learning_rate": 2.2548494661837063e-06,
|
|
"loss": 0.1541,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08557314425706863,
|
|
"step": 10955,
|
|
"valid_targets_mean": 6933.6,
|
|
"valid_targets_min": 4915
|
|
},
|
|
{
|
|
"epoch": 6.041896361631753,
|
|
"grad_norm": 0.4072830273617978,
|
|
"learning_rate": 2.2421856564329515e-06,
|
|
"loss": 0.1588,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07488608360290527,
|
|
"step": 10960,
|
|
"valid_targets_mean": 6407.2,
|
|
"valid_targets_min": 4966
|
|
},
|
|
{
|
|
"epoch": 6.044652701212789,
|
|
"grad_norm": 0.3802207750916098,
|
|
"learning_rate": 2.2295553963893313e-06,
|
|
"loss": 0.1613,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0788707435131073,
|
|
"step": 10965,
|
|
"valid_targets_mean": 7489.0,
|
|
"valid_targets_min": 5451
|
|
},
|
|
{
|
|
"epoch": 6.047409040793826,
|
|
"grad_norm": 0.37499240248617777,
|
|
"learning_rate": 2.2169587099151002e-06,
|
|
"loss": 0.1569,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07427690923213959,
|
|
"step": 10970,
|
|
"valid_targets_mean": 7392.1,
|
|
"valid_targets_min": 5195
|
|
},
|
|
{
|
|
"epoch": 6.050165380374862,
|
|
"grad_norm": 0.4247131373609521,
|
|
"learning_rate": 2.204395620809088e-06,
|
|
"loss": 0.1664,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10309232771396637,
|
|
"step": 10975,
|
|
"valid_targets_mean": 7050.0,
|
|
"valid_targets_min": 3733
|
|
},
|
|
{
|
|
"epoch": 6.052921719955899,
|
|
"grad_norm": 0.4859770661005047,
|
|
"learning_rate": 2.191866152806643e-06,
|
|
"loss": 0.1831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09220490604639053,
|
|
"step": 10980,
|
|
"valid_targets_mean": 6038.4,
|
|
"valid_targets_min": 5143
|
|
},
|
|
{
|
|
"epoch": 6.055678059536935,
|
|
"grad_norm": 0.41202433473228667,
|
|
"learning_rate": 2.1793703295795998e-06,
|
|
"loss": 0.1286,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07725677639245987,
|
|
"step": 10985,
|
|
"valid_targets_mean": 7356.9,
|
|
"valid_targets_min": 5765
|
|
},
|
|
{
|
|
"epoch": 6.058434399117972,
|
|
"grad_norm": 0.3852750465929006,
|
|
"learning_rate": 2.1669081747362196e-06,
|
|
"loss": 0.1547,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07314597070217133,
|
|
"step": 10990,
|
|
"valid_targets_mean": 7668.2,
|
|
"valid_targets_min": 4640
|
|
},
|
|
{
|
|
"epoch": 6.061190738699008,
|
|
"grad_norm": 0.41501665428931866,
|
|
"learning_rate": 2.1544797118211604e-06,
|
|
"loss": 0.1569,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08686096966266632,
|
|
"step": 10995,
|
|
"valid_targets_mean": 7774.5,
|
|
"valid_targets_min": 5483
|
|
},
|
|
{
|
|
"epoch": 6.063947078280044,
|
|
"grad_norm": 0.4222019819678538,
|
|
"learning_rate": 2.1420849643154294e-06,
|
|
"loss": 0.1645,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07948994636535645,
|
|
"step": 11000,
|
|
"valid_targets_mean": 7311.8,
|
|
"valid_targets_min": 6424
|
|
},
|
|
{
|
|
"epoch": 6.06670341786108,
|
|
"grad_norm": 0.4066041233934624,
|
|
"learning_rate": 2.129723955636329e-06,
|
|
"loss": 0.1595,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07008326053619385,
|
|
"step": 11005,
|
|
"valid_targets_mean": 6788.4,
|
|
"valid_targets_min": 4371
|
|
},
|
|
{
|
|
"epoch": 6.069459757442117,
|
|
"grad_norm": 0.3830746989318652,
|
|
"learning_rate": 2.1173967091374227e-06,
|
|
"loss": 0.1579,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06926591694355011,
|
|
"step": 11010,
|
|
"valid_targets_mean": 7830.8,
|
|
"valid_targets_min": 5371
|
|
},
|
|
{
|
|
"epoch": 6.072216097023153,
|
|
"grad_norm": 0.36184386986403333,
|
|
"learning_rate": 2.105103248108482e-06,
|
|
"loss": 0.1392,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0650649145245552,
|
|
"step": 11015,
|
|
"valid_targets_mean": 8286.8,
|
|
"valid_targets_min": 4818
|
|
},
|
|
{
|
|
"epoch": 6.07497243660419,
|
|
"grad_norm": 0.35168990165997804,
|
|
"learning_rate": 2.0928435957754557e-06,
|
|
"loss": 0.1414,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05969775468111038,
|
|
"step": 11020,
|
|
"valid_targets_mean": 7339.1,
|
|
"valid_targets_min": 5995
|
|
},
|
|
{
|
|
"epoch": 6.077728776185226,
|
|
"grad_norm": 0.42449203156193316,
|
|
"learning_rate": 2.080617775300413e-06,
|
|
"loss": 0.1455,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0784744918346405,
|
|
"step": 11025,
|
|
"valid_targets_mean": 6405.4,
|
|
"valid_targets_min": 4206
|
|
},
|
|
{
|
|
"epoch": 6.080485115766263,
|
|
"grad_norm": 1.1642498569394675,
|
|
"learning_rate": 2.0684258097815047e-06,
|
|
"loss": 0.1416,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03938277065753937,
|
|
"step": 11030,
|
|
"valid_targets_mean": 246.4,
|
|
"valid_targets_min": 136
|
|
},
|
|
{
|
|
"epoch": 6.083241455347299,
|
|
"grad_norm": 0.4097688700253261,
|
|
"learning_rate": 2.0562677222529204e-06,
|
|
"loss": 0.1634,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07718682289123535,
|
|
"step": 11035,
|
|
"valid_targets_mean": 6778.4,
|
|
"valid_targets_min": 5133
|
|
},
|
|
{
|
|
"epoch": 6.0859977949283355,
|
|
"grad_norm": 0.48269327719049643,
|
|
"learning_rate": 2.0441435356848436e-06,
|
|
"loss": 0.1701,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08896273374557495,
|
|
"step": 11040,
|
|
"valid_targets_mean": 6963.0,
|
|
"valid_targets_min": 5072
|
|
},
|
|
{
|
|
"epoch": 6.088754134509371,
|
|
"grad_norm": 0.42928980143660495,
|
|
"learning_rate": 2.0320532729834187e-06,
|
|
"loss": 0.1669,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08975011855363846,
|
|
"step": 11045,
|
|
"valid_targets_mean": 6662.5,
|
|
"valid_targets_min": 5036
|
|
},
|
|
{
|
|
"epoch": 6.091510474090408,
|
|
"grad_norm": 0.4438270722010837,
|
|
"learning_rate": 2.019996956990682e-06,
|
|
"loss": 0.1635,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09847986698150635,
|
|
"step": 11050,
|
|
"valid_targets_mean": 7389.0,
|
|
"valid_targets_min": 5546
|
|
},
|
|
{
|
|
"epoch": 6.094266813671444,
|
|
"grad_norm": 0.40160285257645645,
|
|
"learning_rate": 2.0079746104845445e-06,
|
|
"loss": 0.1592,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08063308894634247,
|
|
"step": 11055,
|
|
"valid_targets_mean": 7507.0,
|
|
"valid_targets_min": 5051
|
|
},
|
|
{
|
|
"epoch": 6.097023153252481,
|
|
"grad_norm": 0.5129852689298762,
|
|
"learning_rate": 1.9959862561787347e-06,
|
|
"loss": 0.1646,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07456132024526596,
|
|
"step": 11060,
|
|
"valid_targets_mean": 7192.6,
|
|
"valid_targets_min": 5809
|
|
},
|
|
{
|
|
"epoch": 6.099779492833517,
|
|
"grad_norm": 0.39847737310881465,
|
|
"learning_rate": 1.984031916722762e-06,
|
|
"loss": 0.1595,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.079062819480896,
|
|
"step": 11065,
|
|
"valid_targets_mean": 7255.8,
|
|
"valid_targets_min": 5119
|
|
},
|
|
{
|
|
"epoch": 6.102535832414554,
|
|
"grad_norm": 0.42213307776809994,
|
|
"learning_rate": 1.972111614701873e-06,
|
|
"loss": 0.1682,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.077930748462677,
|
|
"step": 11070,
|
|
"valid_targets_mean": 6710.9,
|
|
"valid_targets_min": 4029
|
|
},
|
|
{
|
|
"epoch": 6.10529217199559,
|
|
"grad_norm": 0.47903337190828876,
|
|
"learning_rate": 1.9602253726370056e-06,
|
|
"loss": 0.1629,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08942839503288269,
|
|
"step": 11075,
|
|
"valid_targets_mean": 6299.6,
|
|
"valid_targets_min": 5464
|
|
},
|
|
{
|
|
"epoch": 6.1080485115766265,
|
|
"grad_norm": 0.421532761083628,
|
|
"learning_rate": 1.9483732129847466e-06,
|
|
"loss": 0.1641,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07850450277328491,
|
|
"step": 11080,
|
|
"valid_targets_mean": 6239.8,
|
|
"valid_targets_min": 4663
|
|
},
|
|
{
|
|
"epoch": 6.1108048511576625,
|
|
"grad_norm": 0.4439673204885292,
|
|
"learning_rate": 1.9365551581372964e-06,
|
|
"loss": 0.1607,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08328081667423248,
|
|
"step": 11085,
|
|
"valid_targets_mean": 5759.1,
|
|
"valid_targets_min": 4989
|
|
},
|
|
{
|
|
"epoch": 6.113561190738699,
|
|
"grad_norm": 0.4743079148411868,
|
|
"learning_rate": 1.924771230422413e-06,
|
|
"loss": 0.1527,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06001581251621246,
|
|
"step": 11090,
|
|
"valid_targets_mean": 3368.0,
|
|
"valid_targets_min": 2264
|
|
},
|
|
{
|
|
"epoch": 6.116317530319735,
|
|
"grad_norm": 0.4277594069452669,
|
|
"learning_rate": 1.913021452103392e-06,
|
|
"loss": 0.1323,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07851102203130722,
|
|
"step": 11095,
|
|
"valid_targets_mean": 6208.5,
|
|
"valid_targets_min": 5061
|
|
},
|
|
{
|
|
"epoch": 6.119073869900772,
|
|
"grad_norm": 0.41966840543183376,
|
|
"learning_rate": 1.901305845378998e-06,
|
|
"loss": 0.1584,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08095232397317886,
|
|
"step": 11100,
|
|
"valid_targets_mean": 6299.4,
|
|
"valid_targets_min": 4566
|
|
},
|
|
{
|
|
"epoch": 6.121830209481808,
|
|
"grad_norm": 0.42717286152477624,
|
|
"learning_rate": 1.8896244323834412e-06,
|
|
"loss": 0.1579,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08530595153570175,
|
|
"step": 11105,
|
|
"valid_targets_mean": 7560.9,
|
|
"valid_targets_min": 4931
|
|
},
|
|
{
|
|
"epoch": 6.124586549062845,
|
|
"grad_norm": 0.408159752482954,
|
|
"learning_rate": 1.877977235186328e-06,
|
|
"loss": 0.1579,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06795749813318253,
|
|
"step": 11110,
|
|
"valid_targets_mean": 5546.4,
|
|
"valid_targets_min": 4426
|
|
},
|
|
{
|
|
"epoch": 6.127342888643881,
|
|
"grad_norm": 0.43968386974912516,
|
|
"learning_rate": 1.8663642757926227e-06,
|
|
"loss": 0.1583,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07765524089336395,
|
|
"step": 11115,
|
|
"valid_targets_mean": 5400.6,
|
|
"valid_targets_min": 4401
|
|
},
|
|
{
|
|
"epoch": 6.130099228224918,
|
|
"grad_norm": 0.42043336766184386,
|
|
"learning_rate": 1.8547855761426014e-06,
|
|
"loss": 0.1611,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08248654752969742,
|
|
"step": 11120,
|
|
"valid_targets_mean": 6330.9,
|
|
"valid_targets_min": 5178
|
|
},
|
|
{
|
|
"epoch": 6.1328555678059535,
|
|
"grad_norm": 0.4279032315513361,
|
|
"learning_rate": 1.843241158111817e-06,
|
|
"loss": 0.1617,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07567909359931946,
|
|
"step": 11125,
|
|
"valid_targets_mean": 5793.9,
|
|
"valid_targets_min": 5097
|
|
},
|
|
{
|
|
"epoch": 6.13561190738699,
|
|
"grad_norm": 0.7114816228670157,
|
|
"learning_rate": 1.8317310435110535e-06,
|
|
"loss": 0.1218,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07074061036109924,
|
|
"step": 11130,
|
|
"valid_targets_mean": 1627.8,
|
|
"valid_targets_min": 136
|
|
},
|
|
{
|
|
"epoch": 6.138368246968026,
|
|
"grad_norm": 0.4171344468528038,
|
|
"learning_rate": 1.820255254086285e-06,
|
|
"loss": 0.1553,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08032190054655075,
|
|
"step": 11135,
|
|
"valid_targets_mean": 6442.6,
|
|
"valid_targets_min": 4719
|
|
},
|
|
{
|
|
"epoch": 6.141124586549063,
|
|
"grad_norm": 0.3998645585287632,
|
|
"learning_rate": 1.8088138115186304e-06,
|
|
"loss": 0.155,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0855802595615387,
|
|
"step": 11140,
|
|
"valid_targets_mean": 7150.1,
|
|
"valid_targets_min": 5326
|
|
},
|
|
{
|
|
"epoch": 6.143880926130099,
|
|
"grad_norm": 0.40710182345280616,
|
|
"learning_rate": 1.7974067374243298e-06,
|
|
"loss": 0.1624,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08003983646631241,
|
|
"step": 11145,
|
|
"valid_targets_mean": 7101.2,
|
|
"valid_targets_min": 4056
|
|
},
|
|
{
|
|
"epoch": 6.146637265711136,
|
|
"grad_norm": 0.4079946513826592,
|
|
"learning_rate": 1.786034053354684e-06,
|
|
"loss": 0.1538,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07565173506736755,
|
|
"step": 11150,
|
|
"valid_targets_mean": 6992.9,
|
|
"valid_targets_min": 4903
|
|
},
|
|
{
|
|
"epoch": 6.149393605292172,
|
|
"grad_norm": 0.4432797950815951,
|
|
"learning_rate": 1.7746957807960141e-06,
|
|
"loss": 0.1503,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0809486135840416,
|
|
"step": 11155,
|
|
"valid_targets_mean": 5869.9,
|
|
"valid_targets_min": 4747
|
|
},
|
|
{
|
|
"epoch": 6.152149944873209,
|
|
"grad_norm": 0.44677870469753533,
|
|
"learning_rate": 1.763391941169641e-06,
|
|
"loss": 0.1574,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08046580851078033,
|
|
"step": 11160,
|
|
"valid_targets_mean": 6258.1,
|
|
"valid_targets_min": 4588
|
|
},
|
|
{
|
|
"epoch": 6.154906284454245,
|
|
"grad_norm": 0.8477875436560176,
|
|
"learning_rate": 1.7521225558318212e-06,
|
|
"loss": 0.1554,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059198521077632904,
|
|
"step": 11165,
|
|
"valid_targets_mean": 1159.1,
|
|
"valid_targets_min": 787
|
|
},
|
|
{
|
|
"epoch": 6.157662624035281,
|
|
"grad_norm": 0.7816434235496538,
|
|
"learning_rate": 1.7408876460737234e-06,
|
|
"loss": 0.1417,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06842900812625885,
|
|
"step": 11170,
|
|
"valid_targets_mean": 1495.9,
|
|
"valid_targets_min": 956
|
|
},
|
|
{
|
|
"epoch": 6.160418963616317,
|
|
"grad_norm": 0.7138840414387904,
|
|
"learning_rate": 1.7296872331213776e-06,
|
|
"loss": 0.1309,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0593431293964386,
|
|
"step": 11175,
|
|
"valid_targets_mean": 1366.0,
|
|
"valid_targets_min": 795
|
|
},
|
|
{
|
|
"epoch": 6.163175303197354,
|
|
"grad_norm": 0.817140250047042,
|
|
"learning_rate": 1.7185213381356414e-06,
|
|
"loss": 0.1306,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06614295393228531,
|
|
"step": 11180,
|
|
"valid_targets_mean": 1429.9,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 6.16593164277839,
|
|
"grad_norm": 0.7603975523380472,
|
|
"learning_rate": 1.7073899822121532e-06,
|
|
"loss": 0.1326,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06398025155067444,
|
|
"step": 11185,
|
|
"valid_targets_mean": 1574.1,
|
|
"valid_targets_min": 637
|
|
},
|
|
{
|
|
"epoch": 6.168687982359427,
|
|
"grad_norm": 0.6944260352926153,
|
|
"learning_rate": 1.6962931863813103e-06,
|
|
"loss": 0.128,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06333683431148529,
|
|
"step": 11190,
|
|
"valid_targets_mean": 1737.8,
|
|
"valid_targets_min": 723
|
|
},
|
|
{
|
|
"epoch": 6.171444321940463,
|
|
"grad_norm": 0.7511518496646499,
|
|
"learning_rate": 1.6852309716082027e-06,
|
|
"loss": 0.1318,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05191802233457565,
|
|
"step": 11195,
|
|
"valid_targets_mean": 1306.0,
|
|
"valid_targets_min": 760
|
|
},
|
|
{
|
|
"epoch": 6.1742006615215,
|
|
"grad_norm": 0.8365524259933473,
|
|
"learning_rate": 1.6742033587925921e-06,
|
|
"loss": 0.1164,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06547710299491882,
|
|
"step": 11200,
|
|
"valid_targets_mean": 1363.9,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 6.176957001102536,
|
|
"grad_norm": 0.7699685231852054,
|
|
"learning_rate": 1.6632103687688617e-06,
|
|
"loss": 0.1357,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06085358187556267,
|
|
"step": 11205,
|
|
"valid_targets_mean": 1488.0,
|
|
"valid_targets_min": 866
|
|
},
|
|
{
|
|
"epoch": 6.1797133406835725,
|
|
"grad_norm": 0.8051575683922286,
|
|
"learning_rate": 1.652252022305989e-06,
|
|
"loss": 0.1296,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07382594794034958,
|
|
"step": 11210,
|
|
"valid_targets_mean": 1986.4,
|
|
"valid_targets_min": 788
|
|
},
|
|
{
|
|
"epoch": 6.182469680264608,
|
|
"grad_norm": 0.835449958961355,
|
|
"learning_rate": 1.6413283401074975e-06,
|
|
"loss": 0.1208,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0608520545065403,
|
|
"step": 11215,
|
|
"valid_targets_mean": 1518.4,
|
|
"valid_targets_min": 895
|
|
},
|
|
{
|
|
"epoch": 6.185226019845645,
|
|
"grad_norm": 0.7438098872774863,
|
|
"learning_rate": 1.6304393428114185e-06,
|
|
"loss": 0.1257,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06338074803352356,
|
|
"step": 11220,
|
|
"valid_targets_mean": 1784.0,
|
|
"valid_targets_min": 1005
|
|
},
|
|
{
|
|
"epoch": 6.187982359426681,
|
|
"grad_norm": 0.7136049137126398,
|
|
"learning_rate": 1.6195850509902534e-06,
|
|
"loss": 0.121,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05041872337460518,
|
|
"step": 11225,
|
|
"valid_targets_mean": 1389.1,
|
|
"valid_targets_min": 716
|
|
},
|
|
{
|
|
"epoch": 6.190738699007718,
|
|
"grad_norm": 0.7970816192388164,
|
|
"learning_rate": 1.6087654851509339e-06,
|
|
"loss": 0.1131,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04992698132991791,
|
|
"step": 11230,
|
|
"valid_targets_mean": 1125.1,
|
|
"valid_targets_min": 658
|
|
},
|
|
{
|
|
"epoch": 6.193495038588754,
|
|
"grad_norm": 0.777624346239686,
|
|
"learning_rate": 1.597980665734784e-06,
|
|
"loss": 0.1129,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055556513369083405,
|
|
"step": 11235,
|
|
"valid_targets_mean": 1460.8,
|
|
"valid_targets_min": 639
|
|
},
|
|
{
|
|
"epoch": 6.196251378169791,
|
|
"grad_norm": 0.7838459882133969,
|
|
"learning_rate": 1.5872306131174896e-06,
|
|
"loss": 0.1183,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04152761399745941,
|
|
"step": 11240,
|
|
"valid_targets_mean": 1127.0,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 6.199007717750827,
|
|
"grad_norm": 0.8438034097882097,
|
|
"learning_rate": 1.5765153476090444e-06,
|
|
"loss": 0.1255,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.053877003490924835,
|
|
"step": 11245,
|
|
"valid_targets_mean": 1318.9,
|
|
"valid_targets_min": 629
|
|
},
|
|
{
|
|
"epoch": 6.2017640573318635,
|
|
"grad_norm": 0.9607418166307967,
|
|
"learning_rate": 1.5658348894537167e-06,
|
|
"loss": 0.1286,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0717051550745964,
|
|
"step": 11250,
|
|
"valid_targets_mean": 1605.4,
|
|
"valid_targets_min": 622
|
|
},
|
|
{
|
|
"epoch": 6.2045203969128995,
|
|
"grad_norm": 0.7672895497178146,
|
|
"learning_rate": 1.5551892588300188e-06,
|
|
"loss": 0.1204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04917332902550697,
|
|
"step": 11255,
|
|
"valid_targets_mean": 1396.2,
|
|
"valid_targets_min": 939
|
|
},
|
|
{
|
|
"epoch": 6.207276736493936,
|
|
"grad_norm": 0.7673931974992443,
|
|
"learning_rate": 1.5445784758506599e-06,
|
|
"loss": 0.1226,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05353057384490967,
|
|
"step": 11260,
|
|
"valid_targets_mean": 1333.2,
|
|
"valid_targets_min": 775
|
|
},
|
|
{
|
|
"epoch": 6.210033076074972,
|
|
"grad_norm": 1.008305635626877,
|
|
"learning_rate": 1.5340025605625152e-06,
|
|
"loss": 0.123,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06336316466331482,
|
|
"step": 11265,
|
|
"valid_targets_mean": 1520.2,
|
|
"valid_targets_min": 620
|
|
},
|
|
{
|
|
"epoch": 6.212789415656009,
|
|
"grad_norm": 0.8082642551206577,
|
|
"learning_rate": 1.523461532946584e-06,
|
|
"loss": 0.1216,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.060453642159700394,
|
|
"step": 11270,
|
|
"valid_targets_mean": 1436.1,
|
|
"valid_targets_min": 690
|
|
},
|
|
{
|
|
"epoch": 6.215545755237045,
|
|
"grad_norm": 0.7834092473524754,
|
|
"learning_rate": 1.5129554129179514e-06,
|
|
"loss": 0.1164,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03289002925157547,
|
|
"step": 11275,
|
|
"valid_targets_mean": 846.4,
|
|
"valid_targets_min": 610
|
|
},
|
|
{
|
|
"epoch": 6.218302094818082,
|
|
"grad_norm": 0.9179313652337541,
|
|
"learning_rate": 1.5024842203257506e-06,
|
|
"loss": 0.114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.062398478388786316,
|
|
"step": 11280,
|
|
"valid_targets_mean": 1519.8,
|
|
"valid_targets_min": 792
|
|
},
|
|
{
|
|
"epoch": 6.221058434399118,
|
|
"grad_norm": 0.7299948986571572,
|
|
"learning_rate": 1.4920479749531302e-06,
|
|
"loss": 0.1137,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06160613149404526,
|
|
"step": 11285,
|
|
"valid_targets_mean": 1771.8,
|
|
"valid_targets_min": 561
|
|
},
|
|
{
|
|
"epoch": 6.223814773980155,
|
|
"grad_norm": 0.7369697313938885,
|
|
"learning_rate": 1.4816466965172094e-06,
|
|
"loss": 0.1184,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06131540611386299,
|
|
"step": 11290,
|
|
"valid_targets_mean": 1734.2,
|
|
"valid_targets_min": 1336
|
|
},
|
|
{
|
|
"epoch": 6.2265711135611905,
|
|
"grad_norm": 0.8619978796089565,
|
|
"learning_rate": 1.4712804046690466e-06,
|
|
"loss": 0.1205,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07115498185157776,
|
|
"step": 11295,
|
|
"valid_targets_mean": 1868.2,
|
|
"valid_targets_min": 1103
|
|
},
|
|
{
|
|
"epoch": 6.229327453142227,
|
|
"grad_norm": 0.8195687993482541,
|
|
"learning_rate": 1.460949118993602e-06,
|
|
"loss": 0.1262,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05849746614694595,
|
|
"step": 11300,
|
|
"valid_targets_mean": 1414.5,
|
|
"valid_targets_min": 677
|
|
},
|
|
{
|
|
"epoch": 6.232083792723263,
|
|
"grad_norm": 0.8033301430434138,
|
|
"learning_rate": 1.4506528590096957e-06,
|
|
"loss": 0.1145,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07568148523569107,
|
|
"step": 11305,
|
|
"valid_targets_mean": 1962.0,
|
|
"valid_targets_min": 978
|
|
},
|
|
{
|
|
"epoch": 6.2348401323043,
|
|
"grad_norm": 0.8056431643167312,
|
|
"learning_rate": 1.4403916441699762e-06,
|
|
"loss": 0.1203,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057772550731897354,
|
|
"step": 11310,
|
|
"valid_targets_mean": 1562.5,
|
|
"valid_targets_min": 935
|
|
},
|
|
{
|
|
"epoch": 6.237596471885336,
|
|
"grad_norm": 0.8264655064638785,
|
|
"learning_rate": 1.43016549386088e-06,
|
|
"loss": 0.1179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0661739706993103,
|
|
"step": 11315,
|
|
"valid_targets_mean": 1639.5,
|
|
"valid_targets_min": 803
|
|
},
|
|
{
|
|
"epoch": 6.240352811466373,
|
|
"grad_norm": 0.8038392873312592,
|
|
"learning_rate": 1.419974427402595e-06,
|
|
"loss": 0.121,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057862259447574615,
|
|
"step": 11320,
|
|
"valid_targets_mean": 1568.8,
|
|
"valid_targets_min": 1015
|
|
},
|
|
{
|
|
"epoch": 6.243109151047409,
|
|
"grad_norm": 0.8045691261312088,
|
|
"learning_rate": 1.4098184640490331e-06,
|
|
"loss": 0.114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.048569343984127045,
|
|
"step": 11325,
|
|
"valid_targets_mean": 1119.9,
|
|
"valid_targets_min": 881
|
|
},
|
|
{
|
|
"epoch": 6.245865490628446,
|
|
"grad_norm": 0.7407317637917925,
|
|
"learning_rate": 1.3996976229877768e-06,
|
|
"loss": 0.1116,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06087819114327431,
|
|
"step": 11330,
|
|
"valid_targets_mean": 1812.2,
|
|
"valid_targets_min": 1167
|
|
},
|
|
{
|
|
"epoch": 6.248621830209482,
|
|
"grad_norm": 0.8471710050745777,
|
|
"learning_rate": 1.3896119233400595e-06,
|
|
"loss": 0.107,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055279236286878586,
|
|
"step": 11335,
|
|
"valid_targets_mean": 1413.4,
|
|
"valid_targets_min": 676
|
|
},
|
|
{
|
|
"epoch": 6.251378169790518,
|
|
"grad_norm": 0.8442180157205643,
|
|
"learning_rate": 1.3795613841607148e-06,
|
|
"loss": 0.117,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046165142208337784,
|
|
"step": 11340,
|
|
"valid_targets_mean": 1048.5,
|
|
"valid_targets_min": 675
|
|
},
|
|
{
|
|
"epoch": 6.254134509371554,
|
|
"grad_norm": 0.757545564874326,
|
|
"learning_rate": 1.369546024438153e-06,
|
|
"loss": 0.1133,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052129071205854416,
|
|
"step": 11345,
|
|
"valid_targets_mean": 1424.6,
|
|
"valid_targets_min": 793
|
|
},
|
|
{
|
|
"epoch": 6.256890848952591,
|
|
"grad_norm": 0.7796226939132678,
|
|
"learning_rate": 1.3595658630943186e-06,
|
|
"loss": 0.115,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0571492537856102,
|
|
"step": 11350,
|
|
"valid_targets_mean": 1482.8,
|
|
"valid_targets_min": 741
|
|
},
|
|
{
|
|
"epoch": 6.259647188533627,
|
|
"grad_norm": 0.8386128191654393,
|
|
"learning_rate": 1.34962091898466e-06,
|
|
"loss": 0.1126,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06870017200708389,
|
|
"step": 11355,
|
|
"valid_targets_mean": 1727.5,
|
|
"valid_targets_min": 740
|
|
},
|
|
{
|
|
"epoch": 6.262403528114664,
|
|
"grad_norm": 1.2480903710158788,
|
|
"learning_rate": 1.3397112108980826e-06,
|
|
"loss": 0.119,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0706290602684021,
|
|
"step": 11360,
|
|
"valid_targets_mean": 1706.5,
|
|
"valid_targets_min": 1154
|
|
},
|
|
{
|
|
"epoch": 6.2651598676957,
|
|
"grad_norm": 0.8023820827103393,
|
|
"learning_rate": 1.329836757556926e-06,
|
|
"loss": 0.1073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06883151829242706,
|
|
"step": 11365,
|
|
"valid_targets_mean": 1717.0,
|
|
"valid_targets_min": 858
|
|
},
|
|
{
|
|
"epoch": 6.267916207276737,
|
|
"grad_norm": 0.8540060008332424,
|
|
"learning_rate": 1.3199975776169183e-06,
|
|
"loss": 0.1208,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06844878196716309,
|
|
"step": 11370,
|
|
"valid_targets_mean": 1628.0,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 6.270672546857773,
|
|
"grad_norm": 0.7553979795561787,
|
|
"learning_rate": 1.3101936896671562e-06,
|
|
"loss": 0.118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04996376112103462,
|
|
"step": 11375,
|
|
"valid_targets_mean": 1371.6,
|
|
"valid_targets_min": 1053
|
|
},
|
|
{
|
|
"epoch": 6.2734288864388095,
|
|
"grad_norm": 0.8853317840379998,
|
|
"learning_rate": 1.3004251122300481e-06,
|
|
"loss": 0.1242,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.060767095535993576,
|
|
"step": 11380,
|
|
"valid_targets_mean": 1571.4,
|
|
"valid_targets_min": 617
|
|
},
|
|
{
|
|
"epoch": 6.276185226019845,
|
|
"grad_norm": 0.754762050968365,
|
|
"learning_rate": 1.2906918637612975e-06,
|
|
"loss": 0.109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06336519122123718,
|
|
"step": 11385,
|
|
"valid_targets_mean": 1611.0,
|
|
"valid_targets_min": 626
|
|
},
|
|
{
|
|
"epoch": 6.278941565600882,
|
|
"grad_norm": 0.8842397384135064,
|
|
"learning_rate": 1.2809939626498547e-06,
|
|
"loss": 0.1134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06171749532222748,
|
|
"step": 11390,
|
|
"valid_targets_mean": 1875.5,
|
|
"valid_targets_min": 1166
|
|
},
|
|
{
|
|
"epoch": 6.281697905181918,
|
|
"grad_norm": 0.8212158933239772,
|
|
"learning_rate": 1.2713314272178967e-06,
|
|
"loss": 0.1132,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05758538842201233,
|
|
"step": 11395,
|
|
"valid_targets_mean": 1362.9,
|
|
"valid_targets_min": 696
|
|
},
|
|
{
|
|
"epoch": 6.284454244762955,
|
|
"grad_norm": 0.807746941011723,
|
|
"learning_rate": 1.2617042757207787e-06,
|
|
"loss": 0.1074,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0454978346824646,
|
|
"step": 11400,
|
|
"valid_targets_mean": 1171.2,
|
|
"valid_targets_min": 678
|
|
},
|
|
{
|
|
"epoch": 6.287210584343991,
|
|
"grad_norm": 0.8285847351702328,
|
|
"learning_rate": 1.2521125263470046e-06,
|
|
"loss": 0.1094,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.057846799492836,
|
|
"step": 11405,
|
|
"valid_targets_mean": 1548.8,
|
|
"valid_targets_min": 502
|
|
},
|
|
{
|
|
"epoch": 6.289966923925028,
|
|
"grad_norm": 0.7599152204606122,
|
|
"learning_rate": 1.2425561972181987e-06,
|
|
"loss": 0.112,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.061956629157066345,
|
|
"step": 11410,
|
|
"valid_targets_mean": 1877.4,
|
|
"valid_targets_min": 1330
|
|
},
|
|
{
|
|
"epoch": 6.292723263506064,
|
|
"grad_norm": 0.7727349638970457,
|
|
"learning_rate": 1.2330353063890588e-06,
|
|
"loss": 0.1114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04131048917770386,
|
|
"step": 11415,
|
|
"valid_targets_mean": 1301.9,
|
|
"valid_targets_min": 659
|
|
},
|
|
{
|
|
"epoch": 6.2954796030871005,
|
|
"grad_norm": 0.856138903436397,
|
|
"learning_rate": 1.223549871847336e-06,
|
|
"loss": 0.1149,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045268163084983826,
|
|
"step": 11420,
|
|
"valid_targets_mean": 1325.8,
|
|
"valid_targets_min": 955
|
|
},
|
|
{
|
|
"epoch": 6.2982359426681365,
|
|
"grad_norm": 0.7558641470758759,
|
|
"learning_rate": 1.2140999115137953e-06,
|
|
"loss": 0.1062,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.044058140367269516,
|
|
"step": 11425,
|
|
"valid_targets_mean": 1461.5,
|
|
"valid_targets_min": 962
|
|
},
|
|
{
|
|
"epoch": 6.300992282249173,
|
|
"grad_norm": 0.8326308589565506,
|
|
"learning_rate": 1.2046854432421706e-06,
|
|
"loss": 0.1156,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05293414741754532,
|
|
"step": 11430,
|
|
"valid_targets_mean": 1423.1,
|
|
"valid_targets_min": 792
|
|
},
|
|
{
|
|
"epoch": 6.303748621830209,
|
|
"grad_norm": 0.8355811923809932,
|
|
"learning_rate": 1.1953064848191497e-06,
|
|
"loss": 0.1091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06813247501850128,
|
|
"step": 11435,
|
|
"valid_targets_mean": 1629.1,
|
|
"valid_targets_min": 916
|
|
},
|
|
{
|
|
"epoch": 6.306504961411246,
|
|
"grad_norm": 0.8016395068939323,
|
|
"learning_rate": 1.1859630539643297e-06,
|
|
"loss": 0.114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04623547941446304,
|
|
"step": 11440,
|
|
"valid_targets_mean": 1243.4,
|
|
"valid_targets_min": 871
|
|
},
|
|
{
|
|
"epoch": 6.309261300992282,
|
|
"grad_norm": 0.8506172960165423,
|
|
"learning_rate": 1.176655168330183e-06,
|
|
"loss": 0.1137,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05749363824725151,
|
|
"step": 11445,
|
|
"valid_targets_mean": 1896.4,
|
|
"valid_targets_min": 1216
|
|
},
|
|
{
|
|
"epoch": 6.312017640573319,
|
|
"grad_norm": 0.8454401526137229,
|
|
"learning_rate": 1.167382845502032e-06,
|
|
"loss": 0.1015,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04869765043258667,
|
|
"step": 11450,
|
|
"valid_targets_mean": 1387.0,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 6.314773980154355,
|
|
"grad_norm": 0.8421950673667783,
|
|
"learning_rate": 1.1581461029980078e-06,
|
|
"loss": 0.1091,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054183073341846466,
|
|
"step": 11455,
|
|
"valid_targets_mean": 1290.0,
|
|
"valid_targets_min": 602
|
|
},
|
|
{
|
|
"epoch": 6.317530319735392,
|
|
"grad_norm": 0.7680387413960023,
|
|
"learning_rate": 1.148944958269016e-06,
|
|
"loss": 0.1169,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041217636317014694,
|
|
"step": 11460,
|
|
"valid_targets_mean": 1410.0,
|
|
"valid_targets_min": 745
|
|
},
|
|
{
|
|
"epoch": 6.3202866593164275,
|
|
"grad_norm": 0.875541457448063,
|
|
"learning_rate": 1.1397794286987107e-06,
|
|
"loss": 0.1062,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046386100351810455,
|
|
"step": 11465,
|
|
"valid_targets_mean": 1057.5,
|
|
"valid_targets_min": 722
|
|
},
|
|
{
|
|
"epoch": 6.323042998897464,
|
|
"grad_norm": 0.7746842236082632,
|
|
"learning_rate": 1.1306495316034648e-06,
|
|
"loss": 0.1115,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06386857479810715,
|
|
"step": 11470,
|
|
"valid_targets_mean": 1625.4,
|
|
"valid_targets_min": 1323
|
|
},
|
|
{
|
|
"epoch": 6.3257993384785,
|
|
"grad_norm": 0.7790673476639862,
|
|
"learning_rate": 1.1215552842323208e-06,
|
|
"loss": 0.1141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05024179816246033,
|
|
"step": 11475,
|
|
"valid_targets_mean": 1408.6,
|
|
"valid_targets_min": 901
|
|
},
|
|
{
|
|
"epoch": 6.328555678059537,
|
|
"grad_norm": 0.8100644666948812,
|
|
"learning_rate": 1.1124967037669764e-06,
|
|
"loss": 0.1119,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05275323987007141,
|
|
"step": 11480,
|
|
"valid_targets_mean": 1553.9,
|
|
"valid_targets_min": 802
|
|
},
|
|
{
|
|
"epoch": 6.331312017640573,
|
|
"grad_norm": 0.8629408449981302,
|
|
"learning_rate": 1.1034738073217356e-06,
|
|
"loss": 0.1127,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05800532549619675,
|
|
"step": 11485,
|
|
"valid_targets_mean": 1436.0,
|
|
"valid_targets_min": 1078
|
|
},
|
|
{
|
|
"epoch": 6.33406835722161,
|
|
"grad_norm": 0.8512344310170323,
|
|
"learning_rate": 1.0944866119434948e-06,
|
|
"loss": 0.108,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04837712645530701,
|
|
"step": 11490,
|
|
"valid_targets_mean": 1443.0,
|
|
"valid_targets_min": 1258
|
|
},
|
|
{
|
|
"epoch": 6.336824696802646,
|
|
"grad_norm": 0.8308390613316642,
|
|
"learning_rate": 1.0855351346116927e-06,
|
|
"loss": 0.1128,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05595929175615311,
|
|
"step": 11495,
|
|
"valid_targets_mean": 1482.1,
|
|
"valid_targets_min": 843
|
|
},
|
|
{
|
|
"epoch": 6.339581036383683,
|
|
"grad_norm": 0.8916099423962629,
|
|
"learning_rate": 1.0766193922382895e-06,
|
|
"loss": 0.1096,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054496049880981445,
|
|
"step": 11500,
|
|
"valid_targets_mean": 1415.9,
|
|
"valid_targets_min": 666
|
|
},
|
|
{
|
|
"epoch": 6.342337375964719,
|
|
"grad_norm": 0.9092783184100754,
|
|
"learning_rate": 1.0677394016677333e-06,
|
|
"loss": 0.1167,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06710590422153473,
|
|
"step": 11505,
|
|
"valid_targets_mean": 1815.6,
|
|
"valid_targets_min": 1039
|
|
},
|
|
{
|
|
"epoch": 6.345093715545755,
|
|
"grad_norm": 0.8333995636996362,
|
|
"learning_rate": 1.058895179676922e-06,
|
|
"loss": 0.1143,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04844500124454498,
|
|
"step": 11510,
|
|
"valid_targets_mean": 1323.5,
|
|
"valid_targets_min": 775
|
|
},
|
|
{
|
|
"epoch": 6.347850055126791,
|
|
"grad_norm": 0.9487824978086212,
|
|
"learning_rate": 1.050086742975176e-06,
|
|
"loss": 0.1128,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.059179145842790604,
|
|
"step": 11515,
|
|
"valid_targets_mean": 1817.0,
|
|
"valid_targets_min": 1280
|
|
},
|
|
{
|
|
"epoch": 6.350606394707828,
|
|
"grad_norm": 0.8311738815150421,
|
|
"learning_rate": 1.0413141082042167e-06,
|
|
"loss": 0.1071,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04196677356958389,
|
|
"step": 11520,
|
|
"valid_targets_mean": 922.0,
|
|
"valid_targets_min": 619
|
|
},
|
|
{
|
|
"epoch": 6.353362734288864,
|
|
"grad_norm": 0.806357576655406,
|
|
"learning_rate": 1.0325772919381127e-06,
|
|
"loss": 0.1075,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04718027636408806,
|
|
"step": 11525,
|
|
"valid_targets_mean": 1475.8,
|
|
"valid_targets_min": 655
|
|
},
|
|
{
|
|
"epoch": 6.356119073869901,
|
|
"grad_norm": 0.8356477608303491,
|
|
"learning_rate": 1.0238763106832694e-06,
|
|
"loss": 0.1083,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04415200650691986,
|
|
"step": 11530,
|
|
"valid_targets_mean": 1410.9,
|
|
"valid_targets_min": 694
|
|
},
|
|
{
|
|
"epoch": 6.358875413450937,
|
|
"grad_norm": 0.8301058780419746,
|
|
"learning_rate": 1.015211180878386e-06,
|
|
"loss": 0.1045,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.049798063933849335,
|
|
"step": 11535,
|
|
"valid_targets_mean": 1651.0,
|
|
"valid_targets_min": 814
|
|
},
|
|
{
|
|
"epoch": 6.361631753031974,
|
|
"grad_norm": 0.8315130749302949,
|
|
"learning_rate": 1.0065819188944247e-06,
|
|
"loss": 0.1037,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04856701195240021,
|
|
"step": 11540,
|
|
"valid_targets_mean": 1547.8,
|
|
"valid_targets_min": 849
|
|
},
|
|
{
|
|
"epoch": 6.36438809261301,
|
|
"grad_norm": 0.807277165113739,
|
|
"learning_rate": 9.97988541034589e-07,
|
|
"loss": 0.1065,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0515897162258625,
|
|
"step": 11545,
|
|
"valid_targets_mean": 1349.2,
|
|
"valid_targets_min": 746
|
|
},
|
|
{
|
|
"epoch": 6.3671444321940465,
|
|
"grad_norm": 0.7909631249430272,
|
|
"learning_rate": 9.894310635342852e-07,
|
|
"loss": 0.1163,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04889138787984848,
|
|
"step": 11550,
|
|
"valid_targets_mean": 1415.6,
|
|
"valid_targets_min": 929
|
|
},
|
|
{
|
|
"epoch": 6.369900771775082,
|
|
"grad_norm": 0.8480869616739024,
|
|
"learning_rate": 9.809095025610894e-07,
|
|
"loss": 0.107,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.054064054042100906,
|
|
"step": 11555,
|
|
"valid_targets_mean": 1389.9,
|
|
"valid_targets_min": 826
|
|
},
|
|
{
|
|
"epoch": 6.372657111356119,
|
|
"grad_norm": 0.8134859126893602,
|
|
"learning_rate": 9.724238742147207e-07,
|
|
"loss": 0.1059,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04554691165685654,
|
|
"step": 11560,
|
|
"valid_targets_mean": 1378.1,
|
|
"valid_targets_min": 908
|
|
},
|
|
{
|
|
"epoch": 6.375413450937155,
|
|
"grad_norm": 0.7439656404959266,
|
|
"learning_rate": 9.63974194527022e-07,
|
|
"loss": 0.1095,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05691155046224594,
|
|
"step": 11565,
|
|
"valid_targets_mean": 1676.0,
|
|
"valid_targets_min": 1090
|
|
},
|
|
{
|
|
"epoch": 6.378169790518192,
|
|
"grad_norm": 0.8729025840949677,
|
|
"learning_rate": 9.555604794619056e-07,
|
|
"loss": 0.1008,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04898655787110329,
|
|
"step": 11570,
|
|
"valid_targets_mean": 1115.2,
|
|
"valid_targets_min": 719
|
|
},
|
|
{
|
|
"epoch": 6.380926130099228,
|
|
"grad_norm": 0.7788807783458898,
|
|
"learning_rate": 9.471827449153425e-07,
|
|
"loss": 0.1093,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.052582141011953354,
|
|
"step": 11575,
|
|
"valid_targets_mean": 1533.0,
|
|
"valid_targets_min": 609
|
|
},
|
|
{
|
|
"epoch": 6.383682469680265,
|
|
"grad_norm": 0.7821629179685586,
|
|
"learning_rate": 9.38841006715323e-07,
|
|
"loss": 0.1055,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.055352941155433655,
|
|
"step": 11580,
|
|
"valid_targets_mean": 1564.6,
|
|
"valid_targets_min": 710
|
|
},
|
|
{
|
|
"epoch": 6.386438809261301,
|
|
"grad_norm": 0.8476245966212717,
|
|
"learning_rate": 9.305352806218338e-07,
|
|
"loss": 0.1114,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.051344215869903564,
|
|
"step": 11585,
|
|
"valid_targets_mean": 1533.1,
|
|
"valid_targets_min": 873
|
|
},
|
|
{
|
|
"epoch": 6.3891951488423375,
|
|
"grad_norm": 0.8008545618821022,
|
|
"learning_rate": 9.222655823268201e-07,
|
|
"loss": 0.109,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05269902944564819,
|
|
"step": 11590,
|
|
"valid_targets_mean": 1465.5,
|
|
"valid_targets_min": 764
|
|
},
|
|
{
|
|
"epoch": 6.3919514884233735,
|
|
"grad_norm": 0.8389524717526698,
|
|
"learning_rate": 9.140319274541621e-07,
|
|
"loss": 0.1071,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04479958862066269,
|
|
"step": 11595,
|
|
"valid_targets_mean": 1249.2,
|
|
"valid_targets_min": 815
|
|
},
|
|
{
|
|
"epoch": 6.39470782800441,
|
|
"grad_norm": 0.8593865650181137,
|
|
"learning_rate": 9.058343315596407e-07,
|
|
"loss": 0.1089,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.05146509408950806,
|
|
"step": 11600,
|
|
"valid_targets_mean": 1477.4,
|
|
"valid_targets_min": 1179
|
|
},
|
|
{
|
|
"epoch": 6.397464167585446,
|
|
"grad_norm": 0.9176074085995714,
|
|
"learning_rate": 8.976728101309162e-07,
|
|
"loss": 0.1992,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10465699434280396,
|
|
"step": 11605,
|
|
"valid_targets_mean": 4555.2,
|
|
"valid_targets_min": 700
|
|
},
|
|
{
|
|
"epoch": 6.400220507166483,
|
|
"grad_norm": 0.9449246452088759,
|
|
"learning_rate": 8.895473785874875e-07,
|
|
"loss": 0.219,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12931451201438904,
|
|
"step": 11610,
|
|
"valid_targets_mean": 5515.5,
|
|
"valid_targets_min": 1004
|
|
},
|
|
{
|
|
"epoch": 6.402976846747519,
|
|
"grad_norm": 0.773857313274555,
|
|
"learning_rate": 8.814580522806748e-07,
|
|
"loss": 0.2144,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08493387699127197,
|
|
"step": 11615,
|
|
"valid_targets_mean": 4310.8,
|
|
"valid_targets_min": 2049
|
|
},
|
|
{
|
|
"epoch": 6.405733186328556,
|
|
"grad_norm": 0.7918580057742662,
|
|
"learning_rate": 8.734048464935841e-07,
|
|
"loss": 0.2101,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08387384563684464,
|
|
"step": 11620,
|
|
"valid_targets_mean": 3495.1,
|
|
"valid_targets_min": 1559
|
|
},
|
|
{
|
|
"epoch": 6.408489525909592,
|
|
"grad_norm": 0.7497232558586551,
|
|
"learning_rate": 8.653877764410734e-07,
|
|
"loss": 0.2087,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1300373375415802,
|
|
"step": 11625,
|
|
"valid_targets_mean": 5011.6,
|
|
"valid_targets_min": 1969
|
|
},
|
|
{
|
|
"epoch": 6.411245865490629,
|
|
"grad_norm": 0.7257678193954185,
|
|
"learning_rate": 8.574068572697359e-07,
|
|
"loss": 0.214,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1075010821223259,
|
|
"step": 11630,
|
|
"valid_targets_mean": 6144.0,
|
|
"valid_targets_min": 1649
|
|
},
|
|
{
|
|
"epoch": 6.4140022050716645,
|
|
"grad_norm": 0.7008508691299757,
|
|
"learning_rate": 8.494621040578632e-07,
|
|
"loss": 0.2022,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11478772759437561,
|
|
"step": 11635,
|
|
"valid_targets_mean": 7660.9,
|
|
"valid_targets_min": 2225
|
|
},
|
|
{
|
|
"epoch": 6.416758544652701,
|
|
"grad_norm": 0.6510824509551777,
|
|
"learning_rate": 8.415535318154178e-07,
|
|
"loss": 0.2021,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06924179196357727,
|
|
"step": 11640,
|
|
"valid_targets_mean": 3117.5,
|
|
"valid_targets_min": 466
|
|
},
|
|
{
|
|
"epoch": 6.419514884233737,
|
|
"grad_norm": 0.646089132095851,
|
|
"learning_rate": 8.336811554840074e-07,
|
|
"loss": 0.2118,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12336082011461258,
|
|
"step": 11645,
|
|
"valid_targets_mean": 6088.8,
|
|
"valid_targets_min": 1747
|
|
},
|
|
{
|
|
"epoch": 6.422271223814774,
|
|
"grad_norm": 0.6306370928998536,
|
|
"learning_rate": 8.258449899368526e-07,
|
|
"loss": 0.2093,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09987568855285645,
|
|
"step": 11650,
|
|
"valid_targets_mean": 4500.0,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 6.42502756339581,
|
|
"grad_norm": 0.6466288811586338,
|
|
"learning_rate": 8.180450499787662e-07,
|
|
"loss": 0.204,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10034269094467163,
|
|
"step": 11655,
|
|
"valid_targets_mean": 3932.4,
|
|
"valid_targets_min": 1221
|
|
},
|
|
{
|
|
"epoch": 6.427783902976847,
|
|
"grad_norm": 0.5735423975472115,
|
|
"learning_rate": 8.102813503461115e-07,
|
|
"loss": 0.2072,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.108061283826828,
|
|
"step": 11660,
|
|
"valid_targets_mean": 5551.0,
|
|
"valid_targets_min": 1550
|
|
},
|
|
{
|
|
"epoch": 6.430540242557883,
|
|
"grad_norm": 0.6046605200372734,
|
|
"learning_rate": 8.025539057067954e-07,
|
|
"loss": 0.2042,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10980994254350662,
|
|
"step": 11665,
|
|
"valid_targets_mean": 4318.0,
|
|
"valid_targets_min": 1546
|
|
},
|
|
{
|
|
"epoch": 6.43329658213892,
|
|
"grad_norm": 0.6305184940295764,
|
|
"learning_rate": 7.948627306602174e-07,
|
|
"loss": 0.1994,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10546945035457611,
|
|
"step": 11670,
|
|
"valid_targets_mean": 4605.4,
|
|
"valid_targets_min": 1956
|
|
},
|
|
{
|
|
"epoch": 6.436052921719956,
|
|
"grad_norm": 0.5947982027416145,
|
|
"learning_rate": 7.872078397372607e-07,
|
|
"loss": 0.1964,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0815252959728241,
|
|
"step": 11675,
|
|
"valid_targets_mean": 3389.5,
|
|
"valid_targets_min": 1538
|
|
},
|
|
{
|
|
"epoch": 6.438809261300992,
|
|
"grad_norm": 0.6807347272565658,
|
|
"learning_rate": 7.795892474002497e-07,
|
|
"loss": 0.197,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10113245993852615,
|
|
"step": 11680,
|
|
"valid_targets_mean": 3408.9,
|
|
"valid_targets_min": 1639
|
|
},
|
|
{
|
|
"epoch": 6.441565600882028,
|
|
"grad_norm": 0.6491583085672764,
|
|
"learning_rate": 7.720069680429399e-07,
|
|
"loss": 0.1991,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09409807622432709,
|
|
"step": 11685,
|
|
"valid_targets_mean": 3183.5,
|
|
"valid_targets_min": 935
|
|
},
|
|
{
|
|
"epoch": 6.444321940463065,
|
|
"grad_norm": 0.6716017143490126,
|
|
"learning_rate": 7.644610159904719e-07,
|
|
"loss": 0.2021,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10768577456474304,
|
|
"step": 11690,
|
|
"valid_targets_mean": 3576.2,
|
|
"valid_targets_min": 1266
|
|
},
|
|
{
|
|
"epoch": 6.447078280044101,
|
|
"grad_norm": 0.678915511303888,
|
|
"learning_rate": 7.569514054993576e-07,
|
|
"loss": 0.1979,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07824469357728958,
|
|
"step": 11695,
|
|
"valid_targets_mean": 2092.0,
|
|
"valid_targets_min": 1170
|
|
},
|
|
{
|
|
"epoch": 6.449834619625138,
|
|
"grad_norm": 0.5858357127061398,
|
|
"learning_rate": 7.494781507574478e-07,
|
|
"loss": 0.2007,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09551960229873657,
|
|
"step": 11700,
|
|
"valid_targets_mean": 4129.1,
|
|
"valid_targets_min": 1668
|
|
},
|
|
{
|
|
"epoch": 6.452590959206174,
|
|
"grad_norm": 0.6536341332706154,
|
|
"learning_rate": 7.42041265883906e-07,
|
|
"loss": 0.1971,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09203163534402847,
|
|
"step": 11705,
|
|
"valid_targets_mean": 3367.0,
|
|
"valid_targets_min": 1329
|
|
},
|
|
{
|
|
"epoch": 6.455347298787211,
|
|
"grad_norm": 0.6795829626456635,
|
|
"learning_rate": 7.346407649291887e-07,
|
|
"loss": 0.1876,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11640680581331253,
|
|
"step": 11710,
|
|
"valid_targets_mean": 3943.2,
|
|
"valid_targets_min": 2108
|
|
},
|
|
{
|
|
"epoch": 6.458103638368247,
|
|
"grad_norm": 0.6482003760919435,
|
|
"learning_rate": 7.272766618750049e-07,
|
|
"loss": 0.1788,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08040355145931244,
|
|
"step": 11715,
|
|
"valid_targets_mean": 4311.9,
|
|
"valid_targets_min": 2214
|
|
},
|
|
{
|
|
"epoch": 6.4608599779492835,
|
|
"grad_norm": 0.7440207435500588,
|
|
"learning_rate": 7.19948970634301e-07,
|
|
"loss": 0.1952,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1044035479426384,
|
|
"step": 11720,
|
|
"valid_targets_mean": 3457.4,
|
|
"valid_targets_min": 1561
|
|
},
|
|
{
|
|
"epoch": 6.463616317530319,
|
|
"grad_norm": 0.6344714155903496,
|
|
"learning_rate": 7.126577050512274e-07,
|
|
"loss": 0.1984,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09968722611665726,
|
|
"step": 11725,
|
|
"valid_targets_mean": 3474.9,
|
|
"valid_targets_min": 771
|
|
},
|
|
{
|
|
"epoch": 6.466372657111356,
|
|
"grad_norm": 0.742025713479001,
|
|
"learning_rate": 7.05402878901118e-07,
|
|
"loss": 0.1992,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09637433290481567,
|
|
"step": 11730,
|
|
"valid_targets_mean": 4008.4,
|
|
"valid_targets_min": 1770
|
|
},
|
|
{
|
|
"epoch": 6.469128996692392,
|
|
"grad_norm": 0.7526829785345676,
|
|
"learning_rate": 6.981845058904646e-07,
|
|
"loss": 0.1857,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07750971615314484,
|
|
"step": 11735,
|
|
"valid_targets_mean": 2430.8,
|
|
"valid_targets_min": 994
|
|
},
|
|
{
|
|
"epoch": 6.471885336273429,
|
|
"grad_norm": 0.6321894503162707,
|
|
"learning_rate": 6.91002599656887e-07,
|
|
"loss": 0.1868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09305538237094879,
|
|
"step": 11740,
|
|
"valid_targets_mean": 3961.1,
|
|
"valid_targets_min": 1738
|
|
},
|
|
{
|
|
"epoch": 6.474641675854465,
|
|
"grad_norm": 0.6708597982593111,
|
|
"learning_rate": 6.838571737691024e-07,
|
|
"loss": 0.1913,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08937305212020874,
|
|
"step": 11745,
|
|
"valid_targets_mean": 3741.0,
|
|
"valid_targets_min": 862
|
|
},
|
|
{
|
|
"epoch": 6.477398015435502,
|
|
"grad_norm": 0.6499275659291807,
|
|
"learning_rate": 6.767482417269145e-07,
|
|
"loss": 0.1881,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08986106514930725,
|
|
"step": 11750,
|
|
"valid_targets_mean": 3093.5,
|
|
"valid_targets_min": 1408
|
|
},
|
|
{
|
|
"epoch": 6.480154355016538,
|
|
"grad_norm": 0.5849415776568183,
|
|
"learning_rate": 6.69675816961175e-07,
|
|
"loss": 0.1813,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08190472424030304,
|
|
"step": 11755,
|
|
"valid_targets_mean": 4835.2,
|
|
"valid_targets_min": 2110
|
|
},
|
|
{
|
|
"epoch": 6.4829106945975745,
|
|
"grad_norm": 0.65553427744922,
|
|
"learning_rate": 6.626399128337647e-07,
|
|
"loss": 0.1919,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08421658724546432,
|
|
"step": 11760,
|
|
"valid_targets_mean": 3404.2,
|
|
"valid_targets_min": 1317
|
|
},
|
|
{
|
|
"epoch": 6.4856670341786105,
|
|
"grad_norm": 0.592439353371577,
|
|
"learning_rate": 6.556405426375634e-07,
|
|
"loss": 0.1886,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09729952365159988,
|
|
"step": 11765,
|
|
"valid_targets_mean": 4223.4,
|
|
"valid_targets_min": 1863
|
|
},
|
|
{
|
|
"epoch": 6.488423373759647,
|
|
"grad_norm": 0.6723522803319498,
|
|
"learning_rate": 6.486777195964334e-07,
|
|
"loss": 0.1884,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10911525785923004,
|
|
"step": 11770,
|
|
"valid_targets_mean": 3774.9,
|
|
"valid_targets_min": 2594
|
|
},
|
|
{
|
|
"epoch": 6.491179713340683,
|
|
"grad_norm": 0.6513073354297014,
|
|
"learning_rate": 6.417514568651783e-07,
|
|
"loss": 0.1861,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08294127881526947,
|
|
"step": 11775,
|
|
"valid_targets_mean": 3062.8,
|
|
"valid_targets_min": 1443
|
|
},
|
|
{
|
|
"epoch": 6.49393605292172,
|
|
"grad_norm": 0.6915676778793636,
|
|
"learning_rate": 6.348617675295399e-07,
|
|
"loss": 0.1893,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09108276665210724,
|
|
"step": 11780,
|
|
"valid_targets_mean": 3855.1,
|
|
"valid_targets_min": 1268
|
|
},
|
|
{
|
|
"epoch": 6.496692392502756,
|
|
"grad_norm": 0.5980086158170506,
|
|
"learning_rate": 6.280086646061567e-07,
|
|
"loss": 0.1915,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10236439853906631,
|
|
"step": 11785,
|
|
"valid_targets_mean": 5580.1,
|
|
"valid_targets_min": 2023
|
|
},
|
|
{
|
|
"epoch": 6.499448732083793,
|
|
"grad_norm": 0.6193873598614877,
|
|
"learning_rate": 6.211921610425431e-07,
|
|
"loss": 0.1852,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10233017057180405,
|
|
"step": 11790,
|
|
"valid_targets_mean": 4074.2,
|
|
"valid_targets_min": 1922
|
|
},
|
|
{
|
|
"epoch": 6.502205071664829,
|
|
"grad_norm": 0.6497140939363708,
|
|
"learning_rate": 6.144122697170707e-07,
|
|
"loss": 0.1948,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1002703532576561,
|
|
"step": 11795,
|
|
"valid_targets_mean": 3672.6,
|
|
"valid_targets_min": 2089
|
|
},
|
|
{
|
|
"epoch": 6.504961411245866,
|
|
"grad_norm": 0.6859752307139635,
|
|
"learning_rate": 6.076690034389355e-07,
|
|
"loss": 0.1934,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12542329728603363,
|
|
"step": 11800,
|
|
"valid_targets_mean": 4324.4,
|
|
"valid_targets_min": 1363
|
|
},
|
|
{
|
|
"epoch": 6.5077177508269015,
|
|
"grad_norm": 0.6628387168604822,
|
|
"learning_rate": 6.009623749481442e-07,
|
|
"loss": 0.1813,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09685531258583069,
|
|
"step": 11805,
|
|
"valid_targets_mean": 3350.5,
|
|
"valid_targets_min": 1087
|
|
},
|
|
{
|
|
"epoch": 6.510474090407938,
|
|
"grad_norm": 0.6355519829422334,
|
|
"learning_rate": 5.94292396915479e-07,
|
|
"loss": 0.1906,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11752857267856598,
|
|
"step": 11810,
|
|
"valid_targets_mean": 4624.4,
|
|
"valid_targets_min": 2476
|
|
},
|
|
{
|
|
"epoch": 6.513230429988974,
|
|
"grad_norm": 0.6487773894934018,
|
|
"learning_rate": 5.876590819424777e-07,
|
|
"loss": 0.1922,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1067754477262497,
|
|
"step": 11815,
|
|
"valid_targets_mean": 4545.9,
|
|
"valid_targets_min": 2311
|
|
},
|
|
{
|
|
"epoch": 6.515986769570011,
|
|
"grad_norm": 0.634694380376294,
|
|
"learning_rate": 5.810624425614175e-07,
|
|
"loss": 0.1852,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09914753586053848,
|
|
"step": 11820,
|
|
"valid_targets_mean": 3913.2,
|
|
"valid_targets_min": 2010
|
|
},
|
|
{
|
|
"epoch": 6.518743109151047,
|
|
"grad_norm": 0.5933522994140181,
|
|
"learning_rate": 5.745024912352759e-07,
|
|
"loss": 0.1821,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06522028893232346,
|
|
"step": 11825,
|
|
"valid_targets_mean": 2885.6,
|
|
"valid_targets_min": 1294
|
|
},
|
|
{
|
|
"epoch": 6.521499448732084,
|
|
"grad_norm": 0.6755619370844584,
|
|
"learning_rate": 5.679792403577234e-07,
|
|
"loss": 0.1926,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11111055314540863,
|
|
"step": 11830,
|
|
"valid_targets_mean": 3632.0,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 6.52425578831312,
|
|
"grad_norm": 0.6927068446110023,
|
|
"learning_rate": 5.614927022530881e-07,
|
|
"loss": 0.188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08472685515880585,
|
|
"step": 11835,
|
|
"valid_targets_mean": 2715.8,
|
|
"valid_targets_min": 1582
|
|
},
|
|
{
|
|
"epoch": 6.527012127894157,
|
|
"grad_norm": 0.6988293977660577,
|
|
"learning_rate": 5.550428891763382e-07,
|
|
"loss": 0.1859,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09875528514385223,
|
|
"step": 11840,
|
|
"valid_targets_mean": 3353.0,
|
|
"valid_targets_min": 2403
|
|
},
|
|
{
|
|
"epoch": 6.529768467475193,
|
|
"grad_norm": 0.6273384763309744,
|
|
"learning_rate": 5.486298133130574e-07,
|
|
"loss": 0.1916,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09804292768239975,
|
|
"step": 11845,
|
|
"valid_targets_mean": 4056.6,
|
|
"valid_targets_min": 2288
|
|
},
|
|
{
|
|
"epoch": 6.532524807056229,
|
|
"grad_norm": 0.6451506294727752,
|
|
"learning_rate": 5.422534867794227e-07,
|
|
"loss": 0.1935,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12458132207393646,
|
|
"step": 11850,
|
|
"valid_targets_mean": 5186.2,
|
|
"valid_targets_min": 2187
|
|
},
|
|
{
|
|
"epoch": 6.535281146637265,
|
|
"grad_norm": 0.6674053889520452,
|
|
"learning_rate": 5.359139216221843e-07,
|
|
"loss": 0.1886,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11639135330915451,
|
|
"step": 11855,
|
|
"valid_targets_mean": 4289.0,
|
|
"valid_targets_min": 2606
|
|
},
|
|
{
|
|
"epoch": 6.538037486218302,
|
|
"grad_norm": 0.6726664995218427,
|
|
"learning_rate": 5.296111298186324e-07,
|
|
"loss": 0.1982,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.13616593182086945,
|
|
"step": 11860,
|
|
"valid_targets_mean": 4530.1,
|
|
"valid_targets_min": 2009
|
|
},
|
|
{
|
|
"epoch": 6.540793825799338,
|
|
"grad_norm": 0.5982888428943166,
|
|
"learning_rate": 5.23345123276584e-07,
|
|
"loss": 0.1781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07824790477752686,
|
|
"step": 11865,
|
|
"valid_targets_mean": 3603.1,
|
|
"valid_targets_min": 1902
|
|
},
|
|
{
|
|
"epoch": 6.543550165380375,
|
|
"grad_norm": 0.6439790549528115,
|
|
"learning_rate": 5.171159138343607e-07,
|
|
"loss": 0.1792,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10575902462005615,
|
|
"step": 11870,
|
|
"valid_targets_mean": 3794.1,
|
|
"valid_targets_min": 2261
|
|
},
|
|
{
|
|
"epoch": 6.546306504961411,
|
|
"grad_norm": 0.5879830132568188,
|
|
"learning_rate": 5.109235132607593e-07,
|
|
"loss": 0.1846,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08203437179327011,
|
|
"step": 11875,
|
|
"valid_targets_mean": 4446.4,
|
|
"valid_targets_min": 902
|
|
},
|
|
{
|
|
"epoch": 6.549062844542448,
|
|
"grad_norm": 0.7049133128520148,
|
|
"learning_rate": 5.047679332550392e-07,
|
|
"loss": 0.1912,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.11049512028694153,
|
|
"step": 11880,
|
|
"valid_targets_mean": 3782.0,
|
|
"valid_targets_min": 1218
|
|
},
|
|
{
|
|
"epoch": 6.551819184123484,
|
|
"grad_norm": 0.6402981702549861,
|
|
"learning_rate": 4.986491854468933e-07,
|
|
"loss": 0.1908,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09350930154323578,
|
|
"step": 11885,
|
|
"valid_targets_mean": 4039.6,
|
|
"valid_targets_min": 1658
|
|
},
|
|
{
|
|
"epoch": 6.5545755237045205,
|
|
"grad_norm": 0.6646866548946285,
|
|
"learning_rate": 4.925672813964211e-07,
|
|
"loss": 0.1893,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0934898778796196,
|
|
"step": 11890,
|
|
"valid_targets_mean": 3492.2,
|
|
"valid_targets_min": 1421
|
|
},
|
|
{
|
|
"epoch": 6.557331863285556,
|
|
"grad_norm": 0.6020260218496069,
|
|
"learning_rate": 4.865222325941244e-07,
|
|
"loss": 0.1868,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07506995648145676,
|
|
"step": 11895,
|
|
"valid_targets_mean": 3909.8,
|
|
"valid_targets_min": 2315
|
|
},
|
|
{
|
|
"epoch": 6.560088202866593,
|
|
"grad_norm": 0.6290719045147388,
|
|
"learning_rate": 4.805140504608629e-07,
|
|
"loss": 0.1974,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08875058591365814,
|
|
"step": 11900,
|
|
"valid_targets_mean": 3644.5,
|
|
"valid_targets_min": 1187
|
|
},
|
|
{
|
|
"epoch": 6.56284454244763,
|
|
"grad_norm": 0.5942470606565454,
|
|
"learning_rate": 4.7454274634785914e-07,
|
|
"loss": 0.1941,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07791933417320251,
|
|
"step": 11905,
|
|
"valid_targets_mean": 3410.8,
|
|
"valid_targets_min": 616
|
|
},
|
|
{
|
|
"epoch": 6.565600882028666,
|
|
"grad_norm": 0.67540829739137,
|
|
"learning_rate": 4.686083315366463e-07,
|
|
"loss": 0.1883,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08648917078971863,
|
|
"step": 11910,
|
|
"valid_targets_mean": 2662.8,
|
|
"valid_targets_min": 1363
|
|
},
|
|
{
|
|
"epoch": 6.568357221609702,
|
|
"grad_norm": 0.63323653327467,
|
|
"learning_rate": 4.627108172390737e-07,
|
|
"loss": 0.1838,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09705701470375061,
|
|
"step": 11915,
|
|
"valid_targets_mean": 3497.0,
|
|
"valid_targets_min": 1393
|
|
},
|
|
{
|
|
"epoch": 6.571113561190739,
|
|
"grad_norm": 0.6144191040593426,
|
|
"learning_rate": 4.5685021459727305e-07,
|
|
"loss": 0.178,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09148149937391281,
|
|
"step": 11920,
|
|
"valid_targets_mean": 4659.9,
|
|
"valid_targets_min": 1414
|
|
},
|
|
{
|
|
"epoch": 6.573869900771776,
|
|
"grad_norm": 0.6160456591816065,
|
|
"learning_rate": 4.510265346836318e-07,
|
|
"loss": 0.1773,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09587536752223969,
|
|
"step": 11925,
|
|
"valid_targets_mean": 3795.8,
|
|
"valid_targets_min": 620
|
|
},
|
|
{
|
|
"epoch": 6.5766262403528115,
|
|
"grad_norm": 0.6008971160887345,
|
|
"learning_rate": 4.4523978850079087e-07,
|
|
"loss": 0.174,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06938239932060242,
|
|
"step": 11930,
|
|
"valid_targets_mean": 3543.5,
|
|
"valid_targets_min": 1286
|
|
},
|
|
{
|
|
"epoch": 6.5793825799338475,
|
|
"grad_norm": 0.7309277039332485,
|
|
"learning_rate": 4.3948998698160496e-07,
|
|
"loss": 0.1713,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07072250545024872,
|
|
"step": 11935,
|
|
"valid_targets_mean": 2120.8,
|
|
"valid_targets_min": 641
|
|
},
|
|
{
|
|
"epoch": 6.582138919514884,
|
|
"grad_norm": 0.6979346136216782,
|
|
"learning_rate": 4.3377714098913115e-07,
|
|
"loss": 0.1717,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0994897112250328,
|
|
"step": 11940,
|
|
"valid_targets_mean": 3445.9,
|
|
"valid_targets_min": 2096
|
|
},
|
|
{
|
|
"epoch": 6.584895259095921,
|
|
"grad_norm": 0.6723001936631122,
|
|
"learning_rate": 4.281012613166069e-07,
|
|
"loss": 0.1869,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.1332911252975464,
|
|
"step": 11945,
|
|
"valid_targets_mean": 4586.4,
|
|
"valid_targets_min": 2783
|
|
},
|
|
{
|
|
"epoch": 6.587651598676957,
|
|
"grad_norm": 0.6422451921833976,
|
|
"learning_rate": 4.2246235868742993e-07,
|
|
"loss": 0.1968,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08041521161794662,
|
|
"step": 11950,
|
|
"valid_targets_mean": 3270.9,
|
|
"valid_targets_min": 1502
|
|
},
|
|
{
|
|
"epoch": 6.590407938257993,
|
|
"grad_norm": 0.735852275377291,
|
|
"learning_rate": 4.1686044375513824e-07,
|
|
"loss": 0.179,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10889242589473724,
|
|
"step": 11955,
|
|
"valid_targets_mean": 3216.4,
|
|
"valid_targets_min": 986
|
|
},
|
|
{
|
|
"epoch": 6.59316427783903,
|
|
"grad_norm": 0.6059050413169834,
|
|
"learning_rate": 4.1129552710338806e-07,
|
|
"loss": 0.1853,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0799802839756012,
|
|
"step": 11960,
|
|
"valid_targets_mean": 3652.0,
|
|
"valid_targets_min": 1605
|
|
},
|
|
{
|
|
"epoch": 6.595920617420067,
|
|
"grad_norm": 0.609617554592497,
|
|
"learning_rate": 4.0576761924593364e-07,
|
|
"loss": 0.186,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0994764119386673,
|
|
"step": 11965,
|
|
"valid_targets_mean": 4645.4,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 6.598676957001103,
|
|
"grad_norm": 0.5430993562702443,
|
|
"learning_rate": 4.00276730626612e-07,
|
|
"loss": 0.1792,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07788201421499252,
|
|
"step": 11970,
|
|
"valid_targets_mean": 5431.8,
|
|
"valid_targets_min": 1762
|
|
},
|
|
{
|
|
"epoch": 6.6014332965821385,
|
|
"grad_norm": 0.5463431264535845,
|
|
"learning_rate": 3.948228716193181e-07,
|
|
"loss": 0.1859,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08277466893196106,
|
|
"step": 11975,
|
|
"valid_targets_mean": 4498.6,
|
|
"valid_targets_min": 2379
|
|
},
|
|
{
|
|
"epoch": 6.604189636163175,
|
|
"grad_norm": 0.7239435036122048,
|
|
"learning_rate": 3.8940605252798525e-07,
|
|
"loss": 0.1769,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08240791410207748,
|
|
"step": 11980,
|
|
"valid_targets_mean": 2672.5,
|
|
"valid_targets_min": 1531
|
|
},
|
|
{
|
|
"epoch": 6.606945975744212,
|
|
"grad_norm": 0.6407112652275979,
|
|
"learning_rate": 3.840262835865716e-07,
|
|
"loss": 0.1841,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06895279884338379,
|
|
"step": 11985,
|
|
"valid_targets_mean": 3118.6,
|
|
"valid_targets_min": 1471
|
|
},
|
|
{
|
|
"epoch": 6.609702315325248,
|
|
"grad_norm": 0.7060189967326987,
|
|
"learning_rate": 3.7868357495903563e-07,
|
|
"loss": 0.1795,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09895411133766174,
|
|
"step": 11990,
|
|
"valid_targets_mean": 3860.9,
|
|
"valid_targets_min": 1697
|
|
},
|
|
{
|
|
"epoch": 6.612458654906284,
|
|
"grad_norm": 0.6371515759837318,
|
|
"learning_rate": 3.7337793673931423e-07,
|
|
"loss": 0.1869,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08876638114452362,
|
|
"step": 11995,
|
|
"valid_targets_mean": 3230.8,
|
|
"valid_targets_min": 1667
|
|
},
|
|
{
|
|
"epoch": 6.615214994487321,
|
|
"grad_norm": 0.7289358631582122,
|
|
"learning_rate": 3.681093789513135e-07,
|
|
"loss": 0.1912,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08727249503135681,
|
|
"step": 12000,
|
|
"valid_targets_mean": 3201.0,
|
|
"valid_targets_min": 1316
|
|
},
|
|
{
|
|
"epoch": 6.617971334068358,
|
|
"grad_norm": 0.719491615480887,
|
|
"learning_rate": 3.6287791154887785e-07,
|
|
"loss": 0.1879,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10242816805839539,
|
|
"step": 12005,
|
|
"valid_targets_mean": 2652.2,
|
|
"valid_targets_min": 1001
|
|
},
|
|
{
|
|
"epoch": 6.620727673649394,
|
|
"grad_norm": 0.6690614702555859,
|
|
"learning_rate": 3.5768354441578333e-07,
|
|
"loss": 0.1974,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0773639902472496,
|
|
"step": 12010,
|
|
"valid_targets_mean": 2528.5,
|
|
"valid_targets_min": 843
|
|
},
|
|
{
|
|
"epoch": 6.62348401323043,
|
|
"grad_norm": 0.595843625275714,
|
|
"learning_rate": 3.525262873657065e-07,
|
|
"loss": 0.1787,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09580498933792114,
|
|
"step": 12015,
|
|
"valid_targets_mean": 4447.8,
|
|
"valid_targets_min": 2600
|
|
},
|
|
{
|
|
"epoch": 6.626240352811466,
|
|
"grad_norm": 0.6588911816334713,
|
|
"learning_rate": 3.474061501422155e-07,
|
|
"loss": 0.2004,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08399167656898499,
|
|
"step": 12020,
|
|
"valid_targets_mean": 3453.4,
|
|
"valid_targets_min": 1241
|
|
},
|
|
{
|
|
"epoch": 6.628996692392503,
|
|
"grad_norm": 0.7216651705193683,
|
|
"learning_rate": 3.423231424187501e-07,
|
|
"loss": 0.1947,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.12225198745727539,
|
|
"step": 12025,
|
|
"valid_targets_mean": 3746.0,
|
|
"valid_targets_min": 990
|
|
},
|
|
{
|
|
"epoch": 6.631753031973539,
|
|
"grad_norm": 0.695339874477122,
|
|
"learning_rate": 3.3727727379859523e-07,
|
|
"loss": 0.1856,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09461720287799835,
|
|
"step": 12030,
|
|
"valid_targets_mean": 3809.4,
|
|
"valid_targets_min": 1131
|
|
},
|
|
{
|
|
"epoch": 6.634509371554575,
|
|
"grad_norm": 0.6612199346286155,
|
|
"learning_rate": 3.32268553814874e-07,
|
|
"loss": 0.1862,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10540962219238281,
|
|
"step": 12035,
|
|
"valid_targets_mean": 4040.2,
|
|
"valid_targets_min": 2248
|
|
},
|
|
{
|
|
"epoch": 6.637265711135612,
|
|
"grad_norm": 0.6335950438579119,
|
|
"learning_rate": 3.2729699193052357e-07,
|
|
"loss": 0.1852,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07712060213088989,
|
|
"step": 12040,
|
|
"valid_targets_mean": 3809.2,
|
|
"valid_targets_min": 1909
|
|
},
|
|
{
|
|
"epoch": 6.640022050716649,
|
|
"grad_norm": 0.641221874707116,
|
|
"learning_rate": 3.223625975382838e-07,
|
|
"loss": 0.1886,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0928414836525917,
|
|
"step": 12045,
|
|
"valid_targets_mean": 3358.8,
|
|
"valid_targets_min": 2760
|
|
},
|
|
{
|
|
"epoch": 6.642778390297685,
|
|
"grad_norm": 0.765916878281289,
|
|
"learning_rate": 3.174653799606642e-07,
|
|
"loss": 0.1914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09122616052627563,
|
|
"step": 12050,
|
|
"valid_targets_mean": 2591.8,
|
|
"valid_targets_min": 616
|
|
},
|
|
{
|
|
"epoch": 6.645534729878721,
|
|
"grad_norm": 0.5825621448408331,
|
|
"learning_rate": 3.126053484499436e-07,
|
|
"loss": 0.1748,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08199020475149155,
|
|
"step": 12055,
|
|
"valid_targets_mean": 3763.9,
|
|
"valid_targets_min": 3138
|
|
},
|
|
{
|
|
"epoch": 6.6482910694597575,
|
|
"grad_norm": 0.6313029810205866,
|
|
"learning_rate": 3.0778251218814394e-07,
|
|
"loss": 0.1802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09305316209793091,
|
|
"step": 12060,
|
|
"valid_targets_mean": 3772.1,
|
|
"valid_targets_min": 1843
|
|
},
|
|
{
|
|
"epoch": 6.651047409040794,
|
|
"grad_norm": 0.6495587088172492,
|
|
"learning_rate": 3.0299688028701204e-07,
|
|
"loss": 0.1831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09014217555522919,
|
|
"step": 12065,
|
|
"valid_targets_mean": 3888.1,
|
|
"valid_targets_min": 2322
|
|
},
|
|
{
|
|
"epoch": 6.65380374862183,
|
|
"grad_norm": 0.7300503514237874,
|
|
"learning_rate": 2.9824846178801105e-07,
|
|
"loss": 0.1848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09146483242511749,
|
|
"step": 12070,
|
|
"valid_targets_mean": 2851.5,
|
|
"valid_targets_min": 1064
|
|
},
|
|
{
|
|
"epoch": 6.656560088202866,
|
|
"grad_norm": 0.5817267617146338,
|
|
"learning_rate": 2.93537265662287e-07,
|
|
"loss": 0.1804,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08441506326198578,
|
|
"step": 12075,
|
|
"valid_targets_mean": 3490.5,
|
|
"valid_targets_min": 1710
|
|
},
|
|
{
|
|
"epoch": 6.659316427783903,
|
|
"grad_norm": 0.6613496252635918,
|
|
"learning_rate": 2.888633008106734e-07,
|
|
"loss": 0.1885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09310775250196457,
|
|
"step": 12080,
|
|
"valid_targets_mean": 3912.6,
|
|
"valid_targets_min": 981
|
|
},
|
|
{
|
|
"epoch": 6.66207276736494,
|
|
"grad_norm": 0.6166171657507659,
|
|
"learning_rate": 2.842265760636531e-07,
|
|
"loss": 0.1838,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10195156186819077,
|
|
"step": 12085,
|
|
"valid_targets_mean": 4673.2,
|
|
"valid_targets_min": 3322
|
|
},
|
|
{
|
|
"epoch": 6.664829106945976,
|
|
"grad_norm": 0.7068932755892615,
|
|
"learning_rate": 2.796271001813566e-07,
|
|
"loss": 0.1823,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09147079288959503,
|
|
"step": 12090,
|
|
"valid_targets_mean": 3132.5,
|
|
"valid_targets_min": 1350
|
|
},
|
|
{
|
|
"epoch": 6.667585446527012,
|
|
"grad_norm": 0.6163809973483977,
|
|
"learning_rate": 2.75064881853544e-07,
|
|
"loss": 0.1871,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08081193268299103,
|
|
"step": 12095,
|
|
"valid_targets_mean": 3979.8,
|
|
"valid_targets_min": 2268
|
|
},
|
|
{
|
|
"epoch": 6.6703417861080485,
|
|
"grad_norm": 0.6666869771394652,
|
|
"learning_rate": 2.705399296995759e-07,
|
|
"loss": 0.1825,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08253160119056702,
|
|
"step": 12100,
|
|
"valid_targets_mean": 3196.8,
|
|
"valid_targets_min": 2098
|
|
},
|
|
{
|
|
"epoch": 6.673098125689085,
|
|
"grad_norm": 0.6378104281543062,
|
|
"learning_rate": 2.660522522684117e-07,
|
|
"loss": 0.1708,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07568346709012985,
|
|
"step": 12105,
|
|
"valid_targets_mean": 3781.6,
|
|
"valid_targets_min": 1686
|
|
},
|
|
{
|
|
"epoch": 6.675854465270121,
|
|
"grad_norm": 0.6596524414343626,
|
|
"learning_rate": 2.6160185803858926e-07,
|
|
"loss": 0.1789,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09766026586294174,
|
|
"step": 12110,
|
|
"valid_targets_mean": 3496.2,
|
|
"valid_targets_min": 2243
|
|
},
|
|
{
|
|
"epoch": 6.678610804851157,
|
|
"grad_norm": 0.6695916606017187,
|
|
"learning_rate": 2.5718875541820285e-07,
|
|
"loss": 0.1757,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08377499878406525,
|
|
"step": 12115,
|
|
"valid_targets_mean": 3654.4,
|
|
"valid_targets_min": 903
|
|
},
|
|
{
|
|
"epoch": 6.681367144432194,
|
|
"grad_norm": 0.7855575654365269,
|
|
"learning_rate": 2.5281295274489413e-07,
|
|
"loss": 0.1734,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10766284912824631,
|
|
"step": 12120,
|
|
"valid_targets_mean": 4804.6,
|
|
"valid_targets_min": 1866
|
|
},
|
|
{
|
|
"epoch": 6.684123484013231,
|
|
"grad_norm": 0.6271708443516495,
|
|
"learning_rate": 2.484744582858389e-07,
|
|
"loss": 0.1693,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09076222777366638,
|
|
"step": 12125,
|
|
"valid_targets_mean": 3809.4,
|
|
"valid_targets_min": 872
|
|
},
|
|
{
|
|
"epoch": 6.686879823594267,
|
|
"grad_norm": 0.6352773890096619,
|
|
"learning_rate": 2.4417328023771835e-07,
|
|
"loss": 0.1743,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08445511758327484,
|
|
"step": 12130,
|
|
"valid_targets_mean": 3245.6,
|
|
"valid_targets_min": 1694
|
|
},
|
|
{
|
|
"epoch": 6.689636163175303,
|
|
"grad_norm": 0.6639767377893608,
|
|
"learning_rate": 2.399094267267188e-07,
|
|
"loss": 0.1745,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0702781081199646,
|
|
"step": 12135,
|
|
"valid_targets_mean": 2389.2,
|
|
"valid_targets_min": 897
|
|
},
|
|
{
|
|
"epoch": 6.69239250275634,
|
|
"grad_norm": 0.6677703453634342,
|
|
"learning_rate": 2.3568290580850307e-07,
|
|
"loss": 0.1897,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08483712375164032,
|
|
"step": 12140,
|
|
"valid_targets_mean": 3098.4,
|
|
"valid_targets_min": 1695
|
|
},
|
|
{
|
|
"epoch": 6.695148842337376,
|
|
"grad_norm": 0.6602896344700285,
|
|
"learning_rate": 2.3149372546821258e-07,
|
|
"loss": 0.1737,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07802101969718933,
|
|
"step": 12145,
|
|
"valid_targets_mean": 3205.5,
|
|
"valid_targets_min": 1710
|
|
},
|
|
{
|
|
"epoch": 6.697905181918412,
|
|
"grad_norm": 0.609306347009806,
|
|
"learning_rate": 2.273418936204297e-07,
|
|
"loss": 0.1737,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07542389631271362,
|
|
"step": 12150,
|
|
"valid_targets_mean": 3492.4,
|
|
"valid_targets_min": 992
|
|
},
|
|
{
|
|
"epoch": 6.700661521499448,
|
|
"grad_norm": 0.6033456863536918,
|
|
"learning_rate": 2.2322741810918202e-07,
|
|
"loss": 0.1709,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08896403014659882,
|
|
"step": 12155,
|
|
"valid_targets_mean": 4175.6,
|
|
"valid_targets_min": 1673
|
|
},
|
|
{
|
|
"epoch": 6.703417861080485,
|
|
"grad_norm": 0.5646107531505503,
|
|
"learning_rate": 2.1915030670791594e-07,
|
|
"loss": 0.1823,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07576298713684082,
|
|
"step": 12160,
|
|
"valid_targets_mean": 3773.9,
|
|
"valid_targets_min": 2251
|
|
},
|
|
{
|
|
"epoch": 6.706174200661522,
|
|
"grad_norm": 0.6391919384804599,
|
|
"learning_rate": 2.1511056711948751e-07,
|
|
"loss": 0.1841,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.09295916557312012,
|
|
"step": 12165,
|
|
"valid_targets_mean": 4170.9,
|
|
"valid_targets_min": 2888
|
|
},
|
|
{
|
|
"epoch": 6.708930540242558,
|
|
"grad_norm": 0.6215347597916574,
|
|
"learning_rate": 2.1110820697614943e-07,
|
|
"loss": 0.1836,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0980633944272995,
|
|
"step": 12170,
|
|
"valid_targets_mean": 4478.4,
|
|
"valid_targets_min": 2960
|
|
},
|
|
{
|
|
"epoch": 6.711686879823594,
|
|
"grad_norm": 0.6038768397251708,
|
|
"learning_rate": 2.071432338395285e-07,
|
|
"loss": 0.1782,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07454174757003784,
|
|
"step": 12175,
|
|
"valid_targets_mean": 3730.8,
|
|
"valid_targets_min": 1815
|
|
},
|
|
{
|
|
"epoch": 6.714443219404631,
|
|
"grad_norm": 0.6114021264061849,
|
|
"learning_rate": 2.0321565520061925e-07,
|
|
"loss": 0.1814,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10640440881252289,
|
|
"step": 12180,
|
|
"valid_targets_mean": 4718.6,
|
|
"valid_targets_min": 1725
|
|
},
|
|
{
|
|
"epoch": 6.7171995589856675,
|
|
"grad_norm": 0.5702306543094484,
|
|
"learning_rate": 1.9932547847976823e-07,
|
|
"loss": 0.1748,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07278351485729218,
|
|
"step": 12185,
|
|
"valid_targets_mean": 3794.4,
|
|
"valid_targets_min": 1951
|
|
},
|
|
{
|
|
"epoch": 6.719955898566703,
|
|
"grad_norm": 0.5262401383057111,
|
|
"learning_rate": 1.9547271102665854e-07,
|
|
"loss": 0.1435,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06443116068840027,
|
|
"step": 12190,
|
|
"valid_targets_mean": 2261.1,
|
|
"valid_targets_min": 331
|
|
},
|
|
{
|
|
"epoch": 6.722712238147739,
|
|
"grad_norm": 0.39543555825528565,
|
|
"learning_rate": 1.91657360120292e-07,
|
|
"loss": 0.0832,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.036877550184726715,
|
|
"step": 12195,
|
|
"valid_targets_mean": 3604.4,
|
|
"valid_targets_min": 708
|
|
},
|
|
{
|
|
"epoch": 6.725468577728776,
|
|
"grad_norm": 0.4533679609766993,
|
|
"learning_rate": 1.878794329689848e-07,
|
|
"loss": 0.0743,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04005257785320282,
|
|
"step": 12200,
|
|
"valid_targets_mean": 2933.4,
|
|
"valid_targets_min": 729
|
|
},
|
|
{
|
|
"epoch": 6.728224917309813,
|
|
"grad_norm": 0.4694968455970161,
|
|
"learning_rate": 1.8413893671034744e-07,
|
|
"loss": 0.0794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07012099772691727,
|
|
"step": 12205,
|
|
"valid_targets_mean": 2718.9,
|
|
"valid_targets_min": 921
|
|
},
|
|
{
|
|
"epoch": 6.730981256890849,
|
|
"grad_norm": 0.5315576434983758,
|
|
"learning_rate": 1.804358784112692e-07,
|
|
"loss": 0.073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030594170093536377,
|
|
"step": 12210,
|
|
"valid_targets_mean": 2827.9,
|
|
"valid_targets_min": 813
|
|
},
|
|
{
|
|
"epoch": 6.733737596471885,
|
|
"grad_norm": 0.5205679683902596,
|
|
"learning_rate": 1.7677026506791372e-07,
|
|
"loss": 0.0781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04112879931926727,
|
|
"step": 12215,
|
|
"valid_targets_mean": 2900.5,
|
|
"valid_targets_min": 854
|
|
},
|
|
{
|
|
"epoch": 6.736493936052922,
|
|
"grad_norm": 0.8280268595739756,
|
|
"learning_rate": 1.731421036056946e-07,
|
|
"loss": 0.1175,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04723259061574936,
|
|
"step": 12220,
|
|
"valid_targets_mean": 1034.4,
|
|
"valid_targets_min": 536
|
|
},
|
|
{
|
|
"epoch": 6.7392502756339585,
|
|
"grad_norm": 0.4349223224567377,
|
|
"learning_rate": 1.6955140087927314e-07,
|
|
"loss": 0.0875,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03537938743829727,
|
|
"step": 12225,
|
|
"valid_targets_mean": 2992.8,
|
|
"valid_targets_min": 915
|
|
},
|
|
{
|
|
"epoch": 6.7420066152149944,
|
|
"grad_norm": 0.4679762315987108,
|
|
"learning_rate": 1.6599816367253385e-07,
|
|
"loss": 0.0831,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042808398604393005,
|
|
"step": 12230,
|
|
"valid_targets_mean": 3356.0,
|
|
"valid_targets_min": 2281
|
|
},
|
|
{
|
|
"epoch": 6.74476295479603,
|
|
"grad_norm": 0.3679296511990159,
|
|
"learning_rate": 1.6248239869858463e-07,
|
|
"loss": 0.1104,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02819836139678955,
|
|
"step": 12235,
|
|
"valid_targets_mean": 3776.1,
|
|
"valid_targets_min": 958
|
|
},
|
|
{
|
|
"epoch": 6.747519294377067,
|
|
"grad_norm": 0.38532272463757417,
|
|
"learning_rate": 1.5900411259973214e-07,
|
|
"loss": 0.0635,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.042835891246795654,
|
|
"step": 12240,
|
|
"valid_targets_mean": 4237.5,
|
|
"valid_targets_min": 2082
|
|
},
|
|
{
|
|
"epoch": 6.750275633958104,
|
|
"grad_norm": 0.42378237385596157,
|
|
"learning_rate": 1.5556331194747755e-07,
|
|
"loss": 0.0729,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04055206850171089,
|
|
"step": 12245,
|
|
"valid_targets_mean": 4378.9,
|
|
"valid_targets_min": 3337
|
|
},
|
|
{
|
|
"epoch": 6.75303197353914,
|
|
"grad_norm": 0.3897659481224976,
|
|
"learning_rate": 1.5216000324250079e-07,
|
|
"loss": 0.0726,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03475230559706688,
|
|
"step": 12250,
|
|
"valid_targets_mean": 3934.4,
|
|
"valid_targets_min": 515
|
|
},
|
|
{
|
|
"epoch": 6.755788313120176,
|
|
"grad_norm": 0.6302737096058082,
|
|
"learning_rate": 1.4879419291464747e-07,
|
|
"loss": 0.1141,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0403238907456398,
|
|
"step": 12255,
|
|
"valid_targets_mean": 1878.6,
|
|
"valid_targets_min": 546
|
|
},
|
|
{
|
|
"epoch": 6.758544652701213,
|
|
"grad_norm": 0.3934556542603552,
|
|
"learning_rate": 1.4546588732291532e-07,
|
|
"loss": 0.0674,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03330637887120247,
|
|
"step": 12260,
|
|
"valid_targets_mean": 3432.2,
|
|
"valid_targets_min": 2715
|
|
},
|
|
{
|
|
"epoch": 6.76130099228225,
|
|
"grad_norm": 0.5102468752393273,
|
|
"learning_rate": 1.4217509275545217e-07,
|
|
"loss": 0.0759,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03271430358290672,
|
|
"step": 12265,
|
|
"valid_targets_mean": 1612.8,
|
|
"valid_targets_min": 515
|
|
},
|
|
{
|
|
"epoch": 6.7640573318632855,
|
|
"grad_norm": 0.38101852309250933,
|
|
"learning_rate": 1.3892181542952688e-07,
|
|
"loss": 0.0646,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.022396264597773552,
|
|
"step": 12270,
|
|
"valid_targets_mean": 3322.8,
|
|
"valid_targets_min": 840
|
|
},
|
|
{
|
|
"epoch": 6.766813671444321,
|
|
"grad_norm": 0.48275004141409406,
|
|
"learning_rate": 1.357060614915362e-07,
|
|
"loss": 0.0769,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03673069179058075,
|
|
"step": 12275,
|
|
"valid_targets_mean": 2764.0,
|
|
"valid_targets_min": 574
|
|
},
|
|
{
|
|
"epoch": 6.769570011025358,
|
|
"grad_norm": 0.4865345996164738,
|
|
"learning_rate": 1.3252783701698026e-07,
|
|
"loss": 0.0781,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0403091199696064,
|
|
"step": 12280,
|
|
"valid_targets_mean": 2077.9,
|
|
"valid_targets_min": 847
|
|
},
|
|
{
|
|
"epoch": 6.772326350606395,
|
|
"grad_norm": 0.47905459850440935,
|
|
"learning_rate": 1.2938714801044916e-07,
|
|
"loss": 0.084,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04171409457921982,
|
|
"step": 12285,
|
|
"valid_targets_mean": 2857.1,
|
|
"valid_targets_min": 770
|
|
},
|
|
{
|
|
"epoch": 6.775082690187431,
|
|
"grad_norm": 0.4244935604071747,
|
|
"learning_rate": 1.2628400040562983e-07,
|
|
"loss": 0.0809,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033787067979574203,
|
|
"step": 12290,
|
|
"valid_targets_mean": 1736.5,
|
|
"valid_targets_min": 611
|
|
},
|
|
{
|
|
"epoch": 6.777839029768468,
|
|
"grad_norm": 0.5648133197314156,
|
|
"learning_rate": 1.2321840006527475e-07,
|
|
"loss": 0.1188,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.08631548285484314,
|
|
"step": 12295,
|
|
"valid_targets_mean": 3779.8,
|
|
"valid_targets_min": 2812
|
|
},
|
|
{
|
|
"epoch": 6.780595369349504,
|
|
"grad_norm": 0.33126659553565285,
|
|
"learning_rate": 1.2019035278119983e-07,
|
|
"loss": 0.0693,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02871672809123993,
|
|
"step": 12300,
|
|
"valid_targets_mean": 4061.4,
|
|
"valid_targets_min": 3722
|
|
},
|
|
{
|
|
"epoch": 6.783351708930541,
|
|
"grad_norm": 0.4618865727380007,
|
|
"learning_rate": 1.1719986427427111e-07,
|
|
"loss": 0.0772,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035096172243356705,
|
|
"step": 12305,
|
|
"valid_targets_mean": 1199.0,
|
|
"valid_targets_min": 711
|
|
},
|
|
{
|
|
"epoch": 6.786108048511577,
|
|
"grad_norm": 0.41378111300222226,
|
|
"learning_rate": 1.1424694019439575e-07,
|
|
"loss": 0.1032,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.033986933529376984,
|
|
"step": 12310,
|
|
"valid_targets_mean": 3407.8,
|
|
"valid_targets_min": 1803
|
|
},
|
|
{
|
|
"epoch": 6.788864388092613,
|
|
"grad_norm": 0.460190395123742,
|
|
"learning_rate": 1.1133158612051331e-07,
|
|
"loss": 0.0897,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03641940653324127,
|
|
"step": 12315,
|
|
"valid_targets_mean": 3789.4,
|
|
"valid_targets_min": 3167
|
|
},
|
|
{
|
|
"epoch": 6.791620727673649,
|
|
"grad_norm": 0.7333666005378974,
|
|
"learning_rate": 1.084538075605801e-07,
|
|
"loss": 0.1015,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.10202132910490036,
|
|
"step": 12320,
|
|
"valid_targets_mean": 1682.0,
|
|
"valid_targets_min": 915
|
|
},
|
|
{
|
|
"epoch": 6.794377067254686,
|
|
"grad_norm": 0.4660584687367797,
|
|
"learning_rate": 1.0561360995156034e-07,
|
|
"loss": 0.243,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04186289384961128,
|
|
"step": 12325,
|
|
"valid_targets_mean": 3663.0,
|
|
"valid_targets_min": 2876
|
|
},
|
|
{
|
|
"epoch": 6.797133406835722,
|
|
"grad_norm": 0.42696874694936704,
|
|
"learning_rate": 1.0281099865941946e-07,
|
|
"loss": 0.0843,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.036597203463315964,
|
|
"step": 12330,
|
|
"valid_targets_mean": 3140.6,
|
|
"valid_targets_min": 828
|
|
},
|
|
{
|
|
"epoch": 6.799889746416759,
|
|
"grad_norm": 0.511197393512323,
|
|
"learning_rate": 1.0004597897911084e-07,
|
|
"loss": 0.0827,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06011876091361046,
|
|
"step": 12335,
|
|
"valid_targets_mean": 3329.5,
|
|
"valid_targets_min": 1794
|
|
},
|
|
{
|
|
"epoch": 6.802646085997795,
|
|
"grad_norm": 0.4704393932639716,
|
|
"learning_rate": 9.731855613456687e-08,
|
|
"loss": 0.1702,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04856371507048607,
|
|
"step": 12340,
|
|
"valid_targets_mean": 3535.5,
|
|
"valid_targets_min": 2771
|
|
},
|
|
{
|
|
"epoch": 6.805402425578832,
|
|
"grad_norm": 0.2949390739081591,
|
|
"learning_rate": 9.462873527868343e-08,
|
|
"loss": 0.0628,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02964148297905922,
|
|
"step": 12345,
|
|
"valid_targets_mean": 4594.6,
|
|
"valid_targets_min": 2655
|
|
},
|
|
{
|
|
"epoch": 6.808158765159868,
|
|
"grad_norm": 0.38824674581202684,
|
|
"learning_rate": 9.197652149332215e-08,
|
|
"loss": 0.0844,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.020141052082180977,
|
|
"step": 12350,
|
|
"valid_targets_mean": 2743.9,
|
|
"valid_targets_min": 557
|
|
},
|
|
{
|
|
"epoch": 6.8109151047409044,
|
|
"grad_norm": 0.4607638649146544,
|
|
"learning_rate": 8.936191978929254e-08,
|
|
"loss": 0.0822,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039801470935344696,
|
|
"step": 12355,
|
|
"valid_targets_mean": 4249.6,
|
|
"valid_targets_min": 1606
|
|
},
|
|
{
|
|
"epoch": 6.81367144432194,
|
|
"grad_norm": 0.42503459708289154,
|
|
"learning_rate": 8.678493510633656e-08,
|
|
"loss": 0.0694,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03275343030691147,
|
|
"step": 12360,
|
|
"valid_targets_mean": 3570.9,
|
|
"valid_targets_min": 2163
|
|
},
|
|
{
|
|
"epoch": 6.816427783902977,
|
|
"grad_norm": 0.44543136746163786,
|
|
"learning_rate": 8.424557231313746e-08,
|
|
"loss": 0.069,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.034869201481342316,
|
|
"step": 12365,
|
|
"valid_targets_mean": 3057.5,
|
|
"valid_targets_min": 533
|
|
},
|
|
{
|
|
"epoch": 6.819184123484013,
|
|
"grad_norm": 0.4180718806441837,
|
|
"learning_rate": 8.174383620729087e-08,
|
|
"loss": 0.0709,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03462996706366539,
|
|
"step": 12370,
|
|
"valid_targets_mean": 3622.0,
|
|
"valid_targets_min": 2249
|
|
},
|
|
{
|
|
"epoch": 6.82194046306505,
|
|
"grad_norm": 0.396136362244719,
|
|
"learning_rate": 7.927973151531155e-08,
|
|
"loss": 0.0671,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.028120435774326324,
|
|
"step": 12375,
|
|
"valid_targets_mean": 2801.5,
|
|
"valid_targets_min": 817
|
|
},
|
|
{
|
|
"epoch": 6.824696802646086,
|
|
"grad_norm": 0.8093765091038903,
|
|
"learning_rate": 7.685326289261108e-08,
|
|
"loss": 0.1089,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.107778400182724,
|
|
"step": 12380,
|
|
"valid_targets_mean": 2015.1,
|
|
"valid_targets_min": 859
|
|
},
|
|
{
|
|
"epoch": 6.827453142227123,
|
|
"grad_norm": 0.369838844252055,
|
|
"learning_rate": 7.446443492350242e-08,
|
|
"loss": 0.0675,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03551428020000458,
|
|
"step": 12385,
|
|
"valid_targets_mean": 3601.5,
|
|
"valid_targets_min": 2709
|
|
},
|
|
{
|
|
"epoch": 6.830209481808159,
|
|
"grad_norm": 0.6450879560147654,
|
|
"learning_rate": 7.211325212117759e-08,
|
|
"loss": 0.0644,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04826982691884041,
|
|
"step": 12390,
|
|
"valid_targets_mean": 1605.5,
|
|
"valid_targets_min": 781
|
|
},
|
|
{
|
|
"epoch": 6.8329658213891955,
|
|
"grad_norm": 0.34377627230542573,
|
|
"learning_rate": 6.979971892770998e-08,
|
|
"loss": 0.0634,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02662857249379158,
|
|
"step": 12395,
|
|
"valid_targets_mean": 3703.2,
|
|
"valid_targets_min": 2343
|
|
},
|
|
{
|
|
"epoch": 6.835722160970231,
|
|
"grad_norm": 0.386532054790279,
|
|
"learning_rate": 6.752383971404097e-08,
|
|
"loss": 0.0643,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026853736490011215,
|
|
"step": 12400,
|
|
"valid_targets_mean": 1990.2,
|
|
"valid_targets_min": 545
|
|
},
|
|
{
|
|
"epoch": 6.838478500551268,
|
|
"grad_norm": 0.407646005165065,
|
|
"learning_rate": 6.528561877997108e-08,
|
|
"loss": 0.1072,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03396901860833168,
|
|
"step": 12405,
|
|
"valid_targets_mean": 3322.1,
|
|
"valid_targets_min": 904
|
|
},
|
|
{
|
|
"epoch": 6.841234840132304,
|
|
"grad_norm": 0.3968501139297692,
|
|
"learning_rate": 6.308506035415552e-08,
|
|
"loss": 0.0937,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06823492050170898,
|
|
"step": 12410,
|
|
"valid_targets_mean": 3514.5,
|
|
"valid_targets_min": 2184
|
|
},
|
|
{
|
|
"epoch": 6.843991179713341,
|
|
"grad_norm": 0.34653346597932905,
|
|
"learning_rate": 6.092216859409306e-08,
|
|
"loss": 0.0779,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.022607002407312393,
|
|
"step": 12415,
|
|
"valid_targets_mean": 1680.6,
|
|
"valid_targets_min": 595
|
|
},
|
|
{
|
|
"epoch": 6.846747519294377,
|
|
"grad_norm": 1.1497671161907015,
|
|
"learning_rate": 5.879694758611942e-08,
|
|
"loss": 0.0902,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0782558023929596,
|
|
"step": 12420,
|
|
"valid_targets_mean": 1323.4,
|
|
"valid_targets_min": 575
|
|
},
|
|
{
|
|
"epoch": 6.849503858875414,
|
|
"grad_norm": 0.6112375724719856,
|
|
"learning_rate": 5.6709401345396154e-08,
|
|
"loss": 0.1134,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04306153953075409,
|
|
"step": 12425,
|
|
"valid_targets_mean": 1671.9,
|
|
"valid_targets_min": 691
|
|
},
|
|
{
|
|
"epoch": 6.85226019845645,
|
|
"grad_norm": 0.4988687136819589,
|
|
"learning_rate": 5.4659533815912826e-08,
|
|
"loss": 0.0794,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.045069798827171326,
|
|
"step": 12430,
|
|
"valid_targets_mean": 1879.5,
|
|
"valid_targets_min": 687
|
|
},
|
|
{
|
|
"epoch": 6.855016538037487,
|
|
"grad_norm": 0.3882565790042688,
|
|
"learning_rate": 5.264734887046707e-08,
|
|
"loss": 0.0914,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030401194468140602,
|
|
"step": 12435,
|
|
"valid_targets_mean": 3679.5,
|
|
"valid_targets_min": 2015
|
|
},
|
|
{
|
|
"epoch": 6.8577728776185225,
|
|
"grad_norm": 0.6010030768784288,
|
|
"learning_rate": 5.06728503106646e-08,
|
|
"loss": 0.0742,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03887934610247612,
|
|
"step": 12440,
|
|
"valid_targets_mean": 1351.2,
|
|
"valid_targets_min": 520
|
|
},
|
|
{
|
|
"epoch": 6.860529217199559,
|
|
"grad_norm": 0.33127583612110834,
|
|
"learning_rate": 4.8736041866912496e-08,
|
|
"loss": 0.0729,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.027286695316433907,
|
|
"step": 12445,
|
|
"valid_targets_mean": 4111.8,
|
|
"valid_targets_min": 466
|
|
},
|
|
{
|
|
"epoch": 6.863285556780595,
|
|
"grad_norm": 0.30931058080356744,
|
|
"learning_rate": 4.683692719840594e-08,
|
|
"loss": 0.0784,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02128816768527031,
|
|
"step": 12450,
|
|
"valid_targets_mean": 5544.5,
|
|
"valid_targets_min": 4219
|
|
},
|
|
{
|
|
"epoch": 6.866041896361632,
|
|
"grad_norm": 0.3581233686224879,
|
|
"learning_rate": 4.497550989313038e-08,
|
|
"loss": 0.0682,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.024580489844083786,
|
|
"step": 12455,
|
|
"valid_targets_mean": 3800.1,
|
|
"valid_targets_min": 456
|
|
},
|
|
{
|
|
"epoch": 6.868798235942668,
|
|
"grad_norm": 0.38389781780672394,
|
|
"learning_rate": 4.315179346784826e-08,
|
|
"loss": 0.0665,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030943233519792557,
|
|
"step": 12460,
|
|
"valid_targets_mean": 3487.1,
|
|
"valid_targets_min": 976
|
|
},
|
|
{
|
|
"epoch": 6.871554575523705,
|
|
"grad_norm": 0.4403145144513486,
|
|
"learning_rate": 4.136578136809233e-08,
|
|
"loss": 0.0753,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.046636179089546204,
|
|
"step": 12465,
|
|
"valid_targets_mean": 2431.6,
|
|
"valid_targets_min": 591
|
|
},
|
|
{
|
|
"epoch": 6.874310915104741,
|
|
"grad_norm": 0.4181932833212492,
|
|
"learning_rate": 3.96174769681612e-08,
|
|
"loss": 0.0817,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0463918074965477,
|
|
"step": 12470,
|
|
"valid_targets_mean": 3660.8,
|
|
"valid_targets_min": 2753
|
|
},
|
|
{
|
|
"epoch": 6.877067254685778,
|
|
"grad_norm": 0.40243478423363044,
|
|
"learning_rate": 3.790688357111494e-08,
|
|
"loss": 0.0848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04377498850226402,
|
|
"step": 12475,
|
|
"valid_targets_mean": 3690.9,
|
|
"valid_targets_min": 2892
|
|
},
|
|
{
|
|
"epoch": 6.879823594266814,
|
|
"grad_norm": 0.4003064076016153,
|
|
"learning_rate": 3.62340044087639e-08,
|
|
"loss": 0.073,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03098774328827858,
|
|
"step": 12480,
|
|
"valid_targets_mean": 3447.0,
|
|
"valid_targets_min": 1093
|
|
},
|
|
{
|
|
"epoch": 6.88257993384785,
|
|
"grad_norm": 0.4358853986734758,
|
|
"learning_rate": 3.45988426416688e-08,
|
|
"loss": 0.071,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04536157473921776,
|
|
"step": 12485,
|
|
"valid_targets_mean": 3798.8,
|
|
"valid_targets_min": 1861
|
|
},
|
|
{
|
|
"epoch": 6.885336273428886,
|
|
"grad_norm": 0.3910370394563129,
|
|
"learning_rate": 3.3001401359125106e-08,
|
|
"loss": 0.0788,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02842230536043644,
|
|
"step": 12490,
|
|
"valid_targets_mean": 3005.0,
|
|
"valid_targets_min": 720
|
|
},
|
|
{
|
|
"epoch": 6.888092613009923,
|
|
"grad_norm": 0.3325990569830016,
|
|
"learning_rate": 3.1441683579167545e-08,
|
|
"loss": 0.0594,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026847798377275467,
|
|
"step": 12495,
|
|
"valid_targets_mean": 3797.5,
|
|
"valid_targets_min": 2761
|
|
},
|
|
{
|
|
"epoch": 6.890848952590959,
|
|
"grad_norm": 0.49238440629594865,
|
|
"learning_rate": 2.991969224856117e-08,
|
|
"loss": 0.0791,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029867786914110184,
|
|
"step": 12500,
|
|
"valid_targets_mean": 2420.0,
|
|
"valid_targets_min": 588
|
|
},
|
|
{
|
|
"epoch": 6.893605292171996,
|
|
"grad_norm": 0.4461244518156673,
|
|
"learning_rate": 2.8435430242792496e-08,
|
|
"loss": 0.0741,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03694469481706619,
|
|
"step": 12505,
|
|
"valid_targets_mean": 3433.5,
|
|
"valid_targets_min": 596
|
|
},
|
|
{
|
|
"epoch": 6.896361631753032,
|
|
"grad_norm": 0.44218541196510724,
|
|
"learning_rate": 2.6988900366067274e-08,
|
|
"loss": 0.0848,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04154649004340172,
|
|
"step": 12510,
|
|
"valid_targets_mean": 3530.0,
|
|
"valid_targets_min": 1636
|
|
},
|
|
{
|
|
"epoch": 6.899117971334069,
|
|
"grad_norm": 0.46344153748813605,
|
|
"learning_rate": 2.558010535130162e-08,
|
|
"loss": 0.0765,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03794536739587784,
|
|
"step": 12515,
|
|
"valid_targets_mean": 1950.8,
|
|
"valid_targets_min": 454
|
|
},
|
|
{
|
|
"epoch": 6.901874310915105,
|
|
"grad_norm": 0.7690549424645567,
|
|
"learning_rate": 2.4209047860122015e-08,
|
|
"loss": 0.0826,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0450926348567009,
|
|
"step": 12520,
|
|
"valid_targets_mean": 819.5,
|
|
"valid_targets_min": 609
|
|
},
|
|
{
|
|
"epoch": 6.904630650496141,
|
|
"grad_norm": 0.7507289498884353,
|
|
"learning_rate": 2.2875730482856405e-08,
|
|
"loss": 0.0768,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.06610219180583954,
|
|
"step": 12525,
|
|
"valid_targets_mean": 2364.8,
|
|
"valid_targets_min": 893
|
|
},
|
|
{
|
|
"epoch": 6.907386990077177,
|
|
"grad_norm": 0.4155390484079837,
|
|
"learning_rate": 2.1580155738534226e-08,
|
|
"loss": 0.0829,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.030742565169930458,
|
|
"step": 12530,
|
|
"valid_targets_mean": 3029.5,
|
|
"valid_targets_min": 1140
|
|
},
|
|
{
|
|
"epoch": 6.910143329658214,
|
|
"grad_norm": 0.455966338770568,
|
|
"learning_rate": 2.0322326074875277e-08,
|
|
"loss": 0.0761,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04529863968491554,
|
|
"step": 12535,
|
|
"valid_targets_mean": 3029.6,
|
|
"valid_targets_min": 1502
|
|
},
|
|
{
|
|
"epoch": 6.91289966923925,
|
|
"grad_norm": 0.5878871867331752,
|
|
"learning_rate": 1.9102243868285298e-08,
|
|
"loss": 0.0672,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.041006460785865784,
|
|
"step": 12540,
|
|
"valid_targets_mean": 1295.8,
|
|
"valid_targets_min": 624
|
|
},
|
|
{
|
|
"epoch": 6.915656008820287,
|
|
"grad_norm": 0.5976068964105508,
|
|
"learning_rate": 1.791991142385596e-08,
|
|
"loss": 0.0851,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.048554107546806335,
|
|
"step": 12545,
|
|
"valid_targets_mean": 1481.6,
|
|
"valid_targets_min": 687
|
|
},
|
|
{
|
|
"epoch": 6.918412348401323,
|
|
"grad_norm": 0.3619404922287913,
|
|
"learning_rate": 1.677533097536266e-08,
|
|
"loss": 0.0732,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02396014705300331,
|
|
"step": 12550,
|
|
"valid_targets_mean": 3963.5,
|
|
"valid_targets_min": 3651
|
|
},
|
|
{
|
|
"epoch": 6.92116868798236,
|
|
"grad_norm": 0.3667170722909769,
|
|
"learning_rate": 1.5668504685251163e-08,
|
|
"loss": 0.0654,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.043025754392147064,
|
|
"step": 12555,
|
|
"valid_targets_mean": 5041.0,
|
|
"valid_targets_min": 3813
|
|
},
|
|
{
|
|
"epoch": 6.923925027563396,
|
|
"grad_norm": 0.3607427507710741,
|
|
"learning_rate": 1.4599434644639866e-08,
|
|
"loss": 0.0726,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03428248316049576,
|
|
"step": 12560,
|
|
"valid_targets_mean": 2392.4,
|
|
"valid_targets_min": 757
|
|
},
|
|
{
|
|
"epoch": 6.9266813671444325,
|
|
"grad_norm": 0.40390372873299646,
|
|
"learning_rate": 1.3568122873313106e-08,
|
|
"loss": 0.0637,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035549458116292953,
|
|
"step": 12565,
|
|
"valid_targets_mean": 3566.1,
|
|
"valid_targets_min": 664
|
|
},
|
|
{
|
|
"epoch": 6.929437706725468,
|
|
"grad_norm": 0.39914053467229454,
|
|
"learning_rate": 1.2574571319723395e-08,
|
|
"loss": 0.0615,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04623376205563545,
|
|
"step": 12570,
|
|
"valid_targets_mean": 4990.0,
|
|
"valid_targets_min": 4088
|
|
},
|
|
{
|
|
"epoch": 6.932194046306505,
|
|
"grad_norm": 0.3745950705926087,
|
|
"learning_rate": 1.161878186098031e-08,
|
|
"loss": 0.0802,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.031329452991485596,
|
|
"step": 12575,
|
|
"valid_targets_mean": 3248.4,
|
|
"valid_targets_min": 779
|
|
},
|
|
{
|
|
"epoch": 6.934950385887541,
|
|
"grad_norm": 0.5137781308989937,
|
|
"learning_rate": 1.070075630284828e-08,
|
|
"loss": 0.0885,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03825901448726654,
|
|
"step": 12580,
|
|
"valid_targets_mean": 3469.2,
|
|
"valid_targets_min": 1084
|
|
},
|
|
{
|
|
"epoch": 6.937706725468578,
|
|
"grad_norm": 0.457523288553172,
|
|
"learning_rate": 9.820496379748801e-09,
|
|
"loss": 0.0724,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.0451371893286705,
|
|
"step": 12585,
|
|
"valid_targets_mean": 3227.9,
|
|
"valid_targets_min": 700
|
|
},
|
|
{
|
|
"epoch": 6.940463065049614,
|
|
"grad_norm": 0.4368051293694344,
|
|
"learning_rate": 8.978003754744891e-09,
|
|
"loss": 0.0764,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.027652854099869728,
|
|
"step": 12590,
|
|
"valid_targets_mean": 2815.9,
|
|
"valid_targets_min": 769
|
|
},
|
|
{
|
|
"epoch": 6.943219404630651,
|
|
"grad_norm": 0.33439153596930243,
|
|
"learning_rate": 8.17328001955886e-09,
|
|
"loss": 0.092,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.026871275156736374,
|
|
"step": 12595,
|
|
"valid_targets_mean": 3541.8,
|
|
"valid_targets_min": 982
|
|
},
|
|
{
|
|
"epoch": 6.945975744211687,
|
|
"grad_norm": 0.4609353945789452,
|
|
"learning_rate": 7.406326694543442e-09,
|
|
"loss": 0.0694,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04127993434667587,
|
|
"step": 12600,
|
|
"valid_targets_mean": 3189.4,
|
|
"valid_targets_min": 1043
|
|
},
|
|
{
|
|
"epoch": 6.948732083792724,
|
|
"grad_norm": 0.39732565516863816,
|
|
"learning_rate": 6.677145228699555e-09,
|
|
"loss": 0.0748,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03324468806385994,
|
|
"step": 12605,
|
|
"valid_targets_mean": 3380.1,
|
|
"valid_targets_min": 1067
|
|
},
|
|
{
|
|
"epoch": 6.9514884233737595,
|
|
"grad_norm": 0.42311130630578014,
|
|
"learning_rate": 5.985736999665204e-09,
|
|
"loss": 0.0715,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.058023326098918915,
|
|
"step": 12610,
|
|
"valid_targets_mean": 3092.6,
|
|
"valid_targets_min": 901
|
|
},
|
|
{
|
|
"epoch": 6.954244762954796,
|
|
"grad_norm": 0.6274996391578916,
|
|
"learning_rate": 5.332103313713255e-09,
|
|
"loss": 0.1122,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04706891253590584,
|
|
"step": 12615,
|
|
"valid_targets_mean": 1287.2,
|
|
"valid_targets_min": 567
|
|
},
|
|
{
|
|
"epoch": 6.957001102535832,
|
|
"grad_norm": 0.5221086656770177,
|
|
"learning_rate": 4.716245405747e-09,
|
|
"loss": 0.0878,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.047883741557598114,
|
|
"step": 12620,
|
|
"valid_targets_mean": 3460.5,
|
|
"valid_targets_min": 831
|
|
},
|
|
{
|
|
"epoch": 6.959757442116869,
|
|
"grad_norm": 0.4634443093744723,
|
|
"learning_rate": 4.138164439304593e-09,
|
|
"loss": 0.0702,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029030289500951767,
|
|
"step": 12625,
|
|
"valid_targets_mean": 2148.4,
|
|
"valid_targets_min": 632
|
|
},
|
|
{
|
|
"epoch": 6.962513781697905,
|
|
"grad_norm": 0.386145850063163,
|
|
"learning_rate": 3.5978615065479504e-09,
|
|
"loss": 0.0702,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.035001419484615326,
|
|
"step": 12630,
|
|
"valid_targets_mean": 2754.4,
|
|
"valid_targets_min": 1583
|
|
},
|
|
{
|
|
"epoch": 6.965270121278942,
|
|
"grad_norm": 0.3923364744929002,
|
|
"learning_rate": 3.0953376282694126e-09,
|
|
"loss": 0.0688,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029471293091773987,
|
|
"step": 12635,
|
|
"valid_targets_mean": 3267.4,
|
|
"valid_targets_min": 796
|
|
},
|
|
{
|
|
"epoch": 6.968026460859978,
|
|
"grad_norm": 0.3905166842367607,
|
|
"learning_rate": 2.6305937538828595e-09,
|
|
"loss": 0.0631,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02997518517076969,
|
|
"step": 12640,
|
|
"valid_targets_mean": 3208.9,
|
|
"valid_targets_min": 952
|
|
},
|
|
{
|
|
"epoch": 6.970782800441015,
|
|
"grad_norm": 0.5400194832042085,
|
|
"learning_rate": 2.203630761423714e-09,
|
|
"loss": 0.086,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.037696488201618195,
|
|
"step": 12645,
|
|
"valid_targets_mean": 2115.5,
|
|
"valid_targets_min": 885
|
|
},
|
|
{
|
|
"epoch": 6.9735391400220506,
|
|
"grad_norm": 0.41825412741101065,
|
|
"learning_rate": 1.81444945755338e-09,
|
|
"loss": 0.0623,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03784322366118431,
|
|
"step": 12650,
|
|
"valid_targets_mean": 3298.5,
|
|
"valid_targets_min": 1018
|
|
},
|
|
{
|
|
"epoch": 6.976295479603087,
|
|
"grad_norm": 0.46181704909513704,
|
|
"learning_rate": 1.4630505775459213e-09,
|
|
"loss": 0.0681,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03504685312509537,
|
|
"step": 12655,
|
|
"valid_targets_mean": 2170.4,
|
|
"valid_targets_min": 782
|
|
},
|
|
{
|
|
"epoch": 6.979051819184123,
|
|
"grad_norm": 0.46455926630735983,
|
|
"learning_rate": 1.1494347852991639e-09,
|
|
"loss": 0.0694,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04397851228713989,
|
|
"step": 12660,
|
|
"valid_targets_mean": 3052.4,
|
|
"valid_targets_min": 2563
|
|
},
|
|
{
|
|
"epoch": 6.98180815876516,
|
|
"grad_norm": 0.3523323196516239,
|
|
"learning_rate": 8.736026733213721e-10,
|
|
"loss": 0.0725,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.03489045426249504,
|
|
"step": 12665,
|
|
"valid_targets_mean": 3972.9,
|
|
"valid_targets_min": 2264
|
|
},
|
|
{
|
|
"epoch": 6.984564498346196,
|
|
"grad_norm": 0.4257124485105558,
|
|
"learning_rate": 6.355547627445724e-10,
|
|
"loss": 0.0621,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.039443500339984894,
|
|
"step": 12670,
|
|
"valid_targets_mean": 4005.2,
|
|
"valid_targets_min": 2423
|
|
},
|
|
{
|
|
"epoch": 6.987320837927233,
|
|
"grad_norm": 0.40889617547906687,
|
|
"learning_rate": 4.352915033090099e-10,
|
|
"loss": 0.068,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04439264163374901,
|
|
"step": 12675,
|
|
"valid_targets_mean": 4350.4,
|
|
"valid_targets_min": 2545
|
|
},
|
|
{
|
|
"epoch": 6.990077177508269,
|
|
"grad_norm": 0.5824100358118545,
|
|
"learning_rate": 2.728132733720301e-10,
|
|
"loss": 0.1021,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.07505124062299728,
|
|
"step": 12680,
|
|
"valid_targets_mean": 2195.4,
|
|
"valid_targets_min": 1523
|
|
},
|
|
{
|
|
"epoch": 6.992833517089306,
|
|
"grad_norm": 0.3241444796490329,
|
|
"learning_rate": 1.4812037989919702e-10,
|
|
"loss": 0.1191,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.019839102402329445,
|
|
"step": 12685,
|
|
"valid_targets_mean": 3601.6,
|
|
"valid_targets_min": 1092
|
|
},
|
|
{
|
|
"epoch": 6.995589856670342,
|
|
"grad_norm": 0.37947487969533694,
|
|
"learning_rate": 6.121305847317516e-11,
|
|
"loss": 0.069,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.02677541971206665,
|
|
"step": 12690,
|
|
"valid_targets_mean": 3158.4,
|
|
"valid_targets_min": 835
|
|
},
|
|
{
|
|
"epoch": 6.998346196251378,
|
|
"grad_norm": 0.35921445277273123,
|
|
"learning_rate": 1.2091473289288502e-11,
|
|
"loss": 0.0651,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.029492974281311035,
|
|
"step": 12695,
|
|
"valid_targets_mean": 3701.5,
|
|
"valid_targets_min": 2543
|
|
},
|
|
{
|
|
"epoch": 7.0,
|
|
"loss_nan_ranks": 0,
|
|
"loss_rank_avg": 0.04253735765814781,
|
|
"step": 12698,
|
|
"total_flos": 3.879727127867687e+18,
|
|
"train_loss": 0.18411178431570632,
|
|
"train_runtime": 81354.623,
|
|
"train_samples_per_second": 2.497,
|
|
"train_steps_per_second": 0.156,
|
|
"valid_targets_mean": 4458.2,
|
|
"valid_targets_min": 1268
|
|
}
|
|
],
|
|
"logging_steps": 5,
|
|
"max_steps": 12698,
|
|
"num_input_tokens_seen": 0,
|
|
"num_train_epochs": 7,
|
|
"save_steps": 1500,
|
|
"stateful_callbacks": {
|
|
"TrainerControl": {
|
|
"args": {
|
|
"should_epoch_stop": false,
|
|
"should_evaluate": false,
|
|
"should_log": false,
|
|
"should_save": true,
|
|
"should_training_stop": true
|
|
},
|
|
"attributes": {}
|
|
}
|
|
},
|
|
"total_flos": 3.879727127867687e+18,
|
|
"train_batch_size": 1,
|
|
"trial_name": null,
|
|
"trial_params": null
|
|
}
|