{
  "best_global_step": 24000,
  "best_metric": 0.15508916974067688,
  "best_model_checkpoint": "./patchtst_tsmixup_final/checkpoint-24000",
  "epoch": 2.285292577250869,
  "eval_steps": 1000,
  "global_step": 24000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.004761224586963767,
      "grad_norm": 2.066147804260254,
      "learning_rate": 4.9000000000000005e-06,
      "loss": 0.5158,
      "step": 50
    },
    {
      "epoch": 0.009522449173927534,
      "grad_norm": 1.1728284358978271,
      "learning_rate": 9.900000000000002e-06,
      "loss": 0.4068,
      "step": 100
    },
    {
      "epoch": 0.014283673760891302,
      "grad_norm": 1.0786150693893433,
      "learning_rate": 1.49e-05,
      "loss": 0.3107,
      "step": 150
    },
    {
      "epoch": 0.01904489834785507,
      "grad_norm": 1.0987305641174316,
      "learning_rate": 1.9900000000000003e-05,
      "loss": 0.2444,
      "step": 200
    },
    {
      "epoch": 0.023806122934818836,
      "grad_norm": 0.9221014976501465,
      "learning_rate": 2.4900000000000002e-05,
      "loss": 0.2106,
      "step": 250
    },
    {
      "epoch": 0.028567347521782603,
      "grad_norm": 1.1374775171279907,
      "learning_rate": 2.9900000000000002e-05,
      "loss": 0.1931,
      "step": 300
    },
    {
      "epoch": 0.03332857210874637,
      "grad_norm": 1.2504770755767822,
      "learning_rate": 3.49e-05,
      "loss": 0.1971,
      "step": 350
    },
    {
      "epoch": 0.03808979669571014,
      "grad_norm": 0.8367598056793213,
      "learning_rate": 3.99e-05,
      "loss": 0.1918,
      "step": 400
    },
    {
      "epoch": 0.0428510212826739,
      "grad_norm": 0.9541486501693726,
      "learning_rate": 4.49e-05,
      "loss": 0.192,
      "step": 450
    },
    {
      "epoch": 0.04761224586963767,
      "grad_norm": 0.9134742021560669,
      "learning_rate": 4.99e-05,
      "loss": 0.1875,
      "step": 500
    },
    {
      "epoch": 0.052373470456601436,
      "grad_norm": 2.1370866298675537,
      "learning_rate": 5.4900000000000006e-05,
      "loss": 0.19,
      "step": 550
    },
    {
      "epoch": 0.057134695043565206,
      "grad_norm": 1.8110992908477783,
      "learning_rate": 5.99e-05,
      "loss": 0.1902,
      "step": 600
    },
    {
      "epoch": 0.06189591963052897,
      "grad_norm": 0.8074783682823181,
      "learning_rate": 6.49e-05,
      "loss": 0.1867,
      "step": 650
    },
    {
      "epoch": 0.06665714421749273,
      "grad_norm": 1.8686498403549194,
      "learning_rate": 6.99e-05,
      "loss": 0.1839,
      "step": 700
    },
    {
      "epoch": 0.07141836880445651,
      "grad_norm": 0.9226100444793701,
      "learning_rate": 7.49e-05,
      "loss": 0.1832,
      "step": 750
    },
    {
      "epoch": 0.07617959339142027,
      "grad_norm": 1.0714315176010132,
      "learning_rate": 7.99e-05,
      "loss": 0.1862,
      "step": 800
    },
    {
      "epoch": 0.08094081797838404,
      "grad_norm": 0.7746614813804626,
      "learning_rate": 8.49e-05,
      "loss": 0.1836,
      "step": 850
    },
    {
      "epoch": 0.0857020425653478,
      "grad_norm": 0.8154539465904236,
      "learning_rate": 8.99e-05,
      "loss": 0.1863,
      "step": 900
    },
    {
      "epoch": 0.09046326715231158,
      "grad_norm": 0.7890446186065674,
      "learning_rate": 9.49e-05,
      "loss": 0.1865,
      "step": 950
    },
    {
      "epoch": 0.09522449173927534,
      "grad_norm": 0.9673619270324707,
      "learning_rate": 9.99e-05,
      "loss": 0.1797,
      "step": 1000
    },
    {
      "epoch": 0.09522449173927534,
      "eval_loss": 0.17560431361198425,
      "eval_mae": 0.7397370934486389,
      "eval_mse": 447.359619140625,
      "eval_rmse": 21.15087750285139,
      "eval_runtime": 58.709,
      "eval_samples_per_second": 10175.863,
      "eval_smape": 90.89710712432861,
      "eval_steps_per_second": 19.878,
      "step": 1000
    },
    {
      "epoch": 0.09998571632623911,
      "grad_norm": 0.8122360110282898,
      "learning_rate": 9.999532932990182e-05,
      "loss": 0.1813,
      "step": 1050
    },
    {
      "epoch": 0.10474694091320287,
      "grad_norm": 1.058344841003418,
      "learning_rate": 9.999056334000573e-05,
      "loss": 0.1797,
      "step": 1100
    },
    {
      "epoch": 0.10950816550016665,
      "grad_norm": 0.9517110586166382,
      "learning_rate": 9.998579735010962e-05,
      "loss": 0.1816,
      "step": 1150
    },
    {
      "epoch": 0.11426939008713041,
      "grad_norm": 1.0868945121765137,
      "learning_rate": 9.998103136021352e-05,
      "loss": 0.1786,
      "step": 1200
    },
    {
      "epoch": 0.11903061467409418,
      "grad_norm": 1.1176084280014038,
      "learning_rate": 9.997626537031743e-05,
      "loss": 0.1766,
      "step": 1250
    },
    {
      "epoch": 0.12379183926105794,
      "grad_norm": 1.489853858947754,
      "learning_rate": 9.997149938042131e-05,
      "loss": 0.1763,
      "step": 1300
    },
    {
      "epoch": 0.1285530638480217,
      "grad_norm": 1.1758699417114258,
      "learning_rate": 9.996673339052521e-05,
      "loss": 0.1808,
      "step": 1350
    },
    {
      "epoch": 0.13331428843498547,
      "grad_norm": 1.3600101470947266,
      "learning_rate": 9.996196740062911e-05,
      "loss": 0.177,
      "step": 1400
    },
    {
      "epoch": 0.13807551302194926,
      "grad_norm": 0.7045785784721375,
      "learning_rate": 9.995720141073301e-05,
      "loss": 0.1781,
      "step": 1450
    },
    {
      "epoch": 0.14283673760891302,
      "grad_norm": 1.386483907699585,
      "learning_rate": 9.995243542083691e-05,
      "loss": 0.1773,
      "step": 1500
    },
    {
      "epoch": 0.1475979621958768,
      "grad_norm": 1.3609524965286255,
      "learning_rate": 9.994766943094081e-05,
      "loss": 0.1777,
      "step": 1550
    },
    {
      "epoch": 0.15235918678284055,
      "grad_norm": 0.7538266777992249,
      "learning_rate": 9.994290344104471e-05,
      "loss": 0.1771,
      "step": 1600
    },
    {
      "epoch": 0.1571204113698043,
      "grad_norm": 0.6669739484786987,
      "learning_rate": 9.993813745114861e-05,
      "loss": 0.1757,
      "step": 1650
    },
    {
      "epoch": 0.16188163595676808,
      "grad_norm": 1.321413278579712,
      "learning_rate": 9.993337146125251e-05,
      "loss": 0.1729,
      "step": 1700
    },
    {
      "epoch": 0.16664286054373184,
      "grad_norm": 0.7625342011451721,
      "learning_rate": 9.992860547135641e-05,
      "loss": 0.1744,
      "step": 1750
    },
    {
      "epoch": 0.1714040851306956,
      "grad_norm": 1.4134427309036255,
      "learning_rate": 9.99238394814603e-05,
      "loss": 0.1726,
      "step": 1800
    },
    {
      "epoch": 0.17616530971765937,
      "grad_norm": 0.8000154495239258,
      "learning_rate": 9.991907349156421e-05,
      "loss": 0.175,
      "step": 1850
    },
    {
      "epoch": 0.18092653430462316,
      "grad_norm": 0.8776112198829651,
      "learning_rate": 9.99143075016681e-05,
      "loss": 0.1742,
      "step": 1900
    },
    {
      "epoch": 0.18568775889158692,
      "grad_norm": 0.9763234853744507,
      "learning_rate": 9.9909541511772e-05,
      "loss": 0.1753,
      "step": 1950
    },
    {
      "epoch": 0.1904489834785507,
      "grad_norm": 0.6149079203605652,
      "learning_rate": 9.99047755218759e-05,
      "loss": 0.1709,
      "step": 2000
    },
    {
      "epoch": 0.1904489834785507,
      "eval_loss": 0.1690738946199417,
      "eval_mae": 0.7152817249298096,
      "eval_mse": 425.0924072265625,
      "eval_rmse": 20.617769210721185,
      "eval_runtime": 60.02,
      "eval_samples_per_second": 9953.597,
      "eval_smape": 112.30494976043701,
      "eval_steps_per_second": 19.444,
      "step": 2000
    },
    {
      "epoch": 0.19521020806551445,
      "grad_norm": 0.8438695669174194,
      "learning_rate": 9.99000095319798e-05,
      "loss": 0.1753,
      "step": 2050
    },
    {
      "epoch": 0.19997143265247821,
      "grad_norm": 0.8969755172729492,
      "learning_rate": 9.98952435420837e-05,
      "loss": 0.1745,
      "step": 2100
    },
    {
      "epoch": 0.20473265723944198,
      "grad_norm": 0.9189475178718567,
      "learning_rate": 9.98904775521876e-05,
      "loss": 0.1709,
      "step": 2150
    },
    {
      "epoch": 0.20949388182640574,
      "grad_norm": 0.8711380362510681,
      "learning_rate": 9.98857115622915e-05,
      "loss": 0.1692,
      "step": 2200
    },
    {
      "epoch": 0.2142551064133695,
      "grad_norm": 0.7816225290298462,
      "learning_rate": 9.98809455723954e-05,
      "loss": 0.1709,
      "step": 2250
    },
    {
      "epoch": 0.2190163310003333,
      "grad_norm": 0.6408753395080566,
      "learning_rate": 9.987617958249928e-05,
      "loss": 0.1707,
      "step": 2300
    },
    {
      "epoch": 0.22377755558729706,
      "grad_norm": 0.7021253705024719,
      "learning_rate": 9.987141359260319e-05,
      "loss": 0.1739,
      "step": 2350
    },
    {
      "epoch": 0.22853878017426082,
      "grad_norm": 0.9026205539703369,
      "learning_rate": 9.986664760270709e-05,
      "loss": 0.1705,
      "step": 2400
    },
    {
      "epoch": 0.2333000047612246,
      "grad_norm": 0.6956352591514587,
      "learning_rate": 9.986188161281098e-05,
      "loss": 0.1705,
      "step": 2450
    },
    {
      "epoch": 0.23806122934818835,
      "grad_norm": 0.7024583220481873,
      "learning_rate": 9.985711562291489e-05,
      "loss": 0.1718,
      "step": 2500
    },
    {
      "epoch": 0.24282245393515212,
      "grad_norm": 0.6184080839157104,
      "learning_rate": 9.985234963301878e-05,
      "loss": 0.1697,
      "step": 2550
    },
    {
      "epoch": 0.24758367852211588,
      "grad_norm": 0.9684680700302124,
      "learning_rate": 9.984758364312268e-05,
      "loss": 0.1696,
      "step": 2600
    },
    {
      "epoch": 0.25234490310907964,
      "grad_norm": 0.8625733852386475,
      "learning_rate": 9.984281765322659e-05,
      "loss": 0.1731,
      "step": 2650
    },
    {
      "epoch": 0.2571061276960434,
      "grad_norm": 0.6156722903251648,
      "learning_rate": 9.983805166333048e-05,
      "loss": 0.1731,
      "step": 2700
    },
    {
      "epoch": 0.26186735228300717,
      "grad_norm": 0.7371954917907715,
      "learning_rate": 9.983328567343438e-05,
      "loss": 0.173,
      "step": 2750
    },
    {
      "epoch": 0.26662857686997093,
      "grad_norm": 0.658812940120697,
      "learning_rate": 9.982851968353828e-05,
      "loss": 0.1691,
      "step": 2800
    },
    {
      "epoch": 0.2713898014569347,
      "grad_norm": 0.8242245316505432,
      "learning_rate": 9.982375369364217e-05,
      "loss": 0.1715,
      "step": 2850
    },
    {
      "epoch": 0.2761510260438985,
      "grad_norm": 0.6063680052757263,
      "learning_rate": 9.981898770374607e-05,
      "loss": 0.1722,
      "step": 2900
    },
    {
      "epoch": 0.2809122506308623,
      "grad_norm": 0.7695409655570984,
      "learning_rate": 9.981422171384997e-05,
      "loss": 0.1712,
      "step": 2950
    },
    {
      "epoch": 0.28567347521782605,
      "grad_norm": 0.665675163269043,
      "learning_rate": 9.980945572395387e-05,
      "loss": 0.1722,
      "step": 3000
    },
    {
      "epoch": 0.28567347521782605,
      "eval_loss": 0.16621138155460358,
      "eval_mae": 0.7008960247039795,
      "eval_mse": 516.21533203125,
      "eval_rmse": 22.72037262087156,
      "eval_runtime": 62.5998,
      "eval_samples_per_second": 9543.397,
      "eval_smape": 89.52359557151794,
      "eval_steps_per_second": 18.642,
      "step": 3000
    },
    {
      "epoch": 0.2904346998047898,
      "grad_norm": 0.7732612490653992,
      "learning_rate": 9.980468973405776e-05,
      "loss": 0.1708,
      "step": 3050
    },
    {
      "epoch": 0.2951959243917536,
      "grad_norm": 1.0461828708648682,
      "learning_rate": 9.979992374416167e-05,
      "loss": 0.1692,
      "step": 3100
    },
    {
      "epoch": 0.29995714897871734,
      "grad_norm": 0.6057863831520081,
      "learning_rate": 9.979515775426557e-05,
      "loss": 0.1681,
      "step": 3150
    },
    {
      "epoch": 0.3047183735656811,
      "grad_norm": 0.5380491018295288,
      "learning_rate": 9.979039176436946e-05,
      "loss": 0.1687,
      "step": 3200
    },
    {
      "epoch": 0.30947959815264486,
      "grad_norm": 0.7164149284362793,
      "learning_rate": 9.978562577447337e-05,
      "loss": 0.1678,
      "step": 3250
    },
    {
      "epoch": 0.3142408227396086,
      "grad_norm": 0.9223781228065491,
      "learning_rate": 9.978085978457726e-05,
      "loss": 0.1698,
      "step": 3300
    },
    {
      "epoch": 0.3190020473265724,
      "grad_norm": 0.645452082157135,
      "learning_rate": 9.977609379468116e-05,
      "loss": 0.17,
      "step": 3350
    },
    {
      "epoch": 0.32376327191353615,
      "grad_norm": 0.6378370523452759,
      "learning_rate": 9.977132780478507e-05,
      "loss": 0.1673,
      "step": 3400
    },
    {
      "epoch": 0.3285244965004999,
      "grad_norm": 0.511216402053833,
      "learning_rate": 9.976656181488896e-05,
      "loss": 0.1678,
      "step": 3450
    },
    {
      "epoch": 0.3332857210874637,
      "grad_norm": 0.64838707447052,
      "learning_rate": 9.976179582499286e-05,
      "loss": 0.1675,
      "step": 3500
    },
    {
      "epoch": 0.33804694567442745,
      "grad_norm": 0.6467918753623962,
      "learning_rate": 9.975702983509676e-05,
      "loss": 0.1699,
      "step": 3550
    },
    {
      "epoch": 0.3428081702613912,
      "grad_norm": 0.6198284029960632,
      "learning_rate": 9.975226384520065e-05,
      "loss": 0.169,
      "step": 3600
    },
    {
      "epoch": 0.347569394848355,
      "grad_norm": 0.6328741312026978,
      "learning_rate": 9.974749785530455e-05,
      "loss": 0.1685,
      "step": 3650
    },
    {
      "epoch": 0.35233061943531874,
      "grad_norm": 0.8264518976211548,
      "learning_rate": 9.974273186540844e-05,
      "loss": 0.1731,
      "step": 3700
    },
    {
      "epoch": 0.35709184402228256,
      "grad_norm": 0.7238495945930481,
      "learning_rate": 9.973796587551235e-05,
      "loss": 0.1674,
      "step": 3750
    },
    {
      "epoch": 0.3618530686092463,
      "grad_norm": 0.6243422031402588,
      "learning_rate": 9.973319988561624e-05,
      "loss": 0.1699,
      "step": 3800
    },
    {
      "epoch": 0.3666142931962101,
      "grad_norm": 0.76638263463974,
      "learning_rate": 9.972843389572014e-05,
      "loss": 0.1707,
      "step": 3850
    },
    {
      "epoch": 0.37137551778317385,
      "grad_norm": 0.5346329212188721,
      "learning_rate": 9.972366790582405e-05,
      "loss": 0.1669,
      "step": 3900
    },
    {
      "epoch": 0.3761367423701376,
      "grad_norm": 0.6198967695236206,
      "learning_rate": 9.971890191592794e-05,
      "loss": 0.1663,
      "step": 3950
    },
    {
      "epoch": 0.3808979669571014,
      "grad_norm": 0.936530590057373,
      "learning_rate": 9.971413592603184e-05,
      "loss": 0.1694,
      "step": 4000
    },
    {
      "epoch": 0.3808979669571014,
      "eval_loss": 0.16426624357700348,
      "eval_mae": 0.6708112359046936,
      "eval_mse": 321.2046813964844,
      "eval_rmse": 17.922184057655596,
      "eval_runtime": 59.1224,
      "eval_samples_per_second": 10104.719,
      "eval_smape": 93.0514931678772,
      "eval_steps_per_second": 19.739,
      "step": 4000
    },
    {
      "epoch": 0.38565919154406514,
      "grad_norm": 0.5151750445365906,
      "learning_rate": 9.970936993613574e-05,
      "loss": 0.1676,
      "step": 4050
    },
    {
      "epoch": 0.3904204161310289,
      "grad_norm": 0.8430535793304443,
      "learning_rate": 9.970460394623964e-05,
      "loss": 0.1699,
      "step": 4100
    },
    {
      "epoch": 0.39518164071799267,
      "grad_norm": 0.7711997628211975,
      "learning_rate": 9.969983795634354e-05,
      "loss": 0.1664,
      "step": 4150
    },
    {
      "epoch": 0.39994286530495643,
      "grad_norm": 0.5547206401824951,
      "learning_rate": 9.969507196644744e-05,
      "loss": 0.1679,
      "step": 4200
    },
    {
      "epoch": 0.4047040898919202,
      "grad_norm": 0.7514538764953613,
      "learning_rate": 9.969030597655134e-05,
      "loss": 0.1685,
      "step": 4250
    },
    {
      "epoch": 0.40946531447888396,
      "grad_norm": 0.6667706370353699,
      "learning_rate": 9.968553998665524e-05,
      "loss": 0.1691,
      "step": 4300
    },
    {
      "epoch": 0.4142265390658477,
      "grad_norm": 0.5886721611022949,
      "learning_rate": 9.968077399675914e-05,
      "loss": 0.1649,
      "step": 4350
    },
    {
      "epoch": 0.4189877636528115,
      "grad_norm": 0.5160133838653564,
      "learning_rate": 9.967600800686303e-05,
      "loss": 0.1673,
      "step": 4400
    },
    {
      "epoch": 0.42374898823977525,
      "grad_norm": 0.6817535758018494,
      "learning_rate": 9.967124201696692e-05,
      "loss": 0.1687,
      "step": 4450
    },
    {
      "epoch": 0.428510212826739,
      "grad_norm": 0.5424938201904297,
      "learning_rate": 9.966647602707083e-05,
      "loss": 0.1687,
      "step": 4500
    },
    {
      "epoch": 0.43327143741370283,
      "grad_norm": 0.483815461397171,
      "learning_rate": 9.966171003717473e-05,
      "loss": 0.1673,
      "step": 4550
    },
    {
      "epoch": 0.4380326620006666,
      "grad_norm": 0.555853009223938,
      "learning_rate": 9.965694404727862e-05,
      "loss": 0.1667,
      "step": 4600
    },
    {
      "epoch": 0.44279388658763036,
      "grad_norm": 0.45629894733428955,
      "learning_rate": 9.965217805738253e-05,
      "loss": 0.1678,
      "step": 4650
    },
    {
      "epoch": 0.4475551111745941,
      "grad_norm": 0.47480854392051697,
      "learning_rate": 9.964741206748642e-05,
      "loss": 0.1651,
      "step": 4700
    },
    {
      "epoch": 0.4523163357615579,
      "grad_norm": 0.5411455631256104,
      "learning_rate": 9.964264607759032e-05,
      "loss": 0.168,
      "step": 4750
    },
    {
      "epoch": 0.45707756034852165,
      "grad_norm": 0.7176097631454468,
      "learning_rate": 9.963788008769422e-05,
      "loss": 0.1654,
      "step": 4800
    },
    {
      "epoch": 0.4618387849354854,
      "grad_norm": 0.7010710835456848,
      "learning_rate": 9.963311409779812e-05,
      "loss": 0.165,
      "step": 4850
    },
    {
      "epoch": 0.4666000095224492,
      "grad_norm": 0.5730286240577698,
      "learning_rate": 9.962834810790202e-05,
      "loss": 0.1654,
      "step": 4900
    },
    {
      "epoch": 0.47136123410941294,
      "grad_norm": 0.532320499420166,
      "learning_rate": 9.96235821180059e-05,
      "loss": 0.1665,
      "step": 4950
    },
    {
      "epoch": 0.4761224586963767,
      "grad_norm": 0.5974966883659363,
      "learning_rate": 9.961881612810982e-05,
      "loss": 0.1648,
      "step": 5000
    },
    {
      "epoch": 0.4761224586963767,
      "eval_loss": 0.16261516511440277,
      "eval_mae": 0.6730566620826721,
      "eval_mse": 350.6869812011719,
      "eval_rmse": 18.726638278163325,
      "eval_runtime": 61.1422,
      "eval_samples_per_second": 9770.909,
      "eval_smape": 94.07484531402588,
      "eval_steps_per_second": 19.087,
      "step": 5000
    },
    {
      "epoch": 0.48088368328334047,
      "grad_norm": 0.6836587190628052,
      "learning_rate": 9.961405013821372e-05,
      "loss": 0.164,
      "step": 5050
    },
    {
      "epoch": 0.48564490787030423,
      "grad_norm": 0.6935145854949951,
      "learning_rate": 9.96092841483176e-05,
      "loss": 0.1656,
      "step": 5100
    },
    {
      "epoch": 0.490406132457268,
      "grad_norm": 0.5300805568695068,
      "learning_rate": 9.960451815842151e-05,
      "loss": 0.1661,
      "step": 5150
    },
    {
      "epoch": 0.49516735704423176,
      "grad_norm": 0.6059597134590149,
      "learning_rate": 9.95997521685254e-05,
      "loss": 0.1665,
      "step": 5200
    },
    {
      "epoch": 0.4999285816311955,
      "grad_norm": 0.6202102303504944,
      "learning_rate": 9.95949861786293e-05,
      "loss": 0.1667,
      "step": 5250
    },
    {
      "epoch": 0.5046898062181593,
      "grad_norm": 0.6857314705848694,
      "learning_rate": 9.959022018873321e-05,
      "loss": 0.1648,
      "step": 5300
    },
    {
      "epoch": 0.5094510308051231,
      "grad_norm": 0.5026215314865112,
      "learning_rate": 9.95854541988371e-05,
      "loss": 0.1656,
      "step": 5350
    },
    {
      "epoch": 0.5142122553920868,
      "grad_norm": 0.8072870969772339,
      "learning_rate": 9.9580688208941e-05,
      "loss": 0.1637,
      "step": 5400
    },
    {
      "epoch": 0.5189734799790506,
      "grad_norm": 0.5563872456550598,
      "learning_rate": 9.95759222190449e-05,
      "loss": 0.1665,
      "step": 5450
    },
    {
      "epoch": 0.5237347045660143,
      "grad_norm": 0.4486568868160248,
      "learning_rate": 9.95711562291488e-05,
      "loss": 0.1665,
      "step": 5500
    },
    {
      "epoch": 0.5284959291529782,
      "grad_norm": 0.5072858929634094,
      "learning_rate": 9.95663902392527e-05,
      "loss": 0.1671,
      "step": 5550
    },
    {
      "epoch": 0.5332571537399419,
      "grad_norm": 0.4768078327178955,
      "learning_rate": 9.95616242493566e-05,
      "loss": 0.165,
      "step": 5600
    },
    {
      "epoch": 0.5380183783269057,
      "grad_norm": 0.5484294891357422,
      "learning_rate": 9.95568582594605e-05,
      "loss": 0.1615,
      "step": 5650
    },
    {
      "epoch": 0.5427796029138694,
      "grad_norm": 0.5098631978034973,
      "learning_rate": 9.955209226956438e-05,
      "loss": 0.1663,
      "step": 5700
    },
    {
      "epoch": 0.5475408275008332,
      "grad_norm": 0.5663777589797974,
      "learning_rate": 9.95473262796683e-05,
      "loss": 0.1653,
      "step": 5750
    },
    {
      "epoch": 0.552302052087797,
      "grad_norm": 0.5557841658592224,
      "learning_rate": 9.95425602897722e-05,
      "loss": 0.163,
      "step": 5800
    },
    {
      "epoch": 0.5570632766747607,
      "grad_norm": 0.545656144618988,
      "learning_rate": 9.953779429987608e-05,
      "loss": 0.1619,
      "step": 5850
    },
    {
      "epoch": 0.5618245012617246,
      "grad_norm": 0.6774228811264038,
      "learning_rate": 9.953302830998e-05,
      "loss": 0.1649,
      "step": 5900
    },
    {
      "epoch": 0.5665857258486883,
      "grad_norm": 0.4831783175468445,
      "learning_rate": 9.952826232008388e-05,
      "loss": 0.1626,
      "step": 5950
    },
    {
      "epoch": 0.5713469504356521,
      "grad_norm": 0.46657130122184753,
      "learning_rate": 9.952349633018778e-05,
      "loss": 0.1672,
      "step": 6000
    },
    {
      "epoch": 0.5713469504356521,
      "eval_loss": 0.16118212044239044,
      "eval_mae": 0.6796970963478088,
      "eval_mse": 370.8824768066406,
      "eval_rmse": 19.25830929252723,
      "eval_runtime": 58.3327,
      "eval_samples_per_second": 10241.503,
      "eval_smape": 84.66194272041321,
      "eval_steps_per_second": 20.006,
      "step": 6000
    },
    {
      "epoch": 0.5761081750226158,
      "grad_norm": 0.5452147126197815,
      "learning_rate": 9.95187303402917e-05,
      "loss": 0.1645,
      "step": 6050
    },
    {
      "epoch": 0.5808693996095796,
      "grad_norm": 0.6225939989089966,
      "learning_rate": 9.951396435039558e-05,
      "loss": 0.1629,
      "step": 6100
    },
    {
      "epoch": 0.5856306241965433,
      "grad_norm": 0.618532121181488,
      "learning_rate": 9.950919836049948e-05,
      "loss": 0.1638,
      "step": 6150
    },
    {
      "epoch": 0.5903918487835071,
      "grad_norm": 0.6065341830253601,
      "learning_rate": 9.950443237060338e-05,
      "loss": 0.1672,
      "step": 6200
    },
    {
      "epoch": 0.5951530733704709,
      "grad_norm": 0.7495716214179993,
      "learning_rate": 9.949966638070728e-05,
      "loss": 0.1648,
      "step": 6250
    },
    {
      "epoch": 0.5999142979574347,
      "grad_norm": 0.6554955244064331,
      "learning_rate": 9.949490039081118e-05,
      "loss": 0.1654,
      "step": 6300
    },
    {
      "epoch": 0.6046755225443984,
      "grad_norm": 0.5830172300338745,
      "learning_rate": 9.949013440091506e-05,
      "loss": 0.1629,
      "step": 6350
    },
    {
      "epoch": 0.6094367471313622,
      "grad_norm": 0.5021042823791504,
      "learning_rate": 9.948536841101898e-05,
      "loss": 0.1622,
      "step": 6400
    },
    {
      "epoch": 0.6141979717183259,
      "grad_norm": 0.47169509530067444,
      "learning_rate": 9.948060242112288e-05,
      "loss": 0.1632,
      "step": 6450
    },
    {
      "epoch": 0.6189591963052897,
      "grad_norm": 0.7609395980834961,
      "learning_rate": 9.947583643122676e-05,
      "loss": 0.1641,
      "step": 6500
    },
    {
      "epoch": 0.6237204208922534,
      "grad_norm": 0.5191305875778198,
      "learning_rate": 9.947107044133068e-05,
      "loss": 0.1672,
      "step": 6550
    },
    {
      "epoch": 0.6284816454792173,
      "grad_norm": 0.5454711318016052,
      "learning_rate": 9.946630445143456e-05,
      "loss": 0.1648,
      "step": 6600
    },
    {
      "epoch": 0.6332428700661811,
      "grad_norm": 0.49112918972969055,
      "learning_rate": 9.946153846153846e-05,
      "loss": 0.1648,
      "step": 6650
    },
    {
      "epoch": 0.6380040946531448,
      "grad_norm": 0.4859708249568939,
      "learning_rate": 9.945677247164236e-05,
      "loss": 0.1652,
      "step": 6700
    },
    {
      "epoch": 0.6427653192401086,
      "grad_norm": 0.506971001625061,
      "learning_rate": 9.945200648174626e-05,
      "loss": 0.1623,
      "step": 6750
    },
    {
      "epoch": 0.6475265438270723,
      "grad_norm": 0.5732383131980896,
      "learning_rate": 9.944724049185016e-05,
      "loss": 0.1657,
      "step": 6800
    },
    {
      "epoch": 0.6522877684140361,
      "grad_norm": 0.548362672328949,
      "learning_rate": 9.944247450195406e-05,
      "loss": 0.1628,
      "step": 6850
    },
    {
      "epoch": 0.6570489930009998,
      "grad_norm": 0.5271615982055664,
      "learning_rate": 9.943770851205796e-05,
      "loss": 0.1627,
      "step": 6900
    },
    {
      "epoch": 0.6618102175879637,
      "grad_norm": 0.7555857300758362,
      "learning_rate": 9.943294252216186e-05,
      "loss": 0.1635,
      "step": 6950
    },
    {
      "epoch": 0.6665714421749274,
      "grad_norm": 0.5426679849624634,
      "learning_rate": 9.942817653226576e-05,
      "loss": 0.1623,
      "step": 7000
    },
    {
      "epoch": 0.6665714421749274,
      "eval_loss": 0.1605178564786911,
      "eval_mae": 0.6715303063392639,
      "eval_mse": 400.0790100097656,
      "eval_rmse": 20.001975152713435,
      "eval_runtime": 61.0562,
      "eval_samples_per_second": 9784.671,
      "eval_smape": 89.75983262062073,
      "eval_steps_per_second": 19.114,
      "step": 7000
    },
    {
      "epoch": 0.6713326667618912,
      "grad_norm": 0.5322990417480469,
      "learning_rate": 9.942341054236966e-05,
      "loss": 0.1625,
      "step": 7050
    },
    {
      "epoch": 0.6760938913488549,
      "grad_norm": 0.6016077995300293,
      "learning_rate": 9.941864455247354e-05,
      "loss": 0.1647,
      "step": 7100
    },
    {
      "epoch": 0.6808551159358187,
      "grad_norm": 0.5076338648796082,
      "learning_rate": 9.941387856257746e-05,
      "loss": 0.1618,
      "step": 7150
    },
    {
      "epoch": 0.6856163405227824,
      "grad_norm": 0.5658571124076843,
      "learning_rate": 9.940911257268136e-05,
      "loss": 0.1662,
      "step": 7200
    },
    {
      "epoch": 0.6903775651097462,
      "grad_norm": 0.6107982993125916,
      "learning_rate": 9.940434658278524e-05,
      "loss": 0.1608,
      "step": 7250
    },
    {
      "epoch": 0.69513878969671,
      "grad_norm": 0.4623304307460785,
      "learning_rate": 9.939958059288916e-05,
      "loss": 0.1635,
      "step": 7300
    },
    {
      "epoch": 0.6999000142836738,
      "grad_norm": 0.6437474489212036,
      "learning_rate": 9.939481460299304e-05,
      "loss": 0.1606,
      "step": 7350
    },
    {
      "epoch": 0.7046612388706375,
      "grad_norm": 0.6315158605575562,
      "learning_rate": 9.939004861309694e-05,
      "loss": 0.1615,
      "step": 7400
    },
    {
      "epoch": 0.7094224634576013,
      "grad_norm": 0.6503571271896362,
      "learning_rate": 9.938528262320085e-05,
      "loss": 0.1619,
      "step": 7450
    },
    {
      "epoch": 0.7141836880445651,
      "grad_norm": 0.46252092719078064,
      "learning_rate": 9.938051663330474e-05,
      "loss": 0.1625,
      "step": 7500
    },
    {
      "epoch": 0.7189449126315288,
      "grad_norm": 0.5186336636543274,
      "learning_rate": 9.937575064340864e-05,
      "loss": 0.1628,
      "step": 7550
    },
    {
      "epoch": 0.7237061372184926,
      "grad_norm": 0.5236070156097412,
      "learning_rate": 9.937098465351254e-05,
      "loss": 0.1624,
      "step": 7600
    },
    {
      "epoch": 0.7284673618054563,
      "grad_norm": 0.4777911901473999,
      "learning_rate": 9.936621866361644e-05,
      "loss": 0.1625,
      "step": 7650
    },
    {
      "epoch": 0.7332285863924202,
      "grad_norm": 0.5092161297798157,
      "learning_rate": 9.936145267372034e-05,
      "loss": 0.163,
      "step": 7700
    },
    {
      "epoch": 0.7379898109793839,
      "grad_norm": 0.5161564350128174,
      "learning_rate": 9.935668668382424e-05,
      "loss": 0.1615,
      "step": 7750
    },
    {
      "epoch": 0.7427510355663477,
      "grad_norm": 0.48548147082328796,
      "learning_rate": 9.935192069392814e-05,
      "loss": 0.1592,
      "step": 7800
    },
    {
      "epoch": 0.7475122601533114,
      "grad_norm": 0.6095620393753052,
      "learning_rate": 9.934715470403202e-05,
      "loss": 0.1613,
      "step": 7850
    },
    {
      "epoch": 0.7522734847402752,
      "grad_norm": 0.49965670704841614,
      "learning_rate": 9.934238871413592e-05,
      "loss": 0.163,
      "step": 7900
    },
    {
      "epoch": 0.7570347093272389,
      "grad_norm": 0.5934204459190369,
      "learning_rate": 9.933762272423984e-05,
      "loss": 0.1638,
      "step": 7950
    },
    {
      "epoch": 0.7617959339142027,
      "grad_norm": 0.6522780060768127,
      "learning_rate": 9.933285673434372e-05,
      "loss": 0.1638,
      "step": 8000
    },
    {
      "epoch": 0.7617959339142027,
      "eval_loss": 0.16129492223262787,
      "eval_mae": 0.6771246194839478,
      "eval_mse": 387.6971435546875,
      "eval_rmse": 19.69002649959333,
      "eval_runtime": 57.6529,
      "eval_samples_per_second": 10362.273,
      "eval_smape": 122.37988710403442,
      "eval_steps_per_second": 20.242,
      "step": 8000
    },
    {
      "epoch": 0.7665571585011665,
      "grad_norm": 0.47631314396858215,
      "learning_rate": 9.932809074444762e-05,
      "loss": 0.1594,
      "step": 8050
    },
    {
      "epoch": 0.7713183830881303,
      "grad_norm": 0.4288536310195923,
      "learning_rate": 9.932332475455152e-05,
      "loss": 0.1627,
      "step": 8100
    },
    {
      "epoch": 0.776079607675094,
      "grad_norm": 0.4548576772212982,
      "learning_rate": 9.931855876465542e-05,
      "loss": 0.1638,
      "step": 8150
    },
    {
      "epoch": 0.7808408322620578,
      "grad_norm": 0.5950626730918884,
      "learning_rate": 9.931379277475932e-05,
      "loss": 0.1646,
      "step": 8200
    },
    {
      "epoch": 0.7856020568490216,
      "grad_norm": 0.5772454738616943,
      "learning_rate": 9.930902678486322e-05,
      "loss": 0.1629,
      "step": 8250
    },
    {
      "epoch": 0.7903632814359853,
      "grad_norm": 0.5833305716514587,
      "learning_rate": 9.930426079496712e-05,
      "loss": 0.1635,
      "step": 8300
    },
    {
      "epoch": 0.7951245060229492,
      "grad_norm": 0.4767976701259613,
      "learning_rate": 9.929949480507102e-05,
      "loss": 0.1621,
      "step": 8350
    },
    {
      "epoch": 0.7998857306099129,
      "grad_norm": 0.586681604385376,
      "learning_rate": 9.929472881517492e-05,
      "loss": 0.1633,
      "step": 8400
    },
    {
      "epoch": 0.8046469551968767,
      "grad_norm": 0.46445733308792114,
      "learning_rate": 9.928996282527882e-05,
      "loss": 0.1621,
      "step": 8450
    },
    {
      "epoch": 0.8094081797838404,
      "grad_norm": 0.4659370183944702,
      "learning_rate": 9.92851968353827e-05,
      "loss": 0.1644,
      "step": 8500
    },
    {
      "epoch": 0.8141694043708042,
      "grad_norm": 0.48823997378349304,
      "learning_rate": 9.928043084548662e-05,
      "loss": 0.1654,
      "step": 8550
    },
    {
      "epoch": 0.8189306289577679,
      "grad_norm": 0.5804855823516846,
      "learning_rate": 9.92756648555905e-05,
      "loss": 0.1624,
      "step": 8600
    },
    {
      "epoch": 0.8236918535447317,
      "grad_norm": 0.4181581139564514,
      "learning_rate": 9.92708988656944e-05,
      "loss": 0.1593,
      "step": 8650
    },
    {
      "epoch": 0.8284530781316954,
      "grad_norm": 0.6322731971740723,
      "learning_rate": 9.926613287579832e-05,
      "loss": 0.1628,
      "step": 8700
    },
    {
      "epoch": 0.8332143027186593,
      "grad_norm": 0.39184707403182983,
      "learning_rate": 9.92613668859022e-05,
      "loss": 0.1639,
      "step": 8750
    },
    {
      "epoch": 0.837975527305623,
      "grad_norm": 0.5011768341064453,
      "learning_rate": 9.92566008960061e-05,
      "loss": 0.1634,
      "step": 8800
    },
    {
      "epoch": 0.8427367518925868,
      "grad_norm": 0.47292882204055786,
      "learning_rate": 9.925183490611e-05,
      "loss": 0.1609,
      "step": 8850
    },
    {
      "epoch": 0.8474979764795505,
      "grad_norm": 0.5086949467658997,
      "learning_rate": 9.92470689162139e-05,
      "loss": 0.1616,
      "step": 8900
    },
    {
      "epoch": 0.8522592010665143,
      "grad_norm": 0.4720959961414337,
      "learning_rate": 9.92423029263178e-05,
      "loss": 0.1611,
      "step": 8950
    },
    {
      "epoch": 0.857020425653478,
      "grad_norm": 0.8099896311759949,
      "learning_rate": 9.92375369364217e-05,
      "loss": 0.1609,
      "step": 9000
    },
    {
      "epoch": 0.857020425653478,
      "eval_loss": 0.16015625,
      "eval_mae": 0.6603007912635803,
      "eval_mse": 335.3426818847656,
      "eval_rmse": 18.31236418065034,
      "eval_runtime": 56.2142,
      "eval_samples_per_second": 10627.482,
      "eval_smape": 109.38767194747925,
      "eval_steps_per_second": 20.76,
      "step": 9000
    },
    {
      "epoch": 0.8617816502404418,
      "grad_norm": 0.5228590369224548,
      "learning_rate": 9.92327709465256e-05,
      "loss": 0.1617,
      "step": 9050
    },
    {
      "epoch": 0.8665428748274057,
      "grad_norm": 0.5515425205230713,
      "learning_rate": 9.92280049566295e-05,
      "loss": 0.1612,
      "step": 9100
    },
    {
      "epoch": 0.8713040994143694,
      "grad_norm": 0.5289241075515747,
      "learning_rate": 9.92232389667334e-05,
      "loss": 0.1646,
      "step": 9150
    },
    {
      "epoch": 0.8760653240013332,
      "grad_norm": 0.5692815780639648,
      "learning_rate": 9.92184729768373e-05,
      "loss": 0.1603,
      "step": 9200
    },
    {
      "epoch": 0.8808265485882969,
      "grad_norm": 0.41486117243766785,
      "learning_rate": 9.921370698694119e-05,
      "loss": 0.1632,
      "step": 9250
    },
    {
      "epoch": 0.8855877731752607,
      "grad_norm": 0.488235741853714,
      "learning_rate": 9.920894099704509e-05,
      "loss": 0.1613,
      "step": 9300
    },
    {
      "epoch": 0.8903489977622244,
      "grad_norm": 0.6576530337333679,
      "learning_rate": 9.9204175007149e-05,
      "loss": 0.1618,
      "step": 9350
    },
    {
      "epoch": 0.8951102223491882,
      "grad_norm": 0.49431854486465454,
      "learning_rate": 9.919940901725288e-05,
      "loss": 0.1618,
      "step": 9400
    },
    {
      "epoch": 0.899871446936152,
      "grad_norm": 0.5491801500320435,
      "learning_rate": 9.919464302735678e-05,
      "loss": 0.162,
      "step": 9450
    },
    {
      "epoch": 0.9046326715231158,
      "grad_norm": 0.5839897990226746,
      "learning_rate": 9.918987703746068e-05,
      "loss": 0.1587,
      "step": 9500
    },
    {
      "epoch": 0.9093938961100795,
      "grad_norm": 0.5631112456321716,
      "learning_rate": 9.918511104756458e-05,
      "loss": 0.1607,
      "step": 9550
    },
    {
      "epoch": 0.9141551206970433,
      "grad_norm": 0.5420098900794983,
      "learning_rate": 9.918034505766848e-05,
      "loss": 0.1642,
      "step": 9600
    },
    {
      "epoch": 0.918916345284007,
      "grad_norm": 0.568087637424469,
      "learning_rate": 9.917557906777238e-05,
      "loss": 0.1624,
      "step": 9650
    },
    {
      "epoch": 0.9236775698709708,
      "grad_norm": 0.5823555588722229,
      "learning_rate": 9.917081307787628e-05,
      "loss": 0.1639,
      "step": 9700
    },
    {
      "epoch": 0.9284387944579345,
      "grad_norm": 0.5538271069526672,
      "learning_rate": 9.916604708798017e-05,
      "loss": 0.1625,
      "step": 9750
    },
    {
      "epoch": 0.9332000190448984,
      "grad_norm": 0.5160115957260132,
      "learning_rate": 9.916128109808408e-05,
      "loss": 0.1582,
      "step": 9800
    },
    {
      "epoch": 0.9379612436318621,
      "grad_norm": 0.48957574367523193,
      "learning_rate": 9.915651510818798e-05,
      "loss": 0.16,
      "step": 9850
    },
    {
      "epoch": 0.9427224682188259,
      "grad_norm": 0.601917564868927,
      "learning_rate": 9.915174911829187e-05,
      "loss": 0.162,
      "step": 9900
    },
    {
      "epoch": 0.9474836928057897,
      "grad_norm": 0.40893155336380005,
      "learning_rate": 9.914698312839578e-05,
      "loss": 0.165,
      "step": 9950
    },
    {
      "epoch": 0.9522449173927534,
      "grad_norm": 0.5392901301383972,
      "learning_rate": 9.914221713849967e-05,
      "loss": 0.1618,
      "step": 10000
    },
    {
      "epoch": 0.9522449173927534,
      "eval_loss": 0.15920588374137878,
      "eval_mae": 0.6688477993011475,
      "eval_mse": 318.14923095703125,
      "eval_rmse": 17.83673823761035,
      "eval_runtime": 59.3096,
      "eval_samples_per_second": 10072.824,
      "eval_smape": 76.33218169212341,
      "eval_steps_per_second": 19.676,
      "step": 10000
    },
    {
      "epoch": 0.9570061419797172,
      "grad_norm": 0.600141704082489,
      "learning_rate": 9.913745114860357e-05,
      "loss": 0.1596,
      "step": 10050
    },
    {
      "epoch": 0.9617673665666809,
      "grad_norm": 0.49183207750320435,
      "learning_rate": 9.913268515870748e-05,
      "loss": 0.1636,
      "step": 10100
    },
    {
      "epoch": 0.9665285911536448,
      "grad_norm": 0.4760478138923645,
      "learning_rate": 9.912791916881136e-05,
      "loss": 0.1584,
      "step": 10150
    },
    {
      "epoch": 0.9712898157406085,
      "grad_norm": 0.4729287624359131,
      "learning_rate": 9.912315317891526e-05,
      "loss": 0.1617,
      "step": 10200
    },
    {
      "epoch": 0.9760510403275723,
      "grad_norm": 0.578062117099762,
      "learning_rate": 9.911838718901916e-05,
      "loss": 0.1606,
      "step": 10250
    },
    {
      "epoch": 0.980812264914536,
      "grad_norm": 0.43438589572906494,
      "learning_rate": 9.911362119912306e-05,
      "loss": 0.1617,
      "step": 10300
    },
    {
      "epoch": 0.9855734895014998,
      "grad_norm": 0.5887606143951416,
      "learning_rate": 9.910885520922696e-05,
      "loss": 0.1616,
      "step": 10350
    },
    {
      "epoch": 0.9903347140884635,
      "grad_norm": 0.4835382401943207,
      "learning_rate": 9.910408921933086e-05,
      "loss": 0.16,
      "step": 10400
    },
    {
      "epoch": 0.9950959386754273,
      "grad_norm": 0.4186175763607025,
      "learning_rate": 9.909932322943476e-05,
      "loss": 0.1636,
      "step": 10450
    },
    {
      "epoch": 0.999857163262391,
      "grad_norm": 0.6031454205513,
      "learning_rate": 9.909455723953865e-05,
      "loss": 0.16,
      "step": 10500
    },
    {
      "epoch": 1.0045707756034852,
      "grad_norm": 0.444181352853775,
      "learning_rate": 9.908979124964256e-05,
      "loss": 0.1588,
      "step": 10550
    },
    {
      "epoch": 1.0093320001904489,
      "grad_norm": 0.474399596452713,
      "learning_rate": 9.908502525974646e-05,
      "loss": 0.1651,
      "step": 10600
    },
    {
      "epoch": 1.0140932247774128,
      "grad_norm": 0.4467703700065613,
      "learning_rate": 9.908025926985035e-05,
      "loss": 0.1598,
      "step": 10650
    },
    {
      "epoch": 1.0188544493643765,
      "grad_norm": 0.5107735395431519,
      "learning_rate": 9.907549327995426e-05,
      "loss": 0.1609,
      "step": 10700
    },
    {
      "epoch": 1.0236156739513402,
      "grad_norm": 0.4578382968902588,
      "learning_rate": 9.907072729005815e-05,
      "loss": 0.1614,
      "step": 10750
    },
    {
      "epoch": 1.028376898538304,
      "grad_norm": 0.4924687445163727,
      "learning_rate": 9.906596130016205e-05,
      "loss": 0.1594,
      "step": 10800
    },
    {
      "epoch": 1.0331381231252679,
      "grad_norm": 0.4967709183692932,
      "learning_rate": 9.906119531026595e-05,
      "loss": 0.1614,
      "step": 10850
    },
    {
      "epoch": 1.0378993477122316,
      "grad_norm": 0.5059126615524292,
      "learning_rate": 9.905642932036984e-05,
      "loss": 0.1622,
      "step": 10900
    },
    {
      "epoch": 1.0426605722991953,
      "grad_norm": 0.5229778289794922,
      "learning_rate": 9.905166333047374e-05,
      "loss": 0.1592,
      "step": 10950
    },
    {
      "epoch": 1.0474217968861592,
      "grad_norm": 0.45143234729766846,
      "learning_rate": 9.904689734057764e-05,
      "loss": 0.1588,
      "step": 11000
    },
    {
      "epoch": 1.0474217968861592,
      "eval_loss": 0.15856099128723145,
      "eval_mae": 0.6628284454345703,
      "eval_mse": 345.36749267578125,
      "eval_rmse": 18.584065558315846,
      "eval_runtime": 57.9218,
      "eval_samples_per_second": 10314.166,
      "eval_smape": 94.50321197509766,
      "eval_steps_per_second": 20.148,
      "step": 11000
    },
    {
      "epoch": 1.052183021473123,
      "grad_norm": 0.7248756885528564,
      "learning_rate": 9.904213135068154e-05,
      "loss": 0.1614,
      "step": 11050
    },
    {
      "epoch": 1.0569442460600866,
      "grad_norm": 0.5051783323287964,
      "learning_rate": 9.903736536078544e-05,
      "loss": 0.1567,
      "step": 11100
    },
    {
      "epoch": 1.0617054706470503,
      "grad_norm": 0.5285612940788269,
      "learning_rate": 9.903259937088933e-05,
      "loss": 0.1613,
      "step": 11150
    },
    {
      "epoch": 1.0664666952340143,
      "grad_norm": 0.5270511507987976,
      "learning_rate": 9.902783338099324e-05,
      "loss": 0.1615,
      "step": 11200
    },
    {
      "epoch": 1.071227919820978,
      "grad_norm": 0.5635538101196289,
      "learning_rate": 9.902306739109714e-05,
      "loss": 0.1606,
      "step": 11250
    },
    {
      "epoch": 1.0759891444079417,
      "grad_norm": 0.5780921578407288,
      "learning_rate": 9.901830140120103e-05,
      "loss": 0.1619,
      "step": 11300
    },
    {
      "epoch": 1.0807503689949054,
      "grad_norm": 0.5149776935577393,
      "learning_rate": 9.901353541130494e-05,
      "loss": 0.1638,
      "step": 11350
    },
    {
      "epoch": 1.0855115935818693,
      "grad_norm": 0.46075335144996643,
      "learning_rate": 9.900876942140883e-05,
      "loss": 0.1597,
      "step": 11400
    },
    {
      "epoch": 1.090272818168833,
      "grad_norm": 0.42272669076919556,
      "learning_rate": 9.900400343151273e-05,
      "loss": 0.16,
      "step": 11450
    },
    {
      "epoch": 1.0950340427557967,
      "grad_norm": 0.5383277535438538,
      "learning_rate": 9.899923744161663e-05,
      "loss": 0.159,
      "step": 11500
    },
    {
      "epoch": 1.0997952673427607,
      "grad_norm": 0.45635008811950684,
      "learning_rate": 9.899447145172053e-05,
      "loss": 0.1596,
      "step": 11550
    },
    {
      "epoch": 1.1045564919297244,
      "grad_norm": 0.4863174259662628,
      "learning_rate": 9.898970546182443e-05,
      "loss": 0.1613,
      "step": 11600
    },
    {
      "epoch": 1.109317716516688,
      "grad_norm": 0.5197418332099915,
      "learning_rate": 9.898493947192832e-05,
      "loss": 0.1601,
      "step": 11650
    },
    {
      "epoch": 1.1140789411036518,
      "grad_norm": 0.6299956440925598,
      "learning_rate": 9.898017348203222e-05,
      "loss": 0.1621,
      "step": 11700
    },
    {
      "epoch": 1.1188401656906157,
      "grad_norm": 0.47120076417922974,
      "learning_rate": 9.897540749213612e-05,
      "loss": 0.1624,
      "step": 11750
    },
    {
      "epoch": 1.1236013902775794,
      "grad_norm": 0.4576870799064636,
      "learning_rate": 9.897064150224002e-05,
      "loss": 0.1629,
      "step": 11800
    },
    {
      "epoch": 1.1283626148645431,
      "grad_norm": 0.531906008720398,
      "learning_rate": 9.896587551234392e-05,
      "loss": 0.1593,
      "step": 11850
    },
    {
      "epoch": 1.1331238394515069,
      "grad_norm": 0.4709097743034363,
      "learning_rate": 9.896110952244781e-05,
      "loss": 0.1613,
      "step": 11900
    },
    {
      "epoch": 1.1378850640384708,
      "grad_norm": 0.5573967695236206,
      "learning_rate": 9.895634353255172e-05,
      "loss": 0.1623,
      "step": 11950
    },
    {
      "epoch": 1.1426462886254345,
      "grad_norm": 0.5798735022544861,
      "learning_rate": 9.895157754265562e-05,
      "loss": 0.1601,
      "step": 12000
    },
    {
      "epoch": 1.1426462886254345,
      "eval_loss": 0.15796181559562683,
      "eval_mae": 0.65403813123703,
      "eval_mse": 326.8865051269531,
      "eval_rmse": 18.0800029072717,
      "eval_runtime": 56.7165,
      "eval_samples_per_second": 10533.356,
      "eval_smape": 81.25044703483582,
      "eval_steps_per_second": 20.576,
      "step": 12000
    },
    {
      "epoch": 1.1474075132123982,
      "grad_norm": 0.5247554183006287,
      "learning_rate": 9.894681155275951e-05,
      "loss": 0.1574,
      "step": 12050
    },
    {
      "epoch": 1.152168737799362,
      "grad_norm": 0.6056792736053467,
      "learning_rate": 9.894204556286342e-05,
      "loss": 0.1583,
      "step": 12100
    },
    {
      "epoch": 1.1569299623863258,
      "grad_norm": 0.5114259123802185,
      "learning_rate": 9.893727957296731e-05,
      "loss": 0.1641,
      "step": 12150
    },
    {
      "epoch": 1.1616911869732895,
      "grad_norm": 0.40764451026916504,
      "learning_rate": 9.89325135830712e-05,
      "loss": 0.1621,
      "step": 12200
    },
    {
      "epoch": 1.1664524115602533,
      "grad_norm": 0.5216367244720459,
      "learning_rate": 9.89277475931751e-05,
      "loss": 0.1597,
      "step": 12250
    },
    {
      "epoch": 1.171213636147217,
      "grad_norm": 0.48652514815330505,
      "learning_rate": 9.8922981603279e-05,
      "loss": 0.1605,
      "step": 12300
    },
    {
      "epoch": 1.175974860734181,
      "grad_norm": 0.46615609526634216,
      "learning_rate": 9.89182156133829e-05,
      "loss": 0.1571,
      "step": 12350
    },
    {
      "epoch": 1.1807360853211446,
      "grad_norm": 0.5413913130760193,
      "learning_rate": 9.891344962348679e-05,
      "loss": 0.1599,
      "step": 12400
    },
    {
      "epoch": 1.1854973099081083,
      "grad_norm": 0.42148470878601074,
      "learning_rate": 9.89086836335907e-05,
      "loss": 0.1605,
      "step": 12450
    },
    {
      "epoch": 1.1902585344950722,
      "grad_norm": 0.7329843044281006,
      "learning_rate": 9.89039176436946e-05,
      "loss": 0.1603,
      "step": 12500
    },
    {
      "epoch": 1.195019759082036,
      "grad_norm": 0.42273497581481934,
      "learning_rate": 9.889915165379849e-05,
      "loss": 0.1617,
      "step": 12550
    },
    {
      "epoch": 1.1997809836689997,
      "grad_norm": 0.5869929790496826,
      "learning_rate": 9.88943856639024e-05,
      "loss": 0.1612,
      "step": 12600
    },
    {
      "epoch": 1.2045422082559634,
      "grad_norm": 0.4401240944862366,
      "learning_rate": 9.888961967400629e-05,
      "loss": 0.1577,
      "step": 12650
    },
    {
      "epoch": 1.2093034328429273,
      "grad_norm": 0.48275691270828247,
      "learning_rate": 9.888485368411019e-05,
      "loss": 0.1595,
      "step": 12700
    },
    {
      "epoch": 1.214064657429891,
      "grad_norm": 0.5482587218284607,
      "learning_rate": 9.88800876942141e-05,
      "loss": 0.16,
      "step": 12750
    },
    {
      "epoch": 1.2188258820168547,
      "grad_norm": 0.6347602009773254,
      "learning_rate": 9.887532170431799e-05,
      "loss": 0.1596,
      "step": 12800
    },
    {
      "epoch": 1.2235871066038184,
      "grad_norm": 0.4231024384498596,
      "learning_rate": 9.887055571442189e-05,
      "loss": 0.1624,
      "step": 12850
    },
    {
      "epoch": 1.2283483311907823,
      "grad_norm": 0.5505658984184265,
      "learning_rate": 9.886578972452579e-05,
      "loss": 0.1582,
      "step": 12900
    },
    {
      "epoch": 1.233109555777746,
      "grad_norm": 0.4918171763420105,
      "learning_rate": 9.886102373462969e-05,
      "loss": 0.1577,
      "step": 12950
    },
    {
      "epoch": 1.2378707803647098,
      "grad_norm": 0.45393800735473633,
      "learning_rate": 9.885625774473359e-05,
      "loss": 0.1585,
      "step": 13000
    },
    {
      "epoch": 1.2378707803647098,
      "eval_loss": 0.15753023326396942,
      "eval_mae": 0.6532484889030457,
      "eval_mse": 279.79644775390625,
      "eval_rmse": 16.72711713816539,
      "eval_runtime": 57.893,
      "eval_samples_per_second": 10319.3,
      "eval_smape": 107.6181173324585,
      "eval_steps_per_second": 20.158,
      "step": 13000
    },
    {
      "epoch": 1.2426320049516735,
      "grad_norm": 0.3777833878993988,
      "learning_rate": 9.885149175483749e-05,
      "loss": 0.1577,
      "step": 13050
    },
    {
      "epoch": 1.2473932295386374,
      "grad_norm": 0.5700230002403259,
      "learning_rate": 9.884672576494139e-05,
      "loss": 0.1587,
      "step": 13100
    },
    {
      "epoch": 1.2521544541256011,
      "grad_norm": 0.4470745027065277,
      "learning_rate": 9.884195977504529e-05,
      "loss": 0.1604,
      "step": 13150
    },
    {
      "epoch": 1.2569156787125648,
      "grad_norm": 0.4185906648635864,
      "learning_rate": 9.883719378514918e-05,
      "loss": 0.1562,
      "step": 13200
    },
    {
      "epoch": 1.2616769032995285,
      "grad_norm": 0.3835722804069519,
      "learning_rate": 9.883242779525308e-05,
      "loss": 0.1583,
      "step": 13250
    },
    {
      "epoch": 1.2664381278864925,
      "grad_norm": 0.45621258020401,
      "learning_rate": 9.882766180535697e-05,
      "loss": 0.1577,
      "step": 13300
    },
    {
      "epoch": 1.2711993524734562,
      "grad_norm": 0.5590381622314453,
      "learning_rate": 9.882289581546088e-05,
      "loss": 0.1608,
      "step": 13350
    },
    {
      "epoch": 1.2759605770604199,
      "grad_norm": 0.5501840114593506,
      "learning_rate": 9.881812982556477e-05,
      "loss": 0.1623,
      "step": 13400
    },
    {
      "epoch": 1.2807218016473838,
      "grad_norm": 0.4480155408382416,
      "learning_rate": 9.881336383566867e-05,
      "loss": 0.1566,
      "step": 13450
    },
    {
      "epoch": 1.2854830262343475,
      "grad_norm": 0.5472989678382874,
      "learning_rate": 9.880859784577258e-05,
      "loss": 0.1613,
      "step": 13500
    },
    {
      "epoch": 1.2902442508213112,
      "grad_norm": 0.48176899552345276,
      "learning_rate": 9.880383185587647e-05,
      "loss": 0.1614,
      "step": 13550
    },
    {
      "epoch": 1.295005475408275,
      "grad_norm": 0.40140554308891296,
      "learning_rate": 9.879906586598037e-05,
      "loss": 0.1586,
      "step": 13600
    },
    {
      "epoch": 1.2997666999952386,
      "grad_norm": 0.5117682218551636,
      "learning_rate": 9.879429987608427e-05,
      "loss": 0.1602,
      "step": 13650
    },
    {
      "epoch": 1.3045279245822026,
      "grad_norm": 0.7169548273086548,
      "learning_rate": 9.878953388618817e-05,
      "loss": 0.1623,
      "step": 13700
    },
    {
      "epoch": 1.3092891491691663,
      "grad_norm": 0.44949373602867126,
      "learning_rate": 9.878476789629207e-05,
      "loss": 0.1601,
      "step": 13750
    },
    {
      "epoch": 1.31405037375613,
      "grad_norm": 0.597175121307373,
      "learning_rate": 9.878000190639595e-05,
      "loss": 0.1606,
      "step": 13800
    },
    {
      "epoch": 1.318811598343094,
      "grad_norm": 0.553895115852356,
      "learning_rate": 9.877523591649987e-05,
      "loss": 0.1617,
      "step": 13850
    },
    {
      "epoch": 1.3235728229300576,
      "grad_norm": 0.5089927315711975,
      "learning_rate": 9.877046992660377e-05,
      "loss": 0.1595,
      "step": 13900
    },
    {
      "epoch": 1.3283340475170213,
      "grad_norm": 0.430867463350296,
      "learning_rate": 9.876570393670765e-05,
      "loss": 0.1582,
      "step": 13950
    },
    {
      "epoch": 1.3330952721039853,
      "grad_norm": 0.39970675110816956,
      "learning_rate": 9.876093794681156e-05,
      "loss": 0.1567,
      "step": 14000
    },
    {
      "epoch": 1.3330952721039853,
      "eval_loss": 0.15753522515296936,
      "eval_mae": 0.6622462272644043,
      "eval_mse": 328.3490295410156,
      "eval_rmse": 18.120403680409982,
      "eval_runtime": 58.0545,
      "eval_samples_per_second": 10290.588,
      "eval_smape": 91.98985695838928,
      "eval_steps_per_second": 20.102,
      "step": 14000
    },
    {
      "epoch": 1.337856496690949,
      "grad_norm": 0.5808453559875488,
      "learning_rate": 9.875617195691545e-05,
      "loss": 0.1589,
      "step": 14050
    },
    {
      "epoch": 1.3426177212779127,
      "grad_norm": 0.4418608546257019,
      "learning_rate": 9.875140596701935e-05,
      "loss": 0.1584,
      "step": 14100
    },
    {
      "epoch": 1.3473789458648764,
      "grad_norm": 0.6623143553733826,
      "learning_rate": 9.874663997712325e-05,
      "loss": 0.1623,
      "step": 14150
    },
    {
      "epoch": 1.35214017045184,
      "grad_norm": 0.4194190204143524,
      "learning_rate": 9.874187398722715e-05,
      "loss": 0.1596,
      "step": 14200
    },
    {
      "epoch": 1.356901395038804,
      "grad_norm": 0.6208611130714417,
      "learning_rate": 9.873710799733105e-05,
      "loss": 0.1583,
      "step": 14250
    },
    {
      "epoch": 1.3616626196257677,
      "grad_norm": 0.4435657262802124,
      "learning_rate": 9.873234200743495e-05,
      "loss": 0.16,
      "step": 14300
    },
    {
      "epoch": 1.3664238442127314,
      "grad_norm": 0.4906177222728729,
      "learning_rate": 9.872757601753885e-05,
      "loss": 0.1581,
      "step": 14350
    },
    {
      "epoch": 1.3711850687996954,
      "grad_norm": 0.5340787172317505,
      "learning_rate": 9.872281002764275e-05,
      "loss": 0.1588,
      "step": 14400
    },
    {
      "epoch": 1.375946293386659,
      "grad_norm": 0.4307633340358734,
      "learning_rate": 9.871804403774665e-05,
      "loss": 0.1585,
      "step": 14450
    },
    {
      "epoch": 1.3807075179736228,
      "grad_norm": 0.49194300174713135,
      "learning_rate": 9.871327804785055e-05,
      "loss": 0.1572,
      "step": 14500
    },
    {
      "epoch": 1.3854687425605867,
      "grad_norm": 0.4391520917415619,
      "learning_rate": 9.870851205795443e-05,
      "loss": 0.1573,
      "step": 14550
    },
    {
      "epoch": 1.3902299671475504,
      "grad_norm": 0.4503444731235504,
      "learning_rate": 9.870374606805835e-05,
      "loss": 0.1577,
      "step": 14600
    },
    {
      "epoch": 1.3949911917345141,
      "grad_norm": 0.5113334655761719,
      "learning_rate": 9.869898007816225e-05,
      "loss": 0.1594,
      "step": 14650
    },
    {
      "epoch": 1.3997524163214778,
      "grad_norm": 0.3871005177497864,
      "learning_rate": 9.869421408826613e-05,
      "loss": 0.1559,
      "step": 14700
    },
    {
      "epoch": 1.4045136409084416,
      "grad_norm": 0.5482053756713867,
      "learning_rate": 9.868944809837004e-05,
      "loss": 0.1592,
      "step": 14750
    },
    {
      "epoch": 1.4092748654954055,
      "grad_norm": 0.45771437883377075,
      "learning_rate": 9.868468210847393e-05,
      "loss": 0.1572,
      "step": 14800
    },
    {
      "epoch": 1.4140360900823692,
      "grad_norm": 0.4550696015357971,
      "learning_rate": 9.867991611857783e-05,
      "loss": 0.1605,
      "step": 14850
    },
    {
      "epoch": 1.418797314669333,
      "grad_norm": 0.6991235613822937,
      "learning_rate": 9.867515012868174e-05,
      "loss": 0.1589,
      "step": 14900
    },
    {
      "epoch": 1.4235585392562968,
      "grad_norm": 0.5531545877456665,
      "learning_rate": 9.867038413878563e-05,
      "loss": 0.1574,
      "step": 14950
    },
    {
      "epoch": 1.4283197638432605,
      "grad_norm": 0.4692751169204712,
      "learning_rate": 9.866561814888953e-05,
      "loss": 0.1592,
      "step": 15000
    },
    {
      "epoch": 1.4283197638432605,
      "eval_loss": 0.15674826502799988,
      "eval_mae": 0.6522776484489441,
      "eval_mse": 376.8973388671875,
      "eval_rmse": 19.41384400027948,
      "eval_runtime": 57.2088,
      "eval_samples_per_second": 10442.706,
      "eval_smape": 89.79519009590149,
      "eval_steps_per_second": 20.399,
      "step": 15000
    },
    {
      "epoch": 1.4330809884302242,
      "grad_norm": 0.41401219367980957,
      "learning_rate": 9.866085215899343e-05,
      "loss": 0.1594,
      "step": 15050
    },
    {
      "epoch": 1.437842213017188,
      "grad_norm": 0.6155771613121033,
      "learning_rate": 9.865608616909733e-05,
      "loss": 0.1597,
      "step": 15100
    },
    {
      "epoch": 1.4426034376041517,
      "grad_norm": 0.6229146718978882,
      "learning_rate": 9.865132017920123e-05,
      "loss": 0.1595,
      "step": 15150
    },
    {
      "epoch": 1.4473646621911156,
      "grad_norm": 0.5311764478683472,
      "learning_rate": 9.864655418930511e-05,
      "loss": 0.1613,
      "step": 15200
    },
    {
      "epoch": 1.4521258867780793,
      "grad_norm": 0.4075564742088318,
      "learning_rate": 9.864178819940903e-05,
      "loss": 0.1581,
      "step": 15250
    },
    {
      "epoch": 1.456887111365043,
      "grad_norm": 0.5737677216529846,
      "learning_rate": 9.863702220951291e-05,
      "loss": 0.1622,
      "step": 15300
    },
    {
      "epoch": 1.461648335952007,
      "grad_norm": 0.5977826118469238,
      "learning_rate": 9.863225621961681e-05,
      "loss": 0.1586,
      "step": 15350
    },
    {
      "epoch": 1.4664095605389706,
      "grad_norm": 0.4717673659324646,
      "learning_rate": 9.862749022972073e-05,
      "loss": 0.1566,
      "step": 15400
    },
    {
      "epoch": 1.4711707851259344,
      "grad_norm": 0.4783164858818054,
      "learning_rate": 9.862272423982461e-05,
      "loss": 0.162,
      "step": 15450
    },
    {
      "epoch": 1.4759320097128983,
      "grad_norm": 0.4709276258945465,
      "learning_rate": 9.861795824992851e-05,
      "loss": 0.1628,
      "step": 15500
    },
    {
      "epoch": 1.480693234299862,
      "grad_norm": 0.46055328845977783,
      "learning_rate": 9.861319226003241e-05,
      "loss": 0.1573,
      "step": 15550
    },
    {
      "epoch": 1.4854544588868257,
      "grad_norm": 0.4675583839416504,
      "learning_rate": 9.860842627013631e-05,
      "loss": 0.1612,
      "step": 15600
    },
    {
      "epoch": 1.4902156834737894,
      "grad_norm": 0.49638524651527405,
      "learning_rate": 9.860366028024021e-05,
      "loss": 0.1565,
      "step": 15650
    },
    {
      "epoch": 1.4949769080607531,
      "grad_norm": 0.4782148003578186,
      "learning_rate": 9.859889429034411e-05,
      "loss": 0.1559,
      "step": 15700
    },
    {
      "epoch": 1.499738132647717,
      "grad_norm": 0.39972543716430664,
      "learning_rate": 9.859412830044801e-05,
      "loss": 0.1597,
      "step": 15750
    },
    {
      "epoch": 1.5044993572346808,
      "grad_norm": 0.46148207783699036,
      "learning_rate": 9.858936231055191e-05,
      "loss": 0.165,
      "step": 15800
    },
    {
      "epoch": 1.5092605818216445,
      "grad_norm": 0.6166296005249023,
      "learning_rate": 9.858459632065581e-05,
      "loss": 0.1601,
      "step": 15850
    },
    {
      "epoch": 1.5140218064086084,
      "grad_norm": 0.6877084970474243,
      "learning_rate": 9.857983033075971e-05,
      "loss": 0.1622,
      "step": 15900
    },
    {
      "epoch": 1.518783030995572,
      "grad_norm": 0.4697044789791107,
      "learning_rate": 9.85750643408636e-05,
      "loss": 0.1591,
      "step": 15950
    },
    {
      "epoch": 1.5235442555825358,
      "grad_norm": 0.4562913179397583,
      "learning_rate": 9.857029835096751e-05,
      "loss": 0.16,
      "step": 16000
    },
    {
      "epoch": 1.5235442555825358,
      "eval_loss": 0.15759705007076263,
      "eval_mae": 0.6579864621162415,
      "eval_mse": 327.5270690917969,
      "eval_rmse": 18.097708945935583,
      "eval_runtime": 59.3074,
      "eval_samples_per_second": 10073.191,
      "eval_smape": 105.73155879974365,
      "eval_steps_per_second": 19.677,
      "step": 16000
    },
    {
      "epoch": 1.5283054801694997,
      "grad_norm": 0.4631439447402954,
      "learning_rate": 9.856553236107139e-05,
      "loss": 0.1596,
      "step": 16050
    },
    {
      "epoch": 1.5330667047564632,
      "grad_norm": 0.5179547667503357,
      "learning_rate": 9.856076637117529e-05,
      "loss": 0.161,
      "step": 16100
    },
    {
      "epoch": 1.5378279293434272,
      "grad_norm": 0.5846447348594666,
      "learning_rate": 9.85560003812792e-05,
      "loss": 0.157,
      "step": 16150
    },
    {
      "epoch": 1.5425891539303909,
      "grad_norm": 0.41618812084198,
      "learning_rate": 9.855123439138309e-05,
      "loss": 0.1604,
      "step": 16200
    },
    {
      "epoch": 1.5473503785173546,
      "grad_norm": 0.5047721862792969,
      "learning_rate": 9.854646840148699e-05,
      "loss": 0.1607,
      "step": 16250
    },
    {
      "epoch": 1.5521116031043185,
      "grad_norm": 0.5991541147232056,
      "learning_rate": 9.854170241159089e-05,
      "loss": 0.1596,
      "step": 16300
    },
    {
      "epoch": 1.5568728276912822,
      "grad_norm": 0.43497374653816223,
      "learning_rate": 9.853693642169479e-05,
      "loss": 0.1577,
      "step": 16350
    },
    {
      "epoch": 1.561634052278246,
      "grad_norm": 0.40854403376579285,
      "learning_rate": 9.853217043179869e-05,
      "loss": 0.1599,
      "step": 16400
    },
    {
      "epoch": 1.5663952768652099,
      "grad_norm": 0.4918048679828644,
      "learning_rate": 9.852740444190259e-05,
      "loss": 0.1603,
      "step": 16450
    },
    {
      "epoch": 1.5711565014521733,
      "grad_norm": 0.3920314908027649,
      "learning_rate": 9.852263845200649e-05,
      "loss": 0.1573,
      "step": 16500
    },
    {
      "epoch": 1.5759177260391373,
      "grad_norm": 0.3981036841869354,
      "learning_rate": 9.851787246211039e-05,
      "loss": 0.1586,
      "step": 16550
    },
    {
      "epoch": 1.5806789506261012,
      "grad_norm": 0.6724033355712891,
      "learning_rate": 9.851310647221427e-05,
      "loss": 0.1577,
      "step": 16600
    },
    {
      "epoch": 1.5854401752130647,
      "grad_norm": 0.45915940403938293,
      "learning_rate": 9.850834048231819e-05,
      "loss": 0.1634,
      "step": 16650
    },
    {
      "epoch": 1.5902013998000286,
      "grad_norm": 0.4456646740436554,
      "learning_rate": 9.850357449242207e-05,
      "loss": 0.1598,
      "step": 16700
    },
    {
      "epoch": 1.5949626243869923,
      "grad_norm": 0.4617702066898346,
      "learning_rate": 9.849880850252597e-05,
      "loss": 0.1588,
      "step": 16750
    },
    {
      "epoch": 1.599723848973956,
      "grad_norm": 0.4858147203922272,
      "learning_rate": 9.849404251262989e-05,
      "loss": 0.1603,
      "step": 16800
    },
    {
      "epoch": 1.60448507356092,
      "grad_norm": 0.5424185395240784,
      "learning_rate": 9.848927652273377e-05,
      "loss": 0.1588,
      "step": 16850
    },
    {
      "epoch": 1.6092462981478837,
      "grad_norm": 0.432170033454895,
      "learning_rate": 9.848451053283767e-05,
      "loss": 0.1593,
      "step": 16900
    },
    {
      "epoch": 1.6140075227348474,
      "grad_norm": 0.403414249420166,
      "learning_rate": 9.847974454294157e-05,
      "loss": 0.1564,
      "step": 16950
    },
    {
      "epoch": 1.6187687473218113,
      "grad_norm": 0.5686953067779541,
      "learning_rate": 9.847497855304547e-05,
      "loss": 0.1586,
      "step": 17000
    },
    {
      "epoch": 1.6187687473218113,
      "eval_loss": 0.15684476494789124,
      "eval_mae": 0.6601889133453369,
      "eval_mse": 399.5775146484375,
      "eval_rmse": 19.989435075770338,
      "eval_runtime": 60.8742,
      "eval_samples_per_second": 9813.93,
      "eval_smape": 88.6057436466217,
      "eval_steps_per_second": 19.171,
      "step": 17000
    },
    {
      "epoch": 1.6235299719087748,
      "grad_norm": 0.472042977809906,
      "learning_rate": 9.847021256314937e-05,
      "loss": 0.1585,
      "step": 17050
    },
    {
      "epoch": 1.6282911964957387,
      "grad_norm": 0.6247090697288513,
      "learning_rate": 9.846544657325327e-05,
      "loss": 0.1583,
      "step": 17100
    },
    {
      "epoch": 1.6330524210827024,
      "grad_norm": 0.4843044579029083,
      "learning_rate": 9.846068058335717e-05,
      "loss": 0.1592,
      "step": 17150
    },
    {
      "epoch": 1.6378136456696661,
      "grad_norm": 0.48187774419784546,
      "learning_rate": 9.845591459346106e-05,
      "loss": 0.1592,
      "step": 17200
    },
    {
      "epoch": 1.64257487025663,
      "grad_norm": 0.4763176441192627,
      "learning_rate": 9.845114860356497e-05,
      "loss": 0.1592,
      "step": 17250
    },
    {
      "epoch": 1.6473360948435938,
      "grad_norm": 0.5375500321388245,
      "learning_rate": 9.844638261366887e-05,
      "loss": 0.159,
      "step": 17300
    },
    {
      "epoch": 1.6520973194305575,
      "grad_norm": 0.4639647305011749,
      "learning_rate": 9.844161662377276e-05,
      "loss": 0.1578,
      "step": 17350
    },
    {
      "epoch": 1.6568585440175214,
      "grad_norm": 0.44177523255348206,
      "learning_rate": 9.843685063387667e-05,
      "loss": 0.1593,
      "step": 17400
    },
    {
      "epoch": 1.6616197686044851,
      "grad_norm": 0.5092636942863464,
      "learning_rate": 9.843208464398055e-05,
      "loss": 0.161,
      "step": 17450
    },
    {
      "epoch": 1.6663809931914488,
      "grad_norm": 0.4873998165130615,
      "learning_rate": 9.842731865408445e-05,
      "loss": 0.1584,
      "step": 17500
    },
    {
      "epoch": 1.6711422177784128,
      "grad_norm": 0.622734546661377,
      "learning_rate": 9.842255266418837e-05,
      "loss": 0.1597,
      "step": 17550
    },
    {
      "epoch": 1.6759034423653763,
      "grad_norm": 0.5134626030921936,
      "learning_rate": 9.841778667429225e-05,
      "loss": 0.1591,
      "step": 17600
    },
    {
      "epoch": 1.6806646669523402,
      "grad_norm": 0.42936229705810547,
      "learning_rate": 9.841302068439615e-05,
      "loss": 0.1575,
      "step": 17650
    },
    {
      "epoch": 1.6854258915393039,
      "grad_norm": 0.44326427578926086,
      "learning_rate": 9.840825469450005e-05,
      "loss": 0.1568,
      "step": 17700
    },
    {
      "epoch": 1.6901871161262676,
      "grad_norm": 0.3937668204307556,
      "learning_rate": 9.840348870460395e-05,
      "loss": 0.1593,
      "step": 17750
    },
    {
      "epoch": 1.6949483407132315,
      "grad_norm": 0.45583653450012207,
      "learning_rate": 9.839872271470785e-05,
      "loss": 0.157,
      "step": 17800
    },
    {
      "epoch": 1.6997095653001952,
      "grad_norm": 0.5357958078384399,
      "learning_rate": 9.839395672481175e-05,
      "loss": 0.1604,
      "step": 17850
    },
    {
      "epoch": 1.704470789887159,
      "grad_norm": 0.421678751707077,
      "learning_rate": 9.838919073491565e-05,
      "loss": 0.1598,
      "step": 17900
    },
    {
      "epoch": 1.7092320144741229,
      "grad_norm": 0.42326951026916504,
      "learning_rate": 9.838442474501954e-05,
      "loss": 0.157,
      "step": 17950
    },
    {
      "epoch": 1.7139932390610864,
      "grad_norm": 0.43803659081459045,
      "learning_rate": 9.837965875512344e-05,
      "loss": 0.1593,
      "step": 18000
    },
    {
      "epoch": 1.7139932390610864,
      "eval_loss": 0.15651458501815796,
      "eval_mae": 0.6603885889053345,
      "eval_mse": 359.5630187988281,
      "eval_rmse": 18.962146998660995,
      "eval_runtime": 60.8886,
      "eval_samples_per_second": 9811.606,
      "eval_smape": 325.50642490386963,
      "eval_steps_per_second": 19.166,
      "step": 18000
    },
    {
      "epoch": 1.7187544636480503,
      "grad_norm": 0.42728474736213684,
      "learning_rate": 9.837489276522735e-05,
      "loss": 0.1588,
      "step": 18050
    },
    {
      "epoch": 1.723515688235014,
      "grad_norm": 0.4929046928882599,
      "learning_rate": 9.837012677533124e-05,
      "loss": 0.1587,
      "step": 18100
    },
    {
      "epoch": 1.7282769128219777,
      "grad_norm": 0.47886890172958374,
      "learning_rate": 9.836536078543513e-05,
      "loss": 0.1572,
      "step": 18150
    },
    {
      "epoch": 1.7330381374089416,
      "grad_norm": 0.4829745590686798,
      "learning_rate": 9.836059479553903e-05,
      "loss": 0.1575,
      "step": 18200
    },
    {
      "epoch": 1.7377993619959053,
      "grad_norm": 0.4256702661514282,
      "learning_rate": 9.835582880564293e-05,
      "loss": 0.1575,
      "step": 18250
    },
    {
      "epoch": 1.742560586582869,
      "grad_norm": 0.4803942143917084,
      "learning_rate": 9.835106281574683e-05,
      "loss": 0.1559,
      "step": 18300
    },
    {
      "epoch": 1.747321811169833,
      "grad_norm": 0.5108392834663391,
      "learning_rate": 9.834629682585073e-05,
      "loss": 0.1558,
      "step": 18350
    },
    {
      "epoch": 1.7520830357567967,
      "grad_norm": 0.5541846752166748,
      "learning_rate": 9.834153083595463e-05,
      "loss": 0.1571,
      "step": 18400
    },
    {
      "epoch": 1.7568442603437604,
      "grad_norm": 0.3996049463748932,
      "learning_rate": 9.833676484605853e-05,
      "loss": 0.1588,
      "step": 18450
    },
    {
      "epoch": 1.7616054849307243,
      "grad_norm": 0.4311594069004059,
      "learning_rate": 9.833199885616243e-05,
      "loss": 0.1566,
      "step": 18500
    },
    {
      "epoch": 1.7663667095176878,
      "grad_norm": 0.45717188715934753,
      "learning_rate": 9.832723286626633e-05,
      "loss": 0.1586,
      "step": 18550
    },
    {
      "epoch": 1.7711279341046517,
      "grad_norm": 0.48945853114128113,
      "learning_rate": 9.832246687637022e-05,
      "loss": 0.1581,
      "step": 18600
    },
    {
      "epoch": 1.7758891586916155,
      "grad_norm": 0.4877380430698395,
      "learning_rate": 9.831770088647413e-05,
      "loss": 0.1594,
      "step": 18650
    },
    {
      "epoch": 1.7806503832785792,
      "grad_norm": 0.47437113523483276,
      "learning_rate": 9.831293489657803e-05,
      "loss": 0.1566,
      "step": 18700
    },
    {
      "epoch": 1.785411607865543,
      "grad_norm": 0.6995478272438049,
      "learning_rate": 9.830816890668192e-05,
      "loss": 0.1564,
      "step": 18750
    },
    {
      "epoch": 1.7901728324525068,
      "grad_norm": 0.4471156895160675,
      "learning_rate": 9.830340291678583e-05,
      "loss": 0.1567,
      "step": 18800
    },
    {
      "epoch": 1.7949340570394705,
      "grad_norm": 0.459011435508728,
      "learning_rate": 9.829863692688972e-05,
      "loss": 0.1599,
      "step": 18850
    },
    {
      "epoch": 1.7996952816264344,
      "grad_norm": 0.4757770001888275,
      "learning_rate": 9.829387093699362e-05,
      "loss": 0.1574,
      "step": 18900
    },
    {
      "epoch": 1.804456506213398,
      "grad_norm": 0.4678910970687866,
      "learning_rate": 9.828910494709751e-05,
      "loss": 0.1604,
      "step": 18950
    },
    {
      "epoch": 1.8092177308003619,
      "grad_norm": 0.4852876663208008,
      "learning_rate": 9.828433895720141e-05,
      "loss": 0.1562,
      "step": 19000
    },
    {
      "epoch": 1.8092177308003619,
      "eval_loss": 0.15658971667289734,
      "eval_mae": 0.6544845700263977,
      "eval_mse": 281.27392578125,
      "eval_rmse": 16.77122314505564,
      "eval_runtime": 61.2247,
      "eval_samples_per_second": 9757.737,
      "eval_smape": 80.4527759552002,
      "eval_steps_per_second": 19.061,
      "step": 19000
    },
    {
      "epoch": 1.8139789553873258,
      "grad_norm": 0.37452152371406555,
      "learning_rate": 9.827957296730531e-05,
      "loss": 0.1561,
      "step": 19050
    },
    {
      "epoch": 1.8187401799742893,
      "grad_norm": 0.4231972396373749,
      "learning_rate": 9.827480697740921e-05,
      "loss": 0.1588,
      "step": 19100
    },
    {
      "epoch": 1.8235014045612532,
      "grad_norm": 0.423755943775177,
      "learning_rate": 9.827004098751311e-05,
      "loss": 0.1566,
      "step": 19150
    },
    {
      "epoch": 1.828262629148217,
      "grad_norm": 0.4381329119205475,
      "learning_rate": 9.826527499761701e-05,
      "loss": 0.1596,
      "step": 19200
    },
    {
      "epoch": 1.8330238537351806,
      "grad_norm": 0.4936196804046631,
      "learning_rate": 9.826050900772091e-05,
      "loss": 0.1598,
      "step": 19250
    },
    {
      "epoch": 1.8377850783221445,
      "grad_norm": 0.6386341452598572,
      "learning_rate": 9.825574301782481e-05,
      "loss": 0.1601,
      "step": 19300
    },
    {
      "epoch": 1.8425463029091083,
      "grad_norm": 0.49802061915397644,
      "learning_rate": 9.82509770279287e-05,
      "loss": 0.1587,
      "step": 19350
    },
    {
      "epoch": 1.847307527496072,
      "grad_norm": 0.4627436697483063,
      "learning_rate": 9.824621103803261e-05,
      "loss": 0.1587,
      "step": 19400
    },
    {
      "epoch": 1.852068752083036,
      "grad_norm": 0.4635562598705292,
      "learning_rate": 9.824144504813651e-05,
      "loss": 0.1532,
      "step": 19450
    },
    {
      "epoch": 1.8568299766699994,
      "grad_norm": 0.4869281053543091,
      "learning_rate": 9.82366790582404e-05,
      "loss": 0.1576,
      "step": 19500
    },
    {
      "epoch": 1.8615912012569633,
      "grad_norm": 0.4622304141521454,
      "learning_rate": 9.82319130683443e-05,
      "loss": 0.1587,
      "step": 19550
    },
    {
      "epoch": 1.866352425843927,
      "grad_norm": 0.42829111218452454,
      "learning_rate": 9.82271470784482e-05,
      "loss": 0.1584,
      "step": 19600
    },
    {
      "epoch": 1.8711136504308907,
      "grad_norm": 0.44152921438217163,
      "learning_rate": 9.82223810885521e-05,
      "loss": 0.1597,
      "step": 19650
    },
    {
      "epoch": 1.8758748750178547,
      "grad_norm": 0.4362374246120453,
      "learning_rate": 9.8217615098656e-05,
      "loss": 0.1535,
      "step": 19700
    },
    {
      "epoch": 1.8806360996048184,
      "grad_norm": 0.42130181193351746,
      "learning_rate": 9.82128491087599e-05,
      "loss": 0.1585,
      "step": 19750
    },
    {
      "epoch": 1.885397324191782,
      "grad_norm": 0.5120296478271484,
      "learning_rate": 9.82080831188638e-05,
      "loss": 0.1583,
      "step": 19800
    },
    {
      "epoch": 1.890158548778746,
      "grad_norm": 0.4205983579158783,
      "learning_rate": 9.820331712896768e-05,
      "loss": 0.1585,
      "step": 19850
    },
    {
      "epoch": 1.8949197733657097,
      "grad_norm": 0.4189218580722809,
      "learning_rate": 9.819855113907159e-05,
      "loss": 0.1561,
      "step": 19900
    },
    {
      "epoch": 1.8996809979526734,
      "grad_norm": 0.6145504117012024,
      "learning_rate": 9.819378514917549e-05,
      "loss": 0.1625,
      "step": 19950
    },
    {
      "epoch": 1.9044422225396374,
      "grad_norm": 0.5396240949630737,
      "learning_rate": 9.818901915927938e-05,
      "loss": 0.1601,
      "step": 20000
    },
    {
      "epoch": 1.9044422225396374,
      "eval_loss": 0.15698538720607758,
      "eval_mae": 0.6542770862579346,
      "eval_mse": 287.357666015625,
      "eval_rmse": 16.951627237985885,
      "eval_runtime": 58.5657,
      "eval_samples_per_second": 10200.761,
      "eval_smape": 79.55442667007446,
      "eval_steps_per_second": 19.926,
      "step": 20000
    },
    {
      "epoch": 1.9092034471266008,
      "grad_norm": 0.45280078053474426,
      "learning_rate": 9.818425316938329e-05,
      "loss": 0.1593,
      "step": 20050
    },
    {
      "epoch": 1.9139646717135648,
      "grad_norm": 0.455552339553833,
      "learning_rate": 9.817948717948718e-05,
      "loss": 0.1579,
      "step": 20100
    },
    {
      "epoch": 1.9187258963005285,
      "grad_norm": 0.44774943590164185,
      "learning_rate": 9.817472118959108e-05,
      "loss": 0.1577,
      "step": 20150
    },
    {
      "epoch": 1.9234871208874922,
      "grad_norm": 0.4146822392940521,
      "learning_rate": 9.816995519969499e-05,
      "loss": 0.1567,
      "step": 20200
    },
    {
      "epoch": 1.9282483454744561,
      "grad_norm": 0.48615461587905884,
      "learning_rate": 9.816518920979888e-05,
      "loss": 0.1596,
      "step": 20250
    },
    {
      "epoch": 1.9330095700614198,
      "grad_norm": 0.5081954598426819,
      "learning_rate": 9.816042321990278e-05,
      "loss": 0.1576,
      "step": 20300
    },
    {
      "epoch": 1.9377707946483835,
      "grad_norm": 0.5155813097953796,
      "learning_rate": 9.815565723000668e-05,
      "loss": 0.1551,
      "step": 20350
    },
    {
      "epoch": 1.9425320192353475,
      "grad_norm": 0.5232491493225098,
      "learning_rate": 9.815089124011058e-05,
      "loss": 0.1597,
      "step": 20400
    },
    {
      "epoch": 1.947293243822311,
      "grad_norm": 0.4078335762023926,
      "learning_rate": 9.814612525021447e-05,
      "loss": 0.1556,
      "step": 20450
    },
    {
      "epoch": 1.9520544684092749,
      "grad_norm": 0.4975152611732483,
      "learning_rate": 9.814135926031837e-05,
      "loss": 0.1591,
      "step": 20500
    },
    {
      "epoch": 1.9568156929962386,
      "grad_norm": 0.6481941938400269,
      "learning_rate": 9.813659327042227e-05,
      "loss": 0.1565,
      "step": 20550
    },
    {
      "epoch": 1.9615769175832023,
      "grad_norm": 0.4526354968547821,
      "learning_rate": 9.813182728052617e-05,
      "loss": 0.1546,
      "step": 20600
    },
    {
      "epoch": 1.9663381421701662,
      "grad_norm": 0.4288252294063568,
      "learning_rate": 9.812706129063007e-05,
      "loss": 0.1551,
      "step": 20650
    },
    {
      "epoch": 1.97109936675713,
      "grad_norm": 0.434865266084671,
      "learning_rate": 9.812229530073397e-05,
      "loss": 0.1557,
      "step": 20700
    },
    {
      "epoch": 1.9758605913440936,
      "grad_norm": 0.4131234586238861,
      "learning_rate": 9.811752931083786e-05,
      "loss": 0.156,
      "step": 20750
    },
    {
      "epoch": 1.9806218159310576,
      "grad_norm": 0.519029438495636,
      "learning_rate": 9.811276332094177e-05,
      "loss": 0.1596,
      "step": 20800
    },
    {
      "epoch": 1.9853830405180213,
      "grad_norm": 0.4048980474472046,
      "learning_rate": 9.810799733104566e-05,
      "loss": 0.153,
      "step": 20850
    },
    {
      "epoch": 1.990144265104985,
      "grad_norm": 0.44951915740966797,
      "learning_rate": 9.810323134114956e-05,
      "loss": 0.1562,
      "step": 20900
    },
    {
      "epoch": 1.994905489691949,
      "grad_norm": 0.40820789337158203,
      "learning_rate": 9.809846535125346e-05,
      "loss": 0.1576,
      "step": 20950
    },
    {
      "epoch": 1.9996667142789124,
      "grad_norm": 0.6201032400131226,
      "learning_rate": 9.809369936135736e-05,
      "loss": 0.1551,
      "step": 21000
    },
    {
      "epoch": 1.9996667142789124,
      "eval_loss": 0.15608514845371246,
      "eval_mae": 0.6444294452667236,
      "eval_mse": 279.2149658203125,
      "eval_rmse": 16.709726682992528,
      "eval_runtime": 54.4875,
      "eval_samples_per_second": 10964.25,
      "eval_smape": 102.60157585144043,
      "eval_steps_per_second": 21.418,
      "step": 21000
    },
    {
      "epoch": 2.004380326620007,
      "grad_norm": 0.4798502027988434,
      "learning_rate": 9.808893337146126e-05,
      "loss": 0.1558,
      "step": 21050
    },
    {
      "epoch": 2.0091415512069704,
      "grad_norm": 0.47718650102615356,
      "learning_rate": 9.808416738156516e-05,
      "loss": 0.1581,
      "step": 21100
    },
    {
      "epoch": 2.0139027757939343,
      "grad_norm": 0.4789866507053375,
      "learning_rate": 9.807940139166906e-05,
      "loss": 0.1558,
      "step": 21150
    },
    {
      "epoch": 2.0186640003808978,
      "grad_norm": 0.5219734907150269,
      "learning_rate": 9.807463540177296e-05,
      "loss": 0.1564,
      "step": 21200
    },
    {
      "epoch": 2.0234252249678617,
      "grad_norm": 0.4981229305267334,
      "learning_rate": 9.806986941187684e-05,
      "loss": 0.1583,
      "step": 21250
    },
    {
      "epoch": 2.0281864495548256,
      "grad_norm": 0.4696637690067291,
      "learning_rate": 9.806510342198075e-05,
      "loss": 0.1568,
      "step": 21300
    },
    {
      "epoch": 2.032947674141789,
      "grad_norm": 0.44064363837242126,
      "learning_rate": 9.806033743208465e-05,
      "loss": 0.1579,
      "step": 21350
    },
    {
      "epoch": 2.037708898728753,
      "grad_norm": 0.5143409967422485,
      "learning_rate": 9.805557144218854e-05,
      "loss": 0.156,
      "step": 21400
    },
    {
      "epoch": 2.042470123315717,
      "grad_norm": 0.49321916699409485,
      "learning_rate": 9.805080545229245e-05,
      "loss": 0.155,
      "step": 21450
    },
    {
      "epoch": 2.0472313479026805,
      "grad_norm": 0.4427430331707001,
      "learning_rate": 9.804603946239634e-05,
      "loss": 0.1566,
      "step": 21500
    },
    {
      "epoch": 2.0519925724896444,
      "grad_norm": 0.45803865790367126,
      "learning_rate": 9.804127347250024e-05,
      "loss": 0.1593,
      "step": 21550
    },
    {
      "epoch": 2.056753797076608,
      "grad_norm": 0.4853162467479706,
      "learning_rate": 9.803650748260415e-05,
      "loss": 0.1575,
      "step": 21600
    },
    {
      "epoch": 2.061515021663572,
      "grad_norm": 0.47955945134162903,
      "learning_rate": 9.803174149270804e-05,
      "loss": 0.1594,
      "step": 21650
    },
    {
      "epoch": 2.0662762462505357,
      "grad_norm": 0.44243359565734863,
      "learning_rate": 9.802697550281194e-05,
      "loss": 0.1597,
      "step": 21700
    },
    {
      "epoch": 2.0710374708374992,
      "grad_norm": 0.5727405548095703,
      "learning_rate": 9.802220951291584e-05,
      "loss": 0.1576,
      "step": 21750
    },
    {
      "epoch": 2.075798695424463,
      "grad_norm": 0.5481444001197815,
      "learning_rate": 9.801744352301974e-05,
      "loss": 0.1547,
      "step": 21800
    },
    {
      "epoch": 2.080559920011427,
      "grad_norm": 0.46369367837905884,
      "learning_rate": 9.801267753312364e-05,
      "loss": 0.1553,
      "step": 21850
    },
    {
      "epoch": 2.0853211445983906,
      "grad_norm": 0.48963698744773865,
      "learning_rate": 9.800791154322754e-05,
      "loss": 0.1579,
      "step": 21900
    },
    {
      "epoch": 2.0900823691853545,
      "grad_norm": 0.5048883557319641,
      "learning_rate": 9.800314555333144e-05,
      "loss": 0.1597,
      "step": 21950
    },
    {
      "epoch": 2.0948435937723184,
      "grad_norm": 0.4893467128276825,
      "learning_rate": 9.799837956343532e-05,
      "loss": 0.1532,
      "step": 22000
    },
    {
      "epoch": 2.0948435937723184,
      "eval_loss": 0.1553574949502945,
      "eval_mae": 0.6453903317451477,
      "eval_mse": 282.9573974609375,
      "eval_rmse": 16.821337564561787,
      "eval_runtime": 60.2846,
      "eval_samples_per_second": 9909.912,
      "eval_smape": 85.01211404800415,
      "eval_steps_per_second": 19.358,
      "step": 22000
    },
    {
      "epoch": 2.099604818359282,
      "grad_norm": 0.46226996183395386,
      "learning_rate": 9.799361357353923e-05,
      "loss": 0.1564,
      "step": 22050
    },
    {
      "epoch": 2.104366042946246,
      "grad_norm": 0.46276605129241943,
      "learning_rate": 9.798884758364313e-05,
      "loss": 0.1562,
      "step": 22100
    },
    {
      "epoch": 2.1091272675332093,
      "grad_norm": 0.4646179974079132,
      "learning_rate": 9.798408159374702e-05,
      "loss": 0.1576,
      "step": 22150
    },
    {
      "epoch": 2.1138884921201733,
      "grad_norm": 0.4897124171257019,
      "learning_rate": 9.797931560385093e-05,
      "loss": 0.1568,
      "step": 22200
    },
    {
      "epoch": 2.118649716707137,
      "grad_norm": 0.43058913946151733,
      "learning_rate": 9.797454961395482e-05,
      "loss": 0.1579,
      "step": 22250
    },
    {
      "epoch": 2.1234109412941007,
      "grad_norm": 0.37743493914604187,
      "learning_rate": 9.796978362405872e-05,
      "loss": 0.1578,
      "step": 22300
    },
    {
      "epoch": 2.1281721658810646,
      "grad_norm": 0.4362202286720276,
      "learning_rate": 9.796501763416263e-05,
      "loss": 0.1579,
      "step": 22350
    },
    {
      "epoch": 2.1329333904680285,
      "grad_norm": 0.3836491107940674,
      "learning_rate": 9.796025164426652e-05,
      "loss": 0.1576,
      "step": 22400
    },
    {
      "epoch": 2.137694615054992,
      "grad_norm": 0.4960291385650635,
      "learning_rate": 9.795548565437042e-05,
      "loss": 0.1563,
      "step": 22450
    },
    {
      "epoch": 2.142455839641956,
      "grad_norm": 0.43519532680511475,
      "learning_rate": 9.795071966447432e-05,
      "loss": 0.1598,
      "step": 22500
    },
    {
      "epoch": 2.1472170642289194,
      "grad_norm": 0.4592967629432678,
      "learning_rate": 9.794595367457822e-05,
      "loss": 0.158,
      "step": 22550
    },
    {
      "epoch": 2.1519782888158834,
      "grad_norm": 0.5372556447982788,
      "learning_rate": 9.794118768468212e-05,
      "loss": 0.1591,
      "step": 22600
    },
    {
      "epoch": 2.1567395134028473,
      "grad_norm": 0.40967512130737305,
      "learning_rate": 9.7936421694786e-05,
      "loss": 0.159,
      "step": 22650
    },
    {
      "epoch": 2.161500737989811,
      "grad_norm": 0.4951903223991394,
      "learning_rate": 9.793165570488992e-05,
      "loss": 0.1559,
      "step": 22700
    },
    {
      "epoch": 2.1662619625767747,
      "grad_norm": 0.467052698135376,
      "learning_rate": 9.79268897149938e-05,
      "loss": 0.1612,
      "step": 22750
    },
    {
      "epoch": 2.1710231871637387,
      "grad_norm": 0.42239800095558167,
      "learning_rate": 9.79221237250977e-05,
      "loss": 0.1597,
      "step": 22800
    },
    {
      "epoch": 2.175784411750702,
      "grad_norm": 0.4856722354888916,
      "learning_rate": 9.791735773520161e-05,
      "loss": 0.1572,
      "step": 22850
    },
    {
      "epoch": 2.180545636337666,
      "grad_norm": 0.47815021872520447,
      "learning_rate": 9.79125917453055e-05,
      "loss": 0.1607,
      "step": 22900
    },
    {
      "epoch": 2.18530686092463,
      "grad_norm": 0.5437564849853516,
      "learning_rate": 9.79078257554094e-05,
      "loss": 0.1587,
      "step": 22950
    },
    {
      "epoch": 2.1900680855115935,
      "grad_norm": 0.42942744493484497,
      "learning_rate": 9.79030597655133e-05,
      "loss": 0.1564,
      "step": 23000
    },
    {
      "epoch": 2.1900680855115935,
      "eval_loss": 0.15535122156143188,
      "eval_mae": 0.6484833359718323,
      "eval_mse": 332.3757629394531,
      "eval_rmse": 18.231175577549934,
      "eval_runtime": 56.6725,
      "eval_samples_per_second": 10541.528,
      "eval_smape": 76.03496313095093,
      "eval_steps_per_second": 20.592,
      "step": 23000
    },
    {
      "epoch": 2.1948293100985574,
      "grad_norm": 0.5147453546524048,
      "learning_rate": 9.78982937756172e-05,
      "loss": 0.1584,
      "step": 23050
    },
    {
      "epoch": 2.1995905346855213,
      "grad_norm": 0.46314072608947754,
      "learning_rate": 9.78935277857211e-05,
      "loss": 0.1585,
      "step": 23100
    },
    {
      "epoch": 2.204351759272485,
      "grad_norm": 0.5470126867294312,
      "learning_rate": 9.7888761795825e-05,
      "loss": 0.1594,
      "step": 23150
    },
    {
      "epoch": 2.2091129838594488,
      "grad_norm": 0.5074204802513123,
      "learning_rate": 9.78839958059289e-05,
      "loss": 0.1576,
      "step": 23200
    },
    {
      "epoch": 2.2138742084464123,
      "grad_norm": 0.3913464844226837,
      "learning_rate": 9.78792298160328e-05,
      "loss": 0.1564,
      "step": 23250
    },
    {
      "epoch": 2.218635433033376,
      "grad_norm": 0.5663050413131714,
      "learning_rate": 9.78744638261367e-05,
      "loss": 0.1607,
      "step": 23300
    },
    {
      "epoch": 2.22339665762034,
      "grad_norm": 0.5126326084136963,
      "learning_rate": 9.78696978362406e-05,
      "loss": 0.1575,
      "step": 23350
    },
    {
      "epoch": 2.2281578822073036,
      "grad_norm": 0.531508207321167,
      "learning_rate": 9.786493184634448e-05,
      "loss": 0.1552,
      "step": 23400
    },
    {
      "epoch": 2.2329191067942675,
      "grad_norm": 0.42278382182121277,
      "learning_rate": 9.78601658564484e-05,
      "loss": 0.1563,
      "step": 23450
    },
    {
      "epoch": 2.2376803313812315,
      "grad_norm": 0.4637637436389923,
      "learning_rate": 9.78553998665523e-05,
      "loss": 0.1589,
      "step": 23500
    },
    {
      "epoch": 2.242441555968195,
      "grad_norm": 0.4949529767036438,
      "learning_rate": 9.785063387665618e-05,
      "loss": 0.1572,
      "step": 23550
    },
    {
      "epoch": 2.247202780555159,
      "grad_norm": 0.5856890678405762,
      "learning_rate": 9.78458678867601e-05,
      "loss": 0.1571,
      "step": 23600
    },
    {
      "epoch": 2.2519640051421224,
      "grad_norm": 0.6656904220581055,
      "learning_rate": 9.784110189686398e-05,
      "loss": 0.1578,
      "step": 23650
    },
    {
      "epoch": 2.2567252297290863,
      "grad_norm": 0.49895474314689636,
      "learning_rate": 9.783633590696788e-05,
      "loss": 0.1544,
      "step": 23700
    },
    {
      "epoch": 2.26148645431605,
      "grad_norm": 0.5674960017204285,
      "learning_rate": 9.783156991707178e-05,
      "loss": 0.1586,
      "step": 23750
    },
    {
      "epoch": 2.2662476789030137,
      "grad_norm": 0.6098377108573914,
      "learning_rate": 9.782680392717568e-05,
      "loss": 0.1563,
      "step": 23800
    },
    {
      "epoch": 2.2710089034899776,
      "grad_norm": 0.5532990097999573,
      "learning_rate": 9.782203793727958e-05,
      "loss": 0.1558,
      "step": 23850
    },
    {
      "epoch": 2.2757701280769416,
      "grad_norm": 0.44679224491119385,
      "learning_rate": 9.781727194738346e-05,
      "loss": 0.1582,
      "step": 23900
    },
    {
      "epoch": 2.280531352663905,
      "grad_norm": 0.5477356314659119,
      "learning_rate": 9.781250595748738e-05,
      "loss": 0.1583,
      "step": 23950
    },
    {
      "epoch": 2.285292577250869,
      "grad_norm": 0.5364423990249634,
      "learning_rate": 9.780773996759128e-05,
      "loss": 0.1568,
      "step": 24000
    },
    {
      "epoch": 2.285292577250869,
      "eval_loss": 0.15508916974067688,
      "eval_mae": 0.6528403162956238,
      "eval_mse": 356.0440979003906,
      "eval_rmse": 18.869130819950097,
      "eval_runtime": 55.4778,
      "eval_samples_per_second": 10768.544,
      "eval_smape": 92.25972294807434,
      "eval_steps_per_second": 21.035,
      "step": 24000
    }
  ],
  "logging_steps": 50,
  "max_steps": 1050100,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 2000,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 3,
        "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": 4.560071489513472e+16,
  "train_batch_size": 256,
  "trial_name": null,
  "trial_params": null
}