{
  "best_global_step": 37500,
  "best_metric": 2.956035614013672,
  "best_model_checkpoint": "/home/u111169/wrkdir/mgh/aav/checkpoints/esm-2_8m-thermo_final_0_2_valid/checkpoint-37500",
  "epoch": 297.6221335992024,
  "eval_steps": 100,
  "global_step": 37500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.07976071784646062,
      "grad_norm": 573.4181518554688,
      "learning_rate": 6e-08,
      "loss": 68.7784,
      "step": 10
    },
    {
      "epoch": 0.15952143569292124,
      "grad_norm": 634.6896362304688,
      "learning_rate": 1.6e-07,
      "loss": 69.1553,
      "step": 20
    },
    {
      "epoch": 0.23928215353938184,
      "grad_norm": 715.6723022460938,
      "learning_rate": 2.5e-07,
      "loss": 65.9578,
      "step": 30
    },
    {
      "epoch": 0.3190428713858425,
      "grad_norm": 547.8534545898438,
      "learning_rate": 3.5e-07,
      "loss": 67.8186,
      "step": 40
    },
    {
      "epoch": 0.3988035892323031,
      "grad_norm": 683.612548828125,
      "learning_rate": 4.5e-07,
      "loss": 60.4619,
      "step": 50
    },
    {
      "epoch": 0.4785643070787637,
      "grad_norm": 555.7923583984375,
      "learning_rate": 4.999992104320635e-07,
      "loss": 62.6769,
      "step": 60
    },
    {
      "epoch": 0.5583250249252243,
      "grad_norm": 1337.35595703125,
      "learning_rate": 4.999928939184958e-07,
      "loss": 59.9945,
      "step": 70
    },
    {
      "epoch": 0.638085742771685,
      "grad_norm": 843.6145629882812,
      "learning_rate": 4.999818085657911e-07,
      "loss": 55.8398,
      "step": 80
    },
    {
      "epoch": 0.7178464606181456,
      "grad_norm": 228.58128356933594,
      "learning_rate": 4.999634912480267e-07,
      "loss": 49.1213,
      "step": 90
    },
    {
      "epoch": 0.7976071784646062,
      "grad_norm": 701.4474487304688,
      "learning_rate": 4.999388583191803e-07,
      "loss": 54.0468,
      "step": 100
    },
    {
      "epoch": 0.7976071784646062,
      "eval_loss": 6.189545154571533,
      "eval_mae": 1.8340115547180176,
      "eval_mse": 6.189545154571533,
      "eval_r2": -0.8648797273635864,
      "eval_rmse": 2.487879650339126,
      "eval_runtime": 9.8346,
      "eval_samples_per_second": 407.846,
      "eval_steps_per_second": 12.812,
      "step": 100
    },
    {
      "epoch": 0.8773678963110668,
      "grad_norm": 479.60687255859375,
      "learning_rate": 4.999079104016307e-07,
      "loss": 48.8717,
      "step": 110
    },
    {
      "epoch": 0.9571286141575274,
      "grad_norm": 171.86424255371094,
      "learning_rate": 4.99870648277312e-07,
      "loss": 46.4531,
      "step": 120
    },
    {
      "epoch": 1.0319042871385842,
      "grad_norm": 294.8586730957031,
      "learning_rate": 4.998270728876944e-07,
      "loss": 46.563,
      "step": 130
    },
    {
      "epoch": 1.111665004985045,
      "grad_norm": 288.3103332519531,
      "learning_rate": 4.997771853337591e-07,
      "loss": 47.1538,
      "step": 140
    },
    {
      "epoch": 1.1914257228315055,
      "grad_norm": 342.15130615234375,
      "learning_rate": 4.997209868759719e-07,
      "loss": 47.3056,
      "step": 150
    },
    {
      "epoch": 1.271186440677966,
      "grad_norm": 403.61468505859375,
      "learning_rate": 4.996584789342507e-07,
      "loss": 48.6825,
      "step": 160
    },
    {
      "epoch": 1.3509471585244266,
      "grad_norm": 587.3318481445312,
      "learning_rate": 4.995896630879293e-07,
      "loss": 43.8702,
      "step": 170
    },
    {
      "epoch": 1.4307078763708874,
      "grad_norm": 380.1310729980469,
      "learning_rate": 4.995145410757182e-07,
      "loss": 40.8683,
      "step": 180
    },
    {
      "epoch": 1.510468594217348,
      "grad_norm": 415.7288818359375,
      "learning_rate": 4.994331147956603e-07,
      "loss": 41.5049,
      "step": 190
    },
    {
      "epoch": 1.5902293120638085,
      "grad_norm": 236.02073669433594,
      "learning_rate": 4.993453863050829e-07,
      "loss": 39.8648,
      "step": 200
    },
    {
      "epoch": 1.5902293120638085,
      "eval_loss": 5.156213283538818,
      "eval_mae": 1.6833100318908691,
      "eval_mse": 5.156213760375977,
      "eval_r2": -0.5535420179367065,
      "eval_rmse": 2.2707297858565156,
      "eval_runtime": 9.5334,
      "eval_samples_per_second": 420.733,
      "eval_steps_per_second": 13.217,
      "step": 200
    },
    {
      "epoch": 1.6699900299102692,
      "grad_norm": 161.90733337402344,
      "learning_rate": 4.992513578205457e-07,
      "loss": 38.1656,
      "step": 210
    },
    {
      "epoch": 1.7497507477567298,
      "grad_norm": 197.21080017089844,
      "learning_rate": 4.991510317177851e-07,
      "loss": 41.23,
      "step": 220
    },
    {
      "epoch": 1.8295114656031903,
      "grad_norm": 457.8110656738281,
      "learning_rate": 4.990444105316537e-07,
      "loss": 36.2166,
      "step": 230
    },
    {
      "epoch": 1.909272183449651,
      "grad_norm": 281.78924560546875,
      "learning_rate": 4.989314969560569e-07,
      "loss": 39.6083,
      "step": 240
    },
    {
      "epoch": 1.9890329012961117,
      "grad_norm": 248.00437927246094,
      "learning_rate": 4.988122938438841e-07,
      "loss": 36.4834,
      "step": 250
    },
    {
      "epoch": 2.0638085742771684,
      "grad_norm": 286.20037841796875,
      "learning_rate": 4.986868042069371e-07,
      "loss": 32.0955,
      "step": 260
    },
    {
      "epoch": 2.143569292123629,
      "grad_norm": 315.5536193847656,
      "learning_rate": 4.985550312158541e-07,
      "loss": 35.8951,
      "step": 270
    },
    {
      "epoch": 2.22333000997009,
      "grad_norm": 406.2935791015625,
      "learning_rate": 4.984169782000289e-07,
      "loss": 33.277,
      "step": 280
    },
    {
      "epoch": 2.30309072781655,
      "grad_norm": 347.6233215332031,
      "learning_rate": 4.982726486475276e-07,
      "loss": 31.5701,
      "step": 290
    },
    {
      "epoch": 2.382851445663011,
      "grad_norm": 145.79922485351562,
      "learning_rate": 4.98122046205e-07,
      "loss": 32.1125,
      "step": 300
    },
    {
      "epoch": 2.382851445663011,
      "eval_loss": 3.720351219177246,
      "eval_mae": 1.4502124786376953,
      "eval_mse": 3.720351457595825,
      "eval_r2": -0.12092375755310059,
      "eval_rmse": 1.928821261184101,
      "eval_runtime": 9.5934,
      "eval_samples_per_second": 418.101,
      "eval_steps_per_second": 13.134,
      "step": 300
    },
    {
      "epoch": 2.4626121635094718,
      "grad_norm": 172.51040649414062,
      "learning_rate": 4.979651746775878e-07,
      "loss": 28.2744,
      "step": 310
    },
    {
      "epoch": 2.542372881355932,
      "grad_norm": 184.6218719482422,
      "learning_rate": 4.978020380288278e-07,
      "loss": 27.2333,
      "step": 320
    },
    {
      "epoch": 2.622133599202393,
      "grad_norm": 84.76827239990234,
      "learning_rate": 4.976326403805526e-07,
      "loss": 31.0476,
      "step": 330
    },
    {
      "epoch": 2.701894317048853,
      "grad_norm": 125.47854614257812,
      "learning_rate": 4.974569860127857e-07,
      "loss": 28.7382,
      "step": 340
    },
    {
      "epoch": 2.781655034895314,
      "grad_norm": 164.69036865234375,
      "learning_rate": 4.972750793636339e-07,
      "loss": 26.0502,
      "step": 350
    },
    {
      "epoch": 2.8614157527417747,
      "grad_norm": 76.86050415039062,
      "learning_rate": 4.970869250291753e-07,
      "loss": 30.6326,
      "step": 360
    },
    {
      "epoch": 2.9411764705882355,
      "grad_norm": 90.23853302001953,
      "learning_rate": 4.968925277633422e-07,
      "loss": 30.7144,
      "step": 370
    },
    {
      "epoch": 3.015952143569292,
      "grad_norm": 106.75017547607422,
      "learning_rate": 4.966918924778022e-07,
      "loss": 25.1257,
      "step": 380
    },
    {
      "epoch": 3.0957128614157527,
      "grad_norm": 147.7263641357422,
      "learning_rate": 4.964850242418333e-07,
      "loss": 28.2507,
      "step": 390
    },
    {
      "epoch": 3.1754735792622135,
      "grad_norm": 94.30587768554688,
      "learning_rate": 4.962719282821961e-07,
      "loss": 29.4349,
      "step": 400
    },
    {
      "epoch": 3.1754735792622135,
      "eval_loss": 3.5484604835510254,
      "eval_mae": 1.4172353744506836,
      "eval_mse": 3.5484604835510254,
      "eval_r2": -0.06913375854492188,
      "eval_rmse": 1.883735778592907,
      "eval_runtime": 9.2043,
      "eval_samples_per_second": 435.777,
      "eval_steps_per_second": 13.689,
      "step": 400
    },
    {
      "epoch": 3.255234297108674,
      "grad_norm": 136.9176788330078,
      "learning_rate": 4.960526099830015e-07,
      "loss": 28.2686,
      "step": 410
    },
    {
      "epoch": 3.3349950149551346,
      "grad_norm": 35.45146560668945,
      "learning_rate": 4.958270748855751e-07,
      "loss": 29.1841,
      "step": 420
    },
    {
      "epoch": 3.4147557328015954,
      "grad_norm": 157.3170928955078,
      "learning_rate": 4.955953286883171e-07,
      "loss": 28.8563,
      "step": 430
    },
    {
      "epoch": 3.4945164506480557,
      "grad_norm": 134.45327758789062,
      "learning_rate": 4.953573772465578e-07,
      "loss": 29.6298,
      "step": 440
    },
    {
      "epoch": 3.5742771684945165,
      "grad_norm": 55.60561752319336,
      "learning_rate": 4.951132265724103e-07,
      "loss": 27.7373,
      "step": 450
    },
    {
      "epoch": 3.6540378863409773,
      "grad_norm": 128.3103485107422,
      "learning_rate": 4.948628828346182e-07,
      "loss": 29.6246,
      "step": 460
    },
    {
      "epoch": 3.7337986041874376,
      "grad_norm": 115.28073120117188,
      "learning_rate": 4.946063523584002e-07,
      "loss": 27.5998,
      "step": 470
    },
    {
      "epoch": 3.8135593220338984,
      "grad_norm": 52.74937057495117,
      "learning_rate": 4.943436416252895e-07,
      "loss": 26.0449,
      "step": 480
    },
    {
      "epoch": 3.8933200398803587,
      "grad_norm": 200.8792724609375,
      "learning_rate": 4.940747572729705e-07,
      "loss": 27.9922,
      "step": 490
    },
    {
      "epoch": 3.9730807577268195,
      "grad_norm": 82.11042022705078,
      "learning_rate": 4.937997060951117e-07,
      "loss": 28.0655,
      "step": 500
    },
    {
      "epoch": 3.9730807577268195,
      "eval_loss": 3.4706692695617676,
      "eval_mae": 1.3996766805648804,
      "eval_mse": 3.4706692695617676,
      "eval_r2": -0.04569566249847412,
      "eval_rmse": 1.8629732337212384,
      "eval_runtime": 9.2717,
      "eval_samples_per_second": 432.606,
      "eval_steps_per_second": 13.59,
      "step": 500
    },
    {
      "epoch": 4.047856430707876,
      "grad_norm": 37.33889389038086,
      "learning_rate": 4.935184950411928e-07,
      "loss": 26.7186,
      "step": 510
    },
    {
      "epoch": 4.127617148554337,
      "grad_norm": 50.46540069580078,
      "learning_rate": 4.932311312163299e-07,
      "loss": 28.5682,
      "step": 520
    },
    {
      "epoch": 4.2073778664007975,
      "grad_norm": 61.525657653808594,
      "learning_rate": 4.929376218810962e-07,
      "loss": 27.7542,
      "step": 530
    },
    {
      "epoch": 4.287138584247258,
      "grad_norm": 69.10831451416016,
      "learning_rate": 4.926379744513378e-07,
      "loss": 31.3671,
      "step": 540
    },
    {
      "epoch": 4.366899302093719,
      "grad_norm": 65.49845123291016,
      "learning_rate": 4.923321964979867e-07,
      "loss": 25.1761,
      "step": 550
    },
    {
      "epoch": 4.44666001994018,
      "grad_norm": 77.61808776855469,
      "learning_rate": 4.920202957468699e-07,
      "loss": 26.5478,
      "step": 560
    },
    {
      "epoch": 4.526420737786641,
      "grad_norm": 73.5521240234375,
      "learning_rate": 4.917022800785133e-07,
      "loss": 26.3101,
      "step": 570
    },
    {
      "epoch": 4.6061814556331,
      "grad_norm": 88.35464477539062,
      "learning_rate": 4.913781575279436e-07,
      "loss": 27.0035,
      "step": 580
    },
    {
      "epoch": 4.685942173479561,
      "grad_norm": 43.666046142578125,
      "learning_rate": 4.910479362844847e-07,
      "loss": 26.8132,
      "step": 590
    },
    {
      "epoch": 4.765702891326022,
      "grad_norm": 67.6700439453125,
      "learning_rate": 4.907116246915507e-07,
      "loss": 30.1273,
      "step": 600
    },
    {
      "epoch": 4.765702891326022,
      "eval_loss": 3.4153451919555664,
      "eval_mae": 1.3900654315948486,
      "eval_mse": 3.4153451919555664,
      "eval_r2": -0.02902674674987793,
      "eval_rmse": 1.8480652564115712,
      "eval_runtime": 9.5716,
      "eval_samples_per_second": 419.053,
      "eval_steps_per_second": 13.164,
      "step": 600
    },
    {
      "epoch": 4.845463609172483,
      "grad_norm": 78.9489974975586,
      "learning_rate": 4.903692312464353e-07,
      "loss": 27.8143,
      "step": 610
    },
    {
      "epoch": 4.9252243270189435,
      "grad_norm": 94.63184356689453,
      "learning_rate": 4.900207646000974e-07,
      "loss": 27.2606,
      "step": 620
    },
    {
      "epoch": 5.0,
      "grad_norm": 58.72568130493164,
      "learning_rate": 4.896662335569419e-07,
      "loss": 24.7743,
      "step": 630
    },
    {
      "epoch": 5.079760717846461,
      "grad_norm": 67.03953552246094,
      "learning_rate": 4.893056470745975e-07,
      "loss": 28.5772,
      "step": 640
    },
    {
      "epoch": 5.1595214356929215,
      "grad_norm": 69.72952270507812,
      "learning_rate": 4.889390142636909e-07,
      "loss": 27.8049,
      "step": 650
    },
    {
      "epoch": 5.239282153539381,
      "grad_norm": 48.28046798706055,
      "learning_rate": 4.885663443876158e-07,
      "loss": 26.902,
      "step": 660
    },
    {
      "epoch": 5.319042871385842,
      "grad_norm": 86.15617370605469,
      "learning_rate": 4.88187646862299e-07,
      "loss": 28.749,
      "step": 670
    },
    {
      "epoch": 5.398803589232303,
      "grad_norm": 65.24939727783203,
      "learning_rate": 4.878029312559633e-07,
      "loss": 25.8315,
      "step": 680
    },
    {
      "epoch": 5.478564307078764,
      "grad_norm": 34.22309112548828,
      "learning_rate": 4.874122072888844e-07,
      "loss": 29.6742,
      "step": 690
    },
    {
      "epoch": 5.5583250249252245,
      "grad_norm": 136.18731689453125,
      "learning_rate": 4.870154848331468e-07,
      "loss": 28.3412,
      "step": 700
    },
    {
      "epoch": 5.5583250249252245,
      "eval_loss": 3.371689796447754,
      "eval_mae": 1.392737865447998,
      "eval_mse": 3.371690034866333,
      "eval_r2": -0.01587367057800293,
      "eval_rmse": 1.8362162276993232,
      "eval_runtime": 9.2257,
      "eval_samples_per_second": 434.766,
      "eval_steps_per_second": 13.658,
      "step": 700
    },
    {
      "epoch": 5.638085742771685,
      "grad_norm": 99.80322265625,
      "learning_rate": 4.866127739123931e-07,
      "loss": 27.3187,
      "step": 710
    },
    {
      "epoch": 5.717846460618146,
      "grad_norm": 62.02863311767578,
      "learning_rate": 4.862040847015712e-07,
      "loss": 25.038,
      "step": 720
    },
    {
      "epoch": 5.797607178464606,
      "grad_norm": 68.26268768310547,
      "learning_rate": 4.857894275266779e-07,
      "loss": 26.4793,
      "step": 730
    },
    {
      "epoch": 5.877367896311067,
      "grad_norm": 51.290000915527344,
      "learning_rate": 4.853688128644969e-07,
      "loss": 27.7466,
      "step": 740
    },
    {
      "epoch": 5.9571286141575275,
      "grad_norm": 56.944007873535156,
      "learning_rate": 4.849422513423343e-07,
      "loss": 24.5587,
      "step": 750
    },
    {
      "epoch": 6.031904287138584,
      "grad_norm": 53.164737701416016,
      "learning_rate": 4.845097537377511e-07,
      "loss": 24.6128,
      "step": 760
    },
    {
      "epoch": 6.111665004985045,
      "grad_norm": 37.63719177246094,
      "learning_rate": 4.840713309782894e-07,
      "loss": 25.3815,
      "step": 770
    },
    {
      "epoch": 6.1914257228315055,
      "grad_norm": 84.40767669677734,
      "learning_rate": 4.836269941411977e-07,
      "loss": 27.9501,
      "step": 780
    },
    {
      "epoch": 6.271186440677966,
      "grad_norm": 40.62728500366211,
      "learning_rate": 4.831767544531501e-07,
      "loss": 26.8201,
      "step": 790
    },
    {
      "epoch": 6.350947158524427,
      "grad_norm": 91.77393341064453,
      "learning_rate": 4.827206232899628e-07,
      "loss": 27.7182,
      "step": 800
    },
    {
      "epoch": 6.350947158524427,
      "eval_loss": 3.340036630630493,
      "eval_mae": 1.385007619857788,
      "eval_mse": 3.340036630630493,
      "eval_r2": -0.006336688995361328,
      "eval_rmse": 1.8275767099168487,
      "eval_runtime": 9.255,
      "eval_samples_per_second": 433.389,
      "eval_steps_per_second": 13.614,
      "step": 800
    },
    {
      "epoch": 6.430707876370887,
      "grad_norm": 25.9021053314209,
      "learning_rate": 4.822586121763069e-07,
      "loss": 26.24,
      "step": 810
    },
    {
      "epoch": 6.510468594217348,
      "grad_norm": 97.57054901123047,
      "learning_rate": 4.817907327854172e-07,
      "loss": 27.2087,
      "step": 820
    },
    {
      "epoch": 6.5902293120638085,
      "grad_norm": 36.019039154052734,
      "learning_rate": 4.813169969387971e-07,
      "loss": 27.6542,
      "step": 830
    },
    {
      "epoch": 6.669990029910269,
      "grad_norm": 34.63457489013672,
      "learning_rate": 4.8083741660592e-07,
      "loss": 27.4211,
      "step": 840
    },
    {
      "epoch": 6.74975074775673,
      "grad_norm": 51.112457275390625,
      "learning_rate": 4.80352003903927e-07,
      "loss": 27.1675,
      "step": 850
    },
    {
      "epoch": 6.829511465603191,
      "grad_norm": 42.236328125,
      "learning_rate": 4.798607710973204e-07,
      "loss": 25.7901,
      "step": 860
    },
    {
      "epoch": 6.909272183449651,
      "grad_norm": 81.07854461669922,
      "learning_rate": 4.793637305976542e-07,
      "loss": 26.0094,
      "step": 870
    },
    {
      "epoch": 6.989032901296111,
      "grad_norm": 22.95639991760254,
      "learning_rate": 4.788608949632202e-07,
      "loss": 27.9139,
      "step": 880
    },
    {
      "epoch": 7.063808574277169,
      "grad_norm": 70.43840026855469,
      "learning_rate": 4.78352276898731e-07,
      "loss": 25.0272,
      "step": 890
    },
    {
      "epoch": 7.14356929212363,
      "grad_norm": 140.94937133789062,
      "learning_rate": 4.77837889254999e-07,
      "loss": 27.7522,
      "step": 900
    },
    {
      "epoch": 7.14356929212363,
      "eval_loss": 3.3163321018218994,
      "eval_mae": 1.3711785078048706,
      "eval_mse": 3.3163323402404785,
      "eval_r2": 0.0008053183555603027,
      "eval_rmse": 1.8210799928175803,
      "eval_runtime": 9.2647,
      "eval_samples_per_second": 432.931,
      "eval_steps_per_second": 13.6,
      "step": 900
    },
    {
      "epoch": 7.2233300099700894,
      "grad_norm": 24.217979431152344,
      "learning_rate": 4.77317745028611e-07,
      "loss": 28.0138,
      "step": 910
    },
    {
      "epoch": 7.30309072781655,
      "grad_norm": 70.4435806274414,
      "learning_rate": 4.767918573616011e-07,
      "loss": 27.8576,
      "step": 920
    },
    {
      "epoch": 7.382851445663011,
      "grad_norm": 177.81341552734375,
      "learning_rate": 4.7626023954111744e-07,
      "loss": 27.0952,
      "step": 930
    },
    {
      "epoch": 7.462612163509472,
      "grad_norm": 32.845672607421875,
      "learning_rate": 4.7572290499908714e-07,
      "loss": 25.4676,
      "step": 940
    },
    {
      "epoch": 7.5423728813559325,
      "grad_norm": 54.082908630371094,
      "learning_rate": 4.751798673118768e-07,
      "loss": 24.4115,
      "step": 950
    },
    {
      "epoch": 7.622133599202392,
      "grad_norm": 59.80043029785156,
      "learning_rate": 4.7463114019994945e-07,
      "loss": 25.2318,
      "step": 960
    },
    {
      "epoch": 7.701894317048853,
      "grad_norm": 58.63386154174805,
      "learning_rate": 4.7407673752751787e-07,
      "loss": 27.5051,
      "step": 970
    },
    {
      "epoch": 7.781655034895314,
      "grad_norm": 55.40766906738281,
      "learning_rate": 4.7351667330219427e-07,
      "loss": 28.0926,
      "step": 980
    },
    {
      "epoch": 7.861415752741775,
      "grad_norm": 113.65251922607422,
      "learning_rate": 4.7295096167463644e-07,
      "loss": 25.48,
      "step": 990
    },
    {
      "epoch": 7.9411764705882355,
      "grad_norm": 101.72802734375,
      "learning_rate": 4.723796169381903e-07,
      "loss": 26.3553,
      "step": 1000
    },
    {
      "epoch": 7.9411764705882355,
      "eval_loss": 3.298834800720215,
      "eval_mae": 1.3671762943267822,
      "eval_mse": 3.2988345623016357,
      "eval_r2": 0.006077289581298828,
      "eval_rmse": 1.8162694079628263,
      "eval_runtime": 9.6354,
      "eval_samples_per_second": 416.278,
      "eval_steps_per_second": 13.077,
      "step": 1000
    },
    {
      "epoch": 8.015952143569292,
      "grad_norm": 28.65477752685547,
      "learning_rate": 4.7180265352852846e-07,
      "loss": 25.2742,
      "step": 1010
    },
    {
      "epoch": 8.095712861415752,
      "grad_norm": 77.14867401123047,
      "learning_rate": 4.7122008602328584e-07,
      "loss": 26.0271,
      "step": 1020
    },
    {
      "epoch": 8.175473579262214,
      "grad_norm": 62.96834182739258,
      "learning_rate": 4.706319291416911e-07,
      "loss": 26.7356,
      "step": 1030
    },
    {
      "epoch": 8.255234297108673,
      "grad_norm": 65.69111633300781,
      "learning_rate": 4.700381977441949e-07,
      "loss": 27.3675,
      "step": 1040
    },
    {
      "epoch": 8.334995014955135,
      "grad_norm": 69.17853546142578,
      "learning_rate": 4.694389068320943e-07,
      "loss": 26.3667,
      "step": 1050
    },
    {
      "epoch": 8.414755732801595,
      "grad_norm": 24.640756607055664,
      "learning_rate": 4.6883407154715384e-07,
      "loss": 25.1328,
      "step": 1060
    },
    {
      "epoch": 8.494516450648057,
      "grad_norm": 164.16966247558594,
      "learning_rate": 4.6822370717122294e-07,
      "loss": 26.6883,
      "step": 1070
    },
    {
      "epoch": 8.574277168494516,
      "grad_norm": 55.141292572021484,
      "learning_rate": 4.676078291258497e-07,
      "loss": 26.6146,
      "step": 1080
    },
    {
      "epoch": 8.654037886340976,
      "grad_norm": 56.901947021484375,
      "learning_rate": 4.6698645297189145e-07,
      "loss": 27.9764,
      "step": 1090
    },
    {
      "epoch": 8.733798604187438,
      "grad_norm": 35.33796310424805,
      "learning_rate": 4.663595944091213e-07,
      "loss": 26.8944,
      "step": 1100
    },
    {
      "epoch": 8.733798604187438,
      "eval_loss": 3.2831168174743652,
      "eval_mae": 1.371146321296692,
      "eval_mse": 3.2831168174743652,
      "eval_r2": 0.010812938213348389,
      "eval_rmse": 1.8119373105806849,
      "eval_runtime": 9.6483,
      "eval_samples_per_second": 415.723,
      "eval_steps_per_second": 13.059,
      "step": 1100
    },
    {
      "epoch": 8.813559322033898,
      "grad_norm": 55.841068267822266,
      "learning_rate": 4.6572726927583185e-07,
      "loss": 27.0192,
      "step": 1110
    },
    {
      "epoch": 8.89332003988036,
      "grad_norm": 70.44389343261719,
      "learning_rate": 4.650894935484346e-07,
      "loss": 25.9054,
      "step": 1120
    },
    {
      "epoch": 8.97308075772682,
      "grad_norm": 37.819854736328125,
      "learning_rate": 4.6444628334105673e-07,
      "loss": 25.7866,
      "step": 1130
    },
    {
      "epoch": 9.047856430707876,
      "grad_norm": 118.53936767578125,
      "learning_rate": 4.6379765490513345e-07,
      "loss": 22.7737,
      "step": 1140
    },
    {
      "epoch": 9.127617148554338,
      "grad_norm": 39.31532287597656,
      "learning_rate": 4.631436246289978e-07,
      "loss": 27.3392,
      "step": 1150
    },
    {
      "epoch": 9.207377866400797,
      "grad_norm": 46.137168884277344,
      "learning_rate": 4.6248420903746646e-07,
      "loss": 27.0129,
      "step": 1160
    },
    {
      "epoch": 9.28713858424726,
      "grad_norm": 49.3219108581543,
      "learning_rate": 4.618194247914221e-07,
      "loss": 27.1411,
      "step": 1170
    },
    {
      "epoch": 9.366899302093719,
      "grad_norm": 80.34762573242188,
      "learning_rate": 4.6114928868739263e-07,
      "loss": 25.0363,
      "step": 1180
    },
    {
      "epoch": 9.446660019940179,
      "grad_norm": 65.08316040039062,
      "learning_rate": 4.6047381765712675e-07,
      "loss": 27.1349,
      "step": 1190
    },
    {
      "epoch": 9.52642073778664,
      "grad_norm": 54.11983871459961,
      "learning_rate": 4.59793028767166e-07,
      "loss": 26.5226,
      "step": 1200
    },
    {
      "epoch": 9.52642073778664,
      "eval_loss": 3.269491195678711,
      "eval_mae": 1.3665425777435303,
      "eval_mse": 3.269491195678711,
      "eval_r2": 0.014918327331542969,
      "eval_rmse": 1.8081734418132325,
      "eval_runtime": 9.2453,
      "eval_samples_per_second": 433.842,
      "eval_steps_per_second": 13.629,
      "step": 1200
    },
    {
      "epoch": 9.6061814556331,
      "grad_norm": 99.34823608398438,
      "learning_rate": 4.591069392184138e-07,
      "loss": 26.6684,
      "step": 1210
    },
    {
      "epoch": 9.685942173479562,
      "grad_norm": 73.60244750976562,
      "learning_rate": 4.5841556634570073e-07,
      "loss": 26.9203,
      "step": 1220
    },
    {
      "epoch": 9.765702891326022,
      "grad_norm": 43.70383071899414,
      "learning_rate": 4.5771892761734633e-07,
      "loss": 26.2737,
      "step": 1230
    },
    {
      "epoch": 9.845463609172482,
      "grad_norm": 25.8046817779541,
      "learning_rate": 4.570170406347182e-07,
      "loss": 24.5509,
      "step": 1240
    },
    {
      "epoch": 9.925224327018944,
      "grad_norm": 92.3177490234375,
      "learning_rate": 4.5630992313178697e-07,
      "loss": 25.4429,
      "step": 1250
    },
    {
      "epoch": 10.0,
      "grad_norm": 52.23972702026367,
      "learning_rate": 4.5559759297467825e-07,
      "loss": 25.9577,
      "step": 1260
    },
    {
      "epoch": 10.07976071784646,
      "grad_norm": 41.57015609741211,
      "learning_rate": 4.548800681612212e-07,
      "loss": 26.6287,
      "step": 1270
    },
    {
      "epoch": 10.159521435692922,
      "grad_norm": 81.64158630371094,
      "learning_rate": 4.541573668204941e-07,
      "loss": 26.0703,
      "step": 1280
    },
    {
      "epoch": 10.239282153539381,
      "grad_norm": 27.914758682250977,
      "learning_rate": 4.5342950721236584e-07,
      "loss": 26.6378,
      "step": 1290
    },
    {
      "epoch": 10.319042871385843,
      "grad_norm": 53.98166275024414,
      "learning_rate": 4.526965077270348e-07,
      "loss": 26.146,
      "step": 1300
    },
    {
      "epoch": 10.319042871385843,
      "eval_loss": 3.2577223777770996,
      "eval_mae": 1.3596596717834473,
      "eval_mse": 3.2577223777770996,
      "eval_r2": 0.01846414804458618,
      "eval_rmse": 1.8049161691826852,
      "eval_runtime": 9.2569,
      "eval_samples_per_second": 433.299,
      "eval_steps_per_second": 13.611,
      "step": 1300
    },
    {
      "epoch": 10.398803589232303,
      "grad_norm": 123.96611785888672,
      "learning_rate": 4.5195838688456434e-07,
      "loss": 29.7088,
      "step": 1310
    },
    {
      "epoch": 10.478564307078763,
      "grad_norm": 62.53553009033203,
      "learning_rate": 4.5121516333441466e-07,
      "loss": 25.1537,
      "step": 1320
    },
    {
      "epoch": 10.558325024925225,
      "grad_norm": 75.32501220703125,
      "learning_rate": 4.5046685585497146e-07,
      "loss": 26.2433,
      "step": 1330
    },
    {
      "epoch": 10.638085742771684,
      "grad_norm": 49.98992156982422,
      "learning_rate": 4.4971348335307203e-07,
      "loss": 26.079,
      "step": 1340
    },
    {
      "epoch": 10.717846460618146,
      "grad_norm": 137.5787811279297,
      "learning_rate": 4.4895506486352686e-07,
      "loss": 24.1009,
      "step": 1350
    },
    {
      "epoch": 10.797607178464606,
      "grad_norm": 75.4178237915039,
      "learning_rate": 4.4819161954863927e-07,
      "loss": 24.9365,
      "step": 1360
    },
    {
      "epoch": 10.877367896311068,
      "grad_norm": 130.9047088623047,
      "learning_rate": 4.4742316669772096e-07,
      "loss": 26.0591,
      "step": 1370
    },
    {
      "epoch": 10.957128614157527,
      "grad_norm": 48.49380874633789,
      "learning_rate": 4.466497257266046e-07,
      "loss": 27.9292,
      "step": 1380
    },
    {
      "epoch": 11.031904287138584,
      "grad_norm": 28.311349868774414,
      "learning_rate": 4.4587131617715346e-07,
      "loss": 23.6428,
      "step": 1390
    },
    {
      "epoch": 11.111665004985046,
      "grad_norm": 140.38153076171875,
      "learning_rate": 4.4508795771676767e-07,
      "loss": 25.9676,
      "step": 1400
    },
    {
      "epoch": 11.111665004985046,
      "eval_loss": 3.247952938079834,
      "eval_mae": 1.3569178581237793,
      "eval_mse": 3.247952938079834,
      "eval_r2": 0.021407663822174072,
      "eval_rmse": 1.8022077954774898,
      "eval_runtime": 9.3541,
      "eval_samples_per_second": 428.795,
      "eval_steps_per_second": 13.47,
      "step": 1400
    },
    {
      "epoch": 11.191425722831505,
      "grad_norm": 30.292320251464844,
      "learning_rate": 4.4429967013788706e-07,
      "loss": 27.9688,
      "step": 1410
    },
    {
      "epoch": 11.271186440677965,
      "grad_norm": 57.311614990234375,
      "learning_rate": 4.4350647335749126e-07,
      "loss": 25.5213,
      "step": 1420
    },
    {
      "epoch": 11.350947158524427,
      "grad_norm": 72.53961944580078,
      "learning_rate": 4.4270838741659653e-07,
      "loss": 26.5346,
      "step": 1430
    },
    {
      "epoch": 11.430707876370887,
      "grad_norm": 91.33097839355469,
      "learning_rate": 4.419054324797492e-07,
      "loss": 25.9467,
      "step": 1440
    },
    {
      "epoch": 11.510468594217349,
      "grad_norm": 49.6580810546875,
      "learning_rate": 4.4109762883451637e-07,
      "loss": 25.582,
      "step": 1450
    },
    {
      "epoch": 11.590229312063808,
      "grad_norm": 55.276771545410156,
      "learning_rate": 4.402849968909732e-07,
      "loss": 25.4052,
      "step": 1460
    },
    {
      "epoch": 11.66999002991027,
      "grad_norm": 33.58572769165039,
      "learning_rate": 4.394675571811873e-07,
      "loss": 26.9084,
      "step": 1470
    },
    {
      "epoch": 11.74975074775673,
      "grad_norm": 119.92269134521484,
      "learning_rate": 4.386453303587002e-07,
      "loss": 26.1702,
      "step": 1480
    },
    {
      "epoch": 11.82951146560319,
      "grad_norm": 41.54932403564453,
      "learning_rate": 4.3781833719800474e-07,
      "loss": 26.7313,
      "step": 1490
    },
    {
      "epoch": 11.909272183449652,
      "grad_norm": 38.98570251464844,
      "learning_rate": 4.369865985940212e-07,
      "loss": 25.5662,
      "step": 1500
    },
    {
      "epoch": 11.909272183449652,
      "eval_loss": 3.238109588623047,
      "eval_mae": 1.3584039211273193,
      "eval_mse": 3.238109588623047,
      "eval_r2": 0.024373412132263184,
      "eval_rmse": 1.7994748091104378,
      "eval_runtime": 9.6443,
      "eval_samples_per_second": 415.893,
      "eval_steps_per_second": 13.065,
      "step": 1500
    },
    {
      "epoch": 11.989032901296111,
      "grad_norm": 82.44010162353516,
      "learning_rate": 4.361501355615685e-07,
      "loss": 27.1074,
      "step": 1510
    },
    {
      "epoch": 12.063808574277168,
      "grad_norm": 84.16273498535156,
      "learning_rate": 4.3530896923483387e-07,
      "loss": 24.2903,
      "step": 1520
    },
    {
      "epoch": 12.14356929212363,
      "grad_norm": 96.39720153808594,
      "learning_rate": 4.3446312086683846e-07,
      "loss": 25.0496,
      "step": 1530
    },
    {
      "epoch": 12.22333000997009,
      "grad_norm": 34.487491607666016,
      "learning_rate": 4.3361261182890054e-07,
      "loss": 26.8909,
      "step": 1540
    },
    {
      "epoch": 12.303090727816551,
      "grad_norm": 38.0224609375,
      "learning_rate": 4.3275746361009547e-07,
      "loss": 26.0252,
      "step": 1550
    },
    {
      "epoch": 12.382851445663011,
      "grad_norm": 21.54315948486328,
      "learning_rate": 4.3189769781671283e-07,
      "loss": 24.2614,
      "step": 1560
    },
    {
      "epoch": 12.46261216350947,
      "grad_norm": 49.494964599609375,
      "learning_rate": 4.3103333617171035e-07,
      "loss": 26.9778,
      "step": 1570
    },
    {
      "epoch": 12.542372881355933,
      "grad_norm": 22.109155654907227,
      "learning_rate": 4.3016440051416537e-07,
      "loss": 24.0684,
      "step": 1580
    },
    {
      "epoch": 12.622133599202392,
      "grad_norm": 37.458683013916016,
      "learning_rate": 4.292909127987226e-07,
      "loss": 26.7605,
      "step": 1590
    },
    {
      "epoch": 12.701894317048854,
      "grad_norm": 192.13340759277344,
      "learning_rate": 4.2841289509503977e-07,
      "loss": 26.3182,
      "step": 1600
    },
    {
      "epoch": 12.701894317048854,
      "eval_loss": 3.2306623458862305,
      "eval_mae": 1.35684072971344,
      "eval_mse": 3.2306623458862305,
      "eval_r2": 0.026617228984832764,
      "eval_rmse": 1.797404335670255,
      "eval_runtime": 9.2307,
      "eval_samples_per_second": 434.528,
      "eval_steps_per_second": 13.65,
      "step": 1600
    },
    {
      "epoch": 12.781655034895314,
      "grad_norm": 102.10505676269531,
      "learning_rate": 4.2753036958723e-07,
      "loss": 26.7662,
      "step": 1610
    },
    {
      "epoch": 12.861415752741774,
      "grad_norm": 29.1319637298584,
      "learning_rate": 4.2664335857330093e-07,
      "loss": 26.8907,
      "step": 1620
    },
    {
      "epoch": 12.941176470588236,
      "grad_norm": 39.71030044555664,
      "learning_rate": 4.2575188446459187e-07,
      "loss": 27.072,
      "step": 1630
    },
    {
      "epoch": 13.015952143569292,
      "grad_norm": 95.07027435302734,
      "learning_rate": 4.248559697852071e-07,
      "loss": 23.2089,
      "step": 1640
    },
    {
      "epoch": 13.095712861415752,
      "grad_norm": 36.529666900634766,
      "learning_rate": 4.239556371714471e-07,
      "loss": 26.037,
      "step": 1650
    },
    {
      "epoch": 13.175473579262214,
      "grad_norm": 47.9398307800293,
      "learning_rate": 4.2305090937123636e-07,
      "loss": 25.8954,
      "step": 1660
    },
    {
      "epoch": 13.255234297108673,
      "grad_norm": 71.37847900390625,
      "learning_rate": 4.221418092435488e-07,
      "loss": 23.7725,
      "step": 1670
    },
    {
      "epoch": 13.334995014955135,
      "grad_norm": 41.88365173339844,
      "learning_rate": 4.2122835975783005e-07,
      "loss": 25.5059,
      "step": 1680
    },
    {
      "epoch": 13.414755732801595,
      "grad_norm": 34.54059600830078,
      "learning_rate": 4.203105839934173e-07,
      "loss": 27.994,
      "step": 1690
    },
    {
      "epoch": 13.494516450648057,
      "grad_norm": 116.55763244628906,
      "learning_rate": 4.1938850513895626e-07,
      "loss": 26.5336,
      "step": 1700
    },
    {
      "epoch": 13.494516450648057,
      "eval_loss": 3.223775625228882,
      "eval_mae": 1.355072021484375,
      "eval_mse": 3.223775863647461,
      "eval_r2": 0.02869206666946411,
      "eval_rmse": 1.7954876395139736,
      "eval_runtime": 9.2588,
      "eval_samples_per_second": 433.21,
      "eval_steps_per_second": 13.609,
      "step": 1700
    },
    {
      "epoch": 13.574277168494516,
      "grad_norm": 68.4130630493164,
      "learning_rate": 4.1846214649181455e-07,
      "loss": 28.2323,
      "step": 1710
    },
    {
      "epoch": 13.654037886340976,
      "grad_norm": 50.52378845214844,
      "learning_rate": 4.175315314574942e-07,
      "loss": 26.1683,
      "step": 1720
    },
    {
      "epoch": 13.733798604187438,
      "grad_norm": 109.47400665283203,
      "learning_rate": 4.1659668354903934e-07,
      "loss": 25.7688,
      "step": 1730
    },
    {
      "epoch": 13.813559322033898,
      "grad_norm": 51.831729888916016,
      "learning_rate": 4.1565762638644255e-07,
      "loss": 25.1936,
      "step": 1740
    },
    {
      "epoch": 13.89332003988036,
      "grad_norm": 97.07423400878906,
      "learning_rate": 4.1471438369604795e-07,
      "loss": 25.2574,
      "step": 1750
    },
    {
      "epoch": 13.97308075772682,
      "grad_norm": 194.64306640625,
      "learning_rate": 4.13766979309952e-07,
      "loss": 25.8043,
      "step": 1760
    },
    {
      "epoch": 14.047856430707876,
      "grad_norm": 166.36204528808594,
      "learning_rate": 4.128154371654007e-07,
      "loss": 24.2276,
      "step": 1770
    },
    {
      "epoch": 14.127617148554338,
      "grad_norm": 28.835357666015625,
      "learning_rate": 4.1185978130418566e-07,
      "loss": 27.2479,
      "step": 1780
    },
    {
      "epoch": 14.207377866400797,
      "grad_norm": 161.6560821533203,
      "learning_rate": 4.109000358720358e-07,
      "loss": 27.8549,
      "step": 1790
    },
    {
      "epoch": 14.28713858424726,
      "grad_norm": 42.74077606201172,
      "learning_rate": 4.0993622511800794e-07,
      "loss": 25.8464,
      "step": 1800
    },
    {
      "epoch": 14.28713858424726,
      "eval_loss": 3.2174415588378906,
      "eval_mae": 1.352028727531433,
      "eval_mse": 3.2174415588378906,
      "eval_r2": 0.03060060739517212,
      "eval_rmse": 1.7937228210729468,
      "eval_runtime": 9.2733,
      "eval_samples_per_second": 432.53,
      "eval_steps_per_second": 13.587,
      "step": 1800
    },
    {
      "epoch": 14.366899302093719,
      "grad_norm": 94.52295684814453,
      "learning_rate": 4.08968373393874e-07,
      "loss": 26.218,
      "step": 1810
    },
    {
      "epoch": 14.446660019940179,
      "grad_norm": 102.33257293701172,
      "learning_rate": 4.0799650515350516e-07,
      "loss": 25.0329,
      "step": 1820
    },
    {
      "epoch": 14.52642073778664,
      "grad_norm": 77.26675415039062,
      "learning_rate": 4.070206449522547e-07,
      "loss": 26.2714,
      "step": 1830
    },
    {
      "epoch": 14.6061814556331,
      "grad_norm": 57.858577728271484,
      "learning_rate": 4.0604081744633745e-07,
      "loss": 25.818,
      "step": 1840
    },
    {
      "epoch": 14.685942173479562,
      "grad_norm": 53.28114318847656,
      "learning_rate": 4.050570473922064e-07,
      "loss": 26.48,
      "step": 1850
    },
    {
      "epoch": 14.765702891326022,
      "grad_norm": 93.24755096435547,
      "learning_rate": 4.040693596459277e-07,
      "loss": 26.9171,
      "step": 1860
    },
    {
      "epoch": 14.845463609172482,
      "grad_norm": 58.55867004394531,
      "learning_rate": 4.0307777916255237e-07,
      "loss": 26.2595,
      "step": 1870
    },
    {
      "epoch": 14.925224327018944,
      "grad_norm": 125.5550765991211,
      "learning_rate": 4.0208233099548573e-07,
      "loss": 23.0974,
      "step": 1880
    },
    {
      "epoch": 15.0,
      "grad_norm": 29.170337677001953,
      "learning_rate": 4.0108304029585474e-07,
      "loss": 23.3274,
      "step": 1890
    },
    {
      "epoch": 15.07976071784646,
      "grad_norm": 55.78076171875,
      "learning_rate": 4.000799323118722e-07,
      "loss": 26.2063,
      "step": 1900
    },
    {
      "epoch": 15.07976071784646,
      "eval_loss": 3.2104992866516113,
      "eval_mae": 1.3506470918655396,
      "eval_mse": 3.2104992866516113,
      "eval_r2": 0.03269225358963013,
      "eval_rmse": 1.7917866186160705,
      "eval_runtime": 9.5781,
      "eval_samples_per_second": 418.77,
      "eval_steps_per_second": 13.155,
      "step": 1900
    },
    {
      "epoch": 15.159521435692922,
      "grad_norm": 40.96506118774414,
      "learning_rate": 3.990730323881988e-07,
      "loss": 23.7108,
      "step": 1910
    },
    {
      "epoch": 15.239282153539381,
      "grad_norm": 68.11751556396484,
      "learning_rate": 3.9806236596530316e-07,
      "loss": 26.1233,
      "step": 1920
    },
    {
      "epoch": 15.319042871385843,
      "grad_norm": 76.66336822509766,
      "learning_rate": 3.9704795857881856e-07,
      "loss": 26.4697,
      "step": 1930
    },
    {
      "epoch": 15.398803589232303,
      "grad_norm": 160.65145874023438,
      "learning_rate": 3.9602983585889794e-07,
      "loss": 24.5925,
      "step": 1940
    },
    {
      "epoch": 15.478564307078763,
      "grad_norm": 33.02718734741211,
      "learning_rate": 3.950080235295666e-07,
      "loss": 24.7061,
      "step": 1950
    },
    {
      "epoch": 15.558325024925225,
      "grad_norm": 119.47199249267578,
      "learning_rate": 3.939825474080717e-07,
      "loss": 29.3217,
      "step": 1960
    },
    {
      "epoch": 15.638085742771684,
      "grad_norm": 72.96931457519531,
      "learning_rate": 3.9295343340423043e-07,
      "loss": 28.4587,
      "step": 1970
    },
    {
      "epoch": 15.717846460618146,
      "grad_norm": 57.194244384765625,
      "learning_rate": 3.9192070751977526e-07,
      "loss": 26.4959,
      "step": 1980
    },
    {
      "epoch": 15.797607178464606,
      "grad_norm": 34.6958122253418,
      "learning_rate": 3.908843958476968e-07,
      "loss": 25.5749,
      "step": 1990
    },
    {
      "epoch": 15.877367896311068,
      "grad_norm": 28.211673736572266,
      "learning_rate": 3.8984452457158466e-07,
      "loss": 25.7622,
      "step": 2000
    },
    {
      "epoch": 15.877367896311068,
      "eval_loss": 3.2040445804595947,
      "eval_mae": 1.3519315719604492,
      "eval_mse": 3.204044818878174,
      "eval_r2": 0.0346369743347168,
      "eval_rmse": 1.789984586212455,
      "eval_runtime": 9.1533,
      "eval_samples_per_second": 438.202,
      "eval_steps_per_second": 13.766,
      "step": 2000
    },
    {
      "epoch": 15.957128614157527,
      "grad_norm": 40.19672775268555,
      "learning_rate": 3.88801119964966e-07,
      "loss": 23.7187,
      "step": 2010
    },
    {
      "epoch": 16.031904287138584,
      "grad_norm": 125.82440185546875,
      "learning_rate": 3.877542083906414e-07,
      "loss": 23.9347,
      "step": 2020
    },
    {
      "epoch": 16.111665004985046,
      "grad_norm": 47.25736618041992,
      "learning_rate": 3.8670381630001914e-07,
      "loss": 26.4004,
      "step": 2030
    },
    {
      "epoch": 16.191425722831504,
      "grad_norm": 150.1745147705078,
      "learning_rate": 3.856499702324466e-07,
      "loss": 26.9673,
      "step": 2040
    },
    {
      "epoch": 16.271186440677965,
      "grad_norm": 135.0413055419922,
      "learning_rate": 3.845926968145398e-07,
      "loss": 26.9109,
      "step": 2050
    },
    {
      "epoch": 16.350947158524427,
      "grad_norm": 100.4754638671875,
      "learning_rate": 3.8353202275951054e-07,
      "loss": 24.002,
      "step": 2060
    },
    {
      "epoch": 16.43070787637089,
      "grad_norm": 95.51912689208984,
      "learning_rate": 3.8246797486649185e-07,
      "loss": 26.8029,
      "step": 2070
    },
    {
      "epoch": 16.510468594217347,
      "grad_norm": 38.787960052490234,
      "learning_rate": 3.8140058001986043e-07,
      "loss": 26.6201,
      "step": 2080
    },
    {
      "epoch": 16.59022931206381,
      "grad_norm": 155.7447967529297,
      "learning_rate": 3.803298651885575e-07,
      "loss": 26.3985,
      "step": 2090
    },
    {
      "epoch": 16.66999002991027,
      "grad_norm": 85.08961486816406,
      "learning_rate": 3.7925585742540766e-07,
      "loss": 24.1255,
      "step": 2100
    },
    {
      "epoch": 16.66999002991027,
      "eval_loss": 3.1988422870635986,
      "eval_mae": 1.3528438806533813,
      "eval_mse": 3.1988422870635986,
      "eval_r2": 0.03620445728302002,
      "eval_rmse": 1.788530762123928,
      "eval_runtime": 9.2596,
      "eval_samples_per_second": 433.172,
      "eval_steps_per_second": 13.608,
      "step": 2100
    },
    {
      "epoch": 16.74975074775673,
      "grad_norm": 74.64730834960938,
      "learning_rate": 3.781785838664349e-07,
      "loss": 25.9132,
      "step": 2110
    },
    {
      "epoch": 16.82951146560319,
      "grad_norm": 106.97097778320312,
      "learning_rate": 3.770980717301775e-07,
      "loss": 24.2703,
      "step": 2120
    },
    {
      "epoch": 16.90927218344965,
      "grad_norm": 78.0794448852539,
      "learning_rate": 3.760143483169999e-07,
      "loss": 27.0184,
      "step": 2130
    },
    {
      "epoch": 16.989032901296113,
      "grad_norm": 28.421470642089844,
      "learning_rate": 3.749274410084032e-07,
      "loss": 24.9771,
      "step": 2140
    },
    {
      "epoch": 17.063808574277168,
      "grad_norm": 106.12614440917969,
      "learning_rate": 3.7383737726633313e-07,
      "loss": 21.3464,
      "step": 2150
    },
    {
      "epoch": 17.14356929212363,
      "grad_norm": 93.90852355957031,
      "learning_rate": 3.7274418463248636e-07,
      "loss": 24.7134,
      "step": 2160
    },
    {
      "epoch": 17.22333000997009,
      "grad_norm": 177.743896484375,
      "learning_rate": 3.7164789072761453e-07,
      "loss": 24.8612,
      "step": 2170
    },
    {
      "epoch": 17.30309072781655,
      "grad_norm": 204.92091369628906,
      "learning_rate": 3.7054852325082645e-07,
      "loss": 26.3242,
      "step": 2180
    },
    {
      "epoch": 17.38285144566301,
      "grad_norm": 44.62697219848633,
      "learning_rate": 3.6944610997888817e-07,
      "loss": 24.804,
      "step": 2190
    },
    {
      "epoch": 17.462612163509473,
      "grad_norm": 38.966094970703125,
      "learning_rate": 3.6834067876552117e-07,
      "loss": 26.5188,
      "step": 2200
    },
    {
      "epoch": 17.462612163509473,
      "eval_loss": 3.194345235824585,
      "eval_mae": 1.3451436758041382,
      "eval_mse": 3.194345235824585,
      "eval_r2": 0.0375593900680542,
      "eval_rmse": 1.7872731284906023,
      "eval_runtime": 9.2905,
      "eval_samples_per_second": 431.733,
      "eval_steps_per_second": 13.562,
      "step": 2200
    },
    {
      "epoch": 17.54237288135593,
      "grad_norm": 51.10791015625,
      "learning_rate": 3.6723225754069874e-07,
      "loss": 25.2638,
      "step": 2210
    },
    {
      "epoch": 17.622133599202392,
      "grad_norm": 140.89447021484375,
      "learning_rate": 3.6612087430994015e-07,
      "loss": 26.0489,
      "step": 2220
    },
    {
      "epoch": 17.701894317048854,
      "grad_norm": 90.82955932617188,
      "learning_rate": 3.650065571536031e-07,
      "loss": 26.7658,
      "step": 2230
    },
    {
      "epoch": 17.781655034895316,
      "grad_norm": 72.15864562988281,
      "learning_rate": 3.638893342261742e-07,
      "loss": 26.9759,
      "step": 2240
    },
    {
      "epoch": 17.861415752741774,
      "grad_norm": 208.88856506347656,
      "learning_rate": 3.6276923375555777e-07,
      "loss": 26.9895,
      "step": 2250
    },
    {
      "epoch": 17.941176470588236,
      "grad_norm": 80.82164001464844,
      "learning_rate": 3.6164628404236253e-07,
      "loss": 27.1941,
      "step": 2260
    },
    {
      "epoch": 18.015952143569294,
      "grad_norm": 28.38602638244629,
      "learning_rate": 3.605205134591864e-07,
      "loss": 24.3996,
      "step": 2270
    },
    {
      "epoch": 18.095712861415752,
      "grad_norm": 59.99515914916992,
      "learning_rate": 3.593919504498999e-07,
      "loss": 26.4284,
      "step": 2280
    },
    {
      "epoch": 18.175473579262214,
      "grad_norm": 51.018592834472656,
      "learning_rate": 3.582606235289273e-07,
      "loss": 26.4741,
      "step": 2290
    },
    {
      "epoch": 18.255234297108675,
      "grad_norm": 66.35299682617188,
      "learning_rate": 3.5712656128052634e-07,
      "loss": 25.902,
      "step": 2300
    },
    {
      "epoch": 18.255234297108675,
      "eval_loss": 3.1882619857788086,
      "eval_mae": 1.3470134735107422,
      "eval_mse": 3.1882619857788086,
      "eval_r2": 0.03939223289489746,
      "eval_rmse": 1.7855704930858396,
      "eval_runtime": 9.6616,
      "eval_samples_per_second": 415.149,
      "eval_steps_per_second": 13.041,
      "step": 2300
    },
    {
      "epoch": 18.334995014955133,
      "grad_norm": 100.36279296875,
      "learning_rate": 3.559897923580657e-07,
      "loss": 22.7732,
      "step": 2310
    },
    {
      "epoch": 18.414755732801595,
      "grad_norm": 51.83559036254883,
      "learning_rate": 3.548503454833015e-07,
      "loss": 26.4812,
      "step": 2320
    },
    {
      "epoch": 18.494516450648057,
      "grad_norm": 34.32014083862305,
      "learning_rate": 3.5370824944565115e-07,
      "loss": 25.8285,
      "step": 2330
    },
    {
      "epoch": 18.57427716849452,
      "grad_norm": 104.95442962646484,
      "learning_rate": 3.525635331014664e-07,
      "loss": 24.9534,
      "step": 2340
    },
    {
      "epoch": 18.654037886340976,
      "grad_norm": 62.482845306396484,
      "learning_rate": 3.5141622537330364e-07,
      "loss": 25.4369,
      "step": 2350
    },
    {
      "epoch": 18.733798604187438,
      "grad_norm": 203.57952880859375,
      "learning_rate": 3.502663552491939e-07,
      "loss": 25.9546,
      "step": 2360
    },
    {
      "epoch": 18.8135593220339,
      "grad_norm": 27.640735626220703,
      "learning_rate": 3.4911395178190995e-07,
      "loss": 25.982,
      "step": 2370
    },
    {
      "epoch": 18.893320039880358,
      "grad_norm": 98.8842544555664,
      "learning_rate": 3.4795904408823217e-07,
      "loss": 26.3763,
      "step": 2380
    },
    {
      "epoch": 18.97308075772682,
      "grad_norm": 144.08743286132812,
      "learning_rate": 3.468016613482132e-07,
      "loss": 26.1877,
      "step": 2390
    },
    {
      "epoch": 19.047856430707878,
      "grad_norm": 31.12162971496582,
      "learning_rate": 3.4564183280444063e-07,
      "loss": 25.5131,
      "step": 2400
    },
    {
      "epoch": 19.047856430707878,
      "eval_loss": 3.1840460300445557,
      "eval_mae": 1.3496177196502686,
      "eval_mse": 3.1840460300445557,
      "eval_r2": 0.04066246747970581,
      "eval_rmse": 1.7843895398831937,
      "eval_runtime": 9.6718,
      "eval_samples_per_second": 414.711,
      "eval_steps_per_second": 13.028,
      "step": 2400
    },
    {
      "epoch": 19.127617148554336,
      "grad_norm": 161.33517456054688,
      "learning_rate": 3.444795877612978e-07,
      "loss": 28.1758,
      "step": 2410
    },
    {
      "epoch": 19.207377866400797,
      "grad_norm": 78.51399230957031,
      "learning_rate": 3.4331495558422387e-07,
      "loss": 25.6685,
      "step": 2420
    },
    {
      "epoch": 19.28713858424726,
      "grad_norm": 49.65394592285156,
      "learning_rate": 3.4214796569897165e-07,
      "loss": 24.6879,
      "step": 2430
    },
    {
      "epoch": 19.366899302093717,
      "grad_norm": 77.83604431152344,
      "learning_rate": 3.4097864759086386e-07,
      "loss": 23.8251,
      "step": 2440
    },
    {
      "epoch": 19.44666001994018,
      "grad_norm": 97.642822265625,
      "learning_rate": 3.398070308040487e-07,
      "loss": 26.196,
      "step": 2450
    },
    {
      "epoch": 19.52642073778664,
      "grad_norm": 111.97957611083984,
      "learning_rate": 3.386331449407529e-07,
      "loss": 27.0219,
      "step": 2460
    },
    {
      "epoch": 19.606181455633102,
      "grad_norm": 43.619384765625,
      "learning_rate": 3.3745701966053407e-07,
      "loss": 26.6193,
      "step": 2470
    },
    {
      "epoch": 19.68594217347956,
      "grad_norm": 83.98363494873047,
      "learning_rate": 3.3627868467953137e-07,
      "loss": 23.5233,
      "step": 2480
    },
    {
      "epoch": 19.765702891326022,
      "grad_norm": 29.117719650268555,
      "learning_rate": 3.3509816976971436e-07,
      "loss": 24.8931,
      "step": 2490
    },
    {
      "epoch": 19.845463609172484,
      "grad_norm": 51.961605072021484,
      "learning_rate": 3.3391550475813087e-07,
      "loss": 24.3229,
      "step": 2500
    },
    {
      "epoch": 19.845463609172484,
      "eval_loss": 3.180414915084839,
      "eval_mae": 1.3414126634597778,
      "eval_mse": 3.180414915084839,
      "eval_r2": 0.041756510734558105,
      "eval_rmse": 1.7833717826311033,
      "eval_runtime": 9.2352,
      "eval_samples_per_second": 434.317,
      "eval_steps_per_second": 13.643,
      "step": 2500
    },
    {
      "epoch": 19.92522432701894,
      "grad_norm": 93.1182632446289,
      "learning_rate": 3.3273071952615374e-07,
      "loss": 27.2673,
      "step": 2510
    },
    {
      "epoch": 20.0,
      "grad_norm": 72.23329162597656,
      "learning_rate": 3.315438440087255e-07,
      "loss": 23.8389,
      "step": 2520
    },
    {
      "epoch": 20.07976071784646,
      "grad_norm": 141.3828887939453,
      "learning_rate": 3.30354908193602e-07,
      "loss": 26.6774,
      "step": 2530
    },
    {
      "epoch": 20.15952143569292,
      "grad_norm": 17.179548263549805,
      "learning_rate": 3.291639421205951e-07,
      "loss": 24.757,
      "step": 2540
    },
    {
      "epoch": 20.23928215353938,
      "grad_norm": 21.29429817199707,
      "learning_rate": 3.2797097588081324e-07,
      "loss": 26.0694,
      "step": 2550
    },
    {
      "epoch": 20.319042871385843,
      "grad_norm": 39.92965316772461,
      "learning_rate": 3.2677603961590126e-07,
      "loss": 26.8808,
      "step": 2560
    },
    {
      "epoch": 20.398803589232305,
      "grad_norm": 71.05562591552734,
      "learning_rate": 3.255791635172791e-07,
      "loss": 24.223,
      "step": 2570
    },
    {
      "epoch": 20.478564307078763,
      "grad_norm": 64.88768005371094,
      "learning_rate": 3.2438037782537875e-07,
      "loss": 26.9973,
      "step": 2580
    },
    {
      "epoch": 20.558325024925225,
      "grad_norm": 20.247940063476562,
      "learning_rate": 3.231797128288801e-07,
      "loss": 24.806,
      "step": 2590
    },
    {
      "epoch": 20.638085742771686,
      "grad_norm": 36.828975677490234,
      "learning_rate": 3.2197719886394616e-07,
      "loss": 25.024,
      "step": 2600
    },
    {
      "epoch": 20.638085742771686,
      "eval_loss": 3.1747312545776367,
      "eval_mae": 1.3484524488449097,
      "eval_mse": 3.1747310161590576,
      "eval_r2": 0.04346907138824463,
      "eval_rmse": 1.7817774878359693,
      "eval_runtime": 9.278,
      "eval_samples_per_second": 432.315,
      "eval_steps_per_second": 13.581,
      "step": 2600
    },
    {
      "epoch": 20.717846460618144,
      "grad_norm": 77.34175872802734,
      "learning_rate": 3.207728663134558e-07,
      "loss": 26.2727,
      "step": 2610
    },
    {
      "epoch": 20.797607178464606,
      "grad_norm": 35.301483154296875,
      "learning_rate": 3.1956674560623677e-07,
      "loss": 26.4504,
      "step": 2620
    },
    {
      "epoch": 20.877367896311068,
      "grad_norm": 107.94976043701172,
      "learning_rate": 3.183588672162965e-07,
      "loss": 25.7567,
      "step": 2630
    },
    {
      "epoch": 20.957128614157526,
      "grad_norm": 156.19349670410156,
      "learning_rate": 3.171492616620524e-07,
      "loss": 24.9266,
      "step": 2640
    },
    {
      "epoch": 21.031904287138584,
      "grad_norm": 52.308631896972656,
      "learning_rate": 3.159379595055605e-07,
      "loss": 23.7358,
      "step": 2650
    },
    {
      "epoch": 21.111665004985046,
      "grad_norm": 82.55281829833984,
      "learning_rate": 3.1472499135174356e-07,
      "loss": 25.9235,
      "step": 2660
    },
    {
      "epoch": 21.191425722831504,
      "grad_norm": 69.1916732788086,
      "learning_rate": 3.135103878476176e-07,
      "loss": 24.3697,
      "step": 2670
    },
    {
      "epoch": 21.271186440677965,
      "grad_norm": 46.42279815673828,
      "learning_rate": 3.122941796815175e-07,
      "loss": 27.117,
      "step": 2680
    },
    {
      "epoch": 21.350947158524427,
      "grad_norm": 140.99227905273438,
      "learning_rate": 3.110763975823219e-07,
      "loss": 24.7398,
      "step": 2690
    },
    {
      "epoch": 21.43070787637089,
      "grad_norm": 52.34705352783203,
      "learning_rate": 3.0985707231867647e-07,
      "loss": 25.9564,
      "step": 2700
    },
    {
      "epoch": 21.43070787637089,
      "eval_loss": 3.1698925495147705,
      "eval_mae": 1.3458428382873535,
      "eval_mse": 3.1698925495147705,
      "eval_r2": 0.04492682218551636,
      "eval_rmse": 1.7804192061182587,
      "eval_runtime": 9.6939,
      "eval_samples_per_second": 413.764,
      "eval_steps_per_second": 12.998,
      "step": 2700
    },
    {
      "epoch": 21.510468594217347,
      "grad_norm": 20.247802734375,
      "learning_rate": 3.08636234698217e-07,
      "loss": 26.2885,
      "step": 2710
    },
    {
      "epoch": 21.59022931206381,
      "grad_norm": 39.07258987426758,
      "learning_rate": 3.074139155667903e-07,
      "loss": 25.4281,
      "step": 2720
    },
    {
      "epoch": 21.66999002991027,
      "grad_norm": 34.06509017944336,
      "learning_rate": 3.061901458076755e-07,
      "loss": 26.7851,
      "step": 2730
    },
    {
      "epoch": 21.74975074775673,
      "grad_norm": 90.38104248046875,
      "learning_rate": 3.049649563408035e-07,
      "loss": 24.766,
      "step": 2740
    },
    {
      "epoch": 21.82951146560319,
      "grad_norm": 115.99075317382812,
      "learning_rate": 3.037383781219755e-07,
      "loss": 26.1506,
      "step": 2750
    },
    {
      "epoch": 21.90927218344965,
      "grad_norm": 36.51300811767578,
      "learning_rate": 3.0251044214208114e-07,
      "loss": 24.8899,
      "step": 2760
    },
    {
      "epoch": 21.989032901296113,
      "grad_norm": 138.48594665527344,
      "learning_rate": 3.012811794263157e-07,
      "loss": 25.5732,
      "step": 2770
    },
    {
      "epoch": 22.063808574277168,
      "grad_norm": 39.69606399536133,
      "learning_rate": 3.000506210333956e-07,
      "loss": 24.5779,
      "step": 2780
    },
    {
      "epoch": 22.14356929212363,
      "grad_norm": 130.00613403320312,
      "learning_rate": 2.988187980547744e-07,
      "loss": 24.9146,
      "step": 2790
    },
    {
      "epoch": 22.22333000997009,
      "grad_norm": 224.0641326904297,
      "learning_rate": 2.975857416138564e-07,
      "loss": 27.6134,
      "step": 2800
    },
    {
      "epoch": 22.22333000997009,
      "eval_loss": 3.166327476501465,
      "eval_mae": 1.3399311304092407,
      "eval_mse": 3.1663272380828857,
      "eval_r2": 0.04600107669830322,
      "eval_rmse": 1.7794176682507359,
      "eval_runtime": 9.242,
      "eval_samples_per_second": 433.999,
      "eval_steps_per_second": 13.633,
      "step": 2800
    },
    {
      "epoch": 22.30309072781655,
      "grad_norm": 52.53582000732422,
      "learning_rate": 2.9635148286521113e-07,
      "loss": 25.1255,
      "step": 2810
    },
    {
      "epoch": 22.38285144566301,
      "grad_norm": 132.7754364013672,
      "learning_rate": 2.9511605299378536e-07,
      "loss": 26.1952,
      "step": 2820
    },
    {
      "epoch": 22.462612163509473,
      "grad_norm": 38.92604446411133,
      "learning_rate": 2.93879483214116e-07,
      "loss": 23.0946,
      "step": 2830
    },
    {
      "epoch": 22.54237288135593,
      "grad_norm": 32.08353805541992,
      "learning_rate": 2.92641804769541e-07,
      "loss": 25.6264,
      "step": 2840
    },
    {
      "epoch": 22.622133599202392,
      "grad_norm": 93.6625747680664,
      "learning_rate": 2.914030489314098e-07,
      "loss": 24.4846,
      "step": 2850
    },
    {
      "epoch": 22.701894317048854,
      "grad_norm": 70.16485595703125,
      "learning_rate": 2.901632469982934e-07,
      "loss": 24.4818,
      "step": 2860
    },
    {
      "epoch": 22.781655034895316,
      "grad_norm": 36.66552734375,
      "learning_rate": 2.8892243029519373e-07,
      "loss": 26.8278,
      "step": 2870
    },
    {
      "epoch": 22.861415752741774,
      "grad_norm": 34.02045440673828,
      "learning_rate": 2.876806301727519e-07,
      "loss": 27.2244,
      "step": 2880
    },
    {
      "epoch": 22.941176470588236,
      "grad_norm": 136.84756469726562,
      "learning_rate": 2.864378780064564e-07,
      "loss": 27.2033,
      "step": 2890
    },
    {
      "epoch": 23.015952143569294,
      "grad_norm": 82.451171875,
      "learning_rate": 2.8519420519585004e-07,
      "loss": 21.5234,
      "step": 2900
    },
    {
      "epoch": 23.015952143569294,
      "eval_loss": 3.1617584228515625,
      "eval_mae": 1.339278221130371,
      "eval_mse": 3.1617584228515625,
      "eval_r2": 0.047377586364746094,
      "eval_rmse": 1.7781334097450514,
      "eval_runtime": 9.3218,
      "eval_samples_per_second": 430.281,
      "eval_steps_per_second": 13.517,
      "step": 2900
    },
    {
      "epoch": 23.095712861415752,
      "grad_norm": 115.3865966796875,
      "learning_rate": 2.839496431637367e-07,
      "loss": 25.5581,
      "step": 2910
    },
    {
      "epoch": 23.175473579262214,
      "grad_norm": 81.9321517944336,
      "learning_rate": 2.827042233553877e-07,
      "loss": 24.1154,
      "step": 2920
    },
    {
      "epoch": 23.255234297108675,
      "grad_norm": 226.57139587402344,
      "learning_rate": 2.814579772377467e-07,
      "loss": 26.7607,
      "step": 2930
    },
    {
      "epoch": 23.334995014955133,
      "grad_norm": 73.78599548339844,
      "learning_rate": 2.802109362986353e-07,
      "loss": 24.8509,
      "step": 2940
    },
    {
      "epoch": 23.414755732801595,
      "grad_norm": 77.19922637939453,
      "learning_rate": 2.7896313204595703e-07,
      "loss": 26.0284,
      "step": 2950
    },
    {
      "epoch": 23.494516450648057,
      "grad_norm": 218.84909057617188,
      "learning_rate": 2.777145960069015e-07,
      "loss": 26.3203,
      "step": 2960
    },
    {
      "epoch": 23.57427716849452,
      "grad_norm": 51.319217681884766,
      "learning_rate": 2.764653597271476e-07,
      "loss": 25.4633,
      "step": 2970
    },
    {
      "epoch": 23.654037886340976,
      "grad_norm": 65.45964050292969,
      "learning_rate": 2.752154547700667e-07,
      "loss": 26.4497,
      "step": 2980
    },
    {
      "epoch": 23.733798604187438,
      "grad_norm": 101.2373275756836,
      "learning_rate": 2.739649127159248e-07,
      "loss": 23.2934,
      "step": 2990
    },
    {
      "epoch": 23.8135593220339,
      "grad_norm": 92.46833801269531,
      "learning_rate": 2.7271376516108545e-07,
      "loss": 26.7957,
      "step": 3000
    },
    {
      "epoch": 23.8135593220339,
      "eval_loss": 3.157956838607788,
      "eval_mae": 1.3402751684188843,
      "eval_mse": 3.157956838607788,
      "eval_r2": 0.04852306842803955,
      "eval_rmse": 1.7770641064991965,
      "eval_runtime": 9.9161,
      "eval_samples_per_second": 404.494,
      "eval_steps_per_second": 12.707,
      "step": 3000
    },
    {
      "epoch": 23.893320039880358,
      "grad_norm": 82.04583740234375,
      "learning_rate": 2.7146204371721024e-07,
      "loss": 26.716,
      "step": 3010
    },
    {
      "epoch": 23.97308075772682,
      "grad_norm": 126.5661849975586,
      "learning_rate": 2.702097800104612e-07,
      "loss": 25.6039,
      "step": 3020
    },
    {
      "epoch": 24.047856430707878,
      "grad_norm": 124.41060638427734,
      "learning_rate": 2.6895700568070086e-07,
      "loss": 23.313,
      "step": 3030
    },
    {
      "epoch": 24.127617148554336,
      "grad_norm": 153.29127502441406,
      "learning_rate": 2.6770375238069356e-07,
      "loss": 26.2654,
      "step": 3040
    },
    {
      "epoch": 24.207377866400797,
      "grad_norm": 74.55110168457031,
      "learning_rate": 2.664500517753049e-07,
      "loss": 25.8805,
      "step": 3050
    },
    {
      "epoch": 24.28713858424726,
      "grad_norm": 190.80921936035156,
      "learning_rate": 2.6519593554070273e-07,
      "loss": 26.5513,
      "step": 3060
    },
    {
      "epoch": 24.366899302093717,
      "grad_norm": 38.84217071533203,
      "learning_rate": 2.6394143536355595e-07,
      "loss": 25.2968,
      "step": 3070
    },
    {
      "epoch": 24.44666001994018,
      "grad_norm": 30.24015998840332,
      "learning_rate": 2.626865829402341e-07,
      "loss": 25.6536,
      "step": 3080
    },
    {
      "epoch": 24.52642073778664,
      "grad_norm": 68.18790435791016,
      "learning_rate": 2.614314099760068e-07,
      "loss": 26.5879,
      "step": 3090
    },
    {
      "epoch": 24.606181455633102,
      "grad_norm": 130.46934509277344,
      "learning_rate": 2.601759481842426e-07,
      "loss": 25.9434,
      "step": 3100
    },
    {
      "epoch": 24.606181455633102,
      "eval_loss": 3.155501127243042,
      "eval_mae": 1.3429107666015625,
      "eval_mse": 3.155501127243042,
      "eval_r2": 0.049262940883636475,
      "eval_rmse": 1.7763730259275619,
      "eval_runtime": 9.2949,
      "eval_samples_per_second": 431.528,
      "eval_steps_per_second": 13.556,
      "step": 3100
    },
    {
      "epoch": 24.68594217347956,
      "grad_norm": 145.48199462890625,
      "learning_rate": 2.5892022928560715e-07,
      "loss": 26.4445,
      "step": 3110
    },
    {
      "epoch": 24.765702891326022,
      "grad_norm": 65.02415466308594,
      "learning_rate": 2.576642850072627e-07,
      "loss": 26.3289,
      "step": 3120
    },
    {
      "epoch": 24.845463609172484,
      "grad_norm": 83.2958984375,
      "learning_rate": 2.564081470820657e-07,
      "loss": 24.1421,
      "step": 3130
    },
    {
      "epoch": 24.92522432701894,
      "grad_norm": 28.578340530395508,
      "learning_rate": 2.551518472477654e-07,
      "loss": 24.9301,
      "step": 3140
    },
    {
      "epoch": 25.0,
      "grad_norm": 8.585793495178223,
      "learning_rate": 2.5389541724620184e-07,
      "loss": 21.1339,
      "step": 3150
    },
    {
      "epoch": 25.07976071784646,
      "grad_norm": 77.31856536865234,
      "learning_rate": 2.526388888225039e-07,
      "loss": 23.9865,
      "step": 3160
    },
    {
      "epoch": 25.15952143569292,
      "grad_norm": 150.84129333496094,
      "learning_rate": 2.5138229372428715e-07,
      "loss": 28.0155,
      "step": 3170
    },
    {
      "epoch": 25.23928215353938,
      "grad_norm": 23.556215286254883,
      "learning_rate": 2.5012566370085186e-07,
      "loss": 24.9378,
      "step": 3180
    },
    {
      "epoch": 25.319042871385843,
      "grad_norm": 183.71372985839844,
      "learning_rate": 2.4886903050238065e-07,
      "loss": 25.0401,
      "step": 3190
    },
    {
      "epoch": 25.398803589232305,
      "grad_norm": 242.62158203125,
      "learning_rate": 2.4761242587913633e-07,
      "loss": 26.1274,
      "step": 3200
    },
    {
      "epoch": 25.398803589232305,
      "eval_loss": 3.1510255336761475,
      "eval_mae": 1.340527892112732,
      "eval_mse": 3.1510257720947266,
      "eval_r2": 0.05061131715774536,
      "eval_rmse": 1.775112889957911,
      "eval_runtime": 9.3595,
      "eval_samples_per_second": 428.548,
      "eval_steps_per_second": 13.462,
      "step": 3200
    },
    {
      "epoch": 25.478564307078763,
      "grad_norm": 49.5938835144043,
      "learning_rate": 2.4635588158065996e-07,
      "loss": 24.9653,
      "step": 3210
    },
    {
      "epoch": 25.558325024925225,
      "grad_norm": 72.53789520263672,
      "learning_rate": 2.450994293549681e-07,
      "loss": 24.6724,
      "step": 3220
    },
    {
      "epoch": 25.638085742771686,
      "grad_norm": 61.71761703491211,
      "learning_rate": 2.438431009477512e-07,
      "loss": 25.9375,
      "step": 3230
    },
    {
      "epoch": 25.717846460618144,
      "grad_norm": 119.82528686523438,
      "learning_rate": 2.425869281015711e-07,
      "loss": 25.5312,
      "step": 3240
    },
    {
      "epoch": 25.797607178464606,
      "grad_norm": 134.8277587890625,
      "learning_rate": 2.413309425550595e-07,
      "loss": 27.2442,
      "step": 3250
    },
    {
      "epoch": 25.877367896311068,
      "grad_norm": 87.44007110595703,
      "learning_rate": 2.400751760421156e-07,
      "loss": 25.8056,
      "step": 3260
    },
    {
      "epoch": 25.957128614157526,
      "grad_norm": 62.196311950683594,
      "learning_rate": 2.3881966029110437e-07,
      "loss": 26.1153,
      "step": 3270
    },
    {
      "epoch": 26.031904287138584,
      "grad_norm": 125.53699493408203,
      "learning_rate": 2.3756442702405522e-07,
      "loss": 22.7786,
      "step": 3280
    },
    {
      "epoch": 26.111665004985046,
      "grad_norm": 97.53251647949219,
      "learning_rate": 2.3630950795585997e-07,
      "loss": 25.4978,
      "step": 3290
    },
    {
      "epoch": 26.191425722831504,
      "grad_norm": 307.3661804199219,
      "learning_rate": 2.350549347934723e-07,
      "loss": 26.4714,
      "step": 3300
    },
    {
      "epoch": 26.191425722831504,
      "eval_loss": 3.1477982997894287,
      "eval_mae": 1.3374989032745361,
      "eval_mse": 3.1477982997894287,
      "eval_r2": 0.051583707332611084,
      "eval_rmse": 1.7742035677422783,
      "eval_runtime": 9.9848,
      "eval_samples_per_second": 401.71,
      "eval_steps_per_second": 12.619,
      "step": 3300
    },
    {
      "epoch": 26.271186440677965,
      "grad_norm": 195.30360412597656,
      "learning_rate": 2.338007392351057e-07,
      "loss": 26.7384,
      "step": 3310
    },
    {
      "epoch": 26.350947158524427,
      "grad_norm": 118.61621856689453,
      "learning_rate": 2.3254695296943337e-07,
      "loss": 26.8368,
      "step": 3320
    },
    {
      "epoch": 26.43070787637089,
      "grad_norm": 67.60140991210938,
      "learning_rate": 2.3129360767478738e-07,
      "loss": 24.9358,
      "step": 3330
    },
    {
      "epoch": 26.510468594217347,
      "grad_norm": 60.42992401123047,
      "learning_rate": 2.300407350183577e-07,
      "loss": 23.5213,
      "step": 3340
    },
    {
      "epoch": 26.59022931206381,
      "grad_norm": 64.20198822021484,
      "learning_rate": 2.2878836665539304e-07,
      "loss": 25.2233,
      "step": 3350
    },
    {
      "epoch": 26.66999002991027,
      "grad_norm": 130.40182495117188,
      "learning_rate": 2.2753653422840017e-07,
      "loss": 23.0435,
      "step": 3360
    },
    {
      "epoch": 26.74975074775673,
      "grad_norm": 171.5955352783203,
      "learning_rate": 2.2628526936634513e-07,
      "loss": 27.6001,
      "step": 3370
    },
    {
      "epoch": 26.82951146560319,
      "grad_norm": 43.720558166503906,
      "learning_rate": 2.2503460368385376e-07,
      "loss": 25.7998,
      "step": 3380
    },
    {
      "epoch": 26.90927218344965,
      "grad_norm": 55.378265380859375,
      "learning_rate": 2.2378456878041273e-07,
      "loss": 26.9747,
      "step": 3390
    },
    {
      "epoch": 26.989032901296113,
      "grad_norm": 60.05023193359375,
      "learning_rate": 2.225351962395716e-07,
      "loss": 22.6912,
      "step": 3400
    },
    {
      "epoch": 26.989032901296113,
      "eval_loss": 3.1460859775543213,
      "eval_mae": 1.3359055519104004,
      "eval_mse": 3.146085739135742,
      "eval_r2": 0.05209970474243164,
      "eval_rmse": 1.773720874076793,
      "eval_runtime": 9.3107,
      "eval_samples_per_second": 430.797,
      "eval_steps_per_second": 13.533,
      "step": 3400
    },
    {
      "epoch": 27.063808574277168,
      "grad_norm": 164.95831298828125,
      "learning_rate": 2.2128651762814442e-07,
      "loss": 25.3825,
      "step": 3410
    },
    {
      "epoch": 27.14356929212363,
      "grad_norm": 133.99693298339844,
      "learning_rate": 2.2003856449541235e-07,
      "loss": 25.5549,
      "step": 3420
    },
    {
      "epoch": 27.22333000997009,
      "grad_norm": 98.84943389892578,
      "learning_rate": 2.1879136837232675e-07,
      "loss": 26.5847,
      "step": 3430
    },
    {
      "epoch": 27.30309072781655,
      "grad_norm": 142.6348114013672,
      "learning_rate": 2.1754496077071186e-07,
      "loss": 26.7426,
      "step": 3440
    },
    {
      "epoch": 27.38285144566301,
      "grad_norm": 79.17008972167969,
      "learning_rate": 2.162993731824694e-07,
      "loss": 27.3471,
      "step": 3450
    },
    {
      "epoch": 27.462612163509473,
      "grad_norm": 51.78144836425781,
      "learning_rate": 2.150546370787823e-07,
      "loss": 23.8934,
      "step": 3460
    },
    {
      "epoch": 27.54237288135593,
      "grad_norm": 250.04104614257812,
      "learning_rate": 2.1381078390931987e-07,
      "loss": 24.5301,
      "step": 3470
    },
    {
      "epoch": 27.622133599202392,
      "grad_norm": 111.69416046142578,
      "learning_rate": 2.1256784510144288e-07,
      "loss": 23.3404,
      "step": 3480
    },
    {
      "epoch": 27.701894317048854,
      "grad_norm": 27.399375915527344,
      "learning_rate": 2.1132585205941006e-07,
      "loss": 25.2206,
      "step": 3490
    },
    {
      "epoch": 27.781655034895316,
      "grad_norm": 111.54031372070312,
      "learning_rate": 2.1008483616358422e-07,
      "loss": 23.6541,
      "step": 3500
    },
    {
      "epoch": 27.781655034895316,
      "eval_loss": 3.142723560333252,
      "eval_mae": 1.3382033109664917,
      "eval_mse": 3.142723798751831,
      "eval_r2": 0.053112685680389404,
      "eval_rmse": 1.7727729123471598,
      "eval_runtime": 9.9648,
      "eval_samples_per_second": 402.517,
      "eval_steps_per_second": 12.645,
      "step": 3500
    },
    {
      "epoch": 27.861415752741774,
      "grad_norm": 124.3917465209961,
      "learning_rate": 2.0884482876963915e-07,
      "loss": 26.0723,
      "step": 3510
    },
    {
      "epoch": 27.941176470588236,
      "grad_norm": 86.09922790527344,
      "learning_rate": 2.07605861207768e-07,
      "loss": 25.5361,
      "step": 3520
    },
    {
      "epoch": 28.015952143569294,
      "grad_norm": 218.16212463378906,
      "learning_rate": 2.063679647818913e-07,
      "loss": 25.9154,
      "step": 3530
    },
    {
      "epoch": 28.095712861415752,
      "grad_norm": 134.55062866210938,
      "learning_rate": 2.0513117076886604e-07,
      "loss": 26.8341,
      "step": 3540
    },
    {
      "epoch": 28.175473579262214,
      "grad_norm": 173.00433349609375,
      "learning_rate": 2.0389551041769575e-07,
      "loss": 23.9771,
      "step": 3550
    },
    {
      "epoch": 28.255234297108675,
      "grad_norm": 102.30731964111328,
      "learning_rate": 2.026610149487403e-07,
      "loss": 24.7714,
      "step": 3560
    },
    {
      "epoch": 28.334995014955133,
      "grad_norm": 58.39441680908203,
      "learning_rate": 2.0142771555292793e-07,
      "loss": 27.4694,
      "step": 3570
    },
    {
      "epoch": 28.414755732801595,
      "grad_norm": 72.3934326171875,
      "learning_rate": 2.0019564339096635e-07,
      "loss": 25.1937,
      "step": 3580
    },
    {
      "epoch": 28.494516450648057,
      "grad_norm": 81.90557098388672,
      "learning_rate": 1.9896482959255598e-07,
      "loss": 26.5238,
      "step": 3590
    },
    {
      "epoch": 28.57427716849452,
      "grad_norm": 78.32896423339844,
      "learning_rate": 1.9773530525560345e-07,
      "loss": 22.4524,
      "step": 3600
    },
    {
      "epoch": 28.57427716849452,
      "eval_loss": 3.140214443206787,
      "eval_mae": 1.3380094766616821,
      "eval_mse": 3.140214443206787,
      "eval_r2": 0.05386871099472046,
      "eval_rmse": 1.7720650222852397,
      "eval_runtime": 9.2577,
      "eval_samples_per_second": 433.26,
      "eval_steps_per_second": 13.61,
      "step": 3600
    },
    {
      "epoch": 28.654037886340976,
      "grad_norm": 238.5843963623047,
      "learning_rate": 1.9650710144543527e-07,
      "loss": 26.2189,
      "step": 3610
    },
    {
      "epoch": 28.733798604187438,
      "grad_norm": 152.3238067626953,
      "learning_rate": 1.952802491940138e-07,
      "loss": 25.3015,
      "step": 3620
    },
    {
      "epoch": 28.8135593220339,
      "grad_norm": 68.33666229248047,
      "learning_rate": 1.9405477949915233e-07,
      "loss": 23.9232,
      "step": 3630
    },
    {
      "epoch": 28.893320039880358,
      "grad_norm": 164.30014038085938,
      "learning_rate": 1.9283072332373251e-07,
      "loss": 25.088,
      "step": 3640
    },
    {
      "epoch": 28.97308075772682,
      "grad_norm": 45.72752380371094,
      "learning_rate": 1.9160811159492183e-07,
      "loss": 26.5676,
      "step": 3650
    },
    {
      "epoch": 29.047856430707878,
      "grad_norm": 187.55999755859375,
      "learning_rate": 1.90386975203392e-07,
      "loss": 22.3533,
      "step": 3660
    },
    {
      "epoch": 29.127617148554336,
      "grad_norm": 90.50538635253906,
      "learning_rate": 1.8916734500253906e-07,
      "loss": 25.7678,
      "step": 3670
    },
    {
      "epoch": 29.207377866400797,
      "grad_norm": 135.21221923828125,
      "learning_rate": 1.8794925180770287e-07,
      "loss": 25.1531,
      "step": 3680
    },
    {
      "epoch": 29.28713858424726,
      "grad_norm": 108.64622497558594,
      "learning_rate": 1.8673272639538973e-07,
      "loss": 24.7013,
      "step": 3690
    },
    {
      "epoch": 29.366899302093717,
      "grad_norm": 94.64742279052734,
      "learning_rate": 1.8551779950249353e-07,
      "loss": 27.9527,
      "step": 3700
    },
    {
      "epoch": 29.366899302093717,
      "eval_loss": 3.140504837036133,
      "eval_mae": 1.331715703010559,
      "eval_mse": 3.140505075454712,
      "eval_r2": 0.053781211376190186,
      "eval_rmse": 1.7721470242208213,
      "eval_runtime": 9.3146,
      "eval_samples_per_second": 430.615,
      "eval_steps_per_second": 13.527,
      "step": 3700
    },
    {
      "epoch": 29.44666001994018,
      "grad_norm": 78.31409454345703,
      "learning_rate": 1.843045018255201e-07,
      "loss": 26.9881,
      "step": 3710
    },
    {
      "epoch": 29.52642073778664,
      "grad_norm": 194.01712036132812,
      "learning_rate": 1.8309286401981133e-07,
      "loss": 24.7781,
      "step": 3720
    },
    {
      "epoch": 29.606181455633102,
      "grad_norm": 154.0265655517578,
      "learning_rate": 1.8188291669877026e-07,
      "loss": 25.0849,
      "step": 3730
    },
    {
      "epoch": 29.68594217347956,
      "grad_norm": 32.36880874633789,
      "learning_rate": 1.8067469043308835e-07,
      "loss": 23.4199,
      "step": 3740
    },
    {
      "epoch": 29.765702891326022,
      "grad_norm": 122.14185333251953,
      "learning_rate": 1.7946821574997228e-07,
      "loss": 26.3242,
      "step": 3750
    },
    {
      "epoch": 29.845463609172484,
      "grad_norm": 159.14520263671875,
      "learning_rate": 1.7826352313237315e-07,
      "loss": 25.3263,
      "step": 3760
    },
    {
      "epoch": 29.92522432701894,
      "grad_norm": 118.12110137939453,
      "learning_rate": 1.770606430182161e-07,
      "loss": 24.5008,
      "step": 3770
    },
    {
      "epoch": 30.0,
      "grad_norm": 62.7591438293457,
      "learning_rate": 1.758596057996314e-07,
      "loss": 25.0837,
      "step": 3780
    },
    {
      "epoch": 30.07976071784646,
      "grad_norm": 43.26075744628906,
      "learning_rate": 1.746604418221866e-07,
      "loss": 26.6551,
      "step": 3790
    },
    {
      "epoch": 30.15952143569292,
      "grad_norm": 110.1336898803711,
      "learning_rate": 1.7346318138411922e-07,
      "loss": 24.0462,
      "step": 3800
    },
    {
      "epoch": 30.15952143569292,
      "eval_loss": 3.136179208755493,
      "eval_mae": 1.335479974746704,
      "eval_mse": 3.136179208755493,
      "eval_r2": 0.05508452653884888,
      "eval_rmse": 1.7709260878860793,
      "eval_runtime": 9.9855,
      "eval_samples_per_second": 401.681,
      "eval_steps_per_second": 12.618,
      "step": 3800
    },
    {
      "epoch": 30.23928215353938,
      "grad_norm": 49.133968353271484,
      "learning_rate": 1.7226785473557232e-07,
      "loss": 25.7876,
      "step": 3810
    },
    {
      "epoch": 30.319042871385843,
      "grad_norm": 46.56238555908203,
      "learning_rate": 1.71074492077829e-07,
      "loss": 26.4264,
      "step": 3820
    },
    {
      "epoch": 30.398803589232305,
      "grad_norm": 52.70458221435547,
      "learning_rate": 1.6988312356255024e-07,
      "loss": 27.3161,
      "step": 3830
    },
    {
      "epoch": 30.478564307078763,
      "grad_norm": 45.012142181396484,
      "learning_rate": 1.6869377929101281e-07,
      "loss": 24.2407,
      "step": 3840
    },
    {
      "epoch": 30.558325024925225,
      "grad_norm": 208.66177368164062,
      "learning_rate": 1.6750648931334826e-07,
      "loss": 25.4931,
      "step": 3850
    },
    {
      "epoch": 30.638085742771686,
      "grad_norm": 68.44234466552734,
      "learning_rate": 1.6632128362778463e-07,
      "loss": 23.7536,
      "step": 3860
    },
    {
      "epoch": 30.717846460618144,
      "grad_norm": 161.40869140625,
      "learning_rate": 1.6513819217988745e-07,
      "loss": 25.3655,
      "step": 3870
    },
    {
      "epoch": 30.797607178464606,
      "grad_norm": 251.0177764892578,
      "learning_rate": 1.63957244861804e-07,
      "loss": 24.2239,
      "step": 3880
    },
    {
      "epoch": 30.877367896311068,
      "grad_norm": 60.6568489074707,
      "learning_rate": 1.6277847151150752e-07,
      "loss": 25.3857,
      "step": 3890
    },
    {
      "epoch": 30.957128614157526,
      "grad_norm": 103.65233612060547,
      "learning_rate": 1.6160190191204342e-07,
      "loss": 26.6993,
      "step": 3900
    },
    {
      "epoch": 30.957128614157526,
      "eval_loss": 3.1340854167938232,
      "eval_mae": 1.3334490060806274,
      "eval_mse": 3.134085178375244,
      "eval_r2": 0.05571550130844116,
      "eval_rmse": 1.770334764493779,
      "eval_runtime": 9.2742,
      "eval_samples_per_second": 432.49,
      "eval_steps_per_second": 13.586,
      "step": 3900
    },
    {
      "epoch": 31.031904287138584,
      "grad_norm": 233.1190185546875,
      "learning_rate": 1.604275657907771e-07,
      "loss": 22.8734,
      "step": 3910
    },
    {
      "epoch": 31.111665004985046,
      "grad_norm": 189.4170684814453,
      "learning_rate": 1.5925549281864225e-07,
      "loss": 24.9487,
      "step": 3920
    },
    {
      "epoch": 31.191425722831504,
      "grad_norm": 67.26908874511719,
      "learning_rate": 1.5808571260939192e-07,
      "loss": 25.9776,
      "step": 3930
    },
    {
      "epoch": 31.271186440677965,
      "grad_norm": 58.05997848510742,
      "learning_rate": 1.5691825471884957e-07,
      "loss": 26.7025,
      "step": 3940
    },
    {
      "epoch": 31.350947158524427,
      "grad_norm": 31.262191772460938,
      "learning_rate": 1.5575314864416272e-07,
      "loss": 23.673,
      "step": 3950
    },
    {
      "epoch": 31.43070787637089,
      "grad_norm": 154.3089141845703,
      "learning_rate": 1.5459042382305793e-07,
      "loss": 27.3652,
      "step": 3960
    },
    {
      "epoch": 31.510468594217347,
      "grad_norm": 48.65864181518555,
      "learning_rate": 1.5343010963309608e-07,
      "loss": 25.3723,
      "step": 3970
    },
    {
      "epoch": 31.59022931206381,
      "grad_norm": 122.0504150390625,
      "learning_rate": 1.5227223539093136e-07,
      "loss": 24.5209,
      "step": 3980
    },
    {
      "epoch": 31.66999002991027,
      "grad_norm": 42.498260498046875,
      "learning_rate": 1.5111683035156937e-07,
      "loss": 23.8867,
      "step": 3990
    },
    {
      "epoch": 31.74975074775673,
      "grad_norm": 65.39444732666016,
      "learning_rate": 1.4996392370762884e-07,
      "loss": 25.3385,
      "step": 4000
    },
    {
      "epoch": 31.74975074775673,
      "eval_loss": 3.13254976272583,
      "eval_mae": 1.3360387086868286,
      "eval_mse": 3.132550001144409,
      "eval_r2": 0.05617797374725342,
      "eval_rmse": 1.7699011275052652,
      "eval_runtime": 9.3296,
      "eval_samples_per_second": 429.924,
      "eval_steps_per_second": 13.505,
      "step": 4000
    },
    {
      "epoch": 31.82951146560319,
      "grad_norm": 39.13220977783203,
      "learning_rate": 1.4881354458860368e-07,
      "loss": 26.1535,
      "step": 4010
    },
    {
      "epoch": 31.90927218344965,
      "grad_norm": 89.6006088256836,
      "learning_rate": 1.4766572206012676e-07,
      "loss": 25.305,
      "step": 4020
    },
    {
      "epoch": 31.989032901296113,
      "grad_norm": 115.49569702148438,
      "learning_rate": 1.4652048512323618e-07,
      "loss": 26.6158,
      "step": 4030
    },
    {
      "epoch": 32.06380857427717,
      "grad_norm": 71.1096420288086,
      "learning_rate": 1.4537786271364166e-07,
      "loss": 25.0504,
      "step": 4040
    },
    {
      "epoch": 32.14356929212363,
      "grad_norm": 67.03511047363281,
      "learning_rate": 1.4423788370099443e-07,
      "loss": 26.0869,
      "step": 4050
    },
    {
      "epoch": 32.22333000997009,
      "grad_norm": 91.05354309082031,
      "learning_rate": 1.43100576888157e-07,
      "loss": 24.8662,
      "step": 4060
    },
    {
      "epoch": 32.30309072781655,
      "grad_norm": 88.30418395996094,
      "learning_rate": 1.4196597101047572e-07,
      "loss": 25.5,
      "step": 4070
    },
    {
      "epoch": 32.38285144566301,
      "grad_norm": 133.8420867919922,
      "learning_rate": 1.4083409473505493e-07,
      "loss": 26.1354,
      "step": 4080
    },
    {
      "epoch": 32.46261216350947,
      "grad_norm": 126.43040466308594,
      "learning_rate": 1.3970497666003225e-07,
      "loss": 23.7797,
      "step": 4090
    },
    {
      "epoch": 32.54237288135593,
      "grad_norm": 273.7431335449219,
      "learning_rate": 1.3857864531385638e-07,
      "loss": 23.6945,
      "step": 4100
    },
    {
      "epoch": 32.54237288135593,
      "eval_loss": 3.132538318634033,
      "eval_mae": 1.3303931951522827,
      "eval_mse": 3.132538318634033,
      "eval_r2": 0.056181490421295166,
      "eval_rmse": 1.7698978271736572,
      "eval_runtime": 9.9463,
      "eval_samples_per_second": 403.265,
      "eval_steps_per_second": 12.668,
      "step": 4100
    },
    {
      "epoch": 32.62213359920239,
      "grad_norm": 40.9385871887207,
      "learning_rate": 1.37455129154566e-07,
      "loss": 24.5797,
      "step": 4110
    },
    {
      "epoch": 32.701894317048854,
      "grad_norm": 126.75995635986328,
      "learning_rate": 1.363344565690711e-07,
      "loss": 26.4947,
      "step": 4120
    },
    {
      "epoch": 32.781655034895316,
      "grad_norm": 218.6414794921875,
      "learning_rate": 1.3521665587243563e-07,
      "loss": 24.3611,
      "step": 4130
    },
    {
      "epoch": 32.86141575274178,
      "grad_norm": 29.135433197021484,
      "learning_rate": 1.3410175530716163e-07,
      "loss": 26.2829,
      "step": 4140
    },
    {
      "epoch": 32.94117647058823,
      "grad_norm": 88.4239501953125,
      "learning_rate": 1.3298978304247638e-07,
      "loss": 25.2017,
      "step": 4150
    },
    {
      "epoch": 33.015952143569294,
      "grad_norm": 73.62918090820312,
      "learning_rate": 1.318807671736201e-07,
      "loss": 25.3726,
      "step": 4160
    },
    {
      "epoch": 33.095712861415755,
      "grad_norm": 112.5821304321289,
      "learning_rate": 1.3077473572113645e-07,
      "loss": 26.48,
      "step": 4170
    },
    {
      "epoch": 33.17547357926221,
      "grad_norm": 121.62039947509766,
      "learning_rate": 1.2967171663016456e-07,
      "loss": 23.0178,
      "step": 4180
    },
    {
      "epoch": 33.25523429710867,
      "grad_norm": 65.80647277832031,
      "learning_rate": 1.285717377697324e-07,
      "loss": 25.4537,
      "step": 4190
    },
    {
      "epoch": 33.33499501495513,
      "grad_norm": 264.9825134277344,
      "learning_rate": 1.2747482693205356e-07,
      "loss": 26.105,
      "step": 4200
    },
    {
      "epoch": 33.33499501495513,
      "eval_loss": 3.1293115615844727,
      "eval_mae": 1.3327357769012451,
      "eval_mse": 3.1293118000030518,
      "eval_r2": 0.05715364217758179,
      "eval_rmse": 1.7689860937845305,
      "eval_runtime": 9.2653,
      "eval_samples_per_second": 432.905,
      "eval_steps_per_second": 13.599,
      "step": 4200
    },
    {
      "epoch": 33.414755732801595,
      "grad_norm": 54.90214157104492,
      "learning_rate": 1.263810118318242e-07,
      "loss": 25.4921,
      "step": 4210
    },
    {
      "epoch": 33.49451645064806,
      "grad_norm": 65.57184600830078,
      "learning_rate": 1.252903201055232e-07,
      "loss": 24.6091,
      "step": 4220
    },
    {
      "epoch": 33.57427716849452,
      "grad_norm": 46.68421936035156,
      "learning_rate": 1.242027793107142e-07,
      "loss": 26.6366,
      "step": 4230
    },
    {
      "epoch": 33.65403788634098,
      "grad_norm": 157.0302734375,
      "learning_rate": 1.2311841692534843e-07,
      "loss": 26.8544,
      "step": 4240
    },
    {
      "epoch": 33.733798604187434,
      "grad_norm": 58.255130767822266,
      "learning_rate": 1.2203726034707158e-07,
      "loss": 26.418,
      "step": 4250
    },
    {
      "epoch": 33.813559322033896,
      "grad_norm": 99.26273345947266,
      "learning_rate": 1.2095933689253047e-07,
      "loss": 24.3008,
      "step": 4260
    },
    {
      "epoch": 33.89332003988036,
      "grad_norm": 158.2358856201172,
      "learning_rate": 1.198846737966838e-07,
      "loss": 25.159,
      "step": 4270
    },
    {
      "epoch": 33.97308075772682,
      "grad_norm": 197.09414672851562,
      "learning_rate": 1.1881329821211325e-07,
      "loss": 23.8238,
      "step": 4280
    },
    {
      "epoch": 34.047856430707874,
      "grad_norm": 129.14163208007812,
      "learning_rate": 1.1774523720833806e-07,
      "loss": 23.8336,
      "step": 4290
    },
    {
      "epoch": 34.127617148554336,
      "grad_norm": 72.48125457763672,
      "learning_rate": 1.1668051777113089e-07,
      "loss": 26.3832,
      "step": 4300
    },
    {
      "epoch": 34.127617148554336,
      "eval_loss": 3.128072738647461,
      "eval_mae": 1.3328605890274048,
      "eval_mse": 3.128072500228882,
      "eval_r2": 0.057527005672454834,
      "eval_rmse": 1.7686357737614835,
      "eval_runtime": 9.3433,
      "eval_samples_per_second": 429.292,
      "eval_steps_per_second": 13.486,
      "step": 4300
    },
    {
      "epoch": 34.2073778664008,
      "grad_norm": 198.8756561279297,
      "learning_rate": 1.1561916680183554e-07,
      "loss": 24.7351,
      "step": 4310
    },
    {
      "epoch": 34.28713858424726,
      "grad_norm": 112.67052459716797,
      "learning_rate": 1.1456121111668809e-07,
      "loss": 27.0009,
      "step": 4320
    },
    {
      "epoch": 34.36689930209372,
      "grad_norm": 48.918434143066406,
      "learning_rate": 1.1350667744613868e-07,
      "loss": 25.6622,
      "step": 4330
    },
    {
      "epoch": 34.44666001994018,
      "grad_norm": 70.8290786743164,
      "learning_rate": 1.1245559243417624e-07,
      "loss": 25.7665,
      "step": 4340
    },
    {
      "epoch": 34.52642073778664,
      "grad_norm": 154.0522003173828,
      "learning_rate": 1.1140798263765583e-07,
      "loss": 25.922,
      "step": 4350
    },
    {
      "epoch": 34.6061814556331,
      "grad_norm": 102.49356842041016,
      "learning_rate": 1.1036387452562681e-07,
      "loss": 26.0428,
      "step": 4360
    },
    {
      "epoch": 34.68594217347956,
      "grad_norm": 101.10346221923828,
      "learning_rate": 1.0932329447866495e-07,
      "loss": 24.2896,
      "step": 4370
    },
    {
      "epoch": 34.76570289132602,
      "grad_norm": 101.76194763183594,
      "learning_rate": 1.0828626878820501e-07,
      "loss": 24.0958,
      "step": 4380
    },
    {
      "epoch": 34.845463609172484,
      "grad_norm": 187.35472106933594,
      "learning_rate": 1.0725282365587729e-07,
      "loss": 24.7648,
      "step": 4390
    },
    {
      "epoch": 34.925224327018945,
      "grad_norm": 77.70819854736328,
      "learning_rate": 1.062229851928449e-07,
      "loss": 24.5196,
      "step": 4400
    },
    {
      "epoch": 34.925224327018945,
      "eval_loss": 3.127197504043579,
      "eval_mae": 1.3322991132736206,
      "eval_mse": 3.127197504043579,
      "eval_r2": 0.05779063701629639,
      "eval_rmse": 1.7683883917407903,
      "eval_runtime": 9.9349,
      "eval_samples_per_second": 403.728,
      "eval_steps_per_second": 12.683,
      "step": 4400
    },
    {
      "epoch": 35.0,
      "grad_norm": 67.70870208740234,
      "learning_rate": 1.051967794191446e-07,
      "loss": 24.2208,
      "step": 4410
    },
    {
      "epoch": 35.07976071784646,
      "grad_norm": 219.30227661132812,
      "learning_rate": 1.041742322630291e-07,
      "loss": 27.2756,
      "step": 4420
    },
    {
      "epoch": 35.15952143569292,
      "grad_norm": 84.1348876953125,
      "learning_rate": 1.0315536956031182e-07,
      "loss": 25.4429,
      "step": 4430
    },
    {
      "epoch": 35.239282153539385,
      "grad_norm": 99.25320434570312,
      "learning_rate": 1.0214021705371453e-07,
      "loss": 23.5892,
      "step": 4440
    },
    {
      "epoch": 35.31904287138584,
      "grad_norm": 42.8789176940918,
      "learning_rate": 1.0112880039221652e-07,
      "loss": 27.1501,
      "step": 4450
    },
    {
      "epoch": 35.3988035892323,
      "grad_norm": 52.440799713134766,
      "learning_rate": 1.0012114513040665e-07,
      "loss": 24.3347,
      "step": 4460
    },
    {
      "epoch": 35.47856430707876,
      "grad_norm": 59.711448669433594,
      "learning_rate": 9.911727672783802e-08,
      "loss": 24.4259,
      "step": 4470
    },
    {
      "epoch": 35.558325024925225,
      "grad_norm": 104.88539123535156,
      "learning_rate": 9.811722054838412e-08,
      "loss": 26.4102,
      "step": 4480
    },
    {
      "epoch": 35.638085742771686,
      "grad_norm": 134.7513427734375,
      "learning_rate": 9.712100185959862e-08,
      "loss": 25.6119,
      "step": 4490
    },
    {
      "epoch": 35.71784646061815,
      "grad_norm": 183.24839782714844,
      "learning_rate": 9.61286458320763e-08,
      "loss": 24.3452,
      "step": 4500
    },
    {
      "epoch": 35.71784646061815,
      "eval_loss": 3.1262333393096924,
      "eval_mae": 1.333138108253479,
      "eval_mse": 3.1262331008911133,
      "eval_r2": 0.058081209659576416,
      "eval_rmse": 1.7681156921681096,
      "eval_runtime": 9.0519,
      "eval_samples_per_second": 443.114,
      "eval_steps_per_second": 13.92,
      "step": 4500
    },
    {
      "epoch": 35.79760717846461,
      "grad_norm": 215.09974670410156,
      "learning_rate": 9.51401775388177e-08,
      "loss": 24.9939,
      "step": 4510
    },
    {
      "epoch": 35.877367896311064,
      "grad_norm": 102.00492858886719,
      "learning_rate": 9.41556219545953e-08,
      "loss": 25.3147,
      "step": 4520
    },
    {
      "epoch": 35.957128614157526,
      "grad_norm": 96.8487319946289,
      "learning_rate": 9.317500395532235e-08,
      "loss": 25.4636,
      "step": 4530
    },
    {
      "epoch": 36.03190428713859,
      "grad_norm": 68.97376251220703,
      "learning_rate": 9.219834831742482e-08,
      "loss": 22.8502,
      "step": 4540
    },
    {
      "epoch": 36.11166500498504,
      "grad_norm": 51.5819206237793,
      "learning_rate": 9.122567971721484e-08,
      "loss": 25.1066,
      "step": 4550
    },
    {
      "epoch": 36.191425722831504,
      "grad_norm": 137.3096466064453,
      "learning_rate": 9.025702273026787e-08,
      "loss": 25.7935,
      "step": 4560
    },
    {
      "epoch": 36.271186440677965,
      "grad_norm": 303.2586669921875,
      "learning_rate": 8.929240183080106e-08,
      "loss": 23.7943,
      "step": 4570
    },
    {
      "epoch": 36.35094715852443,
      "grad_norm": 298.3479309082031,
      "learning_rate": 8.833184139105534e-08,
      "loss": 25.0741,
      "step": 4580
    },
    {
      "epoch": 36.43070787637089,
      "grad_norm": 202.3203125,
      "learning_rate": 8.737536568067974e-08,
      "loss": 23.5568,
      "step": 4590
    },
    {
      "epoch": 36.51046859421735,
      "grad_norm": 118.80555725097656,
      "learning_rate": 8.642299886611759e-08,
      "loss": 25.5861,
      "step": 4600
    },
    {
      "epoch": 36.51046859421735,
      "eval_loss": 3.1254942417144775,
      "eval_mae": 1.3321858644485474,
      "eval_mse": 3.1254942417144775,
      "eval_r2": 0.0583038330078125,
      "eval_rmse": 1.7679067401066373,
      "eval_runtime": 9.063,
      "eval_samples_per_second": 442.571,
      "eval_steps_per_second": 13.903,
      "step": 4600
    },
    {
      "epoch": 36.59022931206381,
      "grad_norm": 145.08285522460938,
      "learning_rate": 8.54747650099967e-08,
      "loss": 27.5149,
      "step": 4610
    },
    {
      "epoch": 36.66999002991027,
      "grad_norm": 164.43328857421875,
      "learning_rate": 8.453068807052064e-08,
      "loss": 26.1562,
      "step": 4620
    },
    {
      "epoch": 36.74975074775673,
      "grad_norm": 46.407039642333984,
      "learning_rate": 8.359079190086402e-08,
      "loss": 26.4529,
      "step": 4630
    },
    {
      "epoch": 36.82951146560319,
      "grad_norm": 102.09172058105469,
      "learning_rate": 8.265510024856951e-08,
      "loss": 25.0347,
      "step": 4640
    },
    {
      "epoch": 36.90927218344965,
      "grad_norm": 119.56255340576172,
      "learning_rate": 8.172363675494772e-08,
      "loss": 25.3944,
      "step": 4650
    },
    {
      "epoch": 36.98903290129611,
      "grad_norm": 220.11642456054688,
      "learning_rate": 8.079642495448033e-08,
      "loss": 25.6366,
      "step": 4660
    },
    {
      "epoch": 37.06380857427717,
      "grad_norm": 172.42779541015625,
      "learning_rate": 7.987348827422483e-08,
      "loss": 25.4652,
      "step": 4670
    },
    {
      "epoch": 37.14356929212363,
      "grad_norm": 95.17620086669922,
      "learning_rate": 7.895485003322327e-08,
      "loss": 25.5158,
      "step": 4680
    },
    {
      "epoch": 37.22333000997009,
      "grad_norm": 81.66320037841797,
      "learning_rate": 7.804053344191255e-08,
      "loss": 25.609,
      "step": 4690
    },
    {
      "epoch": 37.30309072781655,
      "grad_norm": 112.18912506103516,
      "learning_rate": 7.713056160153813e-08,
      "loss": 24.5701,
      "step": 4700
    },
    {
      "epoch": 37.30309072781655,
      "eval_loss": 3.1239125728607178,
      "eval_mae": 1.332130789756775,
      "eval_mse": 3.1239125728607178,
      "eval_r2": 0.05878037214279175,
      "eval_rmse": 1.7674593553631488,
      "eval_runtime": 9.0645,
      "eval_samples_per_second": 442.496,
      "eval_steps_per_second": 13.9,
      "step": 4700
    },
    {
      "epoch": 37.38285144566301,
      "grad_norm": 103.58297729492188,
      "learning_rate": 7.622495750357072e-08,
      "loss": 25.5507,
      "step": 4710
    },
    {
      "epoch": 37.46261216350947,
      "grad_norm": 39.11042022705078,
      "learning_rate": 7.53237440291247e-08,
      "loss": 25.7191,
      "step": 4720
    },
    {
      "epoch": 37.54237288135593,
      "grad_norm": 199.56143188476562,
      "learning_rate": 7.44269439483807e-08,
      "loss": 25.2238,
      "step": 4730
    },
    {
      "epoch": 37.62213359920239,
      "grad_norm": 168.63783264160156,
      "learning_rate": 7.353457992000961e-08,
      "loss": 24.6829,
      "step": 4740
    },
    {
      "epoch": 37.701894317048854,
      "grad_norm": 55.98255157470703,
      "learning_rate": 7.264667449060074e-08,
      "loss": 25.5242,
      "step": 4750
    },
    {
      "epoch": 37.781655034895316,
      "grad_norm": 53.036888122558594,
      "learning_rate": 7.176325009409175e-08,
      "loss": 25.344,
      "step": 4760
    },
    {
      "epoch": 37.86141575274178,
      "grad_norm": 265.2929382324219,
      "learning_rate": 7.088432905120176e-08,
      "loss": 24.8157,
      "step": 4770
    },
    {
      "epoch": 37.94117647058823,
      "grad_norm": 120.75648498535156,
      "learning_rate": 7.000993356886778e-08,
      "loss": 23.7128,
      "step": 4780
    },
    {
      "epoch": 38.015952143569294,
      "grad_norm": 171.9890899658203,
      "learning_rate": 6.914008573968317e-08,
      "loss": 26.0064,
      "step": 4790
    },
    {
      "epoch": 38.095712861415755,
      "grad_norm": 53.58652114868164,
      "learning_rate": 6.827480754133993e-08,
      "loss": 24.0923,
      "step": 4800
    },
    {
      "epoch": 38.095712861415755,
      "eval_loss": 3.123528003692627,
      "eval_mae": 1.3319824934005737,
      "eval_mse": 3.123528003692627,
      "eval_r2": 0.05889630317687988,
      "eval_rmse": 1.767350560498009,
      "eval_runtime": 9.0553,
      "eval_samples_per_second": 442.943,
      "eval_steps_per_second": 13.914,
      "step": 4800
    },
    {
      "epoch": 38.17547357926221,
      "grad_norm": 96.7684097290039,
      "learning_rate": 6.741412083607301e-08,
      "loss": 25.6224,
      "step": 4810
    },
    {
      "epoch": 38.25523429710867,
      "grad_norm": 328.3621826171875,
      "learning_rate": 6.655804737010795e-08,
      "loss": 26.9059,
      "step": 4820
    },
    {
      "epoch": 38.33499501495513,
      "grad_norm": 89.13323211669922,
      "learning_rate": 6.570660877311193e-08,
      "loss": 24.2715,
      "step": 4830
    },
    {
      "epoch": 38.414755732801595,
      "grad_norm": 157.61807250976562,
      "learning_rate": 6.485982655764649e-08,
      "loss": 25.0064,
      "step": 4840
    },
    {
      "epoch": 38.49451645064806,
      "grad_norm": 151.26788330078125,
      "learning_rate": 6.401772211862468e-08,
      "loss": 23.3381,
      "step": 4850
    },
    {
      "epoch": 38.57427716849452,
      "grad_norm": 57.38077926635742,
      "learning_rate": 6.31803167327703e-08,
      "loss": 23.812,
      "step": 4860
    },
    {
      "epoch": 38.65403788634098,
      "grad_norm": 203.0427703857422,
      "learning_rate": 6.23476315580799e-08,
      "loss": 26.5058,
      "step": 4870
    },
    {
      "epoch": 38.733798604187434,
      "grad_norm": 163.02796936035156,
      "learning_rate": 6.151968763328888e-08,
      "loss": 26.1078,
      "step": 4880
    },
    {
      "epoch": 38.813559322033896,
      "grad_norm": 108.4669418334961,
      "learning_rate": 6.069650587733932e-08,
      "loss": 24.9732,
      "step": 4890
    },
    {
      "epoch": 38.89332003988036,
      "grad_norm": 362.93994140625,
      "learning_rate": 5.987810708885196e-08,
      "loss": 26.611,
      "step": 4900
    },
    {
      "epoch": 38.89332003988036,
      "eval_loss": 3.122894525527954,
      "eval_mae": 1.3315399885177612,
      "eval_mse": 3.122894525527954,
      "eval_r2": 0.059087157249450684,
      "eval_rmse": 1.7671713345139894,
      "eval_runtime": 9.0533,
      "eval_samples_per_second": 443.044,
      "eval_steps_per_second": 13.918,
      "step": 4900
    },
    {
      "epoch": 38.97308075772682,
      "grad_norm": 93.97080993652344,
      "learning_rate": 5.906451194560019e-08,
      "loss": 25.4004,
      "step": 4910
    },
    {
      "epoch": 39.047856430707874,
      "grad_norm": 147.77755737304688,
      "learning_rate": 5.825574100398811e-08,
      "loss": 23.691,
      "step": 4920
    },
    {
      "epoch": 39.127617148554336,
      "grad_norm": 246.35997009277344,
      "learning_rate": 5.7451814698530665e-08,
      "loss": 25.2164,
      "step": 4930
    },
    {
      "epoch": 39.2073778664008,
      "grad_norm": 153.21450805664062,
      "learning_rate": 5.6652753341337623e-08,
      "loss": 24.0185,
      "step": 4940
    },
    {
      "epoch": 39.28713858424726,
      "grad_norm": 143.73780822753906,
      "learning_rate": 5.58585771216005e-08,
      "loss": 25.8635,
      "step": 4950
    },
    {
      "epoch": 39.36689930209372,
      "grad_norm": 58.399566650390625,
      "learning_rate": 5.506930610508201e-08,
      "loss": 25.9007,
      "step": 4960
    },
    {
      "epoch": 39.44666001994018,
      "grad_norm": 26.310766220092773,
      "learning_rate": 5.4284960233609536e-08,
      "loss": 23.4994,
      "step": 4970
    },
    {
      "epoch": 39.52642073778664,
      "grad_norm": 146.9088897705078,
      "learning_rate": 5.350555932457107e-08,
      "loss": 26.3982,
      "step": 4980
    },
    {
      "epoch": 39.6061814556331,
      "grad_norm": 132.27772521972656,
      "learning_rate": 5.273112307041436e-08,
      "loss": 25.3715,
      "step": 4990
    },
    {
      "epoch": 39.68594217347956,
      "grad_norm": 55.70192337036133,
      "learning_rate": 5.1961671038149745e-08,
      "loss": 27.1358,
      "step": 5000
    },
    {
      "epoch": 39.68594217347956,
      "eval_loss": 3.1226961612701416,
      "eval_mae": 1.3344401121139526,
      "eval_mse": 3.1226959228515625,
      "eval_r2": 0.0591469407081604,
      "eval_rmse": 1.767115141367863,
      "eval_runtime": 9.0408,
      "eval_samples_per_second": 443.654,
      "eval_steps_per_second": 13.937,
      "step": 5000
    },
    {
      "epoch": 39.76570289132602,
      "grad_norm": 45.21358871459961,
      "learning_rate": 5.119722266885534e-08,
      "loss": 25.5138,
      "step": 5010
    },
    {
      "epoch": 39.845463609172484,
      "grad_norm": 101.61859130859375,
      "learning_rate": 5.043779727718622e-08,
      "loss": 23.5845,
      "step": 5020
    },
    {
      "epoch": 39.925224327018945,
      "grad_norm": 57.091800689697266,
      "learning_rate": 4.968341405088602e-08,
      "loss": 26.287,
      "step": 5030
    },
    {
      "epoch": 40.0,
      "grad_norm": 21.503507614135742,
      "learning_rate": 4.8934092050302636e-08,
      "loss": 25.0861,
      "step": 5040
    },
    {
      "epoch": 40.07976071784646,
      "grad_norm": 46.6530647277832,
      "learning_rate": 4.818985020790603e-08,
      "loss": 26.3137,
      "step": 5050
    },
    {
      "epoch": 40.15952143569292,
      "grad_norm": 110.19184112548828,
      "learning_rate": 4.7450707327810534e-08,
      "loss": 25.7188,
      "step": 5060
    },
    {
      "epoch": 40.239282153539385,
      "grad_norm": 114.34848022460938,
      "learning_rate": 4.6716682085299135e-08,
      "loss": 24.9849,
      "step": 5070
    },
    {
      "epoch": 40.31904287138584,
      "grad_norm": 76.02941131591797,
      "learning_rate": 4.598779302635203e-08,
      "loss": 25.1661,
      "step": 5080
    },
    {
      "epoch": 40.3988035892323,
      "grad_norm": 161.83319091796875,
      "learning_rate": 4.5264058567177906e-08,
      "loss": 24.9227,
      "step": 5090
    },
    {
      "epoch": 40.47856430707876,
      "grad_norm": 110.09339904785156,
      "learning_rate": 4.454549699374874e-08,
      "loss": 23.4286,
      "step": 5100
    },
    {
      "epoch": 40.47856430707876,
      "eval_loss": 3.1224095821380615,
      "eval_mae": 1.3299893140792847,
      "eval_mse": 3.1224091053009033,
      "eval_r2": 0.05923336744308472,
      "eval_rmse": 1.76703398532708,
      "eval_runtime": 9.0409,
      "eval_samples_per_second": 443.649,
      "eval_steps_per_second": 13.937,
      "step": 5100
    },
    {
      "epoch": 40.558325024925225,
      "grad_norm": 51.97368621826172,
      "learning_rate": 4.383212646133741e-08,
      "loss": 24.7696,
      "step": 5110
    },
    {
      "epoch": 40.638085742771686,
      "grad_norm": 203.3778533935547,
      "learning_rate": 4.319454622188645e-08,
      "loss": 24.8723,
      "step": 5120
    },
    {
      "epoch": 40.71784646061815,
      "grad_norm": 96.84310150146484,
      "learning_rate": 4.2491088215078766e-08,
      "loss": 26.5034,
      "step": 5130
    },
    {
      "epoch": 40.79760717846461,
      "grad_norm": 56.1543083190918,
      "learning_rate": 4.17928731562566e-08,
      "loss": 24.4977,
      "step": 5140
    },
    {
      "epoch": 40.877367896311064,
      "grad_norm": 65.11670684814453,
      "learning_rate": 4.109991868661525e-08,
      "loss": 27.2082,
      "step": 5150
    },
    {
      "epoch": 40.957128614157526,
      "grad_norm": 77.25297546386719,
      "learning_rate": 4.041224231443538e-08,
      "loss": 24.6315,
      "step": 5160
    },
    {
      "epoch": 41.03190428713859,
      "grad_norm": 134.2405548095703,
      "learning_rate": 3.972986141464027e-08,
      "loss": 26.111,
      "step": 5170
    },
    {
      "epoch": 41.11166500498504,
      "grad_norm": 53.60357666015625,
      "learning_rate": 3.905279322835728e-08,
      "loss": 25.429,
      "step": 5180
    },
    {
      "epoch": 41.191425722831504,
      "grad_norm": 84.80567932128906,
      "learning_rate": 3.838105486248186e-08,
      "loss": 23.1782,
      "step": 5190
    },
    {
      "epoch": 41.271186440677965,
      "grad_norm": 85.67137908935547,
      "learning_rate": 3.77146632892458e-08,
      "loss": 26.1872,
      "step": 5200
    },
    {
      "epoch": 41.271186440677965,
      "eval_loss": 3.1218793392181396,
      "eval_mae": 1.3309307098388672,
      "eval_mse": 3.1218795776367188,
      "eval_r2": 0.05939292907714844,
      "eval_rmse": 1.7668841438070348,
      "eval_runtime": 9.0453,
      "eval_samples_per_second": 443.435,
      "eval_steps_per_second": 13.93,
      "step": 5200
    },
    {
      "epoch": 41.35094715852443,
      "grad_norm": 66.26146697998047,
      "learning_rate": 3.705363534578779e-08,
      "loss": 24.4168,
      "step": 5210
    },
    {
      "epoch": 41.43070787637089,
      "grad_norm": 84.41664123535156,
      "learning_rate": 3.639798773372854e-08,
      "loss": 24.8912,
      "step": 5220
    },
    {
      "epoch": 41.51046859421735,
      "grad_norm": 154.61509704589844,
      "learning_rate": 3.574773701874853e-08,
      "loss": 25.9521,
      "step": 5230
    },
    {
      "epoch": 41.59022931206381,
      "grad_norm": 180.2846221923828,
      "learning_rate": 3.510289963016946e-08,
      "loss": 25.1476,
      "step": 5240
    },
    {
      "epoch": 41.66999002991027,
      "grad_norm": 39.9274787902832,
      "learning_rate": 3.446349186053929e-08,
      "loss": 24.6444,
      "step": 5250
    },
    {
      "epoch": 41.74975074775673,
      "grad_norm": 60.37506103515625,
      "learning_rate": 3.382952986522039e-08,
      "loss": 28.0159,
      "step": 5260
    },
    {
      "epoch": 41.82951146560319,
      "grad_norm": 220.5933380126953,
      "learning_rate": 3.320102966198155e-08,
      "loss": 23.8454,
      "step": 5270
    },
    {
      "epoch": 41.90927218344965,
      "grad_norm": 68.86772155761719,
      "learning_rate": 3.257800713059325e-08,
      "loss": 25.9979,
      "step": 5280
    },
    {
      "epoch": 41.98903290129611,
      "grad_norm": 271.44146728515625,
      "learning_rate": 3.1960478012426215e-08,
      "loss": 25.2016,
      "step": 5290
    },
    {
      "epoch": 42.06380857427717,
      "grad_norm": 168.9261016845703,
      "learning_rate": 3.134845791005391e-08,
      "loss": 23.1833,
      "step": 5300
    },
    {
      "epoch": 42.06380857427717,
      "eval_loss": 3.1214566230773926,
      "eval_mae": 1.3316025733947754,
      "eval_mse": 3.1214566230773926,
      "eval_r2": 0.059520423412323,
      "eval_rmse": 1.7667644503660902,
      "eval_runtime": 9.0542,
      "eval_samples_per_second": 442.999,
      "eval_steps_per_second": 13.916,
      "step": 5300
    },
    {
      "epoch": 42.14356929212363,
      "grad_norm": 194.90185546875,
      "learning_rate": 3.074196228685833e-08,
      "loss": 22.7641,
      "step": 5310
    },
    {
      "epoch": 42.22333000997009,
      "grad_norm": 67.69960021972656,
      "learning_rate": 3.0141006466639166e-08,
      "loss": 25.6798,
      "step": 5320
    },
    {
      "epoch": 42.30309072781655,
      "grad_norm": 37.08260726928711,
      "learning_rate": 2.954560563322686e-08,
      "loss": 26.3616,
      "step": 5330
    },
    {
      "epoch": 42.38285144566301,
      "grad_norm": 79.85238647460938,
      "learning_rate": 2.8955774830098534e-08,
      "loss": 25.6113,
      "step": 5340
    },
    {
      "epoch": 42.46261216350947,
      "grad_norm": 93.74951934814453,
      "learning_rate": 2.8371528959998475e-08,
      "loss": 26.3547,
      "step": 5350
    },
    {
      "epoch": 42.54237288135593,
      "grad_norm": 133.0259552001953,
      "learning_rate": 2.7792882784561046e-08,
      "loss": 25.1082,
      "step": 5360
    },
    {
      "epoch": 42.62213359920239,
      "grad_norm": 161.1022491455078,
      "learning_rate": 2.721985092393822e-08,
      "loss": 25.5645,
      "step": 5370
    },
    {
      "epoch": 42.701894317048854,
      "grad_norm": 104.63832092285156,
      "learning_rate": 2.66524478564297e-08,
      "loss": 26.045,
      "step": 5380
    },
    {
      "epoch": 42.781655034895316,
      "grad_norm": 153.9829559326172,
      "learning_rate": 2.60906879181175e-08,
      "loss": 26.1566,
      "step": 5390
    },
    {
      "epoch": 42.86141575274178,
      "grad_norm": 88.28377532958984,
      "learning_rate": 2.5534585302503576e-08,
      "loss": 25.2555,
      "step": 5400
    },
    {
      "epoch": 42.86141575274178,
      "eval_loss": 3.1208252906799316,
      "eval_mae": 1.331380009651184,
      "eval_mse": 3.1208252906799316,
      "eval_r2": 0.05971062183380127,
      "eval_rmse": 1.766585772239755,
      "eval_runtime": 9.0542,
      "eval_samples_per_second": 443.0,
      "eval_steps_per_second": 13.916,
      "step": 5400
    },
    {
      "epoch": 42.94117647058823,
      "grad_norm": 147.16366577148438,
      "learning_rate": 2.498415406015114e-08,
      "loss": 24.5054,
      "step": 5410
    },
    {
      "epoch": 43.015952143569294,
      "grad_norm": 234.27850341796875,
      "learning_rate": 2.4439408098329754e-08,
      "loss": 24.1556,
      "step": 5420
    },
    {
      "epoch": 43.095712861415755,
      "grad_norm": 66.06194305419922,
      "learning_rate": 2.3900361180663907e-08,
      "loss": 24.8309,
      "step": 5430
    },
    {
      "epoch": 43.17547357926221,
      "grad_norm": 100.03466796875,
      "learning_rate": 2.336702692678527e-08,
      "loss": 23.5658,
      "step": 5440
    },
    {
      "epoch": 43.25523429710867,
      "grad_norm": 102.8028564453125,
      "learning_rate": 2.2839418811988697e-08,
      "loss": 27.2123,
      "step": 5450
    },
    {
      "epoch": 43.33499501495513,
      "grad_norm": 141.64247131347656,
      "learning_rate": 2.2317550166891435e-08,
      "loss": 25.9267,
      "step": 5460
    },
    {
      "epoch": 43.414755732801595,
      "grad_norm": 185.2597198486328,
      "learning_rate": 2.1801434177096762e-08,
      "loss": 24.6855,
      "step": 5470
    },
    {
      "epoch": 43.49451645064806,
      "grad_norm": 198.41571044921875,
      "learning_rate": 2.1291083882860355e-08,
      "loss": 26.7005,
      "step": 5480
    },
    {
      "epoch": 43.57427716849452,
      "grad_norm": 55.69032669067383,
      "learning_rate": 2.0786512178761207e-08,
      "loss": 22.5448,
      "step": 5490
    },
    {
      "epoch": 43.65403788634098,
      "grad_norm": 88.44397735595703,
      "learning_rate": 2.0287731813375702e-08,
      "loss": 25.6887,
      "step": 5500
    },
    {
      "epoch": 43.65403788634098,
      "eval_loss": 3.1209118366241455,
      "eval_mae": 1.3314346075057983,
      "eval_mse": 3.1209118366241455,
      "eval_r2": 0.05968451499938965,
      "eval_rmse": 1.7666102673267088,
      "eval_runtime": 9.0592,
      "eval_samples_per_second": 442.753,
      "eval_steps_per_second": 13.908,
      "step": 5500
    },
    {
      "epoch": 43.733798604187434,
      "grad_norm": 84.4006118774414,
      "learning_rate": 1.9794755388955363e-08,
      "loss": 26.3546,
      "step": 5510
    },
    {
      "epoch": 43.813559322033896,
      "grad_norm": 121.09300231933594,
      "learning_rate": 1.9307595361108654e-08,
      "loss": 24.7965,
      "step": 5520
    },
    {
      "epoch": 43.89332003988036,
      "grad_norm": 55.71806716918945,
      "learning_rate": 1.882626403848611e-08,
      "loss": 24.5852,
      "step": 5530
    },
    {
      "epoch": 43.97308075772682,
      "grad_norm": 51.23670959472656,
      "learning_rate": 1.8350773582469502e-08,
      "loss": 27.0388,
      "step": 5540
    },
    {
      "epoch": 44.047856430707874,
      "grad_norm": 179.18077087402344,
      "learning_rate": 1.7881136006864388e-08,
      "loss": 22.9556,
      "step": 5550
    },
    {
      "epoch": 44.127617148554336,
      "grad_norm": 59.81156921386719,
      "learning_rate": 1.7417363177596733e-08,
      "loss": 27.1492,
      "step": 5560
    },
    {
      "epoch": 44.2073778664008,
      "grad_norm": 61.35225296020508,
      "learning_rate": 1.6959466812413086e-08,
      "loss": 25.5271,
      "step": 5570
    },
    {
      "epoch": 44.28713858424726,
      "grad_norm": 25.337446212768555,
      "learning_rate": 1.6507458480584297e-08,
      "loss": 24.9083,
      "step": 5580
    },
    {
      "epoch": 44.36689930209372,
      "grad_norm": 127.35221862792969,
      "learning_rate": 1.6061349602613573e-08,
      "loss": 24.5705,
      "step": 5590
    },
    {
      "epoch": 44.44666001994018,
      "grad_norm": 141.15887451171875,
      "learning_rate": 1.562115144994755e-08,
      "loss": 24.4761,
      "step": 5600
    },
    {
      "epoch": 44.44666001994018,
      "eval_loss": 3.120940923690796,
      "eval_mae": 1.331071376800537,
      "eval_mse": 3.120941162109375,
      "eval_r2": 0.05967569351196289,
      "eval_rmse": 1.7666185672378107,
      "eval_runtime": 14.0638,
      "eval_samples_per_second": 285.2,
      "eval_steps_per_second": 8.959,
      "step": 5600
    },
    {
      "epoch": 44.52642073778664,
      "grad_norm": 270.5105895996094,
      "learning_rate": 1.51868751446918e-08,
      "loss": 25.0175,
      "step": 5610
    },
    {
      "epoch": 44.6061814556331,
      "grad_norm": 158.59524536132812,
      "learning_rate": 1.4758531659329747e-08,
      "loss": 25.5007,
      "step": 5620
    },
    {
      "epoch": 44.68594217347956,
      "grad_norm": 108.44390869140625,
      "learning_rate": 1.433613181644519e-08,
      "loss": 26.3937,
      "step": 5630
    },
    {
      "epoch": 44.76570289132602,
      "grad_norm": 67.45941162109375,
      "learning_rate": 1.3919686288449305e-08,
      "loss": 25.3727,
      "step": 5640
    },
    {
      "epoch": 44.845463609172484,
      "grad_norm": 173.1921844482422,
      "learning_rate": 1.3509205597310586e-08,
      "loss": 24.0026,
      "step": 5650
    },
    {
      "epoch": 44.925224327018945,
      "grad_norm": 332.9158630371094,
      "learning_rate": 1.3104700114289185e-08,
      "loss": 25.2033,
      "step": 5660
    },
    {
      "epoch": 45.0,
      "grad_norm": 18.381704330444336,
      "learning_rate": 1.2706180059674915e-08,
      "loss": 25.1115,
      "step": 5670
    },
    {
      "epoch": 45.07976071784646,
      "grad_norm": 36.26863098144531,
      "learning_rate": 1.231365550252883e-08,
      "loss": 24.8773,
      "step": 5680
    },
    {
      "epoch": 45.15952143569292,
      "grad_norm": 150.63861083984375,
      "learning_rate": 1.1927136360429108e-08,
      "loss": 24.7349,
      "step": 5690
    },
    {
      "epoch": 45.239282153539385,
      "grad_norm": 135.55755615234375,
      "learning_rate": 1.154663239922013e-08,
      "loss": 25.1996,
      "step": 5700
    },
    {
      "epoch": 45.239282153539385,
      "eval_loss": 3.120323896408081,
      "eval_mae": 1.3315786123275757,
      "eval_mse": 3.120323896408081,
      "eval_r2": 0.05986166000366211,
      "eval_rmse": 1.7664438560022453,
      "eval_runtime": 9.0365,
      "eval_samples_per_second": 443.865,
      "eval_steps_per_second": 13.943,
      "step": 5700
    },
    {
      "epoch": 45.31904287138584,
      "grad_norm": 92.68675994873047,
      "learning_rate": 1.1172153232766073e-08,
      "loss": 26.8737,
      "step": 5710
    },
    {
      "epoch": 45.3988035892323,
      "grad_norm": 75.78144073486328,
      "learning_rate": 1.080370832270769e-08,
      "loss": 24.521,
      "step": 5720
    },
    {
      "epoch": 45.47856430707876,
      "grad_norm": 30.703096389770508,
      "learning_rate": 1.0441306978223513e-08,
      "loss": 25.99,
      "step": 5730
    },
    {
      "epoch": 45.558325024925225,
      "grad_norm": 132.82154846191406,
      "learning_rate": 1.0084958355794537e-08,
      "loss": 26.4817,
      "step": 5740
    },
    {
      "epoch": 45.638085742771686,
      "grad_norm": 111.67121887207031,
      "learning_rate": 9.734671458972749e-09,
      "loss": 26.3786,
      "step": 5750
    },
    {
      "epoch": 45.71784646061815,
      "grad_norm": 74.11526489257812,
      "learning_rate": 9.390455138153924e-09,
      "loss": 24.5637,
      "step": 5760
    },
    {
      "epoch": 45.79760717846461,
      "grad_norm": 64.91019439697266,
      "learning_rate": 9.052318090353728e-09,
      "loss": 23.873,
      "step": 5770
    },
    {
      "epoch": 45.877367896311064,
      "grad_norm": 99.60811614990234,
      "learning_rate": 8.720268858988105e-09,
      "loss": 24.5365,
      "step": 5780
    },
    {
      "epoch": 45.957128614157526,
      "grad_norm": 116.65836334228516,
      "learning_rate": 8.394315833657544e-09,
      "loss": 24.941,
      "step": 5790
    },
    {
      "epoch": 46.03190428713859,
      "grad_norm": 65.44205474853516,
      "learning_rate": 8.074467249934796e-09,
      "loss": 25.0614,
      "step": 5800
    },
    {
      "epoch": 46.03190428713859,
      "eval_loss": 3.120579481124878,
      "eval_mae": 1.3315255641937256,
      "eval_mse": 3.120579481124878,
      "eval_r2": 0.059784650802612305,
      "eval_rmse": 1.7665161989421092,
      "eval_runtime": 9.0413,
      "eval_samples_per_second": 443.631,
      "eval_steps_per_second": 13.936,
      "step": 5800
    },
    {
      "epoch": 46.11166500498504,
      "grad_norm": 87.28318786621094,
      "learning_rate": 7.760731189157156e-09,
      "loss": 25.5226,
      "step": 5810
    },
    {
      "epoch": 46.191425722831504,
      "grad_norm": 251.4473114013672,
      "learning_rate": 7.453115578221897e-09,
      "loss": 26.1362,
      "step": 5820
    },
    {
      "epoch": 46.271186440677965,
      "grad_norm": 58.065940856933594,
      "learning_rate": 7.15162818938636e-09,
      "loss": 25.5341,
      "step": 5830
    },
    {
      "epoch": 46.35094715852443,
      "grad_norm": 78.00777435302734,
      "learning_rate": 6.8562766400712676e-09,
      "loss": 25.3617,
      "step": 5840
    },
    {
      "epoch": 46.43070787637089,
      "grad_norm": 62.271507263183594,
      "learning_rate": 6.56706839266849e-09,
      "loss": 25.0898,
      "step": 5850
    },
    {
      "epoch": 46.51046859421735,
      "grad_norm": 45.634952545166016,
      "learning_rate": 6.284010754352448e-09,
      "loss": 24.7997,
      "step": 5860
    },
    {
      "epoch": 46.59022931206381,
      "grad_norm": 89.99065399169922,
      "learning_rate": 6.007110876895261e-09,
      "loss": 24.4887,
      "step": 5870
    },
    {
      "epoch": 46.66999002991027,
      "grad_norm": 37.37367248535156,
      "learning_rate": 5.736375756486444e-09,
      "loss": 25.5751,
      "step": 5880
    },
    {
      "epoch": 46.74975074775673,
      "grad_norm": 123.9768295288086,
      "learning_rate": 5.471812233555834e-09,
      "loss": 24.7693,
      "step": 5890
    },
    {
      "epoch": 46.82951146560319,
      "grad_norm": 130.57916259765625,
      "learning_rate": 5.2134269926008005e-09,
      "loss": 24.1577,
      "step": 5900
    },
    {
      "epoch": 46.82951146560319,
      "eval_loss": 3.1201043128967285,
      "eval_mae": 1.3313286304473877,
      "eval_mse": 3.1201043128967285,
      "eval_r2": 0.05992782115936279,
      "eval_rmse": 1.766381700793101,
      "eval_runtime": 9.0588,
      "eval_samples_per_second": 442.774,
      "eval_steps_per_second": 13.909,
      "step": 5900
    },
    {
      "epoch": 46.90927218344965,
      "grad_norm": 136.50294494628906,
      "learning_rate": 4.961226562017584e-09,
      "loss": 27.04,
      "step": 5910
    },
    {
      "epoch": 46.98903290129611,
      "grad_norm": 162.00277709960938,
      "learning_rate": 4.71521731393601e-09,
      "loss": 25.3749,
      "step": 5920
    },
    {
      "epoch": 47.06380857427717,
      "grad_norm": 155.60572814941406,
      "learning_rate": 4.4754054640587515e-09,
      "loss": 25.0107,
      "step": 5930
    },
    {
      "epoch": 47.14356929212363,
      "grad_norm": 141.03631591796875,
      "learning_rate": 4.241797071504127e-09,
      "loss": 26.8127,
      "step": 5940
    },
    {
      "epoch": 47.22333000997009,
      "grad_norm": 51.73516082763672,
      "learning_rate": 4.014398038653111e-09,
      "loss": 24.7251,
      "step": 5950
    },
    {
      "epoch": 47.30309072781655,
      "grad_norm": 256.9667663574219,
      "learning_rate": 3.793214111000198e-09,
      "loss": 26.5637,
      "step": 5960
    },
    {
      "epoch": 47.38285144566301,
      "grad_norm": 47.89490509033203,
      "learning_rate": 3.57825087700811e-09,
      "loss": 25.4085,
      "step": 5970
    },
    {
      "epoch": 47.46261216350947,
      "grad_norm": 144.0662078857422,
      "learning_rate": 3.369513767966764e-09,
      "loss": 24.9812,
      "step": 5980
    },
    {
      "epoch": 47.54237288135593,
      "grad_norm": 59.69084167480469,
      "learning_rate": 3.167008057855941e-09,
      "loss": 24.4607,
      "step": 5990
    },
    {
      "epoch": 47.62213359920239,
      "grad_norm": 57.48725509643555,
      "learning_rate": 2.97073886321203e-09,
      "loss": 23.9325,
      "step": 6000
    },
    {
      "epoch": 47.62213359920239,
      "eval_loss": 3.1205005645751953,
      "eval_mae": 1.3313699960708618,
      "eval_mse": 3.1205005645751953,
      "eval_r2": 0.059808433055877686,
      "eval_rmse": 1.7664938620259045,
      "eval_runtime": 9.0615,
      "eval_samples_per_second": 442.643,
      "eval_steps_per_second": 13.905,
      "step": 6000
    },
    {
      "epoch": 47.701894317048854,
      "grad_norm": 173.4394073486328,
      "learning_rate": 2.780711142998854e-09,
      "loss": 24.7271,
      "step": 6010
    },
    {
      "epoch": 47.781655034895316,
      "grad_norm": 82.74398803710938,
      "learning_rate": 2.596929698482242e-09,
      "loss": 26.1529,
      "step": 6020
    },
    {
      "epoch": 47.86141575274178,
      "grad_norm": 130.01808166503906,
      "learning_rate": 2.419399173108849e-09,
      "loss": 24.504,
      "step": 6030
    },
    {
      "epoch": 47.94117647058823,
      "grad_norm": 43.647499084472656,
      "learning_rate": 2.2481240523886668e-09,
      "loss": 25.6795,
      "step": 6040
    },
    {
      "epoch": 48.015952143569294,
      "grad_norm": 56.47761535644531,
      "learning_rate": 2.0831086637819182e-09,
      "loss": 23.4324,
      "step": 6050
    },
    {
      "epoch": 48.095712861415755,
      "grad_norm": 155.72174072265625,
      "learning_rate": 1.9243571765895083e-09,
      "loss": 25.0166,
      "step": 6060
    },
    {
      "epoch": 48.17547357926221,
      "grad_norm": 111.24287414550781,
      "learning_rate": 1.7718736018478008e-09,
      "loss": 24.6729,
      "step": 6070
    },
    {
      "epoch": 48.25523429710867,
      "grad_norm": 64.62834167480469,
      "learning_rate": 1.6256617922272287e-09,
      "loss": 27.8513,
      "step": 6080
    },
    {
      "epoch": 48.33499501495513,
      "grad_norm": 78.91862487792969,
      "learning_rate": 1.4857254419349818e-09,
      "loss": 26.6801,
      "step": 6090
    },
    {
      "epoch": 48.414755732801595,
      "grad_norm": 120.07483673095703,
      "learning_rate": 1.352068086621666e-09,
      "loss": 26.0575,
      "step": 6100
    },
    {
      "epoch": 48.414755732801595,
      "eval_loss": 3.1204397678375244,
      "eval_mae": 1.331369161605835,
      "eval_mse": 3.1204397678375244,
      "eval_r2": 0.05982673168182373,
      "eval_rmse": 1.7664766536350047,
      "eval_runtime": 9.0446,
      "eval_samples_per_second": 443.468,
      "eval_steps_per_second": 13.931,
      "step": 6100
    },
    {
      "epoch": 48.49451645064806,
      "grad_norm": 249.7754669189453,
      "learning_rate": 1.2246931032919017e-09,
      "loss": 23.2525,
      "step": 6110
    },
    {
      "epoch": 48.57427716849452,
      "grad_norm": 73.22221374511719,
      "learning_rate": 1.103603710219142e-09,
      "loss": 26.0894,
      "step": 6120
    },
    {
      "epoch": 48.65403788634098,
      "grad_norm": 118.6893310546875,
      "learning_rate": 9.888029668642105e-10,
      "loss": 24.5071,
      "step": 6130
    },
    {
      "epoch": 48.733798604187434,
      "grad_norm": 47.15996170043945,
      "learning_rate": 8.802937737980298e-10,
      "loss": 24.6825,
      "step": 6140
    },
    {
      "epoch": 48.813559322033896,
      "grad_norm": 129.1663055419922,
      "learning_rate": 7.780788726284848e-10,
      "loss": 25.1016,
      "step": 6150
    },
    {
      "epoch": 48.89332003988036,
      "grad_norm": 158.0771942138672,
      "learning_rate": 6.821608459309513e-10,
      "loss": 26.7324,
      "step": 6160
    },
    {
      "epoch": 48.97308075772682,
      "grad_norm": 100.31987762451172,
      "learning_rate": 5.925421171831535e-10,
      "loss": 23.5741,
      "step": 6170
    },
    {
      "epoch": 49.047856430707874,
      "grad_norm": 143.13925170898438,
      "learning_rate": 5.092249507038793e-10,
      "loss": 23.3466,
      "step": 6180
    },
    {
      "epoch": 49.127617148554336,
      "grad_norm": 145.27874755859375,
      "learning_rate": 4.322114515958597e-10,
      "loss": 28.8833,
      "step": 6190
    },
    {
      "epoch": 49.2073778664008,
      "grad_norm": 60.93461608886719,
      "learning_rate": 3.6150356569250584e-10,
      "loss": 25.5008,
      "step": 6200
    },
    {
      "epoch": 49.2073778664008,
      "eval_loss": 3.1204824447631836,
      "eval_mae": 1.3314094543457031,
      "eval_mse": 3.1204826831817627,
      "eval_r2": 0.059813857078552246,
      "eval_rmse": 1.766488800751865,
      "eval_runtime": 9.0554,
      "eval_samples_per_second": 442.938,
      "eval_steps_per_second": 13.914,
      "step": 6200
    },
    {
      "epoch": 49.28713858424726,
      "grad_norm": 143.7784423828125,
      "learning_rate": 2.9710307950867044e-10,
      "loss": 26.3828,
      "step": 6210
    },
    {
      "epoch": 49.36689930209372,
      "grad_norm": 141.80580139160156,
      "learning_rate": 2.390116201957393e-10,
      "loss": 26.8702,
      "step": 6220
    },
    {
      "epoch": 49.44666001994018,
      "grad_norm": 56.256038665771484,
      "learning_rate": 1.8723065550030337e-10,
      "loss": 26.5281,
      "step": 6230
    },
    {
      "epoch": 49.52642073778664,
      "grad_norm": 117.0438232421875,
      "learning_rate": 1.4176149372713274e-10,
      "loss": 24.9371,
      "step": 6240
    },
    {
      "epoch": 49.6061814556331,
      "grad_norm": 50.82411575317383,
      "learning_rate": 1.0260528370617527e-10,
      "loss": 23.6784,
      "step": 6250
    },
    {
      "epoch": 49.68594217347956,
      "grad_norm": 38.99748229980469,
      "learning_rate": 6.976301476346869e-11,
      "loss": 23.8564,
      "step": 6260
    },
    {
      "epoch": 49.76570289132602,
      "grad_norm": 101.93575286865234,
      "learning_rate": 4.323551669621617e-11,
      "loss": 25.8426,
      "step": 6270
    },
    {
      "epoch": 49.845463609172484,
      "grad_norm": 90.21463012695312,
      "learning_rate": 2.3023459751719822e-11,
      "loss": 22.9579,
      "step": 6280
    },
    {
      "epoch": 49.925224327018945,
      "grad_norm": 92.85131072998047,
      "learning_rate": 9.127354610505289e-12,
      "loss": 26.2049,
      "step": 6290
    },
    {
      "epoch": 50.0,
      "grad_norm": 53.79079055786133,
      "learning_rate": 1.5475523734709283e-12,
      "loss": 20.9661,
      "step": 6300
    },
    {
      "epoch": 50.0,
      "eval_loss": 3.120378017425537,
      "eval_mae": 1.3313935995101929,
      "eval_mse": 3.120378017425537,
      "eval_r2": 0.05984538793563843,
      "eval_rmse": 1.7664591751369565,
      "eval_runtime": 9.0621,
      "eval_samples_per_second": 442.611,
      "eval_steps_per_second": 13.904,
      "step": 6300
    },
    {
      "epoch": 50.07976071784646,
      "grad_norm": 76.04975891113281,
      "learning_rate": 4.2753687772899825e-07,
      "loss": 22.6056,
      "step": 6310
    },
    {
      "epoch": 50.15952143569292,
      "grad_norm": 184.59645080566406,
      "learning_rate": 4.273168737213282e-07,
      "loss": 25.5011,
      "step": 6320
    },
    {
      "epoch": 50.239282153539385,
      "grad_norm": 184.94509887695312,
      "learning_rate": 4.270965930362009e-07,
      "loss": 25.8012,
      "step": 6330
    },
    {
      "epoch": 50.31904287138584,
      "grad_norm": 49.95936584472656,
      "learning_rate": 4.2687603601733256e-07,
      "loss": 26.2851,
      "step": 6340
    },
    {
      "epoch": 50.3988035892323,
      "grad_norm": 116.41128540039062,
      "learning_rate": 4.266552030088708e-07,
      "loss": 25.493,
      "step": 6350
    },
    {
      "epoch": 50.47856430707876,
      "grad_norm": 122.41911315917969,
      "learning_rate": 4.2643409435539355e-07,
      "loss": 24.7171,
      "step": 6360
    },
    {
      "epoch": 50.558325024925225,
      "grad_norm": 143.13278198242188,
      "learning_rate": 4.2621271040190934e-07,
      "loss": 25.5837,
      "step": 6370
    },
    {
      "epoch": 50.638085742771686,
      "grad_norm": 49.593753814697266,
      "learning_rate": 4.259910514938556e-07,
      "loss": 24.2556,
      "step": 6380
    },
    {
      "epoch": 50.71784646061815,
      "grad_norm": 336.3773498535156,
      "learning_rate": 4.2576911797709936e-07,
      "loss": 27.6305,
      "step": 6390
    },
    {
      "epoch": 50.79760717846461,
      "grad_norm": 80.11558532714844,
      "learning_rate": 4.255469101979359e-07,
      "loss": 25.54,
      "step": 6400
    },
    {
      "epoch": 50.79760717846461,
      "eval_loss": 3.118216037750244,
      "eval_mae": 1.333853840827942,
      "eval_mse": 3.118216037750244,
      "eval_r2": 0.06049680709838867,
      "eval_rmse": 1.7658471161882174,
      "eval_runtime": 9.1405,
      "eval_samples_per_second": 438.817,
      "eval_steps_per_second": 13.785,
      "step": 6400
    },
    {
      "epoch": 50.877367896311064,
      "grad_norm": 62.02965545654297,
      "learning_rate": 4.2532442850308846e-07,
      "loss": 23.936,
      "step": 6410
    },
    {
      "epoch": 50.957128614157526,
      "grad_norm": 92.97734069824219,
      "learning_rate": 4.2510167323970767e-07,
      "loss": 26.374,
      "step": 6420
    },
    {
      "epoch": 51.03190428713859,
      "grad_norm": 93.20958709716797,
      "learning_rate": 4.2487864475537115e-07,
      "loss": 24.8878,
      "step": 6430
    },
    {
      "epoch": 51.11166500498504,
      "grad_norm": 34.5943603515625,
      "learning_rate": 4.246553433980826e-07,
      "loss": 24.1878,
      "step": 6440
    },
    {
      "epoch": 51.191425722831504,
      "grad_norm": 58.39374923706055,
      "learning_rate": 4.2443176951627183e-07,
      "loss": 26.4275,
      "step": 6450
    },
    {
      "epoch": 51.271186440677965,
      "grad_norm": 172.89662170410156,
      "learning_rate": 4.2420792345879354e-07,
      "loss": 25.2616,
      "step": 6460
    },
    {
      "epoch": 51.35094715852443,
      "grad_norm": 86.6563720703125,
      "learning_rate": 4.239838055749273e-07,
      "loss": 26.9444,
      "step": 6470
    },
    {
      "epoch": 51.43070787637089,
      "grad_norm": 29.669885635375977,
      "learning_rate": 4.2375941621437687e-07,
      "loss": 24.9411,
      "step": 6480
    },
    {
      "epoch": 51.51046859421735,
      "grad_norm": 64.25094604492188,
      "learning_rate": 4.2353475572726955e-07,
      "loss": 22.4844,
      "step": 6490
    },
    {
      "epoch": 51.59022931206381,
      "grad_norm": 65.33932495117188,
      "learning_rate": 4.2330982446415557e-07,
      "loss": 25.193,
      "step": 6500
    },
    {
      "epoch": 51.59022931206381,
      "eval_loss": 3.115053653717041,
      "eval_mae": 1.3340669870376587,
      "eval_mse": 3.115053415298462,
      "eval_r2": 0.06144958734512329,
      "eval_rmse": 1.7649513917664876,
      "eval_runtime": 9.103,
      "eval_samples_per_second": 440.624,
      "eval_steps_per_second": 13.842,
      "step": 6500
    },
    {
      "epoch": 51.66999002991027,
      "grad_norm": 31.864547729492188,
      "learning_rate": 4.23084622776008e-07,
      "loss": 25.3146,
      "step": 6510
    },
    {
      "epoch": 51.74975074775673,
      "grad_norm": 90.17521667480469,
      "learning_rate": 4.228591510142214e-07,
      "loss": 26.0056,
      "step": 6520
    },
    {
      "epoch": 51.82951146560319,
      "grad_norm": 81.95307922363281,
      "learning_rate": 4.226334095306122e-07,
      "loss": 25.6727,
      "step": 6530
    },
    {
      "epoch": 51.90927218344965,
      "grad_norm": 328.4442138671875,
      "learning_rate": 4.224073986774175e-07,
      "loss": 25.8714,
      "step": 6540
    },
    {
      "epoch": 51.98903290129611,
      "grad_norm": 150.25486755371094,
      "learning_rate": 4.221811188072945e-07,
      "loss": 24.787,
      "step": 6550
    },
    {
      "epoch": 52.06380857427717,
      "grad_norm": 53.980255126953125,
      "learning_rate": 4.219545702733205e-07,
      "loss": 24.6424,
      "step": 6560
    },
    {
      "epoch": 52.14356929212363,
      "grad_norm": 63.03770446777344,
      "learning_rate": 4.217277534289919e-07,
      "loss": 23.153,
      "step": 6570
    },
    {
      "epoch": 52.22333000997009,
      "grad_norm": 62.21890640258789,
      "learning_rate": 4.2150066862822374e-07,
      "loss": 27.1061,
      "step": 6580
    },
    {
      "epoch": 52.30309072781655,
      "grad_norm": 62.28493118286133,
      "learning_rate": 4.212733162253491e-07,
      "loss": 25.3507,
      "step": 6590
    },
    {
      "epoch": 52.38285144566301,
      "grad_norm": 96.04239654541016,
      "learning_rate": 4.210456965751187e-07,
      "loss": 24.8468,
      "step": 6600
    },
    {
      "epoch": 52.38285144566301,
      "eval_loss": 3.11200213432312,
      "eval_mae": 1.3248567581176758,
      "eval_mse": 3.11200213432312,
      "eval_r2": 0.06236898899078369,
      "eval_rmse": 1.7640867706332135,
      "eval_runtime": 9.0953,
      "eval_samples_per_second": 440.996,
      "eval_steps_per_second": 13.853,
      "step": 6600
    },
    {
      "epoch": 52.46261216350947,
      "grad_norm": 303.4186096191406,
      "learning_rate": 4.208178100327002e-07,
      "loss": 23.6352,
      "step": 6610
    },
    {
      "epoch": 52.54237288135593,
      "grad_norm": 66.25389099121094,
      "learning_rate": 4.205896569536779e-07,
      "loss": 26.611,
      "step": 6620
    },
    {
      "epoch": 52.62213359920239,
      "grad_norm": 38.68503952026367,
      "learning_rate": 4.203612376940516e-07,
      "loss": 25.7687,
      "step": 6630
    },
    {
      "epoch": 52.701894317048854,
      "grad_norm": 391.4928894042969,
      "learning_rate": 4.201325526102369e-07,
      "loss": 26.5341,
      "step": 6640
    },
    {
      "epoch": 52.781655034895316,
      "grad_norm": 51.631919860839844,
      "learning_rate": 4.1990360205906386e-07,
      "loss": 25.1104,
      "step": 6650
    },
    {
      "epoch": 52.86141575274178,
      "grad_norm": 208.6623992919922,
      "learning_rate": 4.1967438639777676e-07,
      "loss": 25.5387,
      "step": 6660
    },
    {
      "epoch": 52.94117647058823,
      "grad_norm": 90.05245971679688,
      "learning_rate": 4.194449059840338e-07,
      "loss": 25.5684,
      "step": 6670
    },
    {
      "epoch": 53.015952143569294,
      "grad_norm": 96.74432373046875,
      "learning_rate": 4.192151611759061e-07,
      "loss": 21.4243,
      "step": 6680
    },
    {
      "epoch": 53.095712861415755,
      "grad_norm": 258.4862060546875,
      "learning_rate": 4.1898515233187737e-07,
      "loss": 26.9519,
      "step": 6690
    },
    {
      "epoch": 53.17547357926221,
      "grad_norm": 66.87359619140625,
      "learning_rate": 4.187548798108431e-07,
      "loss": 25.9816,
      "step": 6700
    },
    {
      "epoch": 53.17547357926221,
      "eval_loss": 3.109802722930908,
      "eval_mae": 1.3243112564086914,
      "eval_mse": 3.109802722930908,
      "eval_r2": 0.06303161382675171,
      "eval_rmse": 1.7634632751863328,
      "eval_runtime": 9.0994,
      "eval_samples_per_second": 440.8,
      "eval_steps_per_second": 13.847,
      "step": 6700
    },
    {
      "epoch": 53.25523429710867,
      "grad_norm": 322.5094909667969,
      "learning_rate": 4.185243439721108e-07,
      "loss": 25.8368,
      "step": 6710
    },
    {
      "epoch": 53.33499501495513,
      "grad_norm": 75.15873718261719,
      "learning_rate": 4.1829354517539804e-07,
      "loss": 24.9007,
      "step": 6720
    },
    {
      "epoch": 53.414755732801595,
      "grad_norm": 85.91637420654297,
      "learning_rate": 4.1806248378083344e-07,
      "loss": 24.7186,
      "step": 6730
    },
    {
      "epoch": 53.49451645064806,
      "grad_norm": 248.0203857421875,
      "learning_rate": 4.178311601489549e-07,
      "loss": 24.6466,
      "step": 6740
    },
    {
      "epoch": 53.57427716849452,
      "grad_norm": 68.95528411865234,
      "learning_rate": 4.1759957464070967e-07,
      "loss": 25.2751,
      "step": 6750
    },
    {
      "epoch": 53.65403788634098,
      "grad_norm": 150.67889404296875,
      "learning_rate": 4.173677276174536e-07,
      "loss": 24.6001,
      "step": 6760
    },
    {
      "epoch": 53.733798604187434,
      "grad_norm": 169.43186950683594,
      "learning_rate": 4.171356194409506e-07,
      "loss": 27.6572,
      "step": 6770
    },
    {
      "epoch": 53.813559322033896,
      "grad_norm": 97.16729736328125,
      "learning_rate": 4.1690325047337206e-07,
      "loss": 24.71,
      "step": 6780
    },
    {
      "epoch": 53.89332003988036,
      "grad_norm": 53.780696868896484,
      "learning_rate": 4.166706210772963e-07,
      "loss": 24.9079,
      "step": 6790
    },
    {
      "epoch": 53.97308075772682,
      "grad_norm": 291.2892761230469,
      "learning_rate": 4.1643773161570805e-07,
      "loss": 23.6774,
      "step": 6800
    },
    {
      "epoch": 53.97308075772682,
      "eval_loss": 3.1034927368164062,
      "eval_mae": 1.3314313888549805,
      "eval_mse": 3.1034929752349854,
      "eval_r2": 0.06493276357650757,
      "eval_rmse": 1.7616733452132904,
      "eval_runtime": 9.1214,
      "eval_samples_per_second": 439.733,
      "eval_steps_per_second": 13.814,
      "step": 6800
    },
    {
      "epoch": 54.047856430707874,
      "grad_norm": 231.18553161621094,
      "learning_rate": 4.162045824519977e-07,
      "loss": 25.4095,
      "step": 6810
    },
    {
      "epoch": 54.127617148554336,
      "grad_norm": 223.30455017089844,
      "learning_rate": 4.1597117394996117e-07,
      "loss": 27.0908,
      "step": 6820
    },
    {
      "epoch": 54.2073778664008,
      "grad_norm": 106.4852066040039,
      "learning_rate": 4.1573750647379857e-07,
      "loss": 25.5567,
      "step": 6830
    },
    {
      "epoch": 54.28713858424726,
      "grad_norm": 215.9276580810547,
      "learning_rate": 4.155035803881145e-07,
      "loss": 24.7604,
      "step": 6840
    },
    {
      "epoch": 54.36689930209372,
      "grad_norm": 135.57931518554688,
      "learning_rate": 4.1526939605791695e-07,
      "loss": 25.3572,
      "step": 6850
    },
    {
      "epoch": 54.44666001994018,
      "grad_norm": 117.22618103027344,
      "learning_rate": 4.1503495384861674e-07,
      "loss": 24.2714,
      "step": 6860
    },
    {
      "epoch": 54.52642073778664,
      "grad_norm": 114.61555480957031,
      "learning_rate": 4.148002541260274e-07,
      "loss": 25.8058,
      "step": 6870
    },
    {
      "epoch": 54.6061814556331,
      "grad_norm": 119.7732162475586,
      "learning_rate": 4.145652972563638e-07,
      "loss": 22.5095,
      "step": 6880
    },
    {
      "epoch": 54.68594217347956,
      "grad_norm": 135.6663360595703,
      "learning_rate": 4.143300836062425e-07,
      "loss": 26.633,
      "step": 6890
    },
    {
      "epoch": 54.76570289132602,
      "grad_norm": 99.55790710449219,
      "learning_rate": 4.140946135426805e-07,
      "loss": 24.2669,
      "step": 6900
    },
    {
      "epoch": 54.76570289132602,
      "eval_loss": 3.099317789077759,
      "eval_mae": 1.330242395401001,
      "eval_mse": 3.099317789077759,
      "eval_r2": 0.06619071960449219,
      "eval_rmse": 1.7604879406226441,
      "eval_runtime": 9.076,
      "eval_samples_per_second": 441.932,
      "eval_steps_per_second": 13.883,
      "step": 6900
    },
    {
      "epoch": 54.845463609172484,
      "grad_norm": 70.54505920410156,
      "learning_rate": 4.1385888743309495e-07,
      "loss": 25.2153,
      "step": 6910
    },
    {
      "epoch": 54.925224327018945,
      "grad_norm": 105.45303344726562,
      "learning_rate": 4.1362290564530244e-07,
      "loss": 25.4413,
      "step": 6920
    },
    {
      "epoch": 55.0,
      "grad_norm": 109.5523681640625,
      "learning_rate": 4.133866685475187e-07,
      "loss": 22.7176,
      "step": 6930
    },
    {
      "epoch": 55.07976071784646,
      "grad_norm": 120.8057861328125,
      "learning_rate": 4.131501765083576e-07,
      "loss": 23.361,
      "step": 6940
    },
    {
      "epoch": 55.15952143569292,
      "grad_norm": 89.11048126220703,
      "learning_rate": 4.12913429896831e-07,
      "loss": 24.928,
      "step": 6950
    },
    {
      "epoch": 55.239282153539385,
      "grad_norm": 123.72354888916016,
      "learning_rate": 4.126764290823479e-07,
      "loss": 25.3552,
      "step": 6960
    },
    {
      "epoch": 55.31904287138584,
      "grad_norm": 71.4775161743164,
      "learning_rate": 4.1243917443471385e-07,
      "loss": 25.8639,
      "step": 6970
    },
    {
      "epoch": 55.3988035892323,
      "grad_norm": 179.25799560546875,
      "learning_rate": 4.122016663241307e-07,
      "loss": 25.6156,
      "step": 6980
    },
    {
      "epoch": 55.47856430707876,
      "grad_norm": 33.811058044433594,
      "learning_rate": 4.119639051211957e-07,
      "loss": 24.9542,
      "step": 6990
    },
    {
      "epoch": 55.558325024925225,
      "grad_norm": 171.1689910888672,
      "learning_rate": 4.117258911969009e-07,
      "loss": 25.5526,
      "step": 7000
    },
    {
      "epoch": 55.558325024925225,
      "eval_loss": 3.0987918376922607,
      "eval_mae": 1.3330014944076538,
      "eval_mse": 3.0987918376922607,
      "eval_r2": 0.06634920835494995,
      "eval_rmse": 1.760338557690611,
      "eval_runtime": 9.0725,
      "eval_samples_per_second": 442.105,
      "eval_steps_per_second": 13.888,
      "step": 7000
    },
    {
      "epoch": 55.638085742771686,
      "grad_norm": 92.38328552246094,
      "learning_rate": 4.114876249226327e-07,
      "loss": 26.2241,
      "step": 7010
    },
    {
      "epoch": 55.71784646061815,
      "grad_norm": 348.17333984375,
      "learning_rate": 4.112491066701715e-07,
      "loss": 27.2409,
      "step": 7020
    },
    {
      "epoch": 55.79760717846461,
      "grad_norm": 254.02542114257812,
      "learning_rate": 4.110103368116906e-07,
      "loss": 24.6501,
      "step": 7030
    },
    {
      "epoch": 55.877367896311064,
      "grad_norm": 219.8361358642578,
      "learning_rate": 4.1077131571975595e-07,
      "loss": 26.0709,
      "step": 7040
    },
    {
      "epoch": 55.957128614157526,
      "grad_norm": 98.12411499023438,
      "learning_rate": 4.1053204376732557e-07,
      "loss": 24.2047,
      "step": 7050
    },
    {
      "epoch": 56.03190428713859,
      "grad_norm": 366.5306396484375,
      "learning_rate": 4.1029252132774907e-07,
      "loss": 23.0486,
      "step": 7060
    },
    {
      "epoch": 56.11166500498504,
      "grad_norm": 176.13414001464844,
      "learning_rate": 4.100527487747665e-07,
      "loss": 25.716,
      "step": 7070
    },
    {
      "epoch": 56.191425722831504,
      "grad_norm": 84.32244873046875,
      "learning_rate": 4.0981272648250854e-07,
      "loss": 27.4791,
      "step": 7080
    },
    {
      "epoch": 56.271186440677965,
      "grad_norm": 140.40318298339844,
      "learning_rate": 4.0957245482549543e-07,
      "loss": 24.0629,
      "step": 7090
    },
    {
      "epoch": 56.35094715852443,
      "grad_norm": 234.8149871826172,
      "learning_rate": 4.0933193417863644e-07,
      "loss": 25.9714,
      "step": 7100
    },
    {
      "epoch": 56.35094715852443,
      "eval_loss": 3.095909833908081,
      "eval_mae": 1.3204256296157837,
      "eval_mse": 3.09591007232666,
      "eval_r2": 0.06721740961074829,
      "eval_rmse": 1.7595198414131794,
      "eval_runtime": 9.086,
      "eval_samples_per_second": 441.45,
      "eval_steps_per_second": 13.868,
      "step": 7100
    },
    {
      "epoch": 56.43070787637089,
      "grad_norm": 65.00196838378906,
      "learning_rate": 4.0909116491722943e-07,
      "loss": 23.4879,
      "step": 7110
    },
    {
      "epoch": 56.51046859421735,
      "grad_norm": 197.07687377929688,
      "learning_rate": 4.088501474169603e-07,
      "loss": 25.5584,
      "step": 7120
    },
    {
      "epoch": 56.59022931206381,
      "grad_norm": 60.80377197265625,
      "learning_rate": 4.0860888205390197e-07,
      "loss": 26.0502,
      "step": 7130
    },
    {
      "epoch": 56.66999002991027,
      "grad_norm": 110.6761703491211,
      "learning_rate": 4.083673692045145e-07,
      "loss": 25.3108,
      "step": 7140
    },
    {
      "epoch": 56.74975074775673,
      "grad_norm": 116.65326690673828,
      "learning_rate": 4.0814979635070945e-07,
      "loss": 23.8434,
      "step": 7150
    },
    {
      "epoch": 56.82951146560319,
      "grad_norm": 84.38224029541016,
      "learning_rate": 4.079078143158259e-07,
      "loss": 23.5383,
      "step": 7160
    },
    {
      "epoch": 56.90927218344965,
      "grad_norm": 151.78848266601562,
      "learning_rate": 4.076655858885284e-07,
      "loss": 28.0889,
      "step": 7170
    },
    {
      "epoch": 56.98903290129611,
      "grad_norm": 271.9244689941406,
      "learning_rate": 4.0742311144677964e-07,
      "loss": 22.9197,
      "step": 7180
    },
    {
      "epoch": 57.06380857427717,
      "grad_norm": 108.05704498291016,
      "learning_rate": 4.071803913689259e-07,
      "loss": 21.7839,
      "step": 7190
    },
    {
      "epoch": 57.14356929212363,
      "grad_norm": 219.2751922607422,
      "learning_rate": 4.0693742603369703e-07,
      "loss": 26.7117,
      "step": 7200
    },
    {
      "epoch": 57.14356929212363,
      "eval_loss": 3.090022087097168,
      "eval_mae": 1.3286364078521729,
      "eval_mse": 3.0900213718414307,
      "eval_r2": 0.06899166107177734,
      "eval_rmse": 1.7578456621220848,
      "eval_runtime": 9.0823,
      "eval_samples_per_second": 441.628,
      "eval_steps_per_second": 13.873,
      "step": 7200
    },
    {
      "epoch": 57.22333000997009,
      "grad_norm": 154.8135986328125,
      "learning_rate": 4.0669421582020544e-07,
      "loss": 26.9373,
      "step": 7210
    },
    {
      "epoch": 57.30309072781655,
      "grad_norm": 94.78240203857422,
      "learning_rate": 4.0645076110794566e-07,
      "loss": 24.1204,
      "step": 7220
    },
    {
      "epoch": 57.38285144566301,
      "grad_norm": 147.9627227783203,
      "learning_rate": 4.062070622767938e-07,
      "loss": 25.5562,
      "step": 7230
    },
    {
      "epoch": 57.46261216350947,
      "grad_norm": 209.6161651611328,
      "learning_rate": 4.0596311970700665e-07,
      "loss": 24.3868,
      "step": 7240
    },
    {
      "epoch": 57.54237288135593,
      "grad_norm": 72.75938415527344,
      "learning_rate": 4.057189337792216e-07,
      "loss": 25.9237,
      "step": 7250
    },
    {
      "epoch": 57.62213359920239,
      "grad_norm": 241.09524536132812,
      "learning_rate": 4.0547450487445556e-07,
      "loss": 24.6371,
      "step": 7260
    },
    {
      "epoch": 57.701894317048854,
      "grad_norm": 136.7520294189453,
      "learning_rate": 4.0522983337410477e-07,
      "loss": 25.3354,
      "step": 7270
    },
    {
      "epoch": 57.781655034895316,
      "grad_norm": 203.94277954101562,
      "learning_rate": 4.0498491965994377e-07,
      "loss": 24.1224,
      "step": 7280
    },
    {
      "epoch": 57.86141575274178,
      "grad_norm": 61.92915725708008,
      "learning_rate": 4.047397641141252e-07,
      "loss": 25.7716,
      "step": 7290
    },
    {
      "epoch": 57.94117647058823,
      "grad_norm": 59.817298889160156,
      "learning_rate": 4.0449436711917897e-07,
      "loss": 24.8604,
      "step": 7300
    },
    {
      "epoch": 57.94117647058823,
      "eval_loss": 3.088762044906616,
      "eval_mae": 1.318792462348938,
      "eval_mse": 3.088761806488037,
      "eval_r2": 0.06937116384506226,
      "eval_rmse": 1.7574873559966333,
      "eval_runtime": 9.0877,
      "eval_samples_per_second": 441.366,
      "eval_steps_per_second": 13.865,
      "step": 7300
    },
    {
      "epoch": 58.015952143569294,
      "grad_norm": 94.05938720703125,
      "learning_rate": 4.042487290580118e-07,
      "loss": 23.8257,
      "step": 7310
    },
    {
      "epoch": 58.095712861415755,
      "grad_norm": 54.836669921875,
      "learning_rate": 4.040028503139065e-07,
      "loss": 27.0251,
      "step": 7320
    },
    {
      "epoch": 58.17547357926221,
      "grad_norm": 150.6589813232422,
      "learning_rate": 4.037567312705216e-07,
      "loss": 24.3402,
      "step": 7330
    },
    {
      "epoch": 58.25523429710867,
      "grad_norm": 400.803955078125,
      "learning_rate": 4.035103723118901e-07,
      "loss": 23.3259,
      "step": 7340
    },
    {
      "epoch": 58.33499501495513,
      "grad_norm": 128.96749877929688,
      "learning_rate": 4.032637738224201e-07,
      "loss": 24.5398,
      "step": 7350
    },
    {
      "epoch": 58.414755732801595,
      "grad_norm": 235.96865844726562,
      "learning_rate": 4.030169361868927e-07,
      "loss": 23.7088,
      "step": 7360
    },
    {
      "epoch": 58.49451645064806,
      "grad_norm": 75.51648712158203,
      "learning_rate": 4.0276985979046276e-07,
      "loss": 27.9453,
      "step": 7370
    },
    {
      "epoch": 58.57427716849452,
      "grad_norm": 84.60974884033203,
      "learning_rate": 4.025225450186573e-07,
      "loss": 24.5951,
      "step": 7380
    },
    {
      "epoch": 58.65403788634098,
      "grad_norm": 157.3654022216797,
      "learning_rate": 4.0227499225737537e-07,
      "loss": 25.4845,
      "step": 7390
    },
    {
      "epoch": 58.733798604187434,
      "grad_norm": 164.58987426757812,
      "learning_rate": 4.020272018928876e-07,
      "loss": 21.6683,
      "step": 7400
    },
    {
      "epoch": 58.733798604187434,
      "eval_loss": 3.0835020542144775,
      "eval_mae": 1.3220032453536987,
      "eval_mse": 3.0835020542144775,
      "eval_r2": 0.07095593214035034,
      "eval_rmse": 1.7559903343169283,
      "eval_runtime": 9.0819,
      "eval_samples_per_second": 441.648,
      "eval_steps_per_second": 13.874,
      "step": 7400
    },
    {
      "epoch": 58.813559322033896,
      "grad_norm": 92.88033294677734,
      "learning_rate": 4.0177917431183503e-07,
      "loss": 27.4979,
      "step": 7410
    },
    {
      "epoch": 58.89332003988036,
      "grad_norm": 404.1705322265625,
      "learning_rate": 4.0153090990122906e-07,
      "loss": 23.6773,
      "step": 7420
    },
    {
      "epoch": 58.97308075772682,
      "grad_norm": 193.15248107910156,
      "learning_rate": 4.0128240904845056e-07,
      "loss": 26.0743,
      "step": 7430
    },
    {
      "epoch": 59.047856430707874,
      "grad_norm": 395.1861877441406,
      "learning_rate": 4.0103367214124927e-07,
      "loss": 26.1012,
      "step": 7440
    },
    {
      "epoch": 59.127617148554336,
      "grad_norm": 56.67768859863281,
      "learning_rate": 4.007846995677434e-07,
      "loss": 25.2291,
      "step": 7450
    },
    {
      "epoch": 59.2073778664008,
      "grad_norm": 194.84588623046875,
      "learning_rate": 4.005354917164189e-07,
      "loss": 24.8622,
      "step": 7460
    },
    {
      "epoch": 59.28713858424726,
      "grad_norm": 164.9954833984375,
      "learning_rate": 4.0028604897612855e-07,
      "loss": 24.4034,
      "step": 7470
    },
    {
      "epoch": 59.36689930209372,
      "grad_norm": 91.07857513427734,
      "learning_rate": 4.0003637173609195e-07,
      "loss": 25.145,
      "step": 7480
    },
    {
      "epoch": 59.44666001994018,
      "grad_norm": 255.81103515625,
      "learning_rate": 3.9978646038589453e-07,
      "loss": 26.5347,
      "step": 7490
    },
    {
      "epoch": 59.52642073778664,
      "grad_norm": 91.14227294921875,
      "learning_rate": 3.995363153154869e-07,
      "loss": 25.1552,
      "step": 7500
    },
    {
      "epoch": 59.52642073778664,
      "eval_loss": 3.0817205905914307,
      "eval_mae": 1.3204900026321411,
      "eval_mse": 3.0817203521728516,
      "eval_r2": 0.07149273157119751,
      "eval_rmse": 1.7554829398694969,
      "eval_runtime": 9.1094,
      "eval_samples_per_second": 440.315,
      "eval_steps_per_second": 13.832,
      "step": 7500
    },
    {
      "epoch": 59.6061814556331,
      "grad_norm": 278.22918701171875,
      "learning_rate": 3.9928593691518453e-07,
      "loss": 26.0224,
      "step": 7510
    },
    {
      "epoch": 59.68594217347956,
      "grad_norm": 78.10284423828125,
      "learning_rate": 3.990353255756667e-07,
      "loss": 23.5015,
      "step": 7520
    },
    {
      "epoch": 59.76570289132602,
      "grad_norm": 85.6133041381836,
      "learning_rate": 3.987844816879765e-07,
      "loss": 22.773,
      "step": 7530
    },
    {
      "epoch": 59.845463609172484,
      "grad_norm": 130.89801025390625,
      "learning_rate": 3.9853340564351965e-07,
      "loss": 25.6739,
      "step": 7540
    },
    {
      "epoch": 59.925224327018945,
      "grad_norm": 308.4039306640625,
      "learning_rate": 3.982820978340642e-07,
      "loss": 24.2176,
      "step": 7550
    },
    {
      "epoch": 60.0,
      "grad_norm": 216.79660034179688,
      "learning_rate": 3.9803055865173974e-07,
      "loss": 25.2403,
      "step": 7560
    },
    {
      "epoch": 60.07976071784646,
      "grad_norm": 84.96922302246094,
      "learning_rate": 3.97778788489037e-07,
      "loss": 25.8891,
      "step": 7570
    },
    {
      "epoch": 60.15952143569292,
      "grad_norm": 169.84339904785156,
      "learning_rate": 3.975267877388071e-07,
      "loss": 24.7238,
      "step": 7580
    },
    {
      "epoch": 60.239282153539385,
      "grad_norm": 49.10818862915039,
      "learning_rate": 3.972745567942609e-07,
      "loss": 26.6454,
      "step": 7590
    },
    {
      "epoch": 60.31904287138584,
      "grad_norm": 108.38031005859375,
      "learning_rate": 3.9702209604896853e-07,
      "loss": 24.8128,
      "step": 7600
    },
    {
      "epoch": 60.31904287138584,
      "eval_loss": 3.0798652172088623,
      "eval_mae": 1.3183189630508423,
      "eval_mse": 3.0798652172088623,
      "eval_r2": 0.07205164432525635,
      "eval_rmse": 1.754954477246878,
      "eval_runtime": 9.1086,
      "eval_samples_per_second": 440.353,
      "eval_steps_per_second": 13.833,
      "step": 7600
    },
    {
      "epoch": 60.3988035892323,
      "grad_norm": 38.540428161621094,
      "learning_rate": 3.967694058968586e-07,
      "loss": 26.0966,
      "step": 7610
    },
    {
      "epoch": 60.47856430707876,
      "grad_norm": 88.73666381835938,
      "learning_rate": 3.965164867322177e-07,
      "loss": 22.8718,
      "step": 7620
    },
    {
      "epoch": 60.558325024925225,
      "grad_norm": 106.24761199951172,
      "learning_rate": 3.9626333894968986e-07,
      "loss": 26.8745,
      "step": 7630
    },
    {
      "epoch": 60.638085742771686,
      "grad_norm": 251.55950927734375,
      "learning_rate": 3.960099629442757e-07,
      "loss": 24.539,
      "step": 7640
    },
    {
      "epoch": 60.71784646061815,
      "grad_norm": 125.47510528564453,
      "learning_rate": 3.957563591113319e-07,
      "loss": 24.5211,
      "step": 7650
    },
    {
      "epoch": 60.79760717846461,
      "grad_norm": 29.564306259155273,
      "learning_rate": 3.9550252784657097e-07,
      "loss": 24.385,
      "step": 7660
    },
    {
      "epoch": 60.877367896311064,
      "grad_norm": 111.65841674804688,
      "learning_rate": 3.9524846954605985e-07,
      "loss": 24.8668,
      "step": 7670
    },
    {
      "epoch": 60.957128614157526,
      "grad_norm": 300.87506103515625,
      "learning_rate": 3.9499418460622e-07,
      "loss": 25.2645,
      "step": 7680
    },
    {
      "epoch": 61.03190428713859,
      "grad_norm": 47.71473693847656,
      "learning_rate": 3.9473967342382657e-07,
      "loss": 24.3245,
      "step": 7690
    },
    {
      "epoch": 61.11166500498504,
      "grad_norm": 217.0494384765625,
      "learning_rate": 3.944849363960075e-07,
      "loss": 24.9409,
      "step": 7700
    },
    {
      "epoch": 61.11166500498504,
      "eval_loss": 3.0771336555480957,
      "eval_mae": 1.3210324048995972,
      "eval_mse": 3.0771336555480957,
      "eval_r2": 0.07287466526031494,
      "eval_rmse": 1.754176061730434,
      "eval_runtime": 9.1043,
      "eval_samples_per_second": 440.562,
      "eval_steps_per_second": 13.84,
      "step": 7700
    },
    {
      "epoch": 61.191425722831504,
      "grad_norm": 127.66291809082031,
      "learning_rate": 3.942299739202434e-07,
      "loss": 25.6905,
      "step": 7710
    },
    {
      "epoch": 61.271186440677965,
      "grad_norm": 89.4798812866211,
      "learning_rate": 3.9397478639436633e-07,
      "loss": 23.6006,
      "step": 7720
    },
    {
      "epoch": 61.35094715852443,
      "grad_norm": 42.74026870727539,
      "learning_rate": 3.937193742165598e-07,
      "loss": 25.0052,
      "step": 7730
    },
    {
      "epoch": 61.43070787637089,
      "grad_norm": 358.45074462890625,
      "learning_rate": 3.9346373778535776e-07,
      "loss": 25.0835,
      "step": 7740
    },
    {
      "epoch": 61.51046859421735,
      "grad_norm": 209.05079650878906,
      "learning_rate": 3.9320787749964397e-07,
      "loss": 26.9053,
      "step": 7750
    },
    {
      "epoch": 61.59022931206381,
      "grad_norm": 123.0264663696289,
      "learning_rate": 3.929517937586516e-07,
      "loss": 24.7298,
      "step": 7760
    },
    {
      "epoch": 61.66999002991027,
      "grad_norm": 431.9937744140625,
      "learning_rate": 3.926954869619625e-07,
      "loss": 24.2174,
      "step": 7770
    },
    {
      "epoch": 61.74975074775673,
      "grad_norm": 199.51316833496094,
      "learning_rate": 3.9243895750950664e-07,
      "loss": 23.6582,
      "step": 7780
    },
    {
      "epoch": 61.82951146560319,
      "grad_norm": 459.22894287109375,
      "learning_rate": 3.9218220580156116e-07,
      "loss": 24.6488,
      "step": 7790
    },
    {
      "epoch": 61.90927218344965,
      "grad_norm": 65.73253631591797,
      "learning_rate": 3.9192523223875006e-07,
      "loss": 25.5879,
      "step": 7800
    },
    {
      "epoch": 61.90927218344965,
      "eval_loss": 3.0825397968292236,
      "eval_mae": 1.3149493932724,
      "eval_mse": 3.0825395584106445,
      "eval_r2": 0.07124590873718262,
      "eval_rmse": 1.7557162522488208,
      "eval_runtime": 9.0657,
      "eval_samples_per_second": 442.435,
      "eval_steps_per_second": 13.898,
      "step": 7800
    },
    {
      "epoch": 61.98903290129611,
      "grad_norm": 253.05636596679688,
      "learning_rate": 3.916680372220438e-07,
      "loss": 25.8713,
      "step": 7810
    },
    {
      "epoch": 62.06380857427717,
      "grad_norm": 163.3178253173828,
      "learning_rate": 3.914106211527581e-07,
      "loss": 24.4312,
      "step": 7820
    },
    {
      "epoch": 62.14356929212363,
      "grad_norm": 317.12725830078125,
      "learning_rate": 3.911529844325536e-07,
      "loss": 26.369,
      "step": 7830
    },
    {
      "epoch": 62.22333000997009,
      "grad_norm": 71.54997253417969,
      "learning_rate": 3.908951274634355e-07,
      "loss": 25.9222,
      "step": 7840
    },
    {
      "epoch": 62.30309072781655,
      "grad_norm": 171.40684509277344,
      "learning_rate": 3.9063705064775225e-07,
      "loss": 25.9971,
      "step": 7850
    },
    {
      "epoch": 62.38285144566301,
      "grad_norm": 223.0928192138672,
      "learning_rate": 3.903787543881958e-07,
      "loss": 24.2405,
      "step": 7860
    },
    {
      "epoch": 62.46261216350947,
      "grad_norm": 304.8570556640625,
      "learning_rate": 3.901202390878001e-07,
      "loss": 24.8631,
      "step": 7870
    },
    {
      "epoch": 62.54237288135593,
      "grad_norm": 351.0543212890625,
      "learning_rate": 3.898615051499412e-07,
      "loss": 26.5244,
      "step": 7880
    },
    {
      "epoch": 62.62213359920239,
      "grad_norm": 60.6923828125,
      "learning_rate": 3.89602552978336e-07,
      "loss": 25.2301,
      "step": 7890
    },
    {
      "epoch": 62.701894317048854,
      "grad_norm": 103.87010955810547,
      "learning_rate": 3.893433829770423e-07,
      "loss": 22.7804,
      "step": 7900
    },
    {
      "epoch": 62.701894317048854,
      "eval_loss": 3.075007200241089,
      "eval_mae": 1.3185298442840576,
      "eval_mse": 3.075007200241089,
      "eval_r2": 0.07351541519165039,
      "eval_rmse": 1.7535698446999735,
      "eval_runtime": 9.0851,
      "eval_samples_per_second": 441.493,
      "eval_steps_per_second": 13.869,
      "step": 7900
    },
    {
      "epoch": 62.781655034895316,
      "grad_norm": 148.22166442871094,
      "learning_rate": 3.8908399555045746e-07,
      "loss": 24.4337,
      "step": 7910
    },
    {
      "epoch": 62.86141575274178,
      "grad_norm": 56.29420471191406,
      "learning_rate": 3.888243911033182e-07,
      "loss": 24.513,
      "step": 7920
    },
    {
      "epoch": 62.94117647058823,
      "grad_norm": 693.5757446289062,
      "learning_rate": 3.8856457004069987e-07,
      "loss": 25.2368,
      "step": 7930
    },
    {
      "epoch": 63.015952143569294,
      "grad_norm": 330.87225341796875,
      "learning_rate": 3.8830453276801603e-07,
      "loss": 22.621,
      "step": 7940
    },
    {
      "epoch": 63.095712861415755,
      "grad_norm": 104.25527954101562,
      "learning_rate": 3.8804427969101714e-07,
      "loss": 24.8392,
      "step": 7950
    },
    {
      "epoch": 63.17547357926221,
      "grad_norm": 59.79703140258789,
      "learning_rate": 3.8778381121579086e-07,
      "loss": 24.2671,
      "step": 7960
    },
    {
      "epoch": 63.25523429710867,
      "grad_norm": 186.5470733642578,
      "learning_rate": 3.8752312774876075e-07,
      "loss": 25.6232,
      "step": 7970
    },
    {
      "epoch": 63.33499501495513,
      "grad_norm": 264.2695617675781,
      "learning_rate": 3.8726222969668563e-07,
      "loss": 25.2552,
      "step": 7980
    },
    {
      "epoch": 63.414755732801595,
      "grad_norm": 68.55497741699219,
      "learning_rate": 3.8700111746665966e-07,
      "loss": 23.5333,
      "step": 7990
    },
    {
      "epoch": 63.49451645064806,
      "grad_norm": 189.8446044921875,
      "learning_rate": 3.8673979146611065e-07,
      "loss": 25.5833,
      "step": 8000
    },
    {
      "epoch": 63.49451645064806,
      "eval_loss": 3.0853989124298096,
      "eval_mae": 1.312778115272522,
      "eval_mse": 3.0853989124298096,
      "eval_r2": 0.07038438320159912,
      "eval_rmse": 1.756530361943627,
      "eval_runtime": 9.0969,
      "eval_samples_per_second": 440.922,
      "eval_steps_per_second": 13.851,
      "step": 8000
    },
    {
      "epoch": 63.57427716849452,
      "grad_norm": 71.05472564697266,
      "learning_rate": 3.864782521028003e-07,
      "loss": 25.5786,
      "step": 8010
    },
    {
      "epoch": 63.65403788634098,
      "grad_norm": 308.1141662597656,
      "learning_rate": 3.862164997848231e-07,
      "loss": 24.9246,
      "step": 8020
    },
    {
      "epoch": 63.733798604187434,
      "grad_norm": 233.69918823242188,
      "learning_rate": 3.8595453492060593e-07,
      "loss": 26.6499,
      "step": 8030
    },
    {
      "epoch": 63.813559322033896,
      "grad_norm": 284.7898864746094,
      "learning_rate": 3.856923579189072e-07,
      "loss": 22.7096,
      "step": 8040
    },
    {
      "epoch": 63.89332003988036,
      "grad_norm": 46.84286880493164,
      "learning_rate": 3.854299691888164e-07,
      "loss": 26.9125,
      "step": 8050
    },
    {
      "epoch": 63.97308075772682,
      "grad_norm": 390.2908630371094,
      "learning_rate": 3.8516736913975333e-07,
      "loss": 26.2214,
      "step": 8060
    },
    {
      "epoch": 64.04785643070788,
      "grad_norm": 244.6831512451172,
      "learning_rate": 3.8490455818146763e-07,
      "loss": 22.5445,
      "step": 8070
    },
    {
      "epoch": 64.12761714855434,
      "grad_norm": 119.25518035888672,
      "learning_rate": 3.84641536724038e-07,
      "loss": 23.0345,
      "step": 8080
    },
    {
      "epoch": 64.2073778664008,
      "grad_norm": 76.47522735595703,
      "learning_rate": 3.843783051778714e-07,
      "loss": 25.3076,
      "step": 8090
    },
    {
      "epoch": 64.28713858424726,
      "grad_norm": 264.4443664550781,
      "learning_rate": 3.841148639537029e-07,
      "loss": 25.8505,
      "step": 8100
    },
    {
      "epoch": 64.28713858424726,
      "eval_loss": 3.0770719051361084,
      "eval_mae": 1.3144989013671875,
      "eval_mse": 3.0770719051361084,
      "eval_r2": 0.07289326190948486,
      "eval_rmse": 1.7541584606688496,
      "eval_runtime": 9.0998,
      "eval_samples_per_second": 440.78,
      "eval_steps_per_second": 13.847,
      "step": 8100
    },
    {
      "epoch": 64.36689930209371,
      "grad_norm": 404.09783935546875,
      "learning_rate": 3.8385121346259465e-07,
      "loss": 23.8394,
      "step": 8110
    },
    {
      "epoch": 64.44666001994018,
      "grad_norm": 122.11978149414062,
      "learning_rate": 3.8358735411593514e-07,
      "loss": 25.0254,
      "step": 8120
    },
    {
      "epoch": 64.52642073778664,
      "grad_norm": 140.75172424316406,
      "learning_rate": 3.8332328632543903e-07,
      "loss": 26.4085,
      "step": 8130
    },
    {
      "epoch": 64.6061814556331,
      "grad_norm": 115.68319702148438,
      "learning_rate": 3.8305901050314604e-07,
      "loss": 25.4538,
      "step": 8140
    },
    {
      "epoch": 64.68594217347956,
      "grad_norm": 91.7467041015625,
      "learning_rate": 3.827945270614205e-07,
      "loss": 26.4521,
      "step": 8150
    },
    {
      "epoch": 64.76570289132601,
      "grad_norm": 192.41465759277344,
      "learning_rate": 3.825298364129509e-07,
      "loss": 24.4931,
      "step": 8160
    },
    {
      "epoch": 64.84546360917248,
      "grad_norm": 183.51048278808594,
      "learning_rate": 3.822649389707487e-07,
      "loss": 24.4489,
      "step": 8170
    },
    {
      "epoch": 64.92522432701894,
      "grad_norm": 93.11408233642578,
      "learning_rate": 3.819998351481484e-07,
      "loss": 24.1044,
      "step": 8180
    },
    {
      "epoch": 65.0,
      "grad_norm": 95.1280746459961,
      "learning_rate": 3.8173452535880625e-07,
      "loss": 23.9528,
      "step": 8190
    },
    {
      "epoch": 65.07976071784645,
      "grad_norm": 113.3239974975586,
      "learning_rate": 3.814690100167e-07,
      "loss": 24.4158,
      "step": 8200
    },
    {
      "epoch": 65.07976071784645,
      "eval_loss": 3.0694262981414795,
      "eval_mae": 1.322281002998352,
      "eval_mse": 3.0694262981414795,
      "eval_r2": 0.07519686222076416,
      "eval_rmse": 1.751977824671728,
      "eval_runtime": 9.0964,
      "eval_samples_per_second": 440.944,
      "eval_steps_per_second": 13.852,
      "step": 8200
    },
    {
      "epoch": 65.15952143569292,
      "grad_norm": 130.2467041015625,
      "learning_rate": 3.812032895361282e-07,
      "loss": 25.5138,
      "step": 8210
    },
    {
      "epoch": 65.23928215353938,
      "grad_norm": 43.810115814208984,
      "learning_rate": 3.809373643317093e-07,
      "loss": 27.3361,
      "step": 8220
    },
    {
      "epoch": 65.31904287138585,
      "grad_norm": 40.04730224609375,
      "learning_rate": 3.806712348183813e-07,
      "loss": 24.0978,
      "step": 8230
    },
    {
      "epoch": 65.3988035892323,
      "grad_norm": 110.29949188232422,
      "learning_rate": 3.8040490141140113e-07,
      "loss": 23.6178,
      "step": 8240
    },
    {
      "epoch": 65.47856430707877,
      "grad_norm": 298.765625,
      "learning_rate": 3.801383645263437e-07,
      "loss": 24.6665,
      "step": 8250
    },
    {
      "epoch": 65.55832502492522,
      "grad_norm": 292.2880859375,
      "learning_rate": 3.7987162457910127e-07,
      "loss": 24.6888,
      "step": 8260
    },
    {
      "epoch": 65.63808574277168,
      "grad_norm": 221.99192810058594,
      "learning_rate": 3.796046819858834e-07,
      "loss": 25.0066,
      "step": 8270
    },
    {
      "epoch": 65.71784646061815,
      "grad_norm": 441.9041442871094,
      "learning_rate": 3.7933753716321533e-07,
      "loss": 25.6156,
      "step": 8280
    },
    {
      "epoch": 65.7976071784646,
      "grad_norm": 149.19468688964844,
      "learning_rate": 3.790701905279383e-07,
      "loss": 24.9847,
      "step": 8290
    },
    {
      "epoch": 65.87736789631107,
      "grad_norm": 74.34953308105469,
      "learning_rate": 3.7880264249720805e-07,
      "loss": 25.2725,
      "step": 8300
    },
    {
      "epoch": 65.87736789631107,
      "eval_loss": 3.069687843322754,
      "eval_mae": 1.3177374601364136,
      "eval_mse": 3.069688081741333,
      "eval_r2": 0.07511800527572632,
      "eval_rmse": 1.7520525339559123,
      "eval_runtime": 9.0827,
      "eval_samples_per_second": 441.611,
      "eval_steps_per_second": 13.873,
      "step": 8300
    },
    {
      "epoch": 65.95712861415753,
      "grad_norm": 209.19964599609375,
      "learning_rate": 3.78534893488495e-07,
      "loss": 25.0926,
      "step": 8310
    },
    {
      "epoch": 66.03190428713859,
      "grad_norm": 314.91265869140625,
      "learning_rate": 3.782669439195828e-07,
      "loss": 22.914,
      "step": 8320
    },
    {
      "epoch": 66.11166500498504,
      "grad_norm": 370.3639221191406,
      "learning_rate": 3.7799879420856825e-07,
      "loss": 25.2253,
      "step": 8330
    },
    {
      "epoch": 66.19142572283151,
      "grad_norm": 121.62500762939453,
      "learning_rate": 3.777304447738602e-07,
      "loss": 26.6113,
      "step": 8340
    },
    {
      "epoch": 66.27118644067797,
      "grad_norm": 178.5107421875,
      "learning_rate": 3.774618960341796e-07,
      "loss": 23.7175,
      "step": 8350
    },
    {
      "epoch": 66.35094715852442,
      "grad_norm": 282.17498779296875,
      "learning_rate": 3.7719314840855786e-07,
      "loss": 27.1242,
      "step": 8360
    },
    {
      "epoch": 66.43070787637089,
      "grad_norm": 77.5106430053711,
      "learning_rate": 3.7692420231633707e-07,
      "loss": 23.654,
      "step": 8370
    },
    {
      "epoch": 66.51046859421734,
      "grad_norm": 481.6351013183594,
      "learning_rate": 3.766550581771689e-07,
      "loss": 24.4763,
      "step": 8380
    },
    {
      "epoch": 66.59022931206381,
      "grad_norm": 571.66650390625,
      "learning_rate": 3.76385716411014e-07,
      "loss": 24.758,
      "step": 8390
    },
    {
      "epoch": 66.66999002991027,
      "grad_norm": 385.8674011230469,
      "learning_rate": 3.761161774381414e-07,
      "loss": 24.5836,
      "step": 8400
    },
    {
      "epoch": 66.66999002991027,
      "eval_loss": 3.07013201713562,
      "eval_mae": 1.3153153657913208,
      "eval_mse": 3.070132255554199,
      "eval_r2": 0.07498413324356079,
      "eval_rmse": 1.7521792875029083,
      "eval_runtime": 9.061,
      "eval_samples_per_second": 442.668,
      "eval_steps_per_second": 13.906,
      "step": 8400
    },
    {
      "epoch": 66.74975074775674,
      "grad_norm": 127.92801666259766,
      "learning_rate": 3.758464416791279e-07,
      "loss": 25.1251,
      "step": 8410
    },
    {
      "epoch": 66.82951146560319,
      "grad_norm": 459.4199523925781,
      "learning_rate": 3.755765095548574e-07,
      "loss": 25.6292,
      "step": 8420
    },
    {
      "epoch": 66.90927218344964,
      "grad_norm": 123.8630142211914,
      "learning_rate": 3.7530638148652e-07,
      "loss": 25.9062,
      "step": 8430
    },
    {
      "epoch": 66.98903290129611,
      "grad_norm": 66.76941680908203,
      "learning_rate": 3.7503605789561176e-07,
      "loss": 22.7821,
      "step": 8440
    },
    {
      "epoch": 67.06380857427718,
      "grad_norm": 114.29212188720703,
      "learning_rate": 3.7476553920393364e-07,
      "loss": 23.4223,
      "step": 8450
    },
    {
      "epoch": 67.14356929212363,
      "grad_norm": 115.19644927978516,
      "learning_rate": 3.7449482583359116e-07,
      "loss": 26.318,
      "step": 8460
    },
    {
      "epoch": 67.22333000997008,
      "grad_norm": 61.220420837402344,
      "learning_rate": 3.7422391820699367e-07,
      "loss": 23.5731,
      "step": 8470
    },
    {
      "epoch": 67.30309072781655,
      "grad_norm": 76.07804107666016,
      "learning_rate": 3.7395281674685334e-07,
      "loss": 24.1083,
      "step": 8480
    },
    {
      "epoch": 67.38285144566301,
      "grad_norm": 206.1667938232422,
      "learning_rate": 3.736815218761851e-07,
      "loss": 25.1364,
      "step": 8490
    },
    {
      "epoch": 67.46261216350948,
      "grad_norm": 109.1873550415039,
      "learning_rate": 3.734100340183055e-07,
      "loss": 24.0216,
      "step": 8500
    },
    {
      "epoch": 67.46261216350948,
      "eval_loss": 3.066004514694214,
      "eval_mae": 1.3180415630340576,
      "eval_mse": 3.066004514694214,
      "eval_r2": 0.07622784376144409,
      "eval_rmse": 1.7510010036245593,
      "eval_runtime": 9.0849,
      "eval_samples_per_second": 441.501,
      "eval_steps_per_second": 13.869,
      "step": 8500
    },
    {
      "epoch": 67.54237288135593,
      "grad_norm": 313.42364501953125,
      "learning_rate": 3.731383535968323e-07,
      "loss": 25.2754,
      "step": 8510
    },
    {
      "epoch": 67.6221335992024,
      "grad_norm": 327.2785339355469,
      "learning_rate": 3.728664810356836e-07,
      "loss": 23.376,
      "step": 8520
    },
    {
      "epoch": 67.70189431704885,
      "grad_norm": 128.79061889648438,
      "learning_rate": 3.7259441675907753e-07,
      "loss": 25.4697,
      "step": 8530
    },
    {
      "epoch": 67.78165503489531,
      "grad_norm": 306.5848693847656,
      "learning_rate": 3.7232216119153114e-07,
      "loss": 25.0118,
      "step": 8540
    },
    {
      "epoch": 67.86141575274178,
      "grad_norm": 230.59901428222656,
      "learning_rate": 3.720497147578601e-07,
      "loss": 24.424,
      "step": 8550
    },
    {
      "epoch": 67.94117647058823,
      "grad_norm": 385.51251220703125,
      "learning_rate": 3.717770778831777e-07,
      "loss": 26.901,
      "step": 8560
    },
    {
      "epoch": 68.0159521435693,
      "grad_norm": 107.0721206665039,
      "learning_rate": 3.715042509928948e-07,
      "loss": 25.78,
      "step": 8570
    },
    {
      "epoch": 68.09571286141575,
      "grad_norm": 263.0525817871094,
      "learning_rate": 3.712312345127183e-07,
      "loss": 23.3912,
      "step": 8580
    },
    {
      "epoch": 68.17547357926222,
      "grad_norm": 170.27700805664062,
      "learning_rate": 3.709580288686511e-07,
      "loss": 25.7638,
      "step": 8590
    },
    {
      "epoch": 68.25523429710867,
      "grad_norm": 190.2525634765625,
      "learning_rate": 3.7068463448699144e-07,
      "loss": 26.8153,
      "step": 8600
    },
    {
      "epoch": 68.25523429710867,
      "eval_loss": 3.0627195835113525,
      "eval_mae": 1.320117473602295,
      "eval_mse": 3.0627195835113525,
      "eval_r2": 0.07721757888793945,
      "eval_rmse": 1.7500627370215482,
      "eval_runtime": 9.084,
      "eval_samples_per_second": 441.543,
      "eval_steps_per_second": 13.87,
      "step": 8600
    },
    {
      "epoch": 68.33499501495514,
      "grad_norm": 91.61470031738281,
      "learning_rate": 3.704110517943319e-07,
      "loss": 22.9137,
      "step": 8610
    },
    {
      "epoch": 68.4147557328016,
      "grad_norm": 75.98213958740234,
      "learning_rate": 3.701372812175586e-07,
      "loss": 23.1728,
      "step": 8620
    },
    {
      "epoch": 68.49451645064805,
      "grad_norm": 267.9504089355469,
      "learning_rate": 3.698633231838514e-07,
      "loss": 23.4502,
      "step": 8630
    },
    {
      "epoch": 68.57427716849452,
      "grad_norm": 85.68649291992188,
      "learning_rate": 3.6958917812068224e-07,
      "loss": 26.7422,
      "step": 8640
    },
    {
      "epoch": 68.65403788634097,
      "grad_norm": 296.14923095703125,
      "learning_rate": 3.6931484645581524e-07,
      "loss": 23.2774,
      "step": 8650
    },
    {
      "epoch": 68.73379860418744,
      "grad_norm": 160.84161376953125,
      "learning_rate": 3.690403286173051e-07,
      "loss": 25.5654,
      "step": 8660
    },
    {
      "epoch": 68.8135593220339,
      "grad_norm": 243.06686401367188,
      "learning_rate": 3.6876562503349777e-07,
      "loss": 26.0291,
      "step": 8670
    },
    {
      "epoch": 68.89332003988036,
      "grad_norm": 92.43791198730469,
      "learning_rate": 3.684907361330284e-07,
      "loss": 25.0494,
      "step": 8680
    },
    {
      "epoch": 68.97308075772682,
      "grad_norm": 315.2274475097656,
      "learning_rate": 3.6821566234482164e-07,
      "loss": 25.2462,
      "step": 8690
    },
    {
      "epoch": 69.04785643070788,
      "grad_norm": 377.940673828125,
      "learning_rate": 3.679404040980906e-07,
      "loss": 26.6367,
      "step": 8700
    },
    {
      "epoch": 69.04785643070788,
      "eval_loss": 3.0620410442352295,
      "eval_mae": 1.3219993114471436,
      "eval_mse": 3.0620410442352295,
      "eval_r2": 0.07742202281951904,
      "eval_rmse": 1.7498688648682306,
      "eval_runtime": 9.0786,
      "eval_samples_per_second": 441.809,
      "eval_steps_per_second": 13.879,
      "step": 8700
    },
    {
      "epoch": 69.12761714855434,
      "grad_norm": 118.64423370361328,
      "learning_rate": 3.676649618223361e-07,
      "loss": 23.4573,
      "step": 8710
    },
    {
      "epoch": 69.2073778664008,
      "grad_norm": 484.994384765625,
      "learning_rate": 3.67389335947346e-07,
      "loss": 25.0094,
      "step": 8720
    },
    {
      "epoch": 69.28713858424726,
      "grad_norm": 293.2475891113281,
      "learning_rate": 3.671135269031951e-07,
      "loss": 26.0007,
      "step": 8730
    },
    {
      "epoch": 69.36689930209371,
      "grad_norm": 196.64553833007812,
      "learning_rate": 3.668375351202435e-07,
      "loss": 24.5714,
      "step": 8740
    },
    {
      "epoch": 69.44666001994018,
      "grad_norm": 341.39593505859375,
      "learning_rate": 3.665613610291367e-07,
      "loss": 24.8776,
      "step": 8750
    },
    {
      "epoch": 69.52642073778664,
      "grad_norm": 79.00114440917969,
      "learning_rate": 3.662850050608046e-07,
      "loss": 23.8761,
      "step": 8760
    },
    {
      "epoch": 69.6061814556331,
      "grad_norm": 102.5728759765625,
      "learning_rate": 3.660084676464611e-07,
      "loss": 24.649,
      "step": 8770
    },
    {
      "epoch": 69.68594217347956,
      "grad_norm": 297.4561462402344,
      "learning_rate": 3.657317492176027e-07,
      "loss": 26.3547,
      "step": 8780
    },
    {
      "epoch": 69.76570289132601,
      "grad_norm": 214.3977508544922,
      "learning_rate": 3.6545485020600884e-07,
      "loss": 25.4936,
      "step": 8790
    },
    {
      "epoch": 69.84546360917248,
      "grad_norm": 221.43826293945312,
      "learning_rate": 3.6517777104374063e-07,
      "loss": 26.0423,
      "step": 8800
    },
    {
      "epoch": 69.84546360917248,
      "eval_loss": 3.061256170272827,
      "eval_mae": 1.322737216949463,
      "eval_mse": 3.061255931854248,
      "eval_r2": 0.07765859365463257,
      "eval_rmse": 1.7496445158529341,
      "eval_runtime": 9.0972,
      "eval_samples_per_second": 440.906,
      "eval_steps_per_second": 13.85,
      "step": 8800
    },
    {
      "epoch": 69.92522432701894,
      "grad_norm": 119.96556091308594,
      "learning_rate": 3.6490051216314024e-07,
      "loss": 22.6416,
      "step": 8810
    },
    {
      "epoch": 70.0,
      "grad_norm": 139.15509033203125,
      "learning_rate": 3.6462307399683e-07,
      "loss": 23.221,
      "step": 8820
    },
    {
      "epoch": 70.07976071784645,
      "grad_norm": 473.70220947265625,
      "learning_rate": 3.6437322671565705e-07,
      "loss": 27.2323,
      "step": 8830
    },
    {
      "epoch": 70.15952143569292,
      "grad_norm": 51.05868911743164,
      "learning_rate": 3.6409544909937506e-07,
      "loss": 24.7301,
      "step": 8840
    },
    {
      "epoch": 70.23928215353938,
      "grad_norm": 139.17239379882812,
      "learning_rate": 3.638174934535683e-07,
      "loss": 22.8306,
      "step": 8850
    },
    {
      "epoch": 70.31904287138585,
      "grad_norm": 389.852783203125,
      "learning_rate": 3.635393602119468e-07,
      "loss": 23.017,
      "step": 8860
    },
    {
      "epoch": 70.3988035892323,
      "grad_norm": 489.2827453613281,
      "learning_rate": 3.6326104980849716e-07,
      "loss": 24.0965,
      "step": 8870
    },
    {
      "epoch": 70.47856430707877,
      "grad_norm": 314.7647705078125,
      "learning_rate": 3.6298256267748296e-07,
      "loss": 26.8876,
      "step": 8880
    },
    {
      "epoch": 70.55832502492522,
      "grad_norm": 408.3915710449219,
      "learning_rate": 3.627038992534431e-07,
      "loss": 25.819,
      "step": 8890
    },
    {
      "epoch": 70.63808574277168,
      "grad_norm": 126.84314727783203,
      "learning_rate": 3.6242505997119174e-07,
      "loss": 26.666,
      "step": 8900
    },
    {
      "epoch": 70.63808574277168,
      "eval_loss": 3.0710322856903076,
      "eval_mae": 1.3310539722442627,
      "eval_mse": 3.0710322856903076,
      "eval_r2": 0.07471299171447754,
      "eval_rmse": 1.752436100315874,
      "eval_runtime": 9.1199,
      "eval_samples_per_second": 439.808,
      "eval_steps_per_second": 13.816,
      "step": 8900
    },
    {
      "epoch": 70.71784646061815,
      "grad_norm": 358.9391174316406,
      "learning_rate": 3.621460452658176e-07,
      "loss": 25.3601,
      "step": 8910
    },
    {
      "epoch": 70.7976071784646,
      "grad_norm": 293.7834167480469,
      "learning_rate": 3.6186685557268286e-07,
      "loss": 26.088,
      "step": 8920
    },
    {
      "epoch": 70.87736789631107,
      "grad_norm": 43.07698440551758,
      "learning_rate": 3.615874913274228e-07,
      "loss": 24.3503,
      "step": 8930
    },
    {
      "epoch": 70.95712861415753,
      "grad_norm": 155.59963989257812,
      "learning_rate": 3.613079529659452e-07,
      "loss": 22.833,
      "step": 8940
    },
    {
      "epoch": 71.03190428713859,
      "grad_norm": 97.04302978515625,
      "learning_rate": 3.6102824092442933e-07,
      "loss": 22.8683,
      "step": 8950
    },
    {
      "epoch": 71.11166500498504,
      "grad_norm": 115.6922836303711,
      "learning_rate": 3.6074835563932576e-07,
      "loss": 25.648,
      "step": 8960
    },
    {
      "epoch": 71.19142572283151,
      "grad_norm": 118.38312530517578,
      "learning_rate": 3.60468297547355e-07,
      "loss": 25.867,
      "step": 8970
    },
    {
      "epoch": 71.27118644067797,
      "grad_norm": 113.87028503417969,
      "learning_rate": 3.601880670855075e-07,
      "loss": 24.057,
      "step": 8980
    },
    {
      "epoch": 71.35094715852442,
      "grad_norm": 132.84255981445312,
      "learning_rate": 3.599076646910424e-07,
      "loss": 22.2716,
      "step": 8990
    },
    {
      "epoch": 71.43070787637089,
      "grad_norm": 122.4183120727539,
      "learning_rate": 3.5962709080148743e-07,
      "loss": 25.7429,
      "step": 9000
    },
    {
      "epoch": 71.43070787637089,
      "eval_loss": 3.058762550354004,
      "eval_mae": 1.3227847814559937,
      "eval_mse": 3.058762550354004,
      "eval_r2": 0.07840979099273682,
      "eval_rmse": 1.748931831248435,
      "eval_runtime": 9.0988,
      "eval_samples_per_second": 440.826,
      "eval_steps_per_second": 13.848,
      "step": 9000
    },
    {
      "epoch": 71.51046859421734,
      "grad_norm": 62.26708221435547,
      "learning_rate": 3.5934634585463774e-07,
      "loss": 22.6924,
      "step": 9010
    },
    {
      "epoch": 71.59022931206381,
      "grad_norm": 422.3989562988281,
      "learning_rate": 3.590654302885555e-07,
      "loss": 26.4849,
      "step": 9020
    },
    {
      "epoch": 71.66999002991027,
      "grad_norm": 486.0423278808594,
      "learning_rate": 3.587843445415687e-07,
      "loss": 25.1107,
      "step": 9030
    },
    {
      "epoch": 71.74975074775674,
      "grad_norm": 322.81256103515625,
      "learning_rate": 3.5850308905227143e-07,
      "loss": 24.4619,
      "step": 9040
    },
    {
      "epoch": 71.82951146560319,
      "grad_norm": 120.65798950195312,
      "learning_rate": 3.5822166425952236e-07,
      "loss": 25.5674,
      "step": 9050
    },
    {
      "epoch": 71.90927218344964,
      "grad_norm": 301.5791015625,
      "learning_rate": 3.5794007060244443e-07,
      "loss": 26.1748,
      "step": 9060
    },
    {
      "epoch": 71.98903290129611,
      "grad_norm": 84.74629211425781,
      "learning_rate": 3.576583085204238e-07,
      "loss": 24.3629,
      "step": 9070
    },
    {
      "epoch": 72.06380857427718,
      "grad_norm": 330.45648193359375,
      "learning_rate": 3.5737637845310993e-07,
      "loss": 24.3095,
      "step": 9080
    },
    {
      "epoch": 72.14356929212363,
      "grad_norm": 46.20743942260742,
      "learning_rate": 3.570942808404138e-07,
      "loss": 23.7282,
      "step": 9090
    },
    {
      "epoch": 72.22333000997008,
      "grad_norm": 608.0449829101562,
      "learning_rate": 3.5681201612250837e-07,
      "loss": 23.4555,
      "step": 9100
    },
    {
      "epoch": 72.22333000997008,
      "eval_loss": 3.056166410446167,
      "eval_mae": 1.316311001777649,
      "eval_mse": 3.056166410446167,
      "eval_r2": 0.07919204235076904,
      "eval_rmse": 1.7481894664040758,
      "eval_runtime": 9.0777,
      "eval_samples_per_second": 441.853,
      "eval_steps_per_second": 13.88,
      "step": 9100
    },
    {
      "epoch": 72.30309072781655,
      "grad_norm": 306.81744384765625,
      "learning_rate": 3.5652958473982696e-07,
      "loss": 25.5228,
      "step": 9110
    },
    {
      "epoch": 72.38285144566301,
      "grad_norm": 222.90760803222656,
      "learning_rate": 3.562469871330632e-07,
      "loss": 25.3907,
      "step": 9120
    },
    {
      "epoch": 72.46261216350948,
      "grad_norm": 258.46002197265625,
      "learning_rate": 3.559642237431698e-07,
      "loss": 25.5679,
      "step": 9130
    },
    {
      "epoch": 72.54237288135593,
      "grad_norm": 330.5743103027344,
      "learning_rate": 3.5568129501135856e-07,
      "loss": 23.9446,
      "step": 9140
    },
    {
      "epoch": 72.6221335992024,
      "grad_norm": 368.6435546875,
      "learning_rate": 3.553982013790989e-07,
      "loss": 24.0174,
      "step": 9150
    },
    {
      "epoch": 72.70189431704885,
      "grad_norm": 69.02025604248047,
      "learning_rate": 3.5511494328811767e-07,
      "loss": 24.9693,
      "step": 9160
    },
    {
      "epoch": 72.78165503489531,
      "grad_norm": 279.7677001953125,
      "learning_rate": 3.548315211803984e-07,
      "loss": 27.9484,
      "step": 9170
    },
    {
      "epoch": 72.86141575274178,
      "grad_norm": 102.18354797363281,
      "learning_rate": 3.545479354981804e-07,
      "loss": 24.234,
      "step": 9180
    },
    {
      "epoch": 72.94117647058823,
      "grad_norm": 181.75677490234375,
      "learning_rate": 3.542641866839585e-07,
      "loss": 24.5007,
      "step": 9190
    },
    {
      "epoch": 73.0159521435693,
      "grad_norm": 593.4639282226562,
      "learning_rate": 3.5398027518048164e-07,
      "loss": 24.3578,
      "step": 9200
    },
    {
      "epoch": 73.0159521435693,
      "eval_loss": 3.0623042583465576,
      "eval_mae": 1.310591220855713,
      "eval_mse": 3.0623042583465576,
      "eval_r2": 0.077342689037323,
      "eval_rmse": 1.749944072919634,
      "eval_runtime": 9.0868,
      "eval_samples_per_second": 441.41,
      "eval_steps_per_second": 13.866,
      "step": 9200
    },
    {
      "epoch": 73.09571286141575,
      "grad_norm": 263.2080993652344,
      "learning_rate": 3.536962014307531e-07,
      "loss": 26.6369,
      "step": 9210
    },
    {
      "epoch": 73.17547357926222,
      "grad_norm": 169.20558166503906,
      "learning_rate": 3.534119658780288e-07,
      "loss": 26.397,
      "step": 9220
    },
    {
      "epoch": 73.25523429710867,
      "grad_norm": 147.1641082763672,
      "learning_rate": 3.531275689658176e-07,
      "loss": 24.3629,
      "step": 9230
    },
    {
      "epoch": 73.33499501495514,
      "grad_norm": 129.83119201660156,
      "learning_rate": 3.5284301113787996e-07,
      "loss": 25.1567,
      "step": 9240
    },
    {
      "epoch": 73.4147557328016,
      "grad_norm": 364.77545166015625,
      "learning_rate": 3.525582928382273e-07,
      "loss": 27.498,
      "step": 9250
    },
    {
      "epoch": 73.49451645064805,
      "grad_norm": 280.6644592285156,
      "learning_rate": 3.5227341451112156e-07,
      "loss": 24.1225,
      "step": 9260
    },
    {
      "epoch": 73.57427716849452,
      "grad_norm": 337.0046691894531,
      "learning_rate": 3.519883766010744e-07,
      "loss": 22.7654,
      "step": 9270
    },
    {
      "epoch": 73.65403788634097,
      "grad_norm": 85.38751220703125,
      "learning_rate": 3.5170317955284644e-07,
      "loss": 25.1829,
      "step": 9280
    },
    {
      "epoch": 73.73379860418744,
      "grad_norm": 293.2549133300781,
      "learning_rate": 3.514178238114466e-07,
      "loss": 25.0107,
      "step": 9290
    },
    {
      "epoch": 73.8135593220339,
      "grad_norm": 478.9608459472656,
      "learning_rate": 3.511323098221315e-07,
      "loss": 23.8991,
      "step": 9300
    },
    {
      "epoch": 73.8135593220339,
      "eval_loss": 3.056279182434082,
      "eval_mae": 1.3124663829803467,
      "eval_mse": 3.056279420852661,
      "eval_r2": 0.0791580080986023,
      "eval_rmse": 1.7482217882330209,
      "eval_runtime": 9.0865,
      "eval_samples_per_second": 441.424,
      "eval_steps_per_second": 13.867,
      "step": 9300
    },
    {
      "epoch": 73.89332003988036,
      "grad_norm": 317.2913513183594,
      "learning_rate": 3.508466380304046e-07,
      "loss": 23.5169,
      "step": 9310
    },
    {
      "epoch": 73.97308075772682,
      "grad_norm": 217.1324005126953,
      "learning_rate": 3.505608088820155e-07,
      "loss": 23.1523,
      "step": 9320
    },
    {
      "epoch": 74.04785643070788,
      "grad_norm": 103.88399505615234,
      "learning_rate": 3.502748228229596e-07,
      "loss": 21.5205,
      "step": 9330
    },
    {
      "epoch": 74.12761714855434,
      "grad_norm": 106.8578109741211,
      "learning_rate": 3.499886802994768e-07,
      "loss": 23.9702,
      "step": 9340
    },
    {
      "epoch": 74.2073778664008,
      "grad_norm": 567.977294921875,
      "learning_rate": 3.497023817580516e-07,
      "loss": 27.3229,
      "step": 9350
    },
    {
      "epoch": 74.28713858424726,
      "grad_norm": 101.11739349365234,
      "learning_rate": 3.4941592764541127e-07,
      "loss": 25.0238,
      "step": 9360
    },
    {
      "epoch": 74.36689930209371,
      "grad_norm": 84.21052551269531,
      "learning_rate": 3.491293184085266e-07,
      "loss": 23.8544,
      "step": 9370
    },
    {
      "epoch": 74.44666001994018,
      "grad_norm": 391.4809265136719,
      "learning_rate": 3.488425544946098e-07,
      "loss": 24.1024,
      "step": 9380
    },
    {
      "epoch": 74.52642073778664,
      "grad_norm": 82.73574829101562,
      "learning_rate": 3.485556363511148e-07,
      "loss": 27.272,
      "step": 9390
    },
    {
      "epoch": 74.6061814556331,
      "grad_norm": 104.74907684326172,
      "learning_rate": 3.4826856442573603e-07,
      "loss": 23.9074,
      "step": 9400
    },
    {
      "epoch": 74.6061814556331,
      "eval_loss": 3.0523128509521484,
      "eval_mae": 1.320926308631897,
      "eval_mse": 3.0523128509521484,
      "eval_r2": 0.0803530216217041,
      "eval_rmse": 1.7470869614739126,
      "eval_runtime": 9.097,
      "eval_samples_per_second": 440.914,
      "eval_steps_per_second": 13.851,
      "step": 9400
    },
    {
      "epoch": 74.68594217347956,
      "grad_norm": 50.035621643066406,
      "learning_rate": 3.4798133916640794e-07,
      "loss": 24.9706,
      "step": 9410
    },
    {
      "epoch": 74.76570289132601,
      "grad_norm": 165.34947204589844,
      "learning_rate": 3.4769396102130413e-07,
      "loss": 25.6866,
      "step": 9420
    },
    {
      "epoch": 74.84546360917248,
      "grad_norm": 184.37969970703125,
      "learning_rate": 3.474064304388368e-07,
      "loss": 24.5739,
      "step": 9430
    },
    {
      "epoch": 74.92522432701894,
      "grad_norm": 276.6813659667969,
      "learning_rate": 3.4711874786765614e-07,
      "loss": 24.6645,
      "step": 9440
    },
    {
      "epoch": 75.0,
      "grad_norm": 88.17737579345703,
      "learning_rate": 3.468309137566494e-07,
      "loss": 22.8896,
      "step": 9450
    },
    {
      "epoch": 75.07976071784645,
      "grad_norm": 123.72151184082031,
      "learning_rate": 3.465429285549401e-07,
      "loss": 25.006,
      "step": 9460
    },
    {
      "epoch": 75.15952143569292,
      "grad_norm": 141.5162811279297,
      "learning_rate": 3.46254792711888e-07,
      "loss": 24.1427,
      "step": 9470
    },
    {
      "epoch": 75.23928215353938,
      "grad_norm": 191.08383178710938,
      "learning_rate": 3.4596650667708727e-07,
      "loss": 25.8205,
      "step": 9480
    },
    {
      "epoch": 75.31904287138585,
      "grad_norm": 177.44964599609375,
      "learning_rate": 3.4567807090036696e-07,
      "loss": 25.1382,
      "step": 9490
    },
    {
      "epoch": 75.3988035892323,
      "grad_norm": 38.57638931274414,
      "learning_rate": 3.453894858317895e-07,
      "loss": 26.9749,
      "step": 9500
    },
    {
      "epoch": 75.3988035892323,
      "eval_loss": 3.0497334003448486,
      "eval_mae": 1.314713478088379,
      "eval_mse": 3.0497334003448486,
      "eval_r2": 0.0811302661895752,
      "eval_rmse": 1.7463485907300549,
      "eval_runtime": 9.1004,
      "eval_samples_per_second": 440.749,
      "eval_steps_per_second": 13.846,
      "step": 9500
    },
    {
      "epoch": 75.47856430707877,
      "grad_norm": 314.9055480957031,
      "learning_rate": 3.451007519216506e-07,
      "loss": 24.3219,
      "step": 9510
    },
    {
      "epoch": 75.55832502492522,
      "grad_norm": 411.2994384765625,
      "learning_rate": 3.448118696204775e-07,
      "loss": 24.8036,
      "step": 9520
    },
    {
      "epoch": 75.63808574277168,
      "grad_norm": 74.12139129638672,
      "learning_rate": 3.445228393790298e-07,
      "loss": 24.0938,
      "step": 9530
    },
    {
      "epoch": 75.71784646061815,
      "grad_norm": 609.5748901367188,
      "learning_rate": 3.442336616482975e-07,
      "loss": 25.024,
      "step": 9540
    },
    {
      "epoch": 75.7976071784646,
      "grad_norm": 200.36746215820312,
      "learning_rate": 3.439443368795009e-07,
      "loss": 24.2689,
      "step": 9550
    },
    {
      "epoch": 75.87736789631107,
      "grad_norm": 208.6078338623047,
      "learning_rate": 3.436548655240894e-07,
      "loss": 24.5859,
      "step": 9560
    },
    {
      "epoch": 75.95712861415753,
      "grad_norm": 691.5971069335938,
      "learning_rate": 3.433652480337417e-07,
      "loss": 23.3654,
      "step": 9570
    },
    {
      "epoch": 76.03190428713859,
      "grad_norm": 252.24432373046875,
      "learning_rate": 3.43075484860364e-07,
      "loss": 24.1112,
      "step": 9580
    },
    {
      "epoch": 76.11166500498504,
      "grad_norm": 83.96273040771484,
      "learning_rate": 3.427855764560901e-07,
      "loss": 24.871,
      "step": 9590
    },
    {
      "epoch": 76.19142572283151,
      "grad_norm": 276.2619323730469,
      "learning_rate": 3.4249552327328027e-07,
      "loss": 25.5983,
      "step": 9600
    },
    {
      "epoch": 76.19142572283151,
      "eval_loss": 3.0511128902435303,
      "eval_mae": 1.3213589191436768,
      "eval_mse": 3.0511128902435303,
      "eval_r2": 0.08071458339691162,
      "eval_rmse": 1.7467435101478208,
      "eval_runtime": 9.1002,
      "eval_samples_per_second": 440.759,
      "eval_steps_per_second": 13.846,
      "step": 9600
    },
    {
      "epoch": 76.27118644067797,
      "grad_norm": 399.31427001953125,
      "learning_rate": 3.4220532576452085e-07,
      "loss": 25.2765,
      "step": 9610
    },
    {
      "epoch": 76.35094715852442,
      "grad_norm": 464.6283264160156,
      "learning_rate": 3.419149843826231e-07,
      "loss": 25.7047,
      "step": 9620
    },
    {
      "epoch": 76.43070787637089,
      "grad_norm": 209.1724090576172,
      "learning_rate": 3.4162449958062316e-07,
      "loss": 22.4275,
      "step": 9630
    },
    {
      "epoch": 76.51046859421734,
      "grad_norm": 285.52593994140625,
      "learning_rate": 3.413338718117806e-07,
      "loss": 25.4166,
      "step": 9640
    },
    {
      "epoch": 76.59022931206381,
      "grad_norm": 255.88510131835938,
      "learning_rate": 3.410431015295784e-07,
      "loss": 24.787,
      "step": 9650
    },
    {
      "epoch": 76.66999002991027,
      "grad_norm": 137.2398223876953,
      "learning_rate": 3.4075218918772154e-07,
      "loss": 24.8588,
      "step": 9660
    },
    {
      "epoch": 76.74975074775674,
      "grad_norm": 371.6817626953125,
      "learning_rate": 3.404611352401369e-07,
      "loss": 22.7635,
      "step": 9670
    },
    {
      "epoch": 76.82951146560319,
      "grad_norm": 58.25032424926758,
      "learning_rate": 3.4016994014097233e-07,
      "loss": 24.8311,
      "step": 9680
    },
    {
      "epoch": 76.90927218344964,
      "grad_norm": 227.07810974121094,
      "learning_rate": 3.3987860434459583e-07,
      "loss": 24.4209,
      "step": 9690
    },
    {
      "epoch": 76.98903290129611,
      "grad_norm": 142.50608825683594,
      "learning_rate": 3.3958712830559487e-07,
      "loss": 26.006,
      "step": 9700
    },
    {
      "epoch": 76.98903290129611,
      "eval_loss": 3.0453178882598877,
      "eval_mae": 1.3163361549377441,
      "eval_mse": 3.0453178882598877,
      "eval_r2": 0.08246058225631714,
      "eval_rmse": 1.7450839201195705,
      "eval_runtime": 9.0755,
      "eval_samples_per_second": 441.961,
      "eval_steps_per_second": 13.884,
      "step": 9700
    },
    {
      "epoch": 77.06380857427718,
      "grad_norm": 285.72149658203125,
      "learning_rate": 3.3929551247877606e-07,
      "loss": 25.6414,
      "step": 9710
    },
    {
      "epoch": 77.14356929212363,
      "grad_norm": 255.40960693359375,
      "learning_rate": 3.3900375731916374e-07,
      "loss": 22.5476,
      "step": 9720
    },
    {
      "epoch": 77.22333000997008,
      "grad_norm": 208.32432556152344,
      "learning_rate": 3.3871186328199995e-07,
      "loss": 25.8574,
      "step": 9730
    },
    {
      "epoch": 77.30309072781655,
      "grad_norm": 108.06330871582031,
      "learning_rate": 3.384198308227434e-07,
      "loss": 24.569,
      "step": 9740
    },
    {
      "epoch": 77.38285144566301,
      "grad_norm": 337.4876403808594,
      "learning_rate": 3.3812766039706846e-07,
      "loss": 25.0633,
      "step": 9750
    },
    {
      "epoch": 77.46261216350948,
      "grad_norm": 154.1691436767578,
      "learning_rate": 3.3783535246086516e-07,
      "loss": 24.416,
      "step": 9760
    },
    {
      "epoch": 77.54237288135593,
      "grad_norm": 136.4617156982422,
      "learning_rate": 3.3754290747023796e-07,
      "loss": 22.9575,
      "step": 9770
    },
    {
      "epoch": 77.6221335992024,
      "grad_norm": 164.31016540527344,
      "learning_rate": 3.3725032588150517e-07,
      "loss": 27.5521,
      "step": 9780
    },
    {
      "epoch": 77.70189431704885,
      "grad_norm": 422.6942138671875,
      "learning_rate": 3.3695760815119816e-07,
      "loss": 26.1196,
      "step": 9790
    },
    {
      "epoch": 77.78165503489531,
      "grad_norm": 227.1068878173828,
      "learning_rate": 3.36664754736061e-07,
      "loss": 24.7913,
      "step": 9800
    },
    {
      "epoch": 77.78165503489531,
      "eval_loss": 3.0548555850982666,
      "eval_mae": 1.3078426122665405,
      "eval_mse": 3.0548558235168457,
      "eval_r2": 0.079586923122406,
      "eval_rmse": 1.7478145849937416,
      "eval_runtime": 9.0735,
      "eval_samples_per_second": 442.054,
      "eval_steps_per_second": 13.887,
      "step": 9800
    },
    {
      "epoch": 77.86141575274178,
      "grad_norm": 448.2280578613281,
      "learning_rate": 3.3637176609304905e-07,
      "loss": 23.4003,
      "step": 9810
    },
    {
      "epoch": 77.94117647058823,
      "grad_norm": 374.3587951660156,
      "learning_rate": 3.3607864267932913e-07,
      "loss": 23.7702,
      "step": 9820
    },
    {
      "epoch": 78.0159521435693,
      "grad_norm": 439.8951721191406,
      "learning_rate": 3.3578538495227796e-07,
      "loss": 22.7544,
      "step": 9830
    },
    {
      "epoch": 78.09571286141575,
      "grad_norm": 292.9596862792969,
      "learning_rate": 3.354919933694822e-07,
      "loss": 26.3637,
      "step": 9840
    },
    {
      "epoch": 78.17547357926222,
      "grad_norm": 216.87107849121094,
      "learning_rate": 3.35198468388737e-07,
      "loss": 21.3863,
      "step": 9850
    },
    {
      "epoch": 78.25523429710867,
      "grad_norm": 498.3560791015625,
      "learning_rate": 3.3490481046804594e-07,
      "loss": 24.6341,
      "step": 9860
    },
    {
      "epoch": 78.33499501495514,
      "grad_norm": 280.01220703125,
      "learning_rate": 3.3461102006561985e-07,
      "loss": 25.052,
      "step": 9870
    },
    {
      "epoch": 78.4147557328016,
      "grad_norm": 182.12396240234375,
      "learning_rate": 3.343170976398765e-07,
      "loss": 22.8939,
      "step": 9880
    },
    {
      "epoch": 78.49451645064805,
      "grad_norm": 413.0950622558594,
      "learning_rate": 3.340230436494393e-07,
      "loss": 24.4764,
      "step": 9890
    },
    {
      "epoch": 78.57427716849452,
      "grad_norm": 204.4065704345703,
      "learning_rate": 3.337288585531375e-07,
      "loss": 26.2159,
      "step": 9900
    },
    {
      "epoch": 78.57427716849452,
      "eval_loss": 3.047255754470825,
      "eval_mae": 1.3226542472839355,
      "eval_mse": 3.047255754470825,
      "eval_r2": 0.08187675476074219,
      "eval_rmse": 1.745639067639936,
      "eval_runtime": 9.09,
      "eval_samples_per_second": 441.252,
      "eval_steps_per_second": 13.861,
      "step": 9900
    },
    {
      "epoch": 78.65403788634097,
      "grad_norm": 152.0811309814453,
      "learning_rate": 3.3343454281000427e-07,
      "loss": 25.7904,
      "step": 9910
    },
    {
      "epoch": 78.73379860418744,
      "grad_norm": 128.9251708984375,
      "learning_rate": 3.3314009687927717e-07,
      "loss": 26.1647,
      "step": 9920
    },
    {
      "epoch": 78.8135593220339,
      "grad_norm": 136.84832763671875,
      "learning_rate": 3.328455212203966e-07,
      "loss": 25.1437,
      "step": 9930
    },
    {
      "epoch": 78.89332003988036,
      "grad_norm": 158.07907104492188,
      "learning_rate": 3.325508162930056e-07,
      "loss": 23.8966,
      "step": 9940
    },
    {
      "epoch": 78.97308075772682,
      "grad_norm": 363.46539306640625,
      "learning_rate": 3.3225598255694854e-07,
      "loss": 25.4471,
      "step": 9950
    },
    {
      "epoch": 79.04785643070788,
      "grad_norm": 80.3748779296875,
      "learning_rate": 3.319610204722713e-07,
      "loss": 23.815,
      "step": 9960
    },
    {
      "epoch": 79.12761714855434,
      "grad_norm": 132.05052185058594,
      "learning_rate": 3.3166593049921947e-07,
      "loss": 23.0523,
      "step": 9970
    },
    {
      "epoch": 79.2073778664008,
      "grad_norm": 59.55046081542969,
      "learning_rate": 3.313707130982387e-07,
      "loss": 24.2984,
      "step": 9980
    },
    {
      "epoch": 79.28713858424726,
      "grad_norm": 404.2379150390625,
      "learning_rate": 3.310753687299732e-07,
      "loss": 22.0736,
      "step": 9990
    },
    {
      "epoch": 79.36689930209371,
      "grad_norm": 548.7941284179688,
      "learning_rate": 3.3077989785526536e-07,
      "loss": 24.7664,
      "step": 10000
    },
    {
      "epoch": 79.36689930209371,
      "eval_loss": 3.0421335697174072,
      "eval_mae": 1.3189619779586792,
      "eval_mse": 3.0421340465545654,
      "eval_r2": 0.08341985940933228,
      "eval_rmse": 1.7441714498737118,
      "eval_runtime": 9.0942,
      "eval_samples_per_second": 441.052,
      "eval_steps_per_second": 13.855,
      "step": 10000
    },
    {
      "epoch": 79.44666001994018,
      "grad_norm": 132.6288299560547,
      "learning_rate": 3.3048430093515483e-07,
      "loss": 24.042,
      "step": 10010
    },
    {
      "epoch": 79.52642073778664,
      "grad_norm": 94.08177185058594,
      "learning_rate": 3.301885784308782e-07,
      "loss": 22.9365,
      "step": 10020
    },
    {
      "epoch": 79.6061814556331,
      "grad_norm": 486.14080810546875,
      "learning_rate": 3.298927308038678e-07,
      "loss": 24.6133,
      "step": 10030
    },
    {
      "epoch": 79.68594217347956,
      "grad_norm": 149.77662658691406,
      "learning_rate": 3.295967585157512e-07,
      "loss": 25.6675,
      "step": 10040
    },
    {
      "epoch": 79.76570289132601,
      "grad_norm": 269.9638671875,
      "learning_rate": 3.2930066202835065e-07,
      "loss": 27.9391,
      "step": 10050
    },
    {
      "epoch": 79.84546360917248,
      "grad_norm": 112.71063995361328,
      "learning_rate": 3.2900444180368213e-07,
      "loss": 27.5189,
      "step": 10060
    },
    {
      "epoch": 79.92522432701894,
      "grad_norm": 312.2144775390625,
      "learning_rate": 3.287080983039545e-07,
      "loss": 24.9859,
      "step": 10070
    },
    {
      "epoch": 80.0,
      "grad_norm": 385.76629638671875,
      "learning_rate": 3.284116319915693e-07,
      "loss": 22.5648,
      "step": 10080
    },
    {
      "epoch": 80.07976071784645,
      "grad_norm": 402.8841857910156,
      "learning_rate": 3.2811504332911944e-07,
      "loss": 23.9249,
      "step": 10090
    },
    {
      "epoch": 80.15952143569292,
      "grad_norm": 354.6750183105469,
      "learning_rate": 3.278183327793889e-07,
      "loss": 24.9814,
      "step": 10100
    },
    {
      "epoch": 80.15952143569292,
      "eval_loss": 3.04233717918396,
      "eval_mae": 1.3198978900909424,
      "eval_mse": 3.04233717918396,
      "eval_r2": 0.08335870504379272,
      "eval_rmse": 1.7442296807427513,
      "eval_runtime": 9.0921,
      "eval_samples_per_second": 441.155,
      "eval_steps_per_second": 13.858,
      "step": 10100
    },
    {
      "epoch": 80.23928215353938,
      "grad_norm": 125.52257537841797,
      "learning_rate": 3.2752150080535167e-07,
      "loss": 24.5656,
      "step": 10110
    },
    {
      "epoch": 80.31904287138585,
      "grad_norm": 440.5180969238281,
      "learning_rate": 3.2722454787017147e-07,
      "loss": 24.8485,
      "step": 10120
    },
    {
      "epoch": 80.3988035892323,
      "grad_norm": 243.70953369140625,
      "learning_rate": 3.269274744372005e-07,
      "loss": 24.074,
      "step": 10130
    },
    {
      "epoch": 80.47856430707877,
      "grad_norm": 169.30943298339844,
      "learning_rate": 3.2663028096997906e-07,
      "loss": 24.8247,
      "step": 10140
    },
    {
      "epoch": 80.55832502492522,
      "grad_norm": 444.5276184082031,
      "learning_rate": 3.2633296793223493e-07,
      "loss": 25.5621,
      "step": 10150
    },
    {
      "epoch": 80.63808574277168,
      "grad_norm": 171.6515655517578,
      "learning_rate": 3.2603553578788223e-07,
      "loss": 23.3385,
      "step": 10160
    },
    {
      "epoch": 80.71784646061815,
      "grad_norm": 193.34361267089844,
      "learning_rate": 3.257379850010209e-07,
      "loss": 27.2977,
      "step": 10170
    },
    {
      "epoch": 80.7976071784646,
      "grad_norm": 62.21059036254883,
      "learning_rate": 3.2544031603593624e-07,
      "loss": 23.6639,
      "step": 10180
    },
    {
      "epoch": 80.87736789631107,
      "grad_norm": 452.9521179199219,
      "learning_rate": 3.2514252935709774e-07,
      "loss": 24.8742,
      "step": 10190
    },
    {
      "epoch": 80.95712861415753,
      "grad_norm": 547.2555541992188,
      "learning_rate": 3.248446254291588e-07,
      "loss": 24.5573,
      "step": 10200
    },
    {
      "epoch": 80.95712861415753,
      "eval_loss": 3.0389792919158936,
      "eval_mae": 1.3104017972946167,
      "eval_mse": 3.0389792919158936,
      "eval_r2": 0.0843704342842102,
      "eval_rmse": 1.7432668447245514,
      "eval_runtime": 9.0821,
      "eval_samples_per_second": 441.64,
      "eval_steps_per_second": 13.873,
      "step": 10200
    },
    {
      "epoch": 81.03190428713859,
      "grad_norm": 432.891845703125,
      "learning_rate": 3.2454660471695536e-07,
      "loss": 23.64,
      "step": 10210
    },
    {
      "epoch": 81.11166500498504,
      "grad_norm": 121.47529602050781,
      "learning_rate": 3.242484676855061e-07,
      "loss": 26.8345,
      "step": 10220
    },
    {
      "epoch": 81.19142572283151,
      "grad_norm": 951.72412109375,
      "learning_rate": 3.2395021480001083e-07,
      "loss": 25.3862,
      "step": 10230
    },
    {
      "epoch": 81.27118644067797,
      "grad_norm": 143.4444122314453,
      "learning_rate": 3.236518465258502e-07,
      "loss": 23.5563,
      "step": 10240
    },
    {
      "epoch": 81.35094715852442,
      "grad_norm": 130.21266174316406,
      "learning_rate": 3.2335336332858506e-07,
      "loss": 22.0906,
      "step": 10250
    },
    {
      "epoch": 81.43070787637089,
      "grad_norm": 232.81362915039062,
      "learning_rate": 3.230547656739554e-07,
      "loss": 25.2647,
      "step": 10260
    },
    {
      "epoch": 81.51046859421734,
      "grad_norm": 588.8156127929688,
      "learning_rate": 3.227560540278798e-07,
      "loss": 25.4037,
      "step": 10270
    },
    {
      "epoch": 81.59022931206381,
      "grad_norm": 302.1634216308594,
      "learning_rate": 3.2245722885645496e-07,
      "loss": 24.214,
      "step": 10280
    },
    {
      "epoch": 81.66999002991027,
      "grad_norm": 87.46845245361328,
      "learning_rate": 3.2215829062595445e-07,
      "loss": 24.0789,
      "step": 10290
    },
    {
      "epoch": 81.74975074775674,
      "grad_norm": 280.2633361816406,
      "learning_rate": 3.2185923980282835e-07,
      "loss": 23.065,
      "step": 10300
    },
    {
      "epoch": 81.74975074775674,
      "eval_loss": 3.0380895137786865,
      "eval_mae": 1.309239149093628,
      "eval_mse": 3.0380895137786865,
      "eval_r2": 0.08463847637176514,
      "eval_rmse": 1.7430116218140046,
      "eval_runtime": 9.0729,
      "eval_samples_per_second": 442.087,
      "eval_steps_per_second": 13.888,
      "step": 10300
    },
    {
      "epoch": 81.82951146560319,
      "grad_norm": 113.61293029785156,
      "learning_rate": 3.2156007685370247e-07,
      "loss": 24.1365,
      "step": 10310
    },
    {
      "epoch": 81.90927218344964,
      "grad_norm": 299.0325927734375,
      "learning_rate": 3.2126080224537745e-07,
      "loss": 26.3868,
      "step": 10320
    },
    {
      "epoch": 81.98903290129611,
      "grad_norm": 180.27159118652344,
      "learning_rate": 3.2096141644482834e-07,
      "loss": 25.2489,
      "step": 10330
    },
    {
      "epoch": 82.06380857427718,
      "grad_norm": 373.5345458984375,
      "learning_rate": 3.206619199192034e-07,
      "loss": 24.5,
      "step": 10340
    },
    {
      "epoch": 82.14356929212363,
      "grad_norm": 415.2379455566406,
      "learning_rate": 3.203623131358241e-07,
      "loss": 23.3937,
      "step": 10350
    },
    {
      "epoch": 82.22333000997008,
      "grad_norm": 296.4977722167969,
      "learning_rate": 3.2006259656218346e-07,
      "loss": 25.374,
      "step": 10360
    },
    {
      "epoch": 82.30309072781655,
      "grad_norm": 94.92182159423828,
      "learning_rate": 3.1976277066594624e-07,
      "loss": 25.1925,
      "step": 10370
    },
    {
      "epoch": 82.38285144566301,
      "grad_norm": 91.66964721679688,
      "learning_rate": 3.194628359149475e-07,
      "loss": 24.9719,
      "step": 10380
    },
    {
      "epoch": 82.46261216350948,
      "grad_norm": 351.6086120605469,
      "learning_rate": 3.1916279277719225e-07,
      "loss": 23.6032,
      "step": 10390
    },
    {
      "epoch": 82.54237288135593,
      "grad_norm": 298.90740966796875,
      "learning_rate": 3.1886264172085467e-07,
      "loss": 25.0727,
      "step": 10400
    },
    {
      "epoch": 82.54237288135593,
      "eval_loss": 3.036163806915283,
      "eval_mae": 1.3075165748596191,
      "eval_mse": 3.036163806915283,
      "eval_r2": 0.08521872758865356,
      "eval_rmse": 1.7424591263255742,
      "eval_runtime": 9.072,
      "eval_samples_per_second": 442.131,
      "eval_steps_per_second": 13.889,
      "step": 10400
    },
    {
      "epoch": 82.6221335992024,
      "grad_norm": 515.6729125976562,
      "learning_rate": 3.185623832142771e-07,
      "loss": 25.3068,
      "step": 10410
    },
    {
      "epoch": 82.70189431704885,
      "grad_norm": 76.53244018554688,
      "learning_rate": 3.182620177259699e-07,
      "loss": 25.9312,
      "step": 10420
    },
    {
      "epoch": 82.78165503489531,
      "grad_norm": 468.8350830078125,
      "learning_rate": 3.179615457246102e-07,
      "loss": 23.6445,
      "step": 10430
    },
    {
      "epoch": 82.86141575274178,
      "grad_norm": 167.1678924560547,
      "learning_rate": 3.1766096767904113e-07,
      "loss": 23.0468,
      "step": 10440
    },
    {
      "epoch": 82.94117647058823,
      "grad_norm": 134.3607177734375,
      "learning_rate": 3.1736028405827163e-07,
      "loss": 25.7779,
      "step": 10450
    },
    {
      "epoch": 83.0159521435693,
      "grad_norm": 102.10858154296875,
      "learning_rate": 3.1705949533147515e-07,
      "loss": 21.8969,
      "step": 10460
    },
    {
      "epoch": 83.09571286141575,
      "grad_norm": 204.12930297851562,
      "learning_rate": 3.167586019679891e-07,
      "loss": 23.8331,
      "step": 10470
    },
    {
      "epoch": 83.17547357926222,
      "grad_norm": 258.52239990234375,
      "learning_rate": 3.1645760443731444e-07,
      "loss": 24.0841,
      "step": 10480
    },
    {
      "epoch": 83.25523429710867,
      "grad_norm": 129.42250061035156,
      "learning_rate": 3.161565032091144e-07,
      "loss": 25.0963,
      "step": 10490
    },
    {
      "epoch": 83.33499501495514,
      "grad_norm": 234.62936401367188,
      "learning_rate": 3.1585529875321406e-07,
      "loss": 25.2912,
      "step": 10500
    },
    {
      "epoch": 83.33499501495514,
      "eval_loss": 3.034513473510742,
      "eval_mae": 1.3175091743469238,
      "eval_mse": 3.034513473510742,
      "eval_r2": 0.08571594953536987,
      "eval_rmse": 1.7419854975029907,
      "eval_runtime": 9.0687,
      "eval_samples_per_second": 442.289,
      "eval_steps_per_second": 13.894,
      "step": 10500
    },
    {
      "epoch": 83.4147557328016,
      "grad_norm": 234.0610809326172,
      "learning_rate": 3.155539915395997e-07,
      "loss": 24.0329,
      "step": 10510
    },
    {
      "epoch": 83.49451645064805,
      "grad_norm": 201.95191955566406,
      "learning_rate": 3.1525258203841783e-07,
      "loss": 25.5342,
      "step": 10520
    },
    {
      "epoch": 83.57427716849452,
      "grad_norm": 485.1669921875,
      "learning_rate": 3.149510707199745e-07,
      "loss": 23.8518,
      "step": 10530
    },
    {
      "epoch": 83.65403788634097,
      "grad_norm": 212.77342224121094,
      "learning_rate": 3.1464945805473487e-07,
      "loss": 26.4807,
      "step": 10540
    },
    {
      "epoch": 83.73379860418744,
      "grad_norm": 329.3431091308594,
      "learning_rate": 3.1434774451332206e-07,
      "loss": 26.6428,
      "step": 10550
    },
    {
      "epoch": 83.8135593220339,
      "grad_norm": 158.35450744628906,
      "learning_rate": 3.1404593056651654e-07,
      "loss": 23.2614,
      "step": 10560
    },
    {
      "epoch": 83.89332003988036,
      "grad_norm": 67.29667663574219,
      "learning_rate": 3.1374401668525554e-07,
      "loss": 22.8227,
      "step": 10570
    },
    {
      "epoch": 83.97308075772682,
      "grad_norm": 208.1637725830078,
      "learning_rate": 3.134420033406322e-07,
      "loss": 24.3643,
      "step": 10580
    },
    {
      "epoch": 84.04785643070788,
      "grad_norm": 71.72771453857422,
      "learning_rate": 3.1313989100389496e-07,
      "loss": 22.6355,
      "step": 10590
    },
    {
      "epoch": 84.12761714855434,
      "grad_norm": 177.875,
      "learning_rate": 3.1283768014644653e-07,
      "loss": 26.4483,
      "step": 10600
    },
    {
      "epoch": 84.12761714855434,
      "eval_loss": 3.030277967453003,
      "eval_mae": 1.312751054763794,
      "eval_mse": 3.030277967453003,
      "eval_r2": 0.08699202537536621,
      "eval_rmse": 1.7407693607864894,
      "eval_runtime": 9.0905,
      "eval_samples_per_second": 441.231,
      "eval_steps_per_second": 13.861,
      "step": 10600
    },
    {
      "epoch": 84.2073778664008,
      "grad_norm": 70.78500366210938,
      "learning_rate": 3.1253537123984363e-07,
      "loss": 23.6658,
      "step": 10610
    },
    {
      "epoch": 84.28713858424726,
      "grad_norm": 414.3121643066406,
      "learning_rate": 3.122329647557955e-07,
      "loss": 27.3614,
      "step": 10620
    },
    {
      "epoch": 84.36689930209371,
      "grad_norm": 101.00165557861328,
      "learning_rate": 3.1193046116616425e-07,
      "loss": 25.9633,
      "step": 10630
    },
    {
      "epoch": 84.44666001994018,
      "grad_norm": 600.6487426757812,
      "learning_rate": 3.116278609429629e-07,
      "loss": 25.1654,
      "step": 10640
    },
    {
      "epoch": 84.52642073778664,
      "grad_norm": 221.32412719726562,
      "learning_rate": 3.1132516455835576e-07,
      "loss": 24.6665,
      "step": 10650
    },
    {
      "epoch": 84.6061814556331,
      "grad_norm": 91.79109954833984,
      "learning_rate": 3.1102237248465686e-07,
      "loss": 24.0079,
      "step": 10660
    },
    {
      "epoch": 84.68594217347956,
      "grad_norm": 486.3948669433594,
      "learning_rate": 3.107194851943297e-07,
      "loss": 23.8476,
      "step": 10670
    },
    {
      "epoch": 84.76570289132601,
      "grad_norm": 155.642333984375,
      "learning_rate": 3.104165031599861e-07,
      "loss": 21.3082,
      "step": 10680
    },
    {
      "epoch": 84.84546360917248,
      "grad_norm": 319.09063720703125,
      "learning_rate": 3.1011342685438625e-07,
      "loss": 22.9344,
      "step": 10690
    },
    {
      "epoch": 84.92522432701894,
      "grad_norm": 218.4431915283203,
      "learning_rate": 3.0981025675043675e-07,
      "loss": 25.8942,
      "step": 10700
    },
    {
      "epoch": 84.92522432701894,
      "eval_loss": 3.028160810470581,
      "eval_mae": 1.3118433952331543,
      "eval_mse": 3.028160810470581,
      "eval_r2": 0.08762991428375244,
      "eval_rmse": 1.7401611449720917,
      "eval_runtime": 9.0799,
      "eval_samples_per_second": 441.745,
      "eval_steps_per_second": 13.877,
      "step": 10700
    },
    {
      "epoch": 85.0,
      "grad_norm": 211.14683532714844,
      "learning_rate": 3.095069933211912e-07,
      "loss": 23.4049,
      "step": 10710
    },
    {
      "epoch": 85.07976071784645,
      "grad_norm": 242.6691436767578,
      "learning_rate": 3.092036370398484e-07,
      "loss": 25.8015,
      "step": 10720
    },
    {
      "epoch": 85.15952143569292,
      "grad_norm": 625.7755737304688,
      "learning_rate": 3.0890018837975215e-07,
      "loss": 24.2249,
      "step": 10730
    },
    {
      "epoch": 85.23928215353938,
      "grad_norm": 193.0255584716797,
      "learning_rate": 3.0859664781439037e-07,
      "loss": 23.4788,
      "step": 10740
    },
    {
      "epoch": 85.31904287138585,
      "grad_norm": 272.1832275390625,
      "learning_rate": 3.082930158173945e-07,
      "loss": 25.8466,
      "step": 10750
    },
    {
      "epoch": 85.3988035892323,
      "grad_norm": 408.71661376953125,
      "learning_rate": 3.079892928625385e-07,
      "loss": 25.5229,
      "step": 10760
    },
    {
      "epoch": 85.47856430707877,
      "grad_norm": 426.63787841796875,
      "learning_rate": 3.076854794237384e-07,
      "loss": 25.0958,
      "step": 10770
    },
    {
      "epoch": 85.55832502492522,
      "grad_norm": 184.2642364501953,
      "learning_rate": 3.0738157597505127e-07,
      "loss": 23.289,
      "step": 10780
    },
    {
      "epoch": 85.63808574277168,
      "grad_norm": 945.2476196289062,
      "learning_rate": 3.070775829906747e-07,
      "loss": 23.4057,
      "step": 10790
    },
    {
      "epoch": 85.71784646061815,
      "grad_norm": 553.4732666015625,
      "learning_rate": 3.06773500944946e-07,
      "loss": 24.8294,
      "step": 10800
    },
    {
      "epoch": 85.71784646061815,
      "eval_loss": 3.0269877910614014,
      "eval_mae": 1.3075300455093384,
      "eval_mse": 3.0269877910614014,
      "eval_r2": 0.08798336982727051,
      "eval_rmse": 1.7398240689970355,
      "eval_runtime": 9.0905,
      "eval_samples_per_second": 441.23,
      "eval_steps_per_second": 13.861,
      "step": 10800
    },
    {
      "epoch": 85.7976071784646,
      "grad_norm": 112.29386138916016,
      "learning_rate": 3.064693303123415e-07,
      "loss": 24.8945,
      "step": 10810
    },
    {
      "epoch": 85.87736789631107,
      "grad_norm": 720.1716918945312,
      "learning_rate": 3.061650715674755e-07,
      "loss": 24.4036,
      "step": 10820
    },
    {
      "epoch": 85.95712861415753,
      "grad_norm": 325.7308349609375,
      "learning_rate": 3.058911637534923e-07,
      "loss": 24.3189,
      "step": 10830
    },
    {
      "epoch": 86.03190428713859,
      "grad_norm": 230.79962158203125,
      "learning_rate": 3.055867389033845e-07,
      "loss": 21.9985,
      "step": 10840
    },
    {
      "epoch": 86.11166500498504,
      "grad_norm": 496.41143798828125,
      "learning_rate": 3.052822273181721e-07,
      "loss": 26.2874,
      "step": 10850
    },
    {
      "epoch": 86.19142572283151,
      "grad_norm": 704.2967529296875,
      "learning_rate": 3.049776294730017e-07,
      "loss": 24.9313,
      "step": 10860
    },
    {
      "epoch": 86.27118644067797,
      "grad_norm": 557.0463256835938,
      "learning_rate": 3.0467294584315436e-07,
      "loss": 24.8474,
      "step": 10870
    },
    {
      "epoch": 86.35094715852442,
      "grad_norm": 339.0418395996094,
      "learning_rate": 3.043681769040449e-07,
      "loss": 25.2613,
      "step": 10880
    },
    {
      "epoch": 86.43070787637089,
      "grad_norm": 437.7363586425781,
      "learning_rate": 3.040633231312216e-07,
      "loss": 23.9345,
      "step": 10890
    },
    {
      "epoch": 86.51046859421734,
      "grad_norm": 398.3165283203125,
      "learning_rate": 3.0375838500036484e-07,
      "loss": 24.2951,
      "step": 10900
    },
    {
      "epoch": 86.51046859421734,
      "eval_loss": 3.0261216163635254,
      "eval_mae": 1.307518720626831,
      "eval_mse": 3.0261216163635254,
      "eval_r2": 0.08824437856674194,
      "eval_rmse": 1.7395751252428064,
      "eval_runtime": 9.1038,
      "eval_samples_per_second": 440.583,
      "eval_steps_per_second": 13.84,
      "step": 10900
    },
    {
      "epoch": 86.59022931206381,
      "grad_norm": 95.45066833496094,
      "learning_rate": 3.0345336298728646e-07,
      "loss": 23.1709,
      "step": 10910
    },
    {
      "epoch": 86.66999002991027,
      "grad_norm": 105.39664459228516,
      "learning_rate": 3.031482575679297e-07,
      "loss": 24.337,
      "step": 10920
    },
    {
      "epoch": 86.74975074775674,
      "grad_norm": 149.99327087402344,
      "learning_rate": 3.028430692183675e-07,
      "loss": 22.713,
      "step": 10930
    },
    {
      "epoch": 86.82951146560319,
      "grad_norm": 363.6456604003906,
      "learning_rate": 3.025377984148023e-07,
      "loss": 24.9756,
      "step": 10940
    },
    {
      "epoch": 86.90927218344964,
      "grad_norm": 639.7531127929688,
      "learning_rate": 3.0223244563356535e-07,
      "loss": 24.3099,
      "step": 10950
    },
    {
      "epoch": 86.98903290129611,
      "grad_norm": 609.5497436523438,
      "learning_rate": 3.0192701135111576e-07,
      "loss": 26.0549,
      "step": 10960
    },
    {
      "epoch": 87.06380857427718,
      "grad_norm": 71.53561401367188,
      "learning_rate": 3.0162149604403957e-07,
      "loss": 26.1172,
      "step": 10970
    },
    {
      "epoch": 87.14356929212363,
      "grad_norm": 281.8210754394531,
      "learning_rate": 3.0131590018904955e-07,
      "loss": 24.8328,
      "step": 10980
    },
    {
      "epoch": 87.22333000997008,
      "grad_norm": 528.9259033203125,
      "learning_rate": 3.010102242629841e-07,
      "loss": 23.5778,
      "step": 10990
    },
    {
      "epoch": 87.30309072781655,
      "grad_norm": 221.66729736328125,
      "learning_rate": 3.007044687428064e-07,
      "loss": 22.4509,
      "step": 11000
    },
    {
      "epoch": 87.30309072781655,
      "eval_loss": 3.0431015491485596,
      "eval_mae": 1.3019126653671265,
      "eval_mse": 3.0431013107299805,
      "eval_r2": 0.08312851190567017,
      "eval_rmse": 1.7444487125536194,
      "eval_runtime": 9.0783,
      "eval_samples_per_second": 441.823,
      "eval_steps_per_second": 13.879,
      "step": 11000
    },
    {
      "epoch": 87.38285144566301,
      "grad_norm": 143.44786071777344,
      "learning_rate": 3.003986341056038e-07,
      "loss": 25.9225,
      "step": 11010
    },
    {
      "epoch": 87.46261216350948,
      "grad_norm": 234.01194763183594,
      "learning_rate": 3.000927208285876e-07,
      "loss": 24.0189,
      "step": 11020
    },
    {
      "epoch": 87.54237288135593,
      "grad_norm": 247.8142852783203,
      "learning_rate": 2.997867293890912e-07,
      "loss": 24.129,
      "step": 11030
    },
    {
      "epoch": 87.6221335992024,
      "grad_norm": 198.8553924560547,
      "learning_rate": 2.994806602645702e-07,
      "loss": 24.8126,
      "step": 11040
    },
    {
      "epoch": 87.70189431704885,
      "grad_norm": 483.2439880371094,
      "learning_rate": 2.991745139326014e-07,
      "loss": 23.9214,
      "step": 11050
    },
    {
      "epoch": 87.78165503489531,
      "grad_norm": 314.6399841308594,
      "learning_rate": 2.988682908708822e-07,
      "loss": 23.0557,
      "step": 11060
    },
    {
      "epoch": 87.86141575274178,
      "grad_norm": 471.4364929199219,
      "learning_rate": 2.985619915572293e-07,
      "loss": 25.7889,
      "step": 11070
    },
    {
      "epoch": 87.94117647058823,
      "grad_norm": 498.03887939453125,
      "learning_rate": 2.9825561646957914e-07,
      "loss": 25.6226,
      "step": 11080
    },
    {
      "epoch": 88.0159521435693,
      "grad_norm": 281.46551513671875,
      "learning_rate": 2.9794916608598563e-07,
      "loss": 21.588,
      "step": 11090
    },
    {
      "epoch": 88.09571286141575,
      "grad_norm": 768.0538940429688,
      "learning_rate": 2.9764264088462057e-07,
      "loss": 23.5801,
      "step": 11100
    },
    {
      "epoch": 88.09571286141575,
      "eval_loss": 3.024217128753662,
      "eval_mae": 1.3123266696929932,
      "eval_mse": 3.024217128753662,
      "eval_r2": 0.08881819248199463,
      "eval_rmse": 1.7390276388699697,
      "eval_runtime": 9.0429,
      "eval_samples_per_second": 443.554,
      "eval_steps_per_second": 13.934,
      "step": 11100
    },
    {
      "epoch": 88.17547357926222,
      "grad_norm": 146.67355346679688,
      "learning_rate": 2.9733604134377237e-07,
      "loss": 26.1583,
      "step": 11110
    },
    {
      "epoch": 88.25523429710867,
      "grad_norm": 307.28668212890625,
      "learning_rate": 2.970293679418456e-07,
      "loss": 26.4266,
      "step": 11120
    },
    {
      "epoch": 88.33499501495514,
      "grad_norm": 218.9013214111328,
      "learning_rate": 2.967226211573598e-07,
      "loss": 24.7804,
      "step": 11130
    },
    {
      "epoch": 88.4147557328016,
      "grad_norm": 410.3785705566406,
      "learning_rate": 2.964158014689493e-07,
      "loss": 23.9603,
      "step": 11140
    },
    {
      "epoch": 88.49451645064805,
      "grad_norm": 278.336669921875,
      "learning_rate": 2.961089093553619e-07,
      "loss": 26.3096,
      "step": 11150
    },
    {
      "epoch": 88.57427716849452,
      "grad_norm": 621.4981689453125,
      "learning_rate": 2.9580194529545886e-07,
      "loss": 23.0465,
      "step": 11160
    },
    {
      "epoch": 88.65403788634097,
      "grad_norm": 165.18826293945312,
      "learning_rate": 2.95494909768213e-07,
      "loss": 23.998,
      "step": 11170
    },
    {
      "epoch": 88.73379860418744,
      "grad_norm": 293.3717041015625,
      "learning_rate": 2.9518780325270943e-07,
      "loss": 23.3783,
      "step": 11180
    },
    {
      "epoch": 88.8135593220339,
      "grad_norm": 271.273193359375,
      "learning_rate": 2.948806262281435e-07,
      "loss": 24.6496,
      "step": 11190
    },
    {
      "epoch": 88.89332003988036,
      "grad_norm": 340.6837158203125,
      "learning_rate": 2.9457337917382074e-07,
      "loss": 25.3883,
      "step": 11200
    },
    {
      "epoch": 88.89332003988036,
      "eval_loss": 3.0350711345672607,
      "eval_mae": 1.3221527338027954,
      "eval_mse": 3.03507137298584,
      "eval_r2": 0.08554786443710327,
      "eval_rmse": 1.7421456233581163,
      "eval_runtime": 9.137,
      "eval_samples_per_second": 438.983,
      "eval_steps_per_second": 13.79,
      "step": 11200
    },
    {
      "epoch": 88.97308075772682,
      "grad_norm": 373.01025390625,
      "learning_rate": 2.9426606256915596e-07,
      "loss": 22.5712,
      "step": 11210
    },
    {
      "epoch": 89.04785643070788,
      "grad_norm": 82.71126556396484,
      "learning_rate": 2.9395867689367247e-07,
      "loss": 26.1899,
      "step": 11220
    },
    {
      "epoch": 89.12761714855434,
      "grad_norm": 524.1490478515625,
      "learning_rate": 2.936512226270015e-07,
      "loss": 25.1711,
      "step": 11230
    },
    {
      "epoch": 89.2073778664008,
      "grad_norm": 611.0742797851562,
      "learning_rate": 2.9334370024888087e-07,
      "loss": 24.6458,
      "step": 11240
    },
    {
      "epoch": 89.28713858424726,
      "grad_norm": 149.1772003173828,
      "learning_rate": 2.9303611023915536e-07,
      "loss": 23.5635,
      "step": 11250
    },
    {
      "epoch": 89.36689930209371,
      "grad_norm": 138.20068359375,
      "learning_rate": 2.9272845307777467e-07,
      "loss": 24.7863,
      "step": 11260
    },
    {
      "epoch": 89.44666001994018,
      "grad_norm": 285.6581115722656,
      "learning_rate": 2.9242072924479355e-07,
      "loss": 25.1381,
      "step": 11270
    },
    {
      "epoch": 89.52642073778664,
      "grad_norm": 210.21249389648438,
      "learning_rate": 2.921129392203707e-07,
      "loss": 26.7203,
      "step": 11280
    },
    {
      "epoch": 89.6061814556331,
      "grad_norm": 69.21088409423828,
      "learning_rate": 2.9180508348476836e-07,
      "loss": 23.4501,
      "step": 11290
    },
    {
      "epoch": 89.68594217347956,
      "grad_norm": 387.38494873046875,
      "learning_rate": 2.914971625183509e-07,
      "loss": 23.5194,
      "step": 11300
    },
    {
      "epoch": 89.68594217347956,
      "eval_loss": 3.022109031677246,
      "eval_mae": 1.3071075677871704,
      "eval_mse": 3.022109270095825,
      "eval_r2": 0.08945322036743164,
      "eval_rmse": 1.7384214880447795,
      "eval_runtime": 9.0702,
      "eval_samples_per_second": 442.218,
      "eval_steps_per_second": 13.892,
      "step": 11300
    },
    {
      "epoch": 89.76570289132601,
      "grad_norm": 171.12950134277344,
      "learning_rate": 2.9118917680158463e-07,
      "loss": 23.5455,
      "step": 11310
    },
    {
      "epoch": 89.84546360917248,
      "grad_norm": 144.1996612548828,
      "learning_rate": 2.90881126815037e-07,
      "loss": 22.2914,
      "step": 11320
    },
    {
      "epoch": 89.92522432701894,
      "grad_norm": 207.52590942382812,
      "learning_rate": 2.905730130393757e-07,
      "loss": 23.5734,
      "step": 11330
    },
    {
      "epoch": 90.0,
      "grad_norm": 225.84323120117188,
      "learning_rate": 2.9026483595536795e-07,
      "loss": 23.3279,
      "step": 11340
    },
    {
      "epoch": 90.07976071784645,
      "grad_norm": 454.1545104980469,
      "learning_rate": 2.8995659604387967e-07,
      "loss": 24.7852,
      "step": 11350
    },
    {
      "epoch": 90.15952143569292,
      "grad_norm": 412.3381042480469,
      "learning_rate": 2.8964829378587484e-07,
      "loss": 26.0871,
      "step": 11360
    },
    {
      "epoch": 90.23928215353938,
      "grad_norm": 111.9667739868164,
      "learning_rate": 2.8933992966241484e-07,
      "loss": 24.6989,
      "step": 11370
    },
    {
      "epoch": 90.31904287138585,
      "grad_norm": 1320.287841796875,
      "learning_rate": 2.8903150415465735e-07,
      "loss": 24.0967,
      "step": 11380
    },
    {
      "epoch": 90.3988035892323,
      "grad_norm": 128.83531188964844,
      "learning_rate": 2.887230177438563e-07,
      "loss": 24.6157,
      "step": 11390
    },
    {
      "epoch": 90.47856430707877,
      "grad_norm": 792.4503784179688,
      "learning_rate": 2.8841447091135985e-07,
      "loss": 25.8294,
      "step": 11400
    },
    {
      "epoch": 90.47856430707877,
      "eval_loss": 3.02121639251709,
      "eval_mae": 1.3112235069274902,
      "eval_mse": 3.0212161540985107,
      "eval_r2": 0.0897223949432373,
      "eval_rmse": 1.7381645935004288,
      "eval_runtime": 9.0986,
      "eval_samples_per_second": 440.837,
      "eval_steps_per_second": 13.848,
      "step": 11400
    },
    {
      "epoch": 90.55832502492522,
      "grad_norm": 716.572509765625,
      "learning_rate": 2.881058641386113e-07,
      "loss": 24.2307,
      "step": 11410
    },
    {
      "epoch": 90.63808574277168,
      "grad_norm": 1163.026611328125,
      "learning_rate": 2.8779719790714695e-07,
      "loss": 24.2046,
      "step": 11420
    },
    {
      "epoch": 90.71784646061815,
      "grad_norm": 120.25177001953125,
      "learning_rate": 2.8748847269859604e-07,
      "loss": 24.3986,
      "step": 11430
    },
    {
      "epoch": 90.7976071784646,
      "grad_norm": 140.4076385498047,
      "learning_rate": 2.871796889946799e-07,
      "loss": 23.751,
      "step": 11440
    },
    {
      "epoch": 90.87736789631107,
      "grad_norm": 162.21217346191406,
      "learning_rate": 2.868708472772109e-07,
      "loss": 23.8611,
      "step": 11450
    },
    {
      "epoch": 90.95712861415753,
      "grad_norm": 423.1454162597656,
      "learning_rate": 2.865619480280922e-07,
      "loss": 23.9691,
      "step": 11460
    },
    {
      "epoch": 91.03190428713859,
      "grad_norm": 100.34803771972656,
      "learning_rate": 2.862529917293165e-07,
      "loss": 21.8766,
      "step": 11470
    },
    {
      "epoch": 91.11166500498504,
      "grad_norm": 195.40789794921875,
      "learning_rate": 2.8594397886296575e-07,
      "loss": 24.3027,
      "step": 11480
    },
    {
      "epoch": 91.19142572283151,
      "grad_norm": 97.23872375488281,
      "learning_rate": 2.8563490991121006e-07,
      "loss": 23.9407,
      "step": 11490
    },
    {
      "epoch": 91.27118644067797,
      "grad_norm": 614.5779418945312,
      "learning_rate": 2.8532578535630683e-07,
      "loss": 22.0744,
      "step": 11500
    },
    {
      "epoch": 91.27118644067797,
      "eval_loss": 3.0190556049346924,
      "eval_mae": 1.3079864978790283,
      "eval_mse": 3.0190556049346924,
      "eval_r2": 0.09037333726882935,
      "eval_rmse": 1.7375429793057473,
      "eval_runtime": 9.0572,
      "eval_samples_per_second": 442.854,
      "eval_steps_per_second": 13.912,
      "step": 11500
    },
    {
      "epoch": 91.35094715852442,
      "grad_norm": 434.12109375,
      "learning_rate": 2.8501660568060073e-07,
      "loss": 25.5922,
      "step": 11510
    },
    {
      "epoch": 91.43070787637089,
      "grad_norm": 52.7608642578125,
      "learning_rate": 2.8470737136652186e-07,
      "loss": 24.937,
      "step": 11520
    },
    {
      "epoch": 91.51046859421734,
      "grad_norm": 124.67076110839844,
      "learning_rate": 2.84398082896586e-07,
      "loss": 23.3723,
      "step": 11530
    },
    {
      "epoch": 91.59022931206381,
      "grad_norm": 158.92762756347656,
      "learning_rate": 2.8408874075339333e-07,
      "loss": 24.7098,
      "step": 11540
    },
    {
      "epoch": 91.66999002991027,
      "grad_norm": 210.74998474121094,
      "learning_rate": 2.8377934541962766e-07,
      "loss": 24.8712,
      "step": 11550
    },
    {
      "epoch": 91.74975074775674,
      "grad_norm": 453.7698974609375,
      "learning_rate": 2.834698973780558e-07,
      "loss": 23.5147,
      "step": 11560
    },
    {
      "epoch": 91.82951146560319,
      "grad_norm": 74.43639373779297,
      "learning_rate": 2.831603971115271e-07,
      "loss": 22.7803,
      "step": 11570
    },
    {
      "epoch": 91.90927218344964,
      "grad_norm": 875.6751098632812,
      "learning_rate": 2.8285084510297196e-07,
      "loss": 26.8268,
      "step": 11580
    },
    {
      "epoch": 91.98903290129611,
      "grad_norm": 164.56968688964844,
      "learning_rate": 2.8254124183540185e-07,
      "loss": 26.4578,
      "step": 11590
    },
    {
      "epoch": 92.06380857427718,
      "grad_norm": 126.27661895751953,
      "learning_rate": 2.82231587791908e-07,
      "loss": 22.2847,
      "step": 11600
    },
    {
      "epoch": 92.06380857427718,
      "eval_loss": 3.0181376934051514,
      "eval_mae": 1.30584716796875,
      "eval_mse": 3.0181376934051514,
      "eval_r2": 0.09064990282058716,
      "eval_rmse": 1.7372788185565238,
      "eval_runtime": 9.0573,
      "eval_samples_per_second": 442.849,
      "eval_steps_per_second": 13.911,
      "step": 11600
    },
    {
      "epoch": 92.14356929212363,
      "grad_norm": 81.6953353881836,
      "learning_rate": 2.819218834556611e-07,
      "loss": 25.436,
      "step": 11610
    },
    {
      "epoch": 92.22333000997008,
      "grad_norm": 546.46044921875,
      "learning_rate": 2.8161212930991e-07,
      "loss": 24.1348,
      "step": 11620
    },
    {
      "epoch": 92.30309072781655,
      "grad_norm": 314.1280212402344,
      "learning_rate": 2.8130232583798163e-07,
      "loss": 25.676,
      "step": 11630
    },
    {
      "epoch": 92.38285144566301,
      "grad_norm": 1007.1983642578125,
      "learning_rate": 2.809924735232797e-07,
      "loss": 24.7171,
      "step": 11640
    },
    {
      "epoch": 92.46261216350948,
      "grad_norm": 263.078125,
      "learning_rate": 2.806825728492841e-07,
      "loss": 23.8003,
      "step": 11650
    },
    {
      "epoch": 92.54237288135593,
      "grad_norm": 202.86151123046875,
      "learning_rate": 2.803726242995501e-07,
      "loss": 23.4182,
      "step": 11660
    },
    {
      "epoch": 92.6221335992024,
      "grad_norm": 133.31997680664062,
      "learning_rate": 2.8006262835770803e-07,
      "loss": 23.3144,
      "step": 11670
    },
    {
      "epoch": 92.70189431704885,
      "grad_norm": 346.8004150390625,
      "learning_rate": 2.797525855074618e-07,
      "loss": 23.841,
      "step": 11680
    },
    {
      "epoch": 92.78165503489531,
      "grad_norm": 271.6881103515625,
      "learning_rate": 2.7944249623258865e-07,
      "loss": 24.4543,
      "step": 11690
    },
    {
      "epoch": 92.86141575274178,
      "grad_norm": 335.6261901855469,
      "learning_rate": 2.7913236101693825e-07,
      "loss": 25.058,
      "step": 11700
    },
    {
      "epoch": 92.86141575274178,
      "eval_loss": 3.0201170444488525,
      "eval_mae": 1.3131814002990723,
      "eval_mse": 3.0201170444488525,
      "eval_r2": 0.0900534987449646,
      "eval_rmse": 1.7378483951279675,
      "eval_runtime": 9.0569,
      "eval_samples_per_second": 442.868,
      "eval_steps_per_second": 13.912,
      "step": 11700
    },
    {
      "epoch": 92.94117647058823,
      "grad_norm": 81.703857421875,
      "learning_rate": 2.7882218034443205e-07,
      "loss": 25.6484,
      "step": 11710
    },
    {
      "epoch": 93.0159521435693,
      "grad_norm": 325.6644592285156,
      "learning_rate": 2.7851195469906224e-07,
      "loss": 22.4518,
      "step": 11720
    },
    {
      "epoch": 93.09571286141575,
      "grad_norm": 160.6322784423828,
      "learning_rate": 2.7820168456489137e-07,
      "loss": 25.4646,
      "step": 11730
    },
    {
      "epoch": 93.17547357926222,
      "grad_norm": 157.33323669433594,
      "learning_rate": 2.778913704260512e-07,
      "loss": 23.0021,
      "step": 11740
    },
    {
      "epoch": 93.25523429710867,
      "grad_norm": 486.8304748535156,
      "learning_rate": 2.775810127667425e-07,
      "loss": 24.9866,
      "step": 11750
    },
    {
      "epoch": 93.33499501495514,
      "grad_norm": 686.371826171875,
      "learning_rate": 2.772706120712333e-07,
      "loss": 26.729,
      "step": 11760
    },
    {
      "epoch": 93.4147557328016,
      "grad_norm": 607.1654663085938,
      "learning_rate": 2.7696016882385965e-07,
      "loss": 25.4789,
      "step": 11770
    },
    {
      "epoch": 93.49451645064805,
      "grad_norm": 944.3500366210938,
      "learning_rate": 2.766496835090234e-07,
      "loss": 23.3535,
      "step": 11780
    },
    {
      "epoch": 93.57427716849452,
      "grad_norm": 238.7883758544922,
      "learning_rate": 2.763391566111921e-07,
      "loss": 22.0236,
      "step": 11790
    },
    {
      "epoch": 93.65403788634097,
      "grad_norm": 242.50460815429688,
      "learning_rate": 2.7602858861489836e-07,
      "loss": 25.0649,
      "step": 11800
    },
    {
      "epoch": 93.65403788634097,
      "eval_loss": 3.016071081161499,
      "eval_mae": 1.3057998418807983,
      "eval_mse": 3.016071319580078,
      "eval_r2": 0.09127247333526611,
      "eval_rmse": 1.7366840010721807,
      "eval_runtime": 9.0577,
      "eval_samples_per_second": 442.826,
      "eval_steps_per_second": 13.911,
      "step": 11800
    },
    {
      "epoch": 93.73379860418744,
      "grad_norm": 610.6392822265625,
      "learning_rate": 2.757179800047388e-07,
      "loss": 25.7441,
      "step": 11810
    },
    {
      "epoch": 93.8135593220339,
      "grad_norm": 132.06114196777344,
      "learning_rate": 2.7540733126537347e-07,
      "loss": 23.5077,
      "step": 11820
    },
    {
      "epoch": 93.89332003988036,
      "grad_norm": 398.90533447265625,
      "learning_rate": 2.75096642881525e-07,
      "loss": 26.1888,
      "step": 11830
    },
    {
      "epoch": 93.97308075772682,
      "grad_norm": 99.74539184570312,
      "learning_rate": 2.747859153379779e-07,
      "loss": 22.6881,
      "step": 11840
    },
    {
      "epoch": 94.04785643070788,
      "grad_norm": 413.5274963378906,
      "learning_rate": 2.744751491195778e-07,
      "loss": 22.3016,
      "step": 11850
    },
    {
      "epoch": 94.12761714855434,
      "grad_norm": 133.99134826660156,
      "learning_rate": 2.741643447112306e-07,
      "loss": 24.7051,
      "step": 11860
    },
    {
      "epoch": 94.2073778664008,
      "grad_norm": 280.4942321777344,
      "learning_rate": 2.7385350259790174e-07,
      "loss": 24.2716,
      "step": 11870
    },
    {
      "epoch": 94.28713858424726,
      "grad_norm": 116.86431884765625,
      "learning_rate": 2.7354262326461583e-07,
      "loss": 21.2392,
      "step": 11880
    },
    {
      "epoch": 94.36689930209371,
      "grad_norm": 579.6485595703125,
      "learning_rate": 2.732317071964551e-07,
      "loss": 26.4136,
      "step": 11890
    },
    {
      "epoch": 94.44666001994018,
      "grad_norm": 457.6625671386719,
      "learning_rate": 2.7292075487855953e-07,
      "loss": 23.9248,
      "step": 11900
    },
    {
      "epoch": 94.44666001994018,
      "eval_loss": 3.0183889865875244,
      "eval_mae": 1.309451699256897,
      "eval_mse": 3.0183889865875244,
      "eval_r2": 0.09057420492172241,
      "eval_rmse": 1.737351140842727,
      "eval_runtime": 9.0629,
      "eval_samples_per_second": 442.573,
      "eval_steps_per_second": 13.903,
      "step": 11900
    },
    {
      "epoch": 94.52642073778664,
      "grad_norm": 76.22840881347656,
      "learning_rate": 2.726097667961253e-07,
      "loss": 21.309,
      "step": 11910
    },
    {
      "epoch": 94.6061814556331,
      "grad_norm": 411.46966552734375,
      "learning_rate": 2.7229874343440457e-07,
      "loss": 24.1257,
      "step": 11920
    },
    {
      "epoch": 94.68594217347956,
      "grad_norm": 240.69602966308594,
      "learning_rate": 2.719876852787045e-07,
      "loss": 24.3583,
      "step": 11930
    },
    {
      "epoch": 94.76570289132601,
      "grad_norm": 134.96070861816406,
      "learning_rate": 2.716765928143867e-07,
      "loss": 25.1843,
      "step": 11940
    },
    {
      "epoch": 94.84546360917248,
      "grad_norm": 162.1640167236328,
      "learning_rate": 2.7136546652686597e-07,
      "loss": 27.6058,
      "step": 11950
    },
    {
      "epoch": 94.92522432701894,
      "grad_norm": 493.3135070800781,
      "learning_rate": 2.710543069016103e-07,
      "loss": 25.1438,
      "step": 11960
    },
    {
      "epoch": 95.0,
      "grad_norm": 204.4506378173828,
      "learning_rate": 2.707431144241394e-07,
      "loss": 22.6437,
      "step": 11970
    },
    {
      "epoch": 95.07976071784645,
      "grad_norm": 404.43487548828125,
      "learning_rate": 2.7043188958002433e-07,
      "loss": 22.4641,
      "step": 11980
    },
    {
      "epoch": 95.15952143569292,
      "grad_norm": 255.8114013671875,
      "learning_rate": 2.701206328548866e-07,
      "loss": 24.3847,
      "step": 11990
    },
    {
      "epoch": 95.23928215353938,
      "grad_norm": 99.701416015625,
      "learning_rate": 2.698093447343977e-07,
      "loss": 24.5863,
      "step": 12000
    },
    {
      "epoch": 95.23928215353938,
      "eval_loss": 3.018806219100952,
      "eval_mae": 1.302077293395996,
      "eval_mse": 3.018806219100952,
      "eval_r2": 0.09044843912124634,
      "eval_rmse": 1.7374712138913129,
      "eval_runtime": 9.0712,
      "eval_samples_per_second": 442.166,
      "eval_steps_per_second": 13.89,
      "step": 12000
    },
    {
      "epoch": 95.31904287138585,
      "grad_norm": 299.7443542480469,
      "learning_rate": 2.694980257042778e-07,
      "loss": 26.4086,
      "step": 12010
    },
    {
      "epoch": 95.3988035892323,
      "grad_norm": 141.34422302246094,
      "learning_rate": 2.691866762502954e-07,
      "loss": 24.9383,
      "step": 12020
    },
    {
      "epoch": 95.47856430707877,
      "grad_norm": 125.73936462402344,
      "learning_rate": 2.688752968582666e-07,
      "loss": 23.8145,
      "step": 12030
    },
    {
      "epoch": 95.55832502492522,
      "grad_norm": 325.127685546875,
      "learning_rate": 2.6856388801405417e-07,
      "loss": 22.7498,
      "step": 12040
    },
    {
      "epoch": 95.63808574277168,
      "grad_norm": 281.2471008300781,
      "learning_rate": 2.6825245020356656e-07,
      "loss": 23.1331,
      "step": 12050
    },
    {
      "epoch": 95.71784646061815,
      "grad_norm": 340.7892761230469,
      "learning_rate": 2.679409839127579e-07,
      "loss": 23.1226,
      "step": 12060
    },
    {
      "epoch": 95.7976071784646,
      "grad_norm": 173.33863830566406,
      "learning_rate": 2.676294896276264e-07,
      "loss": 26.4354,
      "step": 12070
    },
    {
      "epoch": 95.87736789631107,
      "grad_norm": 119.07532501220703,
      "learning_rate": 2.67317967834214e-07,
      "loss": 24.2303,
      "step": 12080
    },
    {
      "epoch": 95.95712861415753,
      "grad_norm": 769.8329467773438,
      "learning_rate": 2.670064190186056e-07,
      "loss": 25.4267,
      "step": 12090
    },
    {
      "epoch": 96.03190428713859,
      "grad_norm": 285.7257080078125,
      "learning_rate": 2.666948436669284e-07,
      "loss": 22.7673,
      "step": 12100
    },
    {
      "epoch": 96.03190428713859,
      "eval_loss": 3.018017292022705,
      "eval_mae": 1.3130460977554321,
      "eval_mse": 3.018017292022705,
      "eval_r2": 0.0906861424446106,
      "eval_rmse": 1.7372441659198932,
      "eval_runtime": 9.0712,
      "eval_samples_per_second": 442.169,
      "eval_steps_per_second": 13.89,
      "step": 12100
    },
    {
      "epoch": 96.11166500498504,
      "grad_norm": 663.724365234375,
      "learning_rate": 2.663832422653508e-07,
      "loss": 23.2171,
      "step": 12110
    },
    {
      "epoch": 96.19142572283151,
      "grad_norm": 966.6635131835938,
      "learning_rate": 2.66071615300082e-07,
      "loss": 23.5348,
      "step": 12120
    },
    {
      "epoch": 96.27118644067797,
      "grad_norm": 242.81092834472656,
      "learning_rate": 2.6575996325737095e-07,
      "loss": 24.815,
      "step": 12130
    },
    {
      "epoch": 96.35094715852442,
      "grad_norm": 310.6782531738281,
      "learning_rate": 2.654482866235059e-07,
      "loss": 24.0334,
      "step": 12140
    },
    {
      "epoch": 96.43070787637089,
      "grad_norm": 231.919921875,
      "learning_rate": 2.6513658588481316e-07,
      "loss": 22.9641,
      "step": 12150
    },
    {
      "epoch": 96.51046859421734,
      "grad_norm": 422.7674560546875,
      "learning_rate": 2.6482486152765707e-07,
      "loss": 26.6106,
      "step": 12160
    },
    {
      "epoch": 96.59022931206381,
      "grad_norm": 187.1031036376953,
      "learning_rate": 2.6451311403843856e-07,
      "loss": 24.2406,
      "step": 12170
    },
    {
      "epoch": 96.66999002991027,
      "grad_norm": 338.5788269042969,
      "learning_rate": 2.6420134390359473e-07,
      "loss": 22.6762,
      "step": 12180
    },
    {
      "epoch": 96.74975074775674,
      "grad_norm": 251.15664672851562,
      "learning_rate": 2.6388955160959794e-07,
      "loss": 26.2033,
      "step": 12190
    },
    {
      "epoch": 96.82951146560319,
      "grad_norm": 271.2379455566406,
      "learning_rate": 2.6357773764295527e-07,
      "loss": 26.1962,
      "step": 12200
    },
    {
      "epoch": 96.82951146560319,
      "eval_loss": 3.0130105018615723,
      "eval_mae": 1.3046283721923828,
      "eval_mse": 3.0130105018615723,
      "eval_r2": 0.09219461679458618,
      "eval_rmse": 1.735802552671695,
      "eval_runtime": 9.0648,
      "eval_samples_per_second": 442.482,
      "eval_steps_per_second": 13.9,
      "step": 12200
    },
    {
      "epoch": 96.90927218344964,
      "grad_norm": 424.63641357421875,
      "learning_rate": 2.632659024902074e-07,
      "loss": 24.5048,
      "step": 12210
    },
    {
      "epoch": 96.98903290129611,
      "grad_norm": 644.2691650390625,
      "learning_rate": 2.629540466379284e-07,
      "loss": 23.5517,
      "step": 12220
    },
    {
      "epoch": 97.06380857427718,
      "grad_norm": 154.794921875,
      "learning_rate": 2.6264217057272414e-07,
      "loss": 21.9214,
      "step": 12230
    },
    {
      "epoch": 97.14356929212363,
      "grad_norm": 594.4990234375,
      "learning_rate": 2.6233027478123266e-07,
      "loss": 24.9934,
      "step": 12240
    },
    {
      "epoch": 97.22333000997008,
      "grad_norm": 130.84840393066406,
      "learning_rate": 2.6201835975012215e-07,
      "loss": 26.1297,
      "step": 12250
    },
    {
      "epoch": 97.30309072781655,
      "grad_norm": 130.3658447265625,
      "learning_rate": 2.617064259660914e-07,
      "loss": 24.8638,
      "step": 12260
    },
    {
      "epoch": 97.38285144566301,
      "grad_norm": 263.8542175292969,
      "learning_rate": 2.613944739158679e-07,
      "loss": 24.361,
      "step": 12270
    },
    {
      "epoch": 97.46261216350948,
      "grad_norm": 865.0307006835938,
      "learning_rate": 2.610825040862081e-07,
      "loss": 24.5309,
      "step": 12280
    },
    {
      "epoch": 97.54237288135593,
      "grad_norm": 265.0056457519531,
      "learning_rate": 2.6077051696389596e-07,
      "loss": 21.9346,
      "step": 12290
    },
    {
      "epoch": 97.6221335992024,
      "grad_norm": 288.1367492675781,
      "learning_rate": 2.604585130357425e-07,
      "loss": 24.2231,
      "step": 12300
    },
    {
      "epoch": 97.6221335992024,
      "eval_loss": 3.013296365737915,
      "eval_mae": 1.3062692880630493,
      "eval_mse": 3.013295888900757,
      "eval_r2": 0.09210866689682007,
      "eval_rmse": 1.735884756802927,
      "eval_runtime": 9.062,
      "eval_samples_per_second": 442.62,
      "eval_steps_per_second": 13.904,
      "step": 12300
    },
    {
      "epoch": 97.70189431704885,
      "grad_norm": 181.19395446777344,
      "learning_rate": 2.601464927885848e-07,
      "loss": 22.9514,
      "step": 12310
    },
    {
      "epoch": 97.78165503489531,
      "grad_norm": 434.2189025878906,
      "learning_rate": 2.5983445670928584e-07,
      "loss": 24.1628,
      "step": 12320
    },
    {
      "epoch": 97.86141575274178,
      "grad_norm": 972.9837036132812,
      "learning_rate": 2.595224052847327e-07,
      "loss": 25.6507,
      "step": 12330
    },
    {
      "epoch": 97.94117647058823,
      "grad_norm": 75.59383392333984,
      "learning_rate": 2.5921033900183686e-07,
      "loss": 24.7835,
      "step": 12340
    },
    {
      "epoch": 98.0159521435693,
      "grad_norm": 193.68312072753906,
      "learning_rate": 2.5889825834753275e-07,
      "loss": 22.2126,
      "step": 12350
    },
    {
      "epoch": 98.09571286141575,
      "grad_norm": 606.6470336914062,
      "learning_rate": 2.5858616380877744e-07,
      "loss": 26.1878,
      "step": 12360
    },
    {
      "epoch": 98.17547357926222,
      "grad_norm": 162.39686584472656,
      "learning_rate": 2.582740558725495e-07,
      "loss": 25.0336,
      "step": 12370
    },
    {
      "epoch": 98.25523429710867,
      "grad_norm": 252.45741271972656,
      "learning_rate": 2.579619350258482e-07,
      "loss": 23.3831,
      "step": 12380
    },
    {
      "epoch": 98.33499501495514,
      "grad_norm": 379.183837890625,
      "learning_rate": 2.576498017556936e-07,
      "loss": 25.4473,
      "step": 12390
    },
    {
      "epoch": 98.4147557328016,
      "grad_norm": 163.36947631835938,
      "learning_rate": 2.5733765654912444e-07,
      "loss": 23.0562,
      "step": 12400
    },
    {
      "epoch": 98.4147557328016,
      "eval_loss": 3.018754482269287,
      "eval_mae": 1.313119888305664,
      "eval_mse": 3.018754243850708,
      "eval_r2": 0.09046405553817749,
      "eval_rmse": 1.737456256672584,
      "eval_runtime": 9.0448,
      "eval_samples_per_second": 443.459,
      "eval_steps_per_second": 13.931,
      "step": 12400
    },
    {
      "epoch": 98.49451645064805,
      "grad_norm": 354.3284912109375,
      "learning_rate": 2.5702549989319846e-07,
      "loss": 25.1572,
      "step": 12410
    },
    {
      "epoch": 98.57427716849452,
      "grad_norm": 224.61061096191406,
      "learning_rate": 2.5671333227499114e-07,
      "loss": 22.5498,
      "step": 12420
    },
    {
      "epoch": 98.65403788634097,
      "grad_norm": 147.78477478027344,
      "learning_rate": 2.564011541815953e-07,
      "loss": 24.1911,
      "step": 12430
    },
    {
      "epoch": 98.73379860418744,
      "grad_norm": 600.3642578125,
      "learning_rate": 2.560889661001197e-07,
      "loss": 22.6389,
      "step": 12440
    },
    {
      "epoch": 98.8135593220339,
      "grad_norm": 491.30615234375,
      "learning_rate": 2.5577676851768897e-07,
      "loss": 24.767,
      "step": 12450
    },
    {
      "epoch": 98.89332003988036,
      "grad_norm": 287.7174377441406,
      "learning_rate": 2.554645619214426e-07,
      "loss": 25.9285,
      "step": 12460
    },
    {
      "epoch": 98.97308075772682,
      "grad_norm": 302.0243835449219,
      "learning_rate": 2.5515234679853384e-07,
      "loss": 25.2223,
      "step": 12470
    },
    {
      "epoch": 99.04785643070788,
      "grad_norm": 507.0435791015625,
      "learning_rate": 2.548401236361296e-07,
      "loss": 22.9346,
      "step": 12480
    },
    {
      "epoch": 99.12761714855434,
      "grad_norm": 666.8696899414062,
      "learning_rate": 2.545278929214092e-07,
      "loss": 23.28,
      "step": 12490
    },
    {
      "epoch": 99.2073778664008,
      "grad_norm": 985.0255737304688,
      "learning_rate": 2.5421565514156363e-07,
      "loss": 24.0838,
      "step": 12500
    },
    {
      "epoch": 99.2073778664008,
      "eval_loss": 3.0152266025543213,
      "eval_mae": 1.3043137788772583,
      "eval_mse": 3.0152266025543213,
      "eval_r2": 0.09152692556381226,
      "eval_rmse": 1.7364407857898068,
      "eval_runtime": 9.0555,
      "eval_samples_per_second": 442.934,
      "eval_steps_per_second": 13.914,
      "step": 12500
    },
    {
      "epoch": 99.28713858424726,
      "grad_norm": 222.13906860351562,
      "learning_rate": 2.53903410783795e-07,
      "loss": 25.2658,
      "step": 12510
    },
    {
      "epoch": 99.36689930209371,
      "grad_norm": 152.34921264648438,
      "learning_rate": 2.535911603353158e-07,
      "loss": 24.5804,
      "step": 12520
    },
    {
      "epoch": 99.44666001994018,
      "grad_norm": 160.31858825683594,
      "learning_rate": 2.5327890428334787e-07,
      "loss": 23.144,
      "step": 12530
    },
    {
      "epoch": 99.52642073778664,
      "grad_norm": 447.6943359375,
      "learning_rate": 2.529666431151216e-07,
      "loss": 24.772,
      "step": 12540
    },
    {
      "epoch": 99.6061814556331,
      "grad_norm": 331.9676513671875,
      "learning_rate": 2.52654377317876e-07,
      "loss": 26.1398,
      "step": 12550
    },
    {
      "epoch": 99.68594217347956,
      "grad_norm": 339.2489929199219,
      "learning_rate": 2.5234210737885657e-07,
      "loss": 24.3045,
      "step": 12560
    },
    {
      "epoch": 99.76570289132601,
      "grad_norm": 315.4028625488281,
      "learning_rate": 2.5202983378531574e-07,
      "loss": 22.0449,
      "step": 12570
    },
    {
      "epoch": 99.84546360917248,
      "grad_norm": 145.0469207763672,
      "learning_rate": 2.5171755702451145e-07,
      "loss": 24.9994,
      "step": 12580
    },
    {
      "epoch": 99.92522432701894,
      "grad_norm": 112.26505279541016,
      "learning_rate": 2.514052775837068e-07,
      "loss": 24.0593,
      "step": 12590
    },
    {
      "epoch": 100.0,
      "grad_norm": 175.2232208251953,
      "learning_rate": 2.510929959501686e-07,
      "loss": 22.7972,
      "step": 12600
    },
    {
      "epoch": 100.0,
      "eval_loss": 3.0114493370056152,
      "eval_mae": 1.3021821975708008,
      "eval_mse": 3.0114493370056152,
      "eval_r2": 0.09266507625579834,
      "eval_rmse": 1.7353527990024435,
      "eval_runtime": 9.0557,
      "eval_samples_per_second": 442.926,
      "eval_steps_per_second": 13.914,
      "step": 12600
    },
    {
      "epoch": 100.07976071784645,
      "grad_norm": 133.75238037109375,
      "learning_rate": 2.507807126111676e-07,
      "loss": 24.7391,
      "step": 12610
    },
    {
      "epoch": 100.15952143569292,
      "grad_norm": 530.0465698242188,
      "learning_rate": 2.504684280539769e-07,
      "loss": 24.52,
      "step": 12620
    },
    {
      "epoch": 100.23928215353938,
      "grad_norm": 147.8439483642578,
      "learning_rate": 2.501561427658718e-07,
      "loss": 23.4067,
      "step": 12630
    },
    {
      "epoch": 100.31904287138585,
      "grad_norm": 270.1204528808594,
      "learning_rate": 2.498438572341282e-07,
      "loss": 24.5654,
      "step": 12640
    },
    {
      "epoch": 100.3988035892323,
      "grad_norm": 371.5059814453125,
      "learning_rate": 2.49531571946023e-07,
      "loss": 23.2518,
      "step": 12650
    },
    {
      "epoch": 100.47856430707877,
      "grad_norm": 1076.48388671875,
      "learning_rate": 2.4921928738883245e-07,
      "loss": 25.6973,
      "step": 12660
    },
    {
      "epoch": 100.55832502492522,
      "grad_norm": 365.20404052734375,
      "learning_rate": 2.4890700404983143e-07,
      "loss": 25.2121,
      "step": 12670
    },
    {
      "epoch": 100.63808574277168,
      "grad_norm": 402.4895324707031,
      "learning_rate": 2.4859472241629325e-07,
      "loss": 23.8951,
      "step": 12680
    },
    {
      "epoch": 100.71784646061815,
      "grad_norm": 89.0658950805664,
      "learning_rate": 2.4828244297548847e-07,
      "loss": 24.1582,
      "step": 12690
    },
    {
      "epoch": 100.7976071784646,
      "grad_norm": 401.564208984375,
      "learning_rate": 2.4797016621468424e-07,
      "loss": 24.6082,
      "step": 12700
    },
    {
      "epoch": 100.7976071784646,
      "eval_loss": 3.0105321407318115,
      "eval_mae": 1.3034520149230957,
      "eval_mse": 3.0105321407318115,
      "eval_r2": 0.09294140338897705,
      "eval_rmse": 1.7350885109215066,
      "eval_runtime": 9.0681,
      "eval_samples_per_second": 442.319,
      "eval_steps_per_second": 13.895,
      "step": 12700
    },
    {
      "epoch": 100.87736789631107,
      "grad_norm": 164.41128540039062,
      "learning_rate": 2.476578926211434e-07,
      "loss": 24.2216,
      "step": 12710
    },
    {
      "epoch": 100.95712861415753,
      "grad_norm": 408.20245361328125,
      "learning_rate": 2.47345622682124e-07,
      "loss": 22.3349,
      "step": 12720
    },
    {
      "epoch": 101.03190428713859,
      "grad_norm": 675.1469116210938,
      "learning_rate": 2.4703335688487833e-07,
      "loss": 23.9997,
      "step": 12730
    },
    {
      "epoch": 101.11166500498504,
      "grad_norm": 650.2384033203125,
      "learning_rate": 2.467210957166522e-07,
      "loss": 26.1559,
      "step": 12740
    },
    {
      "epoch": 101.19142572283151,
      "grad_norm": 754.37939453125,
      "learning_rate": 2.464088396646842e-07,
      "loss": 22.9195,
      "step": 12750
    },
    {
      "epoch": 101.27118644067797,
      "grad_norm": 78.88389587402344,
      "learning_rate": 2.46096589216205e-07,
      "loss": 25.7293,
      "step": 12760
    },
    {
      "epoch": 101.35094715852442,
      "grad_norm": 853.6570434570312,
      "learning_rate": 2.4578434485843634e-07,
      "loss": 25.0469,
      "step": 12770
    },
    {
      "epoch": 101.43070787637089,
      "grad_norm": 112.20699310302734,
      "learning_rate": 2.4547210707859085e-07,
      "loss": 24.1414,
      "step": 12780
    },
    {
      "epoch": 101.51046859421734,
      "grad_norm": 456.0918884277344,
      "learning_rate": 2.451598763638704e-07,
      "loss": 22.4086,
      "step": 12790
    },
    {
      "epoch": 101.59022931206381,
      "grad_norm": 497.4643859863281,
      "learning_rate": 2.4484765320146613e-07,
      "loss": 25.9025,
      "step": 12800
    },
    {
      "epoch": 101.59022931206381,
      "eval_loss": 3.0093250274658203,
      "eval_mae": 1.304242730140686,
      "eval_mse": 3.0093250274658203,
      "eval_r2": 0.09330511093139648,
      "eval_rmse": 1.7347406225328963,
      "eval_runtime": 9.0622,
      "eval_samples_per_second": 442.609,
      "eval_steps_per_second": 13.904,
      "step": 12800
    },
    {
      "epoch": 101.66999002991027,
      "grad_norm": 251.6475372314453,
      "learning_rate": 2.445354380785574e-07,
      "loss": 25.1064,
      "step": 12810
    },
    {
      "epoch": 101.74975074775674,
      "grad_norm": 1158.7481689453125,
      "learning_rate": 2.4422323148231106e-07,
      "loss": 23.5001,
      "step": 12820
    },
    {
      "epoch": 101.82951146560319,
      "grad_norm": 301.5332336425781,
      "learning_rate": 2.439110338998803e-07,
      "loss": 23.0398,
      "step": 12830
    },
    {
      "epoch": 101.90927218344964,
      "grad_norm": 755.46826171875,
      "learning_rate": 2.435988458184047e-07,
      "loss": 24.2049,
      "step": 12840
    },
    {
      "epoch": 101.98903290129611,
      "grad_norm": 554.9736938476562,
      "learning_rate": 2.433178850710021e-07,
      "loss": 22.9534,
      "step": 12850
    },
    {
      "epoch": 102.06380857427718,
      "grad_norm": 902.829345703125,
      "learning_rate": 2.430057163833565e-07,
      "loss": 21.9611,
      "step": 12860
    },
    {
      "epoch": 102.14356929212363,
      "grad_norm": 263.4960021972656,
      "learning_rate": 2.426935586092837e-07,
      "loss": 26.8969,
      "step": 12870
    },
    {
      "epoch": 102.22333000997008,
      "grad_norm": 396.06732177734375,
      "learning_rate": 2.4238141223586095e-07,
      "loss": 24.1553,
      "step": 12880
    },
    {
      "epoch": 102.30309072781655,
      "grad_norm": 95.65509033203125,
      "learning_rate": 2.420692777501478e-07,
      "loss": 26.1854,
      "step": 12890
    },
    {
      "epoch": 102.38285144566301,
      "grad_norm": 586.4658203125,
      "learning_rate": 2.4175715563918506e-07,
      "loss": 24.511,
      "step": 12900
    },
    {
      "epoch": 102.38285144566301,
      "eval_loss": 3.0080695152282715,
      "eval_mae": 1.3039594888687134,
      "eval_mse": 3.0080695152282715,
      "eval_r2": 0.09368336200714111,
      "eval_rmse": 1.734378711593368,
      "eval_runtime": 9.0565,
      "eval_samples_per_second": 442.885,
      "eval_steps_per_second": 13.913,
      "step": 12900
    },
    {
      "epoch": 102.46261216350948,
      "grad_norm": 553.6513671875,
      "learning_rate": 2.414450463899944e-07,
      "loss": 21.9434,
      "step": 12910
    },
    {
      "epoch": 102.54237288135593,
      "grad_norm": 198.5731201171875,
      "learning_rate": 2.4113295048957736e-07,
      "loss": 26.5607,
      "step": 12920
    },
    {
      "epoch": 102.6221335992024,
      "grad_norm": 322.967041015625,
      "learning_rate": 2.4082086842491485e-07,
      "loss": 22.1324,
      "step": 12930
    },
    {
      "epoch": 102.70189431704885,
      "grad_norm": 430.9654846191406,
      "learning_rate": 2.4050880068296575e-07,
      "loss": 21.6502,
      "step": 12940
    },
    {
      "epoch": 102.78165503489531,
      "grad_norm": 337.82489013671875,
      "learning_rate": 2.401967477506669e-07,
      "loss": 23.6025,
      "step": 12950
    },
    {
      "epoch": 102.86141575274178,
      "grad_norm": 136.6009521484375,
      "learning_rate": 2.398847101149321e-07,
      "loss": 25.4659,
      "step": 12960
    },
    {
      "epoch": 102.94117647058823,
      "grad_norm": 982.8609619140625,
      "learning_rate": 2.3957268826265124e-07,
      "loss": 25.8761,
      "step": 12970
    },
    {
      "epoch": 103.0159521435693,
      "grad_norm": 431.97381591796875,
      "learning_rate": 2.3926068268068915e-07,
      "loss": 20.8754,
      "step": 12980
    },
    {
      "epoch": 103.09571286141575,
      "grad_norm": 145.51699829101562,
      "learning_rate": 2.3894869385588595e-07,
      "loss": 24.1687,
      "step": 12990
    },
    {
      "epoch": 103.17547357926222,
      "grad_norm": 159.59939575195312,
      "learning_rate": 2.3863672227505515e-07,
      "loss": 24.4449,
      "step": 13000
    },
    {
      "epoch": 103.17547357926222,
      "eval_loss": 3.012988805770874,
      "eval_mae": 1.3111497163772583,
      "eval_mse": 3.012989044189453,
      "eval_r2": 0.09220117330551147,
      "eval_rmse": 1.7357963717525893,
      "eval_runtime": 9.0501,
      "eval_samples_per_second": 443.198,
      "eval_steps_per_second": 13.922,
      "step": 13000
    },
    {
      "epoch": 103.25523429710867,
      "grad_norm": 113.45722961425781,
      "learning_rate": 2.383247684249836e-07,
      "loss": 26.0927,
      "step": 13010
    },
    {
      "epoch": 103.33499501495514,
      "grad_norm": 269.3424987792969,
      "learning_rate": 2.380128327924303e-07,
      "loss": 26.8863,
      "step": 13020
    },
    {
      "epoch": 103.4147557328016,
      "grad_norm": 143.92015075683594,
      "learning_rate": 2.377009158641259e-07,
      "loss": 23.3071,
      "step": 13030
    },
    {
      "epoch": 103.49451645064805,
      "grad_norm": 653.4163208007812,
      "learning_rate": 2.3738901812677184e-07,
      "loss": 23.7985,
      "step": 13040
    },
    {
      "epoch": 103.57427716849452,
      "grad_norm": 617.9258422851562,
      "learning_rate": 2.3707714006703978e-07,
      "loss": 24.1426,
      "step": 13050
    },
    {
      "epoch": 103.65403788634097,
      "grad_norm": 101.50848388671875,
      "learning_rate": 2.3676528217157034e-07,
      "loss": 22.7891,
      "step": 13060
    },
    {
      "epoch": 103.73379860418744,
      "grad_norm": 145.66314697265625,
      "learning_rate": 2.3645344492697303e-07,
      "loss": 25.2242,
      "step": 13070
    },
    {
      "epoch": 103.8135593220339,
      "grad_norm": 667.2296752929688,
      "learning_rate": 2.3614162881982492e-07,
      "loss": 21.9651,
      "step": 13080
    },
    {
      "epoch": 103.89332003988036,
      "grad_norm": 245.80865478515625,
      "learning_rate": 2.3582983433667023e-07,
      "loss": 24.2691,
      "step": 13090
    },
    {
      "epoch": 103.97308075772682,
      "grad_norm": 120.25707244873047,
      "learning_rate": 2.3551806196401925e-07,
      "loss": 24.1889,
      "step": 13100
    },
    {
      "epoch": 103.97308075772682,
      "eval_loss": 3.008435010910034,
      "eval_mae": 1.3024288415908813,
      "eval_mse": 3.008435010910034,
      "eval_r2": 0.09357321262359619,
      "eval_rmse": 1.734484076291862,
      "eval_runtime": 9.0327,
      "eval_samples_per_second": 444.052,
      "eval_steps_per_second": 13.949,
      "step": 13100
    },
    {
      "epoch": 104.04785643070788,
      "grad_norm": 273.62322998046875,
      "learning_rate": 2.3520631218834785e-07,
      "loss": 22.3676,
      "step": 13110
    },
    {
      "epoch": 104.12761714855434,
      "grad_norm": 240.7616424560547,
      "learning_rate": 2.348945854960969e-07,
      "loss": 25.3213,
      "step": 13120
    },
    {
      "epoch": 104.2073778664008,
      "grad_norm": 803.4730224609375,
      "learning_rate": 2.34582882373671e-07,
      "loss": 26.0542,
      "step": 13130
    },
    {
      "epoch": 104.28713858424726,
      "grad_norm": 475.0230712890625,
      "learning_rate": 2.3427120330743772e-07,
      "loss": 23.7392,
      "step": 13140
    },
    {
      "epoch": 104.36689930209371,
      "grad_norm": 298.427490234375,
      "learning_rate": 2.3395954878372765e-07,
      "loss": 24.3703,
      "step": 13150
    },
    {
      "epoch": 104.44666001994018,
      "grad_norm": 445.6641845703125,
      "learning_rate": 2.3364791928883273e-07,
      "loss": 25.0728,
      "step": 13160
    },
    {
      "epoch": 104.52642073778664,
      "grad_norm": 1277.1727294921875,
      "learning_rate": 2.3333631530900598e-07,
      "loss": 22.9316,
      "step": 13170
    },
    {
      "epoch": 104.6061814556331,
      "grad_norm": 218.64808654785156,
      "learning_rate": 2.3302473733046046e-07,
      "loss": 23.0187,
      "step": 13180
    },
    {
      "epoch": 104.68594217347956,
      "grad_norm": 415.3343505859375,
      "learning_rate": 2.3271318583936882e-07,
      "loss": 22.822,
      "step": 13190
    },
    {
      "epoch": 104.76570289132601,
      "grad_norm": 667.0796508789062,
      "learning_rate": 2.3240166132186228e-07,
      "loss": 24.5512,
      "step": 13200
    },
    {
      "epoch": 104.76570289132601,
      "eval_loss": 3.019092559814453,
      "eval_mae": 1.2983890771865845,
      "eval_mse": 3.019092559814453,
      "eval_r2": 0.09036219120025635,
      "eval_rmse": 1.7375536135079266,
      "eval_runtime": 9.0355,
      "eval_samples_per_second": 443.914,
      "eval_steps_per_second": 13.945,
      "step": 13200
    },
    {
      "epoch": 104.84546360917248,
      "grad_norm": 70.65787506103516,
      "learning_rate": 2.3209016426403012e-07,
      "loss": 22.9797,
      "step": 13210
    },
    {
      "epoch": 104.92522432701894,
      "grad_norm": 702.3170776367188,
      "learning_rate": 2.3177869515191846e-07,
      "loss": 26.5521,
      "step": 13220
    },
    {
      "epoch": 105.0,
      "grad_norm": 385.10113525390625,
      "learning_rate": 2.3146725447153015e-07,
      "loss": 22.5486,
      "step": 13230
    },
    {
      "epoch": 105.07976071784645,
      "grad_norm": 162.52066040039062,
      "learning_rate": 2.311558427088235e-07,
      "loss": 25.1436,
      "step": 13240
    },
    {
      "epoch": 105.15952143569292,
      "grad_norm": 175.56246948242188,
      "learning_rate": 2.308444603497117e-07,
      "loss": 23.1313,
      "step": 13250
    },
    {
      "epoch": 105.23928215353938,
      "grad_norm": 135.90574645996094,
      "learning_rate": 2.3053310788006232e-07,
      "loss": 23.9574,
      "step": 13260
    },
    {
      "epoch": 105.31904287138585,
      "grad_norm": 741.7534790039062,
      "learning_rate": 2.3022178578569577e-07,
      "loss": 22.7555,
      "step": 13270
    },
    {
      "epoch": 105.3988035892323,
      "grad_norm": 619.00341796875,
      "learning_rate": 2.2991049455238543e-07,
      "loss": 25.2691,
      "step": 13280
    },
    {
      "epoch": 105.47856430707877,
      "grad_norm": 187.55543518066406,
      "learning_rate": 2.2959923466585662e-07,
      "loss": 24.2961,
      "step": 13290
    },
    {
      "epoch": 105.55832502492522,
      "grad_norm": 225.87303161621094,
      "learning_rate": 2.2928800661178556e-07,
      "loss": 24.1411,
      "step": 13300
    },
    {
      "epoch": 105.55832502492522,
      "eval_loss": 3.007848024368286,
      "eval_mae": 1.307215929031372,
      "eval_mse": 3.007847785949707,
      "eval_r2": 0.09375017881393433,
      "eval_rmse": 1.7343147885979946,
      "eval_runtime": 9.0519,
      "eval_samples_per_second": 443.109,
      "eval_steps_per_second": 13.92,
      "step": 13300
    },
    {
      "epoch": 105.63808574277168,
      "grad_norm": 872.1564331054688,
      "learning_rate": 2.289768108757986e-07,
      "loss": 25.6431,
      "step": 13310
    },
    {
      "epoch": 105.71784646061815,
      "grad_norm": 203.1824493408203,
      "learning_rate": 2.2866564794347216e-07,
      "loss": 25.1792,
      "step": 13320
    },
    {
      "epoch": 105.7976071784646,
      "grad_norm": 400.51959228515625,
      "learning_rate": 2.2835451830033105e-07,
      "loss": 24.4278,
      "step": 13330
    },
    {
      "epoch": 105.87736789631107,
      "grad_norm": 103.01773071289062,
      "learning_rate": 2.2804342243184841e-07,
      "loss": 22.2797,
      "step": 13340
    },
    {
      "epoch": 105.95712861415753,
      "grad_norm": 441.1045837402344,
      "learning_rate": 2.2773236082344443e-07,
      "loss": 23.2526,
      "step": 13350
    },
    {
      "epoch": 106.03190428713859,
      "grad_norm": 530.3433837890625,
      "learning_rate": 2.2742133396048602e-07,
      "loss": 22.9374,
      "step": 13360
    },
    {
      "epoch": 106.11166500498504,
      "grad_norm": 233.91416931152344,
      "learning_rate": 2.2711034232828585e-07,
      "loss": 24.1659,
      "step": 13370
    },
    {
      "epoch": 106.19142572283151,
      "grad_norm": 145.86288452148438,
      "learning_rate": 2.2679938641210166e-07,
      "loss": 23.9056,
      "step": 13380
    },
    {
      "epoch": 106.27118644067797,
      "grad_norm": 311.06671142578125,
      "learning_rate": 2.2648846669713525e-07,
      "loss": 24.7324,
      "step": 13390
    },
    {
      "epoch": 106.35094715852442,
      "grad_norm": 700.2940673828125,
      "learning_rate": 2.2617758366853216e-07,
      "loss": 25.6364,
      "step": 13400
    },
    {
      "epoch": 106.35094715852442,
      "eval_loss": 3.0100624561309814,
      "eval_mae": 1.3105032444000244,
      "eval_mse": 3.0100626945495605,
      "eval_r2": 0.09308284521102905,
      "eval_rmse": 1.7349532254644677,
      "eval_runtime": 9.0428,
      "eval_samples_per_second": 443.557,
      "eval_steps_per_second": 13.934,
      "step": 13400
    },
    {
      "epoch": 106.43070787637089,
      "grad_norm": 1145.089111328125,
      "learning_rate": 2.2586673781138056e-07,
      "loss": 23.707,
      "step": 13410
    },
    {
      "epoch": 106.51046859421734,
      "grad_norm": 577.6820068359375,
      "learning_rate": 2.2555592961071085e-07,
      "loss": 23.0162,
      "step": 13420
    },
    {
      "epoch": 106.59022931206381,
      "grad_norm": 449.773193359375,
      "learning_rate": 2.2524515955149423e-07,
      "loss": 25.7381,
      "step": 13430
    },
    {
      "epoch": 106.66999002991027,
      "grad_norm": 265.3964538574219,
      "learning_rate": 2.249344281186427e-07,
      "loss": 22.8916,
      "step": 13440
    },
    {
      "epoch": 106.74975074775674,
      "grad_norm": 185.62965393066406,
      "learning_rate": 2.2462373579700806e-07,
      "loss": 26.5899,
      "step": 13450
    },
    {
      "epoch": 106.82951146560319,
      "grad_norm": 283.2608642578125,
      "learning_rate": 2.2431308307138094e-07,
      "loss": 25.1778,
      "step": 13460
    },
    {
      "epoch": 106.90927218344964,
      "grad_norm": 156.2736053466797,
      "learning_rate": 2.2400247042649004e-07,
      "loss": 23.2291,
      "step": 13470
    },
    {
      "epoch": 106.98903290129611,
      "grad_norm": 154.14137268066406,
      "learning_rate": 2.236918983470018e-07,
      "loss": 22.8961,
      "step": 13480
    },
    {
      "epoch": 107.06380857427718,
      "grad_norm": 113.89803314208984,
      "learning_rate": 2.2338136731751924e-07,
      "loss": 24.3879,
      "step": 13490
    },
    {
      "epoch": 107.14356929212363,
      "grad_norm": 179.24110412597656,
      "learning_rate": 2.2307087782258139e-07,
      "loss": 22.0681,
      "step": 13500
    },
    {
      "epoch": 107.14356929212363,
      "eval_loss": 3.004845142364502,
      "eval_mae": 1.3042222261428833,
      "eval_mse": 3.004845142364502,
      "eval_r2": 0.0946548581123352,
      "eval_rmse": 1.7334489154181907,
      "eval_runtime": 9.0518,
      "eval_samples_per_second": 443.117,
      "eval_steps_per_second": 13.92,
      "step": 13500
    },
    {
      "epoch": 107.22333000997008,
      "grad_norm": 345.2991943359375,
      "learning_rate": 2.2276043034666225e-07,
      "loss": 23.8587,
      "step": 13510
    },
    {
      "epoch": 107.30309072781655,
      "grad_norm": 252.33543395996094,
      "learning_rate": 2.2245002537417053e-07,
      "loss": 24.27,
      "step": 13520
    },
    {
      "epoch": 107.38285144566301,
      "grad_norm": 220.6058349609375,
      "learning_rate": 2.221396633894485e-07,
      "loss": 23.0785,
      "step": 13530
    },
    {
      "epoch": 107.46261216350948,
      "grad_norm": 570.069091796875,
      "learning_rate": 2.2182934487677137e-07,
      "loss": 25.1232,
      "step": 13540
    },
    {
      "epoch": 107.54237288135593,
      "grad_norm": 559.7032470703125,
      "learning_rate": 2.2151907032034643e-07,
      "loss": 24.3478,
      "step": 13550
    },
    {
      "epoch": 107.6221335992024,
      "grad_norm": 784.8019409179688,
      "learning_rate": 2.2120884020431259e-07,
      "loss": 26.435,
      "step": 13560
    },
    {
      "epoch": 107.70189431704885,
      "grad_norm": 526.9161987304688,
      "learning_rate": 2.2089865501273915e-07,
      "loss": 23.47,
      "step": 13570
    },
    {
      "epoch": 107.78165503489531,
      "grad_norm": 127.85801696777344,
      "learning_rate": 2.2058851522962563e-07,
      "loss": 22.3902,
      "step": 13580
    },
    {
      "epoch": 107.86141575274178,
      "grad_norm": 259.3059997558594,
      "learning_rate": 2.2027842133890032e-07,
      "loss": 25.3022,
      "step": 13590
    },
    {
      "epoch": 107.94117647058823,
      "grad_norm": 161.84365844726562,
      "learning_rate": 2.1996837382442016e-07,
      "loss": 24.8369,
      "step": 13600
    },
    {
      "epoch": 107.94117647058823,
      "eval_loss": 3.0051534175872803,
      "eval_mae": 1.299922227859497,
      "eval_mse": 3.0051534175872803,
      "eval_r2": 0.09456199407577515,
      "eval_rmse": 1.7335378327533784,
      "eval_runtime": 9.0565,
      "eval_samples_per_second": 442.888,
      "eval_steps_per_second": 13.913,
      "step": 13600
    },
    {
      "epoch": 108.0159521435693,
      "grad_norm": 451.6736145019531,
      "learning_rate": 2.196583731699697e-07,
      "loss": 21.9785,
      "step": 13610
    },
    {
      "epoch": 108.09571286141575,
      "grad_norm": 244.07496643066406,
      "learning_rate": 2.193484198592605e-07,
      "loss": 24.9731,
      "step": 13620
    },
    {
      "epoch": 108.17547357926222,
      "grad_norm": 86.17716217041016,
      "learning_rate": 2.1903851437592973e-07,
      "loss": 23.5088,
      "step": 13630
    },
    {
      "epoch": 108.25523429710867,
      "grad_norm": 351.5981140136719,
      "learning_rate": 2.1872865720354056e-07,
      "loss": 25.4712,
      "step": 13640
    },
    {
      "epoch": 108.33499501495514,
      "grad_norm": 787.5492553710938,
      "learning_rate": 2.1841884882558038e-07,
      "loss": 23.8798,
      "step": 13650
    },
    {
      "epoch": 108.4147557328016,
      "grad_norm": 407.19866943359375,
      "learning_rate": 2.1810908972546065e-07,
      "loss": 23.8723,
      "step": 13660
    },
    {
      "epoch": 108.49451645064805,
      "grad_norm": 136.1316375732422,
      "learning_rate": 2.177993803865158e-07,
      "loss": 25.1518,
      "step": 13670
    },
    {
      "epoch": 108.57427716849452,
      "grad_norm": 170.29415893554688,
      "learning_rate": 2.174897212920027e-07,
      "loss": 23.5138,
      "step": 13680
    },
    {
      "epoch": 108.65403788634097,
      "grad_norm": 222.90199279785156,
      "learning_rate": 2.1718011292509974e-07,
      "loss": 24.3218,
      "step": 13690
    },
    {
      "epoch": 108.73379860418744,
      "grad_norm": 196.23448181152344,
      "learning_rate": 2.1687055576890632e-07,
      "loss": 23.3596,
      "step": 13700
    },
    {
      "epoch": 108.73379860418744,
      "eval_loss": 3.0122363567352295,
      "eval_mae": 1.2974759340286255,
      "eval_mse": 3.0122363567352295,
      "eval_r2": 0.09242790937423706,
      "eval_rmse": 1.7355795449172675,
      "eval_runtime": 9.1022,
      "eval_samples_per_second": 440.661,
      "eval_steps_per_second": 13.843,
      "step": 13700
    },
    {
      "epoch": 108.8135593220339,
      "grad_norm": 845.6932983398438,
      "learning_rate": 2.1656105030644167e-07,
      "loss": 25.9375,
      "step": 13710
    },
    {
      "epoch": 108.89332003988036,
      "grad_norm": 167.6203155517578,
      "learning_rate": 2.1625159702064465e-07,
      "loss": 22.899,
      "step": 13720
    },
    {
      "epoch": 108.97308075772682,
      "grad_norm": 1029.1893310546875,
      "learning_rate": 2.1594219639437244e-07,
      "loss": 22.4538,
      "step": 13730
    },
    {
      "epoch": 109.04785643070788,
      "grad_norm": 602.4310913085938,
      "learning_rate": 2.1563284891040034e-07,
      "loss": 24.0515,
      "step": 13740
    },
    {
      "epoch": 109.12761714855434,
      "grad_norm": 740.744873046875,
      "learning_rate": 2.1532355505142044e-07,
      "loss": 24.1448,
      "step": 13750
    },
    {
      "epoch": 109.2073778664008,
      "grad_norm": 124.3176498413086,
      "learning_rate": 2.1501431530004133e-07,
      "loss": 25.6922,
      "step": 13760
    },
    {
      "epoch": 109.28713858424726,
      "grad_norm": 1032.07421875,
      "learning_rate": 2.1470513013878705e-07,
      "loss": 23.7145,
      "step": 13770
    },
    {
      "epoch": 109.36689930209371,
      "grad_norm": 714.7402954101562,
      "learning_rate": 2.1439600005009674e-07,
      "loss": 26.4051,
      "step": 13780
    },
    {
      "epoch": 109.44666001994018,
      "grad_norm": 478.6904296875,
      "learning_rate": 2.1408692551632322e-07,
      "loss": 23.6819,
      "step": 13790
    },
    {
      "epoch": 109.52642073778664,
      "grad_norm": 928.071044921875,
      "learning_rate": 2.137779070197328e-07,
      "loss": 22.9517,
      "step": 13800
    },
    {
      "epoch": 109.52642073778664,
      "eval_loss": 3.0034821033477783,
      "eval_mae": 1.3013429641723633,
      "eval_mse": 3.0034821033477783,
      "eval_r2": 0.09506553411483765,
      "eval_rmse": 1.7330557127074069,
      "eval_runtime": 9.1203,
      "eval_samples_per_second": 439.788,
      "eval_steps_per_second": 13.815,
      "step": 13800
    },
    {
      "epoch": 109.6061814556331,
      "grad_norm": 193.4709014892578,
      "learning_rate": 2.1346894504250446e-07,
      "loss": 24.7091,
      "step": 13810
    },
    {
      "epoch": 109.68594217347956,
      "grad_norm": 656.5834350585938,
      "learning_rate": 2.1316004006672888e-07,
      "loss": 22.5456,
      "step": 13820
    },
    {
      "epoch": 109.76570289132601,
      "grad_norm": 1204.993408203125,
      "learning_rate": 2.1285119257440784e-07,
      "loss": 25.7378,
      "step": 13830
    },
    {
      "epoch": 109.84546360917248,
      "grad_norm": 492.1981201171875,
      "learning_rate": 2.125424030474533e-07,
      "loss": 24.8542,
      "step": 13840
    },
    {
      "epoch": 109.92522432701894,
      "grad_norm": 255.29493713378906,
      "learning_rate": 2.1223367196768688e-07,
      "loss": 22.1661,
      "step": 13850
    },
    {
      "epoch": 110.0,
      "grad_norm": 339.08758544921875,
      "learning_rate": 2.1192499981683904e-07,
      "loss": 21.9219,
      "step": 13860
    },
    {
      "epoch": 110.07976071784645,
      "grad_norm": 869.034423828125,
      "learning_rate": 2.1161638707654828e-07,
      "loss": 23.117,
      "step": 13870
    },
    {
      "epoch": 110.15952143569292,
      "grad_norm": 211.4332733154297,
      "learning_rate": 2.1130783422836025e-07,
      "loss": 24.1192,
      "step": 13880
    },
    {
      "epoch": 110.23928215353938,
      "grad_norm": 340.9952697753906,
      "learning_rate": 2.1099934175372725e-07,
      "loss": 23.43,
      "step": 13890
    },
    {
      "epoch": 110.31904287138585,
      "grad_norm": 476.6876220703125,
      "learning_rate": 2.1069091013400742e-07,
      "loss": 23.6174,
      "step": 13900
    },
    {
      "epoch": 110.31904287138585,
      "eval_loss": 3.0029051303863525,
      "eval_mae": 1.2983053922653198,
      "eval_mse": 3.0029051303863525,
      "eval_r2": 0.09523940086364746,
      "eval_rmse": 1.7328892435428043,
      "eval_runtime": 9.0663,
      "eval_samples_per_second": 442.409,
      "eval_steps_per_second": 13.898,
      "step": 13900
    },
    {
      "epoch": 110.3988035892323,
      "grad_norm": 759.9443969726562,
      "learning_rate": 2.1038253985046405e-07,
      "loss": 24.5152,
      "step": 13910
    },
    {
      "epoch": 110.47856430707877,
      "grad_norm": 207.55154418945312,
      "learning_rate": 2.1007423138426435e-07,
      "loss": 24.6316,
      "step": 13920
    },
    {
      "epoch": 110.55832502492522,
      "grad_norm": 635.3157348632812,
      "learning_rate": 2.0976598521647933e-07,
      "loss": 26.1485,
      "step": 13930
    },
    {
      "epoch": 110.63808574277168,
      "grad_norm": 457.217041015625,
      "learning_rate": 2.0945780182808293e-07,
      "loss": 23.8358,
      "step": 13940
    },
    {
      "epoch": 110.71784646061815,
      "grad_norm": 196.58578491210938,
      "learning_rate": 2.0914968169995091e-07,
      "loss": 24.8853,
      "step": 13950
    },
    {
      "epoch": 110.7976071784646,
      "grad_norm": 198.74093627929688,
      "learning_rate": 2.088416253128602e-07,
      "loss": 24.7008,
      "step": 13960
    },
    {
      "epoch": 110.87736789631107,
      "grad_norm": 609.0301513671875,
      "learning_rate": 2.0853363314748866e-07,
      "loss": 23.6362,
      "step": 13970
    },
    {
      "epoch": 110.95712861415753,
      "grad_norm": 733.3483276367188,
      "learning_rate": 2.0822570568441362e-07,
      "loss": 22.5534,
      "step": 13980
    },
    {
      "epoch": 111.03190428713859,
      "grad_norm": 212.95008850097656,
      "learning_rate": 2.0791784340411162e-07,
      "loss": 25.2944,
      "step": 13990
    },
    {
      "epoch": 111.11166500498504,
      "grad_norm": 151.1733856201172,
      "learning_rate": 2.0761004678695737e-07,
      "loss": 26.0677,
      "step": 14000
    },
    {
      "epoch": 111.11166500498504,
      "eval_loss": 3.0067334175109863,
      "eval_mae": 1.3094115257263184,
      "eval_mse": 3.0067334175109863,
      "eval_r2": 0.0940859317779541,
      "eval_rmse": 1.7339934883127406,
      "eval_runtime": 9.0922,
      "eval_samples_per_second": 441.147,
      "eval_steps_per_second": 13.858,
      "step": 14000
    },
    {
      "epoch": 111.19142572283151,
      "grad_norm": 947.1033935546875,
      "learning_rate": 2.0730231631322325e-07,
      "loss": 22.3109,
      "step": 14010
    },
    {
      "epoch": 111.27118644067797,
      "grad_norm": 829.9580688476562,
      "learning_rate": 2.069946524630783e-07,
      "loss": 24.7575,
      "step": 14020
    },
    {
      "epoch": 111.35094715852442,
      "grad_norm": 245.0848388671875,
      "learning_rate": 2.066870557165878e-07,
      "loss": 26.3338,
      "step": 14030
    },
    {
      "epoch": 111.43070787637089,
      "grad_norm": 375.2764892578125,
      "learning_rate": 2.06379526553712e-07,
      "loss": 24.0418,
      "step": 14040
    },
    {
      "epoch": 111.51046859421734,
      "grad_norm": 151.50494384765625,
      "learning_rate": 2.06072065454306e-07,
      "loss": 22.7789,
      "step": 14050
    },
    {
      "epoch": 111.59022931206381,
      "grad_norm": 825.5809936523438,
      "learning_rate": 2.057646728981186e-07,
      "loss": 24.1213,
      "step": 14060
    },
    {
      "epoch": 111.66999002991027,
      "grad_norm": 328.1300048828125,
      "learning_rate": 2.0545734936479166e-07,
      "loss": 24.0528,
      "step": 14070
    },
    {
      "epoch": 111.74975074775674,
      "grad_norm": 421.9128723144531,
      "learning_rate": 2.0515009533385924e-07,
      "loss": 23.4848,
      "step": 14080
    },
    {
      "epoch": 111.82951146560319,
      "grad_norm": 530.5689086914062,
      "learning_rate": 2.0484291128474698e-07,
      "loss": 23.4212,
      "step": 14090
    },
    {
      "epoch": 111.90927218344964,
      "grad_norm": 478.70452880859375,
      "learning_rate": 2.045357976967715e-07,
      "loss": 23.7455,
      "step": 14100
    },
    {
      "epoch": 111.90927218344964,
      "eval_loss": 3.0043723583221436,
      "eval_mae": 1.299153208732605,
      "eval_mse": 3.0043723583221436,
      "eval_r2": 0.09479731321334839,
      "eval_rmse": 1.7333125391348623,
      "eval_runtime": 9.0905,
      "eval_samples_per_second": 441.228,
      "eval_steps_per_second": 13.861,
      "step": 14100
    },
    {
      "epoch": 111.98903290129611,
      "grad_norm": 353.66131591796875,
      "learning_rate": 2.0422875504913946e-07,
      "loss": 22.9223,
      "step": 14110
    },
    {
      "epoch": 112.06380857427718,
      "grad_norm": 609.3904418945312,
      "learning_rate": 2.0392178382094632e-07,
      "loss": 23.194,
      "step": 14120
    },
    {
      "epoch": 112.14356929212363,
      "grad_norm": 857.0011596679688,
      "learning_rate": 2.0361488449117682e-07,
      "loss": 22.9318,
      "step": 14130
    },
    {
      "epoch": 112.22333000997008,
      "grad_norm": 675.8402709960938,
      "learning_rate": 2.0330805753870306e-07,
      "loss": 22.2699,
      "step": 14140
    },
    {
      "epoch": 112.30309072781655,
      "grad_norm": 543.199951171875,
      "learning_rate": 2.0300130344228444e-07,
      "loss": 25.7679,
      "step": 14150
    },
    {
      "epoch": 112.38285144566301,
      "grad_norm": 442.4114685058594,
      "learning_rate": 2.0269462268056642e-07,
      "loss": 26.0135,
      "step": 14160
    },
    {
      "epoch": 112.46261216350948,
      "grad_norm": 256.444580078125,
      "learning_rate": 2.0238801573208028e-07,
      "loss": 21.2013,
      "step": 14170
    },
    {
      "epoch": 112.54237288135593,
      "grad_norm": 509.817138671875,
      "learning_rate": 2.02081483075242e-07,
      "loss": 23.3253,
      "step": 14180
    },
    {
      "epoch": 112.6221335992024,
      "grad_norm": 580.7782592773438,
      "learning_rate": 2.017750251883517e-07,
      "loss": 26.1034,
      "step": 14190
    },
    {
      "epoch": 112.70189431704885,
      "grad_norm": 255.4623260498047,
      "learning_rate": 2.0146864254959276e-07,
      "loss": 24.2658,
      "step": 14200
    },
    {
      "epoch": 112.70189431704885,
      "eval_loss": 3.004279851913452,
      "eval_mae": 1.2976210117340088,
      "eval_mse": 3.004279851913452,
      "eval_r2": 0.09482520818710327,
      "eval_rmse": 1.7332858540683507,
      "eval_runtime": 9.0935,
      "eval_samples_per_second": 441.083,
      "eval_steps_per_second": 13.856,
      "step": 14200
    },
    {
      "epoch": 112.78165503489531,
      "grad_norm": 419.1800537109375,
      "learning_rate": 2.0116233563703117e-07,
      "loss": 25.155,
      "step": 14210
    },
    {
      "epoch": 112.86141575274178,
      "grad_norm": 174.2738800048828,
      "learning_rate": 2.0085610492861484e-07,
      "loss": 23.2744,
      "step": 14220
    },
    {
      "epoch": 112.94117647058823,
      "grad_norm": 141.05087280273438,
      "learning_rate": 2.0054995090217266e-07,
      "loss": 24.9201,
      "step": 14230
    },
    {
      "epoch": 113.0159521435693,
      "grad_norm": 350.5917053222656,
      "learning_rate": 2.0024387403541394e-07,
      "loss": 22.327,
      "step": 14240
    },
    {
      "epoch": 113.09571286141575,
      "grad_norm": 748.1488037109375,
      "learning_rate": 1.999378748059275e-07,
      "loss": 24.5792,
      "step": 14250
    },
    {
      "epoch": 113.17547357926222,
      "grad_norm": 350.0252685546875,
      "learning_rate": 1.9963195369118112e-07,
      "loss": 23.2496,
      "step": 14260
    },
    {
      "epoch": 113.25523429710867,
      "grad_norm": 167.78985595703125,
      "learning_rate": 1.9932611116852088e-07,
      "loss": 26.3902,
      "step": 14270
    },
    {
      "epoch": 113.33499501495514,
      "grad_norm": 418.8966064453125,
      "learning_rate": 1.990203477151697e-07,
      "loss": 25.5467,
      "step": 14280
    },
    {
      "epoch": 113.4147557328016,
      "grad_norm": 363.3697814941406,
      "learning_rate": 1.9871466380822752e-07,
      "loss": 22.481,
      "step": 14290
    },
    {
      "epoch": 113.49451645064805,
      "grad_norm": 301.104736328125,
      "learning_rate": 1.9840905992467018e-07,
      "loss": 21.3924,
      "step": 14300
    },
    {
      "epoch": 113.49451645064805,
      "eval_loss": 3.0020875930786133,
      "eval_mae": 1.3055925369262695,
      "eval_mse": 3.0020875930786133,
      "eval_r2": 0.09548568725585938,
      "eval_rmse": 1.7326533389800203,
      "eval_runtime": 9.0853,
      "eval_samples_per_second": 441.484,
      "eval_steps_per_second": 13.869,
      "step": 14300
    },
    {
      "epoch": 113.57427716849452,
      "grad_norm": 490.4180603027344,
      "learning_rate": 1.9810353654134852e-07,
      "loss": 23.7989,
      "step": 14310
    },
    {
      "epoch": 113.65403788634097,
      "grad_norm": 273.2963562011719,
      "learning_rate": 1.977980941349877e-07,
      "loss": 24.4324,
      "step": 14320
    },
    {
      "epoch": 113.73379860418744,
      "grad_norm": 916.4816284179688,
      "learning_rate": 1.974927331821867e-07,
      "loss": 24.3051,
      "step": 14330
    },
    {
      "epoch": 113.8135593220339,
      "grad_norm": 641.0733642578125,
      "learning_rate": 1.9718745415941735e-07,
      "loss": 25.1196,
      "step": 14340
    },
    {
      "epoch": 113.89332003988036,
      "grad_norm": 709.544677734375,
      "learning_rate": 1.9688225754302362e-07,
      "loss": 24.2441,
      "step": 14350
    },
    {
      "epoch": 113.97308075772682,
      "grad_norm": 720.6399536132812,
      "learning_rate": 1.9657714380922086e-07,
      "loss": 24.4085,
      "step": 14360
    },
    {
      "epoch": 114.04785643070788,
      "grad_norm": 205.85848999023438,
      "learning_rate": 1.962721134340951e-07,
      "loss": 22.0088,
      "step": 14370
    },
    {
      "epoch": 114.12761714855434,
      "grad_norm": 115.24569702148438,
      "learning_rate": 1.9596716689360244e-07,
      "loss": 25.6115,
      "step": 14380
    },
    {
      "epoch": 114.2073778664008,
      "grad_norm": 688.999267578125,
      "learning_rate": 1.95662304663568e-07,
      "loss": 24.6186,
      "step": 14390
    },
    {
      "epoch": 114.28713858424726,
      "grad_norm": 570.3909301757812,
      "learning_rate": 1.9535752721968535e-07,
      "loss": 23.2304,
      "step": 14400
    },
    {
      "epoch": 114.28713858424726,
      "eval_loss": 3.0144083499908447,
      "eval_mae": 1.2939565181732178,
      "eval_mse": 3.0144081115722656,
      "eval_r2": 0.09177356958389282,
      "eval_rmse": 1.7362050891447893,
      "eval_runtime": 9.0826,
      "eval_samples_per_second": 441.612,
      "eval_steps_per_second": 13.873,
      "step": 14400
    },
    {
      "epoch": 114.36689930209371,
      "grad_norm": 666.518798828125,
      "learning_rate": 1.950528350375159e-07,
      "loss": 25.2675,
      "step": 14410
    },
    {
      "epoch": 114.44666001994018,
      "grad_norm": 186.58453369140625,
      "learning_rate": 1.9474822859248784e-07,
      "loss": 24.6354,
      "step": 14420
    },
    {
      "epoch": 114.52642073778664,
      "grad_norm": 414.18603515625,
      "learning_rate": 1.9444370835989585e-07,
      "loss": 25.7347,
      "step": 14430
    },
    {
      "epoch": 114.6061814556331,
      "grad_norm": 576.7401733398438,
      "learning_rate": 1.9413927481489988e-07,
      "loss": 24.0868,
      "step": 14440
    },
    {
      "epoch": 114.68594217347956,
      "grad_norm": 627.0114135742188,
      "learning_rate": 1.9383492843252448e-07,
      "loss": 24.3032,
      "step": 14450
    },
    {
      "epoch": 114.76570289132601,
      "grad_norm": 218.20523071289062,
      "learning_rate": 1.9353066968765851e-07,
      "loss": 22.7358,
      "step": 14460
    },
    {
      "epoch": 114.84546360917248,
      "grad_norm": 161.6814422607422,
      "learning_rate": 1.9322649905505396e-07,
      "loss": 23.4808,
      "step": 14470
    },
    {
      "epoch": 114.92522432701894,
      "grad_norm": 273.088134765625,
      "learning_rate": 1.929224170093253e-07,
      "loss": 22.8789,
      "step": 14480
    },
    {
      "epoch": 115.0,
      "grad_norm": 774.2197265625,
      "learning_rate": 1.926184240249487e-07,
      "loss": 21.4673,
      "step": 14490
    },
    {
      "epoch": 115.07976071784645,
      "grad_norm": 367.72430419921875,
      "learning_rate": 1.9231452057626157e-07,
      "loss": 24.946,
      "step": 14500
    },
    {
      "epoch": 115.07976071784645,
      "eval_loss": 3.001230001449585,
      "eval_mae": 1.3005801439285278,
      "eval_mse": 3.001230001449585,
      "eval_r2": 0.0957440733909607,
      "eval_rmse": 1.7324058420155437,
      "eval_runtime": 9.0947,
      "eval_samples_per_second": 441.027,
      "eval_steps_per_second": 13.854,
      "step": 14500
    },
    {
      "epoch": 115.15952143569292,
      "grad_norm": 261.6848449707031,
      "learning_rate": 1.9201070713746146e-07,
      "loss": 24.619,
      "step": 14510
    },
    {
      "epoch": 115.23928215353938,
      "grad_norm": 253.65447998046875,
      "learning_rate": 1.917069841826055e-07,
      "loss": 24.864,
      "step": 14520
    },
    {
      "epoch": 115.31904287138585,
      "grad_norm": 247.41769409179688,
      "learning_rate": 1.9140335218560963e-07,
      "loss": 23.7294,
      "step": 14530
    },
    {
      "epoch": 115.3988035892323,
      "grad_norm": 742.1896362304688,
      "learning_rate": 1.9109981162024788e-07,
      "loss": 22.7592,
      "step": 14540
    },
    {
      "epoch": 115.47856430707877,
      "grad_norm": 325.4307861328125,
      "learning_rate": 1.907963629601516e-07,
      "loss": 25.2781,
      "step": 14550
    },
    {
      "epoch": 115.55832502492522,
      "grad_norm": 158.41632080078125,
      "learning_rate": 1.904930066788088e-07,
      "loss": 23.2735,
      "step": 14560
    },
    {
      "epoch": 115.63808574277168,
      "grad_norm": 434.5899353027344,
      "learning_rate": 1.9018974324956323e-07,
      "loss": 23.4698,
      "step": 14570
    },
    {
      "epoch": 115.71784646061815,
      "grad_norm": 507.632080078125,
      "learning_rate": 1.8988657314561375e-07,
      "loss": 23.6647,
      "step": 14580
    },
    {
      "epoch": 115.7976071784646,
      "grad_norm": 140.050048828125,
      "learning_rate": 1.895834968400138e-07,
      "loss": 23.897,
      "step": 14590
    },
    {
      "epoch": 115.87736789631107,
      "grad_norm": 319.8278503417969,
      "learning_rate": 1.8928051480567042e-07,
      "loss": 23.4679,
      "step": 14600
    },
    {
      "epoch": 115.87736789631107,
      "eval_loss": 2.999934196472168,
      "eval_mae": 1.3039181232452393,
      "eval_mse": 2.999934196472168,
      "eval_r2": 0.0961344838142395,
      "eval_rmse": 1.7320318116224562,
      "eval_runtime": 9.0889,
      "eval_samples_per_second": 441.307,
      "eval_steps_per_second": 13.863,
      "step": 14600
    },
    {
      "epoch": 115.95712861415753,
      "grad_norm": 186.12600708007812,
      "learning_rate": 1.8897762751534317e-07,
      "loss": 24.3203,
      "step": 14610
    },
    {
      "epoch": 116.03190428713859,
      "grad_norm": 747.21435546875,
      "learning_rate": 1.8867483544164422e-07,
      "loss": 22.5211,
      "step": 14620
    },
    {
      "epoch": 116.11166500498504,
      "grad_norm": 534.8453979492188,
      "learning_rate": 1.8837213905703702e-07,
      "loss": 24.2478,
      "step": 14630
    },
    {
      "epoch": 116.19142572283151,
      "grad_norm": 704.918212890625,
      "learning_rate": 1.880695388338358e-07,
      "loss": 23.4607,
      "step": 14640
    },
    {
      "epoch": 116.27118644067797,
      "grad_norm": 923.6112060546875,
      "learning_rate": 1.8776703524420445e-07,
      "loss": 25.0206,
      "step": 14650
    },
    {
      "epoch": 116.35094715852442,
      "grad_norm": 232.6576385498047,
      "learning_rate": 1.874646287601564e-07,
      "loss": 25.6268,
      "step": 14660
    },
    {
      "epoch": 116.43070787637089,
      "grad_norm": 414.2278137207031,
      "learning_rate": 1.8716231985355342e-07,
      "loss": 21.9105,
      "step": 14670
    },
    {
      "epoch": 116.51046859421734,
      "grad_norm": 250.12484741210938,
      "learning_rate": 1.8686010899610504e-07,
      "loss": 24.9453,
      "step": 14680
    },
    {
      "epoch": 116.59022931206381,
      "grad_norm": 214.85546875,
      "learning_rate": 1.865579966593678e-07,
      "loss": 23.202,
      "step": 14690
    },
    {
      "epoch": 116.66999002991027,
      "grad_norm": 456.5260009765625,
      "learning_rate": 1.8625598331474451e-07,
      "loss": 24.1828,
      "step": 14700
    },
    {
      "epoch": 116.66999002991027,
      "eval_loss": 3.0012688636779785,
      "eval_mae": 1.2985434532165527,
      "eval_mse": 3.0012688636779785,
      "eval_r2": 0.09573233127593994,
      "eval_rmse": 1.7324170582391465,
      "eval_runtime": 9.1014,
      "eval_samples_per_second": 440.7,
      "eval_steps_per_second": 13.844,
      "step": 14700
    },
    {
      "epoch": 116.74975074775674,
      "grad_norm": 446.82501220703125,
      "learning_rate": 1.8595406943348352e-07,
      "loss": 25.0955,
      "step": 14710
    },
    {
      "epoch": 116.82951146560319,
      "grad_norm": 665.4845581054688,
      "learning_rate": 1.85652255486678e-07,
      "loss": 23.4462,
      "step": 14720
    },
    {
      "epoch": 116.90927218344964,
      "grad_norm": 176.94802856445312,
      "learning_rate": 1.853505419452651e-07,
      "loss": 25.5434,
      "step": 14730
    },
    {
      "epoch": 116.98903290129611,
      "grad_norm": 1226.932373046875,
      "learning_rate": 1.8504892928002548e-07,
      "loss": 23.575,
      "step": 14740
    },
    {
      "epoch": 117.06380857427718,
      "grad_norm": 566.0380859375,
      "learning_rate": 1.8474741796158215e-07,
      "loss": 22.8273,
      "step": 14750
    },
    {
      "epoch": 117.14356929212363,
      "grad_norm": 85.70375061035156,
      "learning_rate": 1.8444600846040035e-07,
      "loss": 24.948,
      "step": 14760
    },
    {
      "epoch": 117.22333000997008,
      "grad_norm": 285.765625,
      "learning_rate": 1.8414470124678597e-07,
      "loss": 24.4736,
      "step": 14770
    },
    {
      "epoch": 117.30309072781655,
      "grad_norm": 481.25933837890625,
      "learning_rate": 1.8384349679088558e-07,
      "loss": 25.229,
      "step": 14780
    },
    {
      "epoch": 117.38285144566301,
      "grad_norm": 679.0187377929688,
      "learning_rate": 1.8354239556268552e-07,
      "loss": 23.8175,
      "step": 14790
    },
    {
      "epoch": 117.46261216350948,
      "grad_norm": 844.68115234375,
      "learning_rate": 1.832413980320109e-07,
      "loss": 23.7498,
      "step": 14800
    },
    {
      "epoch": 117.46261216350948,
      "eval_loss": 2.9960711002349854,
      "eval_mae": 1.2975517511367798,
      "eval_mse": 2.9960711002349854,
      "eval_r2": 0.09729844331741333,
      "eval_rmse": 1.730916260318501,
      "eval_runtime": 9.0922,
      "eval_samples_per_second": 441.145,
      "eval_steps_per_second": 13.858,
      "step": 14800
    },
    {
      "epoch": 117.54237288135593,
      "grad_norm": 260.93499755859375,
      "learning_rate": 1.829405046685249e-07,
      "loss": 23.3996,
      "step": 14810
    },
    {
      "epoch": 117.6221335992024,
      "grad_norm": 917.9327392578125,
      "learning_rate": 1.8263971594172837e-07,
      "loss": 25.7007,
      "step": 14820
    },
    {
      "epoch": 117.70189431704885,
      "grad_norm": 245.7829132080078,
      "learning_rate": 1.8233903232095882e-07,
      "loss": 22.9093,
      "step": 14830
    },
    {
      "epoch": 117.78165503489531,
      "grad_norm": 128.6542205810547,
      "learning_rate": 1.8203845427538985e-07,
      "loss": 23.2522,
      "step": 14840
    },
    {
      "epoch": 117.86141575274178,
      "grad_norm": 222.97190856933594,
      "learning_rate": 1.817680246888133e-07,
      "loss": 22.5512,
      "step": 14850
    },
    {
      "epoch": 117.94117647058823,
      "grad_norm": 342.2579345703125,
      "learning_rate": 1.8146764852810764e-07,
      "loss": 25.3012,
      "step": 14860
    },
    {
      "epoch": 118.0159521435693,
      "grad_norm": 458.2828369140625,
      "learning_rate": 1.8116737930227148e-07,
      "loss": 22.3981,
      "step": 14870
    },
    {
      "epoch": 118.09571286141575,
      "grad_norm": 263.3362121582031,
      "learning_rate": 1.8086721747983166e-07,
      "loss": 24.4831,
      "step": 14880
    },
    {
      "epoch": 118.17547357926222,
      "grad_norm": 675.1715087890625,
      "learning_rate": 1.8056716352914753e-07,
      "loss": 24.5227,
      "step": 14890
    },
    {
      "epoch": 118.25523429710867,
      "grad_norm": 270.53472900390625,
      "learning_rate": 1.8026721791841003e-07,
      "loss": 24.3978,
      "step": 14900
    },
    {
      "epoch": 118.25523429710867,
      "eval_loss": 2.9966397285461426,
      "eval_mae": 1.299818515777588,
      "eval_mse": 2.9966397285461426,
      "eval_r2": 0.09712713956832886,
      "eval_rmse": 1.7310805089729775,
      "eval_runtime": 9.1,
      "eval_samples_per_second": 440.772,
      "eval_steps_per_second": 13.846,
      "step": 14900
    },
    {
      "epoch": 118.33499501495514,
      "grad_norm": 407.41754150390625,
      "learning_rate": 1.7996738111564135e-07,
      "loss": 24.0754,
      "step": 14910
    },
    {
      "epoch": 118.4147557328016,
      "grad_norm": 626.1751708984375,
      "learning_rate": 1.7966765358869333e-07,
      "loss": 23.3016,
      "step": 14920
    },
    {
      "epoch": 118.49451645064805,
      "grad_norm": 599.4439086914062,
      "learning_rate": 1.7936803580524782e-07,
      "loss": 23.8976,
      "step": 14930
    },
    {
      "epoch": 118.57427716849452,
      "grad_norm": 399.52374267578125,
      "learning_rate": 1.790685282328152e-07,
      "loss": 23.325,
      "step": 14940
    },
    {
      "epoch": 118.65403788634097,
      "grad_norm": 684.7321166992188,
      "learning_rate": 1.7876913133873396e-07,
      "loss": 24.2629,
      "step": 14950
    },
    {
      "epoch": 118.73379860418744,
      "grad_norm": 535.8107299804688,
      "learning_rate": 1.784698455901698e-07,
      "loss": 23.974,
      "step": 14960
    },
    {
      "epoch": 118.8135593220339,
      "grad_norm": 91.9757080078125,
      "learning_rate": 1.7817067145411507e-07,
      "loss": 22.9046,
      "step": 14970
    },
    {
      "epoch": 118.89332003988036,
      "grad_norm": 530.356689453125,
      "learning_rate": 1.7787160939738795e-07,
      "loss": 23.6471,
      "step": 14980
    },
    {
      "epoch": 118.97308075772682,
      "grad_norm": 458.6141052246094,
      "learning_rate": 1.7757265988663172e-07,
      "loss": 25.7141,
      "step": 14990
    },
    {
      "epoch": 119.04785643070788,
      "grad_norm": 1084.7567138671875,
      "learning_rate": 1.772738233883141e-07,
      "loss": 24.3464,
      "step": 15000
    },
    {
      "epoch": 119.04785643070788,
      "eval_loss": 2.9981815814971924,
      "eval_mae": 1.3042938709259033,
      "eval_mse": 2.9981815814971924,
      "eval_r2": 0.09666258096694946,
      "eval_rmse": 1.7315257957931762,
      "eval_runtime": 9.0853,
      "eval_samples_per_second": 441.483,
      "eval_steps_per_second": 13.869,
      "step": 15000
    },
    {
      "epoch": 119.12761714855434,
      "grad_norm": 122.95572662353516,
      "learning_rate": 1.7697510036872633e-07,
      "loss": 22.7137,
      "step": 15010
    },
    {
      "epoch": 119.2073778664008,
      "grad_norm": 285.8971862792969,
      "learning_rate": 1.766764912939828e-07,
      "loss": 24.7054,
      "step": 15020
    },
    {
      "epoch": 119.28713858424726,
      "grad_norm": 178.3037872314453,
      "learning_rate": 1.7637799663001996e-07,
      "loss": 22.655,
      "step": 15030
    },
    {
      "epoch": 119.36689930209371,
      "grad_norm": 458.0035705566406,
      "learning_rate": 1.760796168425957e-07,
      "loss": 25.3129,
      "step": 15040
    },
    {
      "epoch": 119.44666001994018,
      "grad_norm": 243.63291931152344,
      "learning_rate": 1.7578135239728874e-07,
      "loss": 24.8409,
      "step": 15050
    },
    {
      "epoch": 119.52642073778664,
      "grad_norm": 301.6819152832031,
      "learning_rate": 1.7548320375949782e-07,
      "loss": 27.2402,
      "step": 15060
    },
    {
      "epoch": 119.6061814556331,
      "grad_norm": 568.8759155273438,
      "learning_rate": 1.7518517139444112e-07,
      "loss": 24.0735,
      "step": 15070
    },
    {
      "epoch": 119.68594217347956,
      "grad_norm": 701.4216918945312,
      "learning_rate": 1.74887255767155e-07,
      "loss": 21.9961,
      "step": 15080
    },
    {
      "epoch": 119.76570289132601,
      "grad_norm": 497.3916320800781,
      "learning_rate": 1.7458945734249403e-07,
      "loss": 22.3489,
      "step": 15090
    },
    {
      "epoch": 119.84546360917248,
      "grad_norm": 642.3511962890625,
      "learning_rate": 1.7429177658512982e-07,
      "loss": 22.4557,
      "step": 15100
    },
    {
      "epoch": 119.84546360917248,
      "eval_loss": 3.0010323524475098,
      "eval_mae": 1.2982264757156372,
      "eval_mse": 3.0010321140289307,
      "eval_r2": 0.09580367803573608,
      "eval_rmse": 1.7323487276033456,
      "eval_runtime": 9.1132,
      "eval_samples_per_second": 440.133,
      "eval_steps_per_second": 13.826,
      "step": 15100
    },
    {
      "epoch": 119.92522432701894,
      "grad_norm": 551.1233520507812,
      "learning_rate": 1.739942139595504e-07,
      "loss": 26.5307,
      "step": 15110
    },
    {
      "epoch": 120.0,
      "grad_norm": 235.44761657714844,
      "learning_rate": 1.7369676993005915e-07,
      "loss": 20.6391,
      "step": 15120
    },
    {
      "epoch": 120.07976071784645,
      "grad_norm": 1141.905029296875,
      "learning_rate": 1.733994449607749e-07,
      "loss": 22.791,
      "step": 15130
    },
    {
      "epoch": 120.15952143569292,
      "grad_norm": 582.3103637695312,
      "learning_rate": 1.731022395156304e-07,
      "loss": 24.694,
      "step": 15140
    },
    {
      "epoch": 120.23928215353938,
      "grad_norm": 262.1140441894531,
      "learning_rate": 1.7280515405837203e-07,
      "loss": 23.0256,
      "step": 15150
    },
    {
      "epoch": 120.31904287138585,
      "grad_norm": 274.61083984375,
      "learning_rate": 1.7250818905255876e-07,
      "loss": 22.8228,
      "step": 15160
    },
    {
      "epoch": 120.3988035892323,
      "grad_norm": 195.396484375,
      "learning_rate": 1.7221134496156176e-07,
      "loss": 25.1282,
      "step": 15170
    },
    {
      "epoch": 120.47856430707877,
      "grad_norm": 929.0553588867188,
      "learning_rate": 1.7191462224856356e-07,
      "loss": 27.0083,
      "step": 15180
    },
    {
      "epoch": 120.55832502492522,
      "grad_norm": 284.4768981933594,
      "learning_rate": 1.7161802137655722e-07,
      "loss": 24.9673,
      "step": 15190
    },
    {
      "epoch": 120.63808574277168,
      "grad_norm": 919.1542358398438,
      "learning_rate": 1.713215428083457e-07,
      "loss": 23.3769,
      "step": 15200
    },
    {
      "epoch": 120.63808574277168,
      "eval_loss": 3.0014636516571045,
      "eval_mae": 1.2961853742599487,
      "eval_mse": 3.0014636516571045,
      "eval_r2": 0.09567362070083618,
      "eval_rmse": 1.7324732758854045,
      "eval_runtime": 9.0886,
      "eval_samples_per_second": 441.324,
      "eval_steps_per_second": 13.864,
      "step": 15200
    },
    {
      "epoch": 120.71784646061815,
      "grad_norm": 329.38873291015625,
      "learning_rate": 1.710251870065411e-07,
      "loss": 24.515,
      "step": 15210
    },
    {
      "epoch": 120.7976071784646,
      "grad_norm": 557.4713134765625,
      "learning_rate": 1.70728954433564e-07,
      "loss": 23.7734,
      "step": 15220
    },
    {
      "epoch": 120.87736789631107,
      "grad_norm": 415.6596374511719,
      "learning_rate": 1.7043284555164285e-07,
      "loss": 23.814,
      "step": 15230
    },
    {
      "epoch": 120.95712861415753,
      "grad_norm": 133.07821655273438,
      "learning_rate": 1.701368608228127e-07,
      "loss": 23.5502,
      "step": 15240
    },
    {
      "epoch": 121.03190428713859,
      "grad_norm": 182.43716430664062,
      "learning_rate": 1.6984100070891514e-07,
      "loss": 21.9608,
      "step": 15250
    },
    {
      "epoch": 121.11166500498504,
      "grad_norm": 180.09335327148438,
      "learning_rate": 1.695452656715974e-07,
      "loss": 24.405,
      "step": 15260
    },
    {
      "epoch": 121.19142572283151,
      "grad_norm": 366.22259521484375,
      "learning_rate": 1.6924965617231142e-07,
      "loss": 22.4355,
      "step": 15270
    },
    {
      "epoch": 121.27118644067797,
      "grad_norm": 109.32711791992188,
      "learning_rate": 1.6895417267231327e-07,
      "loss": 25.3911,
      "step": 15280
    },
    {
      "epoch": 121.35094715852442,
      "grad_norm": 182.69952392578125,
      "learning_rate": 1.6865881563266235e-07,
      "loss": 23.423,
      "step": 15290
    },
    {
      "epoch": 121.43070787637089,
      "grad_norm": 882.4513549804688,
      "learning_rate": 1.6836358551422085e-07,
      "loss": 23.9485,
      "step": 15300
    },
    {
      "epoch": 121.43070787637089,
      "eval_loss": 2.995577335357666,
      "eval_mae": 1.3027689456939697,
      "eval_mse": 2.995577335357666,
      "eval_r2": 0.0974472165107727,
      "eval_rmse": 1.7307736233712558,
      "eval_runtime": 9.0979,
      "eval_samples_per_second": 440.873,
      "eval_steps_per_second": 13.849,
      "step": 15300
    },
    {
      "epoch": 121.51046859421734,
      "grad_norm": 252.78250122070312,
      "learning_rate": 1.6806848277765284e-07,
      "loss": 22.1597,
      "step": 15310
    },
    {
      "epoch": 121.59022931206381,
      "grad_norm": 1262.2432861328125,
      "learning_rate": 1.6777350788342377e-07,
      "loss": 23.4783,
      "step": 15320
    },
    {
      "epoch": 121.66999002991027,
      "grad_norm": 442.42547607421875,
      "learning_rate": 1.6747866129179938e-07,
      "loss": 26.3166,
      "step": 15330
    },
    {
      "epoch": 121.74975074775674,
      "grad_norm": 246.87246704101562,
      "learning_rate": 1.6718394346284538e-07,
      "loss": 25.5064,
      "step": 15340
    },
    {
      "epoch": 121.82951146560319,
      "grad_norm": 1119.2406005859375,
      "learning_rate": 1.6688935485642646e-07,
      "loss": 25.2066,
      "step": 15350
    },
    {
      "epoch": 121.90927218344964,
      "grad_norm": 326.2505798339844,
      "learning_rate": 1.665948959322059e-07,
      "loss": 23.6661,
      "step": 15360
    },
    {
      "epoch": 121.98903290129611,
      "grad_norm": 148.08853149414062,
      "learning_rate": 1.663005671496443e-07,
      "loss": 23.4266,
      "step": 15370
    },
    {
      "epoch": 122.06380857427718,
      "grad_norm": 1457.82958984375,
      "learning_rate": 1.6600636896799936e-07,
      "loss": 22.0045,
      "step": 15380
    },
    {
      "epoch": 122.14356929212363,
      "grad_norm": 116.29139709472656,
      "learning_rate": 1.6571230184632507e-07,
      "loss": 24.5288,
      "step": 15390
    },
    {
      "epoch": 122.22333000997008,
      "grad_norm": 959.0142822265625,
      "learning_rate": 1.6541836624347095e-07,
      "loss": 24.791,
      "step": 15400
    },
    {
      "epoch": 122.22333000997008,
      "eval_loss": 2.9969351291656494,
      "eval_mae": 1.2997649908065796,
      "eval_mse": 2.9969353675842285,
      "eval_r2": 0.09703803062438965,
      "eval_rmse": 1.7311658983425675,
      "eval_runtime": 9.0986,
      "eval_samples_per_second": 440.835,
      "eval_steps_per_second": 13.848,
      "step": 15400
    },
    {
      "epoch": 122.30309072781655,
      "grad_norm": 147.13011169433594,
      "learning_rate": 1.6512456261808089e-07,
      "loss": 23.8236,
      "step": 15410
    },
    {
      "epoch": 122.38285144566301,
      "grad_norm": 85.25444793701172,
      "learning_rate": 1.6483089142859335e-07,
      "loss": 25.0383,
      "step": 15420
    },
    {
      "epoch": 122.46261216350948,
      "grad_norm": 730.5651245117188,
      "learning_rate": 1.6453735313324e-07,
      "loss": 23.6549,
      "step": 15430
    },
    {
      "epoch": 122.54237288135593,
      "grad_norm": 222.33164978027344,
      "learning_rate": 1.6424394819004504e-07,
      "loss": 23.545,
      "step": 15440
    },
    {
      "epoch": 122.6221335992024,
      "grad_norm": 675.6375732421875,
      "learning_rate": 1.6395067705682458e-07,
      "loss": 23.9993,
      "step": 15450
    },
    {
      "epoch": 122.70189431704885,
      "grad_norm": 397.6210632324219,
      "learning_rate": 1.636575401911861e-07,
      "loss": 24.0422,
      "step": 15460
    },
    {
      "epoch": 122.78165503489531,
      "grad_norm": 399.5688171386719,
      "learning_rate": 1.6336453805052742e-07,
      "loss": 22.0093,
      "step": 15470
    },
    {
      "epoch": 122.86141575274178,
      "grad_norm": 166.28970336914062,
      "learning_rate": 1.630716710920363e-07,
      "loss": 24.2535,
      "step": 15480
    },
    {
      "epoch": 122.94117647058823,
      "grad_norm": 581.7630004882812,
      "learning_rate": 1.6277893977268934e-07,
      "loss": 25.0971,
      "step": 15490
    },
    {
      "epoch": 123.0159521435693,
      "grad_norm": 343.17578125,
      "learning_rate": 1.6248634454925168e-07,
      "loss": 22.5816,
      "step": 15500
    },
    {
      "epoch": 123.0159521435693,
      "eval_loss": 2.9995081424713135,
      "eval_mae": 1.3060561418533325,
      "eval_mse": 2.9995081424713135,
      "eval_r2": 0.0962628722190857,
      "eval_rmse": 1.7319088147103223,
      "eval_runtime": 9.1165,
      "eval_samples_per_second": 439.969,
      "eval_steps_per_second": 13.821,
      "step": 15500
    },
    {
      "epoch": 123.09571286141575,
      "grad_norm": 189.19554138183594,
      "learning_rate": 1.6219388587827603e-07,
      "loss": 22.5818,
      "step": 15510
    },
    {
      "epoch": 123.17547357926222,
      "grad_norm": 288.20843505859375,
      "learning_rate": 1.619015642161022e-07,
      "loss": 24.2205,
      "step": 15520
    },
    {
      "epoch": 123.25523429710867,
      "grad_norm": 649.839111328125,
      "learning_rate": 1.6160938001885583e-07,
      "loss": 22.5111,
      "step": 15530
    },
    {
      "epoch": 123.33499501495514,
      "grad_norm": 294.1289978027344,
      "learning_rate": 1.6131733374244844e-07,
      "loss": 26.4326,
      "step": 15540
    },
    {
      "epoch": 123.4147557328016,
      "grad_norm": 309.74951171875,
      "learning_rate": 1.6102542584257608e-07,
      "loss": 25.1714,
      "step": 15550
    },
    {
      "epoch": 123.49451645064805,
      "grad_norm": 566.5947265625,
      "learning_rate": 1.607336567747193e-07,
      "loss": 25.061,
      "step": 15560
    },
    {
      "epoch": 123.57427716849452,
      "grad_norm": 135.1038360595703,
      "learning_rate": 1.6044202699414138e-07,
      "loss": 22.7417,
      "step": 15570
    },
    {
      "epoch": 123.65403788634097,
      "grad_norm": 704.9916381835938,
      "learning_rate": 1.6015053695588875e-07,
      "loss": 24.8618,
      "step": 15580
    },
    {
      "epoch": 123.73379860418744,
      "grad_norm": 155.9027557373047,
      "learning_rate": 1.5985918711478974e-07,
      "loss": 23.0754,
      "step": 15590
    },
    {
      "epoch": 123.8135593220339,
      "grad_norm": 139.53753662109375,
      "learning_rate": 1.5956797792545384e-07,
      "loss": 25.0568,
      "step": 15600
    },
    {
      "epoch": 123.8135593220339,
      "eval_loss": 2.999448299407959,
      "eval_mae": 1.2963074445724487,
      "eval_mse": 2.999448299407959,
      "eval_r2": 0.09628087282180786,
      "eval_rmse": 1.731891538003451,
      "eval_runtime": 9.1006,
      "eval_samples_per_second": 440.741,
      "eval_steps_per_second": 13.845,
      "step": 15600
    },
    {
      "epoch": 123.89332003988036,
      "grad_norm": 343.701416015625,
      "learning_rate": 1.5927690984227095e-07,
      "loss": 23.1298,
      "step": 15610
    },
    {
      "epoch": 123.97308075772682,
      "grad_norm": 455.2835388183594,
      "learning_rate": 1.589859833194109e-07,
      "loss": 22.7114,
      "step": 15620
    },
    {
      "epoch": 124.04785643070788,
      "grad_norm": 767.1268920898438,
      "learning_rate": 1.5869519881082277e-07,
      "loss": 22.7328,
      "step": 15630
    },
    {
      "epoch": 124.12761714855434,
      "grad_norm": 378.0350036621094,
      "learning_rate": 1.5840455677023392e-07,
      "loss": 24.4979,
      "step": 15640
    },
    {
      "epoch": 124.2073778664008,
      "grad_norm": 588.4114990234375,
      "learning_rate": 1.5811405765114933e-07,
      "loss": 22.3417,
      "step": 15650
    },
    {
      "epoch": 124.28713858424726,
      "grad_norm": 118.65570831298828,
      "learning_rate": 1.5782370190685113e-07,
      "loss": 23.0127,
      "step": 15660
    },
    {
      "epoch": 124.36689930209371,
      "grad_norm": 1166.2757568359375,
      "learning_rate": 1.575334899903976e-07,
      "loss": 23.5448,
      "step": 15670
    },
    {
      "epoch": 124.44666001994018,
      "grad_norm": 186.2198028564453,
      "learning_rate": 1.5724342235462273e-07,
      "loss": 22.2049,
      "step": 15680
    },
    {
      "epoch": 124.52642073778664,
      "grad_norm": 953.7238159179688,
      "learning_rate": 1.5695349945213525e-07,
      "loss": 24.5845,
      "step": 15690
    },
    {
      "epoch": 124.6061814556331,
      "grad_norm": 590.2252807617188,
      "learning_rate": 1.5666372173531817e-07,
      "loss": 25.8507,
      "step": 15700
    },
    {
      "epoch": 124.6061814556331,
      "eval_loss": 2.9973418712615967,
      "eval_mae": 1.2974629402160645,
      "eval_mse": 2.9973418712615967,
      "eval_r2": 0.096915602684021,
      "eval_rmse": 1.7312833018491216,
      "eval_runtime": 9.1042,
      "eval_samples_per_second": 440.567,
      "eval_steps_per_second": 13.84,
      "step": 15700
    },
    {
      "epoch": 124.68594217347956,
      "grad_norm": 581.5036010742188,
      "learning_rate": 1.5637408965632782e-07,
      "loss": 24.9864,
      "step": 15710
    },
    {
      "epoch": 124.76570289132601,
      "grad_norm": 982.7798461914062,
      "learning_rate": 1.5608460366709358e-07,
      "loss": 23.1186,
      "step": 15720
    },
    {
      "epoch": 124.84546360917248,
      "grad_norm": 523.9783325195312,
      "learning_rate": 1.557952642193164e-07,
      "loss": 25.3386,
      "step": 15730
    },
    {
      "epoch": 124.92522432701894,
      "grad_norm": 323.17022705078125,
      "learning_rate": 1.555060717644689e-07,
      "loss": 24.6167,
      "step": 15740
    },
    {
      "epoch": 125.0,
      "grad_norm": 173.55587768554688,
      "learning_rate": 1.552170267537944e-07,
      "loss": 22.7146,
      "step": 15750
    },
    {
      "epoch": 125.07976071784645,
      "grad_norm": 261.3866882324219,
      "learning_rate": 1.5492812963830604e-07,
      "loss": 24.0152,
      "step": 15760
    },
    {
      "epoch": 125.15952143569292,
      "grad_norm": 348.5075988769531,
      "learning_rate": 1.5463938086878599e-07,
      "loss": 22.3543,
      "step": 15770
    },
    {
      "epoch": 125.23928215353938,
      "grad_norm": 430.7974853515625,
      "learning_rate": 1.543507808957853e-07,
      "loss": 24.7551,
      "step": 15780
    },
    {
      "epoch": 125.31904287138585,
      "grad_norm": 265.10382080078125,
      "learning_rate": 1.540623301696226e-07,
      "loss": 23.8209,
      "step": 15790
    },
    {
      "epoch": 125.3988035892323,
      "grad_norm": 460.537841796875,
      "learning_rate": 1.5377402914038384e-07,
      "loss": 22.5712,
      "step": 15800
    },
    {
      "epoch": 125.3988035892323,
      "eval_loss": 2.99674391746521,
      "eval_mae": 1.2961981296539307,
      "eval_mse": 2.99674391746521,
      "eval_r2": 0.09709572792053223,
      "eval_rmse": 1.7311106023201435,
      "eval_runtime": 9.0932,
      "eval_samples_per_second": 441.097,
      "eval_steps_per_second": 13.856,
      "step": 15800
    },
    {
      "epoch": 125.47856430707877,
      "grad_norm": 145.07672119140625,
      "learning_rate": 1.5348587825792114e-07,
      "loss": 22.8911,
      "step": 15810
    },
    {
      "epoch": 125.55832502492522,
      "grad_norm": 355.4983825683594,
      "learning_rate": 1.5319787797185257e-07,
      "loss": 23.2123,
      "step": 15820
    },
    {
      "epoch": 125.63808574277168,
      "grad_norm": 122.14202117919922,
      "learning_rate": 1.5291002873156105e-07,
      "loss": 23.9367,
      "step": 15830
    },
    {
      "epoch": 125.71784646061815,
      "grad_norm": 623.5941162109375,
      "learning_rate": 1.5262233098619398e-07,
      "loss": 22.5786,
      "step": 15840
    },
    {
      "epoch": 125.7976071784646,
      "grad_norm": 1107.0355224609375,
      "learning_rate": 1.523347851846622e-07,
      "loss": 27.1192,
      "step": 15850
    },
    {
      "epoch": 125.87736789631107,
      "grad_norm": 371.8562927246094,
      "learning_rate": 1.5204739177563953e-07,
      "loss": 25.3864,
      "step": 15860
    },
    {
      "epoch": 125.95712861415753,
      "grad_norm": 474.218505859375,
      "learning_rate": 1.5176015120756203e-07,
      "loss": 24.8445,
      "step": 15870
    },
    {
      "epoch": 126.03190428713859,
      "grad_norm": 1185.120849609375,
      "learning_rate": 1.514730639286273e-07,
      "loss": 23.3745,
      "step": 15880
    },
    {
      "epoch": 126.11166500498504,
      "grad_norm": 113.61933135986328,
      "learning_rate": 1.511861303867938e-07,
      "loss": 25.0345,
      "step": 15890
    },
    {
      "epoch": 126.19142572283151,
      "grad_norm": 493.3686828613281,
      "learning_rate": 1.5089935102977973e-07,
      "loss": 23.1973,
      "step": 15900
    },
    {
      "epoch": 126.19142572283151,
      "eval_loss": 2.9987268447875977,
      "eval_mae": 1.2949258089065552,
      "eval_mse": 2.9987266063690186,
      "eval_r2": 0.09649831056594849,
      "eval_rmse": 1.731683171474799,
      "eval_runtime": 9.0753,
      "eval_samples_per_second": 441.967,
      "eval_steps_per_second": 13.884,
      "step": 15900
    },
    {
      "epoch": 126.27118644067797,
      "grad_norm": 353.9967041015625,
      "learning_rate": 1.5061272630506324e-07,
      "loss": 24.2967,
      "step": 15910
    },
    {
      "epoch": 126.35094715852442,
      "grad_norm": 487.5097961425781,
      "learning_rate": 1.503262566598809e-07,
      "loss": 23.7637,
      "step": 15920
    },
    {
      "epoch": 126.43070787637089,
      "grad_norm": 207.82666015625,
      "learning_rate": 1.500399425412274e-07,
      "loss": 24.2266,
      "step": 15930
    },
    {
      "epoch": 126.51046859421734,
      "grad_norm": 996.2344360351562,
      "learning_rate": 1.497537843958546e-07,
      "loss": 23.555,
      "step": 15940
    },
    {
      "epoch": 126.59022931206381,
      "grad_norm": 193.35317993164062,
      "learning_rate": 1.4946778267027112e-07,
      "loss": 24.4272,
      "step": 15950
    },
    {
      "epoch": 126.66999002991027,
      "grad_norm": 614.0579833984375,
      "learning_rate": 1.4918193781074153e-07,
      "loss": 23.7424,
      "step": 15960
    },
    {
      "epoch": 126.74975074775674,
      "grad_norm": 408.7231750488281,
      "learning_rate": 1.4889625026328563e-07,
      "loss": 24.5934,
      "step": 15970
    },
    {
      "epoch": 126.82951146560319,
      "grad_norm": 658.59814453125,
      "learning_rate": 1.4861072047367758e-07,
      "loss": 22.9295,
      "step": 15980
    },
    {
      "epoch": 126.90927218344964,
      "grad_norm": 465.7789611816406,
      "learning_rate": 1.4832534888744557e-07,
      "loss": 22.9432,
      "step": 15990
    },
    {
      "epoch": 126.98903290129611,
      "grad_norm": 424.2307434082031,
      "learning_rate": 1.4804013594987093e-07,
      "loss": 23.4441,
      "step": 16000
    },
    {
      "epoch": 126.98903290129611,
      "eval_loss": 2.9949264526367188,
      "eval_mae": 1.297987461090088,
      "eval_mse": 2.9949264526367188,
      "eval_r2": 0.09764325618743896,
      "eval_rmse": 1.730585580847338,
      "eval_runtime": 9.0821,
      "eval_samples_per_second": 441.64,
      "eval_steps_per_second": 13.873,
      "step": 16000
    },
    {
      "epoch": 127.06380857427718,
      "grad_norm": 611.9707641601562,
      "learning_rate": 1.4775508210598743e-07,
      "loss": 22.922,
      "step": 16010
    },
    {
      "epoch": 127.14356929212363,
      "grad_norm": 321.8793640136719,
      "learning_rate": 1.4747018780058042e-07,
      "loss": 24.6874,
      "step": 16020
    },
    {
      "epoch": 127.22333000997008,
      "grad_norm": 452.5003662109375,
      "learning_rate": 1.471854534781865e-07,
      "loss": 24.2103,
      "step": 16030
    },
    {
      "epoch": 127.30309072781655,
      "grad_norm": 291.26043701171875,
      "learning_rate": 1.469008795830926e-07,
      "loss": 24.3956,
      "step": 16040
    },
    {
      "epoch": 127.38285144566301,
      "grad_norm": 988.60693359375,
      "learning_rate": 1.4661646655933552e-07,
      "loss": 24.6791,
      "step": 16050
    },
    {
      "epoch": 127.46261216350948,
      "grad_norm": 131.3463134765625,
      "learning_rate": 1.4633221485070052e-07,
      "loss": 22.7122,
      "step": 16060
    },
    {
      "epoch": 127.54237288135593,
      "grad_norm": 216.421875,
      "learning_rate": 1.4604812490072159e-07,
      "loss": 24.6799,
      "step": 16070
    },
    {
      "epoch": 127.6221335992024,
      "grad_norm": 265.0110168457031,
      "learning_rate": 1.4576419715268026e-07,
      "loss": 23.6442,
      "step": 16080
    },
    {
      "epoch": 127.70189431704885,
      "grad_norm": 202.32069396972656,
      "learning_rate": 1.4548043204960496e-07,
      "loss": 25.1982,
      "step": 16090
    },
    {
      "epoch": 127.78165503489531,
      "grad_norm": 834.6726684570312,
      "learning_rate": 1.4519683003427018e-07,
      "loss": 23.8063,
      "step": 16100
    },
    {
      "epoch": 127.78165503489531,
      "eval_loss": 2.9989752769470215,
      "eval_mae": 1.2963449954986572,
      "eval_mse": 2.9989752769470215,
      "eval_r2": 0.0964234471321106,
      "eval_rmse": 1.7317549702388677,
      "eval_runtime": 9.1158,
      "eval_samples_per_second": 440.007,
      "eval_steps_per_second": 13.822,
      "step": 16100
    },
    {
      "epoch": 127.86141575274178,
      "grad_norm": 438.2342529296875,
      "learning_rate": 1.4491339154919595e-07,
      "loss": 24.8793,
      "step": 16110
    },
    {
      "epoch": 127.94117647058823,
      "grad_norm": 500.1650695800781,
      "learning_rate": 1.446301170366474e-07,
      "loss": 22.3995,
      "step": 16120
    },
    {
      "epoch": 128.01595214356928,
      "grad_norm": 660.446533203125,
      "learning_rate": 1.4434700693863377e-07,
      "loss": 22.4161,
      "step": 16130
    },
    {
      "epoch": 128.09571286141576,
      "grad_norm": 554.826416015625,
      "learning_rate": 1.440640616969072e-07,
      "loss": 24.9374,
      "step": 16140
    },
    {
      "epoch": 128.17547357926222,
      "grad_norm": 274.38031005859375,
      "learning_rate": 1.437812817529632e-07,
      "loss": 24.6947,
      "step": 16150
    },
    {
      "epoch": 128.25523429710867,
      "grad_norm": 464.1123046875,
      "learning_rate": 1.4349866754803936e-07,
      "loss": 24.7641,
      "step": 16160
    },
    {
      "epoch": 128.33499501495513,
      "grad_norm": 204.35543823242188,
      "learning_rate": 1.432162195231143e-07,
      "loss": 24.2712,
      "step": 16170
    },
    {
      "epoch": 128.4147557328016,
      "grad_norm": 322.74456787109375,
      "learning_rate": 1.4293393811890737e-07,
      "loss": 24.4455,
      "step": 16180
    },
    {
      "epoch": 128.49451645064806,
      "grad_norm": 171.8980255126953,
      "learning_rate": 1.4265182377587833e-07,
      "loss": 23.6119,
      "step": 16190
    },
    {
      "epoch": 128.57427716849452,
      "grad_norm": 227.2103271484375,
      "learning_rate": 1.4236987693422577e-07,
      "loss": 22.7084,
      "step": 16200
    },
    {
      "epoch": 128.57427716849452,
      "eval_loss": 2.9964678287506104,
      "eval_mae": 1.3037374019622803,
      "eval_mse": 2.9964678287506104,
      "eval_r2": 0.09717893600463867,
      "eval_rmse": 1.731030857249694,
      "eval_runtime": 9.0864,
      "eval_samples_per_second": 441.428,
      "eval_steps_per_second": 13.867,
      "step": 16200
    },
    {
      "epoch": 128.65403788634097,
      "grad_norm": 286.4161376953125,
      "learning_rate": 1.4208809803388739e-07,
      "loss": 23.3087,
      "step": 16210
    },
    {
      "epoch": 128.73379860418743,
      "grad_norm": 116.77186584472656,
      "learning_rate": 1.4180648751453843e-07,
      "loss": 23.4167,
      "step": 16220
    },
    {
      "epoch": 128.8135593220339,
      "grad_norm": 98.98394775390625,
      "learning_rate": 1.4152504581559156e-07,
      "loss": 24.9745,
      "step": 16230
    },
    {
      "epoch": 128.89332003988036,
      "grad_norm": 324.8134460449219,
      "learning_rate": 1.4124377337619617e-07,
      "loss": 23.0645,
      "step": 16240
    },
    {
      "epoch": 128.97308075772682,
      "grad_norm": 120.70247650146484,
      "learning_rate": 1.4096267063523735e-07,
      "loss": 23.2501,
      "step": 16250
    },
    {
      "epoch": 129.04785643070787,
      "grad_norm": 342.9710998535156,
      "learning_rate": 1.4068173803133531e-07,
      "loss": 21.5063,
      "step": 16260
    },
    {
      "epoch": 129.12761714855435,
      "grad_norm": 173.40664672851562,
      "learning_rate": 1.4040097600284522e-07,
      "loss": 23.2956,
      "step": 16270
    },
    {
      "epoch": 129.2073778664008,
      "grad_norm": 1194.4302978515625,
      "learning_rate": 1.4012038498785557e-07,
      "loss": 24.5195,
      "step": 16280
    },
    {
      "epoch": 129.28713858424726,
      "grad_norm": 308.144775390625,
      "learning_rate": 1.398399654241884e-07,
      "loss": 22.852,
      "step": 16290
    },
    {
      "epoch": 129.3668993020937,
      "grad_norm": 166.0808868408203,
      "learning_rate": 1.3955971774939807e-07,
      "loss": 24.0121,
      "step": 16300
    },
    {
      "epoch": 129.3668993020937,
      "eval_loss": 2.9947915077209473,
      "eval_mae": 1.2989612817764282,
      "eval_mse": 2.9947915077209473,
      "eval_r2": 0.09768396615982056,
      "eval_rmse": 1.7305465921843732,
      "eval_runtime": 9.0693,
      "eval_samples_per_second": 442.262,
      "eval_steps_per_second": 13.893,
      "step": 16300
    },
    {
      "epoch": 129.44666001994017,
      "grad_norm": 304.95184326171875,
      "learning_rate": 1.3927964240077055e-07,
      "loss": 26.7386,
      "step": 16310
    },
    {
      "epoch": 129.52642073778665,
      "grad_norm": 172.49710083007812,
      "learning_rate": 1.3899973981532326e-07,
      "loss": 22.5228,
      "step": 16320
    },
    {
      "epoch": 129.6061814556331,
      "grad_norm": 427.22314453125,
      "learning_rate": 1.3872001042980408e-07,
      "loss": 22.4321,
      "step": 16330
    },
    {
      "epoch": 129.68594217347956,
      "grad_norm": 335.5365295410156,
      "learning_rate": 1.3844045468069004e-07,
      "loss": 23.5267,
      "step": 16340
    },
    {
      "epoch": 129.76570289132601,
      "grad_norm": 206.60768127441406,
      "learning_rate": 1.3816107300418783e-07,
      "loss": 25.8498,
      "step": 16350
    },
    {
      "epoch": 129.84546360917247,
      "grad_norm": 352.3175354003906,
      "learning_rate": 1.378818658362325e-07,
      "loss": 26.3323,
      "step": 16360
    },
    {
      "epoch": 129.92522432701895,
      "grad_norm": 115.40145111083984,
      "learning_rate": 1.3760283361248647e-07,
      "loss": 24.2572,
      "step": 16370
    },
    {
      "epoch": 130.0,
      "grad_norm": 190.1355438232422,
      "learning_rate": 1.3732397676833934e-07,
      "loss": 20.9339,
      "step": 16380
    },
    {
      "epoch": 130.07976071784645,
      "grad_norm": 381.51068115234375,
      "learning_rate": 1.370452957389072e-07,
      "loss": 24.8343,
      "step": 16390
    },
    {
      "epoch": 130.1595214356929,
      "grad_norm": 565.8461303710938,
      "learning_rate": 1.367667909590315e-07,
      "loss": 27.1709,
      "step": 16400
    },
    {
      "epoch": 130.1595214356929,
      "eval_loss": 2.9923336505889893,
      "eval_mae": 1.3002452850341797,
      "eval_mse": 2.9923336505889893,
      "eval_r2": 0.09842449426651001,
      "eval_rmse": 1.7298363074548382,
      "eval_runtime": 9.0968,
      "eval_samples_per_second": 440.924,
      "eval_steps_per_second": 13.851,
      "step": 16400
    },
    {
      "epoch": 130.2392821535394,
      "grad_norm": 690.071533203125,
      "learning_rate": 1.3648846286327912e-07,
      "loss": 25.4686,
      "step": 16410
    },
    {
      "epoch": 130.31904287138585,
      "grad_norm": 202.82559204101562,
      "learning_rate": 1.362103118859408e-07,
      "loss": 22.9148,
      "step": 16420
    },
    {
      "epoch": 130.3988035892323,
      "grad_norm": 350.946533203125,
      "learning_rate": 1.3593233846103102e-07,
      "loss": 22.784,
      "step": 16430
    },
    {
      "epoch": 130.47856430707876,
      "grad_norm": 132.73452758789062,
      "learning_rate": 1.3565454302228757e-07,
      "loss": 26.8174,
      "step": 16440
    },
    {
      "epoch": 130.5583250249252,
      "grad_norm": 538.4484252929688,
      "learning_rate": 1.3537692600316997e-07,
      "loss": 22.1968,
      "step": 16450
    },
    {
      "epoch": 130.6380857427717,
      "grad_norm": 918.4005737304688,
      "learning_rate": 1.3509948783685985e-07,
      "loss": 22.7618,
      "step": 16460
    },
    {
      "epoch": 130.71784646061815,
      "grad_norm": 326.4291687011719,
      "learning_rate": 1.348222289562593e-07,
      "loss": 22.8103,
      "step": 16470
    },
    {
      "epoch": 130.7976071784646,
      "grad_norm": 255.4975128173828,
      "learning_rate": 1.345451497939911e-07,
      "loss": 25.1671,
      "step": 16480
    },
    {
      "epoch": 130.87736789631106,
      "grad_norm": 483.3714294433594,
      "learning_rate": 1.3426825078239727e-07,
      "loss": 22.1233,
      "step": 16490
    },
    {
      "epoch": 130.95712861415754,
      "grad_norm": 643.0153198242188,
      "learning_rate": 1.3399153235353895e-07,
      "loss": 23.1952,
      "step": 16500
    },
    {
      "epoch": 130.95712861415754,
      "eval_loss": 3.0047690868377686,
      "eval_mae": 1.2939373254776,
      "eval_mse": 3.0047690868377686,
      "eval_r2": 0.09467774629592896,
      "eval_rmse": 1.7334269776479678,
      "eval_runtime": 9.0875,
      "eval_samples_per_second": 441.374,
      "eval_steps_per_second": 13.865,
      "step": 16500
    },
    {
      "epoch": 131.0319042871386,
      "grad_norm": 323.6212158203125,
      "learning_rate": 1.3371499493919535e-07,
      "loss": 22.6509,
      "step": 16510
    },
    {
      "epoch": 131.11166500498504,
      "grad_norm": 479.4410705566406,
      "learning_rate": 1.3343863897086328e-07,
      "loss": 23.9574,
      "step": 16520
    },
    {
      "epoch": 131.1914257228315,
      "grad_norm": 190.98963928222656,
      "learning_rate": 1.3316246487975653e-07,
      "loss": 23.5103,
      "step": 16530
    },
    {
      "epoch": 131.27118644067798,
      "grad_norm": 318.1726989746094,
      "learning_rate": 1.3288647309680497e-07,
      "loss": 23.319,
      "step": 16540
    },
    {
      "epoch": 131.35094715852443,
      "grad_norm": 867.248046875,
      "learning_rate": 1.3261066405265397e-07,
      "loss": 24.4538,
      "step": 16550
    },
    {
      "epoch": 131.4307078763709,
      "grad_norm": 776.1854858398438,
      "learning_rate": 1.32335038177664e-07,
      "loss": 22.9098,
      "step": 16560
    },
    {
      "epoch": 131.51046859421734,
      "grad_norm": 262.22979736328125,
      "learning_rate": 1.3205959590190937e-07,
      "loss": 22.832,
      "step": 16570
    },
    {
      "epoch": 131.5902293120638,
      "grad_norm": 1199.579833984375,
      "learning_rate": 1.3178433765517833e-07,
      "loss": 25.2951,
      "step": 16580
    },
    {
      "epoch": 131.66999002991028,
      "grad_norm": 213.65872192382812,
      "learning_rate": 1.3150926386697164e-07,
      "loss": 22.2523,
      "step": 16590
    },
    {
      "epoch": 131.74975074775674,
      "grad_norm": 587.2889404296875,
      "learning_rate": 1.3123437496650224e-07,
      "loss": 24.5088,
      "step": 16600
    },
    {
      "epoch": 131.74975074775674,
      "eval_loss": 2.9919657707214355,
      "eval_mae": 1.3000578880310059,
      "eval_mse": 2.9919657707214355,
      "eval_r2": 0.09853535890579224,
      "eval_rmse": 1.729729970464013,
      "eval_runtime": 9.0844,
      "eval_samples_per_second": 441.528,
      "eval_steps_per_second": 13.87,
      "step": 16600
    },
    {
      "epoch": 131.8295114656032,
      "grad_norm": 673.372314453125,
      "learning_rate": 1.3095967138269478e-07,
      "loss": 24.6808,
      "step": 16610
    },
    {
      "epoch": 131.90927218344964,
      "grad_norm": 602.1070556640625,
      "learning_rate": 1.306851535441849e-07,
      "loss": 26.5808,
      "step": 16620
    },
    {
      "epoch": 131.9890329012961,
      "grad_norm": 168.7585906982422,
      "learning_rate": 1.3041082187931774e-07,
      "loss": 22.9096,
      "step": 16630
    },
    {
      "epoch": 132.06380857427718,
      "grad_norm": 326.1978759765625,
      "learning_rate": 1.3013667681614855e-07,
      "loss": 22.6725,
      "step": 16640
    },
    {
      "epoch": 132.14356929212363,
      "grad_norm": 503.9088134765625,
      "learning_rate": 1.298627187824414e-07,
      "loss": 23.5713,
      "step": 16650
    },
    {
      "epoch": 132.22333000997008,
      "grad_norm": 710.6509399414062,
      "learning_rate": 1.2958894820566824e-07,
      "loss": 22.3871,
      "step": 16660
    },
    {
      "epoch": 132.30309072781654,
      "grad_norm": 299.3223571777344,
      "learning_rate": 1.293153655130085e-07,
      "loss": 27.3263,
      "step": 16670
    },
    {
      "epoch": 132.38285144566302,
      "grad_norm": 118.80657958984375,
      "learning_rate": 1.2904197113134884e-07,
      "loss": 24.3728,
      "step": 16680
    },
    {
      "epoch": 132.46261216350948,
      "grad_norm": 922.7802124023438,
      "learning_rate": 1.2876876548728167e-07,
      "loss": 23.758,
      "step": 16690
    },
    {
      "epoch": 132.54237288135593,
      "grad_norm": 142.9518585205078,
      "learning_rate": 1.2849574900710518e-07,
      "loss": 23.1453,
      "step": 16700
    },
    {
      "epoch": 132.54237288135593,
      "eval_loss": 2.992398738861084,
      "eval_mae": 1.295898199081421,
      "eval_mse": 2.992398738861084,
      "eval_r2": 0.0984048843383789,
      "eval_rmse": 1.7298551207719923,
      "eval_runtime": 9.1062,
      "eval_samples_per_second": 440.471,
      "eval_steps_per_second": 13.837,
      "step": 16700
    },
    {
      "epoch": 132.62213359920239,
      "grad_norm": 365.12811279296875,
      "learning_rate": 1.2822292211682229e-07,
      "loss": 21.4491,
      "step": 16710
    },
    {
      "epoch": 132.70189431704884,
      "grad_norm": 153.51211547851562,
      "learning_rate": 1.279502852421399e-07,
      "loss": 23.5518,
      "step": 16720
    },
    {
      "epoch": 132.78165503489532,
      "grad_norm": 169.6097412109375,
      "learning_rate": 1.276778388084689e-07,
      "loss": 23.0115,
      "step": 16730
    },
    {
      "epoch": 132.86141575274178,
      "grad_norm": 417.13189697265625,
      "learning_rate": 1.2740558324092253e-07,
      "loss": 25.8207,
      "step": 16740
    },
    {
      "epoch": 132.94117647058823,
      "grad_norm": 302.108154296875,
      "learning_rate": 1.2713351896431634e-07,
      "loss": 23.1809,
      "step": 16750
    },
    {
      "epoch": 133.01595214356928,
      "grad_norm": 101.93446350097656,
      "learning_rate": 1.2686164640316773e-07,
      "loss": 23.1577,
      "step": 16760
    },
    {
      "epoch": 133.09571286141576,
      "grad_norm": 449.4652099609375,
      "learning_rate": 1.2658996598169446e-07,
      "loss": 23.7753,
      "step": 16770
    },
    {
      "epoch": 133.17547357926222,
      "grad_norm": 599.4085693359375,
      "learning_rate": 1.263184781238149e-07,
      "loss": 23.5495,
      "step": 16780
    },
    {
      "epoch": 133.25523429710867,
      "grad_norm": 615.1130981445312,
      "learning_rate": 1.260471832531467e-07,
      "loss": 27.3507,
      "step": 16790
    },
    {
      "epoch": 133.33499501495513,
      "grad_norm": 333.2358093261719,
      "learning_rate": 1.2577608179300636e-07,
      "loss": 24.9712,
      "step": 16800
    },
    {
      "epoch": 133.33499501495513,
      "eval_loss": 2.9936790466308594,
      "eval_mae": 1.3027596473693848,
      "eval_mse": 2.9936790466308594,
      "eval_r2": 0.09801912307739258,
      "eval_rmse": 1.7302251433356468,
      "eval_runtime": 9.0992,
      "eval_samples_per_second": 440.807,
      "eval_steps_per_second": 13.847,
      "step": 16800
    },
    {
      "epoch": 133.4147557328016,
      "grad_norm": 647.0711059570312,
      "learning_rate": 1.2550517416640874e-07,
      "loss": 22.4761,
      "step": 16810
    },
    {
      "epoch": 133.49451645064806,
      "grad_norm": 369.3785400390625,
      "learning_rate": 1.2523446079606641e-07,
      "loss": 22.5223,
      "step": 16820
    },
    {
      "epoch": 133.57427716849452,
      "grad_norm": 680.6244506835938,
      "learning_rate": 1.2496394210438827e-07,
      "loss": 22.5881,
      "step": 16830
    },
    {
      "epoch": 133.65403788634097,
      "grad_norm": 258.2887878417969,
      "learning_rate": 1.2469361851347996e-07,
      "loss": 23.0046,
      "step": 16840
    },
    {
      "epoch": 133.73379860418743,
      "grad_norm": Infinity,
      "learning_rate": 1.244234904451426e-07,
      "loss": 24.1551,
      "step": 16850
    },
    {
      "epoch": 133.8135593220339,
      "grad_norm": 177.69448852539062,
      "learning_rate": 1.241805427038098e-07,
      "loss": 24.7674,
      "step": 16860
    },
    {
      "epoch": 133.89332003988036,
      "grad_norm": 532.2387084960938,
      "learning_rate": 1.2391078728932575e-07,
      "loss": 25.262,
      "step": 16870
    },
    {
      "epoch": 133.97308075772682,
      "grad_norm": 595.5781860351562,
      "learning_rate": 1.236412286189081e-07,
      "loss": 23.8183,
      "step": 16880
    },
    {
      "epoch": 134.04785643070787,
      "grad_norm": 557.8573608398438,
      "learning_rate": 1.2337186711316429e-07,
      "loss": 24.132,
      "step": 16890
    },
    {
      "epoch": 134.12761714855435,
      "grad_norm": 487.1327209472656,
      "learning_rate": 1.2310270319239403e-07,
      "loss": 23.2251,
      "step": 16900
    },
    {
      "epoch": 134.12761714855435,
      "eval_loss": 2.9906692504882812,
      "eval_mae": 1.2998462915420532,
      "eval_mse": 2.9906692504882812,
      "eval_r2": 0.09892600774765015,
      "eval_rmse": 1.729355154526762,
      "eval_runtime": 9.0997,
      "eval_samples_per_second": 440.782,
      "eval_steps_per_second": 13.847,
      "step": 16900
    },
    {
      "epoch": 134.2073778664008,
      "grad_norm": 404.20843505859375,
      "learning_rate": 1.228337372765891e-07,
      "loss": 23.15,
      "step": 16910
    },
    {
      "epoch": 134.28713858424726,
      "grad_norm": 351.0567321777344,
      "learning_rate": 1.225649697854319e-07,
      "loss": 24.2283,
      "step": 16920
    },
    {
      "epoch": 134.3668993020937,
      "grad_norm": 334.9957580566406,
      "learning_rate": 1.2229640113829562e-07,
      "loss": 23.772,
      "step": 16930
    },
    {
      "epoch": 134.44666001994017,
      "grad_norm": 325.6705627441406,
      "learning_rate": 1.2202803175424286e-07,
      "loss": 22.0053,
      "step": 16940
    },
    {
      "epoch": 134.52642073778665,
      "grad_norm": 649.3629150390625,
      "learning_rate": 1.2175986205202532e-07,
      "loss": 25.0228,
      "step": 16950
    },
    {
      "epoch": 134.6061814556331,
      "grad_norm": 406.55242919921875,
      "learning_rate": 1.214918924500833e-07,
      "loss": 22.5482,
      "step": 16960
    },
    {
      "epoch": 134.68594217347956,
      "grad_norm": 550.158935546875,
      "learning_rate": 1.2122412336654503e-07,
      "loss": 22.7849,
      "step": 16970
    },
    {
      "epoch": 134.76570289132601,
      "grad_norm": 379.20513916015625,
      "learning_rate": 1.209565552192252e-07,
      "loss": 25.6224,
      "step": 16980
    },
    {
      "epoch": 134.84546360917247,
      "grad_norm": 259.1018981933594,
      "learning_rate": 1.2068918842562553e-07,
      "loss": 24.6256,
      "step": 16990
    },
    {
      "epoch": 134.92522432701895,
      "grad_norm": 107.85888671875,
      "learning_rate": 1.204220234029336e-07,
      "loss": 25.328,
      "step": 17000
    },
    {
      "epoch": 134.92522432701895,
      "eval_loss": 2.992455244064331,
      "eval_mae": 1.2952169179916382,
      "eval_mse": 2.992455244064331,
      "eval_r2": 0.09838783740997314,
      "eval_rmse": 1.7298714530462462,
      "eval_runtime": 9.09,
      "eval_samples_per_second": 441.255,
      "eval_steps_per_second": 13.861,
      "step": 17000
    },
    {
      "epoch": 135.0,
      "grad_norm": 210.70713806152344,
      "learning_rate": 1.2015506056802182e-07,
      "loss": 21.6844,
      "step": 17010
    },
    {
      "epoch": 135.07976071784645,
      "grad_norm": 236.12600708007812,
      "learning_rate": 1.198883003374472e-07,
      "loss": 22.9484,
      "step": 17020
    },
    {
      "epoch": 135.1595214356929,
      "grad_norm": 418.2464599609375,
      "learning_rate": 1.1962174312745088e-07,
      "loss": 26.0478,
      "step": 17030
    },
    {
      "epoch": 135.2392821535394,
      "grad_norm": 370.0785827636719,
      "learning_rate": 1.1935538935395679e-07,
      "loss": 22.035,
      "step": 17040
    },
    {
      "epoch": 135.31904287138585,
      "grad_norm": 756.2055053710938,
      "learning_rate": 1.1908923943257185e-07,
      "loss": 24.213,
      "step": 17050
    },
    {
      "epoch": 135.3988035892323,
      "grad_norm": 1857.18408203125,
      "learning_rate": 1.1882329377858461e-07,
      "loss": 23.3173,
      "step": 17060
    },
    {
      "epoch": 135.47856430707876,
      "grad_norm": 541.4359741210938,
      "learning_rate": 1.1855755280696489e-07,
      "loss": 24.9646,
      "step": 17070
    },
    {
      "epoch": 135.5583250249252,
      "grad_norm": 381.83001708984375,
      "learning_rate": 1.1829201693236335e-07,
      "loss": 23.9757,
      "step": 17080
    },
    {
      "epoch": 135.6380857427717,
      "grad_norm": 181.71661376953125,
      "learning_rate": 1.1802668656911064e-07,
      "loss": 24.3972,
      "step": 17090
    },
    {
      "epoch": 135.71784646061815,
      "grad_norm": 664.7344970703125,
      "learning_rate": 1.1776156213121621e-07,
      "loss": 23.7074,
      "step": 17100
    },
    {
      "epoch": 135.71784646061815,
      "eval_loss": 2.9938080310821533,
      "eval_mae": 1.2958862781524658,
      "eval_mse": 2.9938080310821533,
      "eval_r2": 0.09798026084899902,
      "eval_rmse": 1.730262416826463,
      "eval_runtime": 9.0841,
      "eval_samples_per_second": 441.541,
      "eval_steps_per_second": 13.87,
      "step": 17100
    },
    {
      "epoch": 135.7976071784646,
      "grad_norm": 146.8472442626953,
      "learning_rate": 1.1749664403236884e-07,
      "loss": 23.579,
      "step": 17110
    },
    {
      "epoch": 135.87736789631106,
      "grad_norm": 507.60577392578125,
      "learning_rate": 1.1723193268593512e-07,
      "loss": 24.4998,
      "step": 17120
    },
    {
      "epoch": 135.95712861415754,
      "grad_norm": 139.34336853027344,
      "learning_rate": 1.1696742850495891e-07,
      "loss": 23.7718,
      "step": 17130
    },
    {
      "epoch": 136.0319042871386,
      "grad_norm": 365.88885498046875,
      "learning_rate": 1.1670313190216089e-07,
      "loss": 21.9757,
      "step": 17140
    },
    {
      "epoch": 136.11166500498504,
      "grad_norm": 432.8541259765625,
      "learning_rate": 1.1643904328993795e-07,
      "loss": 24.0741,
      "step": 17150
    },
    {
      "epoch": 136.1914257228315,
      "grad_norm": 517.509765625,
      "learning_rate": 1.1617516308036224e-07,
      "loss": 25.0548,
      "step": 17160
    },
    {
      "epoch": 136.27118644067798,
      "grad_norm": 965.6638793945312,
      "learning_rate": 1.1591149168518097e-07,
      "loss": 21.6349,
      "step": 17170
    },
    {
      "epoch": 136.35094715852443,
      "grad_norm": 460.10772705078125,
      "learning_rate": 1.1564802951581537e-07,
      "loss": 25.4607,
      "step": 17180
    },
    {
      "epoch": 136.4307078763709,
      "grad_norm": 679.1116943359375,
      "learning_rate": 1.153847769833601e-07,
      "loss": 23.2777,
      "step": 17190
    },
    {
      "epoch": 136.51046859421734,
      "grad_norm": 374.9620666503906,
      "learning_rate": 1.1512173449858304e-07,
      "loss": 24.2546,
      "step": 17200
    },
    {
      "epoch": 136.51046859421734,
      "eval_loss": 2.9912867546081543,
      "eval_mae": 1.3019976615905762,
      "eval_mse": 2.9912867546081543,
      "eval_r2": 0.09873998165130615,
      "eval_rmse": 1.7295336812586664,
      "eval_runtime": 9.0801,
      "eval_samples_per_second": 441.735,
      "eval_steps_per_second": 13.877,
      "step": 17200
    },
    {
      "epoch": 136.5902293120638,
      "grad_norm": 767.04736328125,
      "learning_rate": 1.1485890247192401e-07,
      "loss": 22.6815,
      "step": 17210
    },
    {
      "epoch": 136.66999002991028,
      "grad_norm": 258.2825012207031,
      "learning_rate": 1.1459628131349447e-07,
      "loss": 23.3092,
      "step": 17220
    },
    {
      "epoch": 136.74975074775674,
      "grad_norm": 148.56033325195312,
      "learning_rate": 1.1433387143307712e-07,
      "loss": 25.3104,
      "step": 17230
    },
    {
      "epoch": 136.8295114656032,
      "grad_norm": 154.46519470214844,
      "learning_rate": 1.1407167324012465e-07,
      "loss": 25.4671,
      "step": 17240
    },
    {
      "epoch": 136.90927218344964,
      "grad_norm": 138.0871124267578,
      "learning_rate": 1.1380968714375974e-07,
      "loss": 23.5792,
      "step": 17250
    },
    {
      "epoch": 136.9890329012961,
      "grad_norm": 355.7171630859375,
      "learning_rate": 1.1354791355277393e-07,
      "loss": 23.2389,
      "step": 17260
    },
    {
      "epoch": 137.06380857427718,
      "grad_norm": 228.95948791503906,
      "learning_rate": 1.132863528756271e-07,
      "loss": 22.5726,
      "step": 17270
    },
    {
      "epoch": 137.14356929212363,
      "grad_norm": 168.3987274169922,
      "learning_rate": 1.1302500552044714e-07,
      "loss": 22.259,
      "step": 17280
    },
    {
      "epoch": 137.22333000997008,
      "grad_norm": 229.8258819580078,
      "learning_rate": 1.1276387189502917e-07,
      "loss": 22.3476,
      "step": 17290
    },
    {
      "epoch": 137.30309072781654,
      "grad_norm": 993.53466796875,
      "learning_rate": 1.1250295240683422e-07,
      "loss": 24.0292,
      "step": 17300
    },
    {
      "epoch": 137.30309072781654,
      "eval_loss": 2.989678382873535,
      "eval_mae": 1.2974820137023926,
      "eval_mse": 2.989678382873535,
      "eval_r2": 0.09922456741333008,
      "eval_rmse": 1.7290686460848035,
      "eval_runtime": 9.0973,
      "eval_samples_per_second": 440.901,
      "eval_steps_per_second": 13.85,
      "step": 17300
    },
    {
      "epoch": 137.38285144566302,
      "grad_norm": 200.87583923339844,
      "learning_rate": 1.1224224746298983e-07,
      "loss": 25.8856,
      "step": 17310
    },
    {
      "epoch": 137.46261216350948,
      "grad_norm": 846.8623657226562,
      "learning_rate": 1.119817574702886e-07,
      "loss": 24.2075,
      "step": 17320
    },
    {
      "epoch": 137.54237288135593,
      "grad_norm": 763.580810546875,
      "learning_rate": 1.1172148283518748e-07,
      "loss": 22.0138,
      "step": 17330
    },
    {
      "epoch": 137.62213359920239,
      "grad_norm": 218.8740234375,
      "learning_rate": 1.1146142396380775e-07,
      "loss": 23.3127,
      "step": 17340
    },
    {
      "epoch": 137.70189431704884,
      "grad_norm": 393.57061767578125,
      "learning_rate": 1.112015812619338e-07,
      "loss": 24.1285,
      "step": 17350
    },
    {
      "epoch": 137.78165503489532,
      "grad_norm": 405.6858825683594,
      "learning_rate": 1.1094195513501262e-07,
      "loss": 24.1266,
      "step": 17360
    },
    {
      "epoch": 137.86141575274178,
      "grad_norm": 251.8614044189453,
      "learning_rate": 1.106825459881536e-07,
      "loss": 25.0081,
      "step": 17370
    },
    {
      "epoch": 137.94117647058823,
      "grad_norm": 253.51295471191406,
      "learning_rate": 1.1042335422612729e-07,
      "loss": 24.7537,
      "step": 17380
    },
    {
      "epoch": 138.01595214356928,
      "grad_norm": 524.7293701171875,
      "learning_rate": 1.1016438025336502e-07,
      "loss": 23.3051,
      "step": 17390
    },
    {
      "epoch": 138.09571286141576,
      "grad_norm": 1051.2930908203125,
      "learning_rate": 1.0990562447395863e-07,
      "loss": 23.7153,
      "step": 17400
    },
    {
      "epoch": 138.09571286141576,
      "eval_loss": 2.991021156311035,
      "eval_mae": 1.2980806827545166,
      "eval_mse": 2.991020917892456,
      "eval_r2": 0.09881997108459473,
      "eval_rmse": 1.7294568274150286,
      "eval_runtime": 9.1055,
      "eval_samples_per_second": 440.505,
      "eval_steps_per_second": 13.838,
      "step": 17400
    },
    {
      "epoch": 138.17547357926222,
      "grad_norm": 1116.9071044921875,
      "learning_rate": 1.09647087291659e-07,
      "loss": 22.9059,
      "step": 17410
    },
    {
      "epoch": 138.25523429710867,
      "grad_norm": 268.7237548828125,
      "learning_rate": 1.0938876910987646e-07,
      "loss": 23.7649,
      "step": 17420
    },
    {
      "epoch": 138.33499501495513,
      "grad_norm": 501.5276184082031,
      "learning_rate": 1.091306703316792e-07,
      "loss": 25.158,
      "step": 17430
    },
    {
      "epoch": 138.4147557328016,
      "grad_norm": 416.10491943359375,
      "learning_rate": 1.0887279135979311e-07,
      "loss": 25.7373,
      "step": 17440
    },
    {
      "epoch": 138.49451645064806,
      "grad_norm": 473.6330871582031,
      "learning_rate": 1.0861513259660132e-07,
      "loss": 24.9958,
      "step": 17450
    },
    {
      "epoch": 138.57427716849452,
      "grad_norm": 683.8408203125,
      "learning_rate": 1.0835769444414344e-07,
      "loss": 23.4249,
      "step": 17460
    },
    {
      "epoch": 138.65403788634097,
      "grad_norm": 564.5418090820312,
      "learning_rate": 1.0810047730411425e-07,
      "loss": 22.257,
      "step": 17470
    },
    {
      "epoch": 138.73379860418743,
      "grad_norm": 1422.2911376953125,
      "learning_rate": 1.0784348157786429e-07,
      "loss": 23.0322,
      "step": 17480
    },
    {
      "epoch": 138.8135593220339,
      "grad_norm": 518.5892333984375,
      "learning_rate": 1.0758670766639846e-07,
      "loss": 24.717,
      "step": 17490
    },
    {
      "epoch": 138.89332003988036,
      "grad_norm": 437.5633544921875,
      "learning_rate": 1.073301559703754e-07,
      "loss": 23.9585,
      "step": 17500
    },
    {
      "epoch": 138.89332003988036,
      "eval_loss": 2.9898457527160645,
      "eval_mae": 1.298498511314392,
      "eval_mse": 2.9898459911346436,
      "eval_r2": 0.09917402267456055,
      "eval_rmse": 1.7291171131923493,
      "eval_runtime": 9.1,
      "eval_samples_per_second": 440.769,
      "eval_steps_per_second": 13.846,
      "step": 17500
    },
    {
      "epoch": 138.97308075772682,
      "grad_norm": 623.111572265625,
      "learning_rate": 1.0707382689010702e-07,
      "loss": 22.954,
      "step": 17510
    },
    {
      "epoch": 139.04785643070787,
      "grad_norm": 375.7094421386719,
      "learning_rate": 1.0681772082555812e-07,
      "loss": 22.7863,
      "step": 17520
    },
    {
      "epoch": 139.12761714855435,
      "grad_norm": 550.0701293945312,
      "learning_rate": 1.065618381763452e-07,
      "loss": 24.1999,
      "step": 17530
    },
    {
      "epoch": 139.2073778664008,
      "grad_norm": 768.4359741210938,
      "learning_rate": 1.0630617934173644e-07,
      "loss": 22.5755,
      "step": 17540
    },
    {
      "epoch": 139.28713858424726,
      "grad_norm": 1316.3155517578125,
      "learning_rate": 1.0605074472065056e-07,
      "loss": 27.1047,
      "step": 17550
    },
    {
      "epoch": 139.3668993020937,
      "grad_norm": 213.6504669189453,
      "learning_rate": 1.0579553471165647e-07,
      "loss": 23.0379,
      "step": 17560
    },
    {
      "epoch": 139.44666001994017,
      "grad_norm": 363.3310852050781,
      "learning_rate": 1.0554054971297271e-07,
      "loss": 22.4877,
      "step": 17570
    },
    {
      "epoch": 139.52642073778665,
      "grad_norm": 393.981689453125,
      "learning_rate": 1.0528579012246686e-07,
      "loss": 25.587,
      "step": 17580
    },
    {
      "epoch": 139.6061814556331,
      "grad_norm": 402.7592468261719,
      "learning_rate": 1.0503125633765422e-07,
      "loss": 21.617,
      "step": 17590
    },
    {
      "epoch": 139.68594217347956,
      "grad_norm": 1331.46728515625,
      "learning_rate": 1.0477694875569831e-07,
      "loss": 24.8125,
      "step": 17600
    },
    {
      "epoch": 139.68594217347956,
      "eval_loss": 2.9967098236083984,
      "eval_mae": 1.3076661825180054,
      "eval_mse": 2.9967100620269775,
      "eval_r2": 0.09710592031478882,
      "eval_rmse": 1.7311008237612786,
      "eval_runtime": 9.0864,
      "eval_samples_per_second": 441.43,
      "eval_steps_per_second": 13.867,
      "step": 17600
    },
    {
      "epoch": 139.76570289132601,
      "grad_norm": 1280.6114501953125,
      "learning_rate": 1.0452286777340957e-07,
      "loss": 22.8116,
      "step": 17610
    },
    {
      "epoch": 139.84546360917247,
      "grad_norm": 479.7171630859375,
      "learning_rate": 1.0426901378724476e-07,
      "loss": 23.9819,
      "step": 17620
    },
    {
      "epoch": 139.92522432701895,
      "grad_norm": 1034.206298828125,
      "learning_rate": 1.0401538719330632e-07,
      "loss": 24.7589,
      "step": 17630
    },
    {
      "epoch": 140.0,
      "grad_norm": 569.5013427734375,
      "learning_rate": 1.0376198838734229e-07,
      "loss": 22.1108,
      "step": 17640
    },
    {
      "epoch": 140.07976071784645,
      "grad_norm": 302.1639709472656,
      "learning_rate": 1.0350881776474482e-07,
      "loss": 23.5557,
      "step": 17650
    },
    {
      "epoch": 140.1595214356929,
      "grad_norm": 673.09326171875,
      "learning_rate": 1.0325587572055042e-07,
      "loss": 22.7825,
      "step": 17660
    },
    {
      "epoch": 140.2392821535394,
      "grad_norm": 493.1293640136719,
      "learning_rate": 1.0300316264943861e-07,
      "loss": 24.1157,
      "step": 17670
    },
    {
      "epoch": 140.31904287138585,
      "grad_norm": 282.7809753417969,
      "learning_rate": 1.0275067894573173e-07,
      "loss": 24.3361,
      "step": 17680
    },
    {
      "epoch": 140.3988035892323,
      "grad_norm": 1053.279052734375,
      "learning_rate": 1.024984250033944e-07,
      "loss": 24.004,
      "step": 17690
    },
    {
      "epoch": 140.47856430707876,
      "grad_norm": 756.5474243164062,
      "learning_rate": 1.0224640121603248e-07,
      "loss": 21.715,
      "step": 17700
    },
    {
      "epoch": 140.47856430707876,
      "eval_loss": 2.9895496368408203,
      "eval_mae": 1.296020746231079,
      "eval_mse": 2.9895496368408203,
      "eval_r2": 0.0992632508277893,
      "eval_rmse": 1.7290314158050513,
      "eval_runtime": 9.0911,
      "eval_samples_per_second": 441.199,
      "eval_steps_per_second": 13.86,
      "step": 17700
    },
    {
      "epoch": 140.5583250249252,
      "grad_norm": 758.3074340820312,
      "learning_rate": 1.0199460797689274e-07,
      "loss": 23.4764,
      "step": 17710
    },
    {
      "epoch": 140.6380857427717,
      "grad_norm": 194.74085998535156,
      "learning_rate": 1.0174304567886246e-07,
      "loss": 23.0024,
      "step": 17720
    },
    {
      "epoch": 140.71784646061815,
      "grad_norm": 564.7682495117188,
      "learning_rate": 1.0149171471446824e-07,
      "loss": 26.9969,
      "step": 17730
    },
    {
      "epoch": 140.7976071784646,
      "grad_norm": 430.6932678222656,
      "learning_rate": 1.0124061547587603e-07,
      "loss": 23.3746,
      "step": 17740
    },
    {
      "epoch": 140.87736789631106,
      "grad_norm": 475.7787170410156,
      "learning_rate": 1.0098974835488996e-07,
      "loss": 24.8886,
      "step": 17750
    },
    {
      "epoch": 140.95712861415754,
      "grad_norm": 1146.912841796875,
      "learning_rate": 1.0073911374295196e-07,
      "loss": 25.6257,
      "step": 17760
    },
    {
      "epoch": 141.0319042871386,
      "grad_norm": 535.0670776367188,
      "learning_rate": 1.0048871203114137e-07,
      "loss": 23.0122,
      "step": 17770
    },
    {
      "epoch": 141.11166500498504,
      "grad_norm": 176.4267120361328,
      "learning_rate": 1.002385436101742e-07,
      "loss": 22.0339,
      "step": 17780
    },
    {
      "epoch": 141.1914257228315,
      "grad_norm": 1003.4795532226562,
      "learning_rate": 9.998860887040189e-08,
      "loss": 25.6698,
      "step": 17790
    },
    {
      "epoch": 141.27118644067798,
      "grad_norm": 579.8351440429688,
      "learning_rate": 9.97389082018118e-08,
      "loss": 22.7872,
      "step": 17800
    },
    {
      "epoch": 141.27118644067798,
      "eval_loss": 2.9897937774658203,
      "eval_mae": 1.297013282775879,
      "eval_mse": 2.9897942543029785,
      "eval_r2": 0.0991896390914917,
      "eval_rmse": 1.7291021526511898,
      "eval_runtime": 9.0748,
      "eval_samples_per_second": 441.995,
      "eval_steps_per_second": 13.885,
      "step": 17800
    },
    {
      "epoch": 141.35094715852443,
      "grad_norm": 669.0322265625,
      "learning_rate": 9.948944199402596e-08,
      "loss": 25.5954,
      "step": 17810
    },
    {
      "epoch": 141.4307078763709,
      "grad_norm": 266.249267578125,
      "learning_rate": 9.924021063630042e-08,
      "loss": 24.4365,
      "step": 17820
    },
    {
      "epoch": 141.51046859421734,
      "grad_norm": 787.0618896484375,
      "learning_rate": 9.899121451752474e-08,
      "loss": 23.2357,
      "step": 17830
    },
    {
      "epoch": 141.5902293120638,
      "grad_norm": 1211.522705078125,
      "learning_rate": 9.874245402622167e-08,
      "loss": 22.2488,
      "step": 17840
    },
    {
      "epoch": 141.66999002991028,
      "grad_norm": 712.261474609375,
      "learning_rate": 9.849392955054601e-08,
      "loss": 23.9956,
      "step": 17850
    },
    {
      "epoch": 141.74975074775674,
      "grad_norm": 92.70515441894531,
      "learning_rate": 9.824564147828457e-08,
      "loss": 22.3979,
      "step": 17860
    },
    {
      "epoch": 141.8295114656032,
      "grad_norm": 221.4393768310547,
      "learning_rate": 9.799759019685507e-08,
      "loss": 22.6697,
      "step": 17870
    },
    {
      "epoch": 141.90927218344964,
      "grad_norm": 320.49127197265625,
      "learning_rate": 9.774977609330575e-08,
      "loss": 24.4215,
      "step": 17880
    },
    {
      "epoch": 141.9890329012961,
      "grad_norm": 216.5183868408203,
      "learning_rate": 9.750219955431504e-08,
      "loss": 25.499,
      "step": 17890
    },
    {
      "epoch": 142.06380857427718,
      "grad_norm": 1004.3078002929688,
      "learning_rate": 9.725486096619037e-08,
      "loss": 22.1796,
      "step": 17900
    },
    {
      "epoch": 142.06380857427718,
      "eval_loss": 3.0014090538024902,
      "eval_mae": 1.3100292682647705,
      "eval_mse": 3.0014090538024902,
      "eval_r2": 0.09569013118743896,
      "eval_rmse": 1.73245751861409,
      "eval_runtime": 9.0857,
      "eval_samples_per_second": 441.461,
      "eval_steps_per_second": 13.868,
      "step": 17900
    },
    {
      "epoch": 142.14356929212363,
      "grad_norm": 897.8434448242188,
      "learning_rate": 9.700776071486794e-08,
      "loss": 23.5536,
      "step": 17910
    },
    {
      "epoch": 142.22333000997008,
      "grad_norm": 592.3159790039062,
      "learning_rate": 9.676089918591226e-08,
      "loss": 25.0882,
      "step": 17920
    },
    {
      "epoch": 142.30309072781654,
      "grad_norm": 381.72515869140625,
      "learning_rate": 9.65142767645151e-08,
      "loss": 23.2826,
      "step": 17930
    },
    {
      "epoch": 142.38285144566302,
      "grad_norm": 295.14013671875,
      "learning_rate": 9.626789383549527e-08,
      "loss": 25.3821,
      "step": 17940
    },
    {
      "epoch": 142.46261216350948,
      "grad_norm": 551.1314697265625,
      "learning_rate": 9.602175078329808e-08,
      "loss": 21.8017,
      "step": 17950
    },
    {
      "epoch": 142.54237288135593,
      "grad_norm": 599.72509765625,
      "learning_rate": 9.577584799199395e-08,
      "loss": 23.8833,
      "step": 17960
    },
    {
      "epoch": 142.62213359920239,
      "grad_norm": 599.3013916015625,
      "learning_rate": 9.5530185845279e-08,
      "loss": 25.899,
      "step": 17970
    },
    {
      "epoch": 142.70189431704884,
      "grad_norm": 288.3550109863281,
      "learning_rate": 9.52847647264737e-08,
      "loss": 22.9422,
      "step": 17980
    },
    {
      "epoch": 142.78165503489532,
      "grad_norm": 1174.3599853515625,
      "learning_rate": 9.503958501852233e-08,
      "loss": 21.962,
      "step": 17990
    },
    {
      "epoch": 142.86141575274178,
      "grad_norm": 667.4810791015625,
      "learning_rate": 9.479464710399243e-08,
      "loss": 24.851,
      "step": 18000
    },
    {
      "epoch": 142.86141575274178,
      "eval_loss": 2.9901320934295654,
      "eval_mae": 1.2980480194091797,
      "eval_mse": 2.9901320934295654,
      "eval_r2": 0.09908783435821533,
      "eval_rmse": 1.7291998419585763,
      "eval_runtime": 9.0739,
      "eval_samples_per_second": 442.036,
      "eval_steps_per_second": 13.886,
      "step": 18000
    },
    {
      "epoch": 142.94117647058823,
      "grad_norm": 338.58465576171875,
      "learning_rate": 9.454995136507451e-08,
      "loss": 23.9936,
      "step": 18010
    },
    {
      "epoch": 143.01595214356928,
      "grad_norm": 1124.7645263671875,
      "learning_rate": 9.430549818358088e-08,
      "loss": 23.218,
      "step": 18020
    },
    {
      "epoch": 143.09571286141576,
      "grad_norm": 153.89315795898438,
      "learning_rate": 9.406128794094576e-08,
      "loss": 25.5854,
      "step": 18030
    },
    {
      "epoch": 143.17547357926222,
      "grad_norm": 217.41473388671875,
      "learning_rate": 9.381732101822393e-08,
      "loss": 22.5631,
      "step": 18040
    },
    {
      "epoch": 143.25523429710867,
      "grad_norm": 642.3425903320312,
      "learning_rate": 9.357359779609058e-08,
      "loss": 23.2905,
      "step": 18050
    },
    {
      "epoch": 143.33499501495513,
      "grad_norm": 984.7265014648438,
      "learning_rate": 9.333011865484086e-08,
      "loss": 24.2562,
      "step": 18060
    },
    {
      "epoch": 143.4147557328016,
      "grad_norm": 294.946533203125,
      "learning_rate": 9.308688397438901e-08,
      "loss": 24.3051,
      "step": 18070
    },
    {
      "epoch": 143.49451645064806,
      "grad_norm": 652.6814575195312,
      "learning_rate": 9.284389413426746e-08,
      "loss": 23.6761,
      "step": 18080
    },
    {
      "epoch": 143.57427716849452,
      "grad_norm": 382.2814636230469,
      "learning_rate": 9.260114951362702e-08,
      "loss": 22.0659,
      "step": 18090
    },
    {
      "epoch": 143.65403788634097,
      "grad_norm": 193.34925842285156,
      "learning_rate": 9.23586504912358e-08,
      "loss": 24.6004,
      "step": 18100
    },
    {
      "epoch": 143.65403788634097,
      "eval_loss": 2.9914941787719727,
      "eval_mae": 1.2961755990982056,
      "eval_mse": 2.9914941787719727,
      "eval_r2": 0.09867745637893677,
      "eval_rmse": 1.7295936455630185,
      "eval_runtime": 9.0808,
      "eval_samples_per_second": 441.702,
      "eval_steps_per_second": 13.875,
      "step": 18100
    },
    {
      "epoch": 143.73379860418743,
      "grad_norm": 255.50914001464844,
      "learning_rate": 9.211639744547858e-08,
      "loss": 21.8854,
      "step": 18110
    },
    {
      "epoch": 143.8135593220339,
      "grad_norm": 798.5938110351562,
      "learning_rate": 9.187439075435624e-08,
      "loss": 25.0725,
      "step": 18120
    },
    {
      "epoch": 143.89332003988036,
      "grad_norm": 554.2741088867188,
      "learning_rate": 9.163263079548556e-08,
      "loss": 25.6367,
      "step": 18130
    },
    {
      "epoch": 143.97308075772682,
      "grad_norm": 278.55767822265625,
      "learning_rate": 9.139111794609799e-08,
      "loss": 24.5959,
      "step": 18140
    },
    {
      "epoch": 144.04785643070787,
      "grad_norm": 290.1928405761719,
      "learning_rate": 9.114985258303972e-08,
      "loss": 20.7683,
      "step": 18150
    },
    {
      "epoch": 144.12761714855435,
      "grad_norm": 957.6152954101562,
      "learning_rate": 9.090883508277058e-08,
      "loss": 24.8539,
      "step": 18160
    },
    {
      "epoch": 144.2073778664008,
      "grad_norm": 201.96681213378906,
      "learning_rate": 9.066806582136357e-08,
      "loss": 23.961,
      "step": 18170
    },
    {
      "epoch": 144.28713858424726,
      "grad_norm": 442.1460876464844,
      "learning_rate": 9.042754517450463e-08,
      "loss": 22.6299,
      "step": 18180
    },
    {
      "epoch": 144.3668993020937,
      "grad_norm": 326.3381042480469,
      "learning_rate": 9.018727351749153e-08,
      "loss": 21.6338,
      "step": 18190
    },
    {
      "epoch": 144.44666001994017,
      "grad_norm": 411.69830322265625,
      "learning_rate": 8.994725122523353e-08,
      "loss": 22.0807,
      "step": 18200
    },
    {
      "epoch": 144.44666001994017,
      "eval_loss": 2.989143133163452,
      "eval_mae": 1.2954165935516357,
      "eval_mse": 2.989143133163452,
      "eval_r2": 0.09938579797744751,
      "eval_rmse": 1.7289138593820839,
      "eval_runtime": 9.07,
      "eval_samples_per_second": 442.228,
      "eval_steps_per_second": 13.892,
      "step": 18200
    },
    {
      "epoch": 144.52642073778665,
      "grad_norm": 568.3926391601562,
      "learning_rate": 8.9707478672251e-08,
      "loss": 25.0817,
      "step": 18210
    },
    {
      "epoch": 144.6061814556331,
      "grad_norm": 781.808349609375,
      "learning_rate": 8.94679562326744e-08,
      "loss": 24.5619,
      "step": 18220
    },
    {
      "epoch": 144.68594217347956,
      "grad_norm": 268.1374816894531,
      "learning_rate": 8.922868428024408e-08,
      "loss": 24.0362,
      "step": 18230
    },
    {
      "epoch": 144.76570289132601,
      "grad_norm": 390.1956787109375,
      "learning_rate": 8.89896631883095e-08,
      "loss": 25.5777,
      "step": 18240
    },
    {
      "epoch": 144.84546360917247,
      "grad_norm": 286.0571594238281,
      "learning_rate": 8.87508933298285e-08,
      "loss": 25.1571,
      "step": 18250
    },
    {
      "epoch": 144.92522432701895,
      "grad_norm": 972.77294921875,
      "learning_rate": 8.851237507736722e-08,
      "loss": 22.4625,
      "step": 18260
    },
    {
      "epoch": 145.0,
      "grad_norm": 560.763427734375,
      "learning_rate": 8.827410880309918e-08,
      "loss": 23.7497,
      "step": 18270
    },
    {
      "epoch": 145.07976071784645,
      "grad_norm": 493.7109680175781,
      "learning_rate": 8.803609487880431e-08,
      "loss": 22.5282,
      "step": 18280
    },
    {
      "epoch": 145.1595214356929,
      "grad_norm": 1222.901611328125,
      "learning_rate": 8.779833367586922e-08,
      "loss": 23.7649,
      "step": 18290
    },
    {
      "epoch": 145.2392821535394,
      "grad_norm": 182.56158447265625,
      "learning_rate": 8.756082556528612e-08,
      "loss": 25.1346,
      "step": 18300
    },
    {
      "epoch": 145.2392821535394,
      "eval_loss": 2.9885616302490234,
      "eval_mae": 1.2971998453140259,
      "eval_mse": 2.9885613918304443,
      "eval_r2": 0.09956103563308716,
      "eval_rmse": 1.7287456122375102,
      "eval_runtime": 9.0854,
      "eval_samples_per_second": 441.479,
      "eval_steps_per_second": 13.868,
      "step": 18300
    },
    {
      "epoch": 145.31904287138585,
      "grad_norm": 313.9729309082031,
      "learning_rate": 8.732357091765219e-08,
      "loss": 24.3856,
      "step": 18310
    },
    {
      "epoch": 145.3988035892323,
      "grad_norm": 387.6947937011719,
      "learning_rate": 8.7086570103169e-08,
      "loss": 22.9606,
      "step": 18320
    },
    {
      "epoch": 145.47856430707876,
      "grad_norm": 251.2086639404297,
      "learning_rate": 8.684982349164244e-08,
      "loss": 22.8897,
      "step": 18330
    },
    {
      "epoch": 145.5583250249252,
      "grad_norm": 218.4263916015625,
      "learning_rate": 8.661333145248129e-08,
      "loss": 24.4742,
      "step": 18340
    },
    {
      "epoch": 145.6380857427717,
      "grad_norm": 355.86907958984375,
      "learning_rate": 8.637709435469756e-08,
      "loss": 23.6665,
      "step": 18350
    },
    {
      "epoch": 145.71784646061815,
      "grad_norm": 143.9767303466797,
      "learning_rate": 8.61411125669051e-08,
      "loss": 23.7774,
      "step": 18360
    },
    {
      "epoch": 145.7976071784646,
      "grad_norm": 142.5809783935547,
      "learning_rate": 8.590538645731948e-08,
      "loss": 24.6402,
      "step": 18370
    },
    {
      "epoch": 145.87736789631106,
      "grad_norm": 287.8326110839844,
      "learning_rate": 8.56699163937575e-08,
      "loss": 24.4677,
      "step": 18380
    },
    {
      "epoch": 145.95712861415754,
      "grad_norm": 124.76634216308594,
      "learning_rate": 8.543470274363623e-08,
      "loss": 24.3025,
      "step": 18390
    },
    {
      "epoch": 146.0319042871386,
      "grad_norm": 717.8211669921875,
      "learning_rate": 8.519974587397263e-08,
      "loss": 21.0815,
      "step": 18400
    },
    {
      "epoch": 146.0319042871386,
      "eval_loss": 2.9937267303466797,
      "eval_mae": 1.29371178150177,
      "eval_mse": 2.9937262535095215,
      "eval_r2": 0.09800493717193604,
      "eval_rmse": 1.730238785113061,
      "eval_runtime": 9.0846,
      "eval_samples_per_second": 441.518,
      "eval_steps_per_second": 13.87,
      "step": 18400
    },
    {
      "epoch": 146.11166500498504,
      "grad_norm": 151.96337890625,
      "learning_rate": 8.496504615138325e-08,
      "loss": 24.258,
      "step": 18410
    },
    {
      "epoch": 146.1914257228315,
      "grad_norm": 306.97119140625,
      "learning_rate": 8.473060394208304e-08,
      "loss": 25.4368,
      "step": 18420
    },
    {
      "epoch": 146.27118644067798,
      "grad_norm": 157.916259765625,
      "learning_rate": 8.44964196118855e-08,
      "loss": 22.126,
      "step": 18430
    },
    {
      "epoch": 146.35094715852443,
      "grad_norm": 317.0219421386719,
      "learning_rate": 8.426249352620144e-08,
      "loss": 23.0198,
      "step": 18440
    },
    {
      "epoch": 146.4307078763709,
      "grad_norm": 213.1689910888672,
      "learning_rate": 8.402882605003884e-08,
      "loss": 26.4119,
      "step": 18450
    },
    {
      "epoch": 146.51046859421734,
      "grad_norm": 183.77806091308594,
      "learning_rate": 8.379541754800215e-08,
      "loss": 23.2543,
      "step": 18460
    },
    {
      "epoch": 146.5902293120638,
      "grad_norm": 148.0422821044922,
      "learning_rate": 8.356226838429198e-08,
      "loss": 23.2553,
      "step": 18470
    },
    {
      "epoch": 146.66999002991028,
      "grad_norm": 533.5089111328125,
      "learning_rate": 8.33293789227037e-08,
      "loss": 24.2542,
      "step": 18480
    },
    {
      "epoch": 146.74975074775674,
      "grad_norm": 433.2427673339844,
      "learning_rate": 8.309674952662791e-08,
      "loss": 24.8082,
      "step": 18490
    },
    {
      "epoch": 146.8295114656032,
      "grad_norm": 509.19036865234375,
      "learning_rate": 8.286438055904943e-08,
      "loss": 24.1001,
      "step": 18500
    },
    {
      "epoch": 146.8295114656032,
      "eval_loss": 2.9880919456481934,
      "eval_mae": 1.2967747449874878,
      "eval_mse": 2.9880919456481934,
      "eval_r2": 0.09970253705978394,
      "eval_rmse": 1.7286098303689568,
      "eval_runtime": 9.0561,
      "eval_samples_per_second": 442.906,
      "eval_steps_per_second": 13.913,
      "step": 18500
    },
    {
      "epoch": 146.90927218344964,
      "grad_norm": 518.4598388671875,
      "learning_rate": 8.263227238254639e-08,
      "loss": 23.1411,
      "step": 18510
    },
    {
      "epoch": 146.9890329012961,
      "grad_norm": 241.46841430664062,
      "learning_rate": 8.240042535929035e-08,
      "loss": 22.9866,
      "step": 18520
    },
    {
      "epoch": 147.06380857427718,
      "grad_norm": 152.85427856445312,
      "learning_rate": 8.216883985104514e-08,
      "loss": 22.618,
      "step": 18530
    },
    {
      "epoch": 147.14356929212363,
      "grad_norm": 417.3365783691406,
      "learning_rate": 8.193751621916655e-08,
      "loss": 22.4225,
      "step": 18540
    },
    {
      "epoch": 147.22333000997008,
      "grad_norm": 257.4353332519531,
      "learning_rate": 8.170645482460187e-08,
      "loss": 23.5731,
      "step": 18550
    },
    {
      "epoch": 147.30309072781654,
      "grad_norm": 996.5963745117188,
      "learning_rate": 8.14756560278893e-08,
      "loss": 23.9687,
      "step": 18560
    },
    {
      "epoch": 147.38285144566302,
      "grad_norm": 633.5903930664062,
      "learning_rate": 8.124512018915685e-08,
      "loss": 24.7502,
      "step": 18570
    },
    {
      "epoch": 147.46261216350948,
      "grad_norm": 803.2030029296875,
      "learning_rate": 8.101484766812264e-08,
      "loss": 24.6082,
      "step": 18580
    },
    {
      "epoch": 147.54237288135593,
      "grad_norm": 143.506591796875,
      "learning_rate": 8.078483882409387e-08,
      "loss": 25.6971,
      "step": 18590
    },
    {
      "epoch": 147.62213359920239,
      "grad_norm": 147.65036010742188,
      "learning_rate": 8.055509401596619e-08,
      "loss": 21.8963,
      "step": 18600
    },
    {
      "epoch": 147.62213359920239,
      "eval_loss": 2.990255117416382,
      "eval_mae": 1.295224905014038,
      "eval_mse": 2.9902548789978027,
      "eval_r2": 0.09905081987380981,
      "eval_rmse": 1.7292353451736413,
      "eval_runtime": 9.0673,
      "eval_samples_per_second": 442.36,
      "eval_steps_per_second": 13.896,
      "step": 18600
    },
    {
      "epoch": 147.70189431704884,
      "grad_norm": 811.623046875,
      "learning_rate": 8.032561360222318e-08,
      "loss": 24.3758,
      "step": 18610
    },
    {
      "epoch": 147.78165503489532,
      "grad_norm": 271.70947265625,
      "learning_rate": 8.009639794093617e-08,
      "loss": 24.3177,
      "step": 18620
    },
    {
      "epoch": 147.86141575274178,
      "grad_norm": 147.8336181640625,
      "learning_rate": 7.986744738976303e-08,
      "loss": 23.4661,
      "step": 18630
    },
    {
      "epoch": 147.94117647058823,
      "grad_norm": 1055.819580078125,
      "learning_rate": 7.963876230594835e-08,
      "loss": 22.1426,
      "step": 18640
    },
    {
      "epoch": 148.01595214356928,
      "grad_norm": 908.6998291015625,
      "learning_rate": 7.941034304632214e-08,
      "loss": 23.6974,
      "step": 18650
    },
    {
      "epoch": 148.09571286141576,
      "grad_norm": 292.4888610839844,
      "learning_rate": 7.918218996729973e-08,
      "loss": 22.3105,
      "step": 18660
    },
    {
      "epoch": 148.17547357926222,
      "grad_norm": 280.0393981933594,
      "learning_rate": 7.89543034248813e-08,
      "loss": 22.5675,
      "step": 18670
    },
    {
      "epoch": 148.25523429710867,
      "grad_norm": 578.9780883789062,
      "learning_rate": 7.872668377465094e-08,
      "loss": 22.3394,
      "step": 18680
    },
    {
      "epoch": 148.33499501495513,
      "grad_norm": 352.46331787109375,
      "learning_rate": 7.849933137177625e-08,
      "loss": 22.8286,
      "step": 18690
    },
    {
      "epoch": 148.4147557328016,
      "grad_norm": 263.3887939453125,
      "learning_rate": 7.827224657100809e-08,
      "loss": 23.9573,
      "step": 18700
    },
    {
      "epoch": 148.4147557328016,
      "eval_loss": 2.989132881164551,
      "eval_mae": 1.2960878610610962,
      "eval_mse": 2.989132881164551,
      "eval_r2": 0.09938883781433105,
      "eval_rmse": 1.7289108945126554,
      "eval_runtime": 9.0662,
      "eval_samples_per_second": 442.414,
      "eval_steps_per_second": 13.898,
      "step": 18700
    },
    {
      "epoch": 148.49451645064806,
      "grad_norm": 498.36920166015625,
      "learning_rate": 7.804542972667943e-08,
      "loss": 23.0484,
      "step": 18710
    },
    {
      "epoch": 148.57427716849452,
      "grad_norm": 172.67092895507812,
      "learning_rate": 7.781888119270552e-08,
      "loss": 22.4251,
      "step": 18720
    },
    {
      "epoch": 148.65403788634097,
      "grad_norm": 711.1735229492188,
      "learning_rate": 7.759260132258258e-08,
      "loss": 24.8158,
      "step": 18730
    },
    {
      "epoch": 148.73379860418743,
      "grad_norm": 180.4683380126953,
      "learning_rate": 7.736659046938776e-08,
      "loss": 26.0731,
      "step": 18740
    },
    {
      "epoch": 148.8135593220339,
      "grad_norm": 219.04393005371094,
      "learning_rate": 7.714084898577853e-08,
      "loss": 24.9301,
      "step": 18750
    },
    {
      "epoch": 148.89332003988036,
      "grad_norm": 310.49652099609375,
      "learning_rate": 7.691537722399208e-08,
      "loss": 26.3574,
      "step": 18760
    },
    {
      "epoch": 148.97308075772682,
      "grad_norm": 840.8336181640625,
      "learning_rate": 7.669017553584439e-08,
      "loss": 25.9603,
      "step": 18770
    },
    {
      "epoch": 149.04785643070787,
      "grad_norm": 538.75634765625,
      "learning_rate": 7.646524427273044e-08,
      "loss": 22.1064,
      "step": 18780
    },
    {
      "epoch": 149.12761714855435,
      "grad_norm": 635.2152709960938,
      "learning_rate": 7.624058378562312e-08,
      "loss": 23.8754,
      "step": 18790
    },
    {
      "epoch": 149.2073778664008,
      "grad_norm": 474.3734130859375,
      "learning_rate": 7.601619442507271e-08,
      "loss": 23.7222,
      "step": 18800
    },
    {
      "epoch": 149.2073778664008,
      "eval_loss": 2.988734245300293,
      "eval_mae": 1.2976834774017334,
      "eval_mse": 2.988734483718872,
      "eval_r2": 0.09950888156890869,
      "eval_rmse": 1.7287956743695514,
      "eval_runtime": 9.084,
      "eval_samples_per_second": 441.548,
      "eval_steps_per_second": 13.871,
      "step": 18800
    },
    {
      "epoch": 149.28713858424726,
      "grad_norm": 629.4293823242188,
      "learning_rate": 7.579207654120645e-08,
      "loss": 24.2276,
      "step": 18810
    },
    {
      "epoch": 149.3668993020937,
      "grad_norm": 448.8154602050781,
      "learning_rate": 7.556823048372817e-08,
      "loss": 24.7479,
      "step": 18820
    },
    {
      "epoch": 149.44666001994017,
      "grad_norm": 151.45706176757812,
      "learning_rate": 7.53446566019173e-08,
      "loss": 24.5128,
      "step": 18830
    },
    {
      "epoch": 149.52642073778665,
      "grad_norm": 187.52345275878906,
      "learning_rate": 7.512135524462881e-08,
      "loss": 22.5111,
      "step": 18840
    },
    {
      "epoch": 149.6061814556331,
      "grad_norm": 450.91326904296875,
      "learning_rate": 7.489832676029232e-08,
      "loss": 23.3607,
      "step": 18850
    },
    {
      "epoch": 149.68594217347956,
      "grad_norm": 212.0082244873047,
      "learning_rate": 7.46755714969115e-08,
      "loss": 23.701,
      "step": 18860
    },
    {
      "epoch": 149.76570289132601,
      "grad_norm": 790.0916137695312,
      "learning_rate": 7.447532565106768e-08,
      "loss": 24.3561,
      "step": 18870
    },
    {
      "epoch": 149.84546360917247,
      "grad_norm": 492.77972412109375,
      "learning_rate": 7.425309046472627e-08,
      "loss": 23.1056,
      "step": 18880
    },
    {
      "epoch": 149.92522432701895,
      "grad_norm": 1054.5626220703125,
      "learning_rate": 7.40311295061389e-08,
      "loss": 22.2082,
      "step": 18890
    },
    {
      "epoch": 150.0,
      "grad_norm": 734.5057983398438,
      "learning_rate": 7.380944312164386e-08,
      "loss": 23.3368,
      "step": 18900
    },
    {
      "epoch": 150.0,
      "eval_loss": 2.988919496536255,
      "eval_mae": 1.2994049787521362,
      "eval_mse": 2.988919734954834,
      "eval_r2": 0.0994531512260437,
      "eval_rmse": 1.728849251656961,
      "eval_runtime": 9.0753,
      "eval_samples_per_second": 441.97,
      "eval_steps_per_second": 13.884,
      "step": 18900
    },
    {
      "epoch": 150.07976071784645,
      "grad_norm": 494.13604736328125,
      "learning_rate": 7.358803165715091e-08,
      "loss": 23.7358,
      "step": 18910
    },
    {
      "epoch": 150.1595214356929,
      "grad_norm": 604.9675903320312,
      "learning_rate": 7.336689545814051e-08,
      "loss": 23.2456,
      "step": 18920
    },
    {
      "epoch": 150.2392821535394,
      "grad_norm": 1062.545654296875,
      "learning_rate": 7.3146034869664e-08,
      "loss": 23.7062,
      "step": 18930
    },
    {
      "epoch": 150.31904287138585,
      "grad_norm": 198.54518127441406,
      "learning_rate": 7.292545023634264e-08,
      "loss": 24.4768,
      "step": 18940
    },
    {
      "epoch": 150.3988035892323,
      "grad_norm": 414.9964904785156,
      "learning_rate": 7.270514190236691e-08,
      "loss": 24.0219,
      "step": 18950
    },
    {
      "epoch": 150.47856430707876,
      "grad_norm": 433.3045349121094,
      "learning_rate": 7.248511021149614e-08,
      "loss": 22.1332,
      "step": 18960
    },
    {
      "epoch": 150.5583250249252,
      "grad_norm": 282.57342529296875,
      "learning_rate": 7.226535550705831e-08,
      "loss": 22.157,
      "step": 18970
    },
    {
      "epoch": 150.6380857427717,
      "grad_norm": 497.9869384765625,
      "learning_rate": 7.204587813194885e-08,
      "loss": 23.2071,
      "step": 18980
    },
    {
      "epoch": 150.71784646061815,
      "grad_norm": 345.4354553222656,
      "learning_rate": 7.182667842863072e-08,
      "loss": 23.0731,
      "step": 18990
    },
    {
      "epoch": 150.7976071784646,
      "grad_norm": 1432.7169189453125,
      "learning_rate": 7.160775673913347e-08,
      "loss": 24.0156,
      "step": 19000
    },
    {
      "epoch": 150.7976071784646,
      "eval_loss": 2.988027572631836,
      "eval_mae": 1.294617772102356,
      "eval_mse": 2.9880270957946777,
      "eval_r2": 0.09972202777862549,
      "eval_rmse": 1.728591072461812,
      "eval_runtime": 9.0673,
      "eval_samples_per_second": 442.36,
      "eval_steps_per_second": 13.896,
      "step": 19000
    },
    {
      "epoch": 150.87736789631106,
      "grad_norm": 461.5654296875,
      "learning_rate": 7.138911340505277e-08,
      "loss": 26.2207,
      "step": 19010
    },
    {
      "epoch": 150.95712861415754,
      "grad_norm": 546.268798828125,
      "learning_rate": 7.117074876755028e-08,
      "loss": 25.6184,
      "step": 19020
    },
    {
      "epoch": 151.0319042871386,
      "grad_norm": 317.89990234375,
      "learning_rate": 7.095266316735246e-08,
      "loss": 21.8385,
      "step": 19030
    },
    {
      "epoch": 151.11166500498504,
      "grad_norm": 225.2183380126953,
      "learning_rate": 7.073485694475042e-08,
      "loss": 22.523,
      "step": 19040
    },
    {
      "epoch": 151.1914257228315,
      "grad_norm": 1056.931884765625,
      "learning_rate": 7.051733043959959e-08,
      "loss": 25.8547,
      "step": 19050
    },
    {
      "epoch": 151.27118644067798,
      "grad_norm": 271.0970458984375,
      "learning_rate": 7.030008399131859e-08,
      "loss": 24.099,
      "step": 19060
    },
    {
      "epoch": 151.35094715852443,
      "grad_norm": 183.71864318847656,
      "learning_rate": 7.008311793888943e-08,
      "loss": 25.2921,
      "step": 19070
    },
    {
      "epoch": 151.4307078763709,
      "grad_norm": 432.2311096191406,
      "learning_rate": 6.98664326208562e-08,
      "loss": 22.6419,
      "step": 19080
    },
    {
      "epoch": 151.51046859421734,
      "grad_norm": 131.0139923095703,
      "learning_rate": 6.965002837532536e-08,
      "loss": 26.2508,
      "step": 19090
    },
    {
      "epoch": 151.5902293120638,
      "grad_norm": 782.22900390625,
      "learning_rate": 6.943390553996436e-08,
      "loss": 23.513,
      "step": 19100
    },
    {
      "epoch": 151.5902293120638,
      "eval_loss": 2.9875853061676025,
      "eval_mae": 1.2965117692947388,
      "eval_mse": 2.9875853061676025,
      "eval_r2": 0.09985512495040894,
      "eval_rmse": 1.7284632788021859,
      "eval_runtime": 9.0747,
      "eval_samples_per_second": 441.998,
      "eval_steps_per_second": 13.885,
      "step": 19100
    },
    {
      "epoch": 151.66999002991028,
      "grad_norm": 759.2440185546875,
      "learning_rate": 6.921806445200198e-08,
      "loss": 22.4186,
      "step": 19110
    },
    {
      "epoch": 151.74975074775674,
      "grad_norm": 289.4839172363281,
      "learning_rate": 6.900250544822705e-08,
      "loss": 25.331,
      "step": 19120
    },
    {
      "epoch": 151.8295114656032,
      "grad_norm": 499.3775329589844,
      "learning_rate": 6.878722886498836e-08,
      "loss": 24.3565,
      "step": 19130
    },
    {
      "epoch": 151.90927218344964,
      "grad_norm": 393.87469482421875,
      "learning_rate": 6.85722350381941e-08,
      "loss": 22.3572,
      "step": 19140
    },
    {
      "epoch": 151.9890329012961,
      "grad_norm": 187.20021057128906,
      "learning_rate": 6.835752430331116e-08,
      "loss": 22.3619,
      "step": 19150
    },
    {
      "epoch": 152.06380857427718,
      "grad_norm": 673.76953125,
      "learning_rate": 6.814309699536466e-08,
      "loss": 22.2599,
      "step": 19160
    },
    {
      "epoch": 152.14356929212363,
      "grad_norm": 531.2612915039062,
      "learning_rate": 6.792895344893768e-08,
      "loss": 24.8074,
      "step": 19170
    },
    {
      "epoch": 152.22333000997008,
      "grad_norm": 632.0092163085938,
      "learning_rate": 6.771509399817024e-08,
      "loss": 25.183,
      "step": 19180
    },
    {
      "epoch": 152.30309072781654,
      "grad_norm": 1241.568115234375,
      "learning_rate": 6.750151897675941e-08,
      "loss": 23.1038,
      "step": 19190
    },
    {
      "epoch": 152.38285144566302,
      "grad_norm": 1045.9813232421875,
      "learning_rate": 6.728822871795817e-08,
      "loss": 22.1253,
      "step": 19200
    },
    {
      "epoch": 152.38285144566302,
      "eval_loss": 2.9882588386535645,
      "eval_mae": 1.2969865798950195,
      "eval_mse": 2.9882588386535645,
      "eval_r2": 0.0996522307395935,
      "eval_rmse": 1.728658103458739,
      "eval_runtime": 9.0569,
      "eval_samples_per_second": 442.869,
      "eval_steps_per_second": 13.912,
      "step": 19200
    },
    {
      "epoch": 152.46261216350948,
      "grad_norm": 290.79876708984375,
      "learning_rate": 6.70752235545752e-08,
      "loss": 23.7293,
      "step": 19210
    },
    {
      "epoch": 152.54237288135593,
      "grad_norm": 307.24053955078125,
      "learning_rate": 6.686250381897448e-08,
      "loss": 23.989,
      "step": 19220
    },
    {
      "epoch": 152.62213359920239,
      "grad_norm": 719.582763671875,
      "learning_rate": 6.665006984307467e-08,
      "loss": 22.8119,
      "step": 19230
    },
    {
      "epoch": 152.70189431704884,
      "grad_norm": 954.9166259765625,
      "learning_rate": 6.643792195834813e-08,
      "loss": 23.2559,
      "step": 19240
    },
    {
      "epoch": 152.78165503489532,
      "grad_norm": 157.11846923828125,
      "learning_rate": 6.622606049582122e-08,
      "loss": 25.1447,
      "step": 19250
    },
    {
      "epoch": 152.86141575274178,
      "grad_norm": 376.54119873046875,
      "learning_rate": 6.601448578607336e-08,
      "loss": 24.9452,
      "step": 19260
    },
    {
      "epoch": 152.94117647058823,
      "grad_norm": 445.1767578125,
      "learning_rate": 6.580319815923629e-08,
      "loss": 24.1425,
      "step": 19270
    },
    {
      "epoch": 153.01595214356928,
      "grad_norm": 590.1676025390625,
      "learning_rate": 6.559219794499396e-08,
      "loss": 22.0261,
      "step": 19280
    },
    {
      "epoch": 153.09571286141576,
      "grad_norm": 406.55609130859375,
      "learning_rate": 6.538148547258191e-08,
      "loss": 24.1728,
      "step": 19290
    },
    {
      "epoch": 153.17547357926222,
      "grad_norm": 123.7415771484375,
      "learning_rate": 6.517106107078652e-08,
      "loss": 24.9433,
      "step": 19300
    },
    {
      "epoch": 153.17547357926222,
      "eval_loss": 2.9871866703033447,
      "eval_mae": 1.2974798679351807,
      "eval_mse": 2.9871866703033447,
      "eval_r2": 0.09997528791427612,
      "eval_rmse": 1.7283479598458595,
      "eval_runtime": 9.0817,
      "eval_samples_per_second": 441.655,
      "eval_steps_per_second": 13.874,
      "step": 19300
    },
    {
      "epoch": 153.25523429710867,
      "grad_norm": 321.2386779785156,
      "learning_rate": 6.496092506794487e-08,
      "loss": 24.3306,
      "step": 19310
    },
    {
      "epoch": 153.33499501495513,
      "grad_norm": 681.6568603515625,
      "learning_rate": 6.475107779194392e-08,
      "loss": 26.5892,
      "step": 19320
    },
    {
      "epoch": 153.4147557328016,
      "grad_norm": 405.7754821777344,
      "learning_rate": 6.454151957022005e-08,
      "loss": 22.8697,
      "step": 19330
    },
    {
      "epoch": 153.49451645064806,
      "grad_norm": 145.44203186035156,
      "learning_rate": 6.433225072975884e-08,
      "loss": 23.3981,
      "step": 19340
    },
    {
      "epoch": 153.57427716849452,
      "grad_norm": 438.10137939453125,
      "learning_rate": 6.41232715970941e-08,
      "loss": 23.5633,
      "step": 19350
    },
    {
      "epoch": 153.65403788634097,
      "grad_norm": 933.510009765625,
      "learning_rate": 6.391458249830766e-08,
      "loss": 23.3836,
      "step": 19360
    },
    {
      "epoch": 153.73379860418743,
      "grad_norm": 486.6509094238281,
      "learning_rate": 6.370618375902891e-08,
      "loss": 22.858,
      "step": 19370
    },
    {
      "epoch": 153.8135593220339,
      "grad_norm": 1071.4417724609375,
      "learning_rate": 6.3498075704434e-08,
      "loss": 22.2621,
      "step": 19380
    },
    {
      "epoch": 153.89332003988036,
      "grad_norm": 136.1456298828125,
      "learning_rate": 6.329025865924561e-08,
      "loss": 23.795,
      "step": 19390
    },
    {
      "epoch": 153.97308075772682,
      "grad_norm": 863.8829345703125,
      "learning_rate": 6.30827329477325e-08,
      "loss": 23.1745,
      "step": 19400
    },
    {
      "epoch": 153.97308075772682,
      "eval_loss": 2.988656520843506,
      "eval_mae": 1.3007216453552246,
      "eval_mse": 2.9886562824249268,
      "eval_r2": 0.09953248500823975,
      "eval_rmse": 1.7287730569467257,
      "eval_runtime": 9.0795,
      "eval_samples_per_second": 441.766,
      "eval_steps_per_second": 13.877,
      "step": 19400
    },
    {
      "epoch": 154.04785643070787,
      "grad_norm": 736.192626953125,
      "learning_rate": 6.287549889370836e-08,
      "loss": 22.2077,
      "step": 19410
    },
    {
      "epoch": 154.12761714855435,
      "grad_norm": 219.8618927001953,
      "learning_rate": 6.266855682053226e-08,
      "loss": 24.9694,
      "step": 19420
    },
    {
      "epoch": 154.2073778664008,
      "grad_norm": 583.2711181640625,
      "learning_rate": 6.24619070511076e-08,
      "loss": 27.841,
      "step": 19430
    },
    {
      "epoch": 154.28713858424726,
      "grad_norm": 660.341552734375,
      "learning_rate": 6.225554990788151e-08,
      "loss": 24.4646,
      "step": 19440
    },
    {
      "epoch": 154.3668993020937,
      "grad_norm": 145.15283203125,
      "learning_rate": 6.204948571284455e-08,
      "loss": 23.7078,
      "step": 19450
    },
    {
      "epoch": 154.44666001994017,
      "grad_norm": 456.0956115722656,
      "learning_rate": 6.184371478753034e-08,
      "loss": 22.9063,
      "step": 19460
    },
    {
      "epoch": 154.52642073778665,
      "grad_norm": 570.6836547851562,
      "learning_rate": 6.163823745301466e-08,
      "loss": 23.8765,
      "step": 19470
    },
    {
      "epoch": 154.6061814556331,
      "grad_norm": 139.70950317382812,
      "learning_rate": 6.14330540299155e-08,
      "loss": 23.5889,
      "step": 19480
    },
    {
      "epoch": 154.68594217347956,
      "grad_norm": 708.5938720703125,
      "learning_rate": 6.122816483839194e-08,
      "loss": 23.7718,
      "step": 19490
    },
    {
      "epoch": 154.76570289132601,
      "grad_norm": 1011.361328125,
      "learning_rate": 6.1023570198144e-08,
      "loss": 23.5985,
      "step": 19500
    },
    {
      "epoch": 154.76570289132601,
      "eval_loss": 2.988107442855835,
      "eval_mae": 1.2986818552017212,
      "eval_mse": 2.988107442855835,
      "eval_r2": 0.09969782829284668,
      "eval_rmse": 1.7286143129269278,
      "eval_runtime": 9.0787,
      "eval_samples_per_second": 441.801,
      "eval_steps_per_second": 13.879,
      "step": 19500
    },
    {
      "epoch": 154.84546360917247,
      "grad_norm": 203.33657836914062,
      "learning_rate": 6.081927042841233e-08,
      "loss": 22.9083,
      "step": 19510
    },
    {
      "epoch": 154.92522432701895,
      "grad_norm": 678.0839233398438,
      "learning_rate": 6.061526584797724e-08,
      "loss": 22.9988,
      "step": 19520
    },
    {
      "epoch": 155.0,
      "grad_norm": 339.16741943359375,
      "learning_rate": 6.04115567751584e-08,
      "loss": 20.8858,
      "step": 19530
    },
    {
      "epoch": 155.07976071784645,
      "grad_norm": 198.96624755859375,
      "learning_rate": 6.020814352781469e-08,
      "loss": 23.7149,
      "step": 19540
    },
    {
      "epoch": 155.1595214356929,
      "grad_norm": 102.24024963378906,
      "learning_rate": 6.000502642334305e-08,
      "loss": 23.6419,
      "step": 19550
    },
    {
      "epoch": 155.2392821535394,
      "grad_norm": 699.8232421875,
      "learning_rate": 5.980220577867859e-08,
      "loss": 22.9911,
      "step": 19560
    },
    {
      "epoch": 155.31904287138585,
      "grad_norm": 709.4566650390625,
      "learning_rate": 5.959968191029363e-08,
      "loss": 23.0959,
      "step": 19570
    },
    {
      "epoch": 155.3988035892323,
      "grad_norm": 254.96136474609375,
      "learning_rate": 5.939745513419761e-08,
      "loss": 25.32,
      "step": 19580
    },
    {
      "epoch": 155.47856430707876,
      "grad_norm": 195.00021362304688,
      "learning_rate": 5.9195525765936226e-08,
      "loss": 25.8602,
      "step": 19590
    },
    {
      "epoch": 155.5583250249252,
      "grad_norm": 591.7045288085938,
      "learning_rate": 5.899389412059128e-08,
      "loss": 22.6448,
      "step": 19600
    },
    {
      "epoch": 155.5583250249252,
      "eval_loss": 2.986642360687256,
      "eval_mae": 1.2976558208465576,
      "eval_mse": 2.986642360687256,
      "eval_r2": 0.10013926029205322,
      "eval_rmse": 1.7281904873847835,
      "eval_runtime": 9.0698,
      "eval_samples_per_second": 442.235,
      "eval_steps_per_second": 13.892,
      "step": 19600
    },
    {
      "epoch": 155.6380857427717,
      "grad_norm": 704.7317504882812,
      "learning_rate": 5.8792560512779864e-08,
      "loss": 23.3964,
      "step": 19610
    },
    {
      "epoch": 155.71784646061815,
      "grad_norm": 186.69126892089844,
      "learning_rate": 5.8591525256654064e-08,
      "loss": 23.1593,
      "step": 19620
    },
    {
      "epoch": 155.7976071784646,
      "grad_norm": 120.61994171142578,
      "learning_rate": 5.8390788665900525e-08,
      "loss": 21.7643,
      "step": 19630
    },
    {
      "epoch": 155.87736789631106,
      "grad_norm": 1568.0780029296875,
      "learning_rate": 5.8190351053739816e-08,
      "loss": 25.4865,
      "step": 19640
    },
    {
      "epoch": 155.95712861415754,
      "grad_norm": 371.3355407714844,
      "learning_rate": 5.7990212732925843e-08,
      "loss": 24.1773,
      "step": 19650
    },
    {
      "epoch": 156.0319042871386,
      "grad_norm": 231.3843231201172,
      "learning_rate": 5.779037401574582e-08,
      "loss": 23.05,
      "step": 19660
    },
    {
      "epoch": 156.11166500498504,
      "grad_norm": 308.94384765625,
      "learning_rate": 5.759083521401914e-08,
      "loss": 25.9338,
      "step": 19670
    },
    {
      "epoch": 156.1914257228315,
      "grad_norm": 1007.2675170898438,
      "learning_rate": 5.739159663909754e-08,
      "loss": 22.6109,
      "step": 19680
    },
    {
      "epoch": 156.27118644067798,
      "grad_norm": 337.06707763671875,
      "learning_rate": 5.719265860186406e-08,
      "loss": 24.2056,
      "step": 19690
    },
    {
      "epoch": 156.35094715852443,
      "grad_norm": 229.7677001953125,
      "learning_rate": 5.699402141273277e-08,
      "loss": 24.7657,
      "step": 19700
    },
    {
      "epoch": 156.35094715852443,
      "eval_loss": 2.9869258403778076,
      "eval_mae": 1.298068642616272,
      "eval_mse": 2.9869258403778076,
      "eval_r2": 0.10005384683609009,
      "eval_rmse": 1.7282725017710048,
      "eval_runtime": 9.0728,
      "eval_samples_per_second": 442.092,
      "eval_steps_per_second": 13.888,
      "step": 19700
    },
    {
      "epoch": 156.4307078763709,
      "grad_norm": 316.57904052734375,
      "learning_rate": 5.6795685381648525e-08,
      "loss": 23.1404,
      "step": 19710
    },
    {
      "epoch": 156.51046859421734,
      "grad_norm": 271.0904235839844,
      "learning_rate": 5.659765081808629e-08,
      "loss": 24.78,
      "step": 19720
    },
    {
      "epoch": 156.5902293120638,
      "grad_norm": 284.07196044921875,
      "learning_rate": 5.639991803105018e-08,
      "loss": 22.9677,
      "step": 19730
    },
    {
      "epoch": 156.66999002991028,
      "grad_norm": 276.0367736816406,
      "learning_rate": 5.62024873290739e-08,
      "loss": 22.588,
      "step": 19740
    },
    {
      "epoch": 156.74975074775674,
      "grad_norm": 96.85620880126953,
      "learning_rate": 5.6005359020219745e-08,
      "loss": 23.1715,
      "step": 19750
    },
    {
      "epoch": 156.8295114656032,
      "grad_norm": 719.615478515625,
      "learning_rate": 5.580853341207795e-08,
      "loss": 23.5608,
      "step": 19760
    },
    {
      "epoch": 156.90927218344964,
      "grad_norm": 189.48875427246094,
      "learning_rate": 5.561201081176648e-08,
      "loss": 23.8051,
      "step": 19770
    },
    {
      "epoch": 156.9890329012961,
      "grad_norm": 208.35182189941406,
      "learning_rate": 5.541579152593073e-08,
      "loss": 24.8919,
      "step": 19780
    },
    {
      "epoch": 157.06380857427718,
      "grad_norm": 410.5036926269531,
      "learning_rate": 5.521987586074248e-08,
      "loss": 20.6443,
      "step": 19790
    },
    {
      "epoch": 157.14356929212363,
      "grad_norm": 713.73095703125,
      "learning_rate": 5.5024264121900087e-08,
      "loss": 24.4717,
      "step": 19800
    },
    {
      "epoch": 157.14356929212363,
      "eval_loss": 2.9873814582824707,
      "eval_mae": 1.2965357303619385,
      "eval_mse": 2.9873812198638916,
      "eval_r2": 0.09991669654846191,
      "eval_rmse": 1.7284042408718776,
      "eval_runtime": 9.072,
      "eval_samples_per_second": 442.13,
      "eval_steps_per_second": 13.889,
      "step": 19800
    },
    {
      "epoch": 157.22333000997008,
      "grad_norm": 355.75177001953125,
      "learning_rate": 5.482895661462744e-08,
      "loss": 25.0636,
      "step": 19810
    },
    {
      "epoch": 157.30309072781654,
      "grad_norm": 200.91207885742188,
      "learning_rate": 5.463395364367371e-08,
      "loss": 23.233,
      "step": 19820
    },
    {
      "epoch": 157.38285144566302,
      "grad_norm": 839.2678833007812,
      "learning_rate": 5.4439255513313104e-08,
      "loss": 25.1534,
      "step": 19830
    },
    {
      "epoch": 157.46261216350948,
      "grad_norm": 794.5970458984375,
      "learning_rate": 5.424486252734395e-08,
      "loss": 23.7456,
      "step": 19840
    },
    {
      "epoch": 157.54237288135593,
      "grad_norm": 797.7697143554688,
      "learning_rate": 5.405077498908847e-08,
      "loss": 22.2091,
      "step": 19850
    },
    {
      "epoch": 157.62213359920239,
      "grad_norm": 429.3815002441406,
      "learning_rate": 5.385699320139248e-08,
      "loss": 24.0605,
      "step": 19860
    },
    {
      "epoch": 157.70189431704884,
      "grad_norm": 488.4406433105469,
      "learning_rate": 5.36635174666244e-08,
      "loss": 22.0634,
      "step": 19870
    },
    {
      "epoch": 157.78165503489532,
      "grad_norm": 588.324951171875,
      "learning_rate": 5.3470348086675386e-08,
      "loss": 26.5165,
      "step": 19880
    },
    {
      "epoch": 157.86141575274178,
      "grad_norm": 530.62646484375,
      "learning_rate": 5.327748536295842e-08,
      "loss": 22.664,
      "step": 19890
    },
    {
      "epoch": 157.94117647058823,
      "grad_norm": 289.956787109375,
      "learning_rate": 5.308492959640795e-08,
      "loss": 23.8528,
      "step": 19900
    },
    {
      "epoch": 157.94117647058823,
      "eval_loss": 2.9881298542022705,
      "eval_mae": 1.2951724529266357,
      "eval_mse": 2.9881298542022705,
      "eval_r2": 0.09969109296798706,
      "eval_rmse": 1.7286207953748187,
      "eval_runtime": 9.0883,
      "eval_samples_per_second": 441.338,
      "eval_steps_per_second": 13.864,
      "step": 19900
    },
    {
      "epoch": 158.01595214356928,
      "grad_norm": 237.19174194335938,
      "learning_rate": 5.2892681087479574e-08,
      "loss": 22.773,
      "step": 19910
    },
    {
      "epoch": 158.09571286141576,
      "grad_norm": 274.02972412109375,
      "learning_rate": 5.2700740136149554e-08,
      "loss": 23.108,
      "step": 19920
    },
    {
      "epoch": 158.17547357926222,
      "grad_norm": 458.33648681640625,
      "learning_rate": 5.250910704191386e-08,
      "loss": 26.5027,
      "step": 19930
    },
    {
      "epoch": 158.25523429710867,
      "grad_norm": 227.19955444335938,
      "learning_rate": 5.2317782103788485e-08,
      "loss": 24.2449,
      "step": 19940
    },
    {
      "epoch": 158.33499501495513,
      "grad_norm": 455.753173828125,
      "learning_rate": 5.2126765620308514e-08,
      "loss": 23.0352,
      "step": 19950
    },
    {
      "epoch": 158.4147557328016,
      "grad_norm": 636.4472045898438,
      "learning_rate": 5.193605788952754e-08,
      "loss": 23.5401,
      "step": 19960
    },
    {
      "epoch": 158.49451645064806,
      "grad_norm": 778.9373779296875,
      "learning_rate": 5.174565920901766e-08,
      "loss": 24.3629,
      "step": 19970
    },
    {
      "epoch": 158.57427716849452,
      "grad_norm": 329.8791809082031,
      "learning_rate": 5.155556987586856e-08,
      "loss": 24.5927,
      "step": 19980
    },
    {
      "epoch": 158.65403788634097,
      "grad_norm": 900.1206665039062,
      "learning_rate": 5.1365790186687184e-08,
      "loss": 24.1182,
      "step": 19990
    },
    {
      "epoch": 158.73379860418743,
      "grad_norm": 1131.921875,
      "learning_rate": 5.1176320437597566e-08,
      "loss": 23.8033,
      "step": 20000
    },
    {
      "epoch": 158.73379860418743,
      "eval_loss": 2.988154172897339,
      "eval_mae": 1.2970765829086304,
      "eval_mse": 2.9881539344787598,
      "eval_r2": 0.09968382120132446,
      "eval_rmse": 1.7286277605310982,
      "eval_runtime": 9.147,
      "eval_samples_per_second": 438.505,
      "eval_steps_per_second": 13.775,
      "step": 20000
    },
    {
      "epoch": 158.8135593220339,
      "grad_norm": 357.07879638671875,
      "learning_rate": 5.098716092423991e-08,
      "loss": 24.5281,
      "step": 20010
    },
    {
      "epoch": 158.89332003988036,
      "grad_norm": 219.1493377685547,
      "learning_rate": 5.079831194177034e-08,
      "loss": 22.6291,
      "step": 20020
    },
    {
      "epoch": 158.97308075772682,
      "grad_norm": 915.48046875,
      "learning_rate": 5.060977378486064e-08,
      "loss": 22.1652,
      "step": 20030
    },
    {
      "epoch": 159.04785643070787,
      "grad_norm": 368.95794677734375,
      "learning_rate": 5.042154674769736e-08,
      "loss": 21.4069,
      "step": 20040
    },
    {
      "epoch": 159.12761714855435,
      "grad_norm": 542.30810546875,
      "learning_rate": 5.023363112398177e-08,
      "loss": 24.379,
      "step": 20050
    },
    {
      "epoch": 159.2073778664008,
      "grad_norm": 768.3283081054688,
      "learning_rate": 5.0046027206929046e-08,
      "loss": 25.799,
      "step": 20060
    },
    {
      "epoch": 159.28713858424726,
      "grad_norm": 226.7152862548828,
      "learning_rate": 4.985873528926826e-08,
      "loss": 24.935,
      "step": 20070
    },
    {
      "epoch": 159.3668993020937,
      "grad_norm": 565.6937866210938,
      "learning_rate": 4.9671755663241304e-08,
      "loss": 22.7557,
      "step": 20080
    },
    {
      "epoch": 159.44666001994017,
      "grad_norm": 238.96612548828125,
      "learning_rate": 4.948508862060316e-08,
      "loss": 23.0391,
      "step": 20090
    },
    {
      "epoch": 159.52642073778665,
      "grad_norm": 343.6816711425781,
      "learning_rate": 4.9298734452620755e-08,
      "loss": 24.5896,
      "step": 20100
    },
    {
      "epoch": 159.52642073778665,
      "eval_loss": 2.989109516143799,
      "eval_mae": 1.294765830039978,
      "eval_mse": 2.989109516143799,
      "eval_r2": 0.09939593076705933,
      "eval_rmse": 1.7289041373493785,
      "eval_runtime": 9.0579,
      "eval_samples_per_second": 442.817,
      "eval_steps_per_second": 13.91,
      "step": 20100
    },
    {
      "epoch": 159.6061814556331,
      "grad_norm": 803.0233764648438,
      "learning_rate": 4.9112693450072905e-08,
      "loss": 22.7036,
      "step": 20110
    },
    {
      "epoch": 159.68594217347956,
      "grad_norm": 220.58651733398438,
      "learning_rate": 4.892696590324996e-08,
      "loss": 25.0648,
      "step": 20120
    },
    {
      "epoch": 159.76570289132601,
      "grad_norm": 197.690673828125,
      "learning_rate": 4.874155210195291e-08,
      "loss": 24.4888,
      "step": 20130
    },
    {
      "epoch": 159.84546360917247,
      "grad_norm": 403.8138732910156,
      "learning_rate": 4.85564523354933e-08,
      "loss": 21.8955,
      "step": 20140
    },
    {
      "epoch": 159.92522432701895,
      "grad_norm": 137.9730224609375,
      "learning_rate": 4.837166689269276e-08,
      "loss": 22.8665,
      "step": 20150
    },
    {
      "epoch": 160.0,
      "grad_norm": 173.60696411132812,
      "learning_rate": 4.818719606188223e-08,
      "loss": 22.3084,
      "step": 20160
    },
    {
      "epoch": 160.07976071784645,
      "grad_norm": 669.76025390625,
      "learning_rate": 4.800304013090206e-08,
      "loss": 24.1283,
      "step": 20170
    },
    {
      "epoch": 160.1595214356929,
      "grad_norm": 423.6705017089844,
      "learning_rate": 4.7819199387100966e-08,
      "loss": 25.3707,
      "step": 20180
    },
    {
      "epoch": 160.2392821535394,
      "grad_norm": 1109.6336669921875,
      "learning_rate": 4.763567411733593e-08,
      "loss": 23.5949,
      "step": 20190
    },
    {
      "epoch": 160.31904287138585,
      "grad_norm": 775.6715087890625,
      "learning_rate": 4.7452464607971784e-08,
      "loss": 23.3331,
      "step": 20200
    },
    {
      "epoch": 160.31904287138585,
      "eval_loss": 2.9872071743011475,
      "eval_mae": 1.2983434200286865,
      "eval_mse": 2.9872071743011475,
      "eval_r2": 0.09996908903121948,
      "eval_rmse": 1.7283538915109797,
      "eval_runtime": 9.0827,
      "eval_samples_per_second": 441.609,
      "eval_steps_per_second": 13.873,
      "step": 20200
    },
    {
      "epoch": 160.3988035892323,
      "grad_norm": 564.7343139648438,
      "learning_rate": 4.726957114488073e-08,
      "loss": 22.8999,
      "step": 20210
    },
    {
      "epoch": 160.47856430707876,
      "grad_norm": 414.7461853027344,
      "learning_rate": 4.7086994013441375e-08,
      "loss": 23.5632,
      "step": 20220
    },
    {
      "epoch": 160.5583250249252,
      "grad_norm": 993.0753173828125,
      "learning_rate": 4.690473349853924e-08,
      "loss": 25.4307,
      "step": 20230
    },
    {
      "epoch": 160.6380857427717,
      "grad_norm": 496.5462646484375,
      "learning_rate": 4.672278988456568e-08,
      "loss": 23.8259,
      "step": 20240
    },
    {
      "epoch": 160.71784646061815,
      "grad_norm": 124.31321716308594,
      "learning_rate": 4.654116345541742e-08,
      "loss": 25.5149,
      "step": 20250
    },
    {
      "epoch": 160.7976071784646,
      "grad_norm": 247.69906616210938,
      "learning_rate": 4.635985449449631e-08,
      "loss": 21.4839,
      "step": 20260
    },
    {
      "epoch": 160.87736789631106,
      "grad_norm": 826.633056640625,
      "learning_rate": 4.617886328470902e-08,
      "loss": 23.509,
      "step": 20270
    },
    {
      "epoch": 160.95712861415754,
      "grad_norm": 447.5196228027344,
      "learning_rate": 4.599819010846614e-08,
      "loss": 22.9887,
      "step": 20280
    },
    {
      "epoch": 161.0319042871386,
      "grad_norm": 264.0032958984375,
      "learning_rate": 4.581783524768232e-08,
      "loss": 23.0958,
      "step": 20290
    },
    {
      "epoch": 161.11166500498504,
      "grad_norm": 990.3718872070312,
      "learning_rate": 4.5637798983775236e-08,
      "loss": 23.5094,
      "step": 20300
    },
    {
      "epoch": 161.11166500498504,
      "eval_loss": 2.988409996032715,
      "eval_mae": 1.3004189729690552,
      "eval_mse": 2.9884097576141357,
      "eval_r2": 0.09960675239562988,
      "eval_rmse": 1.7287017549635726,
      "eval_runtime": 9.0987,
      "eval_samples_per_second": 440.83,
      "eval_steps_per_second": 13.848,
      "step": 20300
    },
    {
      "epoch": 161.1914257228315,
      "grad_norm": 395.7349853515625,
      "learning_rate": 4.545808159766554e-08,
      "loss": 22.8364,
      "step": 20310
    },
    {
      "epoch": 161.27118644067798,
      "grad_norm": 499.4708251953125,
      "learning_rate": 4.527868336977644e-08,
      "loss": 23.2073,
      "step": 20320
    },
    {
      "epoch": 161.35094715852443,
      "grad_norm": 238.495849609375,
      "learning_rate": 4.509960458003301e-08,
      "loss": 23.3997,
      "step": 20330
    },
    {
      "epoch": 161.4307078763709,
      "grad_norm": 423.6451110839844,
      "learning_rate": 4.4920845507861847e-08,
      "loss": 23.6306,
      "step": 20340
    },
    {
      "epoch": 161.51046859421734,
      "grad_norm": 320.92340087890625,
      "learning_rate": 4.4742406432190873e-08,
      "loss": 26.0693,
      "step": 20350
    },
    {
      "epoch": 161.5902293120638,
      "grad_norm": 530.1223754882812,
      "learning_rate": 4.456428763144848e-08,
      "loss": 22.0599,
      "step": 20360
    },
    {
      "epoch": 161.66999002991028,
      "grad_norm": 875.4921264648438,
      "learning_rate": 4.4386489383563516e-08,
      "loss": 23.7559,
      "step": 20370
    },
    {
      "epoch": 161.74975074775674,
      "grad_norm": 733.9068603515625,
      "learning_rate": 4.420901196596452e-08,
      "loss": 23.7832,
      "step": 20380
    },
    {
      "epoch": 161.8295114656032,
      "grad_norm": 239.6742401123047,
      "learning_rate": 4.4031855655579344e-08,
      "loss": 24.1239,
      "step": 20390
    },
    {
      "epoch": 161.90927218344964,
      "grad_norm": 482.89874267578125,
      "learning_rate": 4.385502072883501e-08,
      "loss": 23.7691,
      "step": 20400
    },
    {
      "epoch": 161.90927218344964,
      "eval_loss": 2.986546754837036,
      "eval_mae": 1.2973977327346802,
      "eval_mse": 2.986546754837036,
      "eval_r2": 0.10016804933547974,
      "eval_rmse": 1.7281628264828046,
      "eval_runtime": 9.1006,
      "eval_samples_per_second": 440.741,
      "eval_steps_per_second": 13.845,
      "step": 20400
    },
    {
      "epoch": 161.9890329012961,
      "grad_norm": 102.24373626708984,
      "learning_rate": 4.367850746165708e-08,
      "loss": 24.6804,
      "step": 20410
    },
    {
      "epoch": 162.06380857427718,
      "grad_norm": 172.08877563476562,
      "learning_rate": 4.3502316129468837e-08,
      "loss": 20.5014,
      "step": 20420
    },
    {
      "epoch": 162.14356929212363,
      "grad_norm": 229.16693115234375,
      "learning_rate": 4.3326447007191654e-08,
      "loss": 22.8511,
      "step": 20430
    },
    {
      "epoch": 162.22333000997008,
      "grad_norm": 334.28741455078125,
      "learning_rate": 4.315090036924402e-08,
      "loss": 23.8031,
      "step": 20440
    },
    {
      "epoch": 162.30309072781654,
      "grad_norm": 113.47332000732422,
      "learning_rate": 4.2975676489541094e-08,
      "loss": 23.2761,
      "step": 20450
    },
    {
      "epoch": 162.38285144566302,
      "grad_norm": 1696.001708984375,
      "learning_rate": 4.280077564149454e-08,
      "loss": 23.0712,
      "step": 20460
    },
    {
      "epoch": 162.46261216350948,
      "grad_norm": 429.3016357421875,
      "learning_rate": 4.2626198098011986e-08,
      "loss": 26.6888,
      "step": 20470
    },
    {
      "epoch": 162.54237288135593,
      "grad_norm": 127.84202575683594,
      "learning_rate": 4.245194413149653e-08,
      "loss": 23.5264,
      "step": 20480
    },
    {
      "epoch": 162.62213359920239,
      "grad_norm": 718.2103881835938,
      "learning_rate": 4.227801401384642e-08,
      "loss": 24.2782,
      "step": 20490
    },
    {
      "epoch": 162.70189431704884,
      "grad_norm": 98.80939483642578,
      "learning_rate": 4.210440801645462e-08,
      "loss": 22.7801,
      "step": 20500
    },
    {
      "epoch": 162.70189431704884,
      "eval_loss": 2.986672878265381,
      "eval_mae": 1.296818733215332,
      "eval_mse": 2.98667311668396,
      "eval_r2": 0.10012996196746826,
      "eval_rmse": 1.7281993856855637,
      "eval_runtime": 9.0883,
      "eval_samples_per_second": 441.338,
      "eval_steps_per_second": 13.864,
      "step": 20500
    },
    {
      "epoch": 162.78165503489532,
      "grad_norm": 710.266845703125,
      "learning_rate": 4.193112641020816e-08,
      "loss": 26.5156,
      "step": 20510
    },
    {
      "epoch": 162.86141575274178,
      "grad_norm": 599.1068115234375,
      "learning_rate": 4.175816946548819e-08,
      "loss": 24.2544,
      "step": 20520
    },
    {
      "epoch": 162.94117647058823,
      "grad_norm": 640.97021484375,
      "learning_rate": 4.158553745216903e-08,
      "loss": 21.5817,
      "step": 20530
    },
    {
      "epoch": 163.01595214356928,
      "grad_norm": 779.9424438476562,
      "learning_rate": 4.14132306396181e-08,
      "loss": 24.3248,
      "step": 20540
    },
    {
      "epoch": 163.09571286141576,
      "grad_norm": 733.3753051757812,
      "learning_rate": 4.124124929669537e-08,
      "loss": 25.5774,
      "step": 20550
    },
    {
      "epoch": 163.17547357926222,
      "grad_norm": 497.6435852050781,
      "learning_rate": 4.1069593691753057e-08,
      "loss": 22.9471,
      "step": 20560
    },
    {
      "epoch": 163.25523429710867,
      "grad_norm": 440.9189453125,
      "learning_rate": 4.0898264092634916e-08,
      "loss": 24.5098,
      "step": 20570
    },
    {
      "epoch": 163.33499501495513,
      "grad_norm": 234.5493621826172,
      "learning_rate": 4.0727260766676237e-08,
      "loss": 24.8129,
      "step": 20580
    },
    {
      "epoch": 163.4147557328016,
      "grad_norm": 147.0826873779297,
      "learning_rate": 4.0556583980703025e-08,
      "loss": 23.6257,
      "step": 20590
    },
    {
      "epoch": 163.49451645064806,
      "grad_norm": 217.13864135742188,
      "learning_rate": 4.0386234001031776e-08,
      "loss": 23.653,
      "step": 20600
    },
    {
      "epoch": 163.49451645064806,
      "eval_loss": 2.986043691635132,
      "eval_mae": 1.299048662185669,
      "eval_mse": 2.986043930053711,
      "eval_r2": 0.10031956434249878,
      "eval_rmse": 1.728017340785014,
      "eval_runtime": 9.0988,
      "eval_samples_per_second": 440.828,
      "eval_steps_per_second": 13.848,
      "step": 20600
    },
    {
      "epoch": 163.57427716849452,
      "grad_norm": 666.421630859375,
      "learning_rate": 4.021621109346923e-08,
      "loss": 21.0761,
      "step": 20610
    },
    {
      "epoch": 163.65403788634097,
      "grad_norm": 378.9775390625,
      "learning_rate": 4.0046515523311595e-08,
      "loss": 24.4574,
      "step": 20620
    },
    {
      "epoch": 163.73379860418743,
      "grad_norm": 370.6648864746094,
      "learning_rate": 3.987714755534435e-08,
      "loss": 22.2913,
      "step": 20630
    },
    {
      "epoch": 163.8135593220339,
      "grad_norm": 836.6725463867188,
      "learning_rate": 3.9708107453841926e-08,
      "loss": 26.003,
      "step": 20640
    },
    {
      "epoch": 163.89332003988036,
      "grad_norm": 203.53436279296875,
      "learning_rate": 3.953939548256696e-08,
      "loss": 24.5101,
      "step": 20650
    },
    {
      "epoch": 163.97308075772682,
      "grad_norm": 131.53392028808594,
      "learning_rate": 3.9371011904770295e-08,
      "loss": 22.3992,
      "step": 20660
    },
    {
      "epoch": 164.04785643070787,
      "grad_norm": 511.9137268066406,
      "learning_rate": 3.9202956983190234e-08,
      "loss": 21.8256,
      "step": 20670
    },
    {
      "epoch": 164.12761714855435,
      "grad_norm": 560.97900390625,
      "learning_rate": 3.903523098005221e-08,
      "loss": 24.6533,
      "step": 20680
    },
    {
      "epoch": 164.2073778664008,
      "grad_norm": 486.2814636230469,
      "learning_rate": 3.8867834157068596e-08,
      "loss": 24.5964,
      "step": 20690
    },
    {
      "epoch": 164.28713858424726,
      "grad_norm": 784.7884521484375,
      "learning_rate": 3.8700766775438124e-08,
      "loss": 23.9616,
      "step": 20700
    },
    {
      "epoch": 164.28713858424726,
      "eval_loss": 2.9880242347717285,
      "eval_mae": 1.2942732572555542,
      "eval_mse": 2.9880242347717285,
      "eval_r2": 0.09972292184829712,
      "eval_rmse": 1.7285902449023969,
      "eval_runtime": 9.0853,
      "eval_samples_per_second": 441.481,
      "eval_steps_per_second": 13.869,
      "step": 20700
    },
    {
      "epoch": 164.3668993020937,
      "grad_norm": 169.08421325683594,
      "learning_rate": 3.853402909584516e-08,
      "loss": 23.0227,
      "step": 20710
    },
    {
      "epoch": 164.44666001994017,
      "grad_norm": 1243.542236328125,
      "learning_rate": 3.8367621378460003e-08,
      "loss": 21.9406,
      "step": 20720
    },
    {
      "epoch": 164.52642073778665,
      "grad_norm": 575.0645751953125,
      "learning_rate": 3.820154388293795e-08,
      "loss": 23.3995,
      "step": 20730
    },
    {
      "epoch": 164.6061814556331,
      "grad_norm": 333.740966796875,
      "learning_rate": 3.803579686841901e-08,
      "loss": 23.2316,
      "step": 20740
    },
    {
      "epoch": 164.68594217347956,
      "grad_norm": 812.60986328125,
      "learning_rate": 3.7870380593527484e-08,
      "loss": 24.5726,
      "step": 20750
    },
    {
      "epoch": 164.76570289132601,
      "grad_norm": 127.72270202636719,
      "learning_rate": 3.770529531637173e-08,
      "loss": 25.9085,
      "step": 20760
    },
    {
      "epoch": 164.84546360917247,
      "grad_norm": 478.8529052734375,
      "learning_rate": 3.754054129454348e-08,
      "loss": 24.0288,
      "step": 20770
    },
    {
      "epoch": 164.92522432701895,
      "grad_norm": 464.51904296875,
      "learning_rate": 3.737611878511779e-08,
      "loss": 23.2809,
      "step": 20780
    },
    {
      "epoch": 165.0,
      "grad_norm": 463.0782775878906,
      "learning_rate": 3.7212028044652233e-08,
      "loss": 21.8776,
      "step": 20790
    },
    {
      "epoch": 165.07976071784645,
      "grad_norm": 1210.877685546875,
      "learning_rate": 3.7048269329186754e-08,
      "loss": 25.0118,
      "step": 20800
    },
    {
      "epoch": 165.07976071784645,
      "eval_loss": 2.9878416061401367,
      "eval_mae": 1.2945494651794434,
      "eval_mse": 2.9878413677215576,
      "eval_r2": 0.09977799654006958,
      "eval_rmse": 1.7285373492411316,
      "eval_runtime": 9.1111,
      "eval_samples_per_second": 440.233,
      "eval_steps_per_second": 13.829,
      "step": 20800
    },
    {
      "epoch": 165.1595214356929,
      "grad_norm": 495.2569274902344,
      "learning_rate": 3.688484289424337e-08,
      "loss": 22.6857,
      "step": 20810
    },
    {
      "epoch": 165.2392821535394,
      "grad_norm": 192.6989288330078,
      "learning_rate": 3.672174899482544e-08,
      "loss": 23.7292,
      "step": 20820
    },
    {
      "epoch": 165.31904287138585,
      "grad_norm": 655.8668212890625,
      "learning_rate": 3.6558987885417445e-08,
      "loss": 22.3309,
      "step": 20830
    },
    {
      "epoch": 165.3988035892323,
      "grad_norm": 940.3027954101562,
      "learning_rate": 3.639655981998482e-08,
      "loss": 23.7985,
      "step": 20840
    },
    {
      "epoch": 165.47856430707876,
      "grad_norm": 500.3050231933594,
      "learning_rate": 3.623446505197306e-08,
      "loss": 23.809,
      "step": 20850
    },
    {
      "epoch": 165.5583250249252,
      "grad_norm": 570.357666015625,
      "learning_rate": 3.607270383430777e-08,
      "loss": 24.2061,
      "step": 20860
    },
    {
      "epoch": 165.6380857427717,
      "grad_norm": 756.9074096679688,
      "learning_rate": 3.592740413257877e-08,
      "loss": 23.6694,
      "step": 20870
    },
    {
      "epoch": 165.71784646061815,
      "grad_norm": 497.1473388671875,
      "learning_rate": 3.5766277355517347e-08,
      "loss": 23.7223,
      "step": 20880
    },
    {
      "epoch": 165.7976071784646,
      "grad_norm": 224.22097778320312,
      "learning_rate": 3.5605484859341893e-08,
      "loss": 24.6421,
      "step": 20890
    },
    {
      "epoch": 165.87736789631106,
      "grad_norm": 277.3652038574219,
      "learning_rate": 3.544502689494588e-08,
      "loss": 22.7341,
      "step": 20900
    },
    {
      "epoch": 165.87736789631106,
      "eval_loss": 2.987583875656128,
      "eval_mae": 1.2943426370620728,
      "eval_mse": 2.987583875656128,
      "eval_r2": 0.09985554218292236,
      "eval_rmse": 1.728462864991935,
      "eval_runtime": 9.0908,
      "eval_samples_per_second": 441.213,
      "eval_steps_per_second": 13.86,
      "step": 20900
    },
    {
      "epoch": 165.95712861415754,
      "grad_norm": 834.4371337890625,
      "learning_rate": 3.528490371270101e-08,
      "loss": 25.1784,
      "step": 20910
    },
    {
      "epoch": 166.0319042871386,
      "grad_norm": 747.2515258789062,
      "learning_rate": 3.512511556245631e-08,
      "loss": 23.1242,
      "step": 20920
    },
    {
      "epoch": 166.11166500498504,
      "grad_norm": 206.32666015625,
      "learning_rate": 3.496566269353832e-08,
      "loss": 23.3442,
      "step": 20930
    },
    {
      "epoch": 166.1914257228315,
      "grad_norm": 1007.1183471679688,
      "learning_rate": 3.480654535475025e-08,
      "loss": 21.7909,
      "step": 20940
    },
    {
      "epoch": 166.27118644067798,
      "grad_norm": 145.24815368652344,
      "learning_rate": 3.4647763794371705e-08,
      "loss": 25.2308,
      "step": 20950
    },
    {
      "epoch": 166.35094715852443,
      "grad_norm": 725.6841430664062,
      "learning_rate": 3.448931826015858e-08,
      "loss": 26.5595,
      "step": 20960
    },
    {
      "epoch": 166.4307078763709,
      "grad_norm": 207.36798095703125,
      "learning_rate": 3.43312089993423e-08,
      "loss": 22.0913,
      "step": 20970
    },
    {
      "epoch": 166.51046859421734,
      "grad_norm": 713.8314819335938,
      "learning_rate": 3.417343625862954e-08,
      "loss": 23.4224,
      "step": 20980
    },
    {
      "epoch": 166.5902293120638,
      "grad_norm": 121.8656234741211,
      "learning_rate": 3.401600028420201e-08,
      "loss": 24.7084,
      "step": 20990
    },
    {
      "epoch": 166.66999002991028,
      "grad_norm": 920.5518798828125,
      "learning_rate": 3.385890132171584e-08,
      "loss": 23.9545,
      "step": 21000
    },
    {
      "epoch": 166.66999002991028,
      "eval_loss": 2.9885149002075195,
      "eval_mae": 1.293936848640442,
      "eval_mse": 2.9885149002075195,
      "eval_r2": 0.09957504272460938,
      "eval_rmse": 1.7287321655500945,
      "eval_runtime": 9.0843,
      "eval_samples_per_second": 441.53,
      "eval_steps_per_second": 13.87,
      "step": 21000
    },
    {
      "epoch": 166.74975074775674,
      "grad_norm": 615.0982666015625,
      "learning_rate": 3.370213961630142e-08,
      "loss": 25.1141,
      "step": 21010
    },
    {
      "epoch": 166.8295114656032,
      "grad_norm": 381.2695617675781,
      "learning_rate": 3.354571541256279e-08,
      "loss": 21.3079,
      "step": 21020
    },
    {
      "epoch": 166.90927218344964,
      "grad_norm": 261.74169921875,
      "learning_rate": 3.3389628954577336e-08,
      "loss": 22.5619,
      "step": 21030
    },
    {
      "epoch": 166.9890329012961,
      "grad_norm": 1052.0252685546875,
      "learning_rate": 3.323388048589559e-08,
      "loss": 25.1966,
      "step": 21040
    },
    {
      "epoch": 167.06380857427718,
      "grad_norm": 293.083984375,
      "learning_rate": 3.307847024954069e-08,
      "loss": 23.9578,
      "step": 21050
    },
    {
      "epoch": 167.14356929212363,
      "grad_norm": 624.7679443359375,
      "learning_rate": 3.2923398488007733e-08,
      "loss": 24.2935,
      "step": 21060
    },
    {
      "epoch": 167.22333000997008,
      "grad_norm": 598.4464111328125,
      "learning_rate": 3.2768665443263984e-08,
      "loss": 23.0892,
      "step": 21070
    },
    {
      "epoch": 167.30309072781654,
      "grad_norm": 801.3358154296875,
      "learning_rate": 3.2614271356748156e-08,
      "loss": 23.4058,
      "step": 21080
    },
    {
      "epoch": 167.38285144566302,
      "grad_norm": 103.62716674804688,
      "learning_rate": 3.2460216469369944e-08,
      "loss": 23.3563,
      "step": 21090
    },
    {
      "epoch": 167.46261216350948,
      "grad_norm": 403.87445068359375,
      "learning_rate": 3.230650102150975e-08,
      "loss": 22.5184,
      "step": 21100
    },
    {
      "epoch": 167.46261216350948,
      "eval_loss": 2.9855480194091797,
      "eval_mae": 1.2961645126342773,
      "eval_mse": 2.9855480194091797,
      "eval_r2": 0.10046899318695068,
      "eval_rmse": 1.7278738436035137,
      "eval_runtime": 9.0721,
      "eval_samples_per_second": 442.125,
      "eval_steps_per_second": 13.889,
      "step": 21100
    },
    {
      "epoch": 167.54237288135593,
      "grad_norm": 494.8824157714844,
      "learning_rate": 3.215312525301853e-08,
      "loss": 22.9951,
      "step": 21110
    },
    {
      "epoch": 167.62213359920239,
      "grad_norm": 809.4344482421875,
      "learning_rate": 3.2000089403216966e-08,
      "loss": 23.5586,
      "step": 21120
    },
    {
      "epoch": 167.70189431704884,
      "grad_norm": 136.36012268066406,
      "learning_rate": 3.184739371089562e-08,
      "loss": 24.2447,
      "step": 21130
    },
    {
      "epoch": 167.78165503489532,
      "grad_norm": 115.35761260986328,
      "learning_rate": 3.1695038414314036e-08,
      "loss": 23.0265,
      "step": 21140
    },
    {
      "epoch": 167.86141575274178,
      "grad_norm": 160.2898712158203,
      "learning_rate": 3.154302375120069e-08,
      "loss": 25.8412,
      "step": 21150
    },
    {
      "epoch": 167.94117647058823,
      "grad_norm": 358.9959716796875,
      "learning_rate": 3.1391349958752644e-08,
      "loss": 23.6602,
      "step": 21160
    },
    {
      "epoch": 168.01595214356928,
      "grad_norm": 1159.0592041015625,
      "learning_rate": 3.124001727363512e-08,
      "loss": 21.6542,
      "step": 21170
    },
    {
      "epoch": 168.09571286141576,
      "grad_norm": 204.9313201904297,
      "learning_rate": 3.108902593198076e-08,
      "loss": 21.4206,
      "step": 21180
    },
    {
      "epoch": 168.17547357926222,
      "grad_norm": 379.1077880859375,
      "learning_rate": 3.093837616938996e-08,
      "loss": 22.0864,
      "step": 21190
    },
    {
      "epoch": 168.25523429710867,
      "grad_norm": 534.8294067382812,
      "learning_rate": 3.0788068220929995e-08,
      "loss": 24.5088,
      "step": 21200
    },
    {
      "epoch": 168.25523429710867,
      "eval_loss": 2.9861366748809814,
      "eval_mae": 1.297932505607605,
      "eval_mse": 2.9861366748809814,
      "eval_r2": 0.10029160976409912,
      "eval_rmse": 1.7280441761948626,
      "eval_runtime": 9.0903,
      "eval_samples_per_second": 441.239,
      "eval_steps_per_second": 13.861,
      "step": 21200
    },
    {
      "epoch": 168.33499501495513,
      "grad_norm": 349.81402587890625,
      "learning_rate": 3.0638102321134786e-08,
      "loss": 23.7894,
      "step": 21210
    },
    {
      "epoch": 168.4147557328016,
      "grad_norm": 262.42547607421875,
      "learning_rate": 3.048847870400442e-08,
      "loss": 26.1303,
      "step": 21220
    },
    {
      "epoch": 168.49451645064806,
      "grad_norm": 584.2164916992188,
      "learning_rate": 3.033919760300518e-08,
      "loss": 20.7866,
      "step": 21230
    },
    {
      "epoch": 168.57427716849452,
      "grad_norm": 638.7256469726562,
      "learning_rate": 3.019025925106858e-08,
      "loss": 23.7111,
      "step": 21240
    },
    {
      "epoch": 168.65403788634097,
      "grad_norm": 249.6031951904297,
      "learning_rate": 3.004166388059159e-08,
      "loss": 22.8574,
      "step": 21250
    },
    {
      "epoch": 168.73379860418743,
      "grad_norm": 939.9068603515625,
      "learning_rate": 2.9893411723435875e-08,
      "loss": 25.1982,
      "step": 21260
    },
    {
      "epoch": 168.8135593220339,
      "grad_norm": 1453.6534423828125,
      "learning_rate": 2.9745503010927487e-08,
      "loss": 25.1476,
      "step": 21270
    },
    {
      "epoch": 168.89332003988036,
      "grad_norm": 820.752197265625,
      "learning_rate": 2.9597937973856846e-08,
      "loss": 24.9524,
      "step": 21280
    },
    {
      "epoch": 168.97308075772682,
      "grad_norm": 318.9416809082031,
      "learning_rate": 2.945071684247785e-08,
      "loss": 25.4665,
      "step": 21290
    },
    {
      "epoch": 169.04785643070787,
      "grad_norm": 698.4987182617188,
      "learning_rate": 2.9303839846507834e-08,
      "loss": 22.4213,
      "step": 21300
    },
    {
      "epoch": 169.04785643070787,
      "eval_loss": 2.985820770263672,
      "eval_mae": 1.2957457304000854,
      "eval_mse": 2.985820770263672,
      "eval_r2": 0.10038679838180542,
      "eval_rmse": 1.7279527685280267,
      "eval_runtime": 9.1152,
      "eval_samples_per_second": 440.032,
      "eval_steps_per_second": 13.823,
      "step": 21300
    },
    {
      "epoch": 169.12761714855435,
      "grad_norm": 513.4396362304688,
      "learning_rate": 2.9157307215127353e-08,
      "loss": 22.4686,
      "step": 21310
    },
    {
      "epoch": 169.2073778664008,
      "grad_norm": 747.5562133789062,
      "learning_rate": 2.901111917697932e-08,
      "loss": 24.3735,
      "step": 21320
    },
    {
      "epoch": 169.28713858424726,
      "grad_norm": 326.5459289550781,
      "learning_rate": 2.886527596016927e-08,
      "loss": 24.3784,
      "step": 21330
    },
    {
      "epoch": 169.3668993020937,
      "grad_norm": 464.299072265625,
      "learning_rate": 2.8719777792264505e-08,
      "loss": 23.1418,
      "step": 21340
    },
    {
      "epoch": 169.44666001994017,
      "grad_norm": 176.2613067626953,
      "learning_rate": 2.8574624900293888e-08,
      "loss": 23.1151,
      "step": 21350
    },
    {
      "epoch": 169.52642073778665,
      "grad_norm": 325.4080810546875,
      "learning_rate": 2.842981751074769e-08,
      "loss": 24.0602,
      "step": 21360
    },
    {
      "epoch": 169.6061814556331,
      "grad_norm": 252.9591522216797,
      "learning_rate": 2.8285355849577086e-08,
      "loss": 24.0935,
      "step": 21370
    },
    {
      "epoch": 169.68594217347956,
      "grad_norm": 140.0712127685547,
      "learning_rate": 2.814124014219349e-08,
      "loss": 24.217,
      "step": 21380
    },
    {
      "epoch": 169.76570289132601,
      "grad_norm": 171.5577850341797,
      "learning_rate": 2.7997470613468853e-08,
      "loss": 25.4641,
      "step": 21390
    },
    {
      "epoch": 169.84546360917247,
      "grad_norm": 470.9291076660156,
      "learning_rate": 2.7854047487734843e-08,
      "loss": 23.8713,
      "step": 21400
    },
    {
      "epoch": 169.84546360917247,
      "eval_loss": 2.9879159927368164,
      "eval_mae": 1.294660210609436,
      "eval_mse": 2.9879157543182373,
      "eval_r2": 0.09975558519363403,
      "eval_rmse": 1.72855886631559,
      "eval_runtime": 9.0644,
      "eval_samples_per_second": 442.5,
      "eval_steps_per_second": 13.901,
      "step": 21400
    },
    {
      "epoch": 169.92522432701895,
      "grad_norm": 498.794921875,
      "learning_rate": 2.7710970988782584e-08,
      "loss": 23.0455,
      "step": 21410
    },
    {
      "epoch": 170.0,
      "grad_norm": 135.36593627929688,
      "learning_rate": 2.75682413398623e-08,
      "loss": 22.2191,
      "step": 21420
    },
    {
      "epoch": 170.07976071784645,
      "grad_norm": 1148.9171142578125,
      "learning_rate": 2.742585876368314e-08,
      "loss": 24.7924,
      "step": 21430
    },
    {
      "epoch": 170.1595214356929,
      "grad_norm": 969.6644287109375,
      "learning_rate": 2.728382348241254e-08,
      "loss": 23.3944,
      "step": 21440
    },
    {
      "epoch": 170.2392821535394,
      "grad_norm": 677.8080444335938,
      "learning_rate": 2.7142135717676206e-08,
      "loss": 23.1857,
      "step": 21450
    },
    {
      "epoch": 170.31904287138585,
      "grad_norm": 340.6452331542969,
      "learning_rate": 2.7000795690557436e-08,
      "loss": 25.0963,
      "step": 21460
    },
    {
      "epoch": 170.3988035892323,
      "grad_norm": 1116.0831298828125,
      "learning_rate": 2.685980362159698e-08,
      "loss": 21.595,
      "step": 21470
    },
    {
      "epoch": 170.47856430707876,
      "grad_norm": 704.5206298828125,
      "learning_rate": 2.6719159730792724e-08,
      "loss": 22.0435,
      "step": 21480
    },
    {
      "epoch": 170.5583250249252,
      "grad_norm": 437.9672546386719,
      "learning_rate": 2.657886423759914e-08,
      "loss": 25.7437,
      "step": 21490
    },
    {
      "epoch": 170.6380857427717,
      "grad_norm": 393.89739990234375,
      "learning_rate": 2.6438917360927266e-08,
      "loss": 23.6258,
      "step": 21500
    },
    {
      "epoch": 170.6380857427717,
      "eval_loss": 2.985738515853882,
      "eval_mae": 1.294340968132019,
      "eval_mse": 2.985738515853882,
      "eval_r2": 0.10041159391403198,
      "eval_rmse": 1.7279289672477518,
      "eval_runtime": 9.0977,
      "eval_samples_per_second": 440.882,
      "eval_steps_per_second": 13.85,
      "step": 21500
    },
    {
      "epoch": 170.71784646061815,
      "grad_norm": 555.5964965820312,
      "learning_rate": 2.6299319319143974e-08,
      "loss": 23.184,
      "step": 21510
    },
    {
      "epoch": 170.7976071784646,
      "grad_norm": 287.30377197265625,
      "learning_rate": 2.616007033007189e-08,
      "loss": 23.8837,
      "step": 21520
    },
    {
      "epoch": 170.87736789631106,
      "grad_norm": 648.5049438476562,
      "learning_rate": 2.6021170610989045e-08,
      "loss": 23.3464,
      "step": 21530
    },
    {
      "epoch": 170.95712861415754,
      "grad_norm": 1021.4182739257812,
      "learning_rate": 2.5882620378628567e-08,
      "loss": 25.7258,
      "step": 21540
    },
    {
      "epoch": 171.0319042871386,
      "grad_norm": 339.848388671875,
      "learning_rate": 2.574441984917794e-08,
      "loss": 21.4869,
      "step": 21550
    },
    {
      "epoch": 171.11166500498504,
      "grad_norm": 1136.0574951171875,
      "learning_rate": 2.560656923827931e-08,
      "loss": 23.1689,
      "step": 21560
    },
    {
      "epoch": 171.1914257228315,
      "grad_norm": 198.03421020507812,
      "learning_rate": 2.546906876102878e-08,
      "loss": 23.7328,
      "step": 21570
    },
    {
      "epoch": 171.27118644067798,
      "grad_norm": 463.3050231933594,
      "learning_rate": 2.5331918631975952e-08,
      "loss": 23.6093,
      "step": 21580
    },
    {
      "epoch": 171.35094715852443,
      "grad_norm": 298.0157165527344,
      "learning_rate": 2.5195119065123837e-08,
      "loss": 22.656,
      "step": 21590
    },
    {
      "epoch": 171.4307078763709,
      "grad_norm": 527.3220825195312,
      "learning_rate": 2.505867027392858e-08,
      "loss": 24.8544,
      "step": 21600
    },
    {
      "epoch": 171.4307078763709,
      "eval_loss": 2.985347032546997,
      "eval_mae": 1.2959703207015991,
      "eval_mse": 2.985347032546997,
      "eval_r2": 0.10052955150604248,
      "eval_rmse": 1.7278156824577664,
      "eval_runtime": 9.1181,
      "eval_samples_per_second": 439.893,
      "eval_steps_per_second": 13.819,
      "step": 21600
    },
    {
      "epoch": 171.51046859421734,
      "grad_norm": 379.0373229980469,
      "learning_rate": 2.492257247129878e-08,
      "loss": 24.2283,
      "step": 21610
    },
    {
      "epoch": 171.5902293120638,
      "grad_norm": 256.77093505859375,
      "learning_rate": 2.4786825869595545e-08,
      "loss": 24.2607,
      "step": 21620
    },
    {
      "epoch": 171.66999002991028,
      "grad_norm": 853.0988159179688,
      "learning_rate": 2.4651430680631857e-08,
      "loss": 24.1805,
      "step": 21630
    },
    {
      "epoch": 171.74975074775674,
      "grad_norm": 120.61614990234375,
      "learning_rate": 2.451638711567236e-08,
      "loss": 24.5177,
      "step": 21640
    },
    {
      "epoch": 171.8295114656032,
      "grad_norm": 637.2351684570312,
      "learning_rate": 2.4381695385433192e-08,
      "loss": 23.1552,
      "step": 21650
    },
    {
      "epoch": 171.90927218344964,
      "grad_norm": 93.33638763427734,
      "learning_rate": 2.424735570008149e-08,
      "loss": 24.8734,
      "step": 21660
    },
    {
      "epoch": 171.9890329012961,
      "grad_norm": 102.57537078857422,
      "learning_rate": 2.411336826923477e-08,
      "loss": 23.239,
      "step": 21670
    },
    {
      "epoch": 172.06380857427718,
      "grad_norm": 117.32454681396484,
      "learning_rate": 2.3979733301961265e-08,
      "loss": 22.2166,
      "step": 21680
    },
    {
      "epoch": 172.14356929212363,
      "grad_norm": 666.1997680664062,
      "learning_rate": 2.384645100677912e-08,
      "loss": 24.1691,
      "step": 21690
    },
    {
      "epoch": 172.22333000997008,
      "grad_norm": 317.80279541015625,
      "learning_rate": 2.3713521591656156e-08,
      "loss": 23.974,
      "step": 21700
    },
    {
      "epoch": 172.22333000997008,
      "eval_loss": 2.9877893924713135,
      "eval_mae": 1.2950363159179688,
      "eval_mse": 2.9877893924713135,
      "eval_r2": 0.09979367256164551,
      "eval_rmse": 1.728522314716045,
      "eval_runtime": 9.1078,
      "eval_samples_per_second": 440.394,
      "eval_steps_per_second": 13.834,
      "step": 21700
    },
    {
      "epoch": 172.30309072781654,
      "grad_norm": 220.41844177246094,
      "learning_rate": 2.35809452640095e-08,
      "loss": 24.3645,
      "step": 21710
    },
    {
      "epoch": 172.38285144566302,
      "grad_norm": 176.1417999267578,
      "learning_rate": 2.344872223070557e-08,
      "loss": 24.1271,
      "step": 21720
    },
    {
      "epoch": 172.46261216350948,
      "grad_norm": 750.0230712890625,
      "learning_rate": 2.3316852698059276e-08,
      "loss": 22.9614,
      "step": 21730
    },
    {
      "epoch": 172.54237288135593,
      "grad_norm": 635.2285766601562,
      "learning_rate": 2.3185336871834105e-08,
      "loss": 24.1727,
      "step": 21740
    },
    {
      "epoch": 172.62213359920239,
      "grad_norm": 276.4936218261719,
      "learning_rate": 2.3054174957241547e-08,
      "loss": 22.6836,
      "step": 21750
    },
    {
      "epoch": 172.70189431704884,
      "grad_norm": 410.6717529296875,
      "learning_rate": 2.2923367158940833e-08,
      "loss": 22.6491,
      "step": 21760
    },
    {
      "epoch": 172.78165503489532,
      "grad_norm": 966.279052734375,
      "learning_rate": 2.279291368103878e-08,
      "loss": 24.87,
      "step": 21770
    },
    {
      "epoch": 172.86141575274178,
      "grad_norm": 311.2745056152344,
      "learning_rate": 2.2662814727089253e-08,
      "loss": 24.8305,
      "step": 21780
    },
    {
      "epoch": 172.94117647058823,
      "grad_norm": 182.11019897460938,
      "learning_rate": 2.2533070500092897e-08,
      "loss": 23.1164,
      "step": 21790
    },
    {
      "epoch": 173.01595214356928,
      "grad_norm": 469.53656005859375,
      "learning_rate": 2.2403681202496945e-08,
      "loss": 21.9007,
      "step": 21800
    },
    {
      "epoch": 173.01595214356928,
      "eval_loss": 2.9855799674987793,
      "eval_mae": 1.2968862056732178,
      "eval_mse": 2.9855797290802,
      "eval_r2": 0.10045945644378662,
      "eval_rmse": 1.7278830195010888,
      "eval_runtime": 9.0932,
      "eval_samples_per_second": 441.099,
      "eval_steps_per_second": 13.857,
      "step": 21800
    },
    {
      "epoch": 173.09571286141576,
      "grad_norm": 457.03216552734375,
      "learning_rate": 2.2274647036194723e-08,
      "loss": 24.6794,
      "step": 21810
    },
    {
      "epoch": 173.17547357926222,
      "grad_norm": 439.5225830078125,
      "learning_rate": 2.2145968202525566e-08,
      "loss": 22.7831,
      "step": 21820
    },
    {
      "epoch": 173.25523429710867,
      "grad_norm": 752.224609375,
      "learning_rate": 2.20176449022742e-08,
      "loss": 23.3946,
      "step": 21830
    },
    {
      "epoch": 173.33499501495513,
      "grad_norm": 185.27256774902344,
      "learning_rate": 2.1889677335670626e-08,
      "loss": 25.1106,
      "step": 21840
    },
    {
      "epoch": 173.4147557328016,
      "grad_norm": 573.1307373046875,
      "learning_rate": 2.1762065702389863e-08,
      "loss": 21.3581,
      "step": 21850
    },
    {
      "epoch": 173.49451645064806,
      "grad_norm": 388.6234436035156,
      "learning_rate": 2.1634810201551583e-08,
      "loss": 24.5047,
      "step": 21860
    },
    {
      "epoch": 173.57427716849452,
      "grad_norm": 453.5081787109375,
      "learning_rate": 2.1507911031719506e-08,
      "loss": 23.7577,
      "step": 21870
    },
    {
      "epoch": 173.65403788634097,
      "grad_norm": 386.9514465332031,
      "learning_rate": 2.1381368390901572e-08,
      "loss": 22.9572,
      "step": 21880
    },
    {
      "epoch": 173.73379860418743,
      "grad_norm": 404.28570556640625,
      "learning_rate": 2.125518247654945e-08,
      "loss": 23.3758,
      "step": 21890
    },
    {
      "epoch": 173.8135593220339,
      "grad_norm": 399.77294921875,
      "learning_rate": 2.1129353485558054e-08,
      "loss": 24.8796,
      "step": 21900
    },
    {
      "epoch": 173.8135593220339,
      "eval_loss": 2.9860446453094482,
      "eval_mae": 1.298270344734192,
      "eval_mse": 2.9860446453094482,
      "eval_r2": 0.10031932592391968,
      "eval_rmse": 1.7280175477434967,
      "eval_runtime": 9.1532,
      "eval_samples_per_second": 438.207,
      "eval_steps_per_second": 13.766,
      "step": 21900
    },
    {
      "epoch": 173.89332003988036,
      "grad_norm": 164.7106170654297,
      "learning_rate": 2.1003881614265305e-08,
      "loss": 25.2506,
      "step": 21910
    },
    {
      "epoch": 173.97308075772682,
      "grad_norm": 167.10516357421875,
      "learning_rate": 2.0878767058452152e-08,
      "loss": 24.2321,
      "step": 21920
    },
    {
      "epoch": 174.04785643070787,
      "grad_norm": 169.80325317382812,
      "learning_rate": 2.0754010013341704e-08,
      "loss": 21.6802,
      "step": 21930
    },
    {
      "epoch": 174.12761714855435,
      "grad_norm": 402.775634765625,
      "learning_rate": 2.0629610673599502e-08,
      "loss": 24.4619,
      "step": 21940
    },
    {
      "epoch": 174.2073778664008,
      "grad_norm": 328.2489013671875,
      "learning_rate": 2.050556923333277e-08,
      "loss": 23.9299,
      "step": 21950
    },
    {
      "epoch": 174.28713858424726,
      "grad_norm": 745.198974609375,
      "learning_rate": 2.0381885886090245e-08,
      "loss": 22.8688,
      "step": 21960
    },
    {
      "epoch": 174.3668993020937,
      "grad_norm": 718.9580078125,
      "learning_rate": 2.0258560824862077e-08,
      "loss": 24.202,
      "step": 21970
    },
    {
      "epoch": 174.44666001994017,
      "grad_norm": 568.1629638671875,
      "learning_rate": 2.0135594242079252e-08,
      "loss": 24.3933,
      "step": 21980
    },
    {
      "epoch": 174.52642073778665,
      "grad_norm": 854.043701171875,
      "learning_rate": 2.0012986329613346e-08,
      "loss": 23.1882,
      "step": 21990
    },
    {
      "epoch": 174.6061814556331,
      "grad_norm": 551.8663940429688,
      "learning_rate": 1.9890737278776494e-08,
      "loss": 24.3438,
      "step": 22000
    },
    {
      "epoch": 174.6061814556331,
      "eval_loss": 2.9860429763793945,
      "eval_mae": 1.2983522415161133,
      "eval_mse": 2.9860429763793945,
      "eval_r2": 0.10031986236572266,
      "eval_rmse": 1.7280170648403315,
      "eval_runtime": 9.1231,
      "eval_samples_per_second": 439.654,
      "eval_steps_per_second": 13.811,
      "step": 22000
    },
    {
      "epoch": 174.68594217347956,
      "grad_norm": 310.53900146484375,
      "learning_rate": 1.9768847280320605e-08,
      "loss": 24.7496,
      "step": 22010
    },
    {
      "epoch": 174.76570289132601,
      "grad_norm": 690.4244995117188,
      "learning_rate": 1.964731652443752e-08,
      "loss": 24.1169,
      "step": 22020
    },
    {
      "epoch": 174.84546360917247,
      "grad_norm": 717.24267578125,
      "learning_rate": 1.9526145200758626e-08,
      "loss": 22.478,
      "step": 22030
    },
    {
      "epoch": 174.92522432701895,
      "grad_norm": 82.09710693359375,
      "learning_rate": 1.9405333498354105e-08,
      "loss": 22.5815,
      "step": 22040
    },
    {
      "epoch": 175.0,
      "grad_norm": 272.75677490234375,
      "learning_rate": 1.9284881605733357e-08,
      "loss": 22.7834,
      "step": 22050
    },
    {
      "epoch": 175.07976071784645,
      "grad_norm": 125.09671783447266,
      "learning_rate": 1.9164789710844233e-08,
      "loss": 26.051,
      "step": 22060
    },
    {
      "epoch": 175.1595214356929,
      "grad_norm": 937.7044067382812,
      "learning_rate": 1.9045058001072838e-08,
      "loss": 22.7696,
      "step": 22070
    },
    {
      "epoch": 175.2392821535394,
      "grad_norm": 635.612060546875,
      "learning_rate": 1.892568666324318e-08,
      "loss": 23.5138,
      "step": 22080
    },
    {
      "epoch": 175.31904287138585,
      "grad_norm": 331.49896240234375,
      "learning_rate": 1.8806675883617174e-08,
      "loss": 24.1467,
      "step": 22090
    },
    {
      "epoch": 175.3988035892323,
      "grad_norm": 672.1553344726562,
      "learning_rate": 1.868802584789392e-08,
      "loss": 21.7088,
      "step": 22100
    },
    {
      "epoch": 175.3988035892323,
      "eval_loss": 2.9872848987579346,
      "eval_mae": 1.2949774265289307,
      "eval_mse": 2.9872846603393555,
      "eval_r2": 0.09994572401046753,
      "eval_rmse": 1.7283763075034775,
      "eval_runtime": 9.3776,
      "eval_samples_per_second": 427.72,
      "eval_steps_per_second": 13.436,
      "step": 22100
    },
    {
      "epoch": 175.47856430707876,
      "grad_norm": 190.68898010253906,
      "learning_rate": 1.856973674120979e-08,
      "loss": 23.3435,
      "step": 22110
    },
    {
      "epoch": 175.5583250249252,
      "grad_norm": 517.9446411132812,
      "learning_rate": 1.8451808748137844e-08,
      "loss": 24.4168,
      "step": 22120
    },
    {
      "epoch": 175.6380857427717,
      "grad_norm": 399.7643737792969,
      "learning_rate": 1.833424205268777e-08,
      "loss": 24.571,
      "step": 22130
    },
    {
      "epoch": 175.71784646061815,
      "grad_norm": 825.5567626953125,
      "learning_rate": 1.8217036838305443e-08,
      "loss": 23.1831,
      "step": 22140
    },
    {
      "epoch": 175.7976071784646,
      "grad_norm": 284.5862121582031,
      "learning_rate": 1.810019328787285e-08,
      "loss": 23.4052,
      "step": 22150
    },
    {
      "epoch": 175.87736789631106,
      "grad_norm": 1046.2374267578125,
      "learning_rate": 1.7983711583707346e-08,
      "loss": 23.4903,
      "step": 22160
    },
    {
      "epoch": 175.95712861415754,
      "grad_norm": 478.9454345703125,
      "learning_rate": 1.7867591907561958e-08,
      "loss": 25.1356,
      "step": 22170
    },
    {
      "epoch": 176.0319042871386,
      "grad_norm": 249.65626525878906,
      "learning_rate": 1.775183444062475e-08,
      "loss": 23.2627,
      "step": 22180
    },
    {
      "epoch": 176.11166500498504,
      "grad_norm": 537.0014038085938,
      "learning_rate": 1.763643936351858e-08,
      "loss": 22.7277,
      "step": 22190
    },
    {
      "epoch": 176.1914257228315,
      "grad_norm": 539.47802734375,
      "learning_rate": 1.7521406856300797e-08,
      "loss": 24.1051,
      "step": 22200
    },
    {
      "epoch": 176.1914257228315,
      "eval_loss": 2.9861366748809814,
      "eval_mae": 1.297548770904541,
      "eval_mse": 2.9861366748809814,
      "eval_r2": 0.10029160976409912,
      "eval_rmse": 1.7280441761948626,
      "eval_runtime": 9.1459,
      "eval_samples_per_second": 438.559,
      "eval_steps_per_second": 13.777,
      "step": 22200
    },
    {
      "epoch": 176.27118644067798,
      "grad_norm": 312.5366516113281,
      "learning_rate": 1.740673709846313e-08,
      "loss": 23.4949,
      "step": 22210
    },
    {
      "epoch": 176.35094715852443,
      "grad_norm": 391.32476806640625,
      "learning_rate": 1.729243026893118e-08,
      "loss": 22.7607,
      "step": 22220
    },
    {
      "epoch": 176.4307078763709,
      "grad_norm": 257.8359375,
      "learning_rate": 1.7178486546064353e-08,
      "loss": 25.6291,
      "step": 22230
    },
    {
      "epoch": 176.51046859421734,
      "grad_norm": 715.9336547851562,
      "learning_rate": 1.706490610765543e-08,
      "loss": 24.058,
      "step": 22240
    },
    {
      "epoch": 176.5902293120638,
      "grad_norm": 565.8201904296875,
      "learning_rate": 1.695168913093023e-08,
      "loss": 24.0624,
      "step": 22250
    },
    {
      "epoch": 176.66999002991028,
      "grad_norm": 206.21766662597656,
      "learning_rate": 1.683883579254769e-08,
      "loss": 23.9974,
      "step": 22260
    },
    {
      "epoch": 176.74975074775674,
      "grad_norm": 788.8040161132812,
      "learning_rate": 1.6726346268599145e-08,
      "loss": 21.2712,
      "step": 22270
    },
    {
      "epoch": 176.8295114656032,
      "grad_norm": 308.278564453125,
      "learning_rate": 1.661422073460822e-08,
      "loss": 25.0758,
      "step": 22280
    },
    {
      "epoch": 176.90927218344964,
      "grad_norm": 585.3943481445312,
      "learning_rate": 1.6502459365530797e-08,
      "loss": 23.6577,
      "step": 22290
    },
    {
      "epoch": 176.9890329012961,
      "grad_norm": 492.2779846191406,
      "learning_rate": 1.6391062335754325e-08,
      "loss": 23.3792,
      "step": 22300
    },
    {
      "epoch": 176.9890329012961,
      "eval_loss": 2.9852004051208496,
      "eval_mae": 1.2953109741210938,
      "eval_mse": 2.9852004051208496,
      "eval_r2": 0.10057371854782104,
      "eval_rmse": 1.7277732504934928,
      "eval_runtime": 9.1232,
      "eval_samples_per_second": 439.65,
      "eval_steps_per_second": 13.811,
      "step": 22300
    },
    {
      "epoch": 177.06380857427718,
      "grad_norm": 381.8243713378906,
      "learning_rate": 1.62800298190979e-08,
      "loss": 21.2628,
      "step": 22310
    },
    {
      "epoch": 177.14356929212363,
      "grad_norm": 526.283935546875,
      "learning_rate": 1.6169361988811747e-08,
      "loss": 23.4491,
      "step": 22320
    },
    {
      "epoch": 177.22333000997008,
      "grad_norm": 304.78961181640625,
      "learning_rate": 1.605905901757701e-08,
      "loss": 22.4426,
      "step": 22330
    },
    {
      "epoch": 177.30309072781654,
      "grad_norm": 120.56478881835938,
      "learning_rate": 1.596009844021709e-08,
      "loss": 25.0317,
      "step": 22340
    },
    {
      "epoch": 177.38285144566302,
      "grad_norm": 744.4158935546875,
      "learning_rate": 1.5850489174877747e-08,
      "loss": 24.6723,
      "step": 22350
    },
    {
      "epoch": 177.46261216350948,
      "grad_norm": 1183.1009521484375,
      "learning_rate": 1.5741245266145065e-08,
      "loss": 22.3539,
      "step": 22360
    },
    {
      "epoch": 177.54237288135593,
      "grad_norm": 349.775634765625,
      "learning_rate": 1.563236688447836e-08,
      "loss": 23.4585,
      "step": 22370
    },
    {
      "epoch": 177.62213359920239,
      "grad_norm": 242.05673217773438,
      "learning_rate": 1.5523854199766732e-08,
      "loss": 23.8761,
      "step": 22380
    },
    {
      "epoch": 177.70189431704884,
      "grad_norm": 368.234130859375,
      "learning_rate": 1.541570738132858e-08,
      "loss": 23.9772,
      "step": 22390
    },
    {
      "epoch": 177.78165503489532,
      "grad_norm": 529.9364624023438,
      "learning_rate": 1.5307926597911386e-08,
      "loss": 22.4049,
      "step": 22400
    },
    {
      "epoch": 177.78165503489532,
      "eval_loss": 2.9866483211517334,
      "eval_mae": 1.2959117889404297,
      "eval_mse": 2.9866480827331543,
      "eval_r2": 0.10013753175735474,
      "eval_rmse": 1.7281921428860723,
      "eval_runtime": 9.081,
      "eval_samples_per_second": 441.691,
      "eval_steps_per_second": 13.875,
      "step": 22400
    },
    {
      "epoch": 177.86141575274178,
      "grad_norm": 845.2642211914062,
      "learning_rate": 1.520051201769168e-08,
      "loss": 24.6607,
      "step": 22410
    },
    {
      "epoch": 177.94117647058823,
      "grad_norm": 914.9999389648438,
      "learning_rate": 1.5093463808274338e-08,
      "loss": 25.4712,
      "step": 22420
    },
    {
      "epoch": 178.01595214356928,
      "grad_norm": 224.6628875732422,
      "learning_rate": 1.4986782136692715e-08,
      "loss": 23.3044,
      "step": 22430
    },
    {
      "epoch": 178.09571286141576,
      "grad_norm": 1200.2086181640625,
      "learning_rate": 1.4880467169408245e-08,
      "loss": 23.4508,
      "step": 22440
    },
    {
      "epoch": 178.17547357926222,
      "grad_norm": 532.684814453125,
      "learning_rate": 1.4774519072310032e-08,
      "loss": 23.1079,
      "step": 22450
    },
    {
      "epoch": 178.25523429710867,
      "grad_norm": 854.6232299804688,
      "learning_rate": 1.466893801071495e-08,
      "loss": 23.4958,
      "step": 22460
    },
    {
      "epoch": 178.33499501495513,
      "grad_norm": 631.4081420898438,
      "learning_rate": 1.4563724149366991e-08,
      "loss": 22.0791,
      "step": 22470
    },
    {
      "epoch": 178.4147557328016,
      "grad_norm": 964.5308227539062,
      "learning_rate": 1.4458877652437224e-08,
      "loss": 25.2987,
      "step": 22480
    },
    {
      "epoch": 178.49451645064806,
      "grad_norm": 1529.9703369140625,
      "learning_rate": 1.4354398683523506e-08,
      "loss": 24.4458,
      "step": 22490
    },
    {
      "epoch": 178.57427716849452,
      "grad_norm": 1350.5501708984375,
      "learning_rate": 1.425028740565032e-08,
      "loss": 25.5982,
      "step": 22500
    },
    {
      "epoch": 178.57427716849452,
      "eval_loss": 2.986062526702881,
      "eval_mae": 1.2957606315612793,
      "eval_mse": 2.98606276512146,
      "eval_r2": 0.10031390190124512,
      "eval_rmse": 1.728022790683462,
      "eval_runtime": 9.0926,
      "eval_samples_per_second": 441.13,
      "eval_steps_per_second": 13.857,
      "step": 22500
    },
    {
      "epoch": 178.65403788634097,
      "grad_norm": 275.2545471191406,
      "learning_rate": 1.4146543981268217e-08,
      "loss": 24.822,
      "step": 22510
    },
    {
      "epoch": 178.73379860418743,
      "grad_norm": 538.2731323242188,
      "learning_rate": 1.4043168572253905e-08,
      "loss": 24.7729,
      "step": 22520
    },
    {
      "epoch": 178.8135593220339,
      "grad_norm": 765.3468627929688,
      "learning_rate": 1.3940161339909884e-08,
      "loss": 23.033,
      "step": 22530
    },
    {
      "epoch": 178.89332003988036,
      "grad_norm": 199.8118133544922,
      "learning_rate": 1.3837522444964083e-08,
      "loss": 23.8803,
      "step": 22540
    },
    {
      "epoch": 178.97308075772682,
      "grad_norm": 362.73321533203125,
      "learning_rate": 1.3735252047569673e-08,
      "loss": 23.395,
      "step": 22550
    },
    {
      "epoch": 179.04785643070787,
      "grad_norm": 600.6903686523438,
      "learning_rate": 1.3633350307304975e-08,
      "loss": 20.4186,
      "step": 22560
    },
    {
      "epoch": 179.12761714855435,
      "grad_norm": 92.33502197265625,
      "learning_rate": 1.3531817383172911e-08,
      "loss": 22.7443,
      "step": 22570
    },
    {
      "epoch": 179.2073778664008,
      "grad_norm": 774.8335571289062,
      "learning_rate": 1.3430653433601058e-08,
      "loss": 26.5549,
      "step": 22580
    },
    {
      "epoch": 179.28713858424726,
      "grad_norm": 1109.836181640625,
      "learning_rate": 1.3329858616441175e-08,
      "loss": 24.2557,
      "step": 22590
    },
    {
      "epoch": 179.3668993020937,
      "grad_norm": 440.6700439453125,
      "learning_rate": 1.3229433088968984e-08,
      "loss": 23.0492,
      "step": 22600
    },
    {
      "epoch": 179.3668993020937,
      "eval_loss": 2.987112522125244,
      "eval_mae": 1.295152187347412,
      "eval_mse": 2.9871127605438232,
      "eval_r2": 0.09999752044677734,
      "eval_rmse": 1.7283265780933368,
      "eval_runtime": 9.1234,
      "eval_samples_per_second": 439.64,
      "eval_steps_per_second": 13.811,
      "step": 22600
    },
    {
      "epoch": 179.44666001994017,
      "grad_norm": 427.29559326171875,
      "learning_rate": 1.312937700788419e-08,
      "loss": 23.5974,
      "step": 22610
    },
    {
      "epoch": 179.52642073778665,
      "grad_norm": 238.57940673828125,
      "learning_rate": 1.3029690529309855e-08,
      "loss": 22.2489,
      "step": 22620
    },
    {
      "epoch": 179.6061814556331,
      "grad_norm": 310.37518310546875,
      "learning_rate": 1.2930373808792333e-08,
      "loss": 24.069,
      "step": 22630
    },
    {
      "epoch": 179.68594217347956,
      "grad_norm": 1182.7216796875,
      "learning_rate": 1.283142700130116e-08,
      "loss": 23.8514,
      "step": 22640
    },
    {
      "epoch": 179.76570289132601,
      "grad_norm": 594.5250854492188,
      "learning_rate": 1.27328502612285e-08,
      "loss": 23.4992,
      "step": 22650
    },
    {
      "epoch": 179.84546360917247,
      "grad_norm": 412.5122985839844,
      "learning_rate": 1.2634643742389234e-08,
      "loss": 24.1487,
      "step": 22660
    },
    {
      "epoch": 179.92522432701895,
      "grad_norm": 497.1600646972656,
      "learning_rate": 1.2536807598020421e-08,
      "loss": 23.2468,
      "step": 22670
    },
    {
      "epoch": 180.0,
      "grad_norm": 138.8757781982422,
      "learning_rate": 1.2439341980781365e-08,
      "loss": 22.8057,
      "step": 22680
    },
    {
      "epoch": 180.07976071784645,
      "grad_norm": 786.8070068359375,
      "learning_rate": 1.234224704275308e-08,
      "loss": 25.2827,
      "step": 22690
    },
    {
      "epoch": 180.1595214356929,
      "grad_norm": 284.813720703125,
      "learning_rate": 1.2245522935438263e-08,
      "loss": 23.404,
      "step": 22700
    },
    {
      "epoch": 180.1595214356929,
      "eval_loss": 2.9867138862609863,
      "eval_mae": 1.2956660985946655,
      "eval_mse": 2.9867141246795654,
      "eval_r2": 0.10011768341064453,
      "eval_rmse": 1.7282112500153346,
      "eval_runtime": 9.1925,
      "eval_samples_per_second": 436.332,
      "eval_steps_per_second": 13.707,
      "step": 22700
    },
    {
      "epoch": 180.2392821535394,
      "grad_norm": 392.55401611328125,
      "learning_rate": 1.2149169809760967e-08,
      "loss": 25.6318,
      "step": 22710
    },
    {
      "epoch": 180.31904287138585,
      "grad_norm": 463.6085205078125,
      "learning_rate": 1.2053187816066318e-08,
      "loss": 22.9739,
      "step": 22720
    },
    {
      "epoch": 180.3988035892323,
      "grad_norm": 143.3795166015625,
      "learning_rate": 1.1957577104120487e-08,
      "loss": 23.4349,
      "step": 22730
    },
    {
      "epoch": 180.47856430707876,
      "grad_norm": 236.1634063720703,
      "learning_rate": 1.1862337823110163e-08,
      "loss": 25.8816,
      "step": 22740
    },
    {
      "epoch": 180.5583250249252,
      "grad_norm": 581.2395629882812,
      "learning_rate": 1.176747012164253e-08,
      "loss": 23.6662,
      "step": 22750
    },
    {
      "epoch": 180.6380857427717,
      "grad_norm": 341.58148193359375,
      "learning_rate": 1.1672974147745068e-08,
      "loss": 22.0037,
      "step": 22760
    },
    {
      "epoch": 180.71784646061815,
      "grad_norm": 359.1589050292969,
      "learning_rate": 1.1578850048865024e-08,
      "loss": 22.7009,
      "step": 22770
    },
    {
      "epoch": 180.7976071784646,
      "grad_norm": 1259.181396484375,
      "learning_rate": 1.1485097971869584e-08,
      "loss": 23.9171,
      "step": 22780
    },
    {
      "epoch": 180.87736789631106,
      "grad_norm": 284.3455505371094,
      "learning_rate": 1.1391718063045342e-08,
      "loss": 22.9658,
      "step": 22790
    },
    {
      "epoch": 180.95712861415754,
      "grad_norm": 898.4352416992188,
      "learning_rate": 1.1298710468098188e-08,
      "loss": 23.8032,
      "step": 22800
    },
    {
      "epoch": 180.95712861415754,
      "eval_loss": 2.9856619834899902,
      "eval_mae": 1.2953579425811768,
      "eval_mse": 2.9856619834899902,
      "eval_r2": 0.10043466091156006,
      "eval_rmse": 1.727906821414277,
      "eval_runtime": 9.0627,
      "eval_samples_per_second": 442.582,
      "eval_steps_per_second": 13.903,
      "step": 22800
    },
    {
      "epoch": 181.0319042871386,
      "grad_norm": 1284.8299560546875,
      "learning_rate": 1.1206075332153091e-08,
      "loss": 23.3462,
      "step": 22810
    },
    {
      "epoch": 181.11166500498504,
      "grad_norm": 179.67510986328125,
      "learning_rate": 1.1113812799753958e-08,
      "loss": 22.7255,
      "step": 22820
    },
    {
      "epoch": 181.1914257228315,
      "grad_norm": 978.6626586914062,
      "learning_rate": 1.1021923014862988e-08,
      "loss": 24.1967,
      "step": 22830
    },
    {
      "epoch": 181.27118644067798,
      "grad_norm": 1190.5296630859375,
      "learning_rate": 1.0930406120861103e-08,
      "loss": 22.3215,
      "step": 22840
    },
    {
      "epoch": 181.35094715852443,
      "grad_norm": 299.45947265625,
      "learning_rate": 1.083926226054721e-08,
      "loss": 22.0059,
      "step": 22850
    },
    {
      "epoch": 181.4307078763709,
      "grad_norm": 464.1260070800781,
      "learning_rate": 1.0748491576138191e-08,
      "loss": 23.297,
      "step": 22860
    },
    {
      "epoch": 181.51046859421734,
      "grad_norm": 654.6446533203125,
      "learning_rate": 1.065809420926858e-08,
      "loss": 23.9606,
      "step": 22870
    },
    {
      "epoch": 181.5902293120638,
      "grad_norm": 534.0228881835938,
      "learning_rate": 1.056807030099055e-08,
      "loss": 24.3782,
      "step": 22880
    },
    {
      "epoch": 181.66999002991028,
      "grad_norm": 243.1739959716797,
      "learning_rate": 1.0478419991773346e-08,
      "loss": 23.5468,
      "step": 22890
    },
    {
      "epoch": 181.74975074775674,
      "grad_norm": 296.27593994140625,
      "learning_rate": 1.038914342150346e-08,
      "loss": 23.2563,
      "step": 22900
    },
    {
      "epoch": 181.74975074775674,
      "eval_loss": 2.9857964515686035,
      "eval_mae": 1.2959256172180176,
      "eval_mse": 2.9857964515686035,
      "eval_r2": 0.10039412975311279,
      "eval_rmse": 1.7279457316619071,
      "eval_runtime": 9.1154,
      "eval_samples_per_second": 440.022,
      "eval_steps_per_second": 13.823,
      "step": 22900
    },
    {
      "epoch": 181.8295114656032,
      "grad_norm": 350.1341247558594,
      "learning_rate": 1.03002407294841e-08,
      "loss": 24.5378,
      "step": 22910
    },
    {
      "epoch": 181.90927218344964,
      "grad_norm": 387.22998046875,
      "learning_rate": 1.0211712054435078e-08,
      "loss": 24.8081,
      "step": 22920
    },
    {
      "epoch": 181.9890329012961,
      "grad_norm": 320.7686462402344,
      "learning_rate": 1.0123557534492705e-08,
      "loss": 24.0877,
      "step": 22930
    },
    {
      "epoch": 182.06380857427718,
      "grad_norm": 329.71893310546875,
      "learning_rate": 1.0035777307209364e-08,
      "loss": 24.8788,
      "step": 22940
    },
    {
      "epoch": 182.14356929212363,
      "grad_norm": 362.3857421875,
      "learning_rate": 9.948371509553549e-09,
      "loss": 23.8843,
      "step": 22950
    },
    {
      "epoch": 182.22333000997008,
      "grad_norm": 880.0196533203125,
      "learning_rate": 9.861340277909358e-09,
      "loss": 24.9908,
      "step": 22960
    },
    {
      "epoch": 182.30309072781654,
      "grad_norm": 1081.274169921875,
      "learning_rate": 9.774683748076495e-09,
      "loss": 22.1578,
      "step": 22970
    },
    {
      "epoch": 182.38285144566302,
      "grad_norm": 762.17236328125,
      "learning_rate": 9.688402055269995e-09,
      "loss": 24.5478,
      "step": 22980
    },
    {
      "epoch": 182.46261216350948,
      "grad_norm": 165.5096893310547,
      "learning_rate": 9.602495334120136e-09,
      "loss": 22.1582,
      "step": 22990
    },
    {
      "epoch": 182.54237288135593,
      "grad_norm": 244.72225952148438,
      "learning_rate": 9.516963718671834e-09,
      "loss": 22.9301,
      "step": 23000
    },
    {
      "epoch": 182.54237288135593,
      "eval_loss": 2.9862356185913086,
      "eval_mae": 1.295803427696228,
      "eval_mse": 2.9862356185913086,
      "eval_r2": 0.10026180744171143,
      "eval_rmse": 1.728072804771636,
      "eval_runtime": 9.1456,
      "eval_samples_per_second": 438.573,
      "eval_steps_per_second": 13.777,
      "step": 23000
    },
    {
      "epoch": 182.62213359920239,
      "grad_norm": 373.03662109375,
      "learning_rate": 9.431807342384913e-09,
      "loss": 21.0979,
      "step": 23010
    },
    {
      "epoch": 182.70189431704884,
      "grad_norm": 121.8209457397461,
      "learning_rate": 9.347026338133696e-09,
      "loss": 24.8742,
      "step": 23020
    },
    {
      "epoch": 182.78165503489532,
      "grad_norm": 339.05914306640625,
      "learning_rate": 9.262620838206696e-09,
      "loss": 24.4378,
      "step": 23030
    },
    {
      "epoch": 182.86141575274178,
      "grad_norm": 252.5173797607422,
      "learning_rate": 9.17859097430651e-09,
      "loss": 23.9127,
      "step": 23040
    },
    {
      "epoch": 182.94117647058823,
      "grad_norm": 383.7898864746094,
      "learning_rate": 9.094936877549725e-09,
      "loss": 24.8554,
      "step": 23050
    },
    {
      "epoch": 183.01595214356928,
      "grad_norm": 418.6182861328125,
      "learning_rate": 9.011658678466428e-09,
      "loss": 21.7836,
      "step": 23060
    },
    {
      "epoch": 183.09571286141576,
      "grad_norm": 506.1742248535156,
      "learning_rate": 8.92875650700034e-09,
      "loss": 23.7084,
      "step": 23070
    },
    {
      "epoch": 183.17547357926222,
      "grad_norm": 480.1518249511719,
      "learning_rate": 8.84623049250835e-09,
      "loss": 24.4412,
      "step": 23080
    },
    {
      "epoch": 183.25523429710867,
      "grad_norm": 506.3466491699219,
      "learning_rate": 8.764080763760368e-09,
      "loss": 22.5461,
      "step": 23090
    },
    {
      "epoch": 183.33499501495513,
      "grad_norm": 816.4349975585938,
      "learning_rate": 8.682307448939274e-09,
      "loss": 25.0832,
      "step": 23100
    },
    {
      "epoch": 183.33499501495513,
      "eval_loss": 2.985621929168701,
      "eval_mae": 1.295138955116272,
      "eval_mse": 2.9856221675872803,
      "eval_r2": 0.10044664144515991,
      "eval_rmse": 1.7278952999494155,
      "eval_runtime": 9.0822,
      "eval_samples_per_second": 441.633,
      "eval_steps_per_second": 13.873,
      "step": 23100
    },
    {
      "epoch": 183.4147557328016,
      "grad_norm": 851.5493774414062,
      "learning_rate": 8.60091067564056e-09,
      "loss": 22.8906,
      "step": 23110
    },
    {
      "epoch": 183.49451645064806,
      "grad_norm": 787.2097778320312,
      "learning_rate": 8.519890570872101e-09,
      "loss": 23.4751,
      "step": 23120
    },
    {
      "epoch": 183.57427716849452,
      "grad_norm": 237.1964569091797,
      "learning_rate": 8.439247261054217e-09,
      "loss": 22.4332,
      "step": 23130
    },
    {
      "epoch": 183.65403788634097,
      "grad_norm": 765.3052368164062,
      "learning_rate": 8.358980872019083e-09,
      "loss": 23.7841,
      "step": 23140
    },
    {
      "epoch": 183.73379860418743,
      "grad_norm": 254.69822692871094,
      "learning_rate": 8.279091529010907e-09,
      "loss": 24.593,
      "step": 23150
    },
    {
      "epoch": 183.8135593220339,
      "grad_norm": 487.26171875,
      "learning_rate": 8.199579356685498e-09,
      "loss": 23.5859,
      "step": 23160
    },
    {
      "epoch": 183.89332003988036,
      "grad_norm": 392.75299072265625,
      "learning_rate": 8.120444479110145e-09,
      "loss": 24.5251,
      "step": 23170
    },
    {
      "epoch": 183.97308075772682,
      "grad_norm": 156.13031005859375,
      "learning_rate": 8.041687019763432e-09,
      "loss": 24.6889,
      "step": 23180
    },
    {
      "epoch": 184.04785643070787,
      "grad_norm": 612.9346923828125,
      "learning_rate": 7.96330710153506e-09,
      "loss": 22.7071,
      "step": 23190
    },
    {
      "epoch": 184.12761714855435,
      "grad_norm": 702.3763427734375,
      "learning_rate": 7.885304846725589e-09,
      "loss": 23.8737,
      "step": 23200
    },
    {
      "epoch": 184.12761714855435,
      "eval_loss": 2.9857828617095947,
      "eval_mae": 1.2956427335739136,
      "eval_mse": 2.985783100128174,
      "eval_r2": 0.10039818286895752,
      "eval_rmse": 1.7279418682722443,
      "eval_runtime": 9.1096,
      "eval_samples_per_second": 440.307,
      "eval_steps_per_second": 13.832,
      "step": 23200
    },
    {
      "epoch": 184.2073778664008,
      "grad_norm": 779.8262329101562,
      "learning_rate": 7.807680377046294e-09,
      "loss": 23.3689,
      "step": 23210
    },
    {
      "epoch": 184.28713858424726,
      "grad_norm": 1790.89892578125,
      "learning_rate": 7.730433813619042e-09,
      "loss": 23.9311,
      "step": 23220
    },
    {
      "epoch": 184.3668993020937,
      "grad_norm": 437.471435546875,
      "learning_rate": 7.653565276975915e-09,
      "loss": 23.7945,
      "step": 23230
    },
    {
      "epoch": 184.44666001994017,
      "grad_norm": 653.4305419921875,
      "learning_rate": 7.57707488705922e-09,
      "loss": 25.2285,
      "step": 23240
    },
    {
      "epoch": 184.52642073778665,
      "grad_norm": 432.324462890625,
      "learning_rate": 7.500962763221208e-09,
      "loss": 23.8279,
      "step": 23250
    },
    {
      "epoch": 184.6061814556331,
      "grad_norm": 1161.883544921875,
      "learning_rate": 7.425229024223878e-09,
      "loss": 21.3262,
      "step": 23260
    },
    {
      "epoch": 184.68594217347956,
      "grad_norm": 644.3816528320312,
      "learning_rate": 7.34987378823887e-09,
      "loss": 24.3121,
      "step": 23270
    },
    {
      "epoch": 184.76570289132601,
      "grad_norm": 789.889892578125,
      "learning_rate": 7.274897172847127e-09,
      "loss": 26.7195,
      "step": 23280
    },
    {
      "epoch": 184.84546360917247,
      "grad_norm": 519.7674560546875,
      "learning_rate": 7.200299295038875e-09,
      "loss": 23.2123,
      "step": 23290
    },
    {
      "epoch": 184.92522432701895,
      "grad_norm": 557.431640625,
      "learning_rate": 7.126080271213391e-09,
      "loss": 22.8665,
      "step": 23300
    },
    {
      "epoch": 184.92522432701895,
      "eval_loss": 2.9849884510040283,
      "eval_mae": 1.295828938484192,
      "eval_mse": 2.984988212585449,
      "eval_r2": 0.10063761472702026,
      "eval_rmse": 1.727711843041382,
      "eval_runtime": 9.1194,
      "eval_samples_per_second": 439.833,
      "eval_steps_per_second": 13.817,
      "step": 23300
    },
    {
      "epoch": 185.0,
      "grad_norm": 284.9809875488281,
      "learning_rate": 7.05224021717879e-09,
      "loss": 21.6302,
      "step": 23310
    },
    {
      "epoch": 185.07976071784645,
      "grad_norm": 756.0015258789062,
      "learning_rate": 6.978779248151795e-09,
      "loss": 23.6411,
      "step": 23320
    },
    {
      "epoch": 185.1595214356929,
      "grad_norm": 528.3745727539062,
      "learning_rate": 6.905697478757689e-09,
      "loss": 23.8638,
      "step": 23330
    },
    {
      "epoch": 185.2392821535394,
      "grad_norm": 1139.6221923828125,
      "learning_rate": 6.832995023030086e-09,
      "loss": 22.9624,
      "step": 23340
    },
    {
      "epoch": 185.31904287138585,
      "grad_norm": 281.22491455078125,
      "learning_rate": 6.76067199441066e-09,
      "loss": 22.7068,
      "step": 23350
    },
    {
      "epoch": 185.3988035892323,
      "grad_norm": 179.5164337158203,
      "learning_rate": 6.6887285057491136e-09,
      "loss": 25.3549,
      "step": 23360
    },
    {
      "epoch": 185.47856430707876,
      "grad_norm": 370.2801818847656,
      "learning_rate": 6.617164669302927e-09,
      "loss": 25.2967,
      "step": 23370
    },
    {
      "epoch": 185.5583250249252,
      "grad_norm": 167.0034942626953,
      "learning_rate": 6.545980596737083e-09,
      "loss": 23.1989,
      "step": 23380
    },
    {
      "epoch": 185.6380857427717,
      "grad_norm": 550.1342163085938,
      "learning_rate": 6.475176399124177e-09,
      "loss": 22.0248,
      "step": 23390
    },
    {
      "epoch": 185.71784646061815,
      "grad_norm": 861.5530395507812,
      "learning_rate": 6.4047521869439455e-09,
      "loss": 25.3384,
      "step": 23400
    },
    {
      "epoch": 185.71784646061815,
      "eval_loss": 2.984853744506836,
      "eval_mae": 1.2957940101623535,
      "eval_mse": 2.984853744506836,
      "eval_r2": 0.1006782054901123,
      "eval_rmse": 1.7276729275261669,
      "eval_runtime": 9.04,
      "eval_samples_per_second": 443.694,
      "eval_steps_per_second": 13.938,
      "step": 23400
    },
    {
      "epoch": 185.7976071784646,
      "grad_norm": 235.17025756835938,
      "learning_rate": 6.334708070083156e-09,
      "loss": 23.875,
      "step": 23410
    },
    {
      "epoch": 185.87736789631106,
      "grad_norm": 287.0189514160156,
      "learning_rate": 6.265044157835686e-09,
      "loss": 22.8222,
      "step": 23420
    },
    {
      "epoch": 185.95712861415754,
      "grad_norm": 823.918701171875,
      "learning_rate": 6.195760558901947e-09,
      "loss": 23.0371,
      "step": 23430
    },
    {
      "epoch": 186.0319042871386,
      "grad_norm": 427.4580993652344,
      "learning_rate": 6.126857381389072e-09,
      "loss": 24.1903,
      "step": 23440
    },
    {
      "epoch": 186.11166500498504,
      "grad_norm": 579.4691162109375,
      "learning_rate": 6.058334732810533e-09,
      "loss": 23.3387,
      "step": 23450
    },
    {
      "epoch": 186.1914257228315,
      "grad_norm": 315.81787109375,
      "learning_rate": 5.990192720086052e-09,
      "loss": 24.8437,
      "step": 23460
    },
    {
      "epoch": 186.27118644067798,
      "grad_norm": 462.8539733886719,
      "learning_rate": 5.922431449541465e-09,
      "loss": 25.4003,
      "step": 23470
    },
    {
      "epoch": 186.35094715852443,
      "grad_norm": 638.9884643554688,
      "learning_rate": 5.8550510269084755e-09,
      "loss": 23.8574,
      "step": 23480
    },
    {
      "epoch": 186.4307078763709,
      "grad_norm": 182.34124755859375,
      "learning_rate": 5.78805155732448e-09,
      "loss": 21.1267,
      "step": 23490
    },
    {
      "epoch": 186.51046859421734,
      "grad_norm": 607.1478881835938,
      "learning_rate": 5.721433145332549e-09,
      "loss": 24.8929,
      "step": 23500
    },
    {
      "epoch": 186.51046859421734,
      "eval_loss": 2.9855034351348877,
      "eval_mae": 1.2960774898529053,
      "eval_mse": 2.9855034351348877,
      "eval_r2": 0.10048240423202515,
      "eval_rmse": 1.7278609420711168,
      "eval_runtime": 9.0589,
      "eval_samples_per_second": 442.771,
      "eval_steps_per_second": 13.909,
      "step": 23500
    },
    {
      "epoch": 186.5902293120638,
      "grad_norm": 331.1116027832031,
      "learning_rate": 5.655195894881143e-09,
      "loss": 20.8989,
      "step": 23510
    },
    {
      "epoch": 186.66999002991028,
      "grad_norm": 475.9549865722656,
      "learning_rate": 5.589339909323948e-09,
      "loss": 21.8069,
      "step": 23520
    },
    {
      "epoch": 186.74975074775674,
      "grad_norm": 561.3901977539062,
      "learning_rate": 5.5238652914197105e-09,
      "loss": 25.9163,
      "step": 23530
    },
    {
      "epoch": 186.8295114656032,
      "grad_norm": 763.07373046875,
      "learning_rate": 5.45877214333218e-09,
      "loss": 24.2063,
      "step": 23540
    },
    {
      "epoch": 186.90927218344964,
      "grad_norm": 189.84625244140625,
      "learning_rate": 5.394060566629832e-09,
      "loss": 23.977,
      "step": 23550
    },
    {
      "epoch": 186.9890329012961,
      "grad_norm": 472.99041748046875,
      "learning_rate": 5.329730662285787e-09,
      "loss": 24.547,
      "step": 23560
    },
    {
      "epoch": 187.06380857427718,
      "grad_norm": 863.2728881835938,
      "learning_rate": 5.265782530677582e-09,
      "loss": 20.8558,
      "step": 23570
    },
    {
      "epoch": 187.14356929212363,
      "grad_norm": 216.08444213867188,
      "learning_rate": 5.20221627158704e-09,
      "loss": 24.0826,
      "step": 23580
    },
    {
      "epoch": 187.22333000997008,
      "grad_norm": 852.3370971679688,
      "learning_rate": 5.139031984200237e-09,
      "loss": 24.7418,
      "step": 23590
    },
    {
      "epoch": 187.30309072781654,
      "grad_norm": 320.7305908203125,
      "learning_rate": 5.076229767107115e-09,
      "loss": 24.4025,
      "step": 23600
    },
    {
      "epoch": 187.30309072781654,
      "eval_loss": 2.9850869178771973,
      "eval_mae": 1.2962417602539062,
      "eval_mse": 2.985086679458618,
      "eval_r2": 0.1006079912185669,
      "eval_rmse": 1.7277403391304547,
      "eval_runtime": 9.0865,
      "eval_samples_per_second": 441.425,
      "eval_steps_per_second": 13.867,
      "step": 23600
    },
    {
      "epoch": 187.38285144566302,
      "grad_norm": 186.22344970703125,
      "learning_rate": 5.013809718301454e-09,
      "loss": 25.3328,
      "step": 23610
    },
    {
      "epoch": 187.46261216350948,
      "grad_norm": 1284.268798828125,
      "learning_rate": 4.951771935180815e-09,
      "loss": 23.2086,
      "step": 23620
    },
    {
      "epoch": 187.54237288135593,
      "grad_norm": 425.0491638183594,
      "learning_rate": 4.890116514546211e-09,
      "loss": 22.4389,
      "step": 23630
    },
    {
      "epoch": 187.62213359920239,
      "grad_norm": 851.8677978515625,
      "learning_rate": 4.828843552602047e-09,
      "loss": 24.5933,
      "step": 23640
    },
    {
      "epoch": 187.70189431704884,
      "grad_norm": 434.36322021484375,
      "learning_rate": 4.767953144955927e-09,
      "loss": 22.7603,
      "step": 23650
    },
    {
      "epoch": 187.78165503489532,
      "grad_norm": 668.190673828125,
      "learning_rate": 4.707445386618658e-09,
      "loss": 21.8588,
      "step": 23660
    },
    {
      "epoch": 187.86141575274178,
      "grad_norm": 220.99130249023438,
      "learning_rate": 4.647320372003772e-09,
      "loss": 24.3321,
      "step": 23670
    },
    {
      "epoch": 187.94117647058823,
      "grad_norm": 598.3583374023438,
      "learning_rate": 4.5875781949278044e-09,
      "loss": 25.4108,
      "step": 23680
    },
    {
      "epoch": 188.01595214356928,
      "grad_norm": 570.7030029296875,
      "learning_rate": 4.5282189486097745e-09,
      "loss": 22.1609,
      "step": 23690
    },
    {
      "epoch": 188.09571286141576,
      "grad_norm": 250.53883361816406,
      "learning_rate": 4.469242725671202e-09,
      "loss": 21.5695,
      "step": 23700
    },
    {
      "epoch": 188.09571286141576,
      "eval_loss": 2.9850971698760986,
      "eval_mae": 1.2961443662643433,
      "eval_mse": 2.9850971698760986,
      "eval_r2": 0.1006048321723938,
      "eval_rmse": 1.7277433750057034,
      "eval_runtime": 9.0591,
      "eval_samples_per_second": 442.758,
      "eval_steps_per_second": 13.909,
      "step": 23700
    },
    {
      "epoch": 188.17547357926222,
      "grad_norm": 533.2770385742188,
      "learning_rate": 4.410649618136059e-09,
      "loss": 24.5917,
      "step": 23710
    },
    {
      "epoch": 188.25523429710867,
      "grad_norm": 85.58604431152344,
      "learning_rate": 4.352439717430434e-09,
      "loss": 22.8943,
      "step": 23720
    },
    {
      "epoch": 188.33499501495513,
      "grad_norm": 430.8806457519531,
      "learning_rate": 4.2946131143824785e-09,
      "loss": 24.1939,
      "step": 23730
    },
    {
      "epoch": 188.4147557328016,
      "grad_norm": 426.0600891113281,
      "learning_rate": 4.23716989922232e-09,
      "loss": 22.7485,
      "step": 23740
    },
    {
      "epoch": 188.49451645064806,
      "grad_norm": 138.37684631347656,
      "learning_rate": 4.1801101615817904e-09,
      "loss": 22.0539,
      "step": 23750
    },
    {
      "epoch": 188.57427716849452,
      "grad_norm": 168.79737854003906,
      "learning_rate": 4.1234339904944186e-09,
      "loss": 22.786,
      "step": 23760
    },
    {
      "epoch": 188.65403788634097,
      "grad_norm": 357.03338623046875,
      "learning_rate": 4.067141474395214e-09,
      "loss": 26.0838,
      "step": 23770
    },
    {
      "epoch": 188.73379860418743,
      "grad_norm": 729.7440795898438,
      "learning_rate": 4.011232701120498e-09,
      "loss": 24.8297,
      "step": 23780
    },
    {
      "epoch": 188.8135593220339,
      "grad_norm": 240.0101776123047,
      "learning_rate": 3.955707757907906e-09,
      "loss": 25.7474,
      "step": 23790
    },
    {
      "epoch": 188.89332003988036,
      "grad_norm": 332.90838623046875,
      "learning_rate": 3.90056673139616e-09,
      "loss": 24.8005,
      "step": 23800
    },
    {
      "epoch": 188.89332003988036,
      "eval_loss": 2.9858198165893555,
      "eval_mae": 1.296196460723877,
      "eval_mse": 2.9858198165893555,
      "eval_r2": 0.1003870964050293,
      "eval_rmse": 1.7279524925730323,
      "eval_runtime": 9.0618,
      "eval_samples_per_second": 442.629,
      "eval_steps_per_second": 13.905,
      "step": 23800
    },
    {
      "epoch": 188.97308075772682,
      "grad_norm": 1091.8680419921875,
      "learning_rate": 3.845809707624825e-09,
      "loss": 23.3533,
      "step": 23810
    },
    {
      "epoch": 189.04785643070787,
      "grad_norm": 332.0385437011719,
      "learning_rate": 3.791436772034362e-09,
      "loss": 23.0622,
      "step": 23820
    },
    {
      "epoch": 189.12761714855435,
      "grad_norm": 548.7697143554688,
      "learning_rate": 3.737448009465932e-09,
      "loss": 23.0311,
      "step": 23830
    },
    {
      "epoch": 189.2073778664008,
      "grad_norm": 172.7244415283203,
      "learning_rate": 3.6838435041612324e-09,
      "loss": 23.8273,
      "step": 23840
    },
    {
      "epoch": 189.28713858424726,
      "grad_norm": 104.53250885009766,
      "learning_rate": 3.6306233397623564e-09,
      "loss": 24.7454,
      "step": 23850
    },
    {
      "epoch": 189.3668993020937,
      "grad_norm": 458.9788513183594,
      "learning_rate": 3.5777875993117377e-09,
      "loss": 21.2395,
      "step": 23860
    },
    {
      "epoch": 189.44666001994017,
      "grad_norm": 556.30859375,
      "learning_rate": 3.525336365251902e-09,
      "loss": 24.1263,
      "step": 23870
    },
    {
      "epoch": 189.52642073778665,
      "grad_norm": 238.5287628173828,
      "learning_rate": 3.4732697194254924e-09,
      "loss": 24.3238,
      "step": 23880
    },
    {
      "epoch": 189.6061814556331,
      "grad_norm": 165.72174072265625,
      "learning_rate": 3.421587743074994e-09,
      "loss": 23.7922,
      "step": 23890
    },
    {
      "epoch": 189.68594217347956,
      "grad_norm": 871.0825805664062,
      "learning_rate": 3.3702905168426774e-09,
      "loss": 25.235,
      "step": 23900
    },
    {
      "epoch": 189.68594217347956,
      "eval_loss": 2.9857451915740967,
      "eval_mae": 1.2961997985839844,
      "eval_mse": 2.9857451915740967,
      "eval_r2": 0.10040956735610962,
      "eval_rmse": 1.7279308989580853,
      "eval_runtime": 9.0432,
      "eval_samples_per_second": 443.539,
      "eval_steps_per_second": 13.933,
      "step": 23900
    },
    {
      "epoch": 189.76570289132601,
      "grad_norm": 147.49085998535156,
      "learning_rate": 3.3193781207705163e-09,
      "loss": 23.5739,
      "step": 23910
    },
    {
      "epoch": 189.84546360917247,
      "grad_norm": 585.6244506835938,
      "learning_rate": 3.2688506342999355e-09,
      "loss": 23.4213,
      "step": 23920
    },
    {
      "epoch": 189.92522432701895,
      "grad_norm": 326.77081298828125,
      "learning_rate": 3.2187081362718415e-09,
      "loss": 22.6919,
      "step": 23930
    },
    {
      "epoch": 190.0,
      "grad_norm": 370.1246337890625,
      "learning_rate": 3.168950704926371e-09,
      "loss": 23.1201,
      "step": 23940
    },
    {
      "epoch": 190.07976071784645,
      "grad_norm": 268.3349914550781,
      "learning_rate": 3.1195784179028073e-09,
      "loss": 24.7828,
      "step": 23950
    },
    {
      "epoch": 190.1595214356929,
      "grad_norm": 339.0169982910156,
      "learning_rate": 3.0705913522395264e-09,
      "loss": 25.0338,
      "step": 23960
    },
    {
      "epoch": 190.2392821535394,
      "grad_norm": 451.1507873535156,
      "learning_rate": 3.021989584373802e-09,
      "loss": 23.9076,
      "step": 23970
    },
    {
      "epoch": 190.31904287138585,
      "grad_norm": 342.91912841796875,
      "learning_rate": 2.973773190141693e-09,
      "loss": 23.5402,
      "step": 23980
    },
    {
      "epoch": 190.3988035892323,
      "grad_norm": 587.4170532226562,
      "learning_rate": 2.925942244777879e-09,
      "loss": 25.0114,
      "step": 23990
    },
    {
      "epoch": 190.47856430707876,
      "grad_norm": 331.1037902832031,
      "learning_rate": 2.878496822915799e-09,
      "loss": 23.9958,
      "step": 24000
    },
    {
      "epoch": 190.47856430707876,
      "eval_loss": 2.985790967941284,
      "eval_mae": 1.2959747314453125,
      "eval_mse": 2.985790967941284,
      "eval_r2": 0.10039573907852173,
      "eval_rmse": 1.7279441449136266,
      "eval_runtime": 9.0531,
      "eval_samples_per_second": 443.055,
      "eval_steps_per_second": 13.918,
      "step": 24000
    },
    {
      "epoch": 190.5583250249252,
      "grad_norm": 328.3941955566406,
      "learning_rate": 2.8314369985870388e-09,
      "loss": 22.8027,
      "step": 24010
    },
    {
      "epoch": 190.6380857427717,
      "grad_norm": 265.7846374511719,
      "learning_rate": 2.784762845221805e-09,
      "loss": 24.2178,
      "step": 24020
    },
    {
      "epoch": 190.71784646061815,
      "grad_norm": 330.3060607910156,
      "learning_rate": 2.7384744356483412e-09,
      "loss": 23.3047,
      "step": 24030
    },
    {
      "epoch": 190.7976071784646,
      "grad_norm": 195.12123107910156,
      "learning_rate": 2.6925718420930944e-09,
      "loss": 23.3755,
      "step": 24040
    },
    {
      "epoch": 190.87736789631106,
      "grad_norm": 461.7795715332031,
      "learning_rate": 2.647055136180382e-09,
      "loss": 23.0593,
      "step": 24050
    },
    {
      "epoch": 190.95712861415754,
      "grad_norm": 245.96263122558594,
      "learning_rate": 2.6019243889325304e-09,
      "loss": 22.2629,
      "step": 24060
    },
    {
      "epoch": 191.0319042871386,
      "grad_norm": 347.748046875,
      "learning_rate": 2.557179670769516e-09,
      "loss": 23.5785,
      "step": 24070
    },
    {
      "epoch": 191.11166500498504,
      "grad_norm": 236.30970764160156,
      "learning_rate": 2.512821051509101e-09,
      "loss": 23.5184,
      "step": 24080
    },
    {
      "epoch": 191.1914257228315,
      "grad_norm": 953.8068237304688,
      "learning_rate": 2.468848600366502e-09,
      "loss": 22.7987,
      "step": 24090
    },
    {
      "epoch": 191.27118644067798,
      "grad_norm": 379.2378234863281,
      "learning_rate": 2.425262385954363e-09,
      "loss": 24.1852,
      "step": 24100
    },
    {
      "epoch": 191.27118644067798,
      "eval_loss": 2.985538959503174,
      "eval_mae": 1.2961422204971313,
      "eval_mse": 2.985538959503174,
      "eval_r2": 0.10047167539596558,
      "eval_rmse": 1.727871221909542,
      "eval_runtime": 9.0399,
      "eval_samples_per_second": 443.697,
      "eval_steps_per_second": 13.938,
      "step": 24100
    },
    {
      "epoch": 191.35094715852443,
      "grad_norm": 672.7747802734375,
      "learning_rate": 2.3820624762827535e-09,
      "loss": 23.1245,
      "step": 24110
    },
    {
      "epoch": 191.4307078763709,
      "grad_norm": 362.42791748046875,
      "learning_rate": 2.3392489387588366e-09,
      "loss": 25.7611,
      "step": 24120
    },
    {
      "epoch": 191.51046859421734,
      "grad_norm": 372.2131042480469,
      "learning_rate": 2.296821840187063e-09,
      "loss": 24.1331,
      "step": 24130
    },
    {
      "epoch": 191.5902293120638,
      "grad_norm": 323.49664306640625,
      "learning_rate": 2.2547812467687553e-09,
      "loss": 22.3181,
      "step": 24140
    },
    {
      "epoch": 191.66999002991028,
      "grad_norm": 186.12277221679688,
      "learning_rate": 2.2131272241022457e-09,
      "loss": 24.1584,
      "step": 24150
    },
    {
      "epoch": 191.74975074775674,
      "grad_norm": 604.3494873046875,
      "learning_rate": 2.171859837182599e-09,
      "loss": 23.3094,
      "step": 24160
    },
    {
      "epoch": 191.8295114656032,
      "grad_norm": 772.862060546875,
      "learning_rate": 2.1309791504016683e-09,
      "loss": 23.8219,
      "step": 24170
    },
    {
      "epoch": 191.90927218344964,
      "grad_norm": 657.0350341796875,
      "learning_rate": 2.0904852275478722e-09,
      "loss": 22.7549,
      "step": 24180
    },
    {
      "epoch": 191.9890329012961,
      "grad_norm": 277.4106750488281,
      "learning_rate": 2.050378131806113e-09,
      "loss": 24.5491,
      "step": 24190
    },
    {
      "epoch": 192.06380857427718,
      "grad_norm": 1266.8355712890625,
      "learning_rate": 2.010657925757803e-09,
      "loss": 22.8491,
      "step": 24200
    },
    {
      "epoch": 192.06380857427718,
      "eval_loss": 2.986049175262451,
      "eval_mae": 1.2961747646331787,
      "eval_mse": 2.986049175262451,
      "eval_r2": 0.10031801462173462,
      "eval_rmse": 1.7280188584799794,
      "eval_runtime": 9.0269,
      "eval_samples_per_second": 444.338,
      "eval_steps_per_second": 13.958,
      "step": 24200
    },
    {
      "epoch": 192.14356929212363,
      "grad_norm": 282.1344299316406,
      "learning_rate": 1.9713246713805587e-09,
      "loss": 23.7891,
      "step": 24210
    },
    {
      "epoch": 192.22333000997008,
      "grad_norm": 235.01341247558594,
      "learning_rate": 1.932378430048259e-09,
      "loss": 23.2705,
      "step": 24220
    },
    {
      "epoch": 192.30309072781654,
      "grad_norm": 218.00794982910156,
      "learning_rate": 1.8938192625309324e-09,
      "loss": 23.3808,
      "step": 24230
    },
    {
      "epoch": 192.38285144566302,
      "grad_norm": 204.65672302246094,
      "learning_rate": 1.8556472289946167e-09,
      "loss": 22.4473,
      "step": 24240
    },
    {
      "epoch": 192.46261216350948,
      "grad_norm": 563.5921630859375,
      "learning_rate": 1.8178623890012512e-09,
      "loss": 24.5627,
      "step": 24250
    },
    {
      "epoch": 192.54237288135593,
      "grad_norm": 179.89076232910156,
      "learning_rate": 1.7804648015086465e-09,
      "loss": 24.8393,
      "step": 24260
    },
    {
      "epoch": 192.62213359920239,
      "grad_norm": 831.5360107421875,
      "learning_rate": 1.7434545248703469e-09,
      "loss": 24.1278,
      "step": 24270
    },
    {
      "epoch": 192.70189431704884,
      "grad_norm": 478.6527099609375,
      "learning_rate": 1.7068316168355746e-09,
      "loss": 22.7936,
      "step": 24280
    },
    {
      "epoch": 192.78165503489532,
      "grad_norm": 280.7521667480469,
      "learning_rate": 1.6705961345491182e-09,
      "loss": 21.1134,
      "step": 24290
    },
    {
      "epoch": 192.86141575274178,
      "grad_norm": 206.8756103515625,
      "learning_rate": 1.6347481345511948e-09,
      "loss": 25.9234,
      "step": 24300
    },
    {
      "epoch": 192.86141575274178,
      "eval_loss": 2.9849839210510254,
      "eval_mae": 1.2959469556808472,
      "eval_mse": 2.9849836826324463,
      "eval_r2": 0.1006389856338501,
      "eval_rmse": 1.7277105320719806,
      "eval_runtime": 9.0473,
      "eval_samples_per_second": 443.336,
      "eval_steps_per_second": 13.927,
      "step": 24300
    },
    {
      "epoch": 192.94117647058823,
      "grad_norm": 380.0176086425781,
      "learning_rate": 1.5992876727774774e-09,
      "loss": 25.2448,
      "step": 24310
    },
    {
      "epoch": 193.01595214356928,
      "grad_norm": 120.8132095336914,
      "learning_rate": 1.5642148045588999e-09,
      "loss": 21.4254,
      "step": 24320
    },
    {
      "epoch": 193.09571286141576,
      "grad_norm": 375.27459716796875,
      "learning_rate": 1.5295295846216582e-09,
      "loss": 25.4883,
      "step": 24330
    },
    {
      "epoch": 193.17547357926222,
      "grad_norm": 470.91632080078125,
      "learning_rate": 1.4952320670869879e-09,
      "loss": 23.6084,
      "step": 24340
    },
    {
      "epoch": 193.25523429710867,
      "grad_norm": 140.60874938964844,
      "learning_rate": 1.4613223054712466e-09,
      "loss": 22.8232,
      "step": 24350
    },
    {
      "epoch": 193.33499501495513,
      "grad_norm": 120.59532165527344,
      "learning_rate": 1.4278003526857484e-09,
      "loss": 24.3427,
      "step": 24360
    },
    {
      "epoch": 193.4147557328016,
      "grad_norm": 519.9879150390625,
      "learning_rate": 1.3946662610367088e-09,
      "loss": 22.8856,
      "step": 24370
    },
    {
      "epoch": 193.49451645064806,
      "grad_norm": 597.9413452148438,
      "learning_rate": 1.3619200822250488e-09,
      "loss": 21.1711,
      "step": 24380
    },
    {
      "epoch": 193.57427716849452,
      "grad_norm": 187.56488037109375,
      "learning_rate": 1.3295618673464792e-09,
      "loss": 24.6198,
      "step": 24390
    },
    {
      "epoch": 193.65403788634097,
      "grad_norm": 556.2128295898438,
      "learning_rate": 1.2975916668913345e-09,
      "loss": 24.2869,
      "step": 24400
    },
    {
      "epoch": 193.65403788634097,
      "eval_loss": 2.985672950744629,
      "eval_mae": 1.296151041984558,
      "eval_mse": 2.98567271232605,
      "eval_r2": 0.10043138265609741,
      "eval_rmse": 1.7279099259874775,
      "eval_runtime": 9.017,
      "eval_samples_per_second": 444.829,
      "eval_steps_per_second": 13.974,
      "step": 24400
    },
    {
      "epoch": 193.73379860418743,
      "grad_norm": 558.5076293945312,
      "learning_rate": 1.2660095307445718e-09,
      "loss": 23.3984,
      "step": 24410
    },
    {
      "epoch": 193.8135593220339,
      "grad_norm": 256.103271484375,
      "learning_rate": 1.2348155081854661e-09,
      "loss": 24.0415,
      "step": 24420
    },
    {
      "epoch": 193.89332003988036,
      "grad_norm": 276.30682373046875,
      "learning_rate": 1.2040096478878881e-09,
      "loss": 25.1056,
      "step": 24430
    },
    {
      "epoch": 193.97308075772682,
      "grad_norm": 461.1793212890625,
      "learning_rate": 1.1735919979198871e-09,
      "loss": 23.1343,
      "step": 24440
    },
    {
      "epoch": 194.04785643070787,
      "grad_norm": 491.55267333984375,
      "learning_rate": 1.1435626057438863e-09,
      "loss": 23.5858,
      "step": 24450
    },
    {
      "epoch": 194.12761714855435,
      "grad_norm": 799.1927490234375,
      "learning_rate": 1.1139215182163765e-09,
      "loss": 26.8489,
      "step": 24460
    },
    {
      "epoch": 194.2073778664008,
      "grad_norm": 270.9708251953125,
      "learning_rate": 1.0846687815880274e-09,
      "loss": 22.9998,
      "step": 24470
    },
    {
      "epoch": 194.28713858424726,
      "grad_norm": 754.822509765625,
      "learning_rate": 1.0558044415035217e-09,
      "loss": 24.6583,
      "step": 24480
    },
    {
      "epoch": 194.3668993020937,
      "grad_norm": 329.4441223144531,
      "learning_rate": 1.0273285430015265e-09,
      "loss": 24.6196,
      "step": 24490
    },
    {
      "epoch": 194.44666001994017,
      "grad_norm": 203.95399475097656,
      "learning_rate": 9.992411305145277e-10,
      "loss": 24.4206,
      "step": 24500
    },
    {
      "epoch": 194.44666001994017,
      "eval_loss": 2.9850587844848633,
      "eval_mae": 1.2959555387496948,
      "eval_mse": 2.9850587844848633,
      "eval_r2": 0.10061639547348022,
      "eval_rmse": 1.7277322664362276,
      "eval_runtime": 9.0375,
      "eval_samples_per_second": 443.819,
      "eval_steps_per_second": 13.942,
      "step": 24500
    },
    {
      "epoch": 194.52642073778665,
      "grad_norm": 506.1182556152344,
      "learning_rate": 9.715422478689673e-10,
      "loss": 22.4749,
      "step": 24510
    },
    {
      "epoch": 194.6061814556331,
      "grad_norm": 546.7146606445312,
      "learning_rate": 9.442319382849118e-10,
      "loss": 21.9358,
      "step": 24520
    },
    {
      "epoch": 194.68594217347956,
      "grad_norm": 1289.88427734375,
      "learning_rate": 9.173102443761627e-10,
      "loss": 22.4655,
      "step": 24530
    },
    {
      "epoch": 194.76570289132601,
      "grad_norm": 116.09934997558594,
      "learning_rate": 8.907772081502008e-10,
      "loss": 22.6633,
      "step": 24540
    },
    {
      "epoch": 194.84546360917247,
      "grad_norm": 1326.2725830078125,
      "learning_rate": 8.646328710079364e-10,
      "loss": 24.1759,
      "step": 24550
    },
    {
      "epoch": 194.92522432701895,
      "grad_norm": 685.0480346679688,
      "learning_rate": 8.388772737439043e-10,
      "loss": 24.5467,
      "step": 24560
    },
    {
      "epoch": 195.0,
      "grad_norm": 294.5447998046875,
      "learning_rate": 8.135104565459849e-10,
      "loss": 21.4338,
      "step": 24570
    },
    {
      "epoch": 195.07976071784645,
      "grad_norm": 447.23480224609375,
      "learning_rate": 7.885324589954612e-10,
      "loss": 23.7994,
      "step": 24580
    },
    {
      "epoch": 195.1595214356929,
      "grad_norm": 302.0307312011719,
      "learning_rate": 7.63943320066851e-10,
      "loss": 23.8053,
      "step": 24590
    },
    {
      "epoch": 195.2392821535394,
      "grad_norm": 1048.637451171875,
      "learning_rate": 7.39743078128019e-10,
      "loss": 25.6922,
      "step": 24600
    },
    {
      "epoch": 195.2392821535394,
      "eval_loss": 2.9844701290130615,
      "eval_mae": 1.2956593036651611,
      "eval_mse": 2.9844703674316406,
      "eval_r2": 0.10079365968704224,
      "eval_rmse": 1.7275619720958322,
      "eval_runtime": 9.093,
      "eval_samples_per_second": 441.106,
      "eval_steps_per_second": 13.857,
      "step": 24600
    },
    {
      "epoch": 195.31904287138585,
      "grad_norm": 213.0170135498047,
      "learning_rate": 7.159317709399537e-10,
      "loss": 24.5134,
      "step": 24610
    },
    {
      "epoch": 195.3988035892323,
      "grad_norm": 996.3701782226562,
      "learning_rate": 6.925094356567685e-10,
      "loss": 24.2102,
      "step": 24620
    },
    {
      "epoch": 195.47856430707876,
      "grad_norm": 212.89291381835938,
      "learning_rate": 6.694761088256173e-10,
      "loss": 23.6198,
      "step": 24630
    },
    {
      "epoch": 195.5583250249252,
      "grad_norm": 202.31045532226562,
      "learning_rate": 6.468318263867234e-10,
      "loss": 27.476,
      "step": 24640
    },
    {
      "epoch": 195.6380857427717,
      "grad_norm": 427.94854736328125,
      "learning_rate": 6.24576623673212e-10,
      "loss": 22.7069,
      "step": 24650
    },
    {
      "epoch": 195.71784646061815,
      "grad_norm": 187.37359619140625,
      "learning_rate": 6.027105354111383e-10,
      "loss": 23.5491,
      "step": 24660
    },
    {
      "epoch": 195.7976071784646,
      "grad_norm": 397.1957702636719,
      "learning_rate": 5.812335957193493e-10,
      "loss": 22.5446,
      "step": 24670
    },
    {
      "epoch": 195.87736789631106,
      "grad_norm": 271.0002136230469,
      "learning_rate": 5.601458381095658e-10,
      "loss": 22.1646,
      "step": 24680
    },
    {
      "epoch": 195.95712861415754,
      "grad_norm": 262.593017578125,
      "learning_rate": 5.394472954861895e-10,
      "loss": 22.4295,
      "step": 24690
    },
    {
      "epoch": 196.0319042871386,
      "grad_norm": 190.5734100341797,
      "learning_rate": 5.191380001463019e-10,
      "loss": 20.8607,
      "step": 24700
    },
    {
      "epoch": 196.0319042871386,
      "eval_loss": 2.985541820526123,
      "eval_mae": 1.295915126800537,
      "eval_mse": 2.985541820526123,
      "eval_r2": 0.1004709005355835,
      "eval_rmse": 1.7278720498133313,
      "eval_runtime": 9.0672,
      "eval_samples_per_second": 442.361,
      "eval_steps_per_second": 13.896,
      "step": 24700
    },
    {
      "epoch": 196.11166500498504,
      "grad_norm": 480.9814758300781,
      "learning_rate": 4.992179837796373e-10,
      "loss": 22.7887,
      "step": 24710
    },
    {
      "epoch": 196.1914257228315,
      "grad_norm": 457.35003662109375,
      "learning_rate": 4.79687277468499e-10,
      "loss": 21.2072,
      "step": 24720
    },
    {
      "epoch": 196.27118644067798,
      "grad_norm": 168.44793701171875,
      "learning_rate": 4.6054591168778744e-10,
      "loss": 25.3187,
      "step": 24730
    },
    {
      "epoch": 196.35094715852443,
      "grad_norm": 497.0190124511719,
      "learning_rate": 4.4179391630480566e-10,
      "loss": 22.2018,
      "step": 24740
    },
    {
      "epoch": 196.4307078763709,
      "grad_norm": 282.4808349609375,
      "learning_rate": 4.234313205793427e-10,
      "loss": 24.2384,
      "step": 24750
    },
    {
      "epoch": 196.51046859421734,
      "grad_norm": 573.6471557617188,
      "learning_rate": 4.054581531636181e-10,
      "loss": 24.1769,
      "step": 24760
    },
    {
      "epoch": 196.5902293120638,
      "grad_norm": 178.6867218017578,
      "learning_rate": 3.8787444210214294e-10,
      "loss": 24.7551,
      "step": 24770
    },
    {
      "epoch": 196.66999002991028,
      "grad_norm": 234.75233459472656,
      "learning_rate": 3.7068021483180335e-10,
      "loss": 22.6184,
      "step": 24780
    },
    {
      "epoch": 196.74975074775674,
      "grad_norm": 255.20985412597656,
      "learning_rate": 3.538754981816661e-10,
      "loss": 25.1924,
      "step": 24790
    },
    {
      "epoch": 196.8295114656032,
      "grad_norm": 559.8621215820312,
      "learning_rate": 3.3746031837311727e-10,
      "loss": 24.9777,
      "step": 24800
    },
    {
      "epoch": 196.8295114656032,
      "eval_loss": 2.9856958389282227,
      "eval_mae": 1.2961050271987915,
      "eval_mse": 2.9856958389282227,
      "eval_r2": 0.10042446851730347,
      "eval_rmse": 1.7279166180485164,
      "eval_runtime": 9.0989,
      "eval_samples_per_second": 440.823,
      "eval_steps_per_second": 13.848,
      "step": 24800
    },
    {
      "epoch": 196.90927218344964,
      "grad_norm": 148.0769500732422,
      "learning_rate": 3.2143470101964056e-10,
      "loss": 23.6349,
      "step": 24810
    },
    {
      "epoch": 196.9890329012961,
      "grad_norm": 245.9391326904297,
      "learning_rate": 3.0579867112695556e-10,
      "loss": 24.8288,
      "step": 24820
    },
    {
      "epoch": 197.06380857427718,
      "grad_norm": 774.696044921875,
      "learning_rate": 2.9055225309276843e-10,
      "loss": 21.4276,
      "step": 24830
    },
    {
      "epoch": 197.14356929212363,
      "grad_norm": 493.7986145019531,
      "learning_rate": 2.756954707069936e-10,
      "loss": 22.5439,
      "step": 24840
    },
    {
      "epoch": 197.22333000997008,
      "grad_norm": 185.34024047851562,
      "learning_rate": 2.612283471514487e-10,
      "loss": 24.3253,
      "step": 24850
    },
    {
      "epoch": 197.30309072781654,
      "grad_norm": 507.3541564941406,
      "learning_rate": 2.4715090500004867e-10,
      "loss": 24.4285,
      "step": 24860
    },
    {
      "epoch": 197.38285144566302,
      "grad_norm": 680.4122314453125,
      "learning_rate": 2.33463166218556e-10,
      "loss": 23.7532,
      "step": 24870
    },
    {
      "epoch": 197.46261216350948,
      "grad_norm": 346.16668701171875,
      "learning_rate": 2.201651521647474e-10,
      "loss": 23.7976,
      "step": 24880
    },
    {
      "epoch": 197.54237288135593,
      "grad_norm": 443.5713806152344,
      "learning_rate": 2.0725688358824712e-10,
      "loss": 23.2051,
      "step": 24890
    },
    {
      "epoch": 197.62213359920239,
      "grad_norm": 155.51341247558594,
      "learning_rate": 1.9473838063058246e-10,
      "loss": 25.3578,
      "step": 24900
    },
    {
      "epoch": 197.62213359920239,
      "eval_loss": 2.9856114387512207,
      "eval_mae": 1.2960511445999146,
      "eval_mse": 2.9856114387512207,
      "eval_r2": 0.10044991970062256,
      "eval_rmse": 1.7278921953499358,
      "eval_runtime": 9.1052,
      "eval_samples_per_second": 440.517,
      "eval_steps_per_second": 13.838,
      "step": 24900
    },
    {
      "epoch": 197.70189431704884,
      "grad_norm": 227.46690368652344,
      "learning_rate": 1.826096628250451e-10,
      "loss": 26.1419,
      "step": 24910
    },
    {
      "epoch": 197.78165503489532,
      "grad_norm": 171.3894805908203,
      "learning_rate": 1.708707490967465e-10,
      "loss": 21.9025,
      "step": 24920
    },
    {
      "epoch": 197.86141575274178,
      "grad_norm": 258.0511779785156,
      "learning_rate": 1.595216577625902e-10,
      "loss": 24.4516,
      "step": 24930
    },
    {
      "epoch": 197.94117647058823,
      "grad_norm": 137.69100952148438,
      "learning_rate": 1.4856240653118857e-10,
      "loss": 24.0129,
      "step": 24940
    },
    {
      "epoch": 198.01595214356928,
      "grad_norm": 142.23512268066406,
      "learning_rate": 1.3799301250291828e-10,
      "loss": 20.7065,
      "step": 24950
    },
    {
      "epoch": 198.09571286141576,
      "grad_norm": 110.43464660644531,
      "learning_rate": 1.2781349216972603e-10,
      "loss": 22.97,
      "step": 24960
    },
    {
      "epoch": 198.17547357926222,
      "grad_norm": 322.2825622558594,
      "learning_rate": 1.1802386141532283e-10,
      "loss": 23.2948,
      "step": 24970
    },
    {
      "epoch": 198.25523429710867,
      "grad_norm": 394.9427490234375,
      "learning_rate": 1.0862413551501748e-10,
      "loss": 23.8814,
      "step": 24980
    },
    {
      "epoch": 198.33499501495513,
      "grad_norm": 480.8127136230469,
      "learning_rate": 9.96143291357443e-11,
      "loss": 25.1555,
      "step": 24990
    },
    {
      "epoch": 198.4147557328016,
      "grad_norm": 171.42910766601562,
      "learning_rate": 9.09944563359799e-11,
      "loss": 23.0883,
      "step": 25000
    },
    {
      "epoch": 198.4147557328016,
      "eval_loss": 2.985551357269287,
      "eval_mae": 1.296038031578064,
      "eval_mse": 2.985551118850708,
      "eval_r2": 0.10046803951263428,
      "eval_rmse": 1.7278747404979067,
      "eval_runtime": 9.1263,
      "eval_samples_per_second": 439.499,
      "eval_steps_per_second": 13.806,
      "step": 25000
    },
    {
      "epoch": 198.49451645064806,
      "grad_norm": 1028.7611083984375,
      "learning_rate": 8.276453056582645e-11,
      "loss": 22.364,
      "step": 25010
    },
    {
      "epoch": 198.57427716849452,
      "grad_norm": 570.9407348632812,
      "learning_rate": 7.49245646668728e-11,
      "loss": 21.5703,
      "step": 25020
    },
    {
      "epoch": 198.65403788634097,
      "grad_norm": 351.5626220703125,
      "learning_rate": 6.747457087227792e-11,
      "loss": 26.2493,
      "step": 25030
    },
    {
      "epoch": 198.73379860418743,
      "grad_norm": 409.9494934082031,
      "learning_rate": 6.041456080665974e-11,
      "loss": 23.5306,
      "step": 25040
    },
    {
      "epoch": 198.8135593220339,
      "grad_norm": 191.91433715820312,
      "learning_rate": 5.374454548617846e-11,
      "loss": 23.6993,
      "step": 25050
    },
    {
      "epoch": 198.89332003988036,
      "grad_norm": 383.2036437988281,
      "learning_rate": 4.746453531839778e-11,
      "loss": 24.7173,
      "step": 25060
    },
    {
      "epoch": 198.97308075772682,
      "grad_norm": 282.2154846191406,
      "learning_rate": 4.1574540102395914e-11,
      "loss": 24.869,
      "step": 25070
    },
    {
      "epoch": 199.04785643070787,
      "grad_norm": 209.2111358642578,
      "learning_rate": 3.6074569028654576e-11,
      "loss": 23.1518,
      "step": 25080
    },
    {
      "epoch": 199.12761714855435,
      "grad_norm": 125.33819580078125,
      "learning_rate": 3.096463067905897e-11,
      "loss": 27.0382,
      "step": 25090
    },
    {
      "epoch": 199.2073778664008,
      "grad_norm": 224.3722381591797,
      "learning_rate": 2.6244733027008804e-11,
      "loss": 23.0602,
      "step": 25100
    },
    {
      "epoch": 199.2073778664008,
      "eval_loss": 2.985177755355835,
      "eval_mae": 1.2959330081939697,
      "eval_mse": 2.985177755355835,
      "eval_r2": 0.10058057308197021,
      "eval_rmse": 1.7277666958695075,
      "eval_runtime": 9.1276,
      "eval_samples_per_second": 439.435,
      "eval_steps_per_second": 13.804,
      "step": 25100
    },
    {
      "epoch": 199.28713858424726,
      "grad_norm": 906.6569213867188,
      "learning_rate": 2.1914883437140762e-11,
      "loss": 25.4792,
      "step": 25110
    },
    {
      "epoch": 199.3668993020937,
      "grad_norm": 139.93795776367188,
      "learning_rate": 1.797508866560604e-11,
      "loss": 21.875,
      "step": 25120
    },
    {
      "epoch": 199.44666001994017,
      "grad_norm": 412.5982360839844,
      "learning_rate": 1.442535485987606e-11,
      "loss": 23.1991,
      "step": 25130
    },
    {
      "epoch": 199.52642073778665,
      "grad_norm": 304.7872314453125,
      "learning_rate": 1.1265687558797976e-11,
      "loss": 23.1054,
      "step": 25140
    },
    {
      "epoch": 199.6061814556331,
      "grad_norm": 691.2601928710938,
      "learning_rate": 8.496091692594687e-12,
      "loss": 23.7589,
      "step": 25150
    },
    {
      "epoch": 199.68594217347956,
      "grad_norm": 180.51930236816406,
      "learning_rate": 6.116571582809316e-12,
      "loss": 23.4738,
      "step": 25160
    },
    {
      "epoch": 199.76570289132601,
      "grad_norm": 635.4140014648438,
      "learning_rate": 4.127130942332968e-12,
      "loss": 21.3643,
      "step": 25170
    },
    {
      "epoch": 199.84546360917247,
      "grad_norm": 272.2777099609375,
      "learning_rate": 2.5277728754047323e-12,
      "loss": 22.6734,
      "step": 25180
    },
    {
      "epoch": 199.92522432701895,
      "grad_norm": 191.4556121826172,
      "learning_rate": 1.3184998775839228e-12,
      "loss": 23.7552,
      "step": 25190
    },
    {
      "epoch": 200.0,
      "grad_norm": 638.1273193359375,
      "learning_rate": 4.993138358055926e-13,
      "loss": 24.4562,
      "step": 25200
    },
    {
      "epoch": 200.0,
      "eval_loss": 2.9852001667022705,
      "eval_mae": 1.2959332466125488,
      "eval_mse": 2.9852004051208496,
      "eval_r2": 0.10057371854782104,
      "eval_rmse": 1.7277732504934928,
      "eval_runtime": 9.1119,
      "eval_samples_per_second": 440.193,
      "eval_steps_per_second": 13.828,
      "step": 25200
    },
    {
      "epoch": 200.07976071784645,
      "grad_norm": 233.1142578125,
      "learning_rate": 3.276812958733593e-07,
      "loss": 23.9068,
      "step": 25210
    },
    {
      "epoch": 200.1595214356929,
      "grad_norm": 264.1903991699219,
      "learning_rate": 3.275626967218783e-07,
      "loss": 22.5422,
      "step": 25220
    },
    {
      "epoch": 200.2392821535394,
      "grad_norm": 269.2657165527344,
      "learning_rate": 3.274440782524416e-07,
      "loss": 25.6559,
      "step": 25230
    },
    {
      "epoch": 200.31904287138585,
      "grad_norm": 1040.9169921875,
      "learning_rate": 3.273254404945924e-07,
      "loss": 24.3639,
      "step": 25240
    },
    {
      "epoch": 200.3988035892323,
      "grad_norm": 264.42822265625,
      "learning_rate": 3.272067834778791e-07,
      "loss": 23.6492,
      "step": 25250
    },
    {
      "epoch": 200.47856430707876,
      "grad_norm": 1180.909912109375,
      "learning_rate": 3.2708810723185464e-07,
      "loss": 24.3143,
      "step": 25260
    },
    {
      "epoch": 200.5583250249252,
      "grad_norm": 173.80987548828125,
      "learning_rate": 3.2696941178607686e-07,
      "loss": 22.7227,
      "step": 25270
    },
    {
      "epoch": 200.6380857427717,
      "grad_norm": 121.28042602539062,
      "learning_rate": 3.2685069717010823e-07,
      "loss": 25.8901,
      "step": 25280
    },
    {
      "epoch": 200.71784646061815,
      "grad_norm": 490.2680969238281,
      "learning_rate": 3.267319634135162e-07,
      "loss": 23.3299,
      "step": 25290
    },
    {
      "epoch": 200.7976071784646,
      "grad_norm": 197.51051330566406,
      "learning_rate": 3.2661321054587275e-07,
      "loss": 24.8621,
      "step": 25300
    },
    {
      "epoch": 200.7976071784646,
      "eval_loss": 2.987062454223633,
      "eval_mae": 1.301984429359436,
      "eval_mse": 2.987062454223633,
      "eval_r2": 0.10001271963119507,
      "eval_rmse": 1.7283120245556451,
      "eval_runtime": 9.1207,
      "eval_samples_per_second": 439.769,
      "eval_steps_per_second": 13.815,
      "step": 25300
    },
    {
      "epoch": 200.87736789631106,
      "grad_norm": 230.46575927734375,
      "learning_rate": 3.26494438596755e-07,
      "loss": 22.8219,
      "step": 25310
    },
    {
      "epoch": 200.95712861415754,
      "grad_norm": 307.9410400390625,
      "learning_rate": 3.263756475957444e-07,
      "loss": 22.7135,
      "step": 25320
    },
    {
      "epoch": 201.0319042871386,
      "grad_norm": 593.2626953125,
      "learning_rate": 3.262568375724274e-07,
      "loss": 20.4246,
      "step": 25330
    },
    {
      "epoch": 201.11166500498504,
      "grad_norm": 1156.7459716796875,
      "learning_rate": 3.2613800855639504e-07,
      "loss": 25.0621,
      "step": 25340
    },
    {
      "epoch": 201.1914257228315,
      "grad_norm": 719.022216796875,
      "learning_rate": 3.2601916057724335e-07,
      "loss": 26.0972,
      "step": 25350
    },
    {
      "epoch": 201.27118644067798,
      "grad_norm": 505.98260498046875,
      "learning_rate": 3.2590029366457266e-07,
      "loss": 23.9704,
      "step": 25360
    },
    {
      "epoch": 201.35094715852443,
      "grad_norm": 354.1245422363281,
      "learning_rate": 3.257814078479884e-07,
      "loss": 23.2421,
      "step": 25370
    },
    {
      "epoch": 201.4307078763709,
      "grad_norm": 282.1812744140625,
      "learning_rate": 3.256625031571005e-07,
      "loss": 21.8984,
      "step": 25380
    },
    {
      "epoch": 201.51046859421734,
      "grad_norm": 676.1168212890625,
      "learning_rate": 3.2554357962152377e-07,
      "loss": 23.0126,
      "step": 25390
    },
    {
      "epoch": 201.5902293120638,
      "grad_norm": 1098.18212890625,
      "learning_rate": 3.254246372708775e-07,
      "loss": 25.1928,
      "step": 25400
    },
    {
      "epoch": 201.5902293120638,
      "eval_loss": 3.001655101776123,
      "eval_mae": 1.3126819133758545,
      "eval_mse": 3.001655101776123,
      "eval_r2": 0.09561598300933838,
      "eval_rmse": 1.7325285284162344,
      "eval_runtime": 9.1184,
      "eval_samples_per_second": 439.878,
      "eval_steps_per_second": 13.818,
      "step": 25400
    },
    {
      "epoch": 201.66999002991028,
      "grad_norm": 400.4809265136719,
      "learning_rate": 3.2530567613478574e-07,
      "loss": 23.2146,
      "step": 25410
    },
    {
      "epoch": 201.74975074775674,
      "grad_norm": 1427.6348876953125,
      "learning_rate": 3.2518669624287727e-07,
      "loss": 24.413,
      "step": 25420
    },
    {
      "epoch": 201.8295114656032,
      "grad_norm": 258.5172119140625,
      "learning_rate": 3.250676976247856e-07,
      "loss": 24.6172,
      "step": 25430
    },
    {
      "epoch": 201.90927218344964,
      "grad_norm": 230.9508056640625,
      "learning_rate": 3.2494868031014867e-07,
      "loss": 23.6625,
      "step": 25440
    },
    {
      "epoch": 201.9890329012961,
      "grad_norm": 410.1961364746094,
      "learning_rate": 3.2482964432860925e-07,
      "loss": 23.1106,
      "step": 25450
    },
    {
      "epoch": 202.06380857427718,
      "grad_norm": 651.6411743164062,
      "learning_rate": 3.2471058970981485e-07,
      "loss": 21.1852,
      "step": 25460
    },
    {
      "epoch": 202.14356929212363,
      "grad_norm": 297.3800354003906,
      "learning_rate": 3.2459151648341755e-07,
      "loss": 25.5979,
      "step": 25470
    },
    {
      "epoch": 202.22333000997008,
      "grad_norm": 132.6808624267578,
      "learning_rate": 3.244724246790738e-07,
      "loss": 24.2881,
      "step": 25480
    },
    {
      "epoch": 202.30309072781654,
      "grad_norm": 289.0632629394531,
      "learning_rate": 3.24353314326445e-07,
      "loss": 23.7794,
      "step": 25490
    },
    {
      "epoch": 202.38285144566302,
      "grad_norm": 856.2789306640625,
      "learning_rate": 3.24234185455197e-07,
      "loss": 22.0349,
      "step": 25500
    },
    {
      "epoch": 202.38285144566302,
      "eval_loss": 2.987736940383911,
      "eval_mae": 1.2919999361038208,
      "eval_mse": 2.987736940383911,
      "eval_r2": 0.09980946779251099,
      "eval_rmse": 1.7285071421269602,
      "eval_runtime": 9.1273,
      "eval_samples_per_second": 439.453,
      "eval_steps_per_second": 13.805,
      "step": 25500
    },
    {
      "epoch": 202.46261216350948,
      "grad_norm": 352.2939758300781,
      "learning_rate": 3.241150380950006e-07,
      "loss": 23.3271,
      "step": 25510
    },
    {
      "epoch": 202.54237288135593,
      "grad_norm": 520.0399169921875,
      "learning_rate": 3.239958722755306e-07,
      "loss": 24.8606,
      "step": 25520
    },
    {
      "epoch": 202.62213359920239,
      "grad_norm": 606.8839721679688,
      "learning_rate": 3.2387668802646683e-07,
      "loss": 23.3053,
      "step": 25530
    },
    {
      "epoch": 202.70189431704884,
      "grad_norm": 356.62353515625,
      "learning_rate": 3.2375748537749375e-07,
      "loss": 22.8932,
      "step": 25540
    },
    {
      "epoch": 202.78165503489532,
      "grad_norm": 222.2799530029297,
      "learning_rate": 3.2363826435830013e-07,
      "loss": 24.2675,
      "step": 25550
    },
    {
      "epoch": 202.86141575274178,
      "grad_norm": 1185.181396484375,
      "learning_rate": 3.235190249985795e-07,
      "loss": 22.7562,
      "step": 25560
    },
    {
      "epoch": 202.94117647058823,
      "grad_norm": 575.6671142578125,
      "learning_rate": 3.2339976732802987e-07,
      "loss": 24.7696,
      "step": 25570
    },
    {
      "epoch": 203.01595214356928,
      "grad_norm": 172.9219207763672,
      "learning_rate": 3.232804913763539e-07,
      "loss": 24.11,
      "step": 25580
    },
    {
      "epoch": 203.09571286141576,
      "grad_norm": 1075.917724609375,
      "learning_rate": 3.2316119717325873e-07,
      "loss": 21.5316,
      "step": 25590
    },
    {
      "epoch": 203.17547357926222,
      "grad_norm": 695.1358642578125,
      "learning_rate": 3.2304188474845606e-07,
      "loss": 25.1565,
      "step": 25600
    },
    {
      "epoch": 203.17547357926222,
      "eval_loss": 2.9972896575927734,
      "eval_mae": 1.2894809246063232,
      "eval_mse": 2.9972894191741943,
      "eval_r2": 0.0969313383102417,
      "eval_rmse": 1.7312681534569376,
      "eval_runtime": 9.1254,
      "eval_samples_per_second": 439.541,
      "eval_steps_per_second": 13.808,
      "step": 25600
    },
    {
      "epoch": 203.25523429710867,
      "grad_norm": 467.5833435058594,
      "learning_rate": 3.2292255413166204e-07,
      "loss": 24.2298,
      "step": 25610
    },
    {
      "epoch": 203.33499501495513,
      "grad_norm": 829.6505737304688,
      "learning_rate": 3.2280320535259757e-07,
      "loss": 22.8333,
      "step": 25620
    },
    {
      "epoch": 203.4147557328016,
      "grad_norm": 712.6729736328125,
      "learning_rate": 3.226838384409881e-07,
      "loss": 23.4115,
      "step": 25630
    },
    {
      "epoch": 203.49451645064806,
      "grad_norm": 341.8227844238281,
      "learning_rate": 3.2256445342656304e-07,
      "loss": 22.5,
      "step": 25640
    },
    {
      "epoch": 203.57427716849452,
      "grad_norm": 390.042724609375,
      "learning_rate": 3.2244505033905704e-07,
      "loss": 24.5384,
      "step": 25650
    },
    {
      "epoch": 203.65403788634097,
      "grad_norm": 874.4033813476562,
      "learning_rate": 3.223256292082088e-07,
      "loss": 25.1737,
      "step": 25660
    },
    {
      "epoch": 203.73379860418743,
      "grad_norm": 768.7593383789062,
      "learning_rate": 3.222061900637618e-07,
      "loss": 24.8873,
      "step": 25670
    },
    {
      "epoch": 203.8135593220339,
      "grad_norm": 324.7832946777344,
      "learning_rate": 3.2208673293546356e-07,
      "loss": 24.5665,
      "step": 25680
    },
    {
      "epoch": 203.89332003988036,
      "grad_norm": 377.18609619140625,
      "learning_rate": 3.2196725785306654e-07,
      "loss": 23.7054,
      "step": 25690
    },
    {
      "epoch": 203.97308075772682,
      "grad_norm": 696.3908081054688,
      "learning_rate": 3.218477648463276e-07,
      "loss": 21.9615,
      "step": 25700
    },
    {
      "epoch": 203.97308075772682,
      "eval_loss": 2.9862987995147705,
      "eval_mae": 1.3001036643981934,
      "eval_mse": 2.9862983226776123,
      "eval_r2": 0.10024291276931763,
      "eval_rmse": 1.7280909474554897,
      "eval_runtime": 9.0986,
      "eval_samples_per_second": 440.836,
      "eval_steps_per_second": 13.848,
      "step": 25700
    },
    {
      "epoch": 204.04785643070787,
      "grad_norm": 554.8890991210938,
      "learning_rate": 3.2172825394500774e-07,
      "loss": 22.9764,
      "step": 25710
    },
    {
      "epoch": 204.12761714855435,
      "grad_norm": 612.753662109375,
      "learning_rate": 3.216087251788728e-07,
      "loss": 24.8179,
      "step": 25720
    },
    {
      "epoch": 204.2073778664008,
      "grad_norm": 660.7884521484375,
      "learning_rate": 3.2148917857769276e-07,
      "loss": 23.1987,
      "step": 25730
    },
    {
      "epoch": 204.28713858424726,
      "grad_norm": 749.5087280273438,
      "learning_rate": 3.213696141712424e-07,
      "loss": 22.4109,
      "step": 25740
    },
    {
      "epoch": 204.3668993020937,
      "grad_norm": 783.62744140625,
      "learning_rate": 3.212500319893005e-07,
      "loss": 23.2447,
      "step": 25750
    },
    {
      "epoch": 204.44666001994017,
      "grad_norm": 850.573974609375,
      "learning_rate": 3.2113043206165065e-07,
      "loss": 24.0904,
      "step": 25760
    },
    {
      "epoch": 204.52642073778665,
      "grad_norm": 613.3543701171875,
      "learning_rate": 3.2101081441808067e-07,
      "loss": 22.9554,
      "step": 25770
    },
    {
      "epoch": 204.6061814556331,
      "grad_norm": 321.5801696777344,
      "learning_rate": 3.2089117908838275e-07,
      "loss": 24.8453,
      "step": 25780
    },
    {
      "epoch": 204.68594217347956,
      "grad_norm": 598.156005859375,
      "learning_rate": 3.207715261023536e-07,
      "loss": 22.7975,
      "step": 25790
    },
    {
      "epoch": 204.76570289132601,
      "grad_norm": 295.4639892578125,
      "learning_rate": 3.206518554897943e-07,
      "loss": 22.9623,
      "step": 25800
    },
    {
      "epoch": 204.76570289132601,
      "eval_loss": 2.9853804111480713,
      "eval_mae": 1.2947884798049927,
      "eval_mse": 2.9853804111480713,
      "eval_r2": 0.10051953792572021,
      "eval_rmse": 1.7278253416211,
      "eval_runtime": 9.1201,
      "eval_samples_per_second": 439.799,
      "eval_steps_per_second": 13.816,
      "step": 25800
    },
    {
      "epoch": 204.84546360917247,
      "grad_norm": 192.13729858398438,
      "learning_rate": 3.2053216728051036e-07,
      "loss": 24.6751,
      "step": 25810
    },
    {
      "epoch": 204.92522432701895,
      "grad_norm": 1190.568603515625,
      "learning_rate": 3.204124615043115e-07,
      "loss": 26.1587,
      "step": 25820
    },
    {
      "epoch": 205.0,
      "grad_norm": 304.8810729980469,
      "learning_rate": 3.2029273819101207e-07,
      "loss": 21.0388,
      "step": 25830
    },
    {
      "epoch": 205.07976071784645,
      "grad_norm": 208.4384307861328,
      "learning_rate": 3.201729973704306e-07,
      "loss": 23.0309,
      "step": 25840
    },
    {
      "epoch": 205.1595214356929,
      "grad_norm": 164.4918212890625,
      "learning_rate": 3.200532390723899e-07,
      "loss": 25.1194,
      "step": 25850
    },
    {
      "epoch": 205.2392821535394,
      "grad_norm": 400.9577941894531,
      "learning_rate": 3.1993346332671744e-07,
      "loss": 22.6518,
      "step": 25860
    },
    {
      "epoch": 205.31904287138585,
      "grad_norm": 298.8614501953125,
      "learning_rate": 3.1981367016324477e-07,
      "loss": 24.188,
      "step": 25870
    },
    {
      "epoch": 205.3988035892323,
      "grad_norm": 367.3051452636719,
      "learning_rate": 3.1969385961180803e-07,
      "loss": 24.6883,
      "step": 25880
    },
    {
      "epoch": 205.47856430707876,
      "grad_norm": 600.275146484375,
      "learning_rate": 3.1957403170224723e-07,
      "loss": 22.7788,
      "step": 25890
    },
    {
      "epoch": 205.5583250249252,
      "grad_norm": 344.5098571777344,
      "learning_rate": 3.1945418646440726e-07,
      "loss": 23.5699,
      "step": 25900
    },
    {
      "epoch": 205.5583250249252,
      "eval_loss": 2.983717203140259,
      "eval_mae": 1.297709345817566,
      "eval_mse": 2.9837169647216797,
      "eval_r2": 0.1010206937789917,
      "eval_rmse": 1.7273439045892627,
      "eval_runtime": 9.1253,
      "eval_samples_per_second": 439.55,
      "eval_steps_per_second": 13.808,
      "step": 25900
    },
    {
      "epoch": 205.6380857427717,
      "grad_norm": 954.1924438476562,
      "learning_rate": 3.1933432392813704e-07,
      "loss": 24.4087,
      "step": 25910
    },
    {
      "epoch": 205.71784646061815,
      "grad_norm": 400.8988037109375,
      "learning_rate": 3.1921444412328965e-07,
      "loss": 26.4869,
      "step": 25920
    },
    {
      "epoch": 205.7976071784646,
      "grad_norm": 497.05010986328125,
      "learning_rate": 3.1909454707972286e-07,
      "loss": 22.0453,
      "step": 25930
    },
    {
      "epoch": 205.87736789631106,
      "grad_norm": 645.5272216796875,
      "learning_rate": 3.1897463282729843e-07,
      "loss": 23.4677,
      "step": 25940
    },
    {
      "epoch": 205.95712861415754,
      "grad_norm": 649.5585327148438,
      "learning_rate": 3.1885470139588243e-07,
      "loss": 23.6492,
      "step": 25950
    },
    {
      "epoch": 206.0319042871386,
      "grad_norm": 312.9655456542969,
      "learning_rate": 3.187347528153454e-07,
      "loss": 19.5719,
      "step": 25960
    },
    {
      "epoch": 206.11166500498504,
      "grad_norm": 301.42578125,
      "learning_rate": 3.1861478711556194e-07,
      "loss": 25.0317,
      "step": 25970
    },
    {
      "epoch": 206.1914257228315,
      "grad_norm": 354.2798767089844,
      "learning_rate": 3.184948043264111e-07,
      "loss": 23.4643,
      "step": 25980
    },
    {
      "epoch": 206.27118644067798,
      "grad_norm": 1141.8057861328125,
      "learning_rate": 3.18374804477776e-07,
      "loss": 22.9418,
      "step": 25990
    },
    {
      "epoch": 206.35094715852443,
      "grad_norm": 691.326904296875,
      "learning_rate": 3.182547875995441e-07,
      "loss": 21.9891,
      "step": 26000
    },
    {
      "epoch": 206.35094715852443,
      "eval_loss": 2.9850969314575195,
      "eval_mae": 1.3001097440719604,
      "eval_mse": 2.9850971698760986,
      "eval_r2": 0.1006048321723938,
      "eval_rmse": 1.7277433750057034,
      "eval_runtime": 9.1101,
      "eval_samples_per_second": 440.283,
      "eval_steps_per_second": 13.831,
      "step": 26000
    },
    {
      "epoch": 206.4307078763709,
      "grad_norm": 354.40484619140625,
      "learning_rate": 3.181347537216071e-07,
      "loss": 23.952,
      "step": 26010
    },
    {
      "epoch": 206.51046859421734,
      "grad_norm": 323.9743347167969,
      "learning_rate": 3.18014702873861e-07,
      "loss": 24.5573,
      "step": 26020
    },
    {
      "epoch": 206.5902293120638,
      "grad_norm": 783.04736328125,
      "learning_rate": 3.1789463508620596e-07,
      "loss": 25.0914,
      "step": 26030
    },
    {
      "epoch": 206.66999002991028,
      "grad_norm": 842.0777587890625,
      "learning_rate": 3.177745503885463e-07,
      "loss": 22.0903,
      "step": 26040
    },
    {
      "epoch": 206.74975074775674,
      "grad_norm": 453.91766357421875,
      "learning_rate": 3.176544488107905e-07,
      "loss": 24.8008,
      "step": 26050
    },
    {
      "epoch": 206.8295114656032,
      "grad_norm": 358.0523681640625,
      "learning_rate": 3.175343303828516e-07,
      "loss": 24.3761,
      "step": 26060
    },
    {
      "epoch": 206.90927218344964,
      "grad_norm": 288.92523193359375,
      "learning_rate": 3.1741419513464643e-07,
      "loss": 23.9646,
      "step": 26070
    },
    {
      "epoch": 206.9890329012961,
      "grad_norm": 384.78790283203125,
      "learning_rate": 3.1729404309609616e-07,
      "loss": 24.4646,
      "step": 26080
    },
    {
      "epoch": 207.06380857427718,
      "grad_norm": 440.7989501953125,
      "learning_rate": 3.171738742971263e-07,
      "loss": 20.5563,
      "step": 26090
    },
    {
      "epoch": 207.14356929212363,
      "grad_norm": 186.81570434570312,
      "learning_rate": 3.170536887676662e-07,
      "loss": 23.336,
      "step": 26100
    },
    {
      "epoch": 207.14356929212363,
      "eval_loss": 2.9858384132385254,
      "eval_mae": 1.3001880645751953,
      "eval_mse": 2.9858384132385254,
      "eval_r2": 0.10038143396377563,
      "eval_rmse": 1.7279578736874708,
      "eval_runtime": 9.1198,
      "eval_samples_per_second": 439.813,
      "eval_steps_per_second": 13.816,
      "step": 26100
    },
    {
      "epoch": 207.22333000997008,
      "grad_norm": 370.08123779296875,
      "learning_rate": 3.169334865376497e-07,
      "loss": 23.6388,
      "step": 26110
    },
    {
      "epoch": 207.30309072781654,
      "grad_norm": 1464.3602294921875,
      "learning_rate": 3.1681326763701455e-07,
      "loss": 22.6312,
      "step": 26120
    },
    {
      "epoch": 207.38285144566302,
      "grad_norm": 869.84716796875,
      "learning_rate": 3.166930320957028e-07,
      "loss": 21.8956,
      "step": 26130
    },
    {
      "epoch": 207.46261216350948,
      "grad_norm": 499.814697265625,
      "learning_rate": 3.165727799436607e-07,
      "loss": 24.4855,
      "step": 26140
    },
    {
      "epoch": 207.54237288135593,
      "grad_norm": 283.2919616699219,
      "learning_rate": 3.1645251121083847e-07,
      "loss": 26.9945,
      "step": 26150
    },
    {
      "epoch": 207.62213359920239,
      "grad_norm": 403.3465576171875,
      "learning_rate": 3.163322259271906e-07,
      "loss": 22.2168,
      "step": 26160
    },
    {
      "epoch": 207.70189431704884,
      "grad_norm": 417.0478515625,
      "learning_rate": 3.162119241226755e-07,
      "loss": 23.5436,
      "step": 26170
    },
    {
      "epoch": 207.78165503489532,
      "grad_norm": 1154.0826416015625,
      "learning_rate": 3.1609160582725596e-07,
      "loss": 23.8994,
      "step": 26180
    },
    {
      "epoch": 207.86141575274178,
      "grad_norm": 448.4119567871094,
      "learning_rate": 3.159712710708987e-07,
      "loss": 24.1008,
      "step": 26190
    },
    {
      "epoch": 207.94117647058823,
      "grad_norm": 294.5215148925781,
      "learning_rate": 3.1585091988357465e-07,
      "loss": 25.3929,
      "step": 26200
    },
    {
      "epoch": 207.94117647058823,
      "eval_loss": 2.9880077838897705,
      "eval_mae": 1.2916208505630493,
      "eval_mse": 2.9880077838897705,
      "eval_r2": 0.09972786903381348,
      "eval_rmse": 1.728585486428071,
      "eval_runtime": 9.2105,
      "eval_samples_per_second": 435.482,
      "eval_steps_per_second": 13.68,
      "step": 26200
    },
    {
      "epoch": 208.01595214356928,
      "grad_norm": 662.5714721679688,
      "learning_rate": 3.157305522952586e-07,
      "loss": 24.0349,
      "step": 26210
    },
    {
      "epoch": 208.09571286141576,
      "grad_norm": 311.1929016113281,
      "learning_rate": 3.1561016833592986e-07,
      "loss": 23.6938,
      "step": 26220
    },
    {
      "epoch": 208.17547357926222,
      "grad_norm": 391.3521423339844,
      "learning_rate": 3.154897680355714e-07,
      "loss": 21.3007,
      "step": 26230
    },
    {
      "epoch": 208.25523429710867,
      "grad_norm": 474.2955017089844,
      "learning_rate": 3.153693514241703e-07,
      "loss": 25.5512,
      "step": 26240
    },
    {
      "epoch": 208.33499501495513,
      "grad_norm": 379.4305725097656,
      "learning_rate": 3.15248918531718e-07,
      "loss": 22.6775,
      "step": 26250
    },
    {
      "epoch": 208.4147557328016,
      "grad_norm": 223.09719848632812,
      "learning_rate": 3.151284693882098e-07,
      "loss": 23.6358,
      "step": 26260
    },
    {
      "epoch": 208.49451645064806,
      "grad_norm": 170.5644989013672,
      "learning_rate": 3.15008004023645e-07,
      "loss": 23.4765,
      "step": 26270
    },
    {
      "epoch": 208.57427716849452,
      "grad_norm": 220.80567932128906,
      "learning_rate": 3.14887522468027e-07,
      "loss": 24.6062,
      "step": 26280
    },
    {
      "epoch": 208.65403788634097,
      "grad_norm": 107.43252563476562,
      "learning_rate": 3.147670247513632e-07,
      "loss": 22.696,
      "step": 26290
    },
    {
      "epoch": 208.73379860418743,
      "grad_norm": 384.41485595703125,
      "learning_rate": 3.1464651090366517e-07,
      "loss": 23.3779,
      "step": 26300
    },
    {
      "epoch": 208.73379860418743,
      "eval_loss": 2.990997314453125,
      "eval_mae": 1.2915152311325073,
      "eval_mse": 2.990997314453125,
      "eval_r2": 0.09882712364196777,
      "eval_rmse": 1.729450003455759,
      "eval_runtime": 9.1109,
      "eval_samples_per_second": 440.24,
      "eval_steps_per_second": 13.83,
      "step": 26300
    },
    {
      "epoch": 208.8135593220339,
      "grad_norm": 706.5662231445312,
      "learning_rate": 3.1452598095494836e-07,
      "loss": 24.5741,
      "step": 26310
    },
    {
      "epoch": 208.89332003988036,
      "grad_norm": 936.375732421875,
      "learning_rate": 3.144054349352322e-07,
      "loss": 24.9527,
      "step": 26320
    },
    {
      "epoch": 208.97308075772682,
      "grad_norm": 1180.5230712890625,
      "learning_rate": 3.142848728745401e-07,
      "loss": 24.4996,
      "step": 26330
    },
    {
      "epoch": 209.04785643070787,
      "grad_norm": 676.7444458007812,
      "learning_rate": 3.1416429480289974e-07,
      "loss": 22.0019,
      "step": 26340
    },
    {
      "epoch": 209.12761714855435,
      "grad_norm": 1116.85986328125,
      "learning_rate": 3.1405576087388343e-07,
      "loss": 24.691,
      "step": 26350
    },
    {
      "epoch": 209.2073778664008,
      "grad_norm": 266.63189697265625,
      "learning_rate": 3.139351524641811e-07,
      "loss": 22.1795,
      "step": 26360
    },
    {
      "epoch": 209.28713858424726,
      "grad_norm": 616.2620849609375,
      "learning_rate": 3.138145281306327e-07,
      "loss": 23.662,
      "step": 26370
    },
    {
      "epoch": 209.3668993020937,
      "grad_norm": 704.6864013671875,
      "learning_rate": 3.136938879032811e-07,
      "loss": 22.9147,
      "step": 26380
    },
    {
      "epoch": 209.44666001994017,
      "grad_norm": 351.6438293457031,
      "learning_rate": 3.1357323181217324e-07,
      "loss": 23.9645,
      "step": 26390
    },
    {
      "epoch": 209.52642073778665,
      "grad_norm": 122.1126480102539,
      "learning_rate": 3.1345255988736025e-07,
      "loss": 23.5815,
      "step": 26400
    },
    {
      "epoch": 209.52642073778665,
      "eval_loss": 2.9857232570648193,
      "eval_mae": 1.2909077405929565,
      "eval_mse": 2.9857232570648193,
      "eval_r2": 0.10041612386703491,
      "eval_rmse": 1.7279245519017372,
      "eval_runtime": 9.1149,
      "eval_samples_per_second": 440.05,
      "eval_steps_per_second": 13.824,
      "step": 26400
    },
    {
      "epoch": 209.6061814556331,
      "grad_norm": 895.959228515625,
      "learning_rate": 3.1333187215889674e-07,
      "loss": 24.7816,
      "step": 26410
    },
    {
      "epoch": 209.68594217347956,
      "grad_norm": 397.9737854003906,
      "learning_rate": 3.1321116865684153e-07,
      "loss": 24.0738,
      "step": 26420
    },
    {
      "epoch": 209.76570289132601,
      "grad_norm": 914.4462890625,
      "learning_rate": 3.130904494112574e-07,
      "loss": 22.1512,
      "step": 26430
    },
    {
      "epoch": 209.84546360917247,
      "grad_norm": 432.4562072753906,
      "learning_rate": 3.1296971445221093e-07,
      "loss": 21.9592,
      "step": 26440
    },
    {
      "epoch": 209.92522432701895,
      "grad_norm": 652.2195434570312,
      "learning_rate": 3.128489638097727e-07,
      "loss": 24.3588,
      "step": 26450
    },
    {
      "epoch": 210.0,
      "grad_norm": 416.7950134277344,
      "learning_rate": 3.1272819751401717e-07,
      "loss": 24.7862,
      "step": 26460
    },
    {
      "epoch": 210.07976071784645,
      "grad_norm": 926.331787109375,
      "learning_rate": 3.126074155950226e-07,
      "loss": 24.4526,
      "step": 26470
    },
    {
      "epoch": 210.1595214356929,
      "grad_norm": 847.6224975585938,
      "learning_rate": 3.1248661808287135e-07,
      "loss": 22.3391,
      "step": 26480
    },
    {
      "epoch": 210.2392821535394,
      "grad_norm": 1064.720458984375,
      "learning_rate": 3.1236580500764946e-07,
      "loss": 24.1305,
      "step": 26490
    },
    {
      "epoch": 210.31904287138585,
      "grad_norm": 592.7523193359375,
      "learning_rate": 3.1224497639944697e-07,
      "loss": 22.8022,
      "step": 26500
    },
    {
      "epoch": 210.31904287138585,
      "eval_loss": 2.9844868183135986,
      "eval_mae": 1.2950390577316284,
      "eval_mse": 2.9844868183135986,
      "eval_r2": 0.10078871250152588,
      "eval_rmse": 1.7275667333893643,
      "eval_runtime": 9.1183,
      "eval_samples_per_second": 439.883,
      "eval_steps_per_second": 13.818,
      "step": 26500
    },
    {
      "epoch": 210.3988035892323,
      "grad_norm": 326.2663879394531,
      "learning_rate": 3.121241322883578e-07,
      "loss": 22.7675,
      "step": 26510
    },
    {
      "epoch": 210.47856430707876,
      "grad_norm": 256.7446594238281,
      "learning_rate": 3.120032727044795e-07,
      "loss": 24.0059,
      "step": 26520
    },
    {
      "epoch": 210.5583250249252,
      "grad_norm": 1118.33642578125,
      "learning_rate": 3.1188239767791395e-07,
      "loss": 26.5326,
      "step": 26530
    },
    {
      "epoch": 210.6380857427717,
      "grad_norm": 410.37005615234375,
      "learning_rate": 3.1176150723876625e-07,
      "loss": 22.0273,
      "step": 26540
    },
    {
      "epoch": 210.71784646061815,
      "grad_norm": 141.73388671875,
      "learning_rate": 3.116406014171459e-07,
      "loss": 22.7835,
      "step": 26550
    },
    {
      "epoch": 210.7976071784646,
      "grad_norm": 614.6095581054688,
      "learning_rate": 3.1151968024316593e-07,
      "loss": 22.3267,
      "step": 26560
    },
    {
      "epoch": 210.87736789631106,
      "grad_norm": 1028.78955078125,
      "learning_rate": 3.113987437469433e-07,
      "loss": 25.7587,
      "step": 26570
    },
    {
      "epoch": 210.95712861415754,
      "grad_norm": 416.4407653808594,
      "learning_rate": 3.1127779195859873e-07,
      "loss": 23.3185,
      "step": 26580
    },
    {
      "epoch": 211.0319042871386,
      "grad_norm": 712.4876098632812,
      "learning_rate": 3.111568249082566e-07,
      "loss": 22.3591,
      "step": 26590
    },
    {
      "epoch": 211.11166500498504,
      "grad_norm": 751.4306640625,
      "learning_rate": 3.110358426260456e-07,
      "loss": 22.1697,
      "step": 26600
    },
    {
      "epoch": 211.11166500498504,
      "eval_loss": 2.9798662662506104,
      "eval_mae": 1.2977105379104614,
      "eval_mse": 2.9798665046691895,
      "eval_r2": 0.1021808385848999,
      "eval_rmse": 1.726228983845767,
      "eval_runtime": 9.0913,
      "eval_samples_per_second": 441.192,
      "eval_steps_per_second": 13.859,
      "step": 26600
    },
    {
      "epoch": 211.1914257228315,
      "grad_norm": 240.1824188232422,
      "learning_rate": 3.1091484514209763e-07,
      "loss": 22.6163,
      "step": 26610
    },
    {
      "epoch": 211.27118644067798,
      "grad_norm": 309.4950256347656,
      "learning_rate": 3.107938324865486e-07,
      "loss": 23.2412,
      "step": 26620
    },
    {
      "epoch": 211.35094715852443,
      "grad_norm": 148.27700805664062,
      "learning_rate": 3.106728046895384e-07,
      "loss": 25.1011,
      "step": 26630
    },
    {
      "epoch": 211.4307078763709,
      "grad_norm": 307.0238952636719,
      "learning_rate": 3.105517617812104e-07,
      "loss": 25.1078,
      "step": 26640
    },
    {
      "epoch": 211.51046859421734,
      "grad_norm": 364.7814025878906,
      "learning_rate": 3.1043070379171187e-07,
      "loss": 25.404,
      "step": 26650
    },
    {
      "epoch": 211.5902293120638,
      "grad_norm": 590.2925415039062,
      "learning_rate": 3.103096307511938e-07,
      "loss": 24.4948,
      "step": 26660
    },
    {
      "epoch": 211.66999002991028,
      "grad_norm": 840.1451416015625,
      "learning_rate": 3.101885426898109e-07,
      "loss": 23.8928,
      "step": 26670
    },
    {
      "epoch": 211.74975074775674,
      "grad_norm": 673.6281127929688,
      "learning_rate": 3.1006743963772163e-07,
      "loss": 23.0499,
      "step": 26680
    },
    {
      "epoch": 211.8295114656032,
      "grad_norm": 723.83837890625,
      "learning_rate": 3.099463216250884e-07,
      "loss": 23.7825,
      "step": 26690
    },
    {
      "epoch": 211.90927218344964,
      "grad_norm": 412.1143493652344,
      "learning_rate": 3.0982518868207696e-07,
      "loss": 23.5121,
      "step": 26700
    },
    {
      "epoch": 211.90927218344964,
      "eval_loss": 2.9848618507385254,
      "eval_mae": 1.2893383502960205,
      "eval_mse": 2.9848618507385254,
      "eval_r2": 0.10067570209503174,
      "eval_rmse": 1.7276752735217709,
      "eval_runtime": 9.0701,
      "eval_samples_per_second": 442.222,
      "eval_steps_per_second": 13.892,
      "step": 26700
    },
    {
      "epoch": 211.9890329012961,
      "grad_norm": 192.29212951660156,
      "learning_rate": 3.0970404083885703e-07,
      "loss": 22.379,
      "step": 26710
    },
    {
      "epoch": 212.06380857427718,
      "grad_norm": 435.377685546875,
      "learning_rate": 3.0958287812560213e-07,
      "loss": 22.9865,
      "step": 26720
    },
    {
      "epoch": 212.14356929212363,
      "grad_norm": 551.6611328125,
      "learning_rate": 3.094617005724892e-07,
      "loss": 23.0014,
      "step": 26730
    },
    {
      "epoch": 212.22333000997008,
      "grad_norm": 1402.768798828125,
      "learning_rate": 3.09340508209699e-07,
      "loss": 22.7538,
      "step": 26740
    },
    {
      "epoch": 212.30309072781654,
      "grad_norm": 192.71163940429688,
      "learning_rate": 3.0921930106741606e-07,
      "loss": 23.8688,
      "step": 26750
    },
    {
      "epoch": 212.38285144566302,
      "grad_norm": 152.849609375,
      "learning_rate": 3.0909807917582853e-07,
      "loss": 24.4226,
      "step": 26760
    },
    {
      "epoch": 212.46261216350948,
      "grad_norm": 523.3204345703125,
      "learning_rate": 3.0897684256512824e-07,
      "loss": 24.8181,
      "step": 26770
    },
    {
      "epoch": 212.54237288135593,
      "grad_norm": 883.080078125,
      "learning_rate": 3.088555912655106e-07,
      "loss": 23.5999,
      "step": 26780
    },
    {
      "epoch": 212.62213359920239,
      "grad_norm": 563.0305786132812,
      "learning_rate": 3.087343253071748e-07,
      "loss": 24.0943,
      "step": 26790
    },
    {
      "epoch": 212.70189431704884,
      "grad_norm": 514.7747192382812,
      "learning_rate": 3.0861304472032374e-07,
      "loss": 22.3982,
      "step": 26800
    },
    {
      "epoch": 212.70189431704884,
      "eval_loss": 2.981959104537964,
      "eval_mae": 1.289758324623108,
      "eval_mse": 2.981959104537964,
      "eval_r2": 0.10155034065246582,
      "eval_rmse": 1.726834996326506,
      "eval_runtime": 9.0866,
      "eval_samples_per_second": 441.417,
      "eval_steps_per_second": 13.867,
      "step": 26800
    },
    {
      "epoch": 212.78165503489532,
      "grad_norm": 395.4801025390625,
      "learning_rate": 3.084917495351637e-07,
      "loss": 22.2588,
      "step": 26810
    },
    {
      "epoch": 212.86141575274178,
      "grad_norm": 589.0004272460938,
      "learning_rate": 3.083704397819048e-07,
      "loss": 22.845,
      "step": 26820
    },
    {
      "epoch": 212.94117647058823,
      "grad_norm": 1045.0267333984375,
      "learning_rate": 3.082491154907609e-07,
      "loss": 25.0153,
      "step": 26830
    },
    {
      "epoch": 213.01595214356928,
      "grad_norm": 992.125244140625,
      "learning_rate": 3.08127776691949e-07,
      "loss": 24.5204,
      "step": 26840
    },
    {
      "epoch": 213.09571286141576,
      "grad_norm": 525.619873046875,
      "learning_rate": 3.080064234156904e-07,
      "loss": 24.5004,
      "step": 26850
    },
    {
      "epoch": 213.17547357926222,
      "grad_norm": 349.7850341796875,
      "learning_rate": 3.0788505569220945e-07,
      "loss": 25.3501,
      "step": 26860
    },
    {
      "epoch": 213.25523429710867,
      "grad_norm": 229.71084594726562,
      "learning_rate": 3.0776367355173427e-07,
      "loss": 22.593,
      "step": 26870
    },
    {
      "epoch": 213.33499501495513,
      "grad_norm": 118.41441345214844,
      "learning_rate": 3.076422770244967e-07,
      "loss": 22.9523,
      "step": 26880
    },
    {
      "epoch": 213.4147557328016,
      "grad_norm": 936.93994140625,
      "learning_rate": 3.0752086614073204e-07,
      "loss": 24.3618,
      "step": 26890
    },
    {
      "epoch": 213.49451645064806,
      "grad_norm": 938.4599609375,
      "learning_rate": 3.073994409306791e-07,
      "loss": 23.4791,
      "step": 26900
    },
    {
      "epoch": 213.49451645064806,
      "eval_loss": 2.980309247970581,
      "eval_mae": 1.300426721572876,
      "eval_mse": 2.980309009552002,
      "eval_r2": 0.10204744338989258,
      "eval_rmse": 1.726357150056732,
      "eval_runtime": 9.0902,
      "eval_samples_per_second": 441.247,
      "eval_steps_per_second": 13.861,
      "step": 26900
    },
    {
      "epoch": 213.57427716849452,
      "grad_norm": 547.9718017578125,
      "learning_rate": 3.072780014245804e-07,
      "loss": 24.3147,
      "step": 26910
    },
    {
      "epoch": 213.65403788634097,
      "grad_norm": 753.8389892578125,
      "learning_rate": 3.0715654765268195e-07,
      "loss": 24.3151,
      "step": 26920
    },
    {
      "epoch": 213.73379860418743,
      "grad_norm": 190.3426513671875,
      "learning_rate": 3.0703507964523335e-07,
      "loss": 23.3416,
      "step": 26930
    },
    {
      "epoch": 213.8135593220339,
      "grad_norm": 464.51312255859375,
      "learning_rate": 3.069135974324876e-07,
      "loss": 23.1847,
      "step": 26940
    },
    {
      "epoch": 213.89332003988036,
      "grad_norm": 402.635009765625,
      "learning_rate": 3.067921010447016e-07,
      "loss": 22.662,
      "step": 26950
    },
    {
      "epoch": 213.97308075772682,
      "grad_norm": 920.003173828125,
      "learning_rate": 3.0667059051213534e-07,
      "loss": 22.8849,
      "step": 26960
    },
    {
      "epoch": 214.04785643070787,
      "grad_norm": 486.64398193359375,
      "learning_rate": 3.0654906586505257e-07,
      "loss": 21.9501,
      "step": 26970
    },
    {
      "epoch": 214.12761714855435,
      "grad_norm": 1110.277099609375,
      "learning_rate": 3.064275271337205e-07,
      "loss": 24.034,
      "step": 26980
    },
    {
      "epoch": 214.2073778664008,
      "grad_norm": 1393.9989013671875,
      "learning_rate": 3.0630597434840996e-07,
      "loss": 23.2111,
      "step": 26990
    },
    {
      "epoch": 214.28713858424726,
      "grad_norm": 522.2705688476562,
      "learning_rate": 3.061844075393951e-07,
      "loss": 24.3892,
      "step": 27000
    },
    {
      "epoch": 214.28713858424726,
      "eval_loss": 2.981308698654175,
      "eval_mae": 1.2962442636489868,
      "eval_mse": 2.981308937072754,
      "eval_r2": 0.10174620151519775,
      "eval_rmse": 1.7266467319845,
      "eval_runtime": 9.0887,
      "eval_samples_per_second": 441.319,
      "eval_steps_per_second": 13.863,
      "step": 27000
    },
    {
      "epoch": 214.3668993020937,
      "grad_norm": 609.6259765625,
      "learning_rate": 3.0606282673695367e-07,
      "loss": 23.9073,
      "step": 27010
    },
    {
      "epoch": 214.44666001994017,
      "grad_norm": 791.1295166015625,
      "learning_rate": 3.059412319713669e-07,
      "loss": 25.4684,
      "step": 27020
    },
    {
      "epoch": 214.52642073778665,
      "grad_norm": 429.5850524902344,
      "learning_rate": 3.0581962327291945e-07,
      "loss": 24.0151,
      "step": 27030
    },
    {
      "epoch": 214.6061814556331,
      "grad_norm": 629.2057495117188,
      "learning_rate": 3.0569800067189956e-07,
      "loss": 24.1557,
      "step": 27040
    },
    {
      "epoch": 214.68594217347956,
      "grad_norm": 265.7466125488281,
      "learning_rate": 3.0557636419859874e-07,
      "loss": 23.352,
      "step": 27050
    },
    {
      "epoch": 214.76570289132601,
      "grad_norm": 484.65228271484375,
      "learning_rate": 3.0545471388331213e-07,
      "loss": 23.608,
      "step": 27060
    },
    {
      "epoch": 214.84546360917247,
      "grad_norm": 283.9285583496094,
      "learning_rate": 3.053330497563382e-07,
      "loss": 22.0524,
      "step": 27070
    },
    {
      "epoch": 214.92522432701895,
      "grad_norm": 143.06239318847656,
      "learning_rate": 3.05211371847979e-07,
      "loss": 22.4378,
      "step": 27080
    },
    {
      "epoch": 215.0,
      "grad_norm": 146.85423278808594,
      "learning_rate": 3.0508968018853994e-07,
      "loss": 21.4124,
      "step": 27090
    },
    {
      "epoch": 215.07976071784645,
      "grad_norm": 1108.4443359375,
      "learning_rate": 3.049679748083297e-07,
      "loss": 24.869,
      "step": 27100
    },
    {
      "epoch": 215.07976071784645,
      "eval_loss": 2.987786054611206,
      "eval_mae": 1.2889059782028198,
      "eval_mse": 2.987786054611206,
      "eval_r2": 0.09979468584060669,
      "eval_rmse": 1.7285213491916165,
      "eval_runtime": 9.0675,
      "eval_samples_per_second": 442.348,
      "eval_steps_per_second": 13.896,
      "step": 27100
    },
    {
      "epoch": 215.1595214356929,
      "grad_norm": 340.2200927734375,
      "learning_rate": 3.048462557376607e-07,
      "loss": 22.5566,
      "step": 27110
    },
    {
      "epoch": 215.2392821535394,
      "grad_norm": 342.66363525390625,
      "learning_rate": 3.0472452300684846e-07,
      "loss": 22.7565,
      "step": 27120
    },
    {
      "epoch": 215.31904287138585,
      "grad_norm": 818.2691040039062,
      "learning_rate": 3.0460277664621216e-07,
      "loss": 23.5732,
      "step": 27130
    },
    {
      "epoch": 215.3988035892323,
      "grad_norm": 315.8173828125,
      "learning_rate": 3.0448101668607405e-07,
      "loss": 25.9838,
      "step": 27140
    },
    {
      "epoch": 215.47856430707876,
      "grad_norm": 444.2835998535156,
      "learning_rate": 3.043592431567601e-07,
      "loss": 21.8117,
      "step": 27150
    },
    {
      "epoch": 215.5583250249252,
      "grad_norm": 523.1034545898438,
      "learning_rate": 3.042374560885996e-07,
      "loss": 22.2515,
      "step": 27160
    },
    {
      "epoch": 215.6380857427717,
      "grad_norm": 622.7999267578125,
      "learning_rate": 3.0411565551192497e-07,
      "loss": 23.3136,
      "step": 27170
    },
    {
      "epoch": 215.71784646061815,
      "grad_norm": 701.6234741210938,
      "learning_rate": 3.039938414570723e-07,
      "loss": 25.0017,
      "step": 27180
    },
    {
      "epoch": 215.7976071784646,
      "grad_norm": 909.46337890625,
      "learning_rate": 3.0387201395438075e-07,
      "loss": 23.118,
      "step": 27190
    },
    {
      "epoch": 215.87736789631106,
      "grad_norm": 378.4754333496094,
      "learning_rate": 3.037501730341931e-07,
      "loss": 23.0487,
      "step": 27200
    },
    {
      "epoch": 215.87736789631106,
      "eval_loss": 2.978470802307129,
      "eval_mae": 1.2960094213485718,
      "eval_mse": 2.978471040725708,
      "eval_r2": 0.10260123014450073,
      "eval_rmse": 1.7258247421814614,
      "eval_runtime": 9.0961,
      "eval_samples_per_second": 440.958,
      "eval_steps_per_second": 13.852,
      "step": 27200
    },
    {
      "epoch": 215.95712861415754,
      "grad_norm": 356.2110900878906,
      "learning_rate": 3.0362831872685534e-07,
      "loss": 23.9503,
      "step": 27210
    },
    {
      "epoch": 216.0319042871386,
      "grad_norm": 158.9725799560547,
      "learning_rate": 3.0350645106271674e-07,
      "loss": 23.5513,
      "step": 27220
    },
    {
      "epoch": 216.11166500498504,
      "grad_norm": 451.63873291015625,
      "learning_rate": 3.0338457007212995e-07,
      "loss": 24.9653,
      "step": 27230
    },
    {
      "epoch": 216.1914257228315,
      "grad_norm": 254.16026306152344,
      "learning_rate": 3.0326267578545105e-07,
      "loss": 25.1305,
      "step": 27240
    },
    {
      "epoch": 216.27118644067798,
      "grad_norm": 154.31466674804688,
      "learning_rate": 3.031407682330393e-07,
      "loss": 24.1676,
      "step": 27250
    },
    {
      "epoch": 216.35094715852443,
      "grad_norm": 523.8579711914062,
      "learning_rate": 3.030188474452572e-07,
      "loss": 23.739,
      "step": 27260
    },
    {
      "epoch": 216.4307078763709,
      "grad_norm": 420.2916564941406,
      "learning_rate": 3.028969134524707e-07,
      "loss": 22.0993,
      "step": 27270
    },
    {
      "epoch": 216.51046859421734,
      "grad_norm": 319.1602478027344,
      "learning_rate": 3.0277496628504905e-07,
      "loss": 24.9593,
      "step": 27280
    },
    {
      "epoch": 216.5902293120638,
      "grad_norm": 464.3196105957031,
      "learning_rate": 3.026530059733646e-07,
      "loss": 22.566,
      "step": 27290
    },
    {
      "epoch": 216.66999002991028,
      "grad_norm": 503.96002197265625,
      "learning_rate": 3.0253103254779313e-07,
      "loss": 22.2243,
      "step": 27300
    },
    {
      "epoch": 216.66999002991028,
      "eval_loss": 2.983785390853882,
      "eval_mae": 1.289696216583252,
      "eval_mse": 2.983785390853882,
      "eval_r2": 0.10100007057189941,
      "eval_rmse": 1.7273637112240958,
      "eval_runtime": 9.1023,
      "eval_samples_per_second": 440.66,
      "eval_steps_per_second": 13.843,
      "step": 27300
    },
    {
      "epoch": 216.74975074775674,
      "grad_norm": 250.5771484375,
      "learning_rate": 3.0240904603871364e-07,
      "loss": 22.3307,
      "step": 27310
    },
    {
      "epoch": 216.8295114656032,
      "grad_norm": 230.27337646484375,
      "learning_rate": 3.0228704647650846e-07,
      "loss": 26.9951,
      "step": 27320
    },
    {
      "epoch": 216.90927218344964,
      "grad_norm": 529.7537231445312,
      "learning_rate": 3.0216503389156295e-07,
      "loss": 21.5616,
      "step": 27330
    },
    {
      "epoch": 216.9890329012961,
      "grad_norm": 253.22357177734375,
      "learning_rate": 3.02043008314266e-07,
      "loss": 22.4429,
      "step": 27340
    },
    {
      "epoch": 217.06380857427718,
      "grad_norm": 1507.0018310546875,
      "learning_rate": 3.019209697750095e-07,
      "loss": 24.3405,
      "step": 27350
    },
    {
      "epoch": 217.14356929212363,
      "grad_norm": 398.4322814941406,
      "learning_rate": 3.0179891830418876e-07,
      "loss": 23.6838,
      "step": 27360
    },
    {
      "epoch": 217.22333000997008,
      "grad_norm": 602.019775390625,
      "learning_rate": 3.0167685393220217e-07,
      "loss": 23.0587,
      "step": 27370
    },
    {
      "epoch": 217.30309072781654,
      "grad_norm": 297.0119934082031,
      "learning_rate": 3.015547766894513e-07,
      "loss": 23.8301,
      "step": 27380
    },
    {
      "epoch": 217.38285144566302,
      "grad_norm": 910.6251831054688,
      "learning_rate": 3.0143268660634125e-07,
      "loss": 23.7992,
      "step": 27390
    },
    {
      "epoch": 217.46261216350948,
      "grad_norm": 256.3167419433594,
      "learning_rate": 3.013105837132798e-07,
      "loss": 24.3788,
      "step": 27400
    },
    {
      "epoch": 217.46261216350948,
      "eval_loss": 2.9812726974487305,
      "eval_mae": 1.3002257347106934,
      "eval_mse": 2.9812726974487305,
      "eval_r2": 0.10175710916519165,
      "eval_rmse": 1.7266362377318305,
      "eval_runtime": 9.1172,
      "eval_samples_per_second": 439.936,
      "eval_steps_per_second": 13.82,
      "step": 27400
    },
    {
      "epoch": 217.54237288135593,
      "grad_norm": 384.02593994140625,
      "learning_rate": 3.0118846804067845e-07,
      "loss": 23.1673,
      "step": 27410
    },
    {
      "epoch": 217.62213359920239,
      "grad_norm": 627.162109375,
      "learning_rate": 3.010663396189513e-07,
      "loss": 22.5284,
      "step": 27420
    },
    {
      "epoch": 217.70189431704884,
      "grad_norm": 474.76556396484375,
      "learning_rate": 3.009441984785163e-07,
      "loss": 23.4623,
      "step": 27430
    },
    {
      "epoch": 217.78165503489532,
      "grad_norm": 1055.54296875,
      "learning_rate": 3.0082204464979404e-07,
      "loss": 24.2294,
      "step": 27440
    },
    {
      "epoch": 217.86141575274178,
      "grad_norm": 845.8869018554688,
      "learning_rate": 3.0069987816320845e-07,
      "loss": 23.0862,
      "step": 27450
    },
    {
      "epoch": 217.94117647058823,
      "grad_norm": 661.3668823242188,
      "learning_rate": 3.0057769904918675e-07,
      "loss": 21.4018,
      "step": 27460
    },
    {
      "epoch": 218.01595214356928,
      "grad_norm": 251.98768615722656,
      "learning_rate": 3.0045550733815894e-07,
      "loss": 23.6382,
      "step": 27470
    },
    {
      "epoch": 218.09571286141576,
      "grad_norm": 373.1899719238281,
      "learning_rate": 3.003333030605587e-07,
      "loss": 23.496,
      "step": 27480
    },
    {
      "epoch": 218.17547357926222,
      "grad_norm": 744.9896240234375,
      "learning_rate": 3.0021108624682215e-07,
      "loss": 24.0153,
      "step": 27490
    },
    {
      "epoch": 218.25523429710867,
      "grad_norm": 481.24951171875,
      "learning_rate": 3.000888569273893e-07,
      "loss": 24.6332,
      "step": 27500
    },
    {
      "epoch": 218.25523429710867,
      "eval_loss": 2.977515459060669,
      "eval_mae": 1.291520118713379,
      "eval_mse": 2.977515459060669,
      "eval_r2": 0.10288918018341064,
      "eval_rmse": 1.725547872143995,
      "eval_runtime": 9.1128,
      "eval_samples_per_second": 440.149,
      "eval_steps_per_second": 13.827,
      "step": 27500
    },
    {
      "epoch": 218.33499501495513,
      "grad_norm": 484.2327575683594,
      "learning_rate": 2.999666151327025e-07,
      "loss": 21.8921,
      "step": 27510
    },
    {
      "epoch": 218.4147557328016,
      "grad_norm": 795.8241577148438,
      "learning_rate": 2.998443608932079e-07,
      "loss": 24.2613,
      "step": 27520
    },
    {
      "epoch": 218.49451645064806,
      "grad_norm": 686.586181640625,
      "learning_rate": 2.997220942393543e-07,
      "loss": 22.1268,
      "step": 27530
    },
    {
      "epoch": 218.57427716849452,
      "grad_norm": 200.74305725097656,
      "learning_rate": 2.995998152015937e-07,
      "loss": 24.4024,
      "step": 27540
    },
    {
      "epoch": 218.65403788634097,
      "grad_norm": 304.2043762207031,
      "learning_rate": 2.994775238103814e-07,
      "loss": 26.3741,
      "step": 27550
    },
    {
      "epoch": 218.73379860418743,
      "grad_norm": 668.7415771484375,
      "learning_rate": 2.9935522009617543e-07,
      "loss": 24.6852,
      "step": 27560
    },
    {
      "epoch": 218.8135593220339,
      "grad_norm": 1236.5673828125,
      "learning_rate": 2.9923290408943706e-07,
      "loss": 23.0106,
      "step": 27570
    },
    {
      "epoch": 218.89332003988036,
      "grad_norm": 484.6416320800781,
      "learning_rate": 2.9911057582063067e-07,
      "loss": 23.2765,
      "step": 27580
    },
    {
      "epoch": 218.97308075772682,
      "grad_norm": 399.0691223144531,
      "learning_rate": 2.9898823532022357e-07,
      "loss": 20.9575,
      "step": 27590
    },
    {
      "epoch": 219.04785643070787,
      "grad_norm": 333.0251770019531,
      "learning_rate": 2.9886588261868633e-07,
      "loss": 23.3247,
      "step": 27600
    },
    {
      "epoch": 219.04785643070787,
      "eval_loss": 2.975332021713257,
      "eval_mae": 1.2925277948379517,
      "eval_mse": 2.975332260131836,
      "eval_r2": 0.10354691743850708,
      "eval_rmse": 1.7249151457772745,
      "eval_runtime": 9.0693,
      "eval_samples_per_second": 442.261,
      "eval_steps_per_second": 13.893,
      "step": 27600
    },
    {
      "epoch": 219.12761714855435,
      "grad_norm": 216.87686157226562,
      "learning_rate": 2.987435177464922e-07,
      "loss": 22.6154,
      "step": 27610
    },
    {
      "epoch": 219.2073778664008,
      "grad_norm": 183.9648895263672,
      "learning_rate": 2.9862114073411786e-07,
      "loss": 26.0827,
      "step": 27620
    },
    {
      "epoch": 219.28713858424726,
      "grad_norm": 300.2879943847656,
      "learning_rate": 2.9849875161204277e-07,
      "loss": 21.6577,
      "step": 27630
    },
    {
      "epoch": 219.3668993020937,
      "grad_norm": 945.4083862304688,
      "learning_rate": 2.983763504107495e-07,
      "loss": 23.166,
      "step": 27640
    },
    {
      "epoch": 219.44666001994017,
      "grad_norm": 1042.5501708984375,
      "learning_rate": 2.9825393716072344e-07,
      "loss": 25.3464,
      "step": 27650
    },
    {
      "epoch": 219.52642073778665,
      "grad_norm": 315.9792785644531,
      "learning_rate": 2.981315118924533e-07,
      "loss": 21.6401,
      "step": 27660
    },
    {
      "epoch": 219.6061814556331,
      "grad_norm": 849.44921875,
      "learning_rate": 2.9800907463643063e-07,
      "loss": 25.0194,
      "step": 27670
    },
    {
      "epoch": 219.68594217347956,
      "grad_norm": 761.8780517578125,
      "learning_rate": 2.978866254231498e-07,
      "loss": 22.5823,
      "step": 27680
    },
    {
      "epoch": 219.76570289132601,
      "grad_norm": 815.2864379882812,
      "learning_rate": 2.977641642831085e-07,
      "loss": 22.0128,
      "step": 27690
    },
    {
      "epoch": 219.84546360917247,
      "grad_norm": 210.50242614746094,
      "learning_rate": 2.97641691246807e-07,
      "loss": 24.7933,
      "step": 27700
    },
    {
      "epoch": 219.84546360917247,
      "eval_loss": 2.9780495166778564,
      "eval_mae": 1.2930556535720825,
      "eval_mse": 2.9780492782592773,
      "eval_r2": 0.10272830724716187,
      "eval_rmse": 1.7257025462863747,
      "eval_runtime": 9.0794,
      "eval_samples_per_second": 441.771,
      "eval_steps_per_second": 13.878,
      "step": 27700
    },
    {
      "epoch": 219.92522432701895,
      "grad_norm": 755.2847290039062,
      "learning_rate": 2.975192063447489e-07,
      "loss": 23.8973,
      "step": 27710
    },
    {
      "epoch": 220.0,
      "grad_norm": 189.5387725830078,
      "learning_rate": 2.973967096074405e-07,
      "loss": 22.0406,
      "step": 27720
    },
    {
      "epoch": 220.07976071784645,
      "grad_norm": 973.175537109375,
      "learning_rate": 2.972742010653912e-07,
      "loss": 24.3044,
      "step": 27730
    },
    {
      "epoch": 220.1595214356929,
      "grad_norm": 270.5625305175781,
      "learning_rate": 2.971516807491132e-07,
      "loss": 24.0269,
      "step": 27740
    },
    {
      "epoch": 220.2392821535394,
      "grad_norm": 458.6932067871094,
      "learning_rate": 2.970291486891217e-07,
      "loss": 21.8895,
      "step": 27750
    },
    {
      "epoch": 220.31904287138585,
      "grad_norm": 547.2145385742188,
      "learning_rate": 2.9690660491593505e-07,
      "loss": 23.4881,
      "step": 27760
    },
    {
      "epoch": 220.3988035892323,
      "grad_norm": 706.6233520507812,
      "learning_rate": 2.9678404946007406e-07,
      "loss": 23.3825,
      "step": 27770
    },
    {
      "epoch": 220.47856430707876,
      "grad_norm": 335.0290832519531,
      "learning_rate": 2.9666148235206277e-07,
      "loss": 23.2111,
      "step": 27780
    },
    {
      "epoch": 220.5583250249252,
      "grad_norm": 266.097900390625,
      "learning_rate": 2.9653890362242805e-07,
      "loss": 23.4541,
      "step": 27790
    },
    {
      "epoch": 220.6380857427717,
      "grad_norm": 1648.7161865234375,
      "learning_rate": 2.964163133016998e-07,
      "loss": 24.1183,
      "step": 27800
    },
    {
      "epoch": 220.6380857427717,
      "eval_loss": 2.9777796268463135,
      "eval_mae": 1.3002476692199707,
      "eval_mse": 2.9777796268463135,
      "eval_r2": 0.10280954837799072,
      "eval_rmse": 1.7256244165073447,
      "eval_runtime": 9.1493,
      "eval_samples_per_second": 438.395,
      "eval_steps_per_second": 13.772,
      "step": 27800
    },
    {
      "epoch": 220.71784646061815,
      "grad_norm": 935.0693359375,
      "learning_rate": 2.962937114204104e-07,
      "loss": 23.7185,
      "step": 27810
    },
    {
      "epoch": 220.7976071784646,
      "grad_norm": 214.67987060546875,
      "learning_rate": 2.9617109800909544e-07,
      "loss": 25.207,
      "step": 27820
    },
    {
      "epoch": 220.87736789631106,
      "grad_norm": 528.9149780273438,
      "learning_rate": 2.960484730982935e-07,
      "loss": 23.291,
      "step": 27830
    },
    {
      "epoch": 220.95712861415754,
      "grad_norm": 151.2432861328125,
      "learning_rate": 2.959258367185457e-07,
      "loss": 23.163,
      "step": 27840
    },
    {
      "epoch": 221.0319042871386,
      "grad_norm": 839.2595825195312,
      "learning_rate": 2.9580318890039625e-07,
      "loss": 22.3116,
      "step": 27850
    },
    {
      "epoch": 221.11166500498504,
      "grad_norm": 226.69766235351562,
      "learning_rate": 2.956805296743919e-07,
      "loss": 24.5424,
      "step": 27860
    },
    {
      "epoch": 221.1914257228315,
      "grad_norm": 639.3292846679688,
      "learning_rate": 2.9555785907108274e-07,
      "loss": 23.6998,
      "step": 27870
    },
    {
      "epoch": 221.27118644067798,
      "grad_norm": 384.93487548828125,
      "learning_rate": 2.9543517712102124e-07,
      "loss": 24.3403,
      "step": 27880
    },
    {
      "epoch": 221.35094715852443,
      "grad_norm": 389.61181640625,
      "learning_rate": 2.953247536897466e-07,
      "loss": 24.1667,
      "step": 27890
    },
    {
      "epoch": 221.4307078763709,
      "grad_norm": 174.5286102294922,
      "learning_rate": 2.9520205026503843e-07,
      "loss": 22.76,
      "step": 27900
    },
    {
      "epoch": 221.4307078763709,
      "eval_loss": 2.9929704666137695,
      "eval_mae": 1.289552092552185,
      "eval_mse": 2.9929704666137695,
      "eval_r2": 0.09823262691497803,
      "eval_rmse": 1.7300203659534674,
      "eval_runtime": 9.1064,
      "eval_samples_per_second": 440.457,
      "eval_steps_per_second": 13.836,
      "step": 27900
    },
    {
      "epoch": 221.51046859421734,
      "grad_norm": 390.555908203125,
      "learning_rate": 2.950793355821965e-07,
      "loss": 21.9596,
      "step": 27910
    },
    {
      "epoch": 221.5902293120638,
      "grad_norm": 646.3958740234375,
      "learning_rate": 2.9495660967178455e-07,
      "loss": 24.6812,
      "step": 27920
    },
    {
      "epoch": 221.66999002991028,
      "grad_norm": 215.1212158203125,
      "learning_rate": 2.9483387256436886e-07,
      "loss": 24.6319,
      "step": 27930
    },
    {
      "epoch": 221.74975074775674,
      "grad_norm": 329.55859375,
      "learning_rate": 2.947111242905188e-07,
      "loss": 23.5231,
      "step": 27940
    },
    {
      "epoch": 221.8295114656032,
      "grad_norm": 870.319580078125,
      "learning_rate": 2.945883648808063e-07,
      "loss": 22.3164,
      "step": 27950
    },
    {
      "epoch": 221.90927218344964,
      "grad_norm": 375.7515869140625,
      "learning_rate": 2.9446559436580597e-07,
      "loss": 22.9546,
      "step": 27960
    },
    {
      "epoch": 221.9890329012961,
      "grad_norm": 353.8161926269531,
      "learning_rate": 2.9434281277609564e-07,
      "loss": 23.2956,
      "step": 27970
    },
    {
      "epoch": 222.06380857427718,
      "grad_norm": 753.7135009765625,
      "learning_rate": 2.9422002014225534e-07,
      "loss": 21.7364,
      "step": 27980
    },
    {
      "epoch": 222.14356929212363,
      "grad_norm": 159.11260986328125,
      "learning_rate": 2.9409721649486823e-07,
      "loss": 23.5395,
      "step": 27990
    },
    {
      "epoch": 222.22333000997008,
      "grad_norm": 284.4923095703125,
      "learning_rate": 2.9397440186452e-07,
      "loss": 22.6555,
      "step": 28000
    },
    {
      "epoch": 222.22333000997008,
      "eval_loss": 2.978847026824951,
      "eval_mae": 1.289054274559021,
      "eval_mse": 2.978847026824951,
      "eval_r2": 0.10248798131942749,
      "eval_rmse": 1.7259336681416673,
      "eval_runtime": 9.1089,
      "eval_samples_per_second": 440.34,
      "eval_steps_per_second": 13.833,
      "step": 28000
    },
    {
      "epoch": 222.30309072781654,
      "grad_norm": 795.4385375976562,
      "learning_rate": 2.938515762817992e-07,
      "loss": 24.3238,
      "step": 28010
    },
    {
      "epoch": 222.38285144566302,
      "grad_norm": 1256.491455078125,
      "learning_rate": 2.9372873977729715e-07,
      "loss": 23.0592,
      "step": 28020
    },
    {
      "epoch": 222.46261216350948,
      "grad_norm": 382.3968811035156,
      "learning_rate": 2.936058923816077e-07,
      "loss": 20.7059,
      "step": 28030
    },
    {
      "epoch": 222.54237288135593,
      "grad_norm": 209.00186157226562,
      "learning_rate": 2.934830341253276e-07,
      "loss": 24.3814,
      "step": 28040
    },
    {
      "epoch": 222.62213359920239,
      "grad_norm": 247.73770141601562,
      "learning_rate": 2.9336016503905615e-07,
      "loss": 23.7226,
      "step": 28050
    },
    {
      "epoch": 222.70189431704884,
      "grad_norm": 259.9332580566406,
      "learning_rate": 2.9323728515339554e-07,
      "loss": 24.3797,
      "step": 28060
    },
    {
      "epoch": 222.78165503489532,
      "grad_norm": 490.0107116699219,
      "learning_rate": 2.931143944989505e-07,
      "loss": 23.7283,
      "step": 28070
    },
    {
      "epoch": 222.86141575274178,
      "grad_norm": 457.6901550292969,
      "learning_rate": 2.929914931063283e-07,
      "loss": 25.6897,
      "step": 28080
    },
    {
      "epoch": 222.94117647058823,
      "grad_norm": 647.62939453125,
      "learning_rate": 2.9286858100613934e-07,
      "loss": 22.9767,
      "step": 28090
    },
    {
      "epoch": 223.01595214356928,
      "grad_norm": 668.23193359375,
      "learning_rate": 2.927456582289962e-07,
      "loss": 22.2384,
      "step": 28100
    },
    {
      "epoch": 223.01595214356928,
      "eval_loss": 2.9920876026153564,
      "eval_mae": 1.3085302114486694,
      "eval_mse": 2.9920876026153564,
      "eval_r2": 0.09849858283996582,
      "eval_rmse": 1.729765187132449,
      "eval_runtime": 9.1242,
      "eval_samples_per_second": 439.601,
      "eval_steps_per_second": 13.809,
      "step": 28100
    },
    {
      "epoch": 223.09571286141576,
      "grad_norm": 302.5481262207031,
      "learning_rate": 2.9262272480551453e-07,
      "loss": 22.2556,
      "step": 28110
    },
    {
      "epoch": 223.17547357926222,
      "grad_norm": 153.3507843017578,
      "learning_rate": 2.9249978076631214e-07,
      "loss": 23.7961,
      "step": 28120
    },
    {
      "epoch": 223.25523429710867,
      "grad_norm": 226.6240997314453,
      "learning_rate": 2.923768261420101e-07,
      "loss": 22.6181,
      "step": 28130
    },
    {
      "epoch": 223.33499501495513,
      "grad_norm": 372.73846435546875,
      "learning_rate": 2.922538609632316e-07,
      "loss": 23.4664,
      "step": 28140
    },
    {
      "epoch": 223.4147557328016,
      "grad_norm": 544.7764282226562,
      "learning_rate": 2.921308852606026e-07,
      "loss": 23.3307,
      "step": 28150
    },
    {
      "epoch": 223.49451645064806,
      "grad_norm": 172.4590606689453,
      "learning_rate": 2.920078990647519e-07,
      "loss": 23.3574,
      "step": 28160
    },
    {
      "epoch": 223.57427716849452,
      "grad_norm": 1462.9517822265625,
      "learning_rate": 2.9188490240631057e-07,
      "loss": 24.8231,
      "step": 28170
    },
    {
      "epoch": 223.65403788634097,
      "grad_norm": 255.56784057617188,
      "learning_rate": 2.9176189531591264e-07,
      "loss": 24.927,
      "step": 28180
    },
    {
      "epoch": 223.73379860418743,
      "grad_norm": 543.3631591796875,
      "learning_rate": 2.9163887782419444e-07,
      "loss": 24.3094,
      "step": 28190
    },
    {
      "epoch": 223.8135593220339,
      "grad_norm": 197.2646484375,
      "learning_rate": 2.915158499617951e-07,
      "loss": 22.7073,
      "step": 28200
    },
    {
      "epoch": 223.8135593220339,
      "eval_loss": 2.9783763885498047,
      "eval_mae": 1.2938679456710815,
      "eval_mse": 2.9783763885498047,
      "eval_r2": 0.10262978076934814,
      "eval_rmse": 1.7257973196612066,
      "eval_runtime": 9.1137,
      "eval_samples_per_second": 440.105,
      "eval_steps_per_second": 13.825,
      "step": 28200
    },
    {
      "epoch": 223.89332003988036,
      "grad_norm": 530.704345703125,
      "learning_rate": 2.9139281175935603e-07,
      "loss": 22.9435,
      "step": 28210
    },
    {
      "epoch": 223.97308075772682,
      "grad_norm": 1255.1724853515625,
      "learning_rate": 2.9126976324752174e-07,
      "loss": 23.3073,
      "step": 28220
    },
    {
      "epoch": 224.04785643070787,
      "grad_norm": 917.60205078125,
      "learning_rate": 2.911467044569387e-07,
      "loss": 23.6917,
      "step": 28230
    },
    {
      "epoch": 224.12761714855435,
      "grad_norm": 565.042724609375,
      "learning_rate": 2.910236354182564e-07,
      "loss": 22.5874,
      "step": 28240
    },
    {
      "epoch": 224.2073778664008,
      "grad_norm": 642.9319458007812,
      "learning_rate": 2.909005561621267e-07,
      "loss": 24.358,
      "step": 28250
    },
    {
      "epoch": 224.28713858424726,
      "grad_norm": 1091.9749755859375,
      "learning_rate": 2.9077746671920406e-07,
      "loss": 23.8511,
      "step": 28260
    },
    {
      "epoch": 224.3668993020937,
      "grad_norm": 1002.00537109375,
      "learning_rate": 2.9065436712014536e-07,
      "loss": 23.917,
      "step": 28270
    },
    {
      "epoch": 224.44666001994017,
      "grad_norm": 291.5292053222656,
      "learning_rate": 2.9053125739561e-07,
      "loss": 25.6457,
      "step": 28280
    },
    {
      "epoch": 224.52642073778665,
      "grad_norm": 488.26898193359375,
      "learning_rate": 2.9040813757626027e-07,
      "loss": 22.2785,
      "step": 28290
    },
    {
      "epoch": 224.6061814556331,
      "grad_norm": 1863.35546875,
      "learning_rate": 2.9028500769276044e-07,
      "loss": 22.4234,
      "step": 28300
    },
    {
      "epoch": 224.6061814556331,
      "eval_loss": 2.9912896156311035,
      "eval_mae": 1.3089357614517212,
      "eval_mse": 2.9912896156311035,
      "eval_r2": 0.09873908758163452,
      "eval_rmse": 1.7295345083666598,
      "eval_runtime": 9.1717,
      "eval_samples_per_second": 437.324,
      "eval_steps_per_second": 13.738,
      "step": 28300
    },
    {
      "epoch": 224.68594217347956,
      "grad_norm": 455.0877380371094,
      "learning_rate": 2.9016186777577757e-07,
      "loss": 24.4669,
      "step": 28310
    },
    {
      "epoch": 224.76570289132601,
      "grad_norm": 307.5437316894531,
      "learning_rate": 2.9003871785598135e-07,
      "loss": 23.6653,
      "step": 28320
    },
    {
      "epoch": 224.84546360917247,
      "grad_norm": 477.96917724609375,
      "learning_rate": 2.899155579640436e-07,
      "loss": 22.3114,
      "step": 28330
    },
    {
      "epoch": 224.92522432701895,
      "grad_norm": 408.08612060546875,
      "learning_rate": 2.8979238813063904e-07,
      "loss": 22.7173,
      "step": 28340
    },
    {
      "epoch": 225.0,
      "grad_norm": 688.6211547851562,
      "learning_rate": 2.8966920838644445e-07,
      "loss": 22.0629,
      "step": 28350
    },
    {
      "epoch": 225.07976071784645,
      "grad_norm": 215.12020874023438,
      "learning_rate": 2.895460187621394e-07,
      "loss": 23.4137,
      "step": 28360
    },
    {
      "epoch": 225.1595214356929,
      "grad_norm": 230.41818237304688,
      "learning_rate": 2.8942281928840573e-07,
      "loss": 23.5063,
      "step": 28370
    },
    {
      "epoch": 225.2392821535394,
      "grad_norm": 909.58740234375,
      "learning_rate": 2.8929960999592783e-07,
      "loss": 23.423,
      "step": 28380
    },
    {
      "epoch": 225.31904287138585,
      "grad_norm": 976.6878051757812,
      "learning_rate": 2.8917639091539256e-07,
      "loss": 23.2553,
      "step": 28390
    },
    {
      "epoch": 225.3988035892323,
      "grad_norm": 406.7839660644531,
      "learning_rate": 2.8905316207748916e-07,
      "loss": 23.7409,
      "step": 28400
    },
    {
      "epoch": 225.3988035892323,
      "eval_loss": 2.9810709953308105,
      "eval_mae": 1.2892674207687378,
      "eval_mse": 2.9810709953308105,
      "eval_r2": 0.10181784629821777,
      "eval_rmse": 1.7265778277653199,
      "eval_runtime": 9.0967,
      "eval_samples_per_second": 440.931,
      "eval_steps_per_second": 13.851,
      "step": 28400
    },
    {
      "epoch": 225.47856430707876,
      "grad_norm": 291.175048828125,
      "learning_rate": 2.889299235129092e-07,
      "loss": 25.8447,
      "step": 28410
    },
    {
      "epoch": 225.5583250249252,
      "grad_norm": 547.2565307617188,
      "learning_rate": 2.888066752523469e-07,
      "loss": 25.7814,
      "step": 28420
    },
    {
      "epoch": 225.6380857427717,
      "grad_norm": 370.3689880371094,
      "learning_rate": 2.8868341732649877e-07,
      "loss": 23.7016,
      "step": 28430
    },
    {
      "epoch": 225.71784646061815,
      "grad_norm": 180.25758361816406,
      "learning_rate": 2.885601497660636e-07,
      "loss": 21.226,
      "step": 28440
    },
    {
      "epoch": 225.7976071784646,
      "grad_norm": 1098.9267578125,
      "learning_rate": 2.884368726017428e-07,
      "loss": 23.4176,
      "step": 28450
    },
    {
      "epoch": 225.87736789631106,
      "grad_norm": 535.2625122070312,
      "learning_rate": 2.8831358586424016e-07,
      "loss": 22.0611,
      "step": 28460
    },
    {
      "epoch": 225.95712861415754,
      "grad_norm": 966.4929809570312,
      "learning_rate": 2.881902895842616e-07,
      "loss": 23.7503,
      "step": 28470
    },
    {
      "epoch": 226.0319042871386,
      "grad_norm": 1121.6949462890625,
      "learning_rate": 2.880669837925158e-07,
      "loss": 21.1102,
      "step": 28480
    },
    {
      "epoch": 226.11166500498504,
      "grad_norm": 576.8964233398438,
      "learning_rate": 2.879436685197134e-07,
      "loss": 23.5315,
      "step": 28490
    },
    {
      "epoch": 226.1914257228315,
      "grad_norm": 332.62518310546875,
      "learning_rate": 2.8782034379656786e-07,
      "loss": 22.4616,
      "step": 28500
    },
    {
      "epoch": 226.1914257228315,
      "eval_loss": 2.976295232772827,
      "eval_mae": 1.2895312309265137,
      "eval_mse": 2.976295232772827,
      "eval_r2": 0.10325682163238525,
      "eval_rmse": 1.7251942594307539,
      "eval_runtime": 9.0972,
      "eval_samples_per_second": 440.907,
      "eval_steps_per_second": 13.85,
      "step": 28500
    },
    {
      "epoch": 226.27118644067798,
      "grad_norm": 968.8788452148438,
      "learning_rate": 2.8769700965379447e-07,
      "loss": 23.2409,
      "step": 28510
    },
    {
      "epoch": 226.35094715852443,
      "grad_norm": 423.68475341796875,
      "learning_rate": 2.875736661221112e-07,
      "loss": 24.2685,
      "step": 28520
    },
    {
      "epoch": 226.4307078763709,
      "grad_norm": 540.13037109375,
      "learning_rate": 2.874503132322384e-07,
      "loss": 25.0156,
      "step": 28530
    },
    {
      "epoch": 226.51046859421734,
      "grad_norm": 504.5898132324219,
      "learning_rate": 2.8732695101489864e-07,
      "loss": 26.6463,
      "step": 28540
    },
    {
      "epoch": 226.5902293120638,
      "grad_norm": 1567.54638671875,
      "learning_rate": 2.872035795008167e-07,
      "loss": 22.8101,
      "step": 28550
    },
    {
      "epoch": 226.66999002991028,
      "grad_norm": 460.20635986328125,
      "learning_rate": 2.870801987207198e-07,
      "loss": 23.7526,
      "step": 28560
    },
    {
      "epoch": 226.74975074775674,
      "grad_norm": 661.027587890625,
      "learning_rate": 2.869568087053376e-07,
      "loss": 22.6854,
      "step": 28570
    },
    {
      "epoch": 226.8295114656032,
      "grad_norm": 478.668212890625,
      "learning_rate": 2.868334094854018e-07,
      "loss": 23.7253,
      "step": 28580
    },
    {
      "epoch": 226.90927218344964,
      "grad_norm": 252.7202606201172,
      "learning_rate": 2.867100010916466e-07,
      "loss": 22.2806,
      "step": 28590
    },
    {
      "epoch": 226.9890329012961,
      "grad_norm": 1410.196533203125,
      "learning_rate": 2.8658658355480824e-07,
      "loss": 21.6722,
      "step": 28600
    },
    {
      "epoch": 226.9890329012961,
      "eval_loss": 2.989102840423584,
      "eval_mae": 1.3085246086120605,
      "eval_mse": 2.989103078842163,
      "eval_r2": 0.09939789772033691,
      "eval_rmse": 1.7289022756773047,
      "eval_runtime": 9.0938,
      "eval_samples_per_second": 441.067,
      "eval_steps_per_second": 13.856,
      "step": 28600
    },
    {
      "epoch": 227.06380857427718,
      "grad_norm": 461.0039367675781,
      "learning_rate": 2.864631569056255e-07,
      "loss": 23.4921,
      "step": 28610
    },
    {
      "epoch": 227.14356929212363,
      "grad_norm": 921.9505004882812,
      "learning_rate": 2.863397211748395e-07,
      "loss": 24.8094,
      "step": 28620
    },
    {
      "epoch": 227.22333000997008,
      "grad_norm": 426.26715087890625,
      "learning_rate": 2.8621627639319314e-07,
      "loss": 21.713,
      "step": 28630
    },
    {
      "epoch": 227.30309072781654,
      "grad_norm": 1372.0335693359375,
      "learning_rate": 2.86092822591432e-07,
      "loss": 24.5109,
      "step": 28640
    },
    {
      "epoch": 227.38285144566302,
      "grad_norm": 1721.040283203125,
      "learning_rate": 2.859693598003038e-07,
      "loss": 25.2417,
      "step": 28650
    },
    {
      "epoch": 227.46261216350948,
      "grad_norm": 663.2008056640625,
      "learning_rate": 2.8584588805055854e-07,
      "loss": 22.8608,
      "step": 28660
    },
    {
      "epoch": 227.54237288135593,
      "grad_norm": 262.5372009277344,
      "learning_rate": 2.857224073729483e-07,
      "loss": 23.815,
      "step": 28670
    },
    {
      "epoch": 227.62213359920239,
      "grad_norm": 593.6583862304688,
      "learning_rate": 2.855989177982275e-07,
      "loss": 23.1764,
      "step": 28680
    },
    {
      "epoch": 227.70189431704884,
      "grad_norm": 270.80657958984375,
      "learning_rate": 2.854754193571528e-07,
      "loss": 23.716,
      "step": 28690
    },
    {
      "epoch": 227.78165503489532,
      "grad_norm": 495.4303283691406,
      "learning_rate": 2.8535191208048306e-07,
      "loss": 20.9373,
      "step": 28700
    },
    {
      "epoch": 227.78165503489532,
      "eval_loss": 2.97832989692688,
      "eval_mae": 1.2897449731826782,
      "eval_mse": 2.978330135345459,
      "eval_r2": 0.10264372825622559,
      "eval_rmse": 1.7257839190771997,
      "eval_runtime": 9.105,
      "eval_samples_per_second": 440.528,
      "eval_steps_per_second": 13.839,
      "step": 28700
    },
    {
      "epoch": 227.86141575274178,
      "grad_norm": 1146.97900390625,
      "learning_rate": 2.852283959989792e-07,
      "loss": 23.5481,
      "step": 28710
    },
    {
      "epoch": 227.94117647058823,
      "grad_norm": 376.2318115234375,
      "learning_rate": 2.8510487114340444e-07,
      "loss": 23.3373,
      "step": 28720
    },
    {
      "epoch": 228.01595214356928,
      "grad_norm": 278.4344177246094,
      "learning_rate": 2.849813375445243e-07,
      "loss": 22.1801,
      "step": 28730
    },
    {
      "epoch": 228.09571286141576,
      "grad_norm": 360.8177490234375,
      "learning_rate": 2.848577952331063e-07,
      "loss": 23.7975,
      "step": 28740
    },
    {
      "epoch": 228.17547357926222,
      "grad_norm": 780.6703491210938,
      "learning_rate": 2.847342442399201e-07,
      "loss": 22.6753,
      "step": 28750
    },
    {
      "epoch": 228.25523429710867,
      "grad_norm": 1090.07470703125,
      "learning_rate": 2.846106845957378e-07,
      "loss": 22.1895,
      "step": 28760
    },
    {
      "epoch": 228.33499501495513,
      "grad_norm": 433.43121337890625,
      "learning_rate": 2.844871163313333e-07,
      "loss": 23.5747,
      "step": 28770
    },
    {
      "epoch": 228.4147557328016,
      "grad_norm": 538.4068603515625,
      "learning_rate": 2.843635394774829e-07,
      "loss": 24.5557,
      "step": 28780
    },
    {
      "epoch": 228.49451645064806,
      "grad_norm": 1891.75439453125,
      "learning_rate": 2.84239954064965e-07,
      "loss": 24.9661,
      "step": 28790
    },
    {
      "epoch": 228.57427716849452,
      "grad_norm": 181.74830627441406,
      "learning_rate": 2.8411636012455996e-07,
      "loss": 24.4292,
      "step": 28800
    },
    {
      "epoch": 228.57427716849452,
      "eval_loss": 2.9761176109313965,
      "eval_mae": 1.2910175323486328,
      "eval_mse": 2.9761176109313965,
      "eval_r2": 0.10331028699874878,
      "eval_rmse": 1.7251427798682046,
      "eval_runtime": 9.0926,
      "eval_samples_per_second": 441.129,
      "eval_steps_per_second": 13.857,
      "step": 28800
    },
    {
      "epoch": 228.65403788634097,
      "grad_norm": 230.27720642089844,
      "learning_rate": 2.8399275768705043e-07,
      "loss": 23.979,
      "step": 28810
    },
    {
      "epoch": 228.73379860418743,
      "grad_norm": 303.1219482421875,
      "learning_rate": 2.8386914678322123e-07,
      "loss": 20.456,
      "step": 28820
    },
    {
      "epoch": 228.8135593220339,
      "grad_norm": 983.5076293945312,
      "learning_rate": 2.837455274438591e-07,
      "loss": 22.9024,
      "step": 28830
    },
    {
      "epoch": 228.89332003988036,
      "grad_norm": 265.1124267578125,
      "learning_rate": 2.83621899699753e-07,
      "loss": 23.7586,
      "step": 28840
    },
    {
      "epoch": 228.97308075772682,
      "grad_norm": 951.542236328125,
      "learning_rate": 2.834982635816939e-07,
      "loss": 24.138,
      "step": 28850
    },
    {
      "epoch": 229.04785643070787,
      "grad_norm": 1288.7901611328125,
      "learning_rate": 2.833746191204751e-07,
      "loss": 22.0687,
      "step": 28860
    },
    {
      "epoch": 229.12761714855435,
      "grad_norm": 963.5640869140625,
      "learning_rate": 2.8325096634689166e-07,
      "loss": 22.8958,
      "step": 28870
    },
    {
      "epoch": 229.2073778664008,
      "grad_norm": 606.7884521484375,
      "learning_rate": 2.831273052917408e-07,
      "loss": 24.818,
      "step": 28880
    },
    {
      "epoch": 229.28713858424726,
      "grad_norm": 287.0824279785156,
      "learning_rate": 2.8300363598582187e-07,
      "loss": 24.0999,
      "step": 28890
    },
    {
      "epoch": 229.3668993020937,
      "grad_norm": 1716.82421875,
      "learning_rate": 2.828799584599364e-07,
      "loss": 25.2726,
      "step": 28900
    },
    {
      "epoch": 229.3668993020937,
      "eval_loss": 2.980441093444824,
      "eval_mae": 1.302371621131897,
      "eval_mse": 2.980441093444824,
      "eval_r2": 0.10200768709182739,
      "eval_rmse": 1.7263954047218801,
      "eval_runtime": 9.1039,
      "eval_samples_per_second": 440.579,
      "eval_steps_per_second": 13.84,
      "step": 28900
    },
    {
      "epoch": 229.44666001994017,
      "grad_norm": 333.4840393066406,
      "learning_rate": 2.8275627274488755e-07,
      "loss": 23.8886,
      "step": 28910
    },
    {
      "epoch": 229.52642073778665,
      "grad_norm": 1196.2061767578125,
      "learning_rate": 2.826325788714811e-07,
      "loss": 22.0717,
      "step": 28920
    },
    {
      "epoch": 229.6061814556331,
      "grad_norm": 377.2966003417969,
      "learning_rate": 2.8250887687052425e-07,
      "loss": 24.145,
      "step": 28930
    },
    {
      "epoch": 229.68594217347956,
      "grad_norm": 529.4716186523438,
      "learning_rate": 2.823851667728268e-07,
      "loss": 22.3141,
      "step": 28940
    },
    {
      "epoch": 229.76570289132601,
      "grad_norm": 199.5446319580078,
      "learning_rate": 2.822614486092e-07,
      "loss": 22.8616,
      "step": 28950
    },
    {
      "epoch": 229.84546360917247,
      "grad_norm": 259.5824279785156,
      "learning_rate": 2.8213772241045767e-07,
      "loss": 23.8629,
      "step": 28960
    },
    {
      "epoch": 229.92522432701895,
      "grad_norm": 295.1745910644531,
      "learning_rate": 2.8201398820741516e-07,
      "loss": 23.6198,
      "step": 28970
    },
    {
      "epoch": 230.0,
      "grad_norm": 241.26560974121094,
      "learning_rate": 2.8189024603089016e-07,
      "loss": 20.7605,
      "step": 28980
    },
    {
      "epoch": 230.07976071784645,
      "grad_norm": 640.4485473632812,
      "learning_rate": 2.8176649591170217e-07,
      "loss": 20.1089,
      "step": 28990
    },
    {
      "epoch": 230.1595214356929,
      "grad_norm": 298.4513854980469,
      "learning_rate": 2.816427378806726e-07,
      "loss": 24.6293,
      "step": 29000
    },
    {
      "epoch": 230.1595214356929,
      "eval_loss": 2.9736437797546387,
      "eval_mae": 1.2933244705200195,
      "eval_mse": 2.9736437797546387,
      "eval_r2": 0.10405564308166504,
      "eval_rmse": 1.72442563764131,
      "eval_runtime": 9.0946,
      "eval_samples_per_second": 441.03,
      "eval_steps_per_second": 13.854,
      "step": 29000
    },
    {
      "epoch": 230.2392821535394,
      "grad_norm": 215.43040466308594,
      "learning_rate": 2.815189719686251e-07,
      "loss": 21.8872,
      "step": 29010
    },
    {
      "epoch": 230.31904287138585,
      "grad_norm": 339.0171203613281,
      "learning_rate": 2.813951982063849e-07,
      "loss": 22.4703,
      "step": 29020
    },
    {
      "epoch": 230.3988035892323,
      "grad_norm": 963.80029296875,
      "learning_rate": 2.8127141662477953e-07,
      "loss": 24.7596,
      "step": 29030
    },
    {
      "epoch": 230.47856430707876,
      "grad_norm": 246.45277404785156,
      "learning_rate": 2.811476272546383e-07,
      "loss": 22.9733,
      "step": 29040
    },
    {
      "epoch": 230.5583250249252,
      "grad_norm": 978.0501098632812,
      "learning_rate": 2.8102383012679255e-07,
      "loss": 23.1466,
      "step": 29050
    },
    {
      "epoch": 230.6380857427717,
      "grad_norm": 342.236328125,
      "learning_rate": 2.8090002527207547e-07,
      "loss": 23.5417,
      "step": 29060
    },
    {
      "epoch": 230.71784646061815,
      "grad_norm": 542.9784545898438,
      "learning_rate": 2.807762127213221e-07,
      "loss": 24.5957,
      "step": 29070
    },
    {
      "epoch": 230.7976071784646,
      "grad_norm": 673.6538696289062,
      "learning_rate": 2.8065239250536967e-07,
      "loss": 23.7822,
      "step": 29080
    },
    {
      "epoch": 230.87736789631106,
      "grad_norm": 489.96136474609375,
      "learning_rate": 2.80528564655057e-07,
      "loss": 25.4736,
      "step": 29090
    },
    {
      "epoch": 230.95712861415754,
      "grad_norm": 297.83221435546875,
      "learning_rate": 2.804047292012251e-07,
      "loss": 23.9836,
      "step": 29100
    },
    {
      "epoch": 230.95712861415754,
      "eval_loss": 2.975355863571167,
      "eval_mae": 1.2902168035507202,
      "eval_mse": 2.975355863571167,
      "eval_r2": 0.10353982448577881,
      "eval_rmse": 1.7249219876768824,
      "eval_runtime": 9.1099,
      "eval_samples_per_second": 440.291,
      "eval_steps_per_second": 13.831,
      "step": 29100
    },
    {
      "epoch": 231.0319042871386,
      "grad_norm": 830.2015380859375,
      "learning_rate": 2.802808861747166e-07,
      "loss": 23.1082,
      "step": 29110
    },
    {
      "epoch": 231.11166500498504,
      "grad_norm": 594.7340698242188,
      "learning_rate": 2.801570356063761e-07,
      "loss": 23.6181,
      "step": 29120
    },
    {
      "epoch": 231.1914257228315,
      "grad_norm": 357.4686279296875,
      "learning_rate": 2.8003317752705035e-07,
      "loss": 24.4905,
      "step": 29130
    },
    {
      "epoch": 231.27118644067798,
      "grad_norm": 425.6934509277344,
      "learning_rate": 2.799093119675876e-07,
      "loss": 22.8521,
      "step": 29140
    },
    {
      "epoch": 231.35094715852443,
      "grad_norm": 204.07247924804688,
      "learning_rate": 2.7978543895883813e-07,
      "loss": 23.1028,
      "step": 29150
    },
    {
      "epoch": 231.4307078763709,
      "grad_norm": 956.6591186523438,
      "learning_rate": 2.796615585316541e-07,
      "loss": 22.0427,
      "step": 29160
    },
    {
      "epoch": 231.51046859421734,
      "grad_norm": 947.2444458007812,
      "learning_rate": 2.795376707168894e-07,
      "loss": 23.0949,
      "step": 29170
    },
    {
      "epoch": 231.5902293120638,
      "grad_norm": 689.728271484375,
      "learning_rate": 2.794137755453999e-07,
      "loss": 25.3765,
      "step": 29180
    },
    {
      "epoch": 231.66999002991028,
      "grad_norm": 980.9293212890625,
      "learning_rate": 2.7928987304804315e-07,
      "loss": 24.1371,
      "step": 29190
    },
    {
      "epoch": 231.74975074775674,
      "grad_norm": 501.6930236816406,
      "learning_rate": 2.7916596325567864e-07,
      "loss": 23.0343,
      "step": 29200
    },
    {
      "epoch": 231.74975074775674,
      "eval_loss": 2.989731788635254,
      "eval_mae": 1.2870382070541382,
      "eval_mse": 2.989731788635254,
      "eval_r2": 0.09920841455459595,
      "eval_rmse": 1.72908408952117,
      "eval_runtime": 9.0984,
      "eval_samples_per_second": 440.845,
      "eval_steps_per_second": 13.849,
      "step": 29200
    },
    {
      "epoch": 231.8295114656032,
      "grad_norm": 277.04345703125,
      "learning_rate": 2.7904204619916773e-07,
      "loss": 23.9762,
      "step": 29210
    },
    {
      "epoch": 231.90927218344964,
      "grad_norm": 975.6008911132812,
      "learning_rate": 2.7891812190937343e-07,
      "loss": 23.6602,
      "step": 29220
    },
    {
      "epoch": 231.9890329012961,
      "grad_norm": 1536.518798828125,
      "learning_rate": 2.7879419041716063e-07,
      "loss": 21.931,
      "step": 29230
    },
    {
      "epoch": 232.06380857427718,
      "grad_norm": 378.703125,
      "learning_rate": 2.7867025175339605e-07,
      "loss": 21.9565,
      "step": 29240
    },
    {
      "epoch": 232.14356929212363,
      "grad_norm": 755.17138671875,
      "learning_rate": 2.785463059489481e-07,
      "loss": 23.7374,
      "step": 29250
    },
    {
      "epoch": 232.22333000997008,
      "grad_norm": 478.4455261230469,
      "learning_rate": 2.7842235303468715e-07,
      "loss": 23.8411,
      "step": 29260
    },
    {
      "epoch": 232.30309072781654,
      "grad_norm": 457.9610595703125,
      "learning_rate": 2.78298393041485e-07,
      "loss": 24.6312,
      "step": 29270
    },
    {
      "epoch": 232.38285144566302,
      "grad_norm": 350.7718505859375,
      "learning_rate": 2.7817442600021565e-07,
      "loss": 21.8555,
      "step": 29280
    },
    {
      "epoch": 232.46261216350948,
      "grad_norm": 282.9169006347656,
      "learning_rate": 2.7805045194175455e-07,
      "loss": 22.2663,
      "step": 29290
    },
    {
      "epoch": 232.54237288135593,
      "grad_norm": 893.3816528320312,
      "learning_rate": 2.779264708969789e-07,
      "loss": 22.8199,
      "step": 29300
    },
    {
      "epoch": 232.54237288135593,
      "eval_loss": 2.9826467037200928,
      "eval_mae": 1.2893154621124268,
      "eval_mse": 2.982646942138672,
      "eval_r2": 0.10134309530258179,
      "eval_rmse": 1.7270341461993945,
      "eval_runtime": 9.1576,
      "eval_samples_per_second": 437.995,
      "eval_steps_per_second": 13.759,
      "step": 29300
    },
    {
      "epoch": 232.62213359920239,
      "grad_norm": 1312.9954833984375,
      "learning_rate": 2.778024828967679e-07,
      "loss": 24.4557,
      "step": 29310
    },
    {
      "epoch": 232.70189431704884,
      "grad_norm": 240.9700927734375,
      "learning_rate": 2.776784879720021e-07,
      "loss": 24.1606,
      "step": 29320
    },
    {
      "epoch": 232.78165503489532,
      "grad_norm": 252.70970153808594,
      "learning_rate": 2.775544861535642e-07,
      "loss": 23.4388,
      "step": 29330
    },
    {
      "epoch": 232.86141575274178,
      "grad_norm": 442.65777587890625,
      "learning_rate": 2.774304774723383e-07,
      "loss": 22.6744,
      "step": 29340
    },
    {
      "epoch": 232.94117647058823,
      "grad_norm": 507.1012268066406,
      "learning_rate": 2.7730646195921014e-07,
      "loss": 22.4341,
      "step": 29350
    },
    {
      "epoch": 233.01595214356928,
      "grad_norm": 320.151123046875,
      "learning_rate": 2.771824396450676e-07,
      "loss": 24.6485,
      "step": 29360
    },
    {
      "epoch": 233.09571286141576,
      "grad_norm": 268.9784851074219,
      "learning_rate": 2.770584105607999e-07,
      "loss": 23.6512,
      "step": 29370
    },
    {
      "epoch": 233.17547357926222,
      "grad_norm": 196.93817138671875,
      "learning_rate": 2.7693437473729794e-07,
      "loss": 23.2546,
      "step": 29380
    },
    {
      "epoch": 233.25523429710867,
      "grad_norm": 672.9905395507812,
      "learning_rate": 2.7681033220545434e-07,
      "loss": 21.759,
      "step": 29390
    },
    {
      "epoch": 233.33499501495513,
      "grad_norm": 838.8317260742188,
      "learning_rate": 2.766862829961637e-07,
      "loss": 24.6047,
      "step": 29400
    },
    {
      "epoch": 233.33499501495513,
      "eval_loss": 2.9748923778533936,
      "eval_mae": 1.2966049909591675,
      "eval_mse": 2.9748923778533936,
      "eval_r2": 0.10367947816848755,
      "eval_rmse": 1.7247876326821785,
      "eval_runtime": 9.1814,
      "eval_samples_per_second": 436.86,
      "eval_steps_per_second": 13.723,
      "step": 29400
    },
    {
      "epoch": 233.4147557328016,
      "grad_norm": 738.0784301757812,
      "learning_rate": 2.7656222714032177e-07,
      "loss": 24.5345,
      "step": 29410
    },
    {
      "epoch": 233.49451645064806,
      "grad_norm": 947.1227416992188,
      "learning_rate": 2.764381646688263e-07,
      "loss": 22.5708,
      "step": 29420
    },
    {
      "epoch": 233.57427716849452,
      "grad_norm": 174.71717834472656,
      "learning_rate": 2.763140956125766e-07,
      "loss": 21.0811,
      "step": 29430
    },
    {
      "epoch": 233.65403788634097,
      "grad_norm": 1085.6224365234375,
      "learning_rate": 2.761900200024736e-07,
      "loss": 25.2356,
      "step": 29440
    },
    {
      "epoch": 233.73379860418743,
      "grad_norm": 450.60736083984375,
      "learning_rate": 2.760659378694199e-07,
      "loss": 23.2346,
      "step": 29450
    },
    {
      "epoch": 233.8135593220339,
      "grad_norm": 697.2225952148438,
      "learning_rate": 2.7594184924431967e-07,
      "loss": 23.9659,
      "step": 29460
    },
    {
      "epoch": 233.89332003988036,
      "grad_norm": 706.6405639648438,
      "learning_rate": 2.7581775415807874e-07,
      "loss": 24.0092,
      "step": 29470
    },
    {
      "epoch": 233.97308075772682,
      "grad_norm": 613.9629516601562,
      "learning_rate": 2.756936526416045e-07,
      "loss": 24.363,
      "step": 29480
    },
    {
      "epoch": 234.04785643070787,
      "grad_norm": 964.1812744140625,
      "learning_rate": 2.7556954472580603e-07,
      "loss": 20.0324,
      "step": 29490
    },
    {
      "epoch": 234.12761714855435,
      "grad_norm": 167.4209747314453,
      "learning_rate": 2.7544543044159394e-07,
      "loss": 23.8551,
      "step": 29500
    },
    {
      "epoch": 234.12761714855435,
      "eval_loss": 2.9705073833465576,
      "eval_mae": 1.2916990518569946,
      "eval_mse": 2.9705073833465576,
      "eval_r2": 0.10500067472457886,
      "eval_rmse": 1.7235159945142828,
      "eval_runtime": 9.1593,
      "eval_samples_per_second": 437.915,
      "eval_steps_per_second": 13.756,
      "step": 29500
    },
    {
      "epoch": 234.2073778664008,
      "grad_norm": 148.0955352783203,
      "learning_rate": 2.7532130981988045e-07,
      "loss": 21.6222,
      "step": 29510
    },
    {
      "epoch": 234.28713858424726,
      "grad_norm": 335.8790588378906,
      "learning_rate": 2.751971828915793e-07,
      "loss": 25.0059,
      "step": 29520
    },
    {
      "epoch": 234.3668993020937,
      "grad_norm": 564.429931640625,
      "learning_rate": 2.750730496876059e-07,
      "loss": 23.5836,
      "step": 29530
    },
    {
      "epoch": 234.44666001994017,
      "grad_norm": 883.6195678710938,
      "learning_rate": 2.749489102388772e-07,
      "loss": 23.2661,
      "step": 29540
    },
    {
      "epoch": 234.52642073778665,
      "grad_norm": 898.0303344726562,
      "learning_rate": 2.7482476457631147e-07,
      "loss": 23.4459,
      "step": 29550
    },
    {
      "epoch": 234.6061814556331,
      "grad_norm": 241.63389587402344,
      "learning_rate": 2.7470061273082907e-07,
      "loss": 21.2976,
      "step": 29560
    },
    {
      "epoch": 234.68594217347956,
      "grad_norm": 517.2592163085938,
      "learning_rate": 2.745764547333513e-07,
      "loss": 26.3265,
      "step": 29570
    },
    {
      "epoch": 234.76570289132601,
      "grad_norm": 501.83367919921875,
      "learning_rate": 2.744522906148013e-07,
      "loss": 23.5089,
      "step": 29580
    },
    {
      "epoch": 234.84546360917247,
      "grad_norm": 1145.0814208984375,
      "learning_rate": 2.7432812040610376e-07,
      "loss": 24.3907,
      "step": 29590
    },
    {
      "epoch": 234.92522432701895,
      "grad_norm": 498.02813720703125,
      "learning_rate": 2.7420394413818476e-07,
      "loss": 22.5578,
      "step": 29600
    },
    {
      "epoch": 234.92522432701895,
      "eval_loss": 2.9738264083862305,
      "eval_mae": 1.2939438819885254,
      "eval_mse": 2.9738266468048096,
      "eval_r2": 0.10400056838989258,
      "eval_rmse": 1.724478659422844,
      "eval_runtime": 9.1822,
      "eval_samples_per_second": 436.822,
      "eval_steps_per_second": 13.722,
      "step": 29600
    },
    {
      "epoch": 235.0,
      "grad_norm": 319.5545349121094,
      "learning_rate": 2.74079761841972e-07,
      "loss": 21.1581,
      "step": 29610
    },
    {
      "epoch": 235.07976071784645,
      "grad_norm": 495.12176513671875,
      "learning_rate": 2.739555735483947e-07,
      "loss": 24.0536,
      "step": 29620
    },
    {
      "epoch": 235.1595214356929,
      "grad_norm": 437.9986572265625,
      "learning_rate": 2.738313792883833e-07,
      "loss": 24.0352,
      "step": 29630
    },
    {
      "epoch": 235.2392821535394,
      "grad_norm": 949.5150146484375,
      "learning_rate": 2.7370717909286993e-07,
      "loss": 23.0663,
      "step": 29640
    },
    {
      "epoch": 235.31904287138585,
      "grad_norm": 535.3571166992188,
      "learning_rate": 2.7358297299278844e-07,
      "loss": 24.5055,
      "step": 29650
    },
    {
      "epoch": 235.3988035892323,
      "grad_norm": 832.263427734375,
      "learning_rate": 2.7345876101907377e-07,
      "loss": 23.8656,
      "step": 29660
    },
    {
      "epoch": 235.47856430707876,
      "grad_norm": 398.2953186035156,
      "learning_rate": 2.7333454320266236e-07,
      "loss": 22.129,
      "step": 29670
    },
    {
      "epoch": 235.5583250249252,
      "grad_norm": 399.34210205078125,
      "learning_rate": 2.7321031957449236e-07,
      "loss": 23.5654,
      "step": 29680
    },
    {
      "epoch": 235.6380857427717,
      "grad_norm": 531.9966430664062,
      "learning_rate": 2.730860901655031e-07,
      "loss": 23.9399,
      "step": 29690
    },
    {
      "epoch": 235.71784646061815,
      "grad_norm": 234.51853942871094,
      "learning_rate": 2.729618550066356e-07,
      "loss": 23.2349,
      "step": 29700
    },
    {
      "epoch": 235.71784646061815,
      "eval_loss": 2.9727234840393066,
      "eval_mae": 1.2878901958465576,
      "eval_mse": 2.9727234840393066,
      "eval_r2": 0.10433298349380493,
      "eval_rmse": 1.7241587757626344,
      "eval_runtime": 9.1214,
      "eval_samples_per_second": 439.736,
      "eval_steps_per_second": 13.814,
      "step": 29700
    },
    {
      "epoch": 235.7976071784646,
      "grad_norm": 224.88511657714844,
      "learning_rate": 2.7283761412883194e-07,
      "loss": 21.5041,
      "step": 29710
    },
    {
      "epoch": 235.87736789631106,
      "grad_norm": 459.13385009765625,
      "learning_rate": 2.7271336756303603e-07,
      "loss": 23.8674,
      "step": 29720
    },
    {
      "epoch": 235.95712861415754,
      "grad_norm": 306.1412048339844,
      "learning_rate": 2.7258911534019315e-07,
      "loss": 23.5009,
      "step": 29730
    },
    {
      "epoch": 236.0319042871386,
      "grad_norm": 208.79376220703125,
      "learning_rate": 2.7246485749124955e-07,
      "loss": 21.118,
      "step": 29740
    },
    {
      "epoch": 236.11166500498504,
      "grad_norm": 340.0598449707031,
      "learning_rate": 2.723405940471534e-07,
      "loss": 23.2676,
      "step": 29750
    },
    {
      "epoch": 236.1914257228315,
      "grad_norm": 1169.4322509765625,
      "learning_rate": 2.72216325038854e-07,
      "loss": 23.9456,
      "step": 29760
    },
    {
      "epoch": 236.27118644067798,
      "grad_norm": 575.0592651367188,
      "learning_rate": 2.720920504973022e-07,
      "loss": 21.9234,
      "step": 29770
    },
    {
      "epoch": 236.35094715852443,
      "grad_norm": 1148.34765625,
      "learning_rate": 2.719677704534499e-07,
      "loss": 24.3241,
      "step": 29780
    },
    {
      "epoch": 236.4307078763709,
      "grad_norm": 1039.1551513671875,
      "learning_rate": 2.718434849382508e-07,
      "loss": 23.3575,
      "step": 29790
    },
    {
      "epoch": 236.51046859421734,
      "grad_norm": 545.8091430664062,
      "learning_rate": 2.717191939826597e-07,
      "loss": 22.3778,
      "step": 29800
    },
    {
      "epoch": 236.51046859421734,
      "eval_loss": 2.9748647212982178,
      "eval_mae": 1.288765549659729,
      "eval_mse": 2.9748647212982178,
      "eval_r2": 0.10368776321411133,
      "eval_rmse": 1.7247796152837085,
      "eval_runtime": 9.1834,
      "eval_samples_per_second": 436.766,
      "eval_steps_per_second": 13.72,
      "step": 29800
    },
    {
      "epoch": 236.5902293120638,
      "grad_norm": 186.42568969726562,
      "learning_rate": 2.715948976176327e-07,
      "loss": 24.9648,
      "step": 29810
    },
    {
      "epoch": 236.66999002991028,
      "grad_norm": 611.363037109375,
      "learning_rate": 2.714705958741276e-07,
      "loss": 21.861,
      "step": 29820
    },
    {
      "epoch": 236.74975074775674,
      "grad_norm": 815.4072875976562,
      "learning_rate": 2.7134628878310296e-07,
      "loss": 24.9254,
      "step": 29830
    },
    {
      "epoch": 236.8295114656032,
      "grad_norm": 520.3768920898438,
      "learning_rate": 2.712219763755193e-07,
      "loss": 22.666,
      "step": 29840
    },
    {
      "epoch": 236.90927218344964,
      "grad_norm": 488.80523681640625,
      "learning_rate": 2.710976586823381e-07,
      "loss": 25.7122,
      "step": 29850
    },
    {
      "epoch": 236.9890329012961,
      "grad_norm": 253.51536560058594,
      "learning_rate": 2.709733357345221e-07,
      "loss": 22.0631,
      "step": 29860
    },
    {
      "epoch": 237.06380857427718,
      "grad_norm": 867.0193481445312,
      "learning_rate": 2.7084900756303563e-07,
      "loss": 23.6757,
      "step": 29870
    },
    {
      "epoch": 237.14356929212363,
      "grad_norm": 638.9913940429688,
      "learning_rate": 2.707246741988441e-07,
      "loss": 24.1338,
      "step": 29880
    },
    {
      "epoch": 237.22333000997008,
      "grad_norm": 470.91595458984375,
      "learning_rate": 2.7060033567291437e-07,
      "loss": 23.7709,
      "step": 29890
    },
    {
      "epoch": 237.30309072781654,
      "grad_norm": 347.67950439453125,
      "learning_rate": 2.7047599201621437e-07,
      "loss": 23.0163,
      "step": 29900
    },
    {
      "epoch": 237.30309072781654,
      "eval_loss": 2.9714319705963135,
      "eval_mae": 1.2921468019485474,
      "eval_mse": 2.9714322090148926,
      "eval_r2": 0.10472196340560913,
      "eval_rmse": 1.7237842698594545,
      "eval_runtime": 9.1011,
      "eval_samples_per_second": 440.715,
      "eval_steps_per_second": 13.844,
      "step": 29900
    },
    {
      "epoch": 237.38285144566302,
      "grad_norm": 203.6299591064453,
      "learning_rate": 2.703516432597135e-07,
      "loss": 23.4956,
      "step": 29910
    },
    {
      "epoch": 237.46261216350948,
      "grad_norm": 249.11911010742188,
      "learning_rate": 2.7022728943438234e-07,
      "loss": 24.6135,
      "step": 29920
    },
    {
      "epoch": 237.54237288135593,
      "grad_norm": 563.9261474609375,
      "learning_rate": 2.7010293057119286e-07,
      "loss": 23.2776,
      "step": 29930
    },
    {
      "epoch": 237.62213359920239,
      "grad_norm": 328.15679931640625,
      "learning_rate": 2.69978566701118e-07,
      "loss": 24.4866,
      "step": 29940
    },
    {
      "epoch": 237.70189431704884,
      "grad_norm": 276.6571350097656,
      "learning_rate": 2.6985419785513225e-07,
      "loss": 20.3041,
      "step": 29950
    },
    {
      "epoch": 237.78165503489532,
      "grad_norm": 545.2789306640625,
      "learning_rate": 2.697298240642113e-07,
      "loss": 22.3424,
      "step": 29960
    },
    {
      "epoch": 237.86141575274178,
      "grad_norm": 233.5922393798828,
      "learning_rate": 2.6960544535933185e-07,
      "loss": 24.4458,
      "step": 29970
    },
    {
      "epoch": 237.94117647058823,
      "grad_norm": 205.71714782714844,
      "learning_rate": 2.69481061771472e-07,
      "loss": 23.6896,
      "step": 29980
    },
    {
      "epoch": 238.01595214356928,
      "grad_norm": 831.3458251953125,
      "learning_rate": 2.6935667333161104e-07,
      "loss": 20.1855,
      "step": 29990
    },
    {
      "epoch": 238.09571286141576,
      "grad_norm": 711.1663208007812,
      "learning_rate": 2.692322800707295e-07,
      "loss": 22.7354,
      "step": 30000
    },
    {
      "epoch": 238.09571286141576,
      "eval_loss": 2.971733331680298,
      "eval_mae": 1.2866849899291992,
      "eval_mse": 2.971733331680298,
      "eval_r2": 0.10463130474090576,
      "eval_rmse": 1.723871611135904,
      "eval_runtime": 9.399,
      "eval_samples_per_second": 426.745,
      "eval_steps_per_second": 13.406,
      "step": 30000
    },
    {
      "epoch": 238.17547357926222,
      "grad_norm": 679.4271850585938,
      "learning_rate": 2.6910788201980903e-07,
      "loss": 25.9507,
      "step": 30010
    },
    {
      "epoch": 238.25523429710867,
      "grad_norm": 191.01588439941406,
      "learning_rate": 2.689834792098325e-07,
      "loss": 23.5691,
      "step": 30020
    },
    {
      "epoch": 238.33499501495513,
      "grad_norm": 490.3308410644531,
      "learning_rate": 2.6885907167178406e-07,
      "loss": 21.6508,
      "step": 30030
    },
    {
      "epoch": 238.4147557328016,
      "grad_norm": 448.085205078125,
      "learning_rate": 2.687346594366488e-07,
      "loss": 24.4007,
      "step": 30040
    },
    {
      "epoch": 238.49451645064806,
      "grad_norm": 326.50433349609375,
      "learning_rate": 2.686102425354134e-07,
      "loss": 23.1411,
      "step": 30050
    },
    {
      "epoch": 238.57427716849452,
      "grad_norm": 195.3041229248047,
      "learning_rate": 2.684858209990651e-07,
      "loss": 24.5588,
      "step": 30060
    },
    {
      "epoch": 238.65403788634097,
      "grad_norm": 298.8865966796875,
      "learning_rate": 2.6836139485859286e-07,
      "loss": 21.3608,
      "step": 30070
    },
    {
      "epoch": 238.73379860418743,
      "grad_norm": 387.4641418457031,
      "learning_rate": 2.6823696414498655e-07,
      "loss": 22.479,
      "step": 30080
    },
    {
      "epoch": 238.8135593220339,
      "grad_norm": 341.1193542480469,
      "learning_rate": 2.6811252888923715e-07,
      "loss": 24.1528,
      "step": 30090
    },
    {
      "epoch": 238.89332003988036,
      "grad_norm": 352.9144592285156,
      "learning_rate": 2.6798808912233677e-07,
      "loss": 22.1946,
      "step": 30100
    },
    {
      "epoch": 238.89332003988036,
      "eval_loss": 2.97428035736084,
      "eval_mae": 1.2877657413482666,
      "eval_mse": 2.97428035736084,
      "eval_r2": 0.10386383533477783,
      "eval_rmse": 1.7246102044696476,
      "eval_runtime": 9.1168,
      "eval_samples_per_second": 439.956,
      "eval_steps_per_second": 13.821,
      "step": 30100
    },
    {
      "epoch": 238.97308075772682,
      "grad_norm": 489.48028564453125,
      "learning_rate": 2.6786364487527874e-07,
      "loss": 24.4485,
      "step": 30110
    },
    {
      "epoch": 239.04785643070787,
      "grad_norm": 708.220947265625,
      "learning_rate": 2.6773919617905745e-07,
      "loss": 22.5021,
      "step": 30120
    },
    {
      "epoch": 239.12761714855435,
      "grad_norm": 494.99481201171875,
      "learning_rate": 2.6761474306466837e-07,
      "loss": 22.909,
      "step": 30130
    },
    {
      "epoch": 239.2073778664008,
      "grad_norm": 355.8766174316406,
      "learning_rate": 2.674902855631081e-07,
      "loss": 23.0503,
      "step": 30140
    },
    {
      "epoch": 239.28713858424726,
      "grad_norm": 308.775146484375,
      "learning_rate": 2.673658237053743e-07,
      "loss": 25.3727,
      "step": 30150
    },
    {
      "epoch": 239.3668993020937,
      "grad_norm": 559.693115234375,
      "learning_rate": 2.6724135752246586e-07,
      "loss": 21.4661,
      "step": 30160
    },
    {
      "epoch": 239.44666001994017,
      "grad_norm": 606.2026977539062,
      "learning_rate": 2.6711688704538255e-07,
      "loss": 22.9412,
      "step": 30170
    },
    {
      "epoch": 239.52642073778665,
      "grad_norm": 560.1284790039062,
      "learning_rate": 2.669924123051252e-07,
      "loss": 22.2418,
      "step": 30180
    },
    {
      "epoch": 239.6061814556331,
      "grad_norm": 427.8968505859375,
      "learning_rate": 2.6686793333269604e-07,
      "loss": 24.7176,
      "step": 30190
    },
    {
      "epoch": 239.68594217347956,
      "grad_norm": 1554.723388671875,
      "learning_rate": 2.667434501590978e-07,
      "loss": 24.0749,
      "step": 30200
    },
    {
      "epoch": 239.68594217347956,
      "eval_loss": 2.9791054725646973,
      "eval_mae": 1.3021334409713745,
      "eval_mse": 2.9791057109832764,
      "eval_r2": 0.1024099588394165,
      "eval_rmse": 1.7260086068682499,
      "eval_runtime": 9.1144,
      "eval_samples_per_second": 440.072,
      "eval_steps_per_second": 13.824,
      "step": 30200
    },
    {
      "epoch": 239.76570289132601,
      "grad_norm": 1463.669921875,
      "learning_rate": 2.6661896281533494e-07,
      "loss": 24.3954,
      "step": 30210
    },
    {
      "epoch": 239.84546360917247,
      "grad_norm": 147.11212158203125,
      "learning_rate": 2.664944713324123e-07,
      "loss": 23.247,
      "step": 30220
    },
    {
      "epoch": 239.92522432701895,
      "grad_norm": 724.29248046875,
      "learning_rate": 2.6636997574133603e-07,
      "loss": 23.0793,
      "step": 30230
    },
    {
      "epoch": 240.0,
      "grad_norm": 305.6433410644531,
      "learning_rate": 2.662454760731134e-07,
      "loss": 22.3269,
      "step": 30240
    },
    {
      "epoch": 240.07976071784645,
      "grad_norm": 596.2266235351562,
      "learning_rate": 2.6612097235875263e-07,
      "loss": 23.0829,
      "step": 30250
    },
    {
      "epoch": 240.1595214356929,
      "grad_norm": 612.7228393554688,
      "learning_rate": 2.659964646292629e-07,
      "loss": 23.3136,
      "step": 30260
    },
    {
      "epoch": 240.2392821535394,
      "grad_norm": 508.3787536621094,
      "learning_rate": 2.6587195291565434e-07,
      "loss": 25.2146,
      "step": 30270
    },
    {
      "epoch": 240.31904287138585,
      "grad_norm": 258.1715393066406,
      "learning_rate": 2.6574743724893825e-07,
      "loss": 22.4899,
      "step": 30280
    },
    {
      "epoch": 240.3988035892323,
      "grad_norm": 287.9705810546875,
      "learning_rate": 2.656229176601267e-07,
      "loss": 20.1821,
      "step": 30290
    },
    {
      "epoch": 240.47856430707876,
      "grad_norm": 324.1119384765625,
      "learning_rate": 2.65498394180233e-07,
      "loss": 22.4198,
      "step": 30300
    },
    {
      "epoch": 240.47856430707876,
      "eval_loss": 2.9733307361602783,
      "eval_mae": 1.2874475717544556,
      "eval_mse": 2.9733307361602783,
      "eval_r2": 0.10414999723434448,
      "eval_rmse": 1.724334867756341,
      "eval_runtime": 9.1209,
      "eval_samples_per_second": 439.758,
      "eval_steps_per_second": 13.814,
      "step": 30300
    },
    {
      "epoch": 240.5583250249252,
      "grad_norm": 266.3356018066406,
      "learning_rate": 2.65373866840271e-07,
      "loss": 23.3075,
      "step": 30310
    },
    {
      "epoch": 240.6380857427717,
      "grad_norm": 1106.54248046875,
      "learning_rate": 2.652493356712561e-07,
      "loss": 23.8427,
      "step": 30320
    },
    {
      "epoch": 240.71784646061815,
      "grad_norm": 694.6317749023438,
      "learning_rate": 2.651248007042041e-07,
      "loss": 24.2339,
      "step": 30330
    },
    {
      "epoch": 240.7976071784646,
      "grad_norm": 273.56085205078125,
      "learning_rate": 2.650002619701321e-07,
      "loss": 24.2687,
      "step": 30340
    },
    {
      "epoch": 240.87736789631106,
      "grad_norm": 841.0004272460938,
      "learning_rate": 2.6487571950005803e-07,
      "loss": 24.3471,
      "step": 30350
    },
    {
      "epoch": 240.95712861415754,
      "grad_norm": 784.8743286132812,
      "learning_rate": 2.6475117332500067e-07,
      "loss": 24.1983,
      "step": 30360
    },
    {
      "epoch": 241.0319042871386,
      "grad_norm": 509.60528564453125,
      "learning_rate": 2.6462662347598e-07,
      "loss": 23.3144,
      "step": 30370
    },
    {
      "epoch": 241.11166500498504,
      "grad_norm": 341.0653381347656,
      "learning_rate": 2.6450206998401634e-07,
      "loss": 23.0637,
      "step": 30380
    },
    {
      "epoch": 241.1914257228315,
      "grad_norm": 537.9449462890625,
      "learning_rate": 2.643775128801316e-07,
      "loss": 23.4943,
      "step": 30390
    },
    {
      "epoch": 241.27118644067798,
      "grad_norm": 310.14495849609375,
      "learning_rate": 2.642529521953482e-07,
      "loss": 21.9223,
      "step": 30400
    },
    {
      "epoch": 241.27118644067798,
      "eval_loss": 2.9676897525787354,
      "eval_mae": 1.2947598695755005,
      "eval_mse": 2.9676897525787354,
      "eval_r2": 0.10584962368011475,
      "eval_rmse": 1.722698392806685,
      "eval_runtime": 9.1257,
      "eval_samples_per_second": 439.529,
      "eval_steps_per_second": 13.807,
      "step": 30400
    },
    {
      "epoch": 241.35094715852443,
      "grad_norm": 383.8916320800781,
      "learning_rate": 2.641283879606895e-07,
      "loss": 22.6062,
      "step": 30410
    },
    {
      "epoch": 241.4307078763709,
      "grad_norm": 167.80691528320312,
      "learning_rate": 2.6400382020717983e-07,
      "loss": 23.409,
      "step": 30420
    },
    {
      "epoch": 241.51046859421734,
      "grad_norm": 308.71026611328125,
      "learning_rate": 2.638792489658442e-07,
      "loss": 26.3346,
      "step": 30430
    },
    {
      "epoch": 241.5902293120638,
      "grad_norm": 165.4586944580078,
      "learning_rate": 2.637546742677088e-07,
      "loss": 23.653,
      "step": 30440
    },
    {
      "epoch": 241.66999002991028,
      "grad_norm": 523.208251953125,
      "learning_rate": 2.6363009614380043e-07,
      "loss": 21.2596,
      "step": 30450
    },
    {
      "epoch": 241.74975074775674,
      "grad_norm": 666.1103515625,
      "learning_rate": 2.635055146251469e-07,
      "loss": 23.8289,
      "step": 30460
    },
    {
      "epoch": 241.8295114656032,
      "grad_norm": 384.0069580078125,
      "learning_rate": 2.6338092974277653e-07,
      "loss": 22.3794,
      "step": 30470
    },
    {
      "epoch": 241.90927218344964,
      "grad_norm": 261.10748291015625,
      "learning_rate": 2.6325634152771905e-07,
      "loss": 24.0764,
      "step": 30480
    },
    {
      "epoch": 241.9890329012961,
      "grad_norm": 452.0245361328125,
      "learning_rate": 2.631317500110046e-07,
      "loss": 22.1141,
      "step": 30490
    },
    {
      "epoch": 242.06380857427718,
      "grad_norm": 265.33489990234375,
      "learning_rate": 2.630071552236641e-07,
      "loss": 21.199,
      "step": 30500
    },
    {
      "epoch": 242.06380857427718,
      "eval_loss": 2.969336986541748,
      "eval_mae": 1.2936196327209473,
      "eval_mse": 2.969337224960327,
      "eval_r2": 0.10535317659378052,
      "eval_rmse": 1.7231764926902662,
      "eval_runtime": 9.1267,
      "eval_samples_per_second": 439.479,
      "eval_steps_per_second": 13.806,
      "step": 30500
    },
    {
      "epoch": 242.14356929212363,
      "grad_norm": 301.48162841796875,
      "learning_rate": 2.628825571967297e-07,
      "loss": 22.0537,
      "step": 30510
    },
    {
      "epoch": 242.22333000997008,
      "grad_norm": 328.41217041015625,
      "learning_rate": 2.627579559612339e-07,
      "loss": 22.4829,
      "step": 30520
    },
    {
      "epoch": 242.30309072781654,
      "grad_norm": 632.689453125,
      "learning_rate": 2.626333515482103e-07,
      "loss": 24.6365,
      "step": 30530
    },
    {
      "epoch": 242.38285144566302,
      "grad_norm": 210.31422424316406,
      "learning_rate": 2.6250874398869305e-07,
      "loss": 24.618,
      "step": 30540
    },
    {
      "epoch": 242.46261216350948,
      "grad_norm": 539.3073120117188,
      "learning_rate": 2.6238413331371735e-07,
      "loss": 23.2583,
      "step": 30550
    },
    {
      "epoch": 242.54237288135593,
      "grad_norm": 324.3432312011719,
      "learning_rate": 2.62259519554319e-07,
      "loss": 22.2958,
      "step": 30560
    },
    {
      "epoch": 242.62213359920239,
      "grad_norm": 381.7465515136719,
      "learning_rate": 2.6213490274153454e-07,
      "loss": 26.3655,
      "step": 30570
    },
    {
      "epoch": 242.70189431704884,
      "grad_norm": 668.5147705078125,
      "learning_rate": 2.620102829064014e-07,
      "loss": 21.8175,
      "step": 30580
    },
    {
      "epoch": 242.78165503489532,
      "grad_norm": 686.0501098632812,
      "learning_rate": 2.6188566007995764e-07,
      "loss": 23.2414,
      "step": 30590
    },
    {
      "epoch": 242.86141575274178,
      "grad_norm": 536.3604125976562,
      "learning_rate": 2.617610342932423e-07,
      "loss": 23.2917,
      "step": 30600
    },
    {
      "epoch": 242.86141575274178,
      "eval_loss": 2.97670841217041,
      "eval_mae": 1.2905606031417847,
      "eval_mse": 2.976708173751831,
      "eval_r2": 0.10313242673873901,
      "eval_rmse": 1.725313934839637,
      "eval_runtime": 9.1134,
      "eval_samples_per_second": 440.123,
      "eval_steps_per_second": 13.826,
      "step": 30600
    },
    {
      "epoch": 242.94117647058823,
      "grad_norm": 141.11659240722656,
      "learning_rate": 2.616364055772948e-07,
      "loss": 23.4465,
      "step": 30610
    },
    {
      "epoch": 243.01595214356928,
      "grad_norm": 891.1488647460938,
      "learning_rate": 2.6151177396315554e-07,
      "loss": 22.8134,
      "step": 30620
    },
    {
      "epoch": 243.09571286141576,
      "grad_norm": 703.3218383789062,
      "learning_rate": 2.6138713948186565e-07,
      "loss": 22.6871,
      "step": 30630
    },
    {
      "epoch": 243.17547357926222,
      "grad_norm": 830.7728271484375,
      "learning_rate": 2.6126250216446677e-07,
      "loss": 22.9452,
      "step": 30640
    },
    {
      "epoch": 243.25523429710867,
      "grad_norm": 1052.3709716796875,
      "learning_rate": 2.6113786204200145e-07,
      "loss": 21.548,
      "step": 30650
    },
    {
      "epoch": 243.33499501495513,
      "grad_norm": 476.34375,
      "learning_rate": 2.6101321914551275e-07,
      "loss": 23.4007,
      "step": 30660
    },
    {
      "epoch": 243.4147557328016,
      "grad_norm": 251.74864196777344,
      "learning_rate": 2.6088857350604476e-07,
      "loss": 22.6418,
      "step": 30670
    },
    {
      "epoch": 243.49451645064806,
      "grad_norm": 1233.851806640625,
      "learning_rate": 2.607639251546418e-07,
      "loss": 22.7851,
      "step": 30680
    },
    {
      "epoch": 243.57427716849452,
      "grad_norm": 267.61334228515625,
      "learning_rate": 2.6063927412234925e-07,
      "loss": 24.4046,
      "step": 30690
    },
    {
      "epoch": 243.65403788634097,
      "grad_norm": 589.5751953125,
      "learning_rate": 2.60514620440213e-07,
      "loss": 23.3741,
      "step": 30700
    },
    {
      "epoch": 243.65403788634097,
      "eval_loss": 2.978055953979492,
      "eval_mae": 1.2858299016952515,
      "eval_mse": 2.9780564308166504,
      "eval_r2": 0.10272616147994995,
      "eval_rmse": 1.7257046186461489,
      "eval_runtime": 9.1004,
      "eval_samples_per_second": 440.75,
      "eval_steps_per_second": 13.846,
      "step": 30700
    },
    {
      "epoch": 243.73379860418743,
      "grad_norm": 154.97348022460938,
      "learning_rate": 2.603899641392794e-07,
      "loss": 23.5732,
      "step": 30710
    },
    {
      "epoch": 243.8135593220339,
      "grad_norm": 659.343994140625,
      "learning_rate": 2.6026530525059597e-07,
      "loss": 24.0906,
      "step": 30720
    },
    {
      "epoch": 243.89332003988036,
      "grad_norm": 758.31982421875,
      "learning_rate": 2.6014064380521033e-07,
      "loss": 22.8632,
      "step": 30730
    },
    {
      "epoch": 243.97308075772682,
      "grad_norm": 788.9865112304688,
      "learning_rate": 2.6002844634404446e-07,
      "loss": 25.1245,
      "step": 30740
    },
    {
      "epoch": 244.04785643070787,
      "grad_norm": 1136.2060546875,
      "learning_rate": 2.599037801264639e-07,
      "loss": 24.0369,
      "step": 30750
    },
    {
      "epoch": 244.12761714855435,
      "grad_norm": 432.9892578125,
      "learning_rate": 2.597791114422235e-07,
      "loss": 21.5665,
      "step": 30760
    },
    {
      "epoch": 244.2073778664008,
      "grad_norm": 403.8238220214844,
      "learning_rate": 2.596544403223735e-07,
      "loss": 24.2965,
      "step": 30770
    },
    {
      "epoch": 244.28713858424726,
      "grad_norm": 342.4436340332031,
      "learning_rate": 2.5952976679796496e-07,
      "loss": 24.103,
      "step": 30780
    },
    {
      "epoch": 244.3668993020937,
      "grad_norm": 1157.9874267578125,
      "learning_rate": 2.5940509090004937e-07,
      "loss": 22.7689,
      "step": 30790
    },
    {
      "epoch": 244.44666001994017,
      "grad_norm": 424.2969970703125,
      "learning_rate": 2.592804126596786e-07,
      "loss": 22.9439,
      "step": 30800
    },
    {
      "epoch": 244.44666001994017,
      "eval_loss": 2.968757390975952,
      "eval_mae": 1.2900669574737549,
      "eval_mse": 2.968757390975952,
      "eval_r2": 0.10552793741226196,
      "eval_rmse": 1.7230082388009502,
      "eval_runtime": 9.1274,
      "eval_samples_per_second": 439.444,
      "eval_steps_per_second": 13.805,
      "step": 30800
    },
    {
      "epoch": 244.52642073778665,
      "grad_norm": 566.1431274414062,
      "learning_rate": 2.591557321079054e-07,
      "loss": 24.0186,
      "step": 30810
    },
    {
      "epoch": 244.6061814556331,
      "grad_norm": 745.6699829101562,
      "learning_rate": 2.590310492757832e-07,
      "loss": 22.743,
      "step": 30820
    },
    {
      "epoch": 244.68594217347956,
      "grad_norm": 855.7810668945312,
      "learning_rate": 2.589063641943656e-07,
      "loss": 22.8092,
      "step": 30830
    },
    {
      "epoch": 244.76570289132601,
      "grad_norm": 261.3106689453125,
      "learning_rate": 2.587816768947071e-07,
      "loss": 23.1349,
      "step": 30840
    },
    {
      "epoch": 244.84546360917247,
      "grad_norm": 297.5519104003906,
      "learning_rate": 2.5865698740786255e-07,
      "loss": 24.3311,
      "step": 30850
    },
    {
      "epoch": 244.92522432701895,
      "grad_norm": 1044.4854736328125,
      "learning_rate": 2.5853229576488757e-07,
      "loss": 24.3775,
      "step": 30860
    },
    {
      "epoch": 245.0,
      "grad_norm": 479.5684814453125,
      "learning_rate": 2.584076019968379e-07,
      "loss": 21.1021,
      "step": 30870
    },
    {
      "epoch": 245.07976071784645,
      "grad_norm": 128.2196502685547,
      "learning_rate": 2.5828290613477026e-07,
      "loss": 23.223,
      "step": 30880
    },
    {
      "epoch": 245.1595214356929,
      "grad_norm": 261.00604248046875,
      "learning_rate": 2.581582082097417e-07,
      "loss": 24.8937,
      "step": 30890
    },
    {
      "epoch": 245.2392821535394,
      "grad_norm": 781.410888671875,
      "learning_rate": 2.580335082528097e-07,
      "loss": 23.9844,
      "step": 30900
    },
    {
      "epoch": 245.2392821535394,
      "eval_loss": 2.9844777584075928,
      "eval_mae": 1.305625557899475,
      "eval_mse": 2.9844777584075928,
      "eval_r2": 0.10079145431518555,
      "eval_rmse": 1.7275641112293323,
      "eval_runtime": 9.1463,
      "eval_samples_per_second": 438.536,
      "eval_steps_per_second": 13.776,
      "step": 30900
    },
    {
      "epoch": 245.31904287138585,
      "grad_norm": 877.4627685546875,
      "learning_rate": 2.5790880629503256e-07,
      "loss": 22.4681,
      "step": 30910
    },
    {
      "epoch": 245.3988035892323,
      "grad_norm": 424.7146911621094,
      "learning_rate": 2.577841023674685e-07,
      "loss": 22.4541,
      "step": 30920
    },
    {
      "epoch": 245.47856430707876,
      "grad_norm": 270.17474365234375,
      "learning_rate": 2.5765939650117687e-07,
      "loss": 22.6297,
      "step": 30930
    },
    {
      "epoch": 245.5583250249252,
      "grad_norm": 532.4530029296875,
      "learning_rate": 2.575346887272171e-07,
      "loss": 21.7078,
      "step": 30940
    },
    {
      "epoch": 245.6380857427717,
      "grad_norm": 668.9741821289062,
      "learning_rate": 2.5740997907664923e-07,
      "loss": 22.8294,
      "step": 30950
    },
    {
      "epoch": 245.71784646061815,
      "grad_norm": 687.4274291992188,
      "learning_rate": 2.572852675805338e-07,
      "loss": 24.2809,
      "step": 30960
    },
    {
      "epoch": 245.7976071784646,
      "grad_norm": 290.156982421875,
      "learning_rate": 2.571605542699317e-07,
      "loss": 22.1749,
      "step": 30970
    },
    {
      "epoch": 245.87736789631106,
      "grad_norm": 493.32818603515625,
      "learning_rate": 2.570358391759044e-07,
      "loss": 24.0103,
      "step": 30980
    },
    {
      "epoch": 245.95712861415754,
      "grad_norm": 1171.052978515625,
      "learning_rate": 2.5691112232951375e-07,
      "loss": 24.4049,
      "step": 30990
    },
    {
      "epoch": 246.0319042871386,
      "grad_norm": 510.27520751953125,
      "learning_rate": 2.56786403761822e-07,
      "loss": 22.9206,
      "step": 31000
    },
    {
      "epoch": 246.0319042871386,
      "eval_loss": 2.969852924346924,
      "eval_mae": 1.2927972078323364,
      "eval_mse": 2.969852924346924,
      "eval_r2": 0.10519784688949585,
      "eval_rmse": 1.7233261224582317,
      "eval_runtime": 9.1208,
      "eval_samples_per_second": 439.762,
      "eval_steps_per_second": 13.815,
      "step": 31000
    },
    {
      "epoch": 246.11166500498504,
      "grad_norm": 368.5177307128906,
      "learning_rate": 2.5666168350389173e-07,
      "loss": 22.376,
      "step": 31010
    },
    {
      "epoch": 246.1914257228315,
      "grad_norm": 645.730712890625,
      "learning_rate": 2.565369615867863e-07,
      "loss": 23.4462,
      "step": 31020
    },
    {
      "epoch": 246.27118644067798,
      "grad_norm": 1023.9620361328125,
      "learning_rate": 2.564122380415691e-07,
      "loss": 25.5178,
      "step": 31030
    },
    {
      "epoch": 246.35094715852443,
      "grad_norm": 1058.9161376953125,
      "learning_rate": 2.5628751289930415e-07,
      "loss": 26.1951,
      "step": 31040
    },
    {
      "epoch": 246.4307078763709,
      "grad_norm": 1442.2529296875,
      "learning_rate": 2.5616278619105583e-07,
      "loss": 24.4609,
      "step": 31050
    },
    {
      "epoch": 246.51046859421734,
      "grad_norm": 632.6666870117188,
      "learning_rate": 2.560380579478888e-07,
      "loss": 21.0663,
      "step": 31060
    },
    {
      "epoch": 246.5902293120638,
      "grad_norm": 250.21461486816406,
      "learning_rate": 2.559133282008684e-07,
      "loss": 23.5933,
      "step": 31070
    },
    {
      "epoch": 246.66999002991028,
      "grad_norm": 334.2854919433594,
      "learning_rate": 2.557885969810597e-07,
      "loss": 22.1378,
      "step": 31080
    },
    {
      "epoch": 246.74975074775674,
      "grad_norm": 257.9935607910156,
      "learning_rate": 2.5566386431952893e-07,
      "loss": 23.9687,
      "step": 31090
    },
    {
      "epoch": 246.8295114656032,
      "grad_norm": 1875.7540283203125,
      "learning_rate": 2.5553913024734214e-07,
      "loss": 22.8232,
      "step": 31100
    },
    {
      "epoch": 246.8295114656032,
      "eval_loss": 2.9731228351593018,
      "eval_mae": 1.2972358465194702,
      "eval_mse": 2.9731228351593018,
      "eval_r2": 0.10421264171600342,
      "eval_rmse": 1.7242745822981043,
      "eval_runtime": 9.094,
      "eval_samples_per_second": 441.062,
      "eval_steps_per_second": 13.855,
      "step": 31100
    },
    {
      "epoch": 246.90927218344964,
      "grad_norm": 588.41845703125,
      "learning_rate": 2.5541439479556605e-07,
      "loss": 21.2464,
      "step": 31110
    },
    {
      "epoch": 246.9890329012961,
      "grad_norm": 1064.3218994140625,
      "learning_rate": 2.5528965799526745e-07,
      "loss": 22.7664,
      "step": 31120
    },
    {
      "epoch": 247.06380857427718,
      "grad_norm": 610.049560546875,
      "learning_rate": 2.551649198775136e-07,
      "loss": 21.6621,
      "step": 31130
    },
    {
      "epoch": 247.14356929212363,
      "grad_norm": 938.4146118164062,
      "learning_rate": 2.5504018047337215e-07,
      "loss": 22.9931,
      "step": 31140
    },
    {
      "epoch": 247.22333000997008,
      "grad_norm": 1247.22314453125,
      "learning_rate": 2.5491543981391103e-07,
      "loss": 23.7451,
      "step": 31150
    },
    {
      "epoch": 247.30309072781654,
      "grad_norm": 869.5410766601562,
      "learning_rate": 2.547906979301983e-07,
      "loss": 20.6923,
      "step": 31160
    },
    {
      "epoch": 247.38285144566302,
      "grad_norm": 924.4537353515625,
      "learning_rate": 2.546659548533026e-07,
      "loss": 23.5296,
      "step": 31170
    },
    {
      "epoch": 247.46261216350948,
      "grad_norm": 435.33319091796875,
      "learning_rate": 2.5454121061429275e-07,
      "loss": 24.9323,
      "step": 31180
    },
    {
      "epoch": 247.54237288135593,
      "grad_norm": 399.9759521484375,
      "learning_rate": 2.5441646524423783e-07,
      "loss": 24.7295,
      "step": 31190
    },
    {
      "epoch": 247.62213359920239,
      "grad_norm": 1023.3956909179688,
      "learning_rate": 2.5429171877420714e-07,
      "loss": 24.0972,
      "step": 31200
    },
    {
      "epoch": 247.62213359920239,
      "eval_loss": 2.971273422241211,
      "eval_mae": 1.2890037298202515,
      "eval_mse": 2.971273422241211,
      "eval_r2": 0.10476988554000854,
      "eval_rmse": 1.7237382116322684,
      "eval_runtime": 9.0778,
      "eval_samples_per_second": 441.849,
      "eval_steps_per_second": 13.88,
      "step": 31200
    },
    {
      "epoch": 247.70189431704884,
      "grad_norm": 410.3566589355469,
      "learning_rate": 2.541669712352705e-07,
      "loss": 23.6653,
      "step": 31210
    },
    {
      "epoch": 247.78165503489532,
      "grad_norm": 448.5334167480469,
      "learning_rate": 2.5404222265849774e-07,
      "loss": 23.3972,
      "step": 31220
    },
    {
      "epoch": 247.86141575274178,
      "grad_norm": 693.1231689453125,
      "learning_rate": 2.539174730749592e-07,
      "loss": 21.8258,
      "step": 31230
    },
    {
      "epoch": 247.94117647058823,
      "grad_norm": 941.747314453125,
      "learning_rate": 2.5379272251572504e-07,
      "loss": 24.0123,
      "step": 31240
    },
    {
      "epoch": 248.01595214356928,
      "grad_norm": 394.1688232421875,
      "learning_rate": 2.536679710118661e-07,
      "loss": 20.3078,
      "step": 31250
    },
    {
      "epoch": 248.09571286141576,
      "grad_norm": 871.2073974609375,
      "learning_rate": 2.5354321859445334e-07,
      "loss": 22.2407,
      "step": 31260
    },
    {
      "epoch": 248.17547357926222,
      "grad_norm": 1198.1002197265625,
      "learning_rate": 2.5341846529455787e-07,
      "loss": 22.9835,
      "step": 31270
    },
    {
      "epoch": 248.25523429710867,
      "grad_norm": 676.827392578125,
      "learning_rate": 2.53293711143251e-07,
      "loss": 24.2606,
      "step": 31280
    },
    {
      "epoch": 248.33499501495513,
      "grad_norm": 238.11154174804688,
      "learning_rate": 2.5316895617160433e-07,
      "loss": 22.299,
      "step": 31290
    },
    {
      "epoch": 248.4147557328016,
      "grad_norm": 201.0675048828125,
      "learning_rate": 2.530442004106897e-07,
      "loss": 21.9802,
      "step": 31300
    },
    {
      "epoch": 248.4147557328016,
      "eval_loss": 2.9748613834381104,
      "eval_mae": 1.2876076698303223,
      "eval_mse": 2.9748613834381104,
      "eval_r2": 0.10368883609771729,
      "eval_rmse": 1.7247786476641316,
      "eval_runtime": 9.1422,
      "eval_samples_per_second": 438.735,
      "eval_steps_per_second": 13.782,
      "step": 31300
    },
    {
      "epoch": 248.49451645064806,
      "grad_norm": 424.4210510253906,
      "learning_rate": 2.52919443891579e-07,
      "loss": 24.8464,
      "step": 31310
    },
    {
      "epoch": 248.57427716849452,
      "grad_norm": 366.867431640625,
      "learning_rate": 2.5279468664534443e-07,
      "loss": 24.1916,
      "step": 31320
    },
    {
      "epoch": 248.65403788634097,
      "grad_norm": 620.6328125,
      "learning_rate": 2.5266992870305837e-07,
      "loss": 23.2312,
      "step": 31330
    },
    {
      "epoch": 248.73379860418743,
      "grad_norm": 266.5705261230469,
      "learning_rate": 2.525451700957933e-07,
      "loss": 24.0617,
      "step": 31340
    },
    {
      "epoch": 248.8135593220339,
      "grad_norm": 601.544189453125,
      "learning_rate": 2.524204108546219e-07,
      "loss": 23.0012,
      "step": 31350
    },
    {
      "epoch": 248.89332003988036,
      "grad_norm": 638.3615112304688,
      "learning_rate": 2.5229565101061703e-07,
      "loss": 23.0096,
      "step": 31360
    },
    {
      "epoch": 248.97308075772682,
      "grad_norm": 682.67236328125,
      "learning_rate": 2.521708905948516e-07,
      "loss": 24.3243,
      "step": 31370
    },
    {
      "epoch": 249.04785643070787,
      "grad_norm": 547.697265625,
      "learning_rate": 2.520461296383989e-07,
      "loss": 22.8326,
      "step": 31380
    },
    {
      "epoch": 249.12761714855435,
      "grad_norm": 498.31414794921875,
      "learning_rate": 2.519213681723322e-07,
      "loss": 23.5225,
      "step": 31390
    },
    {
      "epoch": 249.2073778664008,
      "grad_norm": 980.2252807617188,
      "learning_rate": 2.517966062277247e-07,
      "loss": 21.6833,
      "step": 31400
    },
    {
      "epoch": 249.2073778664008,
      "eval_loss": 2.968057632446289,
      "eval_mae": 1.2914273738861084,
      "eval_mse": 2.968057870864868,
      "eval_r2": 0.10573869943618774,
      "eval_rmse": 1.7228052330036812,
      "eval_runtime": 9.1047,
      "eval_samples_per_second": 440.544,
      "eval_steps_per_second": 13.839,
      "step": 31400
    },
    {
      "epoch": 249.28713858424726,
      "grad_norm": 715.7168579101562,
      "learning_rate": 2.5167184383565003e-07,
      "loss": 21.8021,
      "step": 31410
    },
    {
      "epoch": 249.3668993020937,
      "grad_norm": 540.0652465820312,
      "learning_rate": 2.515470810271819e-07,
      "loss": 24.0498,
      "step": 31420
    },
    {
      "epoch": 249.44666001994017,
      "grad_norm": 903.7225952148438,
      "learning_rate": 2.51422317833394e-07,
      "loss": 24.0246,
      "step": 31430
    },
    {
      "epoch": 249.52642073778665,
      "grad_norm": 641.188232421875,
      "learning_rate": 2.512975542853601e-07,
      "loss": 24.5698,
      "step": 31440
    },
    {
      "epoch": 249.6061814556331,
      "grad_norm": 814.0216064453125,
      "learning_rate": 2.5117279041415406e-07,
      "loss": 23.8711,
      "step": 31450
    },
    {
      "epoch": 249.68594217347956,
      "grad_norm": 268.5207824707031,
      "learning_rate": 2.510480262508501e-07,
      "loss": 22.7423,
      "step": 31460
    },
    {
      "epoch": 249.76570289132601,
      "grad_norm": 471.1623229980469,
      "learning_rate": 2.509232618265221e-07,
      "loss": 24.1789,
      "step": 31470
    },
    {
      "epoch": 249.84546360917247,
      "grad_norm": 562.5411376953125,
      "learning_rate": 2.507984971722441e-07,
      "loss": 24.0431,
      "step": 31480
    },
    {
      "epoch": 249.92522432701895,
      "grad_norm": 161.48260498046875,
      "learning_rate": 2.506737323190906e-07,
      "loss": 20.4402,
      "step": 31490
    },
    {
      "epoch": 250.0,
      "grad_norm": 441.6737976074219,
      "learning_rate": 2.5054896729813563e-07,
      "loss": 21.9377,
      "step": 31500
    },
    {
      "epoch": 250.0,
      "eval_loss": 2.9698264598846436,
      "eval_mae": 1.2904868125915527,
      "eval_mse": 2.9698264598846436,
      "eval_r2": 0.10520583391189575,
      "eval_rmse": 1.7233184441317406,
      "eval_runtime": 9.1197,
      "eval_samples_per_second": 439.819,
      "eval_steps_per_second": 13.816,
      "step": 31500
    },
    {
      "epoch": 250.07976071784645,
      "grad_norm": 1081.425048828125,
      "learning_rate": 2.504242021404535e-07,
      "loss": 22.5199,
      "step": 31510
    },
    {
      "epoch": 250.1595214356929,
      "grad_norm": 672.3206787109375,
      "learning_rate": 2.5029943687711853e-07,
      "loss": 22.2325,
      "step": 31520
    },
    {
      "epoch": 250.2392821535394,
      "grad_norm": 266.5413818359375,
      "learning_rate": 2.5017467153920505e-07,
      "loss": 23.8743,
      "step": 31530
    },
    {
      "epoch": 250.31904287138585,
      "grad_norm": 394.8414611816406,
      "learning_rate": 2.5004990615778754e-07,
      "loss": 26.0877,
      "step": 31540
    },
    {
      "epoch": 250.3988035892323,
      "grad_norm": 203.4445037841797,
      "learning_rate": 2.4992514076394025e-07,
      "loss": 23.0688,
      "step": 31550
    },
    {
      "epoch": 250.47856430707876,
      "grad_norm": 208.2809600830078,
      "learning_rate": 2.498003753887375e-07,
      "loss": 23.1918,
      "step": 31560
    },
    {
      "epoch": 250.5583250249252,
      "grad_norm": 502.4659118652344,
      "learning_rate": 2.496756100632538e-07,
      "loss": 24.0362,
      "step": 31570
    },
    {
      "epoch": 250.6380857427717,
      "grad_norm": 310.458740234375,
      "learning_rate": 2.495508448185635e-07,
      "loss": 23.5664,
      "step": 31580
    },
    {
      "epoch": 250.71784646061815,
      "grad_norm": 1150.802001953125,
      "learning_rate": 2.4942607968574093e-07,
      "loss": 21.5783,
      "step": 31590
    },
    {
      "epoch": 250.7976071784646,
      "grad_norm": 693.6259155273438,
      "learning_rate": 2.4930131469586027e-07,
      "loss": 21.2259,
      "step": 31600
    },
    {
      "epoch": 250.7976071784646,
      "eval_loss": 2.965651035308838,
      "eval_mae": 1.2888035774230957,
      "eval_mse": 2.965651035308838,
      "eval_r2": 0.10646384954452515,
      "eval_rmse": 1.722106569091715,
      "eval_runtime": 9.101,
      "eval_samples_per_second": 440.721,
      "eval_steps_per_second": 13.845,
      "step": 31600
    },
    {
      "epoch": 250.87736789631106,
      "grad_norm": 524.42138671875,
      "learning_rate": 2.4917654987999586e-07,
      "loss": 24.0307,
      "step": 31610
    },
    {
      "epoch": 250.95712861415754,
      "grad_norm": 742.5140380859375,
      "learning_rate": 2.49051785269222e-07,
      "loss": 23.6902,
      "step": 31620
    },
    {
      "epoch": 251.0319042871386,
      "grad_norm": 303.7222595214844,
      "learning_rate": 2.489270208946128e-07,
      "loss": 20.7224,
      "step": 31630
    },
    {
      "epoch": 251.11166500498504,
      "grad_norm": 1036.04345703125,
      "learning_rate": 2.488022567872425e-07,
      "loss": 23.9105,
      "step": 31640
    },
    {
      "epoch": 251.1914257228315,
      "grad_norm": 408.6697082519531,
      "learning_rate": 2.4867749297818497e-07,
      "loss": 24.5924,
      "step": 31650
    },
    {
      "epoch": 251.27118644067798,
      "grad_norm": 271.9720764160156,
      "learning_rate": 2.485527294985144e-07,
      "loss": 25.3175,
      "step": 31660
    },
    {
      "epoch": 251.35094715852443,
      "grad_norm": 310.2798767089844,
      "learning_rate": 2.4842796637930447e-07,
      "loss": 23.1924,
      "step": 31670
    },
    {
      "epoch": 251.4307078763709,
      "grad_norm": 293.2641906738281,
      "learning_rate": 2.4830320365162916e-07,
      "loss": 23.8814,
      "step": 31680
    },
    {
      "epoch": 251.51046859421734,
      "grad_norm": 315.9644470214844,
      "learning_rate": 2.481784413465621e-07,
      "loss": 23.1202,
      "step": 31690
    },
    {
      "epoch": 251.5902293120638,
      "grad_norm": 836.3831787109375,
      "learning_rate": 2.480536794951769e-07,
      "loss": 22.7945,
      "step": 31700
    },
    {
      "epoch": 251.5902293120638,
      "eval_loss": 2.972355365753174,
      "eval_mae": 1.3000593185424805,
      "eval_mse": 2.972355365753174,
      "eval_r2": 0.10444384813308716,
      "eval_rmse": 1.7240520194452293,
      "eval_runtime": 9.1011,
      "eval_samples_per_second": 440.717,
      "eval_steps_per_second": 13.845,
      "step": 31700
    },
    {
      "epoch": 251.66999002991028,
      "grad_norm": 268.80438232421875,
      "learning_rate": 2.4792891812854706e-07,
      "loss": 21.6484,
      "step": 31710
    },
    {
      "epoch": 251.74975074775674,
      "grad_norm": 378.1847229003906,
      "learning_rate": 2.4780415727774594e-07,
      "loss": 23.6207,
      "step": 31720
    },
    {
      "epoch": 251.8295114656032,
      "grad_norm": 573.9688720703125,
      "learning_rate": 2.476793969738469e-07,
      "loss": 22.2657,
      "step": 31730
    },
    {
      "epoch": 251.90927218344964,
      "grad_norm": 347.4887390136719,
      "learning_rate": 2.4755463724792284e-07,
      "loss": 24.1973,
      "step": 31740
    },
    {
      "epoch": 251.9890329012961,
      "grad_norm": 410.47625732421875,
      "learning_rate": 2.474298781310469e-07,
      "loss": 21.5562,
      "step": 31750
    },
    {
      "epoch": 252.06380857427718,
      "grad_norm": 423.4188537597656,
      "learning_rate": 2.4730511965429175e-07,
      "loss": 25.7484,
      "step": 31760
    },
    {
      "epoch": 252.14356929212363,
      "grad_norm": 644.38818359375,
      "learning_rate": 2.4718036184873014e-07,
      "loss": 25.0031,
      "step": 31770
    },
    {
      "epoch": 252.22333000997008,
      "grad_norm": 170.18014526367188,
      "learning_rate": 2.4705560474543466e-07,
      "loss": 21.3753,
      "step": 31780
    },
    {
      "epoch": 252.30309072781654,
      "grad_norm": 645.925537109375,
      "learning_rate": 2.4693084837547743e-07,
      "loss": 23.6118,
      "step": 31790
    },
    {
      "epoch": 252.38285144566302,
      "grad_norm": 530.1683349609375,
      "learning_rate": 2.468060927699306e-07,
      "loss": 24.8149,
      "step": 31800
    },
    {
      "epoch": 252.38285144566302,
      "eval_loss": 2.9768521785736084,
      "eval_mae": 1.2854654788970947,
      "eval_mse": 2.9768521785736084,
      "eval_r2": 0.10308897495269775,
      "eval_rmse": 1.7253556672679429,
      "eval_runtime": 9.102,
      "eval_samples_per_second": 440.674,
      "eval_steps_per_second": 13.843,
      "step": 31800
    },
    {
      "epoch": 252.46261216350948,
      "grad_norm": 1012.8935546875,
      "learning_rate": 2.466813379598661e-07,
      "loss": 22.2498,
      "step": 31810
    },
    {
      "epoch": 252.54237288135593,
      "grad_norm": 1013.740478515625,
      "learning_rate": 2.4655658397635587e-07,
      "loss": 23.5394,
      "step": 31820
    },
    {
      "epoch": 252.62213359920239,
      "grad_norm": 495.7828063964844,
      "learning_rate": 2.4643183085047125e-07,
      "loss": 23.3707,
      "step": 31830
    },
    {
      "epoch": 252.70189431704884,
      "grad_norm": 676.0218505859375,
      "learning_rate": 2.463070786132835e-07,
      "loss": 21.1693,
      "step": 31840
    },
    {
      "epoch": 252.78165503489532,
      "grad_norm": 664.9262084960938,
      "learning_rate": 2.4618232729586387e-07,
      "loss": 23.5246,
      "step": 31850
    },
    {
      "epoch": 252.86141575274178,
      "grad_norm": 548.2990112304688,
      "learning_rate": 2.460575769292832e-07,
      "loss": 21.6169,
      "step": 31860
    },
    {
      "epoch": 252.94117647058823,
      "grad_norm": 435.6769104003906,
      "learning_rate": 2.459328275446122e-07,
      "loss": 21.8176,
      "step": 31870
    },
    {
      "epoch": 253.01595214356928,
      "grad_norm": 430.7669372558594,
      "learning_rate": 2.4580807917292106e-07,
      "loss": 23.307,
      "step": 31880
    },
    {
      "epoch": 253.09571286141576,
      "grad_norm": 966.0969848632812,
      "learning_rate": 2.4568333184528e-07,
      "loss": 22.2923,
      "step": 31890
    },
    {
      "epoch": 253.17547357926222,
      "grad_norm": 467.6542663574219,
      "learning_rate": 2.4555858559275894e-07,
      "loss": 22.0797,
      "step": 31900
    },
    {
      "epoch": 253.17547357926222,
      "eval_loss": 2.9655842781066895,
      "eval_mae": 1.2908598184585571,
      "eval_mse": 2.9655845165252686,
      "eval_r2": 0.10648387670516968,
      "eval_rmse": 1.7220872557815614,
      "eval_runtime": 9.1069,
      "eval_samples_per_second": 440.434,
      "eval_steps_per_second": 13.836,
      "step": 31900
    },
    {
      "epoch": 253.25523429710867,
      "grad_norm": 1069.579345703125,
      "learning_rate": 2.4543384044642755e-07,
      "loss": 23.4295,
      "step": 31910
    },
    {
      "epoch": 253.33499501495513,
      "grad_norm": 543.779541015625,
      "learning_rate": 2.4530909643735495e-07,
      "loss": 24.7901,
      "step": 31920
    },
    {
      "epoch": 253.4147557328016,
      "grad_norm": 509.000732421875,
      "learning_rate": 2.4518435359661035e-07,
      "loss": 22.0587,
      "step": 31930
    },
    {
      "epoch": 253.49451645064806,
      "grad_norm": 662.7017211914062,
      "learning_rate": 2.450596119552625e-07,
      "loss": 21.6148,
      "step": 31940
    },
    {
      "epoch": 253.57427716849452,
      "grad_norm": 606.0281372070312,
      "learning_rate": 2.4493487154437976e-07,
      "loss": 22.612,
      "step": 31950
    },
    {
      "epoch": 253.65403788634097,
      "grad_norm": 1402.4447021484375,
      "learning_rate": 2.4481013239503047e-07,
      "loss": 22.1576,
      "step": 31960
    },
    {
      "epoch": 253.73379860418743,
      "grad_norm": 388.6296081542969,
      "learning_rate": 2.4468539453828226e-07,
      "loss": 25.0013,
      "step": 31970
    },
    {
      "epoch": 253.8135593220339,
      "grad_norm": 227.42666625976562,
      "learning_rate": 2.445606580052028e-07,
      "loss": 24.0291,
      "step": 31980
    },
    {
      "epoch": 253.89332003988036,
      "grad_norm": 543.4359741210938,
      "learning_rate": 2.444359228268592e-07,
      "loss": 24.6239,
      "step": 31990
    },
    {
      "epoch": 253.97308075772682,
      "grad_norm": 441.9510803222656,
      "learning_rate": 2.4431118903431825e-07,
      "loss": 22.7114,
      "step": 32000
    },
    {
      "epoch": 253.97308075772682,
      "eval_loss": 2.9745237827301025,
      "eval_mae": 1.2858246564865112,
      "eval_mse": 2.9745237827301025,
      "eval_r2": 0.10379058122634888,
      "eval_rmse": 1.7246807770512498,
      "eval_runtime": 9.1158,
      "eval_samples_per_second": 440.006,
      "eval_steps_per_second": 13.822,
      "step": 32000
    },
    {
      "epoch": 254.04785643070787,
      "grad_norm": 603.4620971679688,
      "learning_rate": 2.441864566586467e-07,
      "loss": 22.6715,
      "step": 32010
    },
    {
      "epoch": 254.12761714855435,
      "grad_norm": 500.43914794921875,
      "learning_rate": 2.440617257309105e-07,
      "loss": 23.7327,
      "step": 32020
    },
    {
      "epoch": 254.2073778664008,
      "grad_norm": 1103.8165283203125,
      "learning_rate": 2.439369962821754e-07,
      "loss": 21.4156,
      "step": 32030
    },
    {
      "epoch": 254.28713858424726,
      "grad_norm": 522.1469116210938,
      "learning_rate": 2.438122683435069e-07,
      "loss": 24.635,
      "step": 32040
    },
    {
      "epoch": 254.3668993020937,
      "grad_norm": 968.4270629882812,
      "learning_rate": 2.436875419459701e-07,
      "loss": 23.3848,
      "step": 32050
    },
    {
      "epoch": 254.44666001994017,
      "grad_norm": 597.3555908203125,
      "learning_rate": 2.4356281712062956e-07,
      "loss": 22.0033,
      "step": 32060
    },
    {
      "epoch": 254.52642073778665,
      "grad_norm": 883.1405639648438,
      "learning_rate": 2.434380938985497e-07,
      "loss": 24.8752,
      "step": 32070
    },
    {
      "epoch": 254.6061814556331,
      "grad_norm": 798.9927978515625,
      "learning_rate": 2.4331337231079415e-07,
      "loss": 22.7152,
      "step": 32080
    },
    {
      "epoch": 254.68594217347956,
      "grad_norm": 477.7481384277344,
      "learning_rate": 2.4318865238842663e-07,
      "loss": 23.3734,
      "step": 32090
    },
    {
      "epoch": 254.76570289132601,
      "grad_norm": 299.9521179199219,
      "learning_rate": 2.430639341625101e-07,
      "loss": 21.8486,
      "step": 32100
    },
    {
      "epoch": 254.76570289132601,
      "eval_loss": 2.9682564735412598,
      "eval_mae": 1.286859393119812,
      "eval_mse": 2.9682564735412598,
      "eval_r2": 0.10567885637283325,
      "eval_rmse": 1.722862871368833,
      "eval_runtime": 9.1177,
      "eval_samples_per_second": 439.914,
      "eval_steps_per_second": 13.819,
      "step": 32100
    },
    {
      "epoch": 254.84546360917247,
      "grad_norm": 597.6689453125,
      "learning_rate": 2.4293921766410707e-07,
      "loss": 23.7895,
      "step": 32110
    },
    {
      "epoch": 254.92522432701895,
      "grad_norm": 776.0618286132812,
      "learning_rate": 2.428145029242798e-07,
      "loss": 23.4706,
      "step": 32120
    },
    {
      "epoch": 255.0,
      "grad_norm": 317.8028259277344,
      "learning_rate": 2.4268978997409015e-07,
      "loss": 23.0663,
      "step": 32130
    },
    {
      "epoch": 255.07976071784645,
      "grad_norm": 1058.2989501953125,
      "learning_rate": 2.4256507884459944e-07,
      "loss": 21.5352,
      "step": 32140
    },
    {
      "epoch": 255.1595214356929,
      "grad_norm": 476.92132568359375,
      "learning_rate": 2.4244036956686844e-07,
      "loss": 23.7225,
      "step": 32150
    },
    {
      "epoch": 255.2392821535394,
      "grad_norm": 864.2227172851562,
      "learning_rate": 2.423156621719575e-07,
      "loss": 22.7905,
      "step": 32160
    },
    {
      "epoch": 255.31904287138585,
      "grad_norm": 186.06930541992188,
      "learning_rate": 2.421909566909266e-07,
      "loss": 24.4443,
      "step": 32170
    },
    {
      "epoch": 255.3988035892323,
      "grad_norm": 454.7611083984375,
      "learning_rate": 2.4206625315483525e-07,
      "loss": 23.8067,
      "step": 32180
    },
    {
      "epoch": 255.47856430707876,
      "grad_norm": 287.2181701660156,
      "learning_rate": 2.4194155159474246e-07,
      "loss": 22.7769,
      "step": 32190
    },
    {
      "epoch": 255.5583250249252,
      "grad_norm": 530.7894897460938,
      "learning_rate": 2.4181685204170647e-07,
      "loss": 23.8572,
      "step": 32200
    },
    {
      "epoch": 255.5583250249252,
      "eval_loss": 2.971975803375244,
      "eval_mae": 1.2996256351470947,
      "eval_mse": 2.971975564956665,
      "eval_r2": 0.10455828905105591,
      "eval_rmse": 1.7239418682068908,
      "eval_runtime": 9.1027,
      "eval_samples_per_second": 440.641,
      "eval_steps_per_second": 13.842,
      "step": 32200
    },
    {
      "epoch": 255.6380857427717,
      "grad_norm": 1025.2225341796875,
      "learning_rate": 2.416921545267855e-07,
      "loss": 24.3539,
      "step": 32210
    },
    {
      "epoch": 255.71784646061815,
      "grad_norm": 1187.16357421875,
      "learning_rate": 2.415674590810369e-07,
      "loss": 24.031,
      "step": 32220
    },
    {
      "epoch": 255.7976071784646,
      "grad_norm": 687.1815795898438,
      "learning_rate": 2.4144276573551767e-07,
      "loss": 22.4337,
      "step": 32230
    },
    {
      "epoch": 255.87736789631106,
      "grad_norm": 203.11148071289062,
      "learning_rate": 2.413180745212843e-07,
      "loss": 23.0997,
      "step": 32240
    },
    {
      "epoch": 255.95712861415754,
      "grad_norm": 590.783935546875,
      "learning_rate": 2.411933854693925e-07,
      "loss": 22.0966,
      "step": 32250
    },
    {
      "epoch": 256.03190428713856,
      "grad_norm": 681.7847900390625,
      "learning_rate": 2.4106869861089786e-07,
      "loss": 22.7958,
      "step": 32260
    },
    {
      "epoch": 256.11166500498507,
      "grad_norm": 388.5810241699219,
      "learning_rate": 2.4094401397685497e-07,
      "loss": 24.661,
      "step": 32270
    },
    {
      "epoch": 256.1914257228315,
      "grad_norm": 329.6586608886719,
      "learning_rate": 2.4081933159831836e-07,
      "loss": 23.5575,
      "step": 32280
    },
    {
      "epoch": 256.271186440678,
      "grad_norm": 1055.8572998046875,
      "learning_rate": 2.406946515063414e-07,
      "loss": 23.048,
      "step": 32290
    },
    {
      "epoch": 256.35094715852443,
      "grad_norm": 590.2950439453125,
      "learning_rate": 2.405699737319775e-07,
      "loss": 23.0763,
      "step": 32300
    },
    {
      "epoch": 256.35094715852443,
      "eval_loss": 2.9687814712524414,
      "eval_mae": 1.2883718013763428,
      "eval_mse": 2.9687812328338623,
      "eval_r2": 0.10552072525024414,
      "eval_rmse": 1.723015157459116,
      "eval_runtime": 9.088,
      "eval_samples_per_second": 441.349,
      "eval_steps_per_second": 13.864,
      "step": 32300
    },
    {
      "epoch": 256.4307078763709,
      "grad_norm": 905.7960815429688,
      "learning_rate": 2.4044529830627907e-07,
      "loss": 22.4415,
      "step": 32310
    },
    {
      "epoch": 256.51046859421734,
      "grad_norm": 131.0567169189453,
      "learning_rate": 2.403206252602981e-07,
      "loss": 25.3957,
      "step": 32320
    },
    {
      "epoch": 256.5902293120638,
      "grad_norm": 357.97283935546875,
      "learning_rate": 2.4019595462508605e-07,
      "loss": 24.214,
      "step": 32330
    },
    {
      "epoch": 256.66999002991025,
      "grad_norm": 562.968994140625,
      "learning_rate": 2.4007128643169364e-07,
      "loss": 21.4636,
      "step": 32340
    },
    {
      "epoch": 256.7497507477567,
      "grad_norm": 686.8287963867188,
      "learning_rate": 2.399466207111709e-07,
      "loss": 22.2345,
      "step": 32350
    },
    {
      "epoch": 256.8295114656032,
      "grad_norm": 1011.892333984375,
      "learning_rate": 2.3982195749456755e-07,
      "loss": 21.5997,
      "step": 32360
    },
    {
      "epoch": 256.9092721834497,
      "grad_norm": 774.2239990234375,
      "learning_rate": 2.3969729681293256e-07,
      "loss": 23.8638,
      "step": 32370
    },
    {
      "epoch": 256.9890329012961,
      "grad_norm": 479.4481506347656,
      "learning_rate": 2.395726386973141e-07,
      "loss": 22.1618,
      "step": 32380
    },
    {
      "epoch": 257.0638085742772,
      "grad_norm": 1316.7379150390625,
      "learning_rate": 2.394479831787598e-07,
      "loss": 22.9087,
      "step": 32390
    },
    {
      "epoch": 257.14356929212363,
      "grad_norm": 496.0229187011719,
      "learning_rate": 2.3932333028831674e-07,
      "loss": 23.5971,
      "step": 32400
    },
    {
      "epoch": 257.14356929212363,
      "eval_loss": 2.9773855209350586,
      "eval_mae": 1.3040708303451538,
      "eval_mse": 2.9773855209350586,
      "eval_r2": 0.10292834043502808,
      "eval_rmse": 1.7255102204667054,
      "eval_runtime": 9.0936,
      "eval_samples_per_second": 441.079,
      "eval_steps_per_second": 13.856,
      "step": 32400
    },
    {
      "epoch": 257.2233300099701,
      "grad_norm": 2317.558349609375,
      "learning_rate": 2.3919868005703136e-07,
      "loss": 22.1975,
      "step": 32410
    },
    {
      "epoch": 257.30309072781654,
      "grad_norm": 139.14430236816406,
      "learning_rate": 2.390740325159492e-07,
      "loss": 22.7621,
      "step": 32420
    },
    {
      "epoch": 257.382851445663,
      "grad_norm": 503.3020324707031,
      "learning_rate": 2.3894938769611535e-07,
      "loss": 24.6859,
      "step": 32430
    },
    {
      "epoch": 257.46261216350945,
      "grad_norm": 765.3264770507812,
      "learning_rate": 2.3882474562857407e-07,
      "loss": 23.1125,
      "step": 32440
    },
    {
      "epoch": 257.54237288135596,
      "grad_norm": 495.54620361328125,
      "learning_rate": 2.3870010634436913e-07,
      "loss": 23.5224,
      "step": 32450
    },
    {
      "epoch": 257.6221335992024,
      "grad_norm": 534.0633544921875,
      "learning_rate": 2.385754698745435e-07,
      "loss": 21.9895,
      "step": 32460
    },
    {
      "epoch": 257.70189431704887,
      "grad_norm": 245.5608367919922,
      "learning_rate": 2.384508362501395e-07,
      "loss": 24.474,
      "step": 32470
    },
    {
      "epoch": 257.7816550348953,
      "grad_norm": 184.3255615234375,
      "learning_rate": 2.383262055021985e-07,
      "loss": 22.3899,
      "step": 32480
    },
    {
      "epoch": 257.8614157527418,
      "grad_norm": 368.9948425292969,
      "learning_rate": 2.3820157766176138e-07,
      "loss": 23.7286,
      "step": 32490
    },
    {
      "epoch": 257.94117647058823,
      "grad_norm": 268.95953369140625,
      "learning_rate": 2.380769527598684e-07,
      "loss": 23.5707,
      "step": 32500
    },
    {
      "epoch": 257.94117647058823,
      "eval_loss": 2.9620044231414795,
      "eval_mae": 1.2903201580047607,
      "eval_mse": 2.9620041847229004,
      "eval_r2": 0.1075626015663147,
      "eval_rmse": 1.7210474092025765,
      "eval_runtime": 9.0723,
      "eval_samples_per_second": 442.117,
      "eval_steps_per_second": 13.888,
      "step": 32500
    },
    {
      "epoch": 258.0159521435693,
      "grad_norm": 369.2539367675781,
      "learning_rate": 2.3795233082755888e-07,
      "loss": 21.1234,
      "step": 32510
    },
    {
      "epoch": 258.09571286141573,
      "grad_norm": 300.13201904296875,
      "learning_rate": 2.3782771189587135e-07,
      "loss": 23.8872,
      "step": 32520
    },
    {
      "epoch": 258.1754735792622,
      "grad_norm": 201.40646362304688,
      "learning_rate": 2.377030959958438e-07,
      "loss": 21.8254,
      "step": 32530
    },
    {
      "epoch": 258.2552342971087,
      "grad_norm": 402.68914794921875,
      "learning_rate": 2.3757848315851342e-07,
      "loss": 22.2791,
      "step": 32540
    },
    {
      "epoch": 258.33499501495515,
      "grad_norm": 332.69671630859375,
      "learning_rate": 2.374538734149165e-07,
      "loss": 24.6208,
      "step": 32550
    },
    {
      "epoch": 258.4147557328016,
      "grad_norm": 356.6276550292969,
      "learning_rate": 2.3732926679608877e-07,
      "loss": 23.3802,
      "step": 32560
    },
    {
      "epoch": 258.49451645064806,
      "grad_norm": 396.5636291503906,
      "learning_rate": 2.3720466333306488e-07,
      "loss": 22.3744,
      "step": 32570
    },
    {
      "epoch": 258.5742771684945,
      "grad_norm": 252.76632690429688,
      "learning_rate": 2.3708006305687905e-07,
      "loss": 22.4334,
      "step": 32580
    },
    {
      "epoch": 258.654037886341,
      "grad_norm": 1583.0286865234375,
      "learning_rate": 2.369554659985644e-07,
      "loss": 22.3503,
      "step": 32590
    },
    {
      "epoch": 258.7337986041874,
      "grad_norm": 1470.625244140625,
      "learning_rate": 2.368308721891534e-07,
      "loss": 24.4704,
      "step": 32600
    },
    {
      "epoch": 258.7337986041874,
      "eval_loss": 2.960970640182495,
      "eval_mae": 1.2870057821273804,
      "eval_mse": 2.960970640182495,
      "eval_r2": 0.10787403583526611,
      "eval_rmse": 1.7207471168601431,
      "eval_runtime": 9.0696,
      "eval_samples_per_second": 442.246,
      "eval_steps_per_second": 13.893,
      "step": 32600
    },
    {
      "epoch": 258.8135593220339,
      "grad_norm": 312.400146484375,
      "learning_rate": 2.3670628165967785e-07,
      "loss": 24.4746,
      "step": 32610
    },
    {
      "epoch": 258.89332003988034,
      "grad_norm": 391.41143798828125,
      "learning_rate": 2.365816944411684e-07,
      "loss": 22.2742,
      "step": 32620
    },
    {
      "epoch": 258.97308075772685,
      "grad_norm": 1520.7578125,
      "learning_rate": 2.3645711056465505e-07,
      "loss": 23.984,
      "step": 32630
    },
    {
      "epoch": 259.0478564307079,
      "grad_norm": 184.56324768066406,
      "learning_rate": 2.36332530061167e-07,
      "loss": 21.9073,
      "step": 32640
    },
    {
      "epoch": 259.12761714855435,
      "grad_norm": 448.33758544921875,
      "learning_rate": 2.3620795296173267e-07,
      "loss": 23.8176,
      "step": 32650
    },
    {
      "epoch": 259.2073778664008,
      "grad_norm": 1436.3748779296875,
      "learning_rate": 2.3608337929737946e-07,
      "loss": 20.6676,
      "step": 32660
    },
    {
      "epoch": 259.28713858424726,
      "grad_norm": 457.44677734375,
      "learning_rate": 2.359588090991339e-07,
      "loss": 21.6573,
      "step": 32670
    },
    {
      "epoch": 259.3668993020937,
      "grad_norm": 1274.454345703125,
      "learning_rate": 2.3583424239802186e-07,
      "loss": 22.477,
      "step": 32680
    },
    {
      "epoch": 259.44666001994017,
      "grad_norm": 594.2935180664062,
      "learning_rate": 2.3570967922506823e-07,
      "loss": 23.9723,
      "step": 32690
    },
    {
      "epoch": 259.5264207377866,
      "grad_norm": 292.4962158203125,
      "learning_rate": 2.355851196112971e-07,
      "loss": 23.6009,
      "step": 32700
    },
    {
      "epoch": 259.5264207377866,
      "eval_loss": 2.97013521194458,
      "eval_mae": 1.2906213998794556,
      "eval_mse": 2.970135450363159,
      "eval_r2": 0.10511273145675659,
      "eval_rmse": 1.7234080916495544,
      "eval_runtime": 9.0839,
      "eval_samples_per_second": 441.551,
      "eval_steps_per_second": 13.871,
      "step": 32700
    },
    {
      "epoch": 259.6061814556331,
      "grad_norm": 685.1624755859375,
      "learning_rate": 2.3546056358773136e-07,
      "loss": 23.991,
      "step": 32710
    },
    {
      "epoch": 259.6859421734796,
      "grad_norm": 976.3839721679688,
      "learning_rate": 2.3533601118539338e-07,
      "loss": 22.3304,
      "step": 32720
    },
    {
      "epoch": 259.76570289132604,
      "grad_norm": 1046.9388427734375,
      "learning_rate": 2.352114624353045e-07,
      "loss": 25.016,
      "step": 32730
    },
    {
      "epoch": 259.8454636091725,
      "grad_norm": 201.44747924804688,
      "learning_rate": 2.350869173684851e-07,
      "loss": 23.8758,
      "step": 32740
    },
    {
      "epoch": 259.92522432701895,
      "grad_norm": 196.49974060058594,
      "learning_rate": 2.349623760159548e-07,
      "loss": 23.2117,
      "step": 32750
    },
    {
      "epoch": 260.0,
      "grad_norm": 646.876220703125,
      "learning_rate": 2.348378384087319e-07,
      "loss": 22.9075,
      "step": 32760
    },
    {
      "epoch": 260.07976071784645,
      "grad_norm": 1059.302978515625,
      "learning_rate": 2.3471330457783435e-07,
      "loss": 23.9309,
      "step": 32770
    },
    {
      "epoch": 260.1595214356929,
      "grad_norm": 585.2809448242188,
      "learning_rate": 2.3458877455427865e-07,
      "loss": 25.4878,
      "step": 32780
    },
    {
      "epoch": 260.23928215353936,
      "grad_norm": 819.2096557617188,
      "learning_rate": 2.3446424836908068e-07,
      "loss": 21.9851,
      "step": 32790
    },
    {
      "epoch": 260.3190428713858,
      "grad_norm": 1196.5474853515625,
      "learning_rate": 2.343397260532551e-07,
      "loss": 21.1927,
      "step": 32800
    },
    {
      "epoch": 260.3190428713858,
      "eval_loss": 2.9644973278045654,
      "eval_mae": 1.293110728263855,
      "eval_mse": 2.9644970893859863,
      "eval_r2": 0.1068115234375,
      "eval_rmse": 1.7217714974368656,
      "eval_runtime": 9.0683,
      "eval_samples_per_second": 442.309,
      "eval_steps_per_second": 13.895,
      "step": 32800
    },
    {
      "epoch": 260.39880358923233,
      "grad_norm": 377.51409912109375,
      "learning_rate": 2.3421520763781583e-07,
      "loss": 21.3956,
      "step": 32810
    },
    {
      "epoch": 260.4785643070788,
      "grad_norm": 215.3258819580078,
      "learning_rate": 2.340906931537758e-07,
      "loss": 24.547,
      "step": 32820
    },
    {
      "epoch": 260.55832502492524,
      "grad_norm": 267.4694519042969,
      "learning_rate": 2.3396618263214671e-07,
      "loss": 22.4404,
      "step": 32830
    },
    {
      "epoch": 260.6380857427717,
      "grad_norm": 803.0358276367188,
      "learning_rate": 2.3384167610393968e-07,
      "loss": 22.7296,
      "step": 32840
    },
    {
      "epoch": 260.71784646061815,
      "grad_norm": 265.7320556640625,
      "learning_rate": 2.3371717360016442e-07,
      "loss": 22.8819,
      "step": 32850
    },
    {
      "epoch": 260.7976071784646,
      "grad_norm": 446.6425476074219,
      "learning_rate": 2.3359267515182992e-07,
      "loss": 23.9664,
      "step": 32860
    },
    {
      "epoch": 260.87736789631106,
      "grad_norm": 1641.0150146484375,
      "learning_rate": 2.33468180789944e-07,
      "loss": 22.9442,
      "step": 32870
    },
    {
      "epoch": 260.9571286141575,
      "grad_norm": 136.3773193359375,
      "learning_rate": 2.3334369054551372e-07,
      "loss": 23.8371,
      "step": 32880
    },
    {
      "epoch": 261.03190428713856,
      "grad_norm": 961.9443359375,
      "learning_rate": 2.3321920444954468e-07,
      "loss": 22.9601,
      "step": 32890
    },
    {
      "epoch": 261.11166500498507,
      "grad_norm": 441.1259765625,
      "learning_rate": 2.3309472253304186e-07,
      "loss": 23.7666,
      "step": 32900
    },
    {
      "epoch": 261.11166500498507,
      "eval_loss": 2.961627721786499,
      "eval_mae": 1.2911022901535034,
      "eval_mse": 2.96162748336792,
      "eval_r2": 0.10767608880996704,
      "eval_rmse": 1.720937966159129,
      "eval_runtime": 9.0616,
      "eval_samples_per_second": 442.638,
      "eval_steps_per_second": 13.905,
      "step": 32900
    },
    {
      "epoch": 261.1914257228315,
      "grad_norm": 1438.0609130859375,
      "learning_rate": 2.329702448270089e-07,
      "loss": 23.3852,
      "step": 32910
    },
    {
      "epoch": 261.271186440678,
      "grad_norm": 283.7239990234375,
      "learning_rate": 2.3284577136244858e-07,
      "loss": 22.6681,
      "step": 32920
    },
    {
      "epoch": 261.35094715852443,
      "grad_norm": 1353.781005859375,
      "learning_rate": 2.3272130217036275e-07,
      "loss": 21.8145,
      "step": 32930
    },
    {
      "epoch": 261.4307078763709,
      "grad_norm": 678.4915771484375,
      "learning_rate": 2.325968372817518e-07,
      "loss": 22.9129,
      "step": 32940
    },
    {
      "epoch": 261.51046859421734,
      "grad_norm": 603.8546142578125,
      "learning_rate": 2.324723767276152e-07,
      "loss": 22.0271,
      "step": 32950
    },
    {
      "epoch": 261.5902293120638,
      "grad_norm": 411.7084045410156,
      "learning_rate": 2.3234792053895156e-07,
      "loss": 23.6058,
      "step": 32960
    },
    {
      "epoch": 261.66999002991025,
      "grad_norm": 695.4976806640625,
      "learning_rate": 2.3222346874675823e-07,
      "loss": 23.9588,
      "step": 32970
    },
    {
      "epoch": 261.7497507477567,
      "grad_norm": 508.5700988769531,
      "learning_rate": 2.320990213820315e-07,
      "loss": 23.453,
      "step": 32980
    },
    {
      "epoch": 261.8295114656032,
      "grad_norm": 157.2057647705078,
      "learning_rate": 2.3197457847576638e-07,
      "loss": 23.8839,
      "step": 32990
    },
    {
      "epoch": 261.9092721834497,
      "grad_norm": 1685.7197265625,
      "learning_rate": 2.31850140058957e-07,
      "loss": 23.4671,
      "step": 33000
    },
    {
      "epoch": 261.9092721834497,
      "eval_loss": 2.977787971496582,
      "eval_mae": 1.2833104133605957,
      "eval_mse": 2.977787971496582,
      "eval_r2": 0.10280704498291016,
      "eval_rmse": 1.725626834369639,
      "eval_runtime": 9.097,
      "eval_samples_per_second": 440.916,
      "eval_steps_per_second": 13.851,
      "step": 33000
    },
    {
      "epoch": 261.9890329012961,
      "grad_norm": 975.1857299804688,
      "learning_rate": 2.317257061625964e-07,
      "loss": 23.8065,
      "step": 33010
    },
    {
      "epoch": 262.0638085742772,
      "grad_norm": 369.8857421875,
      "learning_rate": 2.3160127681767638e-07,
      "loss": 19.9302,
      "step": 33020
    },
    {
      "epoch": 262.14356929212363,
      "grad_norm": 1166.5467529296875,
      "learning_rate": 2.3147685205518742e-07,
      "loss": 22.5537,
      "step": 33030
    },
    {
      "epoch": 262.2233300099701,
      "grad_norm": 767.1246337890625,
      "learning_rate": 2.313524319061192e-07,
      "loss": 22.2758,
      "step": 33040
    },
    {
      "epoch": 262.30309072781654,
      "grad_norm": 984.523681640625,
      "learning_rate": 2.3122801640146018e-07,
      "loss": 23.4407,
      "step": 33050
    },
    {
      "epoch": 262.382851445663,
      "grad_norm": 477.3348693847656,
      "learning_rate": 2.3110360557219744e-07,
      "loss": 23.6952,
      "step": 33060
    },
    {
      "epoch": 262.46261216350945,
      "grad_norm": 435.17425537109375,
      "learning_rate": 2.309791994493172e-07,
      "loss": 24.8507,
      "step": 33070
    },
    {
      "epoch": 262.54237288135596,
      "grad_norm": 1697.8922119140625,
      "learning_rate": 2.308547980638042e-07,
      "loss": 23.2312,
      "step": 33080
    },
    {
      "epoch": 262.6221335992024,
      "grad_norm": 538.4093627929688,
      "learning_rate": 2.3073040144664227e-07,
      "loss": 22.7747,
      "step": 33090
    },
    {
      "epoch": 262.70189431704887,
      "grad_norm": 502.0633850097656,
      "learning_rate": 2.3060600962881386e-07,
      "loss": 24.6969,
      "step": 33100
    },
    {
      "epoch": 262.70189431704887,
      "eval_loss": 2.9655330181121826,
      "eval_mae": 1.2875216007232666,
      "eval_mse": 2.9655327796936035,
      "eval_r2": 0.10649949312210083,
      "eval_rmse": 1.722072234168359,
      "eval_runtime": 9.105,
      "eval_samples_per_second": 440.529,
      "eval_steps_per_second": 13.839,
      "step": 33100
    },
    {
      "epoch": 262.7816550348953,
      "grad_norm": 758.8497314453125,
      "learning_rate": 2.3048162264130041e-07,
      "loss": 23.9142,
      "step": 33110
    },
    {
      "epoch": 262.8614157527418,
      "grad_norm": 669.5830688476562,
      "learning_rate": 2.3035724051508187e-07,
      "loss": 23.165,
      "step": 33120
    },
    {
      "epoch": 262.94117647058823,
      "grad_norm": 499.54779052734375,
      "learning_rate": 2.3023286328113728e-07,
      "loss": 23.156,
      "step": 33130
    },
    {
      "epoch": 263.0159521435693,
      "grad_norm": 369.984619140625,
      "learning_rate": 2.301084909704444e-07,
      "loss": 20.9903,
      "step": 33140
    },
    {
      "epoch": 263.09571286141573,
      "grad_norm": 862.9248657226562,
      "learning_rate": 2.299841236139795e-07,
      "loss": 21.8262,
      "step": 33150
    },
    {
      "epoch": 263.1754735792622,
      "grad_norm": 345.6047058105469,
      "learning_rate": 2.298597612427181e-07,
      "loss": 23.0221,
      "step": 33160
    },
    {
      "epoch": 263.2552342971087,
      "grad_norm": 238.32444763183594,
      "learning_rate": 2.2973540388763393e-07,
      "loss": 22.3937,
      "step": 33170
    },
    {
      "epoch": 263.33499501495515,
      "grad_norm": 738.8783569335938,
      "learning_rate": 2.296110515796999e-07,
      "loss": 25.3277,
      "step": 33180
    },
    {
      "epoch": 263.4147557328016,
      "grad_norm": 1106.19921875,
      "learning_rate": 2.2948670434988743e-07,
      "loss": 22.6522,
      "step": 33190
    },
    {
      "epoch": 263.49451645064806,
      "grad_norm": 758.0292358398438,
      "learning_rate": 2.2936236222916672e-07,
      "loss": 23.0677,
      "step": 33200
    },
    {
      "epoch": 263.49451645064806,
      "eval_loss": 2.9774444103240967,
      "eval_mae": 1.2838081121444702,
      "eval_mse": 2.9774444103240967,
      "eval_r2": 0.10291057825088501,
      "eval_rmse": 1.7255272847231644,
      "eval_runtime": 9.0703,
      "eval_samples_per_second": 442.215,
      "eval_steps_per_second": 13.892,
      "step": 33200
    },
    {
      "epoch": 263.5742771684945,
      "grad_norm": 295.3321228027344,
      "learning_rate": 2.292380252485069e-07,
      "loss": 21.6681,
      "step": 33210
    },
    {
      "epoch": 263.654037886341,
      "grad_norm": 638.5050048828125,
      "learning_rate": 2.2911369343887545e-07,
      "loss": 23.3144,
      "step": 33220
    },
    {
      "epoch": 263.7337986041874,
      "grad_norm": 403.549072265625,
      "learning_rate": 2.2898936683123882e-07,
      "loss": 23.0726,
      "step": 33230
    },
    {
      "epoch": 263.8135593220339,
      "grad_norm": 254.73532104492188,
      "learning_rate": 2.2886504545656204e-07,
      "loss": 24.1246,
      "step": 33240
    },
    {
      "epoch": 263.89332003988034,
      "grad_norm": 839.294189453125,
      "learning_rate": 2.287407293458091e-07,
      "loss": 25.1882,
      "step": 33250
    },
    {
      "epoch": 263.97308075772685,
      "grad_norm": 319.6940612792969,
      "learning_rate": 2.286164185299423e-07,
      "loss": 22.9318,
      "step": 33260
    },
    {
      "epoch": 264.0478564307079,
      "grad_norm": 268.8839111328125,
      "learning_rate": 2.2849211303992272e-07,
      "loss": 19.9687,
      "step": 33270
    },
    {
      "epoch": 264.12761714855435,
      "grad_norm": 591.1959228515625,
      "learning_rate": 2.2836781290671034e-07,
      "loss": 23.6758,
      "step": 33280
    },
    {
      "epoch": 264.2073778664008,
      "grad_norm": 873.9620361328125,
      "learning_rate": 2.2824351816126365e-07,
      "loss": 22.6141,
      "step": 33290
    },
    {
      "epoch": 264.28713858424726,
      "grad_norm": 595.4581298828125,
      "learning_rate": 2.2811922883453989e-07,
      "loss": 24.4165,
      "step": 33300
    },
    {
      "epoch": 264.28713858424726,
      "eval_loss": 2.9677834510803223,
      "eval_mae": 1.286590337753296,
      "eval_mse": 2.967783212661743,
      "eval_r2": 0.10582143068313599,
      "eval_rmse": 1.722725518665624,
      "eval_runtime": 9.0744,
      "eval_samples_per_second": 442.01,
      "eval_steps_per_second": 13.885,
      "step": 33300
    },
    {
      "epoch": 264.3668993020937,
      "grad_norm": 220.30337524414062,
      "learning_rate": 2.2799494495749457e-07,
      "loss": 24.6584,
      "step": 33310
    },
    {
      "epoch": 264.44666001994017,
      "grad_norm": 304.1072082519531,
      "learning_rate": 2.2787066656108238e-07,
      "loss": 23.2196,
      "step": 33320
    },
    {
      "epoch": 264.5264207377866,
      "grad_norm": 1391.931884765625,
      "learning_rate": 2.2774639367625636e-07,
      "loss": 23.9871,
      "step": 33330
    },
    {
      "epoch": 264.6061814556331,
      "grad_norm": 250.1853790283203,
      "learning_rate": 2.2762212633396813e-07,
      "loss": 23.3213,
      "step": 33340
    },
    {
      "epoch": 264.6859421734796,
      "grad_norm": 300.67144775390625,
      "learning_rate": 2.2749786456516823e-07,
      "loss": 21.7446,
      "step": 33350
    },
    {
      "epoch": 264.76570289132604,
      "grad_norm": 405.91485595703125,
      "learning_rate": 2.2737360840080534e-07,
      "loss": 21.5377,
      "step": 33360
    },
    {
      "epoch": 264.8454636091725,
      "grad_norm": 350.0721740722656,
      "learning_rate": 2.2724935787182719e-07,
      "loss": 24.2495,
      "step": 33370
    },
    {
      "epoch": 264.92522432701895,
      "grad_norm": 703.5527954101562,
      "learning_rate": 2.271251130091798e-07,
      "loss": 23.5282,
      "step": 33380
    },
    {
      "epoch": 265.0,
      "grad_norm": 352.90264892578125,
      "learning_rate": 2.2700087384380805e-07,
      "loss": 21.1718,
      "step": 33390
    },
    {
      "epoch": 265.07976071784645,
      "grad_norm": 167.44500732421875,
      "learning_rate": 2.2687664040665508e-07,
      "loss": 25.0428,
      "step": 33400
    },
    {
      "epoch": 265.07976071784645,
      "eval_loss": 2.9678587913513184,
      "eval_mae": 1.283751368522644,
      "eval_mse": 2.9678587913513184,
      "eval_r2": 0.10579860210418701,
      "eval_rmse": 1.722747454315727,
      "eval_runtime": 9.096,
      "eval_samples_per_second": 440.962,
      "eval_steps_per_second": 13.852,
      "step": 33400
    },
    {
      "epoch": 265.1595214356929,
      "grad_norm": 1086.7607421875,
      "learning_rate": 2.2675241272866288e-07,
      "loss": 22.6771,
      "step": 33410
    },
    {
      "epoch": 265.23928215353936,
      "grad_norm": 171.2627716064453,
      "learning_rate": 2.2662819084077185e-07,
      "loss": 22.4117,
      "step": 33420
    },
    {
      "epoch": 265.3190428713858,
      "grad_norm": 365.1992492675781,
      "learning_rate": 2.2650397477392102e-07,
      "loss": 23.4348,
      "step": 33430
    },
    {
      "epoch": 265.39880358923233,
      "grad_norm": 253.0224609375,
      "learning_rate": 2.2637976455904806e-07,
      "loss": 21.7368,
      "step": 33440
    },
    {
      "epoch": 265.4785643070788,
      "grad_norm": 929.4453735351562,
      "learning_rate": 2.2625556022708893e-07,
      "loss": 22.2043,
      "step": 33450
    },
    {
      "epoch": 265.55832502492524,
      "grad_norm": 283.2198181152344,
      "learning_rate": 2.261313618089782e-07,
      "loss": 23.4093,
      "step": 33460
    },
    {
      "epoch": 265.6380857427717,
      "grad_norm": 1322.727294921875,
      "learning_rate": 2.2600716933564916e-07,
      "loss": 23.9993,
      "step": 33470
    },
    {
      "epoch": 265.71784646061815,
      "grad_norm": 414.2745666503906,
      "learning_rate": 2.258829828380336e-07,
      "loss": 22.8635,
      "step": 33480
    },
    {
      "epoch": 265.7976071784646,
      "grad_norm": 1287.536376953125,
      "learning_rate": 2.2575880234706143e-07,
      "loss": 22.029,
      "step": 33490
    },
    {
      "epoch": 265.87736789631106,
      "grad_norm": 283.93707275390625,
      "learning_rate": 2.2563462789366157e-07,
      "loss": 25.2381,
      "step": 33500
    },
    {
      "epoch": 265.87736789631106,
      "eval_loss": 2.9643938541412354,
      "eval_mae": 1.294697880744934,
      "eval_mse": 2.9643938541412354,
      "eval_r2": 0.10684263706207275,
      "eval_rmse": 1.7217415178072566,
      "eval_runtime": 9.0899,
      "eval_samples_per_second": 441.26,
      "eval_steps_per_second": 13.862,
      "step": 33500
    },
    {
      "epoch": 265.9571286141575,
      "grad_norm": 976.2879638671875,
      "learning_rate": 2.2552287607328728e-07,
      "loss": 24.0215,
      "step": 33510
    },
    {
      "epoch": 266.03190428713856,
      "grad_norm": 420.861083984375,
      "learning_rate": 2.2539871317647784e-07,
      "loss": 20.7929,
      "step": 33520
    },
    {
      "epoch": 266.11166500498507,
      "grad_norm": 324.5087585449219,
      "learning_rate": 2.252745564069253e-07,
      "loss": 22.9747,
      "step": 33530
    },
    {
      "epoch": 266.1914257228315,
      "grad_norm": 211.6680908203125,
      "learning_rate": 2.2515040579555258e-07,
      "loss": 22.1335,
      "step": 33540
    },
    {
      "epoch": 266.271186440678,
      "grad_norm": 562.0562133789062,
      "learning_rate": 2.2502626137328074e-07,
      "loss": 22.2194,
      "step": 33550
    },
    {
      "epoch": 266.35094715852443,
      "grad_norm": 475.1398620605469,
      "learning_rate": 2.2490212317102964e-07,
      "loss": 22.0623,
      "step": 33560
    },
    {
      "epoch": 266.4307078763709,
      "grad_norm": 356.03564453125,
      "learning_rate": 2.2477799121971735e-07,
      "loss": 22.0961,
      "step": 33570
    },
    {
      "epoch": 266.51046859421734,
      "grad_norm": 485.0445251464844,
      "learning_rate": 2.2465386555026062e-07,
      "loss": 24.8321,
      "step": 33580
    },
    {
      "epoch": 266.5902293120638,
      "grad_norm": 270.8175964355469,
      "learning_rate": 2.245297461935743e-07,
      "loss": 22.296,
      "step": 33590
    },
    {
      "epoch": 266.66999002991025,
      "grad_norm": 637.0756225585938,
      "learning_rate": 2.2440563318057204e-07,
      "loss": 23.2108,
      "step": 33600
    },
    {
      "epoch": 266.66999002991025,
      "eval_loss": 2.9591786861419678,
      "eval_mae": 1.293260931968689,
      "eval_mse": 2.9591786861419678,
      "eval_r2": 0.1084139347076416,
      "eval_rmse": 1.7202263473572215,
      "eval_runtime": 9.0779,
      "eval_samples_per_second": 441.842,
      "eval_steps_per_second": 13.88,
      "step": 33600
    },
    {
      "epoch": 266.7497507477567,
      "grad_norm": 958.974365234375,
      "learning_rate": 2.242815265421656e-07,
      "loss": 25.1626,
      "step": 33610
    },
    {
      "epoch": 266.8295114656032,
      "grad_norm": 366.7208251953125,
      "learning_rate": 2.2415742630926533e-07,
      "loss": 22.6531,
      "step": 33620
    },
    {
      "epoch": 266.9092721834497,
      "grad_norm": 643.1389770507812,
      "learning_rate": 2.2403333251278005e-07,
      "loss": 22.5973,
      "step": 33630
    },
    {
      "epoch": 266.9890329012961,
      "grad_norm": 793.8681640625,
      "learning_rate": 2.239092451836167e-07,
      "loss": 25.9147,
      "step": 33640
    },
    {
      "epoch": 267.0638085742772,
      "grad_norm": 199.65530395507812,
      "learning_rate": 2.2378516435268083e-07,
      "loss": 21.7216,
      "step": 33650
    },
    {
      "epoch": 267.14356929212363,
      "grad_norm": 470.8975524902344,
      "learning_rate": 2.236610900508763e-07,
      "loss": 23.0512,
      "step": 33660
    },
    {
      "epoch": 267.2233300099701,
      "grad_norm": 669.757080078125,
      "learning_rate": 2.2353702230910548e-07,
      "loss": 21.9978,
      "step": 33670
    },
    {
      "epoch": 267.30309072781654,
      "grad_norm": 1159.8905029296875,
      "learning_rate": 2.2341296115826888e-07,
      "loss": 22.9267,
      "step": 33680
    },
    {
      "epoch": 267.382851445663,
      "grad_norm": 713.3804321289062,
      "learning_rate": 2.2328890662926542e-07,
      "loss": 23.9274,
      "step": 33690
    },
    {
      "epoch": 267.46261216350945,
      "grad_norm": 351.3713073730469,
      "learning_rate": 2.2316485875299245e-07,
      "loss": 24.1228,
      "step": 33700
    },
    {
      "epoch": 267.46261216350945,
      "eval_loss": 2.970219135284424,
      "eval_mae": 1.299340009689331,
      "eval_mse": 2.970219135284424,
      "eval_r2": 0.1050875186920166,
      "eval_rmse": 1.723432370383133,
      "eval_runtime": 9.0756,
      "eval_samples_per_second": 441.955,
      "eval_steps_per_second": 13.883,
      "step": 33700
    },
    {
      "epoch": 267.54237288135596,
      "grad_norm": 586.8211669921875,
      "learning_rate": 2.230408175603458e-07,
      "loss": 22.3554,
      "step": 33710
    },
    {
      "epoch": 267.6221335992024,
      "grad_norm": 1108.204345703125,
      "learning_rate": 2.229167830822194e-07,
      "loss": 24.1477,
      "step": 33720
    },
    {
      "epoch": 267.70189431704887,
      "grad_norm": 360.2843322753906,
      "learning_rate": 2.227927553495054e-07,
      "loss": 22.8538,
      "step": 33730
    },
    {
      "epoch": 267.7816550348953,
      "grad_norm": 742.3580932617188,
      "learning_rate": 2.226687343930946e-07,
      "loss": 24.0058,
      "step": 33740
    },
    {
      "epoch": 267.8614157527418,
      "grad_norm": 1537.28369140625,
      "learning_rate": 2.2254472024387603e-07,
      "loss": 23.1938,
      "step": 33750
    },
    {
      "epoch": 267.94117647058823,
      "grad_norm": 407.59381103515625,
      "learning_rate": 2.2242071293273682e-07,
      "loss": 22.7664,
      "step": 33760
    },
    {
      "epoch": 268.0159521435693,
      "grad_norm": 220.43446350097656,
      "learning_rate": 2.222967124905627e-07,
      "loss": 20.4323,
      "step": 33770
    },
    {
      "epoch": 268.09571286141573,
      "grad_norm": 549.47021484375,
      "learning_rate": 2.2217271894823734e-07,
      "loss": 21.525,
      "step": 33780
    },
    {
      "epoch": 268.1754735792622,
      "grad_norm": 337.71630859375,
      "learning_rate": 2.22048732336643e-07,
      "loss": 24.2887,
      "step": 33790
    },
    {
      "epoch": 268.2552342971087,
      "grad_norm": 461.509521484375,
      "learning_rate": 2.2192475268665996e-07,
      "loss": 23.4435,
      "step": 33800
    },
    {
      "epoch": 268.2552342971087,
      "eval_loss": 2.960710048675537,
      "eval_mae": 1.2898505926132202,
      "eval_mse": 2.960710048675537,
      "eval_r2": 0.1079525351524353,
      "eval_rmse": 1.7206713947397212,
      "eval_runtime": 9.0974,
      "eval_samples_per_second": 440.897,
      "eval_steps_per_second": 13.85,
      "step": 33800
    },
    {
      "epoch": 268.33499501495515,
      "grad_norm": 1379.5404052734375,
      "learning_rate": 2.2180078002916714e-07,
      "loss": 21.4999,
      "step": 33810
    },
    {
      "epoch": 268.4147557328016,
      "grad_norm": 468.10601806640625,
      "learning_rate": 2.216768143950412e-07,
      "loss": 23.6818,
      "step": 33820
    },
    {
      "epoch": 268.49451645064806,
      "grad_norm": 302.5543518066406,
      "learning_rate": 2.2155285581515746e-07,
      "loss": 22.8665,
      "step": 33830
    },
    {
      "epoch": 268.5742771684945,
      "grad_norm": 536.4502563476562,
      "learning_rate": 2.2142890432038938e-07,
      "loss": 23.9307,
      "step": 33840
    },
    {
      "epoch": 268.654037886341,
      "grad_norm": 948.4057006835938,
      "learning_rate": 2.2130495994160854e-07,
      "loss": 22.1223,
      "step": 33850
    },
    {
      "epoch": 268.7337986041874,
      "grad_norm": 306.21533203125,
      "learning_rate": 2.2118102270968494e-07,
      "loss": 24.5972,
      "step": 33860
    },
    {
      "epoch": 268.8135593220339,
      "grad_norm": 840.4362182617188,
      "learning_rate": 2.2105709265548657e-07,
      "loss": 24.0767,
      "step": 33870
    },
    {
      "epoch": 268.89332003988034,
      "grad_norm": 420.22808837890625,
      "learning_rate": 2.209331698098798e-07,
      "loss": 25.0304,
      "step": 33880
    },
    {
      "epoch": 268.97308075772685,
      "grad_norm": 587.11474609375,
      "learning_rate": 2.2080925420372916e-07,
      "loss": 21.3054,
      "step": 33890
    },
    {
      "epoch": 269.0478564307079,
      "grad_norm": 547.7639770507812,
      "learning_rate": 2.2068534586789733e-07,
      "loss": 19.684,
      "step": 33900
    },
    {
      "epoch": 269.0478564307079,
      "eval_loss": 2.9617063999176025,
      "eval_mae": 1.2873121500015259,
      "eval_mse": 2.9617063999176025,
      "eval_r2": 0.10765236616134644,
      "eval_rmse": 1.7209608943603578,
      "eval_runtime": 9.0903,
      "eval_samples_per_second": 441.239,
      "eval_steps_per_second": 13.861,
      "step": 33900
    },
    {
      "epoch": 269.12761714855435,
      "grad_norm": 651.6534423828125,
      "learning_rate": 2.205614448332454e-07,
      "loss": 23.9525,
      "step": 33910
    },
    {
      "epoch": 269.2073778664008,
      "grad_norm": 1018.7666015625,
      "learning_rate": 2.204375511306323e-07,
      "loss": 25.1037,
      "step": 33920
    },
    {
      "epoch": 269.28713858424726,
      "grad_norm": 970.5116577148438,
      "learning_rate": 2.203136647909153e-07,
      "loss": 22.9558,
      "step": 33930
    },
    {
      "epoch": 269.3668993020937,
      "grad_norm": 222.009765625,
      "learning_rate": 2.2018978584494986e-07,
      "loss": 24.3678,
      "step": 33940
    },
    {
      "epoch": 269.44666001994017,
      "grad_norm": 195.72454833984375,
      "learning_rate": 2.2006591432358968e-07,
      "loss": 22.8183,
      "step": 33950
    },
    {
      "epoch": 269.5264207377866,
      "grad_norm": 372.6313171386719,
      "learning_rate": 2.199420502576864e-07,
      "loss": 21.7113,
      "step": 33960
    },
    {
      "epoch": 269.6061814556331,
      "grad_norm": 259.13079833984375,
      "learning_rate": 2.1981819367808982e-07,
      "loss": 24.587,
      "step": 33970
    },
    {
      "epoch": 269.6859421734796,
      "grad_norm": 737.356201171875,
      "learning_rate": 2.1969434461564812e-07,
      "loss": 22.8807,
      "step": 33980
    },
    {
      "epoch": 269.76570289132604,
      "grad_norm": 760.752685546875,
      "learning_rate": 2.195705031012074e-07,
      "loss": 22.7561,
      "step": 33990
    },
    {
      "epoch": 269.8454636091725,
      "grad_norm": 552.60205078125,
      "learning_rate": 2.1944666916561202e-07,
      "loss": 24.186,
      "step": 34000
    },
    {
      "epoch": 269.8454636091725,
      "eval_loss": 2.962271213531494,
      "eval_mae": 1.285582423210144,
      "eval_mse": 2.962270975112915,
      "eval_r2": 0.10748225450515747,
      "eval_rmse": 1.7211249156040114,
      "eval_runtime": 9.072,
      "eval_samples_per_second": 442.129,
      "eval_steps_per_second": 13.889,
      "step": 34000
    },
    {
      "epoch": 269.92522432701895,
      "grad_norm": 737.112548828125,
      "learning_rate": 2.1932284283970417e-07,
      "loss": 23.0482,
      "step": 34010
    },
    {
      "epoch": 270.0,
      "grad_norm": 179.2029266357422,
      "learning_rate": 2.1919902415432446e-07,
      "loss": 19.8481,
      "step": 34020
    },
    {
      "epoch": 270.07976071784645,
      "grad_norm": 638.3309326171875,
      "learning_rate": 2.1907521314031148e-07,
      "loss": 23.9653,
      "step": 34030
    },
    {
      "epoch": 270.1595214356929,
      "grad_norm": 1045.047119140625,
      "learning_rate": 2.1895140982850197e-07,
      "loss": 23.5545,
      "step": 34040
    },
    {
      "epoch": 270.23928215353936,
      "grad_norm": 982.6498413085938,
      "learning_rate": 2.1882761424973047e-07,
      "loss": 22.3798,
      "step": 34050
    },
    {
      "epoch": 270.3190428713858,
      "grad_norm": 318.80615234375,
      "learning_rate": 2.1870382643482995e-07,
      "loss": 21.9895,
      "step": 34060
    },
    {
      "epoch": 270.39880358923233,
      "grad_norm": 856.8302001953125,
      "learning_rate": 2.1858004641463139e-07,
      "loss": 20.5062,
      "step": 34070
    },
    {
      "epoch": 270.4785643070788,
      "grad_norm": 601.9522094726562,
      "learning_rate": 2.1845627421996357e-07,
      "loss": 23.5981,
      "step": 34080
    },
    {
      "epoch": 270.55832502492524,
      "grad_norm": 570.8819580078125,
      "learning_rate": 2.183325098816537e-07,
      "loss": 23.5832,
      "step": 34090
    },
    {
      "epoch": 270.6380857427717,
      "grad_norm": 380.88885498046875,
      "learning_rate": 2.1820875343052662e-07,
      "loss": 23.4134,
      "step": 34100
    },
    {
      "epoch": 270.6380857427717,
      "eval_loss": 2.9615285396575928,
      "eval_mae": 1.2914930582046509,
      "eval_mse": 2.9615285396575928,
      "eval_r2": 0.10770589113235474,
      "eval_rmse": 1.7209092188891293,
      "eval_runtime": 9.0815,
      "eval_samples_per_second": 441.666,
      "eval_steps_per_second": 13.874,
      "step": 34100
    },
    {
      "epoch": 270.71784646061815,
      "grad_norm": 704.3423461914062,
      "learning_rate": 2.1808500489740552e-07,
      "loss": 22.3576,
      "step": 34110
    },
    {
      "epoch": 270.7976071784646,
      "grad_norm": 511.343017578125,
      "learning_rate": 2.179612643131115e-07,
      "loss": 24.4609,
      "step": 34120
    },
    {
      "epoch": 270.87736789631106,
      "grad_norm": 784.0336303710938,
      "learning_rate": 2.178375317084637e-07,
      "loss": 24.3966,
      "step": 34130
    },
    {
      "epoch": 270.9571286141575,
      "grad_norm": 842.5420532226562,
      "learning_rate": 2.1771380711427933e-07,
      "loss": 23.4618,
      "step": 34140
    },
    {
      "epoch": 271.03190428713856,
      "grad_norm": 345.5811462402344,
      "learning_rate": 2.1759009056137346e-07,
      "loss": 22.0152,
      "step": 34150
    },
    {
      "epoch": 271.11166500498507,
      "grad_norm": 315.23004150390625,
      "learning_rate": 2.1746638208055917e-07,
      "loss": 23.3748,
      "step": 34160
    },
    {
      "epoch": 271.1914257228315,
      "grad_norm": 208.2379608154297,
      "learning_rate": 2.1734268170264768e-07,
      "loss": 22.5506,
      "step": 34170
    },
    {
      "epoch": 271.271186440678,
      "grad_norm": 330.3124084472656,
      "learning_rate": 2.1721898945844824e-07,
      "loss": 22.3357,
      "step": 34180
    },
    {
      "epoch": 271.35094715852443,
      "grad_norm": 461.0664978027344,
      "learning_rate": 2.1709530537876772e-07,
      "loss": 22.663,
      "step": 34190
    },
    {
      "epoch": 271.4307078763709,
      "grad_norm": 825.6907958984375,
      "learning_rate": 2.1697162949441133e-07,
      "loss": 24.8071,
      "step": 34200
    },
    {
      "epoch": 271.4307078763709,
      "eval_loss": 2.970717191696167,
      "eval_mae": 1.2850940227508545,
      "eval_mse": 2.970716953277588,
      "eval_r2": 0.10493749380111694,
      "eval_rmse": 1.7235767906529689,
      "eval_runtime": 9.0672,
      "eval_samples_per_second": 442.366,
      "eval_steps_per_second": 13.896,
      "step": 34200
    },
    {
      "epoch": 271.51046859421734,
      "grad_norm": 412.7564697265625,
      "learning_rate": 2.1684796183618198e-07,
      "loss": 22.3887,
      "step": 34210
    },
    {
      "epoch": 271.5902293120638,
      "grad_norm": 289.83331298828125,
      "learning_rate": 2.1672430243488068e-07,
      "loss": 23.1303,
      "step": 34220
    },
    {
      "epoch": 271.66999002991025,
      "grad_norm": 297.7489929199219,
      "learning_rate": 2.166006513213065e-07,
      "loss": 23.5059,
      "step": 34230
    },
    {
      "epoch": 271.7497507477567,
      "grad_norm": 964.3274536132812,
      "learning_rate": 2.1647700852625605e-07,
      "loss": 23.5733,
      "step": 34240
    },
    {
      "epoch": 271.8295114656032,
      "grad_norm": 617.3482055664062,
      "learning_rate": 2.163533740805242e-07,
      "loss": 21.2103,
      "step": 34250
    },
    {
      "epoch": 271.9092721834497,
      "grad_norm": 525.2120361328125,
      "learning_rate": 2.1622974801490363e-07,
      "loss": 22.3432,
      "step": 34260
    },
    {
      "epoch": 271.9890329012961,
      "grad_norm": 508.3149719238281,
      "learning_rate": 2.161061303601851e-07,
      "loss": 25.6466,
      "step": 34270
    },
    {
      "epoch": 272.0638085742772,
      "grad_norm": 1264.3121337890625,
      "learning_rate": 2.1598252114715698e-07,
      "loss": 21.0367,
      "step": 34280
    },
    {
      "epoch": 272.14356929212363,
      "grad_norm": 820.3016357421875,
      "learning_rate": 2.1585892040660565e-07,
      "loss": 21.2672,
      "step": 34290
    },
    {
      "epoch": 272.2233300099701,
      "grad_norm": 486.15887451171875,
      "learning_rate": 2.1573532816931545e-07,
      "loss": 23.8251,
      "step": 34300
    },
    {
      "epoch": 272.2233300099701,
      "eval_loss": 2.9611992835998535,
      "eval_mae": 1.287341594696045,
      "eval_mse": 2.9611990451812744,
      "eval_r2": 0.10780519247055054,
      "eval_rmse": 1.7208134835540063,
      "eval_runtime": 9.1254,
      "eval_samples_per_second": 439.544,
      "eval_steps_per_second": 13.808,
      "step": 34300
    },
    {
      "epoch": 272.30309072781654,
      "grad_norm": 546.13037109375,
      "learning_rate": 2.1561174446606867e-07,
      "loss": 23.5214,
      "step": 34310
    },
    {
      "epoch": 272.382851445663,
      "grad_norm": 800.3176879882812,
      "learning_rate": 2.1548816932764532e-07,
      "loss": 23.1447,
      "step": 34320
    },
    {
      "epoch": 272.46261216350945,
      "grad_norm": 363.2491455078125,
      "learning_rate": 2.153646027848232e-07,
      "loss": 22.5961,
      "step": 34330
    },
    {
      "epoch": 272.54237288135596,
      "grad_norm": 461.95184326171875,
      "learning_rate": 2.1524104486837822e-07,
      "loss": 22.1595,
      "step": 34340
    },
    {
      "epoch": 272.6221335992024,
      "grad_norm": 179.94744873046875,
      "learning_rate": 2.1511749560908403e-07,
      "loss": 24.6815,
      "step": 34350
    },
    {
      "epoch": 272.70189431704887,
      "grad_norm": 446.2342834472656,
      "learning_rate": 2.1499395503771205e-07,
      "loss": 24.2278,
      "step": 34360
    },
    {
      "epoch": 272.7816550348953,
      "grad_norm": 617.9832153320312,
      "learning_rate": 2.1487042318503172e-07,
      "loss": 22.8166,
      "step": 34370
    },
    {
      "epoch": 272.8614157527418,
      "grad_norm": 328.90850830078125,
      "learning_rate": 2.1474690008180995e-07,
      "loss": 23.1608,
      "step": 34380
    },
    {
      "epoch": 272.94117647058823,
      "grad_norm": 463.5259704589844,
      "learning_rate": 2.1462338575881195e-07,
      "loss": 23.3504,
      "step": 34390
    },
    {
      "epoch": 273.0159521435693,
      "grad_norm": 176.9029083251953,
      "learning_rate": 2.1449988024680032e-07,
      "loss": 21.4252,
      "step": 34400
    },
    {
      "epoch": 273.0159521435693,
      "eval_loss": 2.968507766723633,
      "eval_mae": 1.29869544506073,
      "eval_mse": 2.9685075283050537,
      "eval_r2": 0.10560321807861328,
      "eval_rmse": 1.7229357295920975,
      "eval_runtime": 9.0848,
      "eval_samples_per_second": 441.505,
      "eval_steps_per_second": 13.869,
      "step": 34400
    },
    {
      "epoch": 273.09571286141573,
      "grad_norm": 886.4635009765625,
      "learning_rate": 2.1437638357653583e-07,
      "loss": 22.1249,
      "step": 34410
    },
    {
      "epoch": 273.1754735792622,
      "grad_norm": 790.850830078125,
      "learning_rate": 2.142528957787767e-07,
      "loss": 23.062,
      "step": 34420
    },
    {
      "epoch": 273.2552342971087,
      "grad_norm": 862.314697265625,
      "learning_rate": 2.1412941688427915e-07,
      "loss": 21.4663,
      "step": 34430
    },
    {
      "epoch": 273.33499501495515,
      "grad_norm": 408.88726806640625,
      "learning_rate": 2.1400594692379714e-07,
      "loss": 25.0917,
      "step": 34440
    },
    {
      "epoch": 273.4147557328016,
      "grad_norm": 1136.371826171875,
      "learning_rate": 2.1388248592808243e-07,
      "loss": 23.7612,
      "step": 34450
    },
    {
      "epoch": 273.49451645064806,
      "grad_norm": 459.9880065917969,
      "learning_rate": 2.137590339278846e-07,
      "loss": 25.5118,
      "step": 34460
    },
    {
      "epoch": 273.5742771684945,
      "grad_norm": 501.0910949707031,
      "learning_rate": 2.1363559095395073e-07,
      "loss": 22.8884,
      "step": 34470
    },
    {
      "epoch": 273.654037886341,
      "grad_norm": 395.23516845703125,
      "learning_rate": 2.1351215703702587e-07,
      "loss": 24.1507,
      "step": 34480
    },
    {
      "epoch": 273.7337986041874,
      "grad_norm": 1578.13623046875,
      "learning_rate": 2.1338873220785285e-07,
      "loss": 23.6566,
      "step": 34490
    },
    {
      "epoch": 273.8135593220339,
      "grad_norm": 566.108154296875,
      "learning_rate": 2.1326531649717214e-07,
      "loss": 20.345,
      "step": 34500
    },
    {
      "epoch": 273.8135593220339,
      "eval_loss": 2.959744453430176,
      "eval_mae": 1.2896983623504639,
      "eval_mse": 2.959744453430176,
      "eval_r2": 0.10824346542358398,
      "eval_rmse": 1.7203907850922058,
      "eval_runtime": 9.0942,
      "eval_samples_per_second": 441.052,
      "eval_steps_per_second": 13.855,
      "step": 34500
    },
    {
      "epoch": 273.89332003988034,
      "grad_norm": 167.27906799316406,
      "learning_rate": 2.1314190993572196e-07,
      "loss": 24.3434,
      "step": 34510
    },
    {
      "epoch": 273.97308075772685,
      "grad_norm": 1610.5982666015625,
      "learning_rate": 2.1301851255423826e-07,
      "loss": 21.422,
      "step": 34520
    },
    {
      "epoch": 274.0478564307079,
      "grad_norm": 1023.9450073242188,
      "learning_rate": 2.128951243834546e-07,
      "loss": 21.6396,
      "step": 34530
    },
    {
      "epoch": 274.12761714855435,
      "grad_norm": 578.3451538085938,
      "learning_rate": 2.1277174545410246e-07,
      "loss": 23.1165,
      "step": 34540
    },
    {
      "epoch": 274.2073778664008,
      "grad_norm": 815.1294555664062,
      "learning_rate": 2.1264837579691085e-07,
      "loss": 25.2555,
      "step": 34550
    },
    {
      "epoch": 274.28713858424726,
      "grad_norm": 279.6356506347656,
      "learning_rate": 2.1252501544260654e-07,
      "loss": 21.8185,
      "step": 34560
    },
    {
      "epoch": 274.3668993020937,
      "grad_norm": 782.1747436523438,
      "learning_rate": 2.1240166442191388e-07,
      "loss": 23.5705,
      "step": 34570
    },
    {
      "epoch": 274.44666001994017,
      "grad_norm": 440.4573059082031,
      "learning_rate": 2.1227832276555507e-07,
      "loss": 23.1609,
      "step": 34580
    },
    {
      "epoch": 274.5264207377866,
      "grad_norm": 549.3892211914062,
      "learning_rate": 2.121549905042499e-07,
      "loss": 21.4242,
      "step": 34590
    },
    {
      "epoch": 274.6061814556331,
      "grad_norm": 1361.94189453125,
      "learning_rate": 2.120316676687158e-07,
      "loss": 23.3436,
      "step": 34600
    },
    {
      "epoch": 274.6061814556331,
      "eval_loss": 2.9784059524536133,
      "eval_mae": 1.2834292650222778,
      "eval_mse": 2.9784059524536133,
      "eval_r2": 0.10262084007263184,
      "eval_rmse": 1.7258058849284335,
      "eval_runtime": 9.097,
      "eval_samples_per_second": 440.916,
      "eval_steps_per_second": 13.851,
      "step": 34600
    },
    {
      "epoch": 274.6859421734796,
      "grad_norm": 338.5691223144531,
      "learning_rate": 2.1190835428966775e-07,
      "loss": 23.4626,
      "step": 34610
    },
    {
      "epoch": 274.76570289132604,
      "grad_norm": 707.038330078125,
      "learning_rate": 2.1178505039781854e-07,
      "loss": 22.324,
      "step": 34620
    },
    {
      "epoch": 274.8454636091725,
      "grad_norm": 794.2965698242188,
      "learning_rate": 2.1166175602387862e-07,
      "loss": 21.627,
      "step": 34630
    },
    {
      "epoch": 274.92522432701895,
      "grad_norm": 701.9091796875,
      "learning_rate": 2.1153847119855597e-07,
      "loss": 24.6566,
      "step": 34640
    },
    {
      "epoch": 275.0,
      "grad_norm": 420.62335205078125,
      "learning_rate": 2.1141519595255608e-07,
      "loss": 21.9168,
      "step": 34650
    },
    {
      "epoch": 275.07976071784645,
      "grad_norm": 435.24560546875,
      "learning_rate": 2.1129193031658226e-07,
      "loss": 22.1155,
      "step": 34660
    },
    {
      "epoch": 275.1595214356929,
      "grad_norm": 1613.70654296875,
      "learning_rate": 2.1116867432133546e-07,
      "loss": 23.537,
      "step": 34670
    },
    {
      "epoch": 275.23928215353936,
      "grad_norm": 318.3715515136719,
      "learning_rate": 2.1104542799751392e-07,
      "loss": 23.4494,
      "step": 34680
    },
    {
      "epoch": 275.3190428713858,
      "grad_norm": 196.35650634765625,
      "learning_rate": 2.109221913758139e-07,
      "loss": 24.0127,
      "step": 34690
    },
    {
      "epoch": 275.39880358923233,
      "grad_norm": 288.3596496582031,
      "learning_rate": 2.1079896448692884e-07,
      "loss": 21.391,
      "step": 34700
    },
    {
      "epoch": 275.39880358923233,
      "eval_loss": 2.961453676223755,
      "eval_mae": 1.286414384841919,
      "eval_mse": 2.961453676223755,
      "eval_r2": 0.10772848129272461,
      "eval_rmse": 1.7208874676235384,
      "eval_runtime": 9.0785,
      "eval_samples_per_second": 441.815,
      "eval_steps_per_second": 13.879,
      "step": 34700
    },
    {
      "epoch": 275.4785643070788,
      "grad_norm": 565.7955932617188,
      "learning_rate": 2.1067574736154996e-07,
      "loss": 22.4207,
      "step": 34710
    },
    {
      "epoch": 275.55832502492524,
      "grad_norm": 351.7076721191406,
      "learning_rate": 2.1055254003036604e-07,
      "loss": 22.5282,
      "step": 34720
    },
    {
      "epoch": 275.6380857427717,
      "grad_norm": 591.8775024414062,
      "learning_rate": 2.1042934252406342e-07,
      "loss": 23.4614,
      "step": 34730
    },
    {
      "epoch": 275.71784646061815,
      "grad_norm": 387.2130126953125,
      "learning_rate": 2.1030615487332607e-07,
      "loss": 21.8597,
      "step": 34740
    },
    {
      "epoch": 275.7976071784646,
      "grad_norm": 702.9859008789062,
      "learning_rate": 2.1018297710883523e-07,
      "loss": 23.8811,
      "step": 34750
    },
    {
      "epoch": 275.87736789631106,
      "grad_norm": 718.348388671875,
      "learning_rate": 2.1005980926126987e-07,
      "loss": 22.5782,
      "step": 34760
    },
    {
      "epoch": 275.9571286141575,
      "grad_norm": 815.12158203125,
      "learning_rate": 2.0993665136130652e-07,
      "loss": 24.5857,
      "step": 34770
    },
    {
      "epoch": 276.03190428713856,
      "grad_norm": 349.52239990234375,
      "learning_rate": 2.0981350343961934e-07,
      "loss": 21.4769,
      "step": 34780
    },
    {
      "epoch": 276.11166500498507,
      "grad_norm": 413.54833984375,
      "learning_rate": 2.0969036552687969e-07,
      "loss": 23.0135,
      "step": 34790
    },
    {
      "epoch": 276.1914257228315,
      "grad_norm": 314.8276672363281,
      "learning_rate": 2.0956723765375652e-07,
      "loss": 23.4292,
      "step": 34800
    },
    {
      "epoch": 276.1914257228315,
      "eval_loss": 2.9737393856048584,
      "eval_mae": 1.2879493236541748,
      "eval_mse": 2.9737396240234375,
      "eval_r2": 0.10402679443359375,
      "eval_rmse": 1.724453427617991,
      "eval_runtime": 9.0827,
      "eval_samples_per_second": 441.609,
      "eval_steps_per_second": 13.873,
      "step": 34800
    },
    {
      "epoch": 276.271186440678,
      "grad_norm": 204.84500122070312,
      "learning_rate": 2.0944411985091648e-07,
      "loss": 23.4302,
      "step": 34810
    },
    {
      "epoch": 276.35094715852443,
      "grad_norm": 445.2061462402344,
      "learning_rate": 2.0932101214902366e-07,
      "loss": 21.2021,
      "step": 34820
    },
    {
      "epoch": 276.4307078763709,
      "grad_norm": 338.8351745605469,
      "learning_rate": 2.0919791457873947e-07,
      "loss": 23.7226,
      "step": 34830
    },
    {
      "epoch": 276.51046859421734,
      "grad_norm": 590.1646118164062,
      "learning_rate": 2.0907482717072292e-07,
      "loss": 23.4003,
      "step": 34840
    },
    {
      "epoch": 276.5902293120638,
      "grad_norm": 253.9158172607422,
      "learning_rate": 2.089517499556304e-07,
      "loss": 22.5189,
      "step": 34850
    },
    {
      "epoch": 276.66999002991025,
      "grad_norm": 235.29299926757812,
      "learning_rate": 2.0882868296411594e-07,
      "loss": 22.7965,
      "step": 34860
    },
    {
      "epoch": 276.7497507477567,
      "grad_norm": 1468.8023681640625,
      "learning_rate": 2.0870562622683075e-07,
      "loss": 25.4464,
      "step": 34870
    },
    {
      "epoch": 276.8295114656032,
      "grad_norm": 281.73284912109375,
      "learning_rate": 2.085825797744239e-07,
      "loss": 23.6252,
      "step": 34880
    },
    {
      "epoch": 276.9092721834497,
      "grad_norm": 2000.7344970703125,
      "learning_rate": 2.084595436375413e-07,
      "loss": 24.2562,
      "step": 34890
    },
    {
      "epoch": 276.9890329012961,
      "grad_norm": 1607.6239013671875,
      "learning_rate": 2.0833651784682687e-07,
      "loss": 22.7612,
      "step": 34900
    },
    {
      "epoch": 276.9890329012961,
      "eval_loss": 2.9629199504852295,
      "eval_mae": 1.2866618633270264,
      "eval_mse": 2.9629199504852295,
      "eval_r2": 0.10728675127029419,
      "eval_rmse": 1.7213134376066521,
      "eval_runtime": 9.0793,
      "eval_samples_per_second": 441.772,
      "eval_steps_per_second": 13.878,
      "step": 34900
    },
    {
      "epoch": 277.0638085742772,
      "grad_norm": 2094.517578125,
      "learning_rate": 2.0821350243292172e-07,
      "loss": 20.5139,
      "step": 34910
    },
    {
      "epoch": 277.14356929212363,
      "grad_norm": 459.9913024902344,
      "learning_rate": 2.0809049742646434e-07,
      "loss": 23.6329,
      "step": 34920
    },
    {
      "epoch": 277.2233300099701,
      "grad_norm": 976.9309692382812,
      "learning_rate": 2.079675028580905e-07,
      "loss": 22.3355,
      "step": 34930
    },
    {
      "epoch": 277.30309072781654,
      "grad_norm": 324.9153747558594,
      "learning_rate": 2.078445187584337e-07,
      "loss": 23.7818,
      "step": 34940
    },
    {
      "epoch": 277.382851445663,
      "grad_norm": 794.9937744140625,
      "learning_rate": 2.0772154515812464e-07,
      "loss": 23.1308,
      "step": 34950
    },
    {
      "epoch": 277.46261216350945,
      "grad_norm": 552.079345703125,
      "learning_rate": 2.0759858208779136e-07,
      "loss": 22.4865,
      "step": 34960
    },
    {
      "epoch": 277.54237288135596,
      "grad_norm": 181.21153259277344,
      "learning_rate": 2.074756295780595e-07,
      "loss": 21.8313,
      "step": 34970
    },
    {
      "epoch": 277.6221335992024,
      "grad_norm": 497.83392333984375,
      "learning_rate": 2.0735268765955173e-07,
      "loss": 24.0454,
      "step": 34980
    },
    {
      "epoch": 277.70189431704887,
      "grad_norm": 647.4383544921875,
      "learning_rate": 2.0722975636288834e-07,
      "loss": 23.2122,
      "step": 34990
    },
    {
      "epoch": 277.7816550348953,
      "grad_norm": 1826.175048828125,
      "learning_rate": 2.0710683571868688e-07,
      "loss": 22.9969,
      "step": 35000
    },
    {
      "epoch": 277.7816550348953,
      "eval_loss": 2.960876226425171,
      "eval_mae": 1.2907391786575317,
      "eval_mse": 2.960876226425171,
      "eval_r2": 0.10790246725082397,
      "eval_rmse": 1.7207196826982514,
      "eval_runtime": 9.0922,
      "eval_samples_per_second": 441.149,
      "eval_steps_per_second": 13.858,
      "step": 35000
    },
    {
      "epoch": 277.8614157527418,
      "grad_norm": 541.8032836914062,
      "learning_rate": 2.0698392575756238e-07,
      "loss": 22.5628,
      "step": 35010
    },
    {
      "epoch": 277.94117647058823,
      "grad_norm": 661.0296020507812,
      "learning_rate": 2.0686102651012692e-07,
      "loss": 24.3329,
      "step": 35020
    },
    {
      "epoch": 278.0159521435693,
      "grad_norm": 436.8052062988281,
      "learning_rate": 2.0673813800699022e-07,
      "loss": 22.5826,
      "step": 35030
    },
    {
      "epoch": 278.09571286141573,
      "grad_norm": 789.2468872070312,
      "learning_rate": 2.066152602787591e-07,
      "loss": 25.2499,
      "step": 35040
    },
    {
      "epoch": 278.1754735792622,
      "grad_norm": 1029.5194091796875,
      "learning_rate": 2.064923933560378e-07,
      "loss": 23.2788,
      "step": 35050
    },
    {
      "epoch": 278.2552342971087,
      "grad_norm": 276.64434814453125,
      "learning_rate": 2.0636953726942803e-07,
      "loss": 23.5676,
      "step": 35060
    },
    {
      "epoch": 278.33499501495515,
      "grad_norm": 428.64154052734375,
      "learning_rate": 2.0624669204952843e-07,
      "loss": 22.2599,
      "step": 35070
    },
    {
      "epoch": 278.4147557328016,
      "grad_norm": 646.1993408203125,
      "learning_rate": 2.0612385772693517e-07,
      "loss": 22.4207,
      "step": 35080
    },
    {
      "epoch": 278.49451645064806,
      "grad_norm": 725.5416259765625,
      "learning_rate": 2.0600103433224164e-07,
      "loss": 21.0056,
      "step": 35090
    },
    {
      "epoch": 278.5742771684945,
      "grad_norm": 1010.9341430664062,
      "learning_rate": 2.058782218960387e-07,
      "loss": 20.9893,
      "step": 35100
    },
    {
      "epoch": 278.5742771684945,
      "eval_loss": 2.958890438079834,
      "eval_mae": 1.2890186309814453,
      "eval_mse": 2.958890438079834,
      "eval_r2": 0.10850077867507935,
      "eval_rmse": 1.7201425633010288,
      "eval_runtime": 9.0844,
      "eval_samples_per_second": 441.527,
      "eval_steps_per_second": 13.87,
      "step": 35100
    },
    {
      "epoch": 278.654037886341,
      "grad_norm": 703.8577880859375,
      "learning_rate": 2.0575542044891422e-07,
      "loss": 22.8007,
      "step": 35110
    },
    {
      "epoch": 278.7337986041874,
      "grad_norm": 216.5920867919922,
      "learning_rate": 2.0563263002145331e-07,
      "loss": 22.4238,
      "step": 35120
    },
    {
      "epoch": 278.8135593220339,
      "grad_norm": 331.3422546386719,
      "learning_rate": 2.055098506442386e-07,
      "loss": 25.6107,
      "step": 35130
    },
    {
      "epoch": 278.89332003988034,
      "grad_norm": 1468.609130859375,
      "learning_rate": 2.053870823478498e-07,
      "loss": 22.3461,
      "step": 35140
    },
    {
      "epoch": 278.97308075772685,
      "grad_norm": 895.9401245117188,
      "learning_rate": 2.0526432516286394e-07,
      "loss": 24.9516,
      "step": 35150
    },
    {
      "epoch": 279.0478564307079,
      "grad_norm": 588.5226440429688,
      "learning_rate": 2.0514157911985504e-07,
      "loss": 22.7737,
      "step": 35160
    },
    {
      "epoch": 279.12761714855435,
      "grad_norm": 413.3553161621094,
      "learning_rate": 2.0501884424939464e-07,
      "loss": 21.5904,
      "step": 35170
    },
    {
      "epoch": 279.2073778664008,
      "grad_norm": 746.3114013671875,
      "learning_rate": 2.0489612058205147e-07,
      "loss": 23.2071,
      "step": 35180
    },
    {
      "epoch": 279.28713858424726,
      "grad_norm": 1527.2125244140625,
      "learning_rate": 2.0477340814839123e-07,
      "loss": 20.4271,
      "step": 35190
    },
    {
      "epoch": 279.3668993020937,
      "grad_norm": 308.45721435546875,
      "learning_rate": 2.046507069789772e-07,
      "loss": 23.4165,
      "step": 35200
    },
    {
      "epoch": 279.3668993020937,
      "eval_loss": 2.9652814865112305,
      "eval_mae": 1.2923365831375122,
      "eval_mse": 2.9652814865112305,
      "eval_r2": 0.10657519102096558,
      "eval_rmse": 1.7219992701831295,
      "eval_runtime": 9.0553,
      "eval_samples_per_second": 442.945,
      "eval_steps_per_second": 13.915,
      "step": 35200
    },
    {
      "epoch": 279.44666001994017,
      "grad_norm": 956.6458129882812,
      "learning_rate": 2.0452801710436936e-07,
      "loss": 23.4536,
      "step": 35210
    },
    {
      "epoch": 279.5264207377866,
      "grad_norm": 168.210205078125,
      "learning_rate": 2.0440533855512537e-07,
      "loss": 23.5192,
      "step": 35220
    },
    {
      "epoch": 279.6061814556331,
      "grad_norm": 475.0094909667969,
      "learning_rate": 2.0428267136179972e-07,
      "loss": 24.1107,
      "step": 35230
    },
    {
      "epoch": 279.6859421734796,
      "grad_norm": 760.2296752929688,
      "learning_rate": 2.0416001555494433e-07,
      "loss": 23.5051,
      "step": 35240
    },
    {
      "epoch": 279.76570289132604,
      "grad_norm": 164.65597534179688,
      "learning_rate": 2.0403737116510796e-07,
      "loss": 23.0331,
      "step": 35250
    },
    {
      "epoch": 279.8454636091725,
      "grad_norm": 833.2396240234375,
      "learning_rate": 2.0391473822283688e-07,
      "loss": 23.6303,
      "step": 35260
    },
    {
      "epoch": 279.92522432701895,
      "grad_norm": 849.0071411132812,
      "learning_rate": 2.0379211675867436e-07,
      "loss": 22.9304,
      "step": 35270
    },
    {
      "epoch": 280.0,
      "grad_norm": 560.7474365234375,
      "learning_rate": 2.036695068031607e-07,
      "loss": 21.6578,
      "step": 35280
    },
    {
      "epoch": 280.07976071784645,
      "grad_norm": 424.35943603515625,
      "learning_rate": 2.035469083868336e-07,
      "loss": 22.7506,
      "step": 35290
    },
    {
      "epoch": 280.1595214356929,
      "grad_norm": 641.072509765625,
      "learning_rate": 2.0342432154022748e-07,
      "loss": 23.4757,
      "step": 35300
    },
    {
      "epoch": 280.1595214356929,
      "eval_loss": 2.962219715118408,
      "eval_mae": 1.29058039188385,
      "eval_mse": 2.962219476699829,
      "eval_r2": 0.10749781131744385,
      "eval_rmse": 1.7211099548546656,
      "eval_runtime": 9.0929,
      "eval_samples_per_second": 441.115,
      "eval_steps_per_second": 13.857,
      "step": 35300
    },
    {
      "epoch": 280.23928215353936,
      "grad_norm": 586.6846923828125,
      "learning_rate": 2.0330174629387436e-07,
      "loss": 22.5532,
      "step": 35310
    },
    {
      "epoch": 280.3190428713858,
      "grad_norm": 974.9382934570312,
      "learning_rate": 2.0317918267830295e-07,
      "loss": 22.6349,
      "step": 35320
    },
    {
      "epoch": 280.39880358923233,
      "grad_norm": 439.3430480957031,
      "learning_rate": 2.0305663072403934e-07,
      "loss": 22.8622,
      "step": 35330
    },
    {
      "epoch": 280.4785643070788,
      "grad_norm": 869.9444580078125,
      "learning_rate": 2.029340904616067e-07,
      "loss": 24.2681,
      "step": 35340
    },
    {
      "epoch": 280.55832502492524,
      "grad_norm": 1397.7738037109375,
      "learning_rate": 2.0281156192152505e-07,
      "loss": 21.9909,
      "step": 35350
    },
    {
      "epoch": 280.6380857427717,
      "grad_norm": 302.1526184082031,
      "learning_rate": 2.0268904513431169e-07,
      "loss": 22.708,
      "step": 35360
    },
    {
      "epoch": 280.71784646061815,
      "grad_norm": 340.33953857421875,
      "learning_rate": 2.0256654013048091e-07,
      "loss": 22.6771,
      "step": 35370
    },
    {
      "epoch": 280.7976071784646,
      "grad_norm": 488.6512756347656,
      "learning_rate": 2.0244404694054434e-07,
      "loss": 25.2694,
      "step": 35380
    },
    {
      "epoch": 280.87736789631106,
      "grad_norm": 776.7713012695312,
      "learning_rate": 2.0232156559501022e-07,
      "loss": 22.5324,
      "step": 35390
    },
    {
      "epoch": 280.9571286141575,
      "grad_norm": 978.2791137695312,
      "learning_rate": 2.02199096124384e-07,
      "loss": 22.7049,
      "step": 35400
    },
    {
      "epoch": 280.9571286141575,
      "eval_loss": 2.969667434692383,
      "eval_mae": 1.2845220565795898,
      "eval_mse": 2.969667434692383,
      "eval_r2": 0.10525375604629517,
      "eval_rmse": 1.7232723042782248,
      "eval_runtime": 9.0724,
      "eval_samples_per_second": 442.11,
      "eval_steps_per_second": 13.888,
      "step": 35400
    },
    {
      "epoch": 281.03190428713856,
      "grad_norm": 317.99188232421875,
      "learning_rate": 2.0207663855916838e-07,
      "loss": 22.1657,
      "step": 35410
    },
    {
      "epoch": 281.11166500498507,
      "grad_norm": 1002.4553833007812,
      "learning_rate": 2.0195419292986291e-07,
      "loss": 23.7062,
      "step": 35420
    },
    {
      "epoch": 281.1914257228315,
      "grad_norm": 1255.2816162109375,
      "learning_rate": 2.0183175926696425e-07,
      "loss": 22.2218,
      "step": 35430
    },
    {
      "epoch": 281.271186440678,
      "grad_norm": 406.6369323730469,
      "learning_rate": 2.0170933760096583e-07,
      "loss": 22.3458,
      "step": 35440
    },
    {
      "epoch": 281.35094715852443,
      "grad_norm": 1097.8414306640625,
      "learning_rate": 2.0158692796235843e-07,
      "loss": 23.9187,
      "step": 35450
    },
    {
      "epoch": 281.4307078763709,
      "grad_norm": 496.8609313964844,
      "learning_rate": 2.0146453038162974e-07,
      "loss": 24.5831,
      "step": 35460
    },
    {
      "epoch": 281.51046859421734,
      "grad_norm": 425.373291015625,
      "learning_rate": 2.013421448892643e-07,
      "loss": 24.6211,
      "step": 35470
    },
    {
      "epoch": 281.5902293120638,
      "grad_norm": 1004.8867797851562,
      "learning_rate": 2.0121977151574392e-07,
      "loss": 22.9665,
      "step": 35480
    },
    {
      "epoch": 281.66999002991025,
      "grad_norm": 465.9045715332031,
      "learning_rate": 2.0109741029154694e-07,
      "loss": 22.4116,
      "step": 35490
    },
    {
      "epoch": 281.7497507477567,
      "grad_norm": 1068.4315185546875,
      "learning_rate": 2.009750612471492e-07,
      "loss": 23.6084,
      "step": 35500
    },
    {
      "epoch": 281.7497507477567,
      "eval_loss": 2.960203170776367,
      "eval_mae": 1.2882578372955322,
      "eval_mse": 2.960203170776367,
      "eval_r2": 0.10810524225234985,
      "eval_rmse": 1.7205240977028968,
      "eval_runtime": 9.0652,
      "eval_samples_per_second": 442.461,
      "eval_steps_per_second": 13.899,
      "step": 35500
    },
    {
      "epoch": 281.8295114656032,
      "grad_norm": 391.5953674316406,
      "learning_rate": 2.0085272441302304e-07,
      "loss": 23.0571,
      "step": 35510
    },
    {
      "epoch": 281.9092721834497,
      "grad_norm": 420.6859436035156,
      "learning_rate": 2.007303998196382e-07,
      "loss": 21.3119,
      "step": 35520
    },
    {
      "epoch": 281.9890329012961,
      "grad_norm": 542.68115234375,
      "learning_rate": 2.0060808749746085e-07,
      "loss": 22.2611,
      "step": 35530
    },
    {
      "epoch": 282.0638085742772,
      "grad_norm": 252.11354064941406,
      "learning_rate": 2.0048578747695468e-07,
      "loss": 20.2335,
      "step": 35540
    },
    {
      "epoch": 282.14356929212363,
      "grad_norm": 281.56365966796875,
      "learning_rate": 2.0036349978857985e-07,
      "loss": 22.2323,
      "step": 35550
    },
    {
      "epoch": 282.2233300099701,
      "grad_norm": 200.11773681640625,
      "learning_rate": 2.0024122446279373e-07,
      "loss": 24.5286,
      "step": 35560
    },
    {
      "epoch": 282.30309072781654,
      "grad_norm": 246.14796447753906,
      "learning_rate": 2.0011896153005057e-07,
      "loss": 26.2831,
      "step": 35570
    },
    {
      "epoch": 282.382851445663,
      "grad_norm": 1766.3114013671875,
      "learning_rate": 1.999967110208013e-07,
      "loss": 23.6589,
      "step": 35580
    },
    {
      "epoch": 282.46261216350945,
      "grad_norm": 283.08966064453125,
      "learning_rate": 1.9987447296549402e-07,
      "loss": 24.1291,
      "step": 35590
    },
    {
      "epoch": 282.54237288135596,
      "grad_norm": 310.65386962890625,
      "learning_rate": 1.9975224739457367e-07,
      "loss": 20.7151,
      "step": 35600
    },
    {
      "epoch": 282.54237288135596,
      "eval_loss": 2.9648003578186035,
      "eval_mae": 1.284411072731018,
      "eval_mse": 2.9648003578186035,
      "eval_r2": 0.10672014951705933,
      "eval_rmse": 1.721859563907174,
      "eval_runtime": 9.0663,
      "eval_samples_per_second": 442.409,
      "eval_steps_per_second": 13.898,
      "step": 35600
    },
    {
      "epoch": 282.6221335992024,
      "grad_norm": 177.55731201171875,
      "learning_rate": 1.9963003433848208e-07,
      "loss": 22.3219,
      "step": 35610
    },
    {
      "epoch": 282.70189431704887,
      "grad_norm": 1194.229736328125,
      "learning_rate": 1.995078338276578e-07,
      "loss": 21.6743,
      "step": 35620
    },
    {
      "epoch": 282.7816550348953,
      "grad_norm": 941.6058959960938,
      "learning_rate": 1.993856458925365e-07,
      "loss": 23.208,
      "step": 35630
    },
    {
      "epoch": 282.8614157527418,
      "grad_norm": 747.1911010742188,
      "learning_rate": 1.9926347056355054e-07,
      "loss": 24.0022,
      "step": 35640
    },
    {
      "epoch": 282.94117647058823,
      "grad_norm": 341.9564208984375,
      "learning_rate": 1.9914130787112923e-07,
      "loss": 21.1185,
      "step": 35650
    },
    {
      "epoch": 283.0159521435693,
      "grad_norm": 963.5127563476562,
      "learning_rate": 1.9901915784569885e-07,
      "loss": 22.3423,
      "step": 35660
    },
    {
      "epoch": 283.09571286141573,
      "grad_norm": 266.2913513183594,
      "learning_rate": 1.9889702051768216e-07,
      "loss": 23.8738,
      "step": 35670
    },
    {
      "epoch": 283.1754735792622,
      "grad_norm": 421.998291015625,
      "learning_rate": 1.9877489591749907e-07,
      "loss": 21.798,
      "step": 35680
    },
    {
      "epoch": 283.2552342971087,
      "grad_norm": 1275.6529541015625,
      "learning_rate": 1.9865278407556623e-07,
      "loss": 24.3806,
      "step": 35690
    },
    {
      "epoch": 283.33499501495515,
      "grad_norm": 457.0713806152344,
      "learning_rate": 1.985306850222972e-07,
      "loss": 23.3556,
      "step": 35700
    },
    {
      "epoch": 283.33499501495515,
      "eval_loss": 2.9685611724853516,
      "eval_mae": 1.2989680767059326,
      "eval_mse": 2.9685611724853516,
      "eval_r2": 0.10558706521987915,
      "eval_rmse": 1.7229512971890273,
      "eval_runtime": 9.0899,
      "eval_samples_per_second": 441.26,
      "eval_steps_per_second": 13.862,
      "step": 35700
    },
    {
      "epoch": 283.4147557328016,
      "grad_norm": 799.69921875,
      "learning_rate": 1.9840859878810223e-07,
      "loss": 22.4187,
      "step": 35710
    },
    {
      "epoch": 283.49451645064806,
      "grad_norm": 1309.066650390625,
      "learning_rate": 1.982865254033883e-07,
      "loss": 22.0474,
      "step": 35720
    },
    {
      "epoch": 283.5742771684945,
      "grad_norm": 384.88531494140625,
      "learning_rate": 1.981644648985594e-07,
      "loss": 23.8883,
      "step": 35730
    },
    {
      "epoch": 283.654037886341,
      "grad_norm": 886.3473510742188,
      "learning_rate": 1.9804241730401624e-07,
      "loss": 23.0122,
      "step": 35740
    },
    {
      "epoch": 283.7337986041874,
      "grad_norm": 372.2348937988281,
      "learning_rate": 1.9792038265015634e-07,
      "loss": 22.1438,
      "step": 35750
    },
    {
      "epoch": 283.8135593220339,
      "grad_norm": 1198.8360595703125,
      "learning_rate": 1.977983609673738e-07,
      "loss": 24.4856,
      "step": 35760
    },
    {
      "epoch": 283.89332003988034,
      "grad_norm": 403.27880859375,
      "learning_rate": 1.9767635228605966e-07,
      "loss": 22.1644,
      "step": 35770
    },
    {
      "epoch": 283.97308075772685,
      "grad_norm": 347.24029541015625,
      "learning_rate": 1.9755435663660187e-07,
      "loss": 23.0664,
      "step": 35780
    },
    {
      "epoch": 284.0478564307079,
      "grad_norm": 771.271240234375,
      "learning_rate": 1.9743237404938477e-07,
      "loss": 20.9471,
      "step": 35790
    },
    {
      "epoch": 284.12761714855435,
      "grad_norm": 387.5115661621094,
      "learning_rate": 1.9731040455478982e-07,
      "loss": 22.5388,
      "step": 35800
    },
    {
      "epoch": 284.12761714855435,
      "eval_loss": 2.9756717681884766,
      "eval_mae": 1.2824524641036987,
      "eval_mse": 2.9756722450256348,
      "eval_r2": 0.10344451665878296,
      "eval_rmse": 1.7250136941559724,
      "eval_runtime": 9.0561,
      "eval_samples_per_second": 442.906,
      "eval_steps_per_second": 13.913,
      "step": 35800
    },
    {
      "epoch": 284.2073778664008,
      "grad_norm": 687.7453002929688,
      "learning_rate": 1.9718844818319485e-07,
      "loss": 24.1204,
      "step": 35810
    },
    {
      "epoch": 284.28713858424726,
      "grad_norm": 961.2636108398438,
      "learning_rate": 1.970665049649748e-07,
      "loss": 22.2689,
      "step": 35820
    },
    {
      "epoch": 284.3668993020937,
      "grad_norm": 357.94635009765625,
      "learning_rate": 1.96944574930501e-07,
      "loss": 22.8073,
      "step": 35830
    },
    {
      "epoch": 284.44666001994017,
      "grad_norm": 1568.5445556640625,
      "learning_rate": 1.968226581101417e-07,
      "loss": 22.0942,
      "step": 35840
    },
    {
      "epoch": 284.5264207377866,
      "grad_norm": 887.5462036132812,
      "learning_rate": 1.9670075453426192e-07,
      "loss": 25.5706,
      "step": 35850
    },
    {
      "epoch": 284.6061814556331,
      "grad_norm": 520.3413696289062,
      "learning_rate": 1.965788642332231e-07,
      "loss": 23.1674,
      "step": 35860
    },
    {
      "epoch": 284.6859421734796,
      "grad_norm": 478.873779296875,
      "learning_rate": 1.9645698723738353e-07,
      "loss": 23.0428,
      "step": 35870
    },
    {
      "epoch": 284.76570289132604,
      "grad_norm": 605.9752807617188,
      "learning_rate": 1.9633512357709825e-07,
      "loss": 21.5509,
      "step": 35880
    },
    {
      "epoch": 284.8454636091725,
      "grad_norm": 423.40191650390625,
      "learning_rate": 1.9621327328271905e-07,
      "loss": 23.2131,
      "step": 35890
    },
    {
      "epoch": 284.92522432701895,
      "grad_norm": 1224.188232421875,
      "learning_rate": 1.96091436384594e-07,
      "loss": 22.097,
      "step": 35900
    },
    {
      "epoch": 284.92522432701895,
      "eval_loss": 2.964362859725952,
      "eval_mae": 1.2939927577972412,
      "eval_mse": 2.964362621307373,
      "eval_r2": 0.10685199499130249,
      "eval_rmse": 1.7217324476547955,
      "eval_runtime": 9.061,
      "eval_samples_per_second": 442.669,
      "eval_steps_per_second": 13.906,
      "step": 35900
    },
    {
      "epoch": 285.0,
      "grad_norm": 733.3826293945312,
      "learning_rate": 1.9596961291306835e-07,
      "loss": 23.3368,
      "step": 35910
    },
    {
      "epoch": 285.07976071784645,
      "grad_norm": 1493.994140625,
      "learning_rate": 1.9584780289848355e-07,
      "loss": 26.8808,
      "step": 35920
    },
    {
      "epoch": 285.1595214356929,
      "grad_norm": 127.55492401123047,
      "learning_rate": 1.9572600637117805e-07,
      "loss": 23.3882,
      "step": 35930
    },
    {
      "epoch": 285.23928215353936,
      "grad_norm": 853.9454956054688,
      "learning_rate": 1.9560422336148674e-07,
      "loss": 22.2778,
      "step": 35940
    },
    {
      "epoch": 285.3190428713858,
      "grad_norm": 1000.6528930664062,
      "learning_rate": 1.954824538997412e-07,
      "loss": 22.7299,
      "step": 35950
    },
    {
      "epoch": 285.39880358923233,
      "grad_norm": 966.7951049804688,
      "learning_rate": 1.9536069801626956e-07,
      "loss": 22.1884,
      "step": 35960
    },
    {
      "epoch": 285.4785643070788,
      "grad_norm": 700.5227661132812,
      "learning_rate": 1.952389557413967e-07,
      "loss": 25.1173,
      "step": 35970
    },
    {
      "epoch": 285.55832502492524,
      "grad_norm": 1235.1341552734375,
      "learning_rate": 1.9511722710544416e-07,
      "loss": 22.9758,
      "step": 35980
    },
    {
      "epoch": 285.6380857427717,
      "grad_norm": 396.1883544921875,
      "learning_rate": 1.949955121387298e-07,
      "loss": 22.8223,
      "step": 35990
    },
    {
      "epoch": 285.71784646061815,
      "grad_norm": 495.91973876953125,
      "learning_rate": 1.948738108715683e-07,
      "loss": 22.37,
      "step": 36000
    },
    {
      "epoch": 285.71784646061815,
      "eval_loss": 2.9797921180725098,
      "eval_mae": 1.286327838897705,
      "eval_mse": 2.979792356491089,
      "eval_r2": 0.1022031307220459,
      "eval_rmse": 1.7262075067879554,
      "eval_runtime": 9.0814,
      "eval_samples_per_second": 441.673,
      "eval_steps_per_second": 13.875,
      "step": 36000
    },
    {
      "epoch": 285.7976071784646,
      "grad_norm": 313.2388916015625,
      "learning_rate": 1.9475212333427085e-07,
      "loss": 21.9463,
      "step": 36010
    },
    {
      "epoch": 285.87736789631106,
      "grad_norm": 1386.6151123046875,
      "learning_rate": 1.9463044955714535e-07,
      "loss": 21.5289,
      "step": 36020
    },
    {
      "epoch": 285.9571286141575,
      "grad_norm": 923.9249877929688,
      "learning_rate": 1.9450878957049616e-07,
      "loss": 22.8795,
      "step": 36030
    },
    {
      "epoch": 286.03190428713856,
      "grad_norm": 683.2625122070312,
      "learning_rate": 1.9438714340462406e-07,
      "loss": 18.8855,
      "step": 36040
    },
    {
      "epoch": 286.11166500498507,
      "grad_norm": 288.84393310546875,
      "learning_rate": 1.9426551108982663e-07,
      "loss": 25.6166,
      "step": 36050
    },
    {
      "epoch": 286.1914257228315,
      "grad_norm": 405.8291015625,
      "learning_rate": 1.9414389265639802e-07,
      "loss": 23.2314,
      "step": 36060
    },
    {
      "epoch": 286.271186440678,
      "grad_norm": 850.6596069335938,
      "learning_rate": 1.940222881346286e-07,
      "loss": 21.7071,
      "step": 36070
    },
    {
      "epoch": 286.35094715852443,
      "grad_norm": 548.14111328125,
      "learning_rate": 1.939006975548058e-07,
      "loss": 23.1086,
      "step": 36080
    },
    {
      "epoch": 286.4307078763709,
      "grad_norm": 673.26025390625,
      "learning_rate": 1.9377912094721293e-07,
      "loss": 23.034,
      "step": 36090
    },
    {
      "epoch": 286.51046859421734,
      "grad_norm": 414.7325439453125,
      "learning_rate": 1.9365755834213038e-07,
      "loss": 21.6649,
      "step": 36100
    },
    {
      "epoch": 286.51046859421734,
      "eval_loss": 2.957052230834961,
      "eval_mae": 1.289268970489502,
      "eval_mse": 2.957052230834961,
      "eval_r2": 0.10905462503433228,
      "eval_rmse": 1.719608162005217,
      "eval_runtime": 9.0841,
      "eval_samples_per_second": 441.54,
      "eval_steps_per_second": 13.87,
      "step": 36100
    },
    {
      "epoch": 286.5902293120638,
      "grad_norm": 702.4627685546875,
      "learning_rate": 1.9353600976983473e-07,
      "loss": 25.6429,
      "step": 36110
    },
    {
      "epoch": 286.66999002991025,
      "grad_norm": 408.4979248046875,
      "learning_rate": 1.9341447526059928e-07,
      "loss": 24.0296,
      "step": 36120
    },
    {
      "epoch": 286.7497507477567,
      "grad_norm": 440.3926086425781,
      "learning_rate": 1.9329295484469352e-07,
      "loss": 22.6387,
      "step": 36130
    },
    {
      "epoch": 286.8295114656032,
      "grad_norm": 452.042724609375,
      "learning_rate": 1.931714485523838e-07,
      "loss": 22.3395,
      "step": 36140
    },
    {
      "epoch": 286.9092721834497,
      "grad_norm": 333.8067932128906,
      "learning_rate": 1.9304995641393266e-07,
      "loss": 23.0052,
      "step": 36150
    },
    {
      "epoch": 286.9890329012961,
      "grad_norm": 623.7636108398438,
      "learning_rate": 1.929284784595993e-07,
      "loss": 21.9916,
      "step": 36160
    },
    {
      "epoch": 287.0638085742772,
      "grad_norm": 452.311767578125,
      "learning_rate": 1.9280701471963936e-07,
      "loss": 18.6515,
      "step": 36170
    },
    {
      "epoch": 287.14356929212363,
      "grad_norm": 335.5778503417969,
      "learning_rate": 1.9268556522430483e-07,
      "loss": 21.7196,
      "step": 36180
    },
    {
      "epoch": 287.2233300099701,
      "grad_norm": 242.38865661621094,
      "learning_rate": 1.9256413000384415e-07,
      "loss": 22.0252,
      "step": 36190
    },
    {
      "epoch": 287.30309072781654,
      "grad_norm": 641.955322265625,
      "learning_rate": 1.9244270908850234e-07,
      "loss": 22.9962,
      "step": 36200
    },
    {
      "epoch": 287.30309072781654,
      "eval_loss": 2.964320182800293,
      "eval_mae": 1.2858256101608276,
      "eval_mse": 2.964320182800293,
      "eval_r2": 0.1068648099899292,
      "eval_rmse": 1.7217201232489248,
      "eval_runtime": 9.1449,
      "eval_samples_per_second": 438.605,
      "eval_steps_per_second": 13.778,
      "step": 36200
    },
    {
      "epoch": 287.382851445663,
      "grad_norm": 313.9455261230469,
      "learning_rate": 1.9232130250852092e-07,
      "loss": 25.1304,
      "step": 36210
    },
    {
      "epoch": 287.46261216350945,
      "grad_norm": 691.0582275390625,
      "learning_rate": 1.9219991029413754e-07,
      "loss": 22.6834,
      "step": 36220
    },
    {
      "epoch": 287.54237288135596,
      "grad_norm": 430.69793701171875,
      "learning_rate": 1.9207853247558645e-07,
      "loss": 23.7089,
      "step": 36230
    },
    {
      "epoch": 287.6221335992024,
      "grad_norm": 366.560546875,
      "learning_rate": 1.9195716908309833e-07,
      "loss": 23.8384,
      "step": 36240
    },
    {
      "epoch": 287.70189431704887,
      "grad_norm": 492.2204895019531,
      "learning_rate": 1.9183582014690038e-07,
      "loss": 25.214,
      "step": 36250
    },
    {
      "epoch": 287.7816550348953,
      "grad_norm": 1235.56640625,
      "learning_rate": 1.917144856972159e-07,
      "loss": 22.6221,
      "step": 36260
    },
    {
      "epoch": 287.8614157527418,
      "grad_norm": 344.3128662109375,
      "learning_rate": 1.915931657642648e-07,
      "loss": 23.6879,
      "step": 36270
    },
    {
      "epoch": 287.94117647058823,
      "grad_norm": 1383.4229736328125,
      "learning_rate": 1.914718603782633e-07,
      "loss": 23.2404,
      "step": 36280
    },
    {
      "epoch": 288.0159521435693,
      "grad_norm": 192.54356384277344,
      "learning_rate": 1.9135056956942407e-07,
      "loss": 21.8429,
      "step": 36290
    },
    {
      "epoch": 288.09571286141573,
      "grad_norm": 945.7467041015625,
      "learning_rate": 1.91229293367956e-07,
      "loss": 21.5877,
      "step": 36300
    },
    {
      "epoch": 288.09571286141573,
      "eval_loss": 2.9648287296295166,
      "eval_mae": 1.2923362255096436,
      "eval_mse": 2.9648282527923584,
      "eval_r2": 0.106711745262146,
      "eval_rmse": 1.7218676641346042,
      "eval_runtime": 9.063,
      "eval_samples_per_second": 442.567,
      "eval_steps_per_second": 13.903,
      "step": 36300
    },
    {
      "epoch": 288.1754735792622,
      "grad_norm": 1069.1270751953125,
      "learning_rate": 1.9110803180406466e-07,
      "loss": 21.7894,
      "step": 36310
    },
    {
      "epoch": 288.2552342971087,
      "grad_norm": 659.6483764648438,
      "learning_rate": 1.9098678490795142e-07,
      "loss": 23.323,
      "step": 36320
    },
    {
      "epoch": 288.33499501495515,
      "grad_norm": 509.91754150390625,
      "learning_rate": 1.9086555270981456e-07,
      "loss": 20.9569,
      "step": 36330
    },
    {
      "epoch": 288.4147557328016,
      "grad_norm": 292.78265380859375,
      "learning_rate": 1.9074433523984842e-07,
      "loss": 23.8257,
      "step": 36340
    },
    {
      "epoch": 288.49451645064806,
      "grad_norm": 514.0445556640625,
      "learning_rate": 1.9062313252824381e-07,
      "loss": 24.2512,
      "step": 36350
    },
    {
      "epoch": 288.5742771684945,
      "grad_norm": 709.8264770507812,
      "learning_rate": 1.9050194460518757e-07,
      "loss": 23.5609,
      "step": 36360
    },
    {
      "epoch": 288.654037886341,
      "grad_norm": 1047.640625,
      "learning_rate": 1.9038077150086316e-07,
      "loss": 23.5893,
      "step": 36370
    },
    {
      "epoch": 288.7337986041874,
      "grad_norm": 397.6839294433594,
      "learning_rate": 1.902596132454503e-07,
      "loss": 24.5942,
      "step": 36380
    },
    {
      "epoch": 288.8135593220339,
      "grad_norm": 362.4794921875,
      "learning_rate": 1.901384698691249e-07,
      "loss": 22.3205,
      "step": 36390
    },
    {
      "epoch": 288.89332003988034,
      "grad_norm": 250.99215698242188,
      "learning_rate": 1.9001734140205935e-07,
      "loss": 23.9761,
      "step": 36400
    },
    {
      "epoch": 288.89332003988034,
      "eval_loss": 2.959632396697998,
      "eval_mae": 1.2892621755599976,
      "eval_mse": 2.959632396697998,
      "eval_r2": 0.10827720165252686,
      "eval_rmse": 1.7203582175517975,
      "eval_runtime": 9.0867,
      "eval_samples_per_second": 441.415,
      "eval_steps_per_second": 13.866,
      "step": 36400
    },
    {
      "epoch": 288.97308075772685,
      "grad_norm": 458.8247375488281,
      "learning_rate": 1.89896227874422e-07,
      "loss": 22.2856,
      "step": 36410
    },
    {
      "epoch": 289.0478564307079,
      "grad_norm": 227.16421508789062,
      "learning_rate": 1.8977512931637785e-07,
      "loss": 21.3357,
      "step": 36420
    },
    {
      "epoch": 289.12761714855435,
      "grad_norm": 136.99842834472656,
      "learning_rate": 1.8965404575808789e-07,
      "loss": 24.7046,
      "step": 36430
    },
    {
      "epoch": 289.2073778664008,
      "grad_norm": 292.46783447265625,
      "learning_rate": 1.8953297722970954e-07,
      "loss": 24.0002,
      "step": 36440
    },
    {
      "epoch": 289.28713858424726,
      "grad_norm": 437.4063415527344,
      "learning_rate": 1.8941192376139653e-07,
      "loss": 23.0407,
      "step": 36450
    },
    {
      "epoch": 289.3668993020937,
      "grad_norm": 219.44361877441406,
      "learning_rate": 1.8929088538329856e-07,
      "loss": 22.9118,
      "step": 36460
    },
    {
      "epoch": 289.44666001994017,
      "grad_norm": 897.8577880859375,
      "learning_rate": 1.891698621255618e-07,
      "loss": 22.9397,
      "step": 36470
    },
    {
      "epoch": 289.5264207377866,
      "grad_norm": 469.3039245605469,
      "learning_rate": 1.890488540183286e-07,
      "loss": 22.5491,
      "step": 36480
    },
    {
      "epoch": 289.6061814556331,
      "grad_norm": 524.674560546875,
      "learning_rate": 1.8892786109173767e-07,
      "loss": 21.8458,
      "step": 36490
    },
    {
      "epoch": 289.6859421734796,
      "grad_norm": 186.74449157714844,
      "learning_rate": 1.8880688337592363e-07,
      "loss": 22.8131,
      "step": 36500
    },
    {
      "epoch": 289.6859421734796,
      "eval_loss": 2.96305775642395,
      "eval_mae": 1.2956712245941162,
      "eval_mse": 2.96305775642395,
      "eval_r2": 0.10724520683288574,
      "eval_rmse": 1.721353466439694,
      "eval_runtime": 9.1063,
      "eval_samples_per_second": 440.463,
      "eval_steps_per_second": 13.837,
      "step": 36500
    },
    {
      "epoch": 289.76570289132604,
      "grad_norm": 594.638671875,
      "learning_rate": 1.8868592090101748e-07,
      "loss": 21.3693,
      "step": 36510
    },
    {
      "epoch": 289.8454636091725,
      "grad_norm": 1204.589599609375,
      "learning_rate": 1.8856497369714653e-07,
      "loss": 21.8328,
      "step": 36520
    },
    {
      "epoch": 289.92522432701895,
      "grad_norm": 1412.10498046875,
      "learning_rate": 1.884440417944342e-07,
      "loss": 23.0577,
      "step": 36530
    },
    {
      "epoch": 290.0,
      "grad_norm": 484.0316162109375,
      "learning_rate": 1.8832312522300007e-07,
      "loss": 24.1015,
      "step": 36540
    },
    {
      "epoch": 290.07976071784645,
      "grad_norm": 261.65509033203125,
      "learning_rate": 1.8820222401295976e-07,
      "loss": 25.2079,
      "step": 36550
    },
    {
      "epoch": 290.1595214356929,
      "grad_norm": 623.7129516601562,
      "learning_rate": 1.8808133819442539e-07,
      "loss": 21.8983,
      "step": 36560
    },
    {
      "epoch": 290.23928215353936,
      "grad_norm": 189.4542999267578,
      "learning_rate": 1.8796046779750513e-07,
      "loss": 24.0651,
      "step": 36570
    },
    {
      "epoch": 290.3190428713858,
      "grad_norm": 403.1977844238281,
      "learning_rate": 1.8783961285230312e-07,
      "loss": 20.8615,
      "step": 36580
    },
    {
      "epoch": 290.39880358923233,
      "grad_norm": 887.7024536132812,
      "learning_rate": 1.8771877338891987e-07,
      "loss": 23.1835,
      "step": 36590
    },
    {
      "epoch": 290.4785643070788,
      "grad_norm": 153.1597137451172,
      "learning_rate": 1.8759794943745182e-07,
      "loss": 23.9526,
      "step": 36600
    },
    {
      "epoch": 290.4785643070788,
      "eval_loss": 2.969057559967041,
      "eval_mae": 1.282334804534912,
      "eval_mse": 2.96905779838562,
      "eval_r2": 0.10543739795684814,
      "eval_rmse": 1.723095411863667,
      "eval_runtime": 9.099,
      "eval_samples_per_second": 440.816,
      "eval_steps_per_second": 13.848,
      "step": 36600
    },
    {
      "epoch": 290.55832502492524,
      "grad_norm": 206.9164276123047,
      "learning_rate": 1.8747714102799189e-07,
      "loss": 23.3965,
      "step": 36610
    },
    {
      "epoch": 290.6380857427717,
      "grad_norm": 787.9398193359375,
      "learning_rate": 1.8735634819062873e-07,
      "loss": 23.1207,
      "step": 36620
    },
    {
      "epoch": 290.71784646061815,
      "grad_norm": 193.9875030517578,
      "learning_rate": 1.872355709554475e-07,
      "loss": 22.1797,
      "step": 36630
    },
    {
      "epoch": 290.7976071784646,
      "grad_norm": 546.6417236328125,
      "learning_rate": 1.8711480935252904e-07,
      "loss": 22.8325,
      "step": 36640
    },
    {
      "epoch": 290.87736789631106,
      "grad_norm": 1094.3037109375,
      "learning_rate": 1.8699406341195067e-07,
      "loss": 21.845,
      "step": 36650
    },
    {
      "epoch": 290.9571286141575,
      "grad_norm": 223.6046905517578,
      "learning_rate": 1.868733331637857e-07,
      "loss": 24.3477,
      "step": 36660
    },
    {
      "epoch": 291.03190428713856,
      "grad_norm": 439.85516357421875,
      "learning_rate": 1.8675261863810337e-07,
      "loss": 20.7617,
      "step": 36670
    },
    {
      "epoch": 291.11166500498507,
      "grad_norm": 186.5644989013672,
      "learning_rate": 1.8663191986496937e-07,
      "loss": 23.0272,
      "step": 36680
    },
    {
      "epoch": 291.1914257228315,
      "grad_norm": 2083.707275390625,
      "learning_rate": 1.8651123687444498e-07,
      "loss": 24.9641,
      "step": 36690
    },
    {
      "epoch": 291.271186440678,
      "grad_norm": 237.58786010742188,
      "learning_rate": 1.863905696965879e-07,
      "loss": 21.5397,
      "step": 36700
    },
    {
      "epoch": 291.271186440678,
      "eval_loss": 2.959972858428955,
      "eval_mae": 1.2938432693481445,
      "eval_mse": 2.959972620010376,
      "eval_r2": 0.10817474126815796,
      "eval_rmse": 1.7204570962422678,
      "eval_runtime": 9.0858,
      "eval_samples_per_second": 441.459,
      "eval_steps_per_second": 13.868,
      "step": 36700
    },
    {
      "epoch": 291.35094715852443,
      "grad_norm": 370.0028076171875,
      "learning_rate": 1.8626991836145178e-07,
      "loss": 22.8297,
      "step": 36710
    },
    {
      "epoch": 291.4307078763709,
      "grad_norm": 291.2168884277344,
      "learning_rate": 1.8614928289908646e-07,
      "loss": 22.1641,
      "step": 36720
    },
    {
      "epoch": 291.51046859421734,
      "grad_norm": 1112.2530517578125,
      "learning_rate": 1.8602866333953747e-07,
      "loss": 24.578,
      "step": 36730
    },
    {
      "epoch": 291.5902293120638,
      "grad_norm": 402.6103820800781,
      "learning_rate": 1.8590805971284684e-07,
      "loss": 23.6851,
      "step": 36740
    },
    {
      "epoch": 291.66999002991025,
      "grad_norm": 680.296630859375,
      "learning_rate": 1.8578747204905222e-07,
      "loss": 23.5266,
      "step": 36750
    },
    {
      "epoch": 291.7497507477567,
      "grad_norm": 628.4378662109375,
      "learning_rate": 1.8566690037818758e-07,
      "loss": 23.0505,
      "step": 36760
    },
    {
      "epoch": 291.8295114656032,
      "grad_norm": 839.2183837890625,
      "learning_rate": 1.8554634473028287e-07,
      "loss": 22.2709,
      "step": 36770
    },
    {
      "epoch": 291.9092721834497,
      "grad_norm": 1796.494873046875,
      "learning_rate": 1.8542580513536382e-07,
      "loss": 22.0434,
      "step": 36780
    },
    {
      "epoch": 291.9890329012961,
      "grad_norm": 1286.6932373046875,
      "learning_rate": 1.8530528162345234e-07,
      "loss": 23.1104,
      "step": 36790
    },
    {
      "epoch": 292.0638085742772,
      "grad_norm": 1134.0379638671875,
      "learning_rate": 1.8518477422456636e-07,
      "loss": 22.2488,
      "step": 36800
    },
    {
      "epoch": 292.0638085742772,
      "eval_loss": 2.9605071544647217,
      "eval_mae": 1.287654995918274,
      "eval_mse": 2.9605071544647217,
      "eval_r2": 0.10801368951797485,
      "eval_rmse": 1.7206124358683224,
      "eval_runtime": 9.1038,
      "eval_samples_per_second": 440.586,
      "eval_steps_per_second": 13.84,
      "step": 36800
    },
    {
      "epoch": 292.14356929212363,
      "grad_norm": 1212.5570068359375,
      "learning_rate": 1.8506428296871978e-07,
      "loss": 23.1092,
      "step": 36810
    },
    {
      "epoch": 292.2233300099701,
      "grad_norm": 795.9725341796875,
      "learning_rate": 1.8494380788592249e-07,
      "loss": 23.9108,
      "step": 36820
    },
    {
      "epoch": 292.30309072781654,
      "grad_norm": 618.9758911132812,
      "learning_rate": 1.8482334900618007e-07,
      "loss": 22.1557,
      "step": 36830
    },
    {
      "epoch": 292.382851445663,
      "grad_norm": 199.36758422851562,
      "learning_rate": 1.8470290635949448e-07,
      "loss": 21.3498,
      "step": 36840
    },
    {
      "epoch": 292.46261216350945,
      "grad_norm": 897.5830688476562,
      "learning_rate": 1.845824799758635e-07,
      "loss": 23.5089,
      "step": 36850
    },
    {
      "epoch": 292.54237288135596,
      "grad_norm": 375.9431457519531,
      "learning_rate": 1.8446206988528084e-07,
      "loss": 21.8539,
      "step": 36860
    },
    {
      "epoch": 292.6221335992024,
      "grad_norm": 280.3815002441406,
      "learning_rate": 1.8434167611773593e-07,
      "loss": 23.5507,
      "step": 36870
    },
    {
      "epoch": 292.70189431704887,
      "grad_norm": 555.9196166992188,
      "learning_rate": 1.842212987032145e-07,
      "loss": 22.8814,
      "step": 36880
    },
    {
      "epoch": 292.7816550348953,
      "grad_norm": 351.2213439941406,
      "learning_rate": 1.8410093767169805e-07,
      "loss": 23.9657,
      "step": 36890
    },
    {
      "epoch": 292.8614157527418,
      "grad_norm": 439.0758361816406,
      "learning_rate": 1.8398059305316388e-07,
      "loss": 21.1606,
      "step": 36900
    },
    {
      "epoch": 292.8614157527418,
      "eval_loss": 2.956258535385132,
      "eval_mae": 1.2853070497512817,
      "eval_mse": 2.956258535385132,
      "eval_r2": 0.10929375886917114,
      "eval_rmse": 1.719377368521853,
      "eval_runtime": 9.0696,
      "eval_samples_per_second": 442.249,
      "eval_steps_per_second": 13.893,
      "step": 36900
    },
    {
      "epoch": 292.94117647058823,
      "grad_norm": 355.84674072265625,
      "learning_rate": 1.838602648775855e-07,
      "loss": 24.1349,
      "step": 36910
    },
    {
      "epoch": 293.0159521435693,
      "grad_norm": 650.1371459960938,
      "learning_rate": 1.837399531749319e-07,
      "loss": 22.0405,
      "step": 36920
    },
    {
      "epoch": 293.09571286141573,
      "grad_norm": 197.20037841796875,
      "learning_rate": 1.8361965797516844e-07,
      "loss": 23.2535,
      "step": 36930
    },
    {
      "epoch": 293.1754735792622,
      "grad_norm": 447.2896423339844,
      "learning_rate": 1.8349937930825598e-07,
      "loss": 23.4124,
      "step": 36940
    },
    {
      "epoch": 293.2552342971087,
      "grad_norm": 458.4688415527344,
      "learning_rate": 1.8337911720415155e-07,
      "loss": 21.675,
      "step": 36950
    },
    {
      "epoch": 293.33499501495515,
      "grad_norm": 777.9290771484375,
      "learning_rate": 1.832588716928078e-07,
      "loss": 21.99,
      "step": 36960
    },
    {
      "epoch": 293.4147557328016,
      "grad_norm": 377.29107666015625,
      "learning_rate": 1.8313864280417338e-07,
      "loss": 23.5075,
      "step": 36970
    },
    {
      "epoch": 293.49451645064806,
      "grad_norm": 608.5870971679688,
      "learning_rate": 1.8301843056819288e-07,
      "loss": 23.8374,
      "step": 36980
    },
    {
      "epoch": 293.5742771684945,
      "grad_norm": 262.888427734375,
      "learning_rate": 1.8289823501480662e-07,
      "loss": 24.1258,
      "step": 36990
    },
    {
      "epoch": 293.654037886341,
      "grad_norm": 1055.6136474609375,
      "learning_rate": 1.8277805617395086e-07,
      "loss": 24.0822,
      "step": 37000
    },
    {
      "epoch": 293.654037886341,
      "eval_loss": 2.9568257331848145,
      "eval_mae": 1.2914390563964844,
      "eval_mse": 2.9568257331848145,
      "eval_r2": 0.1091228723526001,
      "eval_rmse": 1.7195423034007666,
      "eval_runtime": 9.1967,
      "eval_samples_per_second": 436.135,
      "eval_steps_per_second": 13.701,
      "step": 37000
    },
    {
      "epoch": 293.7337986041874,
      "grad_norm": 379.95611572265625,
      "learning_rate": 1.8265789407555746e-07,
      "loss": 22.5124,
      "step": 37010
    },
    {
      "epoch": 293.8135593220339,
      "grad_norm": 483.12371826171875,
      "learning_rate": 1.8253774874955447e-07,
      "loss": 22.6095,
      "step": 37020
    },
    {
      "epoch": 293.89332003988034,
      "grad_norm": 291.4644775390625,
      "learning_rate": 1.8241762022586542e-07,
      "loss": 22.6208,
      "step": 37030
    },
    {
      "epoch": 293.97308075772685,
      "grad_norm": 459.98358154296875,
      "learning_rate": 1.8229750853440995e-07,
      "loss": 21.8293,
      "step": 37040
    },
    {
      "epoch": 294.0478564307079,
      "grad_norm": 734.4721069335938,
      "learning_rate": 1.821774137051034e-07,
      "loss": 22.4051,
      "step": 37050
    },
    {
      "epoch": 294.12761714855435,
      "grad_norm": 1470.17333984375,
      "learning_rate": 1.8205733576785675e-07,
      "loss": 22.1222,
      "step": 37060
    },
    {
      "epoch": 294.2073778664008,
      "grad_norm": 166.06588745117188,
      "learning_rate": 1.8193727475257695e-07,
      "loss": 23.2323,
      "step": 37070
    },
    {
      "epoch": 294.28713858424726,
      "grad_norm": 1280.9061279296875,
      "learning_rate": 1.8181723068916669e-07,
      "loss": 22.5115,
      "step": 37080
    },
    {
      "epoch": 294.3668993020937,
      "grad_norm": 253.1221466064453,
      "learning_rate": 1.8169720360752455e-07,
      "loss": 22.1583,
      "step": 37090
    },
    {
      "epoch": 294.44666001994017,
      "grad_norm": 348.7823791503906,
      "learning_rate": 1.8157719353754463e-07,
      "loss": 22.086,
      "step": 37100
    },
    {
      "epoch": 294.44666001994017,
      "eval_loss": 2.9736855030059814,
      "eval_mae": 1.2832196950912476,
      "eval_mse": 2.9736857414245605,
      "eval_r2": 0.10404306650161743,
      "eval_rmse": 1.7244378044523845,
      "eval_runtime": 9.0936,
      "eval_samples_per_second": 441.078,
      "eval_steps_per_second": 13.856,
      "step": 37100
    },
    {
      "epoch": 294.5264207377866,
      "grad_norm": 561.4784545898438,
      "learning_rate": 1.8145720050911693e-07,
      "loss": 23.6026,
      "step": 37110
    },
    {
      "epoch": 294.6061814556331,
      "grad_norm": 459.48199462890625,
      "learning_rate": 1.8133722455212725e-07,
      "loss": 24.0519,
      "step": 37120
    },
    {
      "epoch": 294.6859421734796,
      "grad_norm": 614.2304077148438,
      "learning_rate": 1.8121726569645713e-07,
      "loss": 23.5927,
      "step": 37130
    },
    {
      "epoch": 294.76570289132604,
      "grad_norm": 1104.4097900390625,
      "learning_rate": 1.810973239719839e-07,
      "loss": 20.4572,
      "step": 37140
    },
    {
      "epoch": 294.8454636091725,
      "grad_norm": 468.27069091796875,
      "learning_rate": 1.8097739940858028e-07,
      "loss": 23.6405,
      "step": 37150
    },
    {
      "epoch": 294.92522432701895,
      "grad_norm": 1032.1700439453125,
      "learning_rate": 1.8085749203611513e-07,
      "loss": 23.5576,
      "step": 37160
    },
    {
      "epoch": 295.0,
      "grad_norm": 501.94989013671875,
      "learning_rate": 1.8073760188445295e-07,
      "loss": 23.5079,
      "step": 37170
    },
    {
      "epoch": 295.07976071784645,
      "grad_norm": 586.5398559570312,
      "learning_rate": 1.8061772898345385e-07,
      "loss": 22.7941,
      "step": 37180
    },
    {
      "epoch": 295.1595214356929,
      "grad_norm": 555.3281860351562,
      "learning_rate": 1.8049787336297352e-07,
      "loss": 22.9314,
      "step": 37190
    },
    {
      "epoch": 295.23928215353936,
      "grad_norm": 343.9189147949219,
      "learning_rate": 1.8037803505286353e-07,
      "loss": 22.1603,
      "step": 37200
    },
    {
      "epoch": 295.23928215353936,
      "eval_loss": 2.9602396488189697,
      "eval_mae": 1.2886160612106323,
      "eval_mse": 2.9602396488189697,
      "eval_r2": 0.10809427499771118,
      "eval_rmse": 1.7205346985222267,
      "eval_runtime": 9.1032,
      "eval_samples_per_second": 440.612,
      "eval_steps_per_second": 13.841,
      "step": 37200
    },
    {
      "epoch": 295.3190428713858,
      "grad_norm": 389.4377746582031,
      "learning_rate": 1.8025821408297125e-07,
      "loss": 24.6754,
      "step": 37210
    },
    {
      "epoch": 295.39880358923233,
      "grad_norm": 999.9193725585938,
      "learning_rate": 1.801384104831395e-07,
      "loss": 23.0935,
      "step": 37220
    },
    {
      "epoch": 295.4785643070788,
      "grad_norm": 230.59426879882812,
      "learning_rate": 1.800186242832069e-07,
      "loss": 22.4545,
      "step": 37230
    },
    {
      "epoch": 295.55832502492524,
      "grad_norm": 1362.988037109375,
      "learning_rate": 1.798988555130076e-07,
      "loss": 23.5041,
      "step": 37240
    },
    {
      "epoch": 295.6380857427717,
      "grad_norm": 479.1122131347656,
      "learning_rate": 1.7977910420237157e-07,
      "loss": 22.2714,
      "step": 37250
    },
    {
      "epoch": 295.71784646061815,
      "grad_norm": 301.0849304199219,
      "learning_rate": 1.7965937038112433e-07,
      "loss": 23.3649,
      "step": 37260
    },
    {
      "epoch": 295.7976071784646,
      "grad_norm": 782.4691162109375,
      "learning_rate": 1.7953965407908712e-07,
      "loss": 23.3914,
      "step": 37270
    },
    {
      "epoch": 295.87736789631106,
      "grad_norm": 455.13458251953125,
      "learning_rate": 1.7941995532607684e-07,
      "loss": 24.1414,
      "step": 37280
    },
    {
      "epoch": 295.9571286141575,
      "grad_norm": 1449.0589599609375,
      "learning_rate": 1.7930027415190584e-07,
      "loss": 21.9694,
      "step": 37290
    },
    {
      "epoch": 296.03190428713856,
      "grad_norm": 803.688232421875,
      "learning_rate": 1.7918061058638218e-07,
      "loss": 19.707,
      "step": 37300
    },
    {
      "epoch": 296.03190428713856,
      "eval_loss": 2.9595704078674316,
      "eval_mae": 1.285779356956482,
      "eval_mse": 2.9595704078674316,
      "eval_r2": 0.10829591751098633,
      "eval_rmse": 1.7203402012007485,
      "eval_runtime": 9.0877,
      "eval_samples_per_second": 441.365,
      "eval_steps_per_second": 13.865,
      "step": 37300
    },
    {
      "epoch": 296.11166500498507,
      "grad_norm": 189.18777465820312,
      "learning_rate": 1.790609646593096e-07,
      "loss": 24.1162,
      "step": 37310
    },
    {
      "epoch": 296.1914257228315,
      "grad_norm": 689.5313110351562,
      "learning_rate": 1.789413364004876e-07,
      "loss": 23.4291,
      "step": 37320
    },
    {
      "epoch": 296.271186440678,
      "grad_norm": 630.0205688476562,
      "learning_rate": 1.788217258397108e-07,
      "loss": 22.7952,
      "step": 37330
    },
    {
      "epoch": 296.35094715852443,
      "grad_norm": 521.653564453125,
      "learning_rate": 1.7870213300676985e-07,
      "loss": 22.0444,
      "step": 37340
    },
    {
      "epoch": 296.4307078763709,
      "grad_norm": 375.9777526855469,
      "learning_rate": 1.7858255793145075e-07,
      "loss": 22.5383,
      "step": 37350
    },
    {
      "epoch": 296.51046859421734,
      "grad_norm": 742.4426879882812,
      "learning_rate": 1.7846300064353522e-07,
      "loss": 23.0361,
      "step": 37360
    },
    {
      "epoch": 296.5902293120638,
      "grad_norm": 624.1666870117188,
      "learning_rate": 1.7835541431725244e-07,
      "loss": 23.834,
      "step": 37370
    },
    {
      "epoch": 296.66999002991025,
      "grad_norm": 500.9219055175781,
      "learning_rate": 1.782358909074363e-07,
      "loss": 25.7155,
      "step": 37380
    },
    {
      "epoch": 296.7497507477567,
      "grad_norm": 322.7745361328125,
      "learning_rate": 1.7811638537136547e-07,
      "loss": 23.0646,
      "step": 37390
    },
    {
      "epoch": 296.8295114656032,
      "grad_norm": 221.01759338378906,
      "learning_rate": 1.779968977388044e-07,
      "loss": 22.1865,
      "step": 37400
    },
    {
      "epoch": 296.8295114656032,
      "eval_loss": 2.9581120014190674,
      "eval_mae": 1.294386863708496,
      "eval_mse": 2.9581120014190674,
      "eval_r2": 0.10873532295227051,
      "eval_rmse": 1.7199162774446515,
      "eval_runtime": 9.1113,
      "eval_samples_per_second": 440.224,
      "eval_steps_per_second": 13.829,
      "step": 37400
    },
    {
      "epoch": 296.9092721834497,
      "grad_norm": 1115.1885986328125,
      "learning_rate": 1.7787742803951275e-07,
      "loss": 23.3881,
      "step": 37410
    },
    {
      "epoch": 296.9890329012961,
      "grad_norm": 332.14593505859375,
      "learning_rate": 1.7775797630324617e-07,
      "loss": 21.0627,
      "step": 37420
    },
    {
      "epoch": 297.0638085742772,
      "grad_norm": 678.20703125,
      "learning_rate": 1.7763854255975542e-07,
      "loss": 20.6727,
      "step": 37430
    },
    {
      "epoch": 297.14356929212363,
      "grad_norm": 931.6592407226562,
      "learning_rate": 1.77519126838787e-07,
      "loss": 25.005,
      "step": 37440
    },
    {
      "epoch": 297.2233300099701,
      "grad_norm": 253.5115966796875,
      "learning_rate": 1.7739972917008296e-07,
      "loss": 22.7759,
      "step": 37450
    },
    {
      "epoch": 297.30309072781654,
      "grad_norm": 328.89422607421875,
      "learning_rate": 1.772803495833806e-07,
      "loss": 23.3569,
      "step": 37460
    },
    {
      "epoch": 297.382851445663,
      "grad_norm": 252.57533264160156,
      "learning_rate": 1.7716098810841325e-07,
      "loss": 22.2182,
      "step": 37470
    },
    {
      "epoch": 297.46261216350945,
      "grad_norm": 1059.1834716796875,
      "learning_rate": 1.7704164477490903e-07,
      "loss": 22.699,
      "step": 37480
    },
    {
      "epoch": 297.54237288135596,
      "grad_norm": 206.7607879638672,
      "learning_rate": 1.7692231961259193e-07,
      "loss": 23.1777,
      "step": 37490
    },
    {
      "epoch": 297.6221335992024,
      "grad_norm": 553.6284790039062,
      "learning_rate": 1.7680301265118148e-07,
      "loss": 23.6693,
      "step": 37500
    },
    {
      "epoch": 297.6221335992024,
      "eval_loss": 2.956035614013672,
      "eval_mae": 1.2902369499206543,
      "eval_mse": 2.9560353755950928,
      "eval_r2": 0.10936099290847778,
      "eval_rmse": 1.7193124717732646,
      "eval_runtime": 9.0791,
      "eval_samples_per_second": 441.785,
      "eval_steps_per_second": 13.878,
      "step": 37500
    }
  ],
  "logging_steps": 10,
  "max_steps": 63000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 500,
  "save_steps": 100,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 20,
        "early_stopping_threshold": 0.0
      },
      "attributes": {
        "early_stopping_patience_counter": 0
      }
    },
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 1.6224599394638256e+17,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}