{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 46.61787329262039,
  "global_step": 1000000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.02,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 1.6282,
      "step": 500
    },
    {
      "epoch": 0.05,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 1.5335,
      "step": 1000
    },
    {
      "epoch": 0.07,
      "learning_rate": 2.9980000000000003e-06,
      "loss": 1.4907,
      "step": 1500
    },
    {
      "epoch": 0.09,
      "learning_rate": 3.9980000000000005e-06,
      "loss": 1.462,
      "step": 2000
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.998e-06,
      "loss": 1.4412,
      "step": 2500
    },
    {
      "epoch": 0.14,
      "learning_rate": 5.998000000000001e-06,
      "loss": 1.4227,
      "step": 3000
    },
    {
      "epoch": 0.16,
      "learning_rate": 6.998000000000001e-06,
      "loss": 1.4019,
      "step": 3500
    },
    {
      "epoch": 0.19,
      "learning_rate": 7.998e-06,
      "loss": 1.3897,
      "step": 4000
    },
    {
      "epoch": 0.21,
      "learning_rate": 8.998000000000001e-06,
      "loss": 1.3773,
      "step": 4500
    },
    {
      "epoch": 0.23,
      "learning_rate": 9.998000000000002e-06,
      "loss": 1.365,
      "step": 5000
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.0998000000000002e-05,
      "loss": 1.3594,
      "step": 5500
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.1998000000000001e-05,
      "loss": 1.3516,
      "step": 6000
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.2998000000000002e-05,
      "loss": 1.3443,
      "step": 6500
    },
    {
      "epoch": 0.33,
      "learning_rate": 1.3996000000000001e-05,
      "loss": 1.3378,
      "step": 7000
    },
    {
      "epoch": 0.35,
      "learning_rate": 1.4996000000000002e-05,
      "loss": 1.332,
      "step": 7500
    },
    {
      "epoch": 0.37,
      "learning_rate": 1.5996e-05,
      "loss": 1.3253,
      "step": 8000
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.6996e-05,
      "loss": 1.3209,
      "step": 8500
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.7996000000000002e-05,
      "loss": 1.3159,
      "step": 9000
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.8996000000000002e-05,
      "loss": 1.3121,
      "step": 9500
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9996000000000003e-05,
      "loss": 1.3097,
      "step": 10000
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.999611562641374e-05,
      "loss": 1.3029,
      "step": 10500
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.999221565293356e-05,
      "loss": 1.2979,
      "step": 11000
    },
    {
      "epoch": 0.54,
      "learning_rate": 1.9988323479400342e-05,
      "loss": 1.2988,
      "step": 11500
    },
    {
      "epoch": 0.56,
      "learning_rate": 1.998442350592016e-05,
      "loss": 1.2911,
      "step": 12000
    },
    {
      "epoch": 0.58,
      "learning_rate": 1.998052353243998e-05,
      "loss": 1.2904,
      "step": 12500
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.99766235589598e-05,
      "loss": 1.287,
      "step": 13000
    },
    {
      "epoch": 0.63,
      "learning_rate": 1.9972731385426582e-05,
      "loss": 1.2834,
      "step": 13500
    },
    {
      "epoch": 0.65,
      "learning_rate": 1.9968831411946402e-05,
      "loss": 1.2817,
      "step": 14000
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.996493143846622e-05,
      "loss": 1.2766,
      "step": 14500
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.996103146498604e-05,
      "loss": 1.278,
      "step": 15000
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9957131491505857e-05,
      "loss": 1.2743,
      "step": 15500
    },
    {
      "epoch": 0.75,
      "learning_rate": 1.995323931797264e-05,
      "loss": 1.2662,
      "step": 16000
    },
    {
      "epoch": 0.77,
      "learning_rate": 1.994933934449246e-05,
      "loss": 1.2688,
      "step": 16500
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.994543937101228e-05,
      "loss": 1.2672,
      "step": 17000
    },
    {
      "epoch": 0.82,
      "learning_rate": 1.9941547197479058e-05,
      "loss": 1.2679,
      "step": 17500
    },
    {
      "epoch": 0.84,
      "learning_rate": 1.9937647223998878e-05,
      "loss": 1.2626,
      "step": 18000
    },
    {
      "epoch": 0.86,
      "learning_rate": 1.99337472505187e-05,
      "loss": 1.261,
      "step": 18500
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.992984727703852e-05,
      "loss": 1.2627,
      "step": 19000
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9925955103505297e-05,
      "loss": 1.258,
      "step": 19500
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.9922055130025118e-05,
      "loss": 1.2561,
      "step": 20000
    },
    {
      "epoch": 0.96,
      "learning_rate": 1.9918155156544935e-05,
      "loss": 1.2544,
      "step": 20500
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9914255183064755e-05,
      "loss": 1.2541,
      "step": 21000
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.7456339864102125,
      "eval_loss": 1.1044505834579468,
      "eval_runtime": 51.9764,
      "eval_samples_per_second": 898.638,
      "eval_steps_per_second": 14.045,
      "step": 21451
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.9910363009531537e-05,
      "loss": 1.2537,
      "step": 21500
    },
    {
      "epoch": 1.03,
      "learning_rate": 1.9906463036051354e-05,
      "loss": 1.2498,
      "step": 22000
    },
    {
      "epoch": 1.05,
      "learning_rate": 1.9902563062571178e-05,
      "loss": 1.2475,
      "step": 22500
    },
    {
      "epoch": 1.07,
      "learning_rate": 1.9898663089090995e-05,
      "loss": 1.245,
      "step": 23000
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.9894763115610816e-05,
      "loss": 1.2469,
      "step": 23500
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.9890863142130633e-05,
      "loss": 1.2407,
      "step": 24000
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.9886970968597414e-05,
      "loss": 1.2434,
      "step": 24500
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.9883070995117235e-05,
      "loss": 1.2408,
      "step": 25000
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.9879171021637055e-05,
      "loss": 1.2422,
      "step": 25500
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.9875271048156876e-05,
      "loss": 1.2386,
      "step": 26000
    },
    {
      "epoch": 1.24,
      "learning_rate": 1.9871371074676693e-05,
      "loss": 1.2404,
      "step": 26500
    },
    {
      "epoch": 1.26,
      "learning_rate": 1.9867471101196513e-05,
      "loss": 1.2371,
      "step": 27000
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.9863578927663295e-05,
      "loss": 1.237,
      "step": 27500
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.9859678954183112e-05,
      "loss": 1.2337,
      "step": 28000
    },
    {
      "epoch": 1.33,
      "learning_rate": 1.9855778980702932e-05,
      "loss": 1.2346,
      "step": 28500
    },
    {
      "epoch": 1.35,
      "learning_rate": 1.9851879007222753e-05,
      "loss": 1.2327,
      "step": 29000
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.9847979033742573e-05,
      "loss": 1.2328,
      "step": 29500
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.984407906026239e-05,
      "loss": 1.2293,
      "step": 30000
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.9840186886729172e-05,
      "loss": 1.2297,
      "step": 30500
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.9836286913248993e-05,
      "loss": 1.2326,
      "step": 31000
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.983238693976881e-05,
      "loss": 1.2277,
      "step": 31500
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.982848696628863e-05,
      "loss": 1.226,
      "step": 32000
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.982458699280845e-05,
      "loss": 1.2274,
      "step": 32500
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.982069481927523e-05,
      "loss": 1.2254,
      "step": 33000
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.9816794845795053e-05,
      "loss": 1.2219,
      "step": 33500
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.981289487231487e-05,
      "loss": 1.2225,
      "step": 34000
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.980899489883469e-05,
      "loss": 1.2248,
      "step": 34500
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.9805094925354507e-05,
      "loss": 1.2213,
      "step": 35000
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.9801194951874328e-05,
      "loss": 1.2184,
      "step": 35500
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.979731057828807e-05,
      "loss": 1.2193,
      "step": 36000
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.979341060480789e-05,
      "loss": 1.2195,
      "step": 36500
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.9789510631327708e-05,
      "loss": 1.2187,
      "step": 37000
    },
    {
      "epoch": 1.75,
      "learning_rate": 1.978561065784753e-05,
      "loss": 1.2187,
      "step": 37500
    },
    {
      "epoch": 1.77,
      "learning_rate": 1.978171068436735e-05,
      "loss": 1.2148,
      "step": 38000
    },
    {
      "epoch": 1.79,
      "learning_rate": 1.9777818510834127e-05,
      "loss": 1.2172,
      "step": 38500
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.9773918537353947e-05,
      "loss": 1.2152,
      "step": 39000
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.9770018563873768e-05,
      "loss": 1.2155,
      "step": 39500
    },
    {
      "epoch": 1.86,
      "learning_rate": 1.9766118590393585e-05,
      "loss": 1.2146,
      "step": 40000
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.9762218616913406e-05,
      "loss": 1.2145,
      "step": 40500
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9758318643433226e-05,
      "loss": 1.2124,
      "step": 41000
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.9754418669953047e-05,
      "loss": 1.2149,
      "step": 41500
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.9750526496419825e-05,
      "loss": 1.211,
      "step": 42000
    },
    {
      "epoch": 1.98,
      "learning_rate": 1.9746626522939645e-05,
      "loss": 1.2119,
      "step": 42500
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.7521904338282468,
      "eval_loss": 1.0674182176589966,
      "eval_runtime": 51.8656,
      "eval_samples_per_second": 900.559,
      "eval_steps_per_second": 14.075,
      "step": 42902
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.9742726549459466e-05,
      "loss": 1.2103,
      "step": 43000
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.9738826575979283e-05,
      "loss": 1.2087,
      "step": 43500
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.9734926602499103e-05,
      "loss": 1.2046,
      "step": 44000
    },
    {
      "epoch": 2.07,
      "learning_rate": 1.9731026629018924e-05,
      "loss": 1.2088,
      "step": 44500
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.9727134455485702e-05,
      "loss": 1.2051,
      "step": 45000
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.9723234482005526e-05,
      "loss": 1.2042,
      "step": 45500
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.9719334508525343e-05,
      "loss": 1.2057,
      "step": 46000
    },
    {
      "epoch": 2.17,
      "learning_rate": 1.9715434535045164e-05,
      "loss": 1.2071,
      "step": 46500
    },
    {
      "epoch": 2.19,
      "learning_rate": 1.971153456156498e-05,
      "loss": 1.2032,
      "step": 47000
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.9707634588084805e-05,
      "loss": 1.2026,
      "step": 47500
    },
    {
      "epoch": 2.24,
      "learning_rate": 1.9703734614604622e-05,
      "loss": 1.2028,
      "step": 48000
    },
    {
      "epoch": 2.26,
      "learning_rate": 1.9699834641124442e-05,
      "loss": 1.2047,
      "step": 48500
    },
    {
      "epoch": 2.28,
      "learning_rate": 1.969593466764426e-05,
      "loss": 1.2029,
      "step": 49000
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.969203469416408e-05,
      "loss": 1.2008,
      "step": 49500
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.96881347206839e-05,
      "loss": 1.2029,
      "step": 50000
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.968423474720372e-05,
      "loss": 1.2007,
      "step": 50500
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.9680334773723538e-05,
      "loss": 1.2012,
      "step": 51000
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.967644260019032e-05,
      "loss": 1.1965,
      "step": 51500
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.96725504266571e-05,
      "loss": 1.2006,
      "step": 52000
    },
    {
      "epoch": 2.45,
      "learning_rate": 1.966865045317692e-05,
      "loss": 1.1976,
      "step": 52500
    },
    {
      "epoch": 2.47,
      "learning_rate": 1.966475047969674e-05,
      "loss": 1.1971,
      "step": 53000
    },
    {
      "epoch": 2.49,
      "learning_rate": 1.966085050621656e-05,
      "loss": 1.199,
      "step": 53500
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.965695053273638e-05,
      "loss": 1.1955,
      "step": 54000
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.96530505592562e-05,
      "loss": 1.1936,
      "step": 54500
    },
    {
      "epoch": 2.56,
      "learning_rate": 1.9649150585776017e-05,
      "loss": 1.1933,
      "step": 55000
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.96452584122428e-05,
      "loss": 1.195,
      "step": 55500
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.964135843876262e-05,
      "loss": 1.1956,
      "step": 56000
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.9637458465282436e-05,
      "loss": 1.1917,
      "step": 56500
    },
    {
      "epoch": 2.66,
      "learning_rate": 1.9633558491802257e-05,
      "loss": 1.1943,
      "step": 57000
    },
    {
      "epoch": 2.68,
      "learning_rate": 1.9629658518322077e-05,
      "loss": 1.1939,
      "step": 57500
    },
    {
      "epoch": 2.7,
      "learning_rate": 1.9625758544841898e-05,
      "loss": 1.1908,
      "step": 58000
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.9621858571361715e-05,
      "loss": 1.1932,
      "step": 58500
    },
    {
      "epoch": 2.75,
      "learning_rate": 1.9617958597881536e-05,
      "loss": 1.193,
      "step": 59000
    },
    {
      "epoch": 2.77,
      "learning_rate": 1.9614058624401356e-05,
      "loss": 1.1909,
      "step": 59500
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.9610166450868134e-05,
      "loss": 1.191,
      "step": 60000
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.9606282077281876e-05,
      "loss": 1.1927,
      "step": 60500
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.9602382103801697e-05,
      "loss": 1.1912,
      "step": 61000
    },
    {
      "epoch": 2.87,
      "learning_rate": 1.9598482130321514e-05,
      "loss": 1.191,
      "step": 61500
    },
    {
      "epoch": 2.89,
      "learning_rate": 1.9594582156841335e-05,
      "loss": 1.1925,
      "step": 62000
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.9590682183361155e-05,
      "loss": 1.1924,
      "step": 62500
    },
    {
      "epoch": 2.94,
      "learning_rate": 1.9586782209880976e-05,
      "loss": 1.1869,
      "step": 63000
    },
    {
      "epoch": 2.96,
      "learning_rate": 1.9582882236400793e-05,
      "loss": 1.1885,
      "step": 63500
    },
    {
      "epoch": 2.98,
      "learning_rate": 1.9578982262920613e-05,
      "loss": 1.1879,
      "step": 64000
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.7559359305815577,
      "eval_loss": 1.045499563217163,
      "eval_runtime": 52.122,
      "eval_samples_per_second": 896.128,
      "eval_steps_per_second": 14.006,
      "step": 64353
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.9575082289440434e-05,
      "loss": 1.1888,
      "step": 64500
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.9571182315960254e-05,
      "loss": 1.1836,
      "step": 65000
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.9567297942373993e-05,
      "loss": 1.1835,
      "step": 65500
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.9563397968893814e-05,
      "loss": 1.1845,
      "step": 66000
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.955949799541363e-05,
      "loss": 1.184,
      "step": 66500
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.955559802193345e-05,
      "loss": 1.1846,
      "step": 67000
    },
    {
      "epoch": 3.15,
      "learning_rate": 1.9551698048453272e-05,
      "loss": 1.1838,
      "step": 67500
    },
    {
      "epoch": 3.17,
      "learning_rate": 1.9547798074973092e-05,
      "loss": 1.1826,
      "step": 68000
    },
    {
      "epoch": 3.19,
      "learning_rate": 1.954389810149291e-05,
      "loss": 1.1847,
      "step": 68500
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.953999812801273e-05,
      "loss": 1.1823,
      "step": 69000
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.953609815453255e-05,
      "loss": 1.18,
      "step": 69500
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.953219818105237e-05,
      "loss": 1.1831,
      "step": 70000
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.952829820757219e-05,
      "loss": 1.1819,
      "step": 70500
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.952439823409201e-05,
      "loss": 1.1816,
      "step": 71000
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.952049826061183e-05,
      "loss": 1.1805,
      "step": 71500
    },
    {
      "epoch": 3.36,
      "learning_rate": 1.951661388702557e-05,
      "loss": 1.1794,
      "step": 72000
    },
    {
      "epoch": 3.38,
      "learning_rate": 1.951271391354539e-05,
      "loss": 1.1799,
      "step": 72500
    },
    {
      "epoch": 3.4,
      "learning_rate": 1.950881394006521e-05,
      "loss": 1.1812,
      "step": 73000
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.950491396658503e-05,
      "loss": 1.1803,
      "step": 73500
    },
    {
      "epoch": 3.45,
      "learning_rate": 1.950101399310485e-05,
      "loss": 1.1805,
      "step": 74000
    },
    {
      "epoch": 3.47,
      "learning_rate": 1.9497114019624668e-05,
      "loss": 1.1795,
      "step": 74500
    },
    {
      "epoch": 3.5,
      "learning_rate": 1.9493214046144488e-05,
      "loss": 1.1821,
      "step": 75000
    },
    {
      "epoch": 3.52,
      "learning_rate": 1.948932187261127e-05,
      "loss": 1.18,
      "step": 75500
    },
    {
      "epoch": 3.54,
      "learning_rate": 1.9485421899131087e-05,
      "loss": 1.1783,
      "step": 76000
    },
    {
      "epoch": 3.57,
      "learning_rate": 1.9481521925650907e-05,
      "loss": 1.1764,
      "step": 76500
    },
    {
      "epoch": 3.59,
      "learning_rate": 1.9477621952170728e-05,
      "loss": 1.1788,
      "step": 77000
    },
    {
      "epoch": 3.61,
      "learning_rate": 1.9473721978690548e-05,
      "loss": 1.1775,
      "step": 77500
    },
    {
      "epoch": 3.64,
      "learning_rate": 1.9469829805157326e-05,
      "loss": 1.1743,
      "step": 78000
    },
    {
      "epoch": 3.66,
      "learning_rate": 1.9465929831677147e-05,
      "loss": 1.1765,
      "step": 78500
    },
    {
      "epoch": 3.68,
      "learning_rate": 1.9462029858196967e-05,
      "loss": 1.175,
      "step": 79000
    },
    {
      "epoch": 3.71,
      "learning_rate": 1.9458129884716784e-05,
      "loss": 1.1749,
      "step": 79500
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.9454237711183566e-05,
      "loss": 1.1762,
      "step": 80000
    },
    {
      "epoch": 3.75,
      "learning_rate": 1.9450337737703383e-05,
      "loss": 1.1756,
      "step": 80500
    },
    {
      "epoch": 3.78,
      "learning_rate": 1.9446437764223203e-05,
      "loss": 1.1743,
      "step": 81000
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.9442537790743024e-05,
      "loss": 1.1756,
      "step": 81500
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.9438637817262844e-05,
      "loss": 1.175,
      "step": 82000
    },
    {
      "epoch": 3.85,
      "learning_rate": 1.943473784378266e-05,
      "loss": 1.1734,
      "step": 82500
    },
    {
      "epoch": 3.87,
      "learning_rate": 1.9430837870302482e-05,
      "loss": 1.1751,
      "step": 83000
    },
    {
      "epoch": 3.89,
      "learning_rate": 1.9426937896822303e-05,
      "loss": 1.1721,
      "step": 83500
    },
    {
      "epoch": 3.92,
      "learning_rate": 1.9423037923342123e-05,
      "loss": 1.1713,
      "step": 84000
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.9419145749808905e-05,
      "loss": 1.1725,
      "step": 84500
    },
    {
      "epoch": 3.96,
      "learning_rate": 1.9415245776328722e-05,
      "loss": 1.1736,
      "step": 85000
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.9411353602795503e-05,
      "loss": 1.1737,
      "step": 85500
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.7587732837348504,
      "eval_loss": 1.0296095609664917,
      "eval_runtime": 51.9687,
      "eval_samples_per_second": 898.771,
      "eval_steps_per_second": 14.047,
      "step": 85804
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.9407453629315324e-05,
      "loss": 1.1705,
      "step": 86000
    },
    {
      "epoch": 4.03,
      "learning_rate": 1.940355365583514e-05,
      "loss": 1.1717,
      "step": 86500
    },
    {
      "epoch": 4.06,
      "learning_rate": 1.939965368235496e-05,
      "loss": 1.1695,
      "step": 87000
    },
    {
      "epoch": 4.08,
      "learning_rate": 1.9395753708874782e-05,
      "loss": 1.1676,
      "step": 87500
    },
    {
      "epoch": 4.1,
      "learning_rate": 1.9391853735394602e-05,
      "loss": 1.171,
      "step": 88000
    },
    {
      "epoch": 4.13,
      "learning_rate": 1.938795376191442e-05,
      "loss": 1.167,
      "step": 88500
    },
    {
      "epoch": 4.15,
      "learning_rate": 1.938405378843424e-05,
      "loss": 1.1685,
      "step": 89000
    },
    {
      "epoch": 4.17,
      "learning_rate": 1.938015381495406e-05,
      "loss": 1.1715,
      "step": 89500
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.937626164142084e-05,
      "loss": 1.1671,
      "step": 90000
    },
    {
      "epoch": 4.22,
      "learning_rate": 1.937236166794066e-05,
      "loss": 1.17,
      "step": 90500
    },
    {
      "epoch": 4.24,
      "learning_rate": 1.936846169446048e-05,
      "loss": 1.1687,
      "step": 91000
    },
    {
      "epoch": 4.27,
      "learning_rate": 1.93645617209803e-05,
      "loss": 1.1694,
      "step": 91500
    },
    {
      "epoch": 4.29,
      "learning_rate": 1.9360661747500117e-05,
      "loss": 1.1689,
      "step": 92000
    },
    {
      "epoch": 4.02,
      "learning_rate": 1.8334e-05,
      "loss": 1.1672,
      "step": 92500
    },
    {
      "epoch": 4.05,
      "learning_rate": 1.832391919191919e-05,
      "loss": 1.1673,
      "step": 93000
    },
    {
      "epoch": 4.07,
      "learning_rate": 1.8313818181818184e-05,
      "loss": 1.1634,
      "step": 93500
    },
    {
      "epoch": 4.09,
      "learning_rate": 1.8303717171717172e-05,
      "loss": 1.1639,
      "step": 94000
    },
    {
      "epoch": 4.12,
      "learning_rate": 1.8293636363636366e-05,
      "loss": 1.1633,
      "step": 94500
    },
    {
      "epoch": 4.14,
      "learning_rate": 1.8283535353535355e-05,
      "loss": 1.1625,
      "step": 95000
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.8273434343434343e-05,
      "loss": 1.1635,
      "step": 95500
    },
    {
      "epoch": 4.19,
      "learning_rate": 1.8263333333333335e-05,
      "loss": 1.1631,
      "step": 96000
    },
    {
      "epoch": 4.21,
      "learning_rate": 1.8253232323232324e-05,
      "loss": 1.1618,
      "step": 96500
    },
    {
      "epoch": 4.23,
      "learning_rate": 1.8243131313131316e-05,
      "loss": 1.1639,
      "step": 97000
    },
    {
      "epoch": 4.26,
      "learning_rate": 1.8233030303030304e-05,
      "loss": 1.1645,
      "step": 97500
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.8222929292929293e-05,
      "loss": 1.1625,
      "step": 98000
    },
    {
      "epoch": 4.3,
      "learning_rate": 1.8212828282828285e-05,
      "loss": 1.1649,
      "step": 98500
    },
    {
      "epoch": 4.33,
      "learning_rate": 1.8202747474747476e-05,
      "loss": 1.1631,
      "step": 99000
    },
    {
      "epoch": 4.35,
      "learning_rate": 1.8192646464646467e-05,
      "loss": 1.1596,
      "step": 99500
    },
    {
      "epoch": 4.37,
      "learning_rate": 1.8182545454545456e-05,
      "loss": 1.1611,
      "step": 100000
    },
    {
      "epoch": 4.4,
      "learning_rate": 1.8172464646464647e-05,
      "loss": 1.1622,
      "step": 100500
    },
    {
      "epoch": 4.42,
      "learning_rate": 1.816236363636364e-05,
      "loss": 1.162,
      "step": 101000
    },
    {
      "epoch": 4.44,
      "learning_rate": 1.8152262626262627e-05,
      "loss": 1.1634,
      "step": 101500
    },
    {
      "epoch": 4.47,
      "learning_rate": 1.814216161616162e-05,
      "loss": 1.1627,
      "step": 102000
    },
    {
      "epoch": 4.49,
      "learning_rate": 1.8132060606060608e-05,
      "loss": 1.1602,
      "step": 102500
    },
    {
      "epoch": 4.51,
      "learning_rate": 1.8121959595959596e-05,
      "loss": 1.1594,
      "step": 103000
    },
    {
      "epoch": 4.54,
      "learning_rate": 1.811185858585859e-05,
      "loss": 1.1622,
      "step": 103500
    },
    {
      "epoch": 4.56,
      "learning_rate": 1.8101757575757577e-05,
      "loss": 1.1612,
      "step": 104000
    },
    {
      "epoch": 4.58,
      "learning_rate": 1.809165656565657e-05,
      "loss": 1.16,
      "step": 104500
    },
    {
      "epoch": 4.61,
      "learning_rate": 1.8081555555555558e-05,
      "loss": 1.1601,
      "step": 105000
    },
    {
      "epoch": 4.63,
      "learning_rate": 1.8071474747474748e-05,
      "loss": 1.1588,
      "step": 105500
    },
    {
      "epoch": 4.65,
      "learning_rate": 1.806137373737374e-05,
      "loss": 1.1593,
      "step": 106000
    },
    {
      "epoch": 4.68,
      "learning_rate": 1.805127272727273e-05,
      "loss": 1.1572,
      "step": 106500
    },
    {
      "epoch": 4.7,
      "learning_rate": 1.804117171717172e-05,
      "loss": 1.1591,
      "step": 107000
    },
    {
      "epoch": 4.72,
      "learning_rate": 1.803107070707071e-05,
      "loss": 1.1604,
      "step": 107500
    },
    {
      "epoch": 4.75,
      "learning_rate": 1.8020969696969698e-05,
      "loss": 1.1584,
      "step": 108000
    },
    {
      "epoch": 4.77,
      "learning_rate": 1.8010868686868686e-05,
      "loss": 1.1591,
      "step": 108500
    },
    {
      "epoch": 4.79,
      "learning_rate": 1.800076767676768e-05,
      "loss": 1.1589,
      "step": 109000
    },
    {
      "epoch": 4.82,
      "learning_rate": 1.7990666666666667e-05,
      "loss": 1.161,
      "step": 109500
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.798056565656566e-05,
      "loss": 1.1582,
      "step": 110000
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.797048484848485e-05,
      "loss": 1.1572,
      "step": 110500
    },
    {
      "epoch": 4.89,
      "learning_rate": 1.7960383838383838e-05,
      "loss": 1.1556,
      "step": 111000
    },
    {
      "epoch": 4.91,
      "learning_rate": 1.795028282828283e-05,
      "loss": 1.1573,
      "step": 111500
    },
    {
      "epoch": 4.93,
      "learning_rate": 1.794020202020202e-05,
      "loss": 1.1542,
      "step": 112000
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.793010101010101e-05,
      "loss": 1.159,
      "step": 112500
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.792e-05,
      "loss": 1.1571,
      "step": 113000
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.76073263429499,
      "eval_loss": 1.0153971910476685,
      "eval_runtime": 51.0937,
      "eval_samples_per_second": 914.164,
      "eval_steps_per_second": 14.287,
      "step": 113451
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.790989898989899e-05,
      "loss": 1.1573,
      "step": 113500
    },
    {
      "epoch": 5.03,
      "learning_rate": 1.7899797979797982e-05,
      "loss": 1.1528,
      "step": 114000
    },
    {
      "epoch": 5.05,
      "learning_rate": 1.788969696969697e-05,
      "loss": 1.1534,
      "step": 114500
    },
    {
      "epoch": 5.07,
      "learning_rate": 1.7879595959595962e-05,
      "loss": 1.1511,
      "step": 115000
    },
    {
      "epoch": 5.1,
      "learning_rate": 1.786949494949495e-05,
      "loss": 1.154,
      "step": 115500
    },
    {
      "epoch": 5.12,
      "learning_rate": 1.785939393939394e-05,
      "loss": 1.1525,
      "step": 116000
    },
    {
      "epoch": 5.14,
      "learning_rate": 1.7849313131313133e-05,
      "loss": 1.1543,
      "step": 116500
    },
    {
      "epoch": 5.17,
      "learning_rate": 1.7839212121212122e-05,
      "loss": 1.1517,
      "step": 117000
    },
    {
      "epoch": 5.19,
      "learning_rate": 1.7829131313131313e-05,
      "loss": 1.155,
      "step": 117500
    },
    {
      "epoch": 5.21,
      "learning_rate": 1.7819030303030305e-05,
      "loss": 1.1541,
      "step": 118000
    },
    {
      "epoch": 5.24,
      "learning_rate": 1.7808929292929293e-05,
      "loss": 1.1545,
      "step": 118500
    },
    {
      "epoch": 5.26,
      "learning_rate": 1.7798828282828285e-05,
      "loss": 1.1535,
      "step": 119000
    },
    {
      "epoch": 5.28,
      "learning_rate": 1.7788727272727274e-05,
      "loss": 1.1533,
      "step": 119500
    },
    {
      "epoch": 5.31,
      "learning_rate": 1.7778646464646464e-05,
      "loss": 1.1538,
      "step": 120000
    },
    {
      "epoch": 5.33,
      "learning_rate": 1.7768545454545456e-05,
      "loss": 1.1512,
      "step": 120500
    },
    {
      "epoch": 5.35,
      "learning_rate": 1.7758444444444445e-05,
      "loss": 1.1532,
      "step": 121000
    },
    {
      "epoch": 5.38,
      "learning_rate": 1.7748343434343437e-05,
      "loss": 1.1516,
      "step": 121500
    },
    {
      "epoch": 5.4,
      "learning_rate": 1.7738242424242425e-05,
      "loss": 1.1509,
      "step": 122000
    },
    {
      "epoch": 5.42,
      "learning_rate": 1.7728161616161616e-05,
      "loss": 1.1486,
      "step": 122500
    },
    {
      "epoch": 5.45,
      "learning_rate": 1.7718060606060608e-05,
      "loss": 1.1555,
      "step": 123000
    },
    {
      "epoch": 5.47,
      "learning_rate": 1.77079595959596e-05,
      "loss": 1.1524,
      "step": 123500
    },
    {
      "epoch": 5.49,
      "learning_rate": 1.769785858585859e-05,
      "loss": 1.1486,
      "step": 124000
    },
    {
      "epoch": 5.52,
      "learning_rate": 1.7687757575757577e-05,
      "loss": 1.151,
      "step": 124500
    },
    {
      "epoch": 5.54,
      "learning_rate": 1.7677656565656566e-05,
      "loss": 1.1502,
      "step": 125000
    },
    {
      "epoch": 5.56,
      "learning_rate": 1.766757575757576e-05,
      "loss": 1.1477,
      "step": 125500
    },
    {
      "epoch": 5.59,
      "learning_rate": 1.765749494949495e-05,
      "loss": 1.1451,
      "step": 126000
    },
    {
      "epoch": 5.61,
      "learning_rate": 1.764739393939394e-05,
      "loss": 1.1505,
      "step": 126500
    },
    {
      "epoch": 5.63,
      "learning_rate": 1.763729292929293e-05,
      "loss": 1.151,
      "step": 127000
    },
    {
      "epoch": 5.65,
      "learning_rate": 1.7627191919191923e-05,
      "loss": 1.1485,
      "step": 127500
    },
    {
      "epoch": 5.68,
      "learning_rate": 1.761709090909091e-05,
      "loss": 1.1482,
      "step": 128000
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.76069898989899e-05,
      "loss": 1.1489,
      "step": 128500
    },
    {
      "epoch": 5.72,
      "learning_rate": 1.759688888888889e-05,
      "loss": 1.1507,
      "step": 129000
    },
    {
      "epoch": 5.75,
      "learning_rate": 1.758678787878788e-05,
      "loss": 1.1496,
      "step": 129500
    },
    {
      "epoch": 5.77,
      "learning_rate": 1.757668686868687e-05,
      "loss": 1.1481,
      "step": 130000
    },
    {
      "epoch": 5.79,
      "learning_rate": 1.7566606060606063e-05,
      "loss": 1.1473,
      "step": 130500
    },
    {
      "epoch": 5.82,
      "learning_rate": 1.7556505050505052e-05,
      "loss": 1.147,
      "step": 131000
    },
    {
      "epoch": 5.84,
      "learning_rate": 1.754640404040404e-05,
      "loss": 1.1494,
      "step": 131500
    },
    {
      "epoch": 5.86,
      "learning_rate": 1.7536303030303032e-05,
      "loss": 1.148,
      "step": 132000
    },
    {
      "epoch": 5.89,
      "learning_rate": 1.752620202020202e-05,
      "loss": 1.1469,
      "step": 132500
    },
    {
      "epoch": 5.91,
      "learning_rate": 1.7516101010101013e-05,
      "loss": 1.1453,
      "step": 133000
    },
    {
      "epoch": 5.93,
      "learning_rate": 1.7506e-05,
      "loss": 1.1467,
      "step": 133500
    },
    {
      "epoch": 5.96,
      "learning_rate": 1.749589898989899e-05,
      "loss": 1.1465,
      "step": 134000
    },
    {
      "epoch": 5.98,
      "learning_rate": 1.7485797979797982e-05,
      "loss": 1.1455,
      "step": 134500
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.7624844555701037,
      "eval_loss": 1.0080606937408447,
      "eval_runtime": 51.6695,
      "eval_samples_per_second": 903.976,
      "eval_steps_per_second": 14.128,
      "step": 134902
    },
    {
      "epoch": 6.0,
      "learning_rate": 1.747569696969697e-05,
      "loss": 1.1493,
      "step": 135000
    },
    {
      "epoch": 6.03,
      "learning_rate": 1.7465616161616165e-05,
      "loss": 1.1439,
      "step": 135500
    },
    {
      "epoch": 6.05,
      "learning_rate": 1.7455535353535355e-05,
      "loss": 1.1419,
      "step": 136000
    },
    {
      "epoch": 6.07,
      "learning_rate": 1.7445434343434344e-05,
      "loss": 1.1468,
      "step": 136500
    },
    {
      "epoch": 6.1,
      "learning_rate": 1.7435333333333336e-05,
      "loss": 1.1456,
      "step": 137000
    },
    {
      "epoch": 6.12,
      "learning_rate": 1.7425232323232324e-05,
      "loss": 1.1423,
      "step": 137500
    },
    {
      "epoch": 6.14,
      "learning_rate": 1.7415131313131316e-05,
      "loss": 1.1426,
      "step": 138000
    },
    {
      "epoch": 6.17,
      "learning_rate": 1.7405030303030305e-05,
      "loss": 1.1447,
      "step": 138500
    },
    {
      "epoch": 6.19,
      "learning_rate": 1.7394929292929293e-05,
      "loss": 1.1418,
      "step": 139000
    },
    {
      "epoch": 6.21,
      "learning_rate": 1.7384828282828282e-05,
      "loss": 1.1445,
      "step": 139500
    },
    {
      "epoch": 6.24,
      "learning_rate": 1.7374747474747476e-05,
      "loss": 1.1429,
      "step": 140000
    },
    {
      "epoch": 6.26,
      "learning_rate": 1.7364646464646468e-05,
      "loss": 1.1431,
      "step": 140500
    },
    {
      "epoch": 6.28,
      "learning_rate": 1.7354545454545457e-05,
      "loss": 1.143,
      "step": 141000
    },
    {
      "epoch": 6.31,
      "learning_rate": 1.7344444444444445e-05,
      "loss": 1.1422,
      "step": 141500
    },
    {
      "epoch": 6.33,
      "learning_rate": 1.7334343434343434e-05,
      "loss": 1.1426,
      "step": 142000
    },
    {
      "epoch": 6.35,
      "learning_rate": 1.7324242424242426e-05,
      "loss": 1.146,
      "step": 142500
    },
    {
      "epoch": 6.38,
      "learning_rate": 1.7314161616161616e-05,
      "loss": 1.1443,
      "step": 143000
    },
    {
      "epoch": 6.4,
      "learning_rate": 1.730408080808081e-05,
      "loss": 1.1444,
      "step": 143500
    },
    {
      "epoch": 6.42,
      "learning_rate": 1.72939797979798e-05,
      "loss": 1.1433,
      "step": 144000
    },
    {
      "epoch": 6.45,
      "learning_rate": 1.728387878787879e-05,
      "loss": 1.1436,
      "step": 144500
    },
    {
      "epoch": 6.47,
      "learning_rate": 1.727377777777778e-05,
      "loss": 1.1393,
      "step": 145000
    },
    {
      "epoch": 6.49,
      "learning_rate": 1.7263676767676768e-05,
      "loss": 1.1428,
      "step": 145500
    },
    {
      "epoch": 6.52,
      "learning_rate": 1.725357575757576e-05,
      "loss": 1.1432,
      "step": 146000
    },
    {
      "epoch": 6.54,
      "learning_rate": 1.724347474747475e-05,
      "loss": 1.1409,
      "step": 146500
    },
    {
      "epoch": 6.56,
      "learning_rate": 1.723337373737374e-05,
      "loss": 1.1416,
      "step": 147000
    },
    {
      "epoch": 6.59,
      "learning_rate": 1.722327272727273e-05,
      "loss": 1.1394,
      "step": 147500
    },
    {
      "epoch": 6.61,
      "learning_rate": 1.721319191919192e-05,
      "loss": 1.1424,
      "step": 148000
    },
    {
      "epoch": 6.63,
      "learning_rate": 1.720309090909091e-05,
      "loss": 1.1391,
      "step": 148500
    },
    {
      "epoch": 6.66,
      "learning_rate": 1.71929898989899e-05,
      "loss": 1.1405,
      "step": 149000
    },
    {
      "epoch": 6.68,
      "learning_rate": 1.7182888888888892e-05,
      "loss": 1.1418,
      "step": 149500
    },
    {
      "epoch": 6.7,
      "learning_rate": 1.717278787878788e-05,
      "loss": 1.1395,
      "step": 150000
    },
    {
      "epoch": 6.73,
      "learning_rate": 1.716270707070707e-05,
      "loss": 1.1399,
      "step": 150500
    },
    {
      "epoch": 6.75,
      "learning_rate": 1.7152606060606063e-05,
      "loss": 1.1412,
      "step": 151000
    },
    {
      "epoch": 6.77,
      "learning_rate": 1.7142525252525254e-05,
      "loss": 1.1393,
      "step": 151500
    },
    {
      "epoch": 6.8,
      "learning_rate": 1.7132424242424243e-05,
      "loss": 1.1406,
      "step": 152000
    },
    {
      "epoch": 6.82,
      "learning_rate": 1.7122323232323234e-05,
      "loss": 1.1403,
      "step": 152500
    },
    {
      "epoch": 6.84,
      "learning_rate": 1.7112222222222223e-05,
      "loss": 1.1381,
      "step": 153000
    },
    {
      "epoch": 6.87,
      "learning_rate": 1.7102121212121215e-05,
      "loss": 1.1384,
      "step": 153500
    },
    {
      "epoch": 6.89,
      "learning_rate": 1.7092040404040406e-05,
      "loss": 1.14,
      "step": 154000
    },
    {
      "epoch": 6.91,
      "learning_rate": 1.7081939393939394e-05,
      "loss": 1.1395,
      "step": 154500
    },
    {
      "epoch": 6.94,
      "learning_rate": 1.7071838383838386e-05,
      "loss": 1.1412,
      "step": 155000
    },
    {
      "epoch": 6.96,
      "learning_rate": 1.7061737373737375e-05,
      "loss": 1.1391,
      "step": 155500
    },
    {
      "epoch": 6.98,
      "learning_rate": 1.7051636363636367e-05,
      "loss": 1.1403,
      "step": 156000
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.7638281013714997,
      "eval_loss": 1.000293254852295,
      "eval_runtime": 50.4795,
      "eval_samples_per_second": 925.286,
      "eval_steps_per_second": 14.461,
      "step": 156353
    },
    {
      "epoch": 7.01,
      "learning_rate": 1.7041535353535355e-05,
      "loss": 1.1408,
      "step": 156500
    },
    {
      "epoch": 7.03,
      "learning_rate": 1.7031434343434344e-05,
      "loss": 1.1372,
      "step": 157000
    },
    {
      "epoch": 7.05,
      "learning_rate": 1.7021333333333333e-05,
      "loss": 1.1376,
      "step": 157500
    },
    {
      "epoch": 7.08,
      "learning_rate": 1.7011252525252526e-05,
      "loss": 1.1366,
      "step": 158000
    },
    {
      "epoch": 7.1,
      "learning_rate": 1.700115151515152e-05,
      "loss": 1.1381,
      "step": 158500
    },
    {
      "epoch": 7.12,
      "learning_rate": 1.6991050505050507e-05,
      "loss": 1.1353,
      "step": 159000
    },
    {
      "epoch": 7.15,
      "learning_rate": 1.6980949494949496e-05,
      "loss": 1.1357,
      "step": 159500
    },
    {
      "epoch": 7.17,
      "learning_rate": 1.6970848484848484e-05,
      "loss": 1.1377,
      "step": 160000
    },
    {
      "epoch": 7.19,
      "learning_rate": 1.6960747474747476e-05,
      "loss": 1.1385,
      "step": 160500
    },
    {
      "epoch": 7.22,
      "learning_rate": 1.6950646464646465e-05,
      "loss": 1.1352,
      "step": 161000
    },
    {
      "epoch": 7.24,
      "learning_rate": 1.6940545454545457e-05,
      "loss": 1.1338,
      "step": 161500
    },
    {
      "epoch": 7.26,
      "learning_rate": 1.6930444444444445e-05,
      "loss": 1.1334,
      "step": 162000
    },
    {
      "epoch": 7.29,
      "learning_rate": 1.6920343434343437e-05,
      "loss": 1.1377,
      "step": 162500
    },
    {
      "epoch": 7.31,
      "learning_rate": 1.6910242424242426e-05,
      "loss": 1.1383,
      "step": 163000
    },
    {
      "epoch": 7.33,
      "learning_rate": 1.6900141414141415e-05,
      "loss": 1.1359,
      "step": 163500
    },
    {
      "epoch": 7.36,
      "learning_rate": 1.689006060606061e-05,
      "loss": 1.1364,
      "step": 164000
    },
    {
      "epoch": 7.38,
      "learning_rate": 1.6879959595959597e-05,
      "loss": 1.1361,
      "step": 164500
    },
    {
      "epoch": 7.4,
      "learning_rate": 1.6869878787878788e-05,
      "loss": 1.1329,
      "step": 165000
    },
    {
      "epoch": 7.43,
      "learning_rate": 1.685977777777778e-05,
      "loss": 1.1354,
      "step": 165500
    },
    {
      "epoch": 7.45,
      "learning_rate": 1.6849676767676768e-05,
      "loss": 1.1349,
      "step": 166000
    },
    {
      "epoch": 7.47,
      "learning_rate": 1.683957575757576e-05,
      "loss": 1.1366,
      "step": 166500
    },
    {
      "epoch": 7.5,
      "learning_rate": 1.682949494949495e-05,
      "loss": 1.1371,
      "step": 167000
    },
    {
      "epoch": 7.52,
      "learning_rate": 1.6819414141414145e-05,
      "loss": 1.136,
      "step": 167500
    },
    {
      "epoch": 7.54,
      "learning_rate": 1.6809313131313133e-05,
      "loss": 1.1345,
      "step": 168000
    },
    {
      "epoch": 7.57,
      "learning_rate": 1.6799212121212122e-05,
      "loss": 1.1327,
      "step": 168500
    },
    {
      "epoch": 7.59,
      "learning_rate": 1.678911111111111e-05,
      "loss": 1.1352,
      "step": 169000
    },
    {
      "epoch": 7.61,
      "learning_rate": 1.6779010101010102e-05,
      "loss": 1.136,
      "step": 169500
    },
    {
      "epoch": 7.64,
      "learning_rate": 1.676890909090909e-05,
      "loss": 1.1337,
      "step": 170000
    },
    {
      "epoch": 7.66,
      "learning_rate": 1.6758808080808083e-05,
      "loss": 1.1328,
      "step": 170500
    },
    {
      "epoch": 7.68,
      "learning_rate": 1.674870707070707e-05,
      "loss": 1.1328,
      "step": 171000
    },
    {
      "epoch": 7.71,
      "learning_rate": 1.6738626262626262e-05,
      "loss": 1.1352,
      "step": 171500
    },
    {
      "epoch": 7.73,
      "learning_rate": 1.6728525252525254e-05,
      "loss": 1.1324,
      "step": 172000
    },
    {
      "epoch": 7.75,
      "learning_rate": 1.6718424242424246e-05,
      "loss": 1.1319,
      "step": 172500
    },
    {
      "epoch": 7.78,
      "learning_rate": 1.6708323232323235e-05,
      "loss": 1.1307,
      "step": 173000
    },
    {
      "epoch": 7.8,
      "learning_rate": 1.6698222222222223e-05,
      "loss": 1.1326,
      "step": 173500
    },
    {
      "epoch": 7.82,
      "learning_rate": 1.6688121212121212e-05,
      "loss": 1.1344,
      "step": 174000
    },
    {
      "epoch": 7.85,
      "learning_rate": 1.6678020202020204e-05,
      "loss": 1.1328,
      "step": 174500
    },
    {
      "epoch": 7.87,
      "learning_rate": 1.6667919191919192e-05,
      "loss": 1.1312,
      "step": 175000
    },
    {
      "epoch": 7.89,
      "learning_rate": 1.6657838383838386e-05,
      "loss": 1.1341,
      "step": 175500
    },
    {
      "epoch": 7.92,
      "learning_rate": 1.6647737373737375e-05,
      "loss": 1.1319,
      "step": 176000
    },
    {
      "epoch": 7.94,
      "learning_rate": 1.6637636363636364e-05,
      "loss": 1.1302,
      "step": 176500
    },
    {
      "epoch": 7.96,
      "learning_rate": 1.6627535353535356e-05,
      "loss": 1.1324,
      "step": 177000
    },
    {
      "epoch": 7.99,
      "learning_rate": 1.6617434343434344e-05,
      "loss": 1.132,
      "step": 177500
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.7653896931539229,
      "eval_loss": 0.9911502003669739,
      "eval_runtime": 51.7861,
      "eval_samples_per_second": 901.941,
      "eval_steps_per_second": 14.096,
      "step": 177804
    },
    {
      "epoch": 8.01,
      "learning_rate": 1.6607353535353538e-05,
      "loss": 1.1318,
      "step": 178000
    },
    {
      "epoch": 8.03,
      "learning_rate": 1.6597252525252527e-05,
      "loss": 1.1322,
      "step": 178500
    },
    {
      "epoch": 8.06,
      "learning_rate": 1.6587151515151515e-05,
      "loss": 1.1296,
      "step": 179000
    },
    {
      "epoch": 8.08,
      "learning_rate": 1.6577050505050507e-05,
      "loss": 1.1297,
      "step": 179500
    },
    {
      "epoch": 8.1,
      "learning_rate": 1.6566949494949496e-05,
      "loss": 1.13,
      "step": 180000
    },
    {
      "epoch": 8.13,
      "learning_rate": 1.6556848484848488e-05,
      "loss": 1.1265,
      "step": 180500
    },
    {
      "epoch": 8.15,
      "learning_rate": 1.6546747474747476e-05,
      "loss": 1.1291,
      "step": 181000
    },
    {
      "epoch": 8.17,
      "learning_rate": 1.6536666666666667e-05,
      "loss": 1.128,
      "step": 181500
    },
    {
      "epoch": 8.2,
      "learning_rate": 1.652656565656566e-05,
      "loss": 1.1282,
      "step": 182000
    },
    {
      "epoch": 8.22,
      "learning_rate": 1.6516464646464648e-05,
      "loss": 1.1288,
      "step": 182500
    },
    {
      "epoch": 8.24,
      "learning_rate": 1.650636363636364e-05,
      "loss": 1.1293,
      "step": 183000
    },
    {
      "epoch": 8.27,
      "learning_rate": 1.6496262626262628e-05,
      "loss": 1.1319,
      "step": 183500
    },
    {
      "epoch": 8.29,
      "learning_rate": 1.6486161616161617e-05,
      "loss": 1.1302,
      "step": 184000
    },
    {
      "epoch": 8.31,
      "learning_rate": 1.6476060606060605e-05,
      "loss": 1.1283,
      "step": 184500
    },
    {
      "epoch": 8.34,
      "learning_rate": 1.6465959595959597e-05,
      "loss": 1.1317,
      "step": 185000
    },
    {
      "epoch": 8.36,
      "learning_rate": 1.6455858585858586e-05,
      "loss": 1.1303,
      "step": 185500
    },
    {
      "epoch": 8.38,
      "learning_rate": 1.644577777777778e-05,
      "loss": 1.1301,
      "step": 186000
    },
    {
      "epoch": 8.41,
      "learning_rate": 1.643567676767677e-05,
      "loss": 1.129,
      "step": 186500
    },
    {
      "epoch": 8.43,
      "learning_rate": 1.6425595959595962e-05,
      "loss": 1.1279,
      "step": 187000
    },
    {
      "epoch": 8.45,
      "learning_rate": 1.641549494949495e-05,
      "loss": 1.1308,
      "step": 187500
    },
    {
      "epoch": 8.48,
      "learning_rate": 1.640539393939394e-05,
      "loss": 1.1298,
      "step": 188000
    },
    {
      "epoch": 8.5,
      "learning_rate": 1.6395313131313134e-05,
      "loss": 1.1288,
      "step": 188500
    },
    {
      "epoch": 8.52,
      "learning_rate": 1.6385212121212122e-05,
      "loss": 1.1295,
      "step": 189000
    },
    {
      "epoch": 8.55,
      "learning_rate": 1.6375111111111114e-05,
      "loss": 1.1272,
      "step": 189500
    },
    {
      "epoch": 8.57,
      "learning_rate": 1.6365010101010103e-05,
      "loss": 1.1275,
      "step": 190000
    },
    {
      "epoch": 8.59,
      "learning_rate": 1.635490909090909e-05,
      "loss": 1.1264,
      "step": 190500
    },
    {
      "epoch": 8.62,
      "learning_rate": 1.6344828282828285e-05,
      "loss": 1.1293,
      "step": 191000
    },
    {
      "epoch": 8.64,
      "learning_rate": 1.6334727272727274e-05,
      "loss": 1.1282,
      "step": 191500
    },
    {
      "epoch": 8.66,
      "learning_rate": 1.6324626262626266e-05,
      "loss": 1.125,
      "step": 192000
    },
    {
      "epoch": 8.69,
      "learning_rate": 1.6314525252525254e-05,
      "loss": 1.127,
      "step": 192500
    },
    {
      "epoch": 8.71,
      "learning_rate": 1.6304424242424243e-05,
      "loss": 1.1284,
      "step": 193000
    },
    {
      "epoch": 8.73,
      "learning_rate": 1.6294323232323235e-05,
      "loss": 1.1282,
      "step": 193500
    },
    {
      "epoch": 8.76,
      "learning_rate": 1.6284222222222224e-05,
      "loss": 1.1248,
      "step": 194000
    },
    {
      "epoch": 8.78,
      "learning_rate": 1.6274121212121216e-05,
      "loss": 1.1261,
      "step": 194500
    },
    {
      "epoch": 8.8,
      "learning_rate": 1.6264020202020204e-05,
      "loss": 1.129,
      "step": 195000
    },
    {
      "epoch": 8.82,
      "learning_rate": 1.6253939393939395e-05,
      "loss": 1.1254,
      "step": 195500
    },
    {
      "epoch": 8.85,
      "learning_rate": 1.6243838383838387e-05,
      "loss": 1.128,
      "step": 196000
    },
    {
      "epoch": 8.87,
      "learning_rate": 1.6233737373737375e-05,
      "loss": 1.1268,
      "step": 196500
    },
    {
      "epoch": 8.89,
      "learning_rate": 1.6223636363636367e-05,
      "loss": 1.1287,
      "step": 197000
    },
    {
      "epoch": 8.92,
      "learning_rate": 1.6213535353535356e-05,
      "loss": 1.1268,
      "step": 197500
    },
    {
      "epoch": 8.94,
      "learning_rate": 1.6203454545454546e-05,
      "loss": 1.1239,
      "step": 198000
    },
    {
      "epoch": 8.96,
      "learning_rate": 1.619335353535354e-05,
      "loss": 1.1252,
      "step": 198500
    },
    {
      "epoch": 8.99,
      "learning_rate": 1.6183252525252527e-05,
      "loss": 1.1274,
      "step": 199000
    },
    {
      "epoch": 9.0,
      "eval_accuracy": 0.7663617465706959,
      "eval_loss": 0.9863241314888,
      "eval_runtime": 51.5226,
      "eval_samples_per_second": 906.553,
      "eval_steps_per_second": 14.169,
      "step": 199255
    },
    {
      "epoch": 9.01,
      "learning_rate": 1.617315151515152e-05,
      "loss": 1.1273,
      "step": 199500
    },
    {
      "epoch": 9.03,
      "learning_rate": 1.6163050505050508e-05,
      "loss": 1.1243,
      "step": 200000
    },
    {
      "epoch": 9.06,
      "learning_rate": 1.6152949494949496e-05,
      "loss": 1.1257,
      "step": 200500
    },
    {
      "epoch": 9.08,
      "learning_rate": 1.6142848484848485e-05,
      "loss": 1.126,
      "step": 201000
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.6132747474747477e-05,
      "loss": 1.1237,
      "step": 201500
    },
    {
      "epoch": 9.13,
      "learning_rate": 1.6122666666666667e-05,
      "loss": 1.123,
      "step": 202000
    },
    {
      "epoch": 9.15,
      "learning_rate": 1.6112565656565656e-05,
      "loss": 1.1258,
      "step": 202500
    },
    {
      "epoch": 9.17,
      "learning_rate": 1.6102464646464648e-05,
      "loss": 1.1263,
      "step": 203000
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.6092363636363636e-05,
      "loss": 1.1234,
      "step": 203500
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.608228282828283e-05,
      "loss": 1.1243,
      "step": 204000
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.607218181818182e-05,
      "loss": 1.1236,
      "step": 204500
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.6062080808080808e-05,
      "loss": 1.1245,
      "step": 205000
    },
    {
      "epoch": 9.58,
      "learning_rate": 1.60519797979798e-05,
      "loss": 1.1229,
      "step": 205500
    },
    {
      "epoch": 9.6,
      "learning_rate": 1.6041878787878788e-05,
      "loss": 1.1247,
      "step": 206000
    },
    {
      "epoch": 9.63,
      "learning_rate": 1.603177777777778e-05,
      "loss": 1.1224,
      "step": 206500
    },
    {
      "epoch": 9.65,
      "learning_rate": 1.602167676767677e-05,
      "loss": 1.1247,
      "step": 207000
    },
    {
      "epoch": 9.67,
      "learning_rate": 1.601157575757576e-05,
      "loss": 1.1245,
      "step": 207500
    },
    {
      "epoch": 9.7,
      "learning_rate": 1.600147474747475e-05,
      "loss": 1.1219,
      "step": 208000
    },
    {
      "epoch": 9.72,
      "learning_rate": 1.5991373737373738e-05,
      "loss": 1.1259,
      "step": 208500
    },
    {
      "epoch": 9.74,
      "learning_rate": 1.5981272727272726e-05,
      "loss": 1.1239,
      "step": 209000
    },
    {
      "epoch": 9.77,
      "learning_rate": 1.597117171717172e-05,
      "loss": 1.1225,
      "step": 209500
    },
    {
      "epoch": 9.79,
      "learning_rate": 1.596107070707071e-05,
      "loss": 1.1202,
      "step": 210000
    },
    {
      "epoch": 9.81,
      "learning_rate": 1.5950989898989898e-05,
      "loss": 1.1234,
      "step": 210500
    },
    {
      "epoch": 9.84,
      "learning_rate": 1.594088888888889e-05,
      "loss": 1.1228,
      "step": 211000
    },
    {
      "epoch": 9.86,
      "learning_rate": 1.5930808080808084e-05,
      "loss": 1.1222,
      "step": 211500
    },
    {
      "epoch": 9.88,
      "learning_rate": 1.5920707070707072e-05,
      "loss": 1.1185,
      "step": 212000
    },
    {
      "epoch": 9.91,
      "learning_rate": 1.591060606060606e-05,
      "loss": 1.122,
      "step": 212500
    },
    {
      "epoch": 9.93,
      "learning_rate": 1.5900505050505053e-05,
      "loss": 1.1208,
      "step": 213000
    },
    {
      "epoch": 9.95,
      "learning_rate": 1.589040404040404e-05,
      "loss": 1.1223,
      "step": 213500
    },
    {
      "epoch": 9.98,
      "learning_rate": 1.5880303030303033e-05,
      "loss": 1.1231,
      "step": 214000
    },
    {
      "epoch": 10.0,
      "learning_rate": 1.5870222222222224e-05,
      "loss": 1.1235,
      "step": 214500
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.7661592131558347,
      "eval_loss": 0.9864925146102905,
      "eval_runtime": 50.7962,
      "eval_samples_per_second": 919.517,
      "eval_steps_per_second": 14.371,
      "step": 214510
    },
    {
      "epoch": 10.02,
      "learning_rate": 1.5860121212121212e-05,
      "loss": 1.1198,
      "step": 215000
    },
    {
      "epoch": 10.05,
      "learning_rate": 1.5850020202020204e-05,
      "loss": 1.1213,
      "step": 215500
    },
    {
      "epoch": 10.07,
      "learning_rate": 1.5839919191919193e-05,
      "loss": 1.1194,
      "step": 216000
    },
    {
      "epoch": 10.09,
      "learning_rate": 1.5829818181818185e-05,
      "loss": 1.1227,
      "step": 216500
    },
    {
      "epoch": 10.12,
      "learning_rate": 1.5819737373737375e-05,
      "loss": 1.1183,
      "step": 217000
    },
    {
      "epoch": 10.14,
      "learning_rate": 1.5809636363636364e-05,
      "loss": 1.121,
      "step": 217500
    },
    {
      "epoch": 10.16,
      "learning_rate": 1.5799535353535356e-05,
      "loss": 1.12,
      "step": 218000
    },
    {
      "epoch": 10.19,
      "learning_rate": 1.5789454545454547e-05,
      "loss": 1.1189,
      "step": 218500
    },
    {
      "epoch": 10.21,
      "learning_rate": 1.5779353535353535e-05,
      "loss": 1.1207,
      "step": 219000
    },
    {
      "epoch": 10.23,
      "learning_rate": 1.5769252525252527e-05,
      "loss": 1.1208,
      "step": 219500
    },
    {
      "epoch": 10.26,
      "learning_rate": 1.5759151515151516e-05,
      "loss": 1.1191,
      "step": 220000
    },
    {
      "epoch": 10.28,
      "learning_rate": 1.5749050505050508e-05,
      "loss": 1.1214,
      "step": 220500
    },
    {
      "epoch": 10.3,
      "learning_rate": 1.5738949494949496e-05,
      "loss": 1.1201,
      "step": 221000
    },
    {
      "epoch": 10.33,
      "learning_rate": 1.572884848484849e-05,
      "loss": 1.1226,
      "step": 221500
    },
    {
      "epoch": 10.35,
      "learning_rate": 1.5718747474747477e-05,
      "loss": 1.1204,
      "step": 222000
    },
    {
      "epoch": 10.37,
      "learning_rate": 1.5708646464646466e-05,
      "loss": 1.1155,
      "step": 222500
    },
    {
      "epoch": 10.4,
      "learning_rate": 1.5698545454545454e-05,
      "loss": 1.1177,
      "step": 223000
    },
    {
      "epoch": 10.42,
      "learning_rate": 1.5688464646464648e-05,
      "loss": 1.1158,
      "step": 223500
    },
    {
      "epoch": 10.44,
      "learning_rate": 1.5678363636363637e-05,
      "loss": 1.1196,
      "step": 224000
    },
    {
      "epoch": 10.47,
      "learning_rate": 1.566826262626263e-05,
      "loss": 1.1186,
      "step": 224500
    },
    {
      "epoch": 10.49,
      "learning_rate": 1.5658161616161617e-05,
      "loss": 1.1182,
      "step": 225000
    },
    {
      "epoch": 10.51,
      "learning_rate": 1.5648060606060606e-05,
      "loss": 1.1181,
      "step": 225500
    },
    {
      "epoch": 10.54,
      "learning_rate": 1.5637959595959598e-05,
      "loss": 1.1173,
      "step": 226000
    },
    {
      "epoch": 10.56,
      "learning_rate": 1.5627858585858586e-05,
      "loss": 1.1184,
      "step": 226500
    },
    {
      "epoch": 10.58,
      "learning_rate": 1.561775757575758e-05,
      "loss": 1.1204,
      "step": 227000
    },
    {
      "epoch": 10.61,
      "learning_rate": 1.5607656565656567e-05,
      "loss": 1.118,
      "step": 227500
    },
    {
      "epoch": 10.63,
      "learning_rate": 1.5597575757575758e-05,
      "loss": 1.1197,
      "step": 228000
    },
    {
      "epoch": 10.65,
      "learning_rate": 1.558747474747475e-05,
      "loss": 1.1179,
      "step": 228500
    },
    {
      "epoch": 10.68,
      "learning_rate": 1.5577373737373738e-05,
      "loss": 1.1179,
      "step": 229000
    },
    {
      "epoch": 10.7,
      "learning_rate": 1.556727272727273e-05,
      "loss": 1.1168,
      "step": 229500
    },
    {
      "epoch": 10.72,
      "learning_rate": 1.555719191919192e-05,
      "loss": 1.119,
      "step": 230000
    },
    {
      "epoch": 10.75,
      "learning_rate": 1.554709090909091e-05,
      "loss": 1.1187,
      "step": 230500
    },
    {
      "epoch": 10.77,
      "learning_rate": 1.55369898989899e-05,
      "loss": 1.1164,
      "step": 231000
    },
    {
      "epoch": 10.79,
      "learning_rate": 1.552688888888889e-05,
      "loss": 1.1196,
      "step": 231500
    },
    {
      "epoch": 10.82,
      "learning_rate": 1.551680808080808e-05,
      "loss": 1.1174,
      "step": 232000
    },
    {
      "epoch": 10.84,
      "learning_rate": 1.5506707070707072e-05,
      "loss": 1.1171,
      "step": 232500
    },
    {
      "epoch": 10.86,
      "learning_rate": 1.549660606060606e-05,
      "loss": 1.1196,
      "step": 233000
    },
    {
      "epoch": 10.89,
      "learning_rate": 1.5486505050505053e-05,
      "loss": 1.1195,
      "step": 233500
    },
    {
      "epoch": 10.91,
      "learning_rate": 1.547640404040404e-05,
      "loss": 1.1168,
      "step": 234000
    },
    {
      "epoch": 10.93,
      "learning_rate": 1.546630303030303e-05,
      "loss": 1.1169,
      "step": 234500
    },
    {
      "epoch": 10.96,
      "learning_rate": 1.5456202020202022e-05,
      "loss": 1.1161,
      "step": 235000
    },
    {
      "epoch": 10.98,
      "learning_rate": 1.544610101010101e-05,
      "loss": 1.1182,
      "step": 235500
    },
    {
      "epoch": 11.0,
      "eval_accuracy": 0.7675014149457302,
      "eval_loss": 0.9782118797302246,
      "eval_runtime": 51.3505,
      "eval_samples_per_second": 909.592,
      "eval_steps_per_second": 14.216,
      "step": 235961
    },
    {
      "epoch": 11.0,
      "learning_rate": 1.5436000000000003e-05,
      "loss": 1.1192,
      "step": 236000
    },
    {
      "epoch": 11.03,
      "learning_rate": 1.5425919191919193e-05,
      "loss": 1.1156,
      "step": 236500
    },
    {
      "epoch": 11.05,
      "learning_rate": 1.5415818181818182e-05,
      "loss": 1.1169,
      "step": 237000
    },
    {
      "epoch": 11.07,
      "learning_rate": 1.5405717171717174e-05,
      "loss": 1.1143,
      "step": 237500
    },
    {
      "epoch": 11.1,
      "learning_rate": 1.5395616161616162e-05,
      "loss": 1.1162,
      "step": 238000
    },
    {
      "epoch": 11.12,
      "learning_rate": 1.5385515151515154e-05,
      "loss": 1.117,
      "step": 238500
    },
    {
      "epoch": 11.14,
      "learning_rate": 1.5375434343434345e-05,
      "loss": 1.1138,
      "step": 239000
    },
    {
      "epoch": 11.16,
      "learning_rate": 1.5365333333333333e-05,
      "loss": 1.1179,
      "step": 239500
    },
    {
      "epoch": 11.19,
      "learning_rate": 1.5355272727272726e-05,
      "loss": 1.1152,
      "step": 240000
    },
    {
      "epoch": 11.21,
      "learning_rate": 1.5345171717171718e-05,
      "loss": 1.1129,
      "step": 240500
    },
    {
      "epoch": 11.23,
      "learning_rate": 1.5335070707070707e-05,
      "loss": 1.1149,
      "step": 241000
    },
    {
      "epoch": 11.26,
      "learning_rate": 1.53249696969697e-05,
      "loss": 1.1136,
      "step": 241500
    },
    {
      "epoch": 11.28,
      "learning_rate": 1.531486868686869e-05,
      "loss": 1.1158,
      "step": 242000
    },
    {
      "epoch": 11.3,
      "learning_rate": 1.530476767676768e-05,
      "loss": 1.1135,
      "step": 242500
    },
    {
      "epoch": 11.33,
      "learning_rate": 1.5294666666666668e-05,
      "loss": 1.1129,
      "step": 243000
    },
    {
      "epoch": 11.35,
      "learning_rate": 1.5284565656565656e-05,
      "loss": 1.1154,
      "step": 243500
    },
    {
      "epoch": 11.37,
      "learning_rate": 1.5274464646464648e-05,
      "loss": 1.1155,
      "step": 244000
    },
    {
      "epoch": 11.4,
      "learning_rate": 1.5264363636363637e-05,
      "loss": 1.1138,
      "step": 244500
    },
    {
      "epoch": 11.42,
      "learning_rate": 1.5254282828282829e-05,
      "loss": 1.1128,
      "step": 245000
    },
    {
      "epoch": 11.44,
      "learning_rate": 1.524418181818182e-05,
      "loss": 1.1137,
      "step": 245500
    },
    {
      "epoch": 11.47,
      "learning_rate": 1.523408080808081e-05,
      "loss": 1.1123,
      "step": 246000
    },
    {
      "epoch": 11.49,
      "learning_rate": 1.52239797979798e-05,
      "loss": 1.1144,
      "step": 246500
    },
    {
      "epoch": 11.51,
      "learning_rate": 1.5213878787878789e-05,
      "loss": 1.1128,
      "step": 247000
    },
    {
      "epoch": 11.54,
      "learning_rate": 1.5203777777777779e-05,
      "loss": 1.113,
      "step": 247500
    },
    {
      "epoch": 11.56,
      "learning_rate": 1.5193676767676767e-05,
      "loss": 1.1152,
      "step": 248000
    },
    {
      "epoch": 11.58,
      "learning_rate": 1.518357575757576e-05,
      "loss": 1.111,
      "step": 248500
    },
    {
      "epoch": 11.61,
      "learning_rate": 1.5173515151515152e-05,
      "loss": 1.1155,
      "step": 249000
    },
    {
      "epoch": 11.63,
      "learning_rate": 1.5163414141414144e-05,
      "loss": 1.1148,
      "step": 249500
    },
    {
      "epoch": 11.65,
      "learning_rate": 1.5153313131313133e-05,
      "loss": 1.1143,
      "step": 250000
    },
    {
      "epoch": 11.68,
      "learning_rate": 1.5143212121212123e-05,
      "loss": 1.1141,
      "step": 250500
    },
    {
      "epoch": 11.7,
      "learning_rate": 1.5133111111111111e-05,
      "loss": 1.1133,
      "step": 251000
    },
    {
      "epoch": 11.72,
      "learning_rate": 1.5123010101010103e-05,
      "loss": 1.1121,
      "step": 251500
    },
    {
      "epoch": 11.75,
      "learning_rate": 1.5112929292929294e-05,
      "loss": 1.1131,
      "step": 252000
    },
    {
      "epoch": 11.77,
      "learning_rate": 1.5102828282828284e-05,
      "loss": 1.1143,
      "step": 252500
    },
    {
      "epoch": 11.79,
      "learning_rate": 1.5092727272727275e-05,
      "loss": 1.1148,
      "step": 253000
    },
    {
      "epoch": 11.82,
      "learning_rate": 1.5082646464646467e-05,
      "loss": 1.1113,
      "step": 253500
    },
    {
      "epoch": 11.84,
      "learning_rate": 1.5072545454545455e-05,
      "loss": 1.1159,
      "step": 254000
    },
    {
      "epoch": 11.86,
      "learning_rate": 1.5062444444444446e-05,
      "loss": 1.1115,
      "step": 254500
    },
    {
      "epoch": 11.89,
      "learning_rate": 1.5052343434343434e-05,
      "loss": 1.1128,
      "step": 255000
    },
    {
      "epoch": 11.91,
      "learning_rate": 1.5042242424242426e-05,
      "loss": 1.113,
      "step": 255500
    },
    {
      "epoch": 11.93,
      "learning_rate": 1.5032161616161618e-05,
      "loss": 1.1116,
      "step": 256000
    },
    {
      "epoch": 11.96,
      "learning_rate": 1.5022060606060607e-05,
      "loss": 1.11,
      "step": 256500
    },
    {
      "epoch": 11.98,
      "learning_rate": 1.5011959595959597e-05,
      "loss": 1.1151,
      "step": 257000
    },
    {
      "epoch": 12.0,
      "eval_accuracy": 0.7684468930646907,
      "eval_loss": 0.9742035865783691,
      "eval_runtime": 51.4063,
      "eval_samples_per_second": 908.605,
      "eval_steps_per_second": 14.201,
      "step": 257412
    },
    {
      "epoch": 12.0,
      "learning_rate": 1.5001858585858586e-05,
      "loss": 1.1157,
      "step": 257500
    },
    {
      "epoch": 12.03,
      "learning_rate": 1.4991757575757578e-05,
      "loss": 1.1123,
      "step": 258000
    },
    {
      "epoch": 12.05,
      "learning_rate": 1.4981656565656567e-05,
      "loss": 1.1111,
      "step": 258500
    },
    {
      "epoch": 12.07,
      "learning_rate": 1.4971555555555557e-05,
      "loss": 1.1096,
      "step": 259000
    },
    {
      "epoch": 12.1,
      "learning_rate": 1.4961454545454545e-05,
      "loss": 1.1111,
      "step": 259500
    },
    {
      "epoch": 12.12,
      "learning_rate": 1.4951353535353537e-05,
      "loss": 1.114,
      "step": 260000
    },
    {
      "epoch": 12.14,
      "learning_rate": 1.4941272727272728e-05,
      "loss": 1.1104,
      "step": 260500
    },
    {
      "epoch": 12.17,
      "learning_rate": 1.493119191919192e-05,
      "loss": 1.1094,
      "step": 261000
    },
    {
      "epoch": 12.19,
      "learning_rate": 1.4921090909090909e-05,
      "loss": 1.1107,
      "step": 261500
    },
    {
      "epoch": 12.21,
      "learning_rate": 1.49109898989899e-05,
      "loss": 1.1094,
      "step": 262000
    },
    {
      "epoch": 12.24,
      "learning_rate": 1.490088888888889e-05,
      "loss": 1.1117,
      "step": 262500
    },
    {
      "epoch": 12.26,
      "learning_rate": 1.489078787878788e-05,
      "loss": 1.1101,
      "step": 263000
    },
    {
      "epoch": 12.28,
      "learning_rate": 1.4880707070707072e-05,
      "loss": 1.1114,
      "step": 263500
    },
    {
      "epoch": 12.31,
      "learning_rate": 1.487060606060606e-05,
      "loss": 1.1084,
      "step": 264000
    },
    {
      "epoch": 12.33,
      "learning_rate": 1.4860505050505052e-05,
      "loss": 1.1113,
      "step": 264500
    },
    {
      "epoch": 12.35,
      "learning_rate": 1.4850404040404041e-05,
      "loss": 1.1093,
      "step": 265000
    },
    {
      "epoch": 12.38,
      "learning_rate": 1.4840303030303031e-05,
      "loss": 1.1097,
      "step": 265500
    },
    {
      "epoch": 12.4,
      "learning_rate": 1.4830202020202023e-05,
      "loss": 1.111,
      "step": 266000
    },
    {
      "epoch": 12.42,
      "learning_rate": 1.4820101010101012e-05,
      "loss": 1.1097,
      "step": 266500
    },
    {
      "epoch": 12.45,
      "learning_rate": 1.4810000000000002e-05,
      "loss": 1.108,
      "step": 267000
    },
    {
      "epoch": 12.47,
      "learning_rate": 1.479989898989899e-05,
      "loss": 1.1126,
      "step": 267500
    },
    {
      "epoch": 12.49,
      "learning_rate": 1.4789818181818183e-05,
      "loss": 1.1091,
      "step": 268000
    },
    {
      "epoch": 12.52,
      "learning_rate": 1.4779717171717173e-05,
      "loss": 1.1081,
      "step": 268500
    },
    {
      "epoch": 12.54,
      "learning_rate": 1.4769616161616162e-05,
      "loss": 1.1101,
      "step": 269000
    },
    {
      "epoch": 12.56,
      "learning_rate": 1.4759515151515154e-05,
      "loss": 1.1087,
      "step": 269500
    },
    {
      "epoch": 12.59,
      "learning_rate": 1.4749414141414142e-05,
      "loss": 1.1082,
      "step": 270000
    },
    {
      "epoch": 12.61,
      "learning_rate": 1.4739313131313133e-05,
      "loss": 1.1082,
      "step": 270500
    },
    {
      "epoch": 12.63,
      "learning_rate": 1.4729232323232325e-05,
      "loss": 1.1093,
      "step": 271000
    },
    {
      "epoch": 12.66,
      "learning_rate": 1.4719131313131314e-05,
      "loss": 1.109,
      "step": 271500
    },
    {
      "epoch": 12.68,
      "learning_rate": 1.4709030303030306e-05,
      "loss": 1.1084,
      "step": 272000
    },
    {
      "epoch": 12.7,
      "learning_rate": 1.4698929292929294e-05,
      "loss": 1.111,
      "step": 272500
    },
    {
      "epoch": 12.73,
      "learning_rate": 1.4688828282828284e-05,
      "loss": 1.1088,
      "step": 273000
    },
    {
      "epoch": 12.75,
      "learning_rate": 1.4678727272727273e-05,
      "loss": 1.1101,
      "step": 273500
    },
    {
      "epoch": 12.77,
      "learning_rate": 1.4668626262626265e-05,
      "loss": 1.1083,
      "step": 274000
    },
    {
      "epoch": 12.8,
      "learning_rate": 1.4658525252525254e-05,
      "loss": 1.108,
      "step": 274500
    },
    {
      "epoch": 12.82,
      "learning_rate": 1.4648424242424244e-05,
      "loss": 1.1112,
      "step": 275000
    },
    {
      "epoch": 12.84,
      "learning_rate": 1.4638343434343436e-05,
      "loss": 1.1078,
      "step": 275500
    },
    {
      "epoch": 12.87,
      "learning_rate": 1.4628242424242425e-05,
      "loss": 1.1091,
      "step": 276000
    },
    {
      "epoch": 12.89,
      "learning_rate": 1.4618161616161617e-05,
      "loss": 1.1067,
      "step": 276500
    },
    {
      "epoch": 12.91,
      "learning_rate": 1.4608060606060607e-05,
      "loss": 1.1082,
      "step": 277000
    },
    {
      "epoch": 12.94,
      "learning_rate": 1.4597959595959596e-05,
      "loss": 1.1079,
      "step": 277500
    },
    {
      "epoch": 12.96,
      "learning_rate": 1.4587858585858588e-05,
      "loss": 1.1081,
      "step": 278000
    },
    {
      "epoch": 12.98,
      "learning_rate": 1.4577757575757576e-05,
      "loss": 1.1109,
      "step": 278500
    },
    {
      "epoch": 13.0,
      "eval_accuracy": 0.7684372273507248,
      "eval_loss": 0.9741110801696777,
      "eval_runtime": 51.6014,
      "eval_samples_per_second": 905.169,
      "eval_steps_per_second": 14.147,
      "step": 278863
    },
    {
      "epoch": 13.01,
      "learning_rate": 1.4567676767676769e-05,
      "loss": 1.1076,
      "step": 279000
    },
    {
      "epoch": 13.03,
      "learning_rate": 1.4557575757575759e-05,
      "loss": 1.107,
      "step": 279500
    },
    {
      "epoch": 13.05,
      "learning_rate": 1.4547474747474748e-05,
      "loss": 1.1065,
      "step": 280000
    },
    {
      "epoch": 13.08,
      "learning_rate": 1.453737373737374e-05,
      "loss": 1.1072,
      "step": 280500
    },
    {
      "epoch": 13.1,
      "learning_rate": 1.4527272727272728e-05,
      "loss": 1.1066,
      "step": 281000
    },
    {
      "epoch": 13.12,
      "learning_rate": 1.4517212121212122e-05,
      "loss": 1.1024,
      "step": 281500
    },
    {
      "epoch": 13.15,
      "learning_rate": 1.4507111111111111e-05,
      "loss": 1.1094,
      "step": 282000
    },
    {
      "epoch": 13.17,
      "learning_rate": 1.4497010101010103e-05,
      "loss": 1.1071,
      "step": 282500
    },
    {
      "epoch": 13.19,
      "learning_rate": 1.4486909090909092e-05,
      "loss": 1.1039,
      "step": 283000
    },
    {
      "epoch": 13.22,
      "learning_rate": 1.4476808080808082e-05,
      "loss": 1.1071,
      "step": 283500
    },
    {
      "epoch": 13.24,
      "learning_rate": 1.446670707070707e-05,
      "loss": 1.1045,
      "step": 284000
    },
    {
      "epoch": 13.26,
      "learning_rate": 1.4456606060606062e-05,
      "loss": 1.1058,
      "step": 284500
    },
    {
      "epoch": 13.29,
      "learning_rate": 1.4446505050505051e-05,
      "loss": 1.1054,
      "step": 285000
    },
    {
      "epoch": 13.31,
      "learning_rate": 1.4436404040404041e-05,
      "loss": 1.1068,
      "step": 285500
    },
    {
      "epoch": 13.33,
      "learning_rate": 1.4426323232323234e-05,
      "loss": 1.1082,
      "step": 286000
    },
    {
      "epoch": 13.36,
      "learning_rate": 1.4416222222222222e-05,
      "loss": 1.1056,
      "step": 286500
    },
    {
      "epoch": 13.38,
      "learning_rate": 1.4406121212121214e-05,
      "loss": 1.1063,
      "step": 287000
    },
    {
      "epoch": 13.4,
      "learning_rate": 1.4396020202020203e-05,
      "loss": 1.1057,
      "step": 287500
    },
    {
      "epoch": 13.43,
      "learning_rate": 1.4385939393939395e-05,
      "loss": 1.1059,
      "step": 288000
    },
    {
      "epoch": 13.45,
      "learning_rate": 1.4375838383838385e-05,
      "loss": 1.1071,
      "step": 288500
    },
    {
      "epoch": 13.47,
      "learning_rate": 1.4365737373737374e-05,
      "loss": 1.1059,
      "step": 289000
    },
    {
      "epoch": 13.5,
      "learning_rate": 1.4355656565656566e-05,
      "loss": 1.1064,
      "step": 289500
    },
    {
      "epoch": 13.52,
      "learning_rate": 1.4345555555555556e-05,
      "loss": 1.107,
      "step": 290000
    },
    {
      "epoch": 13.54,
      "learning_rate": 1.4335454545454545e-05,
      "loss": 1.1054,
      "step": 290500
    },
    {
      "epoch": 13.57,
      "learning_rate": 1.4325353535353537e-05,
      "loss": 1.1072,
      "step": 291000
    },
    {
      "epoch": 13.59,
      "learning_rate": 1.4315252525252526e-05,
      "loss": 1.1045,
      "step": 291500
    },
    {
      "epoch": 13.61,
      "learning_rate": 1.4305151515151516e-05,
      "loss": 1.1036,
      "step": 292000
    },
    {
      "epoch": 13.64,
      "learning_rate": 1.4295050505050508e-05,
      "loss": 1.105,
      "step": 292500
    },
    {
      "epoch": 13.66,
      "learning_rate": 1.4284949494949496e-05,
      "loss": 1.1059,
      "step": 293000
    },
    {
      "epoch": 13.68,
      "learning_rate": 1.4274848484848487e-05,
      "loss": 1.105,
      "step": 293500
    },
    {
      "epoch": 13.71,
      "learning_rate": 1.4264747474747475e-05,
      "loss": 1.1042,
      "step": 294000
    },
    {
      "epoch": 13.73,
      "learning_rate": 1.4254646464646467e-05,
      "loss": 1.1053,
      "step": 294500
    },
    {
      "epoch": 13.75,
      "learning_rate": 1.4244545454545456e-05,
      "loss": 1.1051,
      "step": 295000
    },
    {
      "epoch": 13.78,
      "learning_rate": 1.4234444444444446e-05,
      "loss": 1.1062,
      "step": 295500
    },
    {
      "epoch": 13.8,
      "learning_rate": 1.4224343434343435e-05,
      "loss": 1.1045,
      "step": 296000
    },
    {
      "epoch": 13.82,
      "learning_rate": 1.4214242424242427e-05,
      "loss": 1.1054,
      "step": 296500
    },
    {
      "epoch": 13.85,
      "learning_rate": 1.4204161616161617e-05,
      "loss": 1.1044,
      "step": 297000
    },
    {
      "epoch": 13.87,
      "learning_rate": 1.419408080808081e-05,
      "loss": 1.1066,
      "step": 297500
    },
    {
      "epoch": 13.89,
      "learning_rate": 1.4183979797979798e-05,
      "loss": 1.1042,
      "step": 298000
    },
    {
      "epoch": 13.92,
      "learning_rate": 1.417389898989899e-05,
      "loss": 1.1051,
      "step": 298500
    },
    {
      "epoch": 13.94,
      "learning_rate": 1.4163797979797982e-05,
      "loss": 1.1027,
      "step": 299000
    },
    {
      "epoch": 13.96,
      "learning_rate": 1.4153696969696971e-05,
      "loss": 1.1043,
      "step": 299500
    },
    {
      "epoch": 13.99,
      "learning_rate": 1.4143595959595961e-05,
      "loss": 1.1042,
      "step": 300000
    },
    {
      "epoch": 14.0,
      "eval_accuracy": 0.7696585567375857,
      "eval_loss": 0.9662981629371643,
      "eval_runtime": 51.573,
      "eval_samples_per_second": 905.668,
      "eval_steps_per_second": 14.155,
      "step": 300314
    },
    {
      "epoch": 14.01,
      "learning_rate": 1.413349494949495e-05,
      "loss": 1.1034,
      "step": 300500
    },
    {
      "epoch": 14.03,
      "learning_rate": 1.4123393939393942e-05,
      "loss": 1.1024,
      "step": 301000
    },
    {
      "epoch": 14.06,
      "learning_rate": 1.411329292929293e-05,
      "loss": 1.1039,
      "step": 301500
    },
    {
      "epoch": 14.08,
      "learning_rate": 1.410319191919192e-05,
      "loss": 1.103,
      "step": 302000
    },
    {
      "epoch": 14.1,
      "learning_rate": 1.409309090909091e-05,
      "loss": 1.1028,
      "step": 302500
    },
    {
      "epoch": 14.13,
      "learning_rate": 1.4083010101010102e-05,
      "loss": 1.1034,
      "step": 303000
    },
    {
      "epoch": 14.15,
      "learning_rate": 1.4072909090909092e-05,
      "loss": 1.1017,
      "step": 303500
    },
    {
      "epoch": 14.17,
      "learning_rate": 1.4062808080808082e-05,
      "loss": 1.1021,
      "step": 304000
    },
    {
      "epoch": 14.2,
      "learning_rate": 1.4052707070707072e-05,
      "loss": 1.1018,
      "step": 304500
    },
    {
      "epoch": 14.22,
      "learning_rate": 1.4042606060606061e-05,
      "loss": 1.1008,
      "step": 305000
    },
    {
      "epoch": 14.24,
      "learning_rate": 1.4032525252525253e-05,
      "loss": 1.1044,
      "step": 305500
    },
    {
      "epoch": 14.27,
      "learning_rate": 1.4022424242424243e-05,
      "loss": 1.1025,
      "step": 306000
    },
    {
      "epoch": 14.29,
      "learning_rate": 1.4012323232323232e-05,
      "loss": 1.1018,
      "step": 306500
    },
    {
      "epoch": 14.31,
      "learning_rate": 1.4002222222222224e-05,
      "loss": 1.1035,
      "step": 307000
    },
    {
      "epoch": 14.33,
      "learning_rate": 1.3992121212121213e-05,
      "loss": 1.1052,
      "step": 307500
    },
    {
      "epoch": 14.36,
      "learning_rate": 1.3982020202020203e-05,
      "loss": 1.1019,
      "step": 308000
    },
    {
      "epoch": 14.38,
      "learning_rate": 1.3971919191919192e-05,
      "loss": 1.1026,
      "step": 308500
    },
    {
      "epoch": 14.4,
      "learning_rate": 1.3961818181818184e-05,
      "loss": 1.1033,
      "step": 309000
    },
    {
      "epoch": 14.43,
      "learning_rate": 1.3951717171717172e-05,
      "loss": 1.1028,
      "step": 309500
    },
    {
      "epoch": 14.45,
      "learning_rate": 1.3941636363636364e-05,
      "loss": 1.1007,
      "step": 310000
    },
    {
      "epoch": 14.47,
      "learning_rate": 1.3931555555555557e-05,
      "loss": 1.1017,
      "step": 310500
    },
    {
      "epoch": 14.5,
      "learning_rate": 1.3921454545454547e-05,
      "loss": 1.1026,
      "step": 311000
    },
    {
      "epoch": 14.52,
      "learning_rate": 1.3911373737373739e-05,
      "loss": 1.1011,
      "step": 311500
    },
    {
      "epoch": 14.54,
      "learning_rate": 1.3901272727272728e-05,
      "loss": 1.104,
      "step": 312000
    },
    {
      "epoch": 14.57,
      "learning_rate": 1.3891171717171718e-05,
      "loss": 1.103,
      "step": 312500
    },
    {
      "epoch": 14.59,
      "learning_rate": 1.3881070707070707e-05,
      "loss": 1.1017,
      "step": 313000
    },
    {
      "epoch": 14.61,
      "learning_rate": 1.3870969696969699e-05,
      "loss": 1.1015,
      "step": 313500
    },
    {
      "epoch": 14.64,
      "learning_rate": 1.3860868686868687e-05,
      "loss": 1.1031,
      "step": 314000
    },
    {
      "epoch": 14.66,
      "learning_rate": 1.3850767676767677e-05,
      "loss": 1.1019,
      "step": 314500
    },
    {
      "epoch": 14.68,
      "learning_rate": 1.3840666666666666e-05,
      "loss": 1.1014,
      "step": 315000
    },
    {
      "epoch": 14.71,
      "learning_rate": 1.3830565656565658e-05,
      "loss": 1.0998,
      "step": 315500
    },
    {
      "epoch": 14.73,
      "learning_rate": 1.3820464646464648e-05,
      "loss": 1.1004,
      "step": 316000
    },
    {
      "epoch": 14.75,
      "learning_rate": 1.3810363636363637e-05,
      "loss": 1.0991,
      "step": 316500
    },
    {
      "epoch": 14.78,
      "learning_rate": 1.3800262626262629e-05,
      "loss": 1.1007,
      "step": 317000
    },
    {
      "epoch": 14.8,
      "learning_rate": 1.3790202020202021e-05,
      "loss": 1.1012,
      "step": 317500
    },
    {
      "epoch": 14.82,
      "learning_rate": 1.378010101010101e-05,
      "loss": 1.1018,
      "step": 318000
    },
    {
      "epoch": 14.85,
      "learning_rate": 1.377e-05,
      "loss": 1.1001,
      "step": 318500
    },
    {
      "epoch": 14.87,
      "learning_rate": 1.3759898989898992e-05,
      "loss": 1.0996,
      "step": 319000
    },
    {
      "epoch": 14.89,
      "learning_rate": 1.3749818181818181e-05,
      "loss": 1.1006,
      "step": 319500
    },
    {
      "epoch": 14.92,
      "learning_rate": 1.3739717171717173e-05,
      "loss": 1.1007,
      "step": 320000
    },
    {
      "epoch": 14.94,
      "learning_rate": 1.3729616161616163e-05,
      "loss": 1.1011,
      "step": 320500
    },
    {
      "epoch": 14.96,
      "learning_rate": 1.3719515151515152e-05,
      "loss": 1.1008,
      "step": 321000
    },
    {
      "epoch": 14.99,
      "learning_rate": 1.3709414141414144e-05,
      "loss": 1.0998,
      "step": 321500
    },
    {
      "epoch": 15.0,
      "eval_accuracy": 0.7700938270723376,
      "eval_loss": 0.9633080959320068,
      "eval_runtime": 51.1078,
      "eval_samples_per_second": 913.911,
      "eval_steps_per_second": 14.284,
      "step": 321765
    },
    {
      "epoch": 15.01,
      "learning_rate": 1.3699313131313133e-05,
      "loss": 1.1028,
      "step": 322000
    },
    {
      "epoch": 15.03,
      "learning_rate": 1.3689212121212123e-05,
      "loss": 1.0999,
      "step": 322500
    },
    {
      "epoch": 15.06,
      "learning_rate": 1.3679111111111111e-05,
      "loss": 1.0985,
      "step": 323000
    },
    {
      "epoch": 15.08,
      "learning_rate": 1.3669010101010103e-05,
      "loss": 1.0983,
      "step": 323500
    },
    {
      "epoch": 15.1,
      "learning_rate": 1.3658909090909092e-05,
      "loss": 1.0978,
      "step": 324000
    },
    {
      "epoch": 15.13,
      "learning_rate": 1.3648808080808082e-05,
      "loss": 1.1012,
      "step": 324500
    },
    {
      "epoch": 15.15,
      "learning_rate": 1.3638727272727275e-05,
      "loss": 1.0985,
      "step": 325000
    },
    {
      "epoch": 15.17,
      "learning_rate": 1.3628626262626263e-05,
      "loss": 1.099,
      "step": 325500
    },
    {
      "epoch": 15.2,
      "learning_rate": 1.3618525252525253e-05,
      "loss": 1.0998,
      "step": 326000
    },
    {
      "epoch": 15.22,
      "learning_rate": 1.3608424242424244e-05,
      "loss": 1.0992,
      "step": 326500
    },
    {
      "epoch": 15.24,
      "learning_rate": 1.3598323232323234e-05,
      "loss": 1.0981,
      "step": 327000
    },
    {
      "epoch": 15.27,
      "learning_rate": 1.3588222222222223e-05,
      "loss": 1.0986,
      "step": 327500
    },
    {
      "epoch": 15.29,
      "learning_rate": 1.3578141414141415e-05,
      "loss": 1.0988,
      "step": 328000
    },
    {
      "epoch": 15.31,
      "learning_rate": 1.3568060606060607e-05,
      "loss": 1.1007,
      "step": 328500
    },
    {
      "epoch": 15.34,
      "learning_rate": 1.3557959595959597e-05,
      "loss": 1.1017,
      "step": 329000
    },
    {
      "epoch": 15.36,
      "learning_rate": 1.3547858585858586e-05,
      "loss": 1.1006,
      "step": 329500
    },
    {
      "epoch": 15.38,
      "learning_rate": 1.3537757575757578e-05,
      "loss": 1.0952,
      "step": 330000
    },
    {
      "epoch": 15.41,
      "learning_rate": 1.3527656565656567e-05,
      "loss": 1.0987,
      "step": 330500
    },
    {
      "epoch": 15.43,
      "learning_rate": 1.3517555555555557e-05,
      "loss": 1.099,
      "step": 331000
    },
    {
      "epoch": 15.45,
      "learning_rate": 1.3507454545454545e-05,
      "loss": 1.099,
      "step": 331500
    },
    {
      "epoch": 15.48,
      "learning_rate": 1.3497353535353537e-05,
      "loss": 1.1001,
      "step": 332000
    },
    {
      "epoch": 15.5,
      "learning_rate": 1.3487252525252526e-05,
      "loss": 1.0952,
      "step": 332500
    },
    {
      "epoch": 15.52,
      "learning_rate": 1.3477151515151516e-05,
      "loss": 1.0984,
      "step": 333000
    },
    {
      "epoch": 15.55,
      "learning_rate": 1.3467050505050505e-05,
      "loss": 1.0977,
      "step": 333500
    },
    {
      "epoch": 15.57,
      "learning_rate": 1.3456969696969697e-05,
      "loss": 1.0982,
      "step": 334000
    },
    {
      "epoch": 15.59,
      "learning_rate": 1.3446868686868687e-05,
      "loss": 1.0958,
      "step": 334500
    },
    {
      "epoch": 15.62,
      "learning_rate": 1.3436767676767678e-05,
      "loss": 1.0998,
      "step": 335000
    },
    {
      "epoch": 15.64,
      "learning_rate": 1.3426666666666668e-05,
      "loss": 1.0982,
      "step": 335500
    },
    {
      "epoch": 15.66,
      "learning_rate": 1.3416565656565657e-05,
      "loss": 1.0989,
      "step": 336000
    },
    {
      "epoch": 15.69,
      "learning_rate": 1.3406464646464647e-05,
      "loss": 1.0972,
      "step": 336500
    },
    {
      "epoch": 15.71,
      "learning_rate": 1.3396363636363639e-05,
      "loss": 1.0982,
      "step": 337000
    },
    {
      "epoch": 15.73,
      "learning_rate": 1.3386262626262627e-05,
      "loss": 1.0968,
      "step": 337500
    },
    {
      "epoch": 15.76,
      "learning_rate": 1.3376161616161618e-05,
      "loss": 1.0973,
      "step": 338000
    },
    {
      "epoch": 15.78,
      "learning_rate": 1.336608080808081e-05,
      "loss": 1.0979,
      "step": 338500
    },
    {
      "epoch": 15.8,
      "learning_rate": 1.3355979797979799e-05,
      "loss": 1.0956,
      "step": 339000
    },
    {
      "epoch": 15.83,
      "learning_rate": 1.334587878787879e-05,
      "loss": 1.0967,
      "step": 339500
    },
    {
      "epoch": 15.85,
      "learning_rate": 1.3335777777777779e-05,
      "loss": 1.0962,
      "step": 340000
    },
    {
      "epoch": 15.87,
      "learning_rate": 1.3325717171717172e-05,
      "loss": 1.098,
      "step": 340500
    },
    {
      "epoch": 15.9,
      "learning_rate": 1.3315616161616162e-05,
      "loss": 1.0988,
      "step": 341000
    },
    {
      "epoch": 15.92,
      "learning_rate": 1.3305535353535354e-05,
      "loss": 1.0984,
      "step": 341500
    },
    {
      "epoch": 15.94,
      "learning_rate": 1.3295434343434343e-05,
      "loss": 1.0968,
      "step": 342000
    },
    {
      "epoch": 15.97,
      "learning_rate": 1.3285333333333335e-05,
      "loss": 1.0953,
      "step": 342500
    },
    {
      "epoch": 15.99,
      "learning_rate": 1.3275232323232323e-05,
      "loss": 1.0992,
      "step": 343000
    },
    {
      "epoch": 16.0,
      "eval_accuracy": 0.770279722036309,
      "eval_loss": 0.961768627166748,
      "eval_runtime": 51.6814,
      "eval_samples_per_second": 903.767,
      "eval_steps_per_second": 14.125,
      "step": 343216
    },
    {
      "epoch": 16.01,
      "learning_rate": 1.3265131313131314e-05,
      "loss": 1.0965,
      "step": 343500
    },
    {
      "epoch": 16.04,
      "learning_rate": 1.3255030303030306e-05,
      "loss": 1.0976,
      "step": 344000
    },
    {
      "epoch": 16.06,
      "learning_rate": 1.3244929292929294e-05,
      "loss": 1.0957,
      "step": 344500
    },
    {
      "epoch": 16.08,
      "learning_rate": 1.3234828282828285e-05,
      "loss": 1.0948,
      "step": 345000
    },
    {
      "epoch": 16.11,
      "learning_rate": 1.3224727272727273e-05,
      "loss": 1.0944,
      "step": 345500
    },
    {
      "epoch": 16.13,
      "learning_rate": 1.3214646464646465e-05,
      "loss": 1.0939,
      "step": 346000
    },
    {
      "epoch": 16.15,
      "learning_rate": 1.3204545454545457e-05,
      "loss": 1.0937,
      "step": 346500
    },
    {
      "epoch": 16.18,
      "learning_rate": 1.3194444444444446e-05,
      "loss": 1.0966,
      "step": 347000
    },
    {
      "epoch": 16.2,
      "learning_rate": 1.3184343434343436e-05,
      "loss": 1.0966,
      "step": 347500
    },
    {
      "epoch": 16.22,
      "learning_rate": 1.3174242424242425e-05,
      "loss": 1.0959,
      "step": 348000
    },
    {
      "epoch": 16.25,
      "learning_rate": 1.3164141414141415e-05,
      "loss": 1.098,
      "step": 348500
    },
    {
      "epoch": 16.27,
      "learning_rate": 1.3154040404040405e-05,
      "loss": 1.0967,
      "step": 349000
    },
    {
      "epoch": 16.29,
      "learning_rate": 1.3143939393939396e-05,
      "loss": 1.0958,
      "step": 349500
    },
    {
      "epoch": 16.32,
      "learning_rate": 1.3133858585858588e-05,
      "loss": 1.0955,
      "step": 350000
    },
    {
      "epoch": 16.34,
      "learning_rate": 1.312377777777778e-05,
      "loss": 1.0974,
      "step": 350500
    },
    {
      "epoch": 16.36,
      "learning_rate": 1.3113676767676769e-05,
      "loss": 1.0951,
      "step": 351000
    },
    {
      "epoch": 16.39,
      "learning_rate": 1.3103575757575759e-05,
      "loss": 1.0959,
      "step": 351500
    },
    {
      "epoch": 16.41,
      "learning_rate": 1.3093474747474748e-05,
      "loss": 1.095,
      "step": 352000
    },
    {
      "epoch": 16.43,
      "learning_rate": 1.308337373737374e-05,
      "loss": 1.0942,
      "step": 352500
    },
    {
      "epoch": 16.46,
      "learning_rate": 1.3073272727272728e-05,
      "loss": 1.0959,
      "step": 353000
    },
    {
      "epoch": 16.48,
      "learning_rate": 1.3063171717171718e-05,
      "loss": 1.0949,
      "step": 353500
    },
    {
      "epoch": 16.5,
      "learning_rate": 1.3053070707070707e-05,
      "loss": 1.0955,
      "step": 354000
    },
    {
      "epoch": 16.53,
      "learning_rate": 1.3042969696969699e-05,
      "loss": 1.0941,
      "step": 354500
    },
    {
      "epoch": 16.55,
      "learning_rate": 1.3032868686868688e-05,
      "loss": 1.0983,
      "step": 355000
    },
    {
      "epoch": 16.57,
      "learning_rate": 1.3022767676767678e-05,
      "loss": 1.0961,
      "step": 355500
    },
    {
      "epoch": 16.6,
      "learning_rate": 1.301268686868687e-05,
      "loss": 1.0971,
      "step": 356000
    },
    {
      "epoch": 16.62,
      "learning_rate": 1.3002585858585859e-05,
      "loss": 1.0962,
      "step": 356500
    },
    {
      "epoch": 16.64,
      "learning_rate": 1.2992484848484849e-05,
      "loss": 1.0965,
      "step": 357000
    },
    {
      "epoch": 16.67,
      "learning_rate": 1.298238383838384e-05,
      "loss": 1.094,
      "step": 357500
    },
    {
      "epoch": 16.69,
      "learning_rate": 1.297230303030303e-05,
      "loss": 1.0953,
      "step": 358000
    },
    {
      "epoch": 16.71,
      "learning_rate": 1.2962202020202022e-05,
      "loss": 1.0932,
      "step": 358500
    },
    {
      "epoch": 16.74,
      "learning_rate": 1.295210101010101e-05,
      "loss": 1.0952,
      "step": 359000
    },
    {
      "epoch": 16.76,
      "learning_rate": 1.2942020202020203e-05,
      "loss": 1.0952,
      "step": 359500
    },
    {
      "epoch": 16.78,
      "learning_rate": 1.2931919191919193e-05,
      "loss": 1.097,
      "step": 360000
    },
    {
      "epoch": 16.81,
      "learning_rate": 1.2921818181818182e-05,
      "loss": 1.0947,
      "step": 360500
    },
    {
      "epoch": 16.83,
      "learning_rate": 1.2911717171717174e-05,
      "loss": 1.0936,
      "step": 361000
    },
    {
      "epoch": 16.85,
      "learning_rate": 1.2901616161616162e-05,
      "loss": 1.0962,
      "step": 361500
    },
    {
      "epoch": 16.88,
      "learning_rate": 1.2891515151515152e-05,
      "loss": 1.0961,
      "step": 362000
    },
    {
      "epoch": 16.9,
      "learning_rate": 1.2881414141414141e-05,
      "loss": 1.0941,
      "step": 362500
    },
    {
      "epoch": 16.92,
      "learning_rate": 1.2871313131313133e-05,
      "loss": 1.0943,
      "step": 363000
    },
    {
      "epoch": 16.95,
      "learning_rate": 1.2861212121212123e-05,
      "loss": 1.0937,
      "step": 363500
    },
    {
      "epoch": 16.97,
      "learning_rate": 1.2851131313131314e-05,
      "loss": 1.0949,
      "step": 364000
    },
    {
      "epoch": 16.99,
      "learning_rate": 1.2841030303030304e-05,
      "loss": 1.0948,
      "step": 364500
    },
    {
      "epoch": 17.0,
      "eval_accuracy": 0.771228150835467,
      "eval_loss": 0.9590633511543274,
      "eval_runtime": 52.0227,
      "eval_samples_per_second": 897.838,
      "eval_steps_per_second": 14.032,
      "step": 364667
    },
    {
      "epoch": 17.02,
      "learning_rate": 1.2830929292929294e-05,
      "loss": 1.095,
      "step": 365000
    },
    {
      "epoch": 17.04,
      "learning_rate": 1.2820828282828283e-05,
      "loss": 1.0942,
      "step": 365500
    },
    {
      "epoch": 17.06,
      "learning_rate": 1.2810727272727275e-05,
      "loss": 1.0963,
      "step": 366000
    },
    {
      "epoch": 17.09,
      "learning_rate": 1.2800646464646467e-05,
      "loss": 1.0902,
      "step": 366500
    },
    {
      "epoch": 17.11,
      "learning_rate": 1.2790545454545456e-05,
      "loss": 1.0918,
      "step": 367000
    },
    {
      "epoch": 17.13,
      "learning_rate": 1.2780444444444446e-05,
      "loss": 1.09,
      "step": 367500
    },
    {
      "epoch": 17.16,
      "learning_rate": 1.2770343434343435e-05,
      "loss": 1.0941,
      "step": 368000
    },
    {
      "epoch": 17.18,
      "learning_rate": 1.2760262626262627e-05,
      "loss": 1.0935,
      "step": 368500
    },
    {
      "epoch": 17.2,
      "learning_rate": 1.275018181818182e-05,
      "loss": 1.0936,
      "step": 369000
    },
    {
      "epoch": 17.23,
      "learning_rate": 1.2740080808080808e-05,
      "loss": 1.0914,
      "step": 369500
    },
    {
      "epoch": 17.25,
      "learning_rate": 1.27299797979798e-05,
      "loss": 1.0932,
      "step": 370000
    },
    {
      "epoch": 17.27,
      "learning_rate": 1.271987878787879e-05,
      "loss": 1.0928,
      "step": 370500
    },
    {
      "epoch": 17.3,
      "learning_rate": 1.2709777777777779e-05,
      "loss": 1.0945,
      "step": 371000
    },
    {
      "epoch": 17.32,
      "learning_rate": 1.2699676767676769e-05,
      "loss": 1.0927,
      "step": 371500
    },
    {
      "epoch": 17.34,
      "learning_rate": 1.2689575757575758e-05,
      "loss": 1.092,
      "step": 372000
    },
    {
      "epoch": 17.37,
      "learning_rate": 1.267947474747475e-05,
      "loss": 1.0883,
      "step": 372500
    },
    {
      "epoch": 17.39,
      "learning_rate": 1.2669393939393942e-05,
      "loss": 1.0947,
      "step": 373000
    },
    {
      "epoch": 17.41,
      "learning_rate": 1.2659313131313134e-05,
      "loss": 1.0914,
      "step": 373500
    },
    {
      "epoch": 17.44,
      "learning_rate": 1.2649212121212123e-05,
      "loss": 1.0928,
      "step": 374000
    },
    {
      "epoch": 17.46,
      "learning_rate": 1.2639111111111113e-05,
      "loss": 1.091,
      "step": 374500
    },
    {
      "epoch": 17.48,
      "learning_rate": 1.2629010101010102e-05,
      "loss": 1.0935,
      "step": 375000
    },
    {
      "epoch": 17.51,
      "learning_rate": 1.2618909090909094e-05,
      "loss": 1.0916,
      "step": 375500
    },
    {
      "epoch": 17.53,
      "learning_rate": 1.2608808080808082e-05,
      "loss": 1.092,
      "step": 376000
    },
    {
      "epoch": 17.55,
      "learning_rate": 1.2598707070707072e-05,
      "loss": 1.0921,
      "step": 376500
    },
    {
      "epoch": 17.57,
      "learning_rate": 1.2588626262626265e-05,
      "loss": 1.093,
      "step": 377000
    },
    {
      "epoch": 17.6,
      "learning_rate": 1.2578525252525253e-05,
      "loss": 1.0916,
      "step": 377500
    },
    {
      "epoch": 17.62,
      "learning_rate": 1.2568424242424244e-05,
      "loss": 1.0924,
      "step": 378000
    },
    {
      "epoch": 17.64,
      "learning_rate": 1.2558323232323232e-05,
      "loss": 1.0902,
      "step": 378500
    },
    {
      "epoch": 17.67,
      "learning_rate": 1.2548222222222224e-05,
      "loss": 1.0931,
      "step": 379000
    },
    {
      "epoch": 17.69,
      "learning_rate": 1.2538121212121213e-05,
      "loss": 1.0909,
      "step": 379500
    },
    {
      "epoch": 17.71,
      "learning_rate": 1.2528020202020203e-05,
      "loss": 1.0907,
      "step": 380000
    },
    {
      "epoch": 17.74,
      "learning_rate": 1.2517919191919192e-05,
      "loss": 1.0899,
      "step": 380500
    },
    {
      "epoch": 17.76,
      "learning_rate": 1.2507818181818184e-05,
      "loss": 1.09,
      "step": 381000
    },
    {
      "epoch": 17.78,
      "learning_rate": 1.2497717171717172e-05,
      "loss": 1.0913,
      "step": 381500
    },
    {
      "epoch": 17.81,
      "learning_rate": 1.2487616161616162e-05,
      "loss": 1.0934,
      "step": 382000
    },
    {
      "epoch": 17.83,
      "learning_rate": 1.2477515151515151e-05,
      "loss": 1.0907,
      "step": 382500
    },
    {
      "epoch": 17.85,
      "learning_rate": 1.2467434343434343e-05,
      "loss": 1.0935,
      "step": 383000
    },
    {
      "epoch": 17.88,
      "learning_rate": 1.2457333333333335e-05,
      "loss": 1.0907,
      "step": 383500
    },
    {
      "epoch": 17.9,
      "learning_rate": 1.2447232323232324e-05,
      "loss": 1.0891,
      "step": 384000
    },
    {
      "epoch": 17.92,
      "learning_rate": 1.2437131313131314e-05,
      "loss": 1.0907,
      "step": 384500
    },
    {
      "epoch": 17.95,
      "learning_rate": 1.2427030303030303e-05,
      "loss": 1.0912,
      "step": 385000
    },
    {
      "epoch": 17.97,
      "learning_rate": 1.2416949494949495e-05,
      "loss": 1.0921,
      "step": 385500
    },
    {
      "epoch": 17.99,
      "learning_rate": 1.2406848484848485e-05,
      "loss": 1.0923,
      "step": 386000
    },
    {
      "epoch": 18.0,
      "eval_accuracy": 0.7719286372361273,
      "eval_loss": 0.9558597207069397,
      "eval_runtime": 52.0187,
      "eval_samples_per_second": 897.907,
      "eval_steps_per_second": 14.033,
      "step": 386118
    },
    {
      "epoch": 18.02,
      "learning_rate": 1.2396747474747476e-05,
      "loss": 1.0908,
      "step": 386500
    },
    {
      "epoch": 18.04,
      "learning_rate": 1.2386646464646466e-05,
      "loss": 1.0878,
      "step": 387000
    },
    {
      "epoch": 18.06,
      "learning_rate": 1.2376545454545456e-05,
      "loss": 1.0926,
      "step": 387500
    },
    {
      "epoch": 18.09,
      "learning_rate": 1.2366444444444445e-05,
      "loss": 1.0909,
      "step": 388000
    },
    {
      "epoch": 18.11,
      "learning_rate": 1.2356363636363637e-05,
      "loss": 1.0909,
      "step": 388500
    },
    {
      "epoch": 18.13,
      "learning_rate": 1.2346262626262626e-05,
      "loss": 1.0893,
      "step": 389000
    },
    {
      "epoch": 18.16,
      "learning_rate": 1.2336161616161618e-05,
      "loss": 1.0897,
      "step": 389500
    },
    {
      "epoch": 18.18,
      "learning_rate": 1.2326060606060608e-05,
      "loss": 1.0903,
      "step": 390000
    },
    {
      "epoch": 18.2,
      "learning_rate": 1.2315959595959596e-05,
      "loss": 1.0884,
      "step": 390500
    },
    {
      "epoch": 18.23,
      "learning_rate": 1.2305858585858588e-05,
      "loss": 1.0915,
      "step": 391000
    },
    {
      "epoch": 18.25,
      "learning_rate": 1.2295757575757577e-05,
      "loss": 1.0895,
      "step": 391500
    },
    {
      "epoch": 18.27,
      "learning_rate": 1.228569696969697e-05,
      "loss": 1.0893,
      "step": 392000
    },
    {
      "epoch": 18.3,
      "learning_rate": 1.227559595959596e-05,
      "loss": 1.0908,
      "step": 392500
    },
    {
      "epoch": 18.32,
      "learning_rate": 1.2265494949494952e-05,
      "loss": 1.0897,
      "step": 393000
    },
    {
      "epoch": 18.34,
      "learning_rate": 1.225539393939394e-05,
      "loss": 1.0879,
      "step": 393500
    },
    {
      "epoch": 18.37,
      "learning_rate": 1.224529292929293e-05,
      "loss": 1.0903,
      "step": 394000
    },
    {
      "epoch": 18.39,
      "learning_rate": 1.223519191919192e-05,
      "loss": 1.0894,
      "step": 394500
    },
    {
      "epoch": 18.41,
      "learning_rate": 1.2225090909090911e-05,
      "loss": 1.0899,
      "step": 395000
    },
    {
      "epoch": 18.44,
      "learning_rate": 1.22149898989899e-05,
      "loss": 1.0924,
      "step": 395500
    },
    {
      "epoch": 18.46,
      "learning_rate": 1.220488888888889e-05,
      "loss": 1.0904,
      "step": 396000
    },
    {
      "epoch": 18.48,
      "learning_rate": 1.2194808080808082e-05,
      "loss": 1.0895,
      "step": 396500
    },
    {
      "epoch": 18.51,
      "learning_rate": 1.2184707070707071e-05,
      "loss": 1.0905,
      "step": 397000
    },
    {
      "epoch": 18.53,
      "learning_rate": 1.2174606060606063e-05,
      "loss": 1.0878,
      "step": 397500
    },
    {
      "epoch": 18.55,
      "learning_rate": 1.2164505050505052e-05,
      "loss": 1.0927,
      "step": 398000
    },
    {
      "epoch": 18.58,
      "learning_rate": 1.2154404040404042e-05,
      "loss": 1.0892,
      "step": 398500
    },
    {
      "epoch": 18.6,
      "learning_rate": 1.214430303030303e-05,
      "loss": 1.0883,
      "step": 399000
    },
    {
      "epoch": 18.62,
      "learning_rate": 1.2134202020202022e-05,
      "loss": 1.0913,
      "step": 399500
    },
    {
      "epoch": 18.65,
      "learning_rate": 1.2124101010101011e-05,
      "loss": 1.0901,
      "step": 400000
    },
    {
      "epoch": 18.67,
      "learning_rate": 1.2114020202020203e-05,
      "loss": 1.0881,
      "step": 400500
    },
    {
      "epoch": 18.69,
      "learning_rate": 1.2103919191919193e-05,
      "loss": 1.0883,
      "step": 401000
    },
    {
      "epoch": 18.72,
      "learning_rate": 1.2093818181818182e-05,
      "loss": 1.0891,
      "step": 401500
    },
    {
      "epoch": 18.74,
      "learning_rate": 1.2083717171717172e-05,
      "loss": 1.0893,
      "step": 402000
    },
    {
      "epoch": 18.76,
      "learning_rate": 1.2073616161616161e-05,
      "loss": 1.0896,
      "step": 402500
    },
    {
      "epoch": 18.79,
      "learning_rate": 1.2063515151515153e-05,
      "loss": 1.0901,
      "step": 403000
    },
    {
      "epoch": 18.81,
      "learning_rate": 1.2053434343434345e-05,
      "loss": 1.0895,
      "step": 403500
    },
    {
      "epoch": 18.83,
      "learning_rate": 1.2043333333333334e-05,
      "loss": 1.0885,
      "step": 404000
    },
    {
      "epoch": 18.86,
      "learning_rate": 1.2033232323232324e-05,
      "loss": 1.087,
      "step": 404500
    },
    {
      "epoch": 18.88,
      "learning_rate": 1.2023131313131313e-05,
      "loss": 1.0859,
      "step": 405000
    },
    {
      "epoch": 18.9,
      "learning_rate": 1.2013030303030305e-05,
      "loss": 1.0865,
      "step": 405500
    },
    {
      "epoch": 18.93,
      "learning_rate": 1.2002929292929293e-05,
      "loss": 1.0872,
      "step": 406000
    },
    {
      "epoch": 18.95,
      "learning_rate": 1.1992828282828284e-05,
      "loss": 1.0868,
      "step": 406500
    },
    {
      "epoch": 18.97,
      "learning_rate": 1.1982727272727276e-05,
      "loss": 1.0886,
      "step": 407000
    },
    {
      "epoch": 19.0,
      "learning_rate": 1.1972646464646464e-05,
      "loss": 1.0891,
      "step": 407500
    },
    {
      "epoch": 19.0,
      "eval_accuracy": 0.7723270650741523,
      "eval_loss": 0.9527531862258911,
      "eval_runtime": 51.779,
      "eval_samples_per_second": 902.064,
      "eval_steps_per_second": 14.098,
      "step": 407569
    },
    {
      "epoch": 19.02,
      "learning_rate": 1.196258585858586e-05,
      "loss": 1.0879,
      "step": 408000
    },
    {
      "epoch": 19.04,
      "learning_rate": 1.1952484848484849e-05,
      "loss": 1.088,
      "step": 408500
    },
    {
      "epoch": 19.07,
      "learning_rate": 1.1942383838383839e-05,
      "loss": 1.0853,
      "step": 409000
    },
    {
      "epoch": 19.09,
      "learning_rate": 1.1932282828282828e-05,
      "loss": 1.0883,
      "step": 409500
    },
    {
      "epoch": 19.11,
      "learning_rate": 1.192218181818182e-05,
      "loss": 1.0873,
      "step": 410000
    },
    {
      "epoch": 19.14,
      "learning_rate": 1.1912080808080808e-05,
      "loss": 1.0872,
      "step": 410500
    },
    {
      "epoch": 19.16,
      "learning_rate": 1.1901979797979799e-05,
      "loss": 1.0888,
      "step": 411000
    },
    {
      "epoch": 19.18,
      "learning_rate": 1.1891878787878787e-05,
      "loss": 1.0868,
      "step": 411500
    },
    {
      "epoch": 19.21,
      "learning_rate": 1.188179797979798e-05,
      "loss": 1.0874,
      "step": 412000
    },
    {
      "epoch": 19.23,
      "learning_rate": 1.1871696969696971e-05,
      "loss": 1.0873,
      "step": 412500
    },
    {
      "epoch": 19.25,
      "learning_rate": 1.186159595959596e-05,
      "loss": 1.0873,
      "step": 413000
    },
    {
      "epoch": 19.28,
      "learning_rate": 1.185149494949495e-05,
      "loss": 1.0868,
      "step": 413500
    },
    {
      "epoch": 19.3,
      "learning_rate": 1.1841393939393942e-05,
      "loss": 1.0882,
      "step": 414000
    },
    {
      "epoch": 19.32,
      "learning_rate": 1.1831292929292931e-05,
      "loss": 1.0872,
      "step": 414500
    },
    {
      "epoch": 19.35,
      "learning_rate": 1.1821191919191921e-05,
      "loss": 1.087,
      "step": 415000
    },
    {
      "epoch": 19.37,
      "learning_rate": 1.181109090909091e-05,
      "loss": 1.0849,
      "step": 415500
    },
    {
      "epoch": 19.39,
      "learning_rate": 1.18009898989899e-05,
      "loss": 1.0864,
      "step": 416000
    },
    {
      "epoch": 19.42,
      "learning_rate": 1.1790909090909092e-05,
      "loss": 1.0889,
      "step": 416500
    },
    {
      "epoch": 19.44,
      "learning_rate": 1.1780808080808081e-05,
      "loss": 1.0864,
      "step": 417000
    },
    {
      "epoch": 19.46,
      "learning_rate": 1.1770707070707073e-05,
      "loss": 1.0864,
      "step": 417500
    },
    {
      "epoch": 19.49,
      "learning_rate": 1.1760606060606061e-05,
      "loss": 1.085,
      "step": 418000
    },
    {
      "epoch": 19.51,
      "learning_rate": 1.1750525252525254e-05,
      "loss": 1.0867,
      "step": 418500
    },
    {
      "epoch": 19.53,
      "learning_rate": 1.1740424242424244e-05,
      "loss": 1.0845,
      "step": 419000
    },
    {
      "epoch": 19.56,
      "learning_rate": 1.1730323232323233e-05,
      "loss": 1.0876,
      "step": 419500
    },
    {
      "epoch": 19.58,
      "learning_rate": 1.1720222222222225e-05,
      "loss": 1.0864,
      "step": 420000
    },
    {
      "epoch": 19.6,
      "learning_rate": 1.1710141414141417e-05,
      "loss": 1.087,
      "step": 420500
    },
    {
      "epoch": 19.63,
      "learning_rate": 1.1700040404040405e-05,
      "loss": 1.0877,
      "step": 421000
    },
    {
      "epoch": 19.65,
      "learning_rate": 1.1689939393939396e-05,
      "loss": 1.0849,
      "step": 421500
    },
    {
      "epoch": 19.67,
      "learning_rate": 1.1679838383838384e-05,
      "loss": 1.0857,
      "step": 422000
    },
    {
      "epoch": 19.7,
      "learning_rate": 1.1669737373737375e-05,
      "loss": 1.0861,
      "step": 422500
    },
    {
      "epoch": 19.72,
      "learning_rate": 1.1659636363636365e-05,
      "loss": 1.0861,
      "step": 423000
    },
    {
      "epoch": 19.74,
      "learning_rate": 1.1649555555555555e-05,
      "loss": 1.0844,
      "step": 423500
    },
    {
      "epoch": 19.77,
      "learning_rate": 1.1639454545454547e-05,
      "loss": 1.0868,
      "step": 424000
    },
    {
      "epoch": 19.79,
      "learning_rate": 1.1629353535353536e-05,
      "loss": 1.0871,
      "step": 424500
    },
    {
      "epoch": 19.81,
      "learning_rate": 1.1619252525252526e-05,
      "loss": 1.0858,
      "step": 425000
    },
    {
      "epoch": 19.84,
      "learning_rate": 1.1609151515151515e-05,
      "loss": 1.0894,
      "step": 425500
    },
    {
      "epoch": 19.86,
      "learning_rate": 1.1599050505050507e-05,
      "loss": 1.0868,
      "step": 426000
    },
    {
      "epoch": 19.88,
      "learning_rate": 1.1588969696969699e-05,
      "loss": 1.0842,
      "step": 426500
    },
    {
      "epoch": 19.91,
      "learning_rate": 1.1578868686868688e-05,
      "loss": 1.086,
      "step": 427000
    },
    {
      "epoch": 19.93,
      "learning_rate": 1.1568767676767678e-05,
      "loss": 1.086,
      "step": 427500
    },
    {
      "epoch": 19.95,
      "learning_rate": 1.1558666666666667e-05,
      "loss": 1.0852,
      "step": 428000
    },
    {
      "epoch": 19.98,
      "learning_rate": 1.1548565656565659e-05,
      "loss": 1.0833,
      "step": 428500
    },
    {
      "epoch": 20.0,
      "learning_rate": 1.153848484848485e-05,
      "loss": 1.0841,
      "step": 429000
    },
    {
      "epoch": 20.0,
      "eval_accuracy": 0.7726146545415603,
      "eval_loss": 0.9511629939079285,
      "eval_runtime": 51.01,
      "eval_samples_per_second": 915.663,
      "eval_steps_per_second": 14.311,
      "step": 429020
    },
    {
      "epoch": 20.02,
      "learning_rate": 1.152838383838384e-05,
      "loss": 1.0812,
      "step": 429500
    },
    {
      "epoch": 20.05,
      "learning_rate": 1.151828282828283e-05,
      "loss": 1.0859,
      "step": 430000
    },
    {
      "epoch": 20.07,
      "learning_rate": 1.1508181818181818e-05,
      "loss": 1.0852,
      "step": 430500
    },
    {
      "epoch": 20.09,
      "learning_rate": 1.1498080808080809e-05,
      "loss": 1.0842,
      "step": 431000
    },
    {
      "epoch": 20.12,
      "learning_rate": 1.1488e-05,
      "loss": 1.0857,
      "step": 431500
    },
    {
      "epoch": 20.14,
      "learning_rate": 1.147789898989899e-05,
      "loss": 1.0843,
      "step": 432000
    },
    {
      "epoch": 20.16,
      "learning_rate": 1.1467797979797981e-05,
      "loss": 1.0847,
      "step": 432500
    },
    {
      "epoch": 20.19,
      "learning_rate": 1.145769696969697e-05,
      "loss": 1.0855,
      "step": 433000
    },
    {
      "epoch": 20.21,
      "learning_rate": 1.144759595959596e-05,
      "loss": 1.084,
      "step": 433500
    },
    {
      "epoch": 20.23,
      "learning_rate": 1.1437494949494949e-05,
      "loss": 1.0861,
      "step": 434000
    },
    {
      "epoch": 20.26,
      "learning_rate": 1.1427414141414141e-05,
      "loss": 1.0836,
      "step": 434500
    },
    {
      "epoch": 20.28,
      "learning_rate": 1.1417313131313133e-05,
      "loss": 1.0858,
      "step": 435000
    },
    {
      "epoch": 20.3,
      "learning_rate": 1.1407232323232325e-05,
      "loss": 1.0867,
      "step": 435500
    },
    {
      "epoch": 20.33,
      "learning_rate": 1.1397131313131314e-05,
      "loss": 1.0863,
      "step": 436000
    },
    {
      "epoch": 20.35,
      "learning_rate": 1.1387030303030304e-05,
      "loss": 1.0827,
      "step": 436500
    },
    {
      "epoch": 20.37,
      "learning_rate": 1.1376929292929293e-05,
      "loss": 1.084,
      "step": 437000
    },
    {
      "epoch": 20.4,
      "learning_rate": 1.1366828282828283e-05,
      "loss": 1.0858,
      "step": 437500
    },
    {
      "epoch": 20.42,
      "learning_rate": 1.1356727272727273e-05,
      "loss": 1.0845,
      "step": 438000
    },
    {
      "epoch": 20.44,
      "learning_rate": 1.1346626262626264e-05,
      "loss": 1.0857,
      "step": 438500
    },
    {
      "epoch": 20.47,
      "learning_rate": 1.1336545454545456e-05,
      "loss": 1.0823,
      "step": 439000
    },
    {
      "epoch": 20.49,
      "learning_rate": 1.1326444444444444e-05,
      "loss": 1.0818,
      "step": 439500
    },
    {
      "epoch": 20.51,
      "learning_rate": 1.1316343434343435e-05,
      "loss": 1.0834,
      "step": 440000
    },
    {
      "epoch": 20.54,
      "learning_rate": 1.1306242424242427e-05,
      "loss": 1.0846,
      "step": 440500
    },
    {
      "epoch": 20.56,
      "learning_rate": 1.1296141414141415e-05,
      "loss": 1.0847,
      "step": 441000
    },
    {
      "epoch": 20.58,
      "learning_rate": 1.1286040404040406e-05,
      "loss": 1.0852,
      "step": 441500
    },
    {
      "epoch": 20.61,
      "learning_rate": 1.1275959595959596e-05,
      "loss": 1.0869,
      "step": 442000
    },
    {
      "epoch": 20.63,
      "learning_rate": 1.1265858585858586e-05,
      "loss": 1.0834,
      "step": 442500
    },
    {
      "epoch": 20.65,
      "learning_rate": 1.1255757575757578e-05,
      "loss": 1.0829,
      "step": 443000
    },
    {
      "epoch": 20.68,
      "learning_rate": 1.1245656565656567e-05,
      "loss": 1.0841,
      "step": 443500
    },
    {
      "epoch": 20.7,
      "learning_rate": 1.123557575757576e-05,
      "loss": 1.0836,
      "step": 444000
    },
    {
      "epoch": 20.72,
      "learning_rate": 1.122547474747475e-05,
      "loss": 1.084,
      "step": 444500
    },
    {
      "epoch": 20.74,
      "learning_rate": 1.1215373737373738e-05,
      "loss": 1.0819,
      "step": 445000
    },
    {
      "epoch": 20.77,
      "learning_rate": 1.1205272727272728e-05,
      "loss": 1.0854,
      "step": 445500
    },
    {
      "epoch": 20.79,
      "learning_rate": 1.1195171717171717e-05,
      "loss": 1.0827,
      "step": 446000
    },
    {
      "epoch": 20.81,
      "learning_rate": 1.1185070707070709e-05,
      "loss": 1.0824,
      "step": 446500
    },
    {
      "epoch": 20.84,
      "learning_rate": 1.1174969696969698e-05,
      "loss": 1.0858,
      "step": 447000
    },
    {
      "epoch": 20.86,
      "learning_rate": 1.1164868686868688e-05,
      "loss": 1.0854,
      "step": 447500
    },
    {
      "epoch": 20.88,
      "learning_rate": 1.1154767676767677e-05,
      "loss": 1.0853,
      "step": 448000
    },
    {
      "epoch": 20.91,
      "learning_rate": 1.1144666666666668e-05,
      "loss": 1.0811,
      "step": 448500
    },
    {
      "epoch": 20.93,
      "learning_rate": 1.1134565656565657e-05,
      "loss": 1.0848,
      "step": 449000
    },
    {
      "epoch": 20.95,
      "learning_rate": 1.1124464646464647e-05,
      "loss": 1.084,
      "step": 449500
    },
    {
      "epoch": 20.98,
      "learning_rate": 1.111438383838384e-05,
      "loss": 1.0824,
      "step": 450000
    },
    {
      "epoch": 21.0,
      "eval_accuracy": 0.7732778119270386,
      "eval_loss": 0.9474597573280334,
      "eval_runtime": 51.6342,
      "eval_samples_per_second": 904.593,
      "eval_steps_per_second": 14.138,
      "step": 450471
    },
    {
      "epoch": 21.0,
      "learning_rate": 1.1104303030303032e-05,
      "loss": 1.0809,
      "step": 450500
    },
    {
      "epoch": 21.02,
      "learning_rate": 1.1094222222222224e-05,
      "loss": 1.0829,
      "step": 451000
    },
    {
      "epoch": 21.05,
      "learning_rate": 1.1084121212121213e-05,
      "loss": 1.0832,
      "step": 451500
    },
    {
      "epoch": 21.07,
      "learning_rate": 1.1074020202020203e-05,
      "loss": 1.0816,
      "step": 452000
    },
    {
      "epoch": 21.09,
      "learning_rate": 1.1063919191919192e-05,
      "loss": 1.0838,
      "step": 452500
    },
    {
      "epoch": 21.12,
      "learning_rate": 1.1053818181818184e-05,
      "loss": 1.0833,
      "step": 453000
    },
    {
      "epoch": 21.14,
      "learning_rate": 1.1043717171717172e-05,
      "loss": 1.0834,
      "step": 453500
    },
    {
      "epoch": 21.16,
      "learning_rate": 1.1033616161616162e-05,
      "loss": 1.0815,
      "step": 454000
    },
    {
      "epoch": 21.19,
      "learning_rate": 1.1023515151515151e-05,
      "loss": 1.0816,
      "step": 454500
    },
    {
      "epoch": 21.21,
      "learning_rate": 1.1013414141414143e-05,
      "loss": 1.0825,
      "step": 455000
    },
    {
      "epoch": 21.23,
      "learning_rate": 1.1003313131313132e-05,
      "loss": 1.0819,
      "step": 455500
    },
    {
      "epoch": 21.26,
      "learning_rate": 1.0993212121212122e-05,
      "loss": 1.0817,
      "step": 456000
    },
    {
      "epoch": 21.28,
      "learning_rate": 1.098311111111111e-05,
      "loss": 1.0819,
      "step": 456500
    },
    {
      "epoch": 21.3,
      "learning_rate": 1.0973010101010102e-05,
      "loss": 1.0837,
      "step": 457000
    },
    {
      "epoch": 21.33,
      "learning_rate": 1.0962929292929295e-05,
      "loss": 1.0828,
      "step": 457500
    },
    {
      "epoch": 21.35,
      "learning_rate": 1.0952828282828283e-05,
      "loss": 1.0786,
      "step": 458000
    },
    {
      "epoch": 21.37,
      "learning_rate": 1.0942727272727274e-05,
      "loss": 1.083,
      "step": 458500
    },
    {
      "epoch": 21.4,
      "learning_rate": 1.0932646464646466e-05,
      "loss": 1.0818,
      "step": 459000
    },
    {
      "epoch": 21.42,
      "learning_rate": 1.0922545454545454e-05,
      "loss": 1.0806,
      "step": 459500
    },
    {
      "epoch": 21.44,
      "learning_rate": 1.0912444444444445e-05,
      "loss": 1.0812,
      "step": 460000
    },
    {
      "epoch": 21.47,
      "learning_rate": 1.0902343434343435e-05,
      "loss": 1.0806,
      "step": 460500
    },
    {
      "epoch": 21.49,
      "learning_rate": 1.0892242424242425e-05,
      "loss": 1.0815,
      "step": 461000
    },
    {
      "epoch": 21.51,
      "learning_rate": 1.0882161616161618e-05,
      "loss": 1.0813,
      "step": 461500
    },
    {
      "epoch": 21.54,
      "learning_rate": 1.0872060606060606e-05,
      "loss": 1.081,
      "step": 462000
    },
    {
      "epoch": 21.56,
      "learning_rate": 1.0861959595959596e-05,
      "loss": 1.0816,
      "step": 462500
    },
    {
      "epoch": 21.58,
      "learning_rate": 1.0851858585858585e-05,
      "loss": 1.083,
      "step": 463000
    },
    {
      "epoch": 21.61,
      "learning_rate": 1.0841757575757577e-05,
      "loss": 1.0831,
      "step": 463500
    },
    {
      "epoch": 21.63,
      "learning_rate": 1.0831656565656567e-05,
      "loss": 1.0815,
      "step": 464000
    },
    {
      "epoch": 21.65,
      "learning_rate": 1.0821555555555556e-05,
      "loss": 1.0795,
      "step": 464500
    },
    {
      "epoch": 21.68,
      "learning_rate": 1.0811474747474748e-05,
      "loss": 1.0809,
      "step": 465000
    },
    {
      "epoch": 21.7,
      "learning_rate": 1.080137373737374e-05,
      "loss": 1.0838,
      "step": 465500
    },
    {
      "epoch": 21.72,
      "learning_rate": 1.0791292929292929e-05,
      "loss": 1.0823,
      "step": 466000
    },
    {
      "epoch": 21.75,
      "learning_rate": 1.078119191919192e-05,
      "loss": 1.0817,
      "step": 466500
    },
    {
      "epoch": 21.77,
      "learning_rate": 1.0771090909090911e-05,
      "loss": 1.0823,
      "step": 467000
    },
    {
      "epoch": 21.79,
      "learning_rate": 1.07609898989899e-05,
      "loss": 1.0806,
      "step": 467500
    },
    {
      "epoch": 21.82,
      "learning_rate": 1.075088888888889e-05,
      "loss": 1.0803,
      "step": 468000
    },
    {
      "epoch": 21.84,
      "learning_rate": 1.0740787878787879e-05,
      "loss": 1.0808,
      "step": 468500
    },
    {
      "epoch": 21.86,
      "learning_rate": 1.073068686868687e-05,
      "loss": 1.0804,
      "step": 469000
    },
    {
      "epoch": 21.89,
      "learning_rate": 1.072058585858586e-05,
      "loss": 1.0818,
      "step": 469500
    },
    {
      "epoch": 21.91,
      "learning_rate": 1.071048484848485e-05,
      "loss": 1.0807,
      "step": 470000
    },
    {
      "epoch": 21.93,
      "learning_rate": 1.0700383838383838e-05,
      "loss": 1.0823,
      "step": 470500
    },
    {
      "epoch": 21.96,
      "learning_rate": 1.069028282828283e-05,
      "loss": 1.0826,
      "step": 471000
    },
    {
      "epoch": 21.98,
      "learning_rate": 1.0680181818181819e-05,
      "loss": 1.0788,
      "step": 471500
    },
    {
      "epoch": 22.0,
      "eval_accuracy": 0.7738636750659805,
      "eval_loss": 0.944471001625061,
      "eval_runtime": 51.9655,
      "eval_samples_per_second": 898.826,
      "eval_steps_per_second": 14.048,
      "step": 471922
    },
    {
      "epoch": 22.0,
      "learning_rate": 1.0670080808080809e-05,
      "loss": 1.0808,
      "step": 472000
    },
    {
      "epoch": 22.03,
      "learning_rate": 1.0660000000000001e-05,
      "loss": 1.0783,
      "step": 472500
    },
    {
      "epoch": 22.05,
      "learning_rate": 1.0649919191919194e-05,
      "loss": 1.0803,
      "step": 473000
    },
    {
      "epoch": 22.07,
      "learning_rate": 1.0639818181818182e-05,
      "loss": 1.081,
      "step": 473500
    },
    {
      "epoch": 22.1,
      "learning_rate": 1.0629717171717172e-05,
      "loss": 1.0814,
      "step": 474000
    },
    {
      "epoch": 22.12,
      "learning_rate": 1.0619616161616163e-05,
      "loss": 1.0816,
      "step": 474500
    },
    {
      "epoch": 22.14,
      "learning_rate": 1.0609535353535353e-05,
      "loss": 1.0763,
      "step": 475000
    },
    {
      "epoch": 22.17,
      "learning_rate": 1.0599434343434345e-05,
      "loss": 1.0782,
      "step": 475500
    },
    {
      "epoch": 22.19,
      "learning_rate": 1.0589333333333334e-05,
      "loss": 1.0822,
      "step": 476000
    },
    {
      "epoch": 22.21,
      "learning_rate": 1.0579252525252526e-05,
      "loss": 1.0778,
      "step": 476500
    },
    {
      "epoch": 22.24,
      "learning_rate": 1.0569151515151516e-05,
      "loss": 1.0764,
      "step": 477000
    },
    {
      "epoch": 22.26,
      "learning_rate": 1.0559050505050505e-05,
      "loss": 1.0814,
      "step": 477500
    },
    {
      "epoch": 22.28,
      "learning_rate": 1.0548949494949497e-05,
      "loss": 1.08,
      "step": 478000
    },
    {
      "epoch": 22.31,
      "learning_rate": 1.0538848484848486e-05,
      "loss": 1.0786,
      "step": 478500
    },
    {
      "epoch": 22.33,
      "learning_rate": 1.0528747474747476e-05,
      "loss": 1.0801,
      "step": 479000
    },
    {
      "epoch": 22.35,
      "learning_rate": 1.0518646464646464e-05,
      "loss": 1.0776,
      "step": 479500
    },
    {
      "epoch": 22.38,
      "learning_rate": 1.0508545454545456e-05,
      "loss": 1.0805,
      "step": 480000
    },
    {
      "epoch": 22.4,
      "learning_rate": 1.0498444444444445e-05,
      "loss": 1.0806,
      "step": 480500
    },
    {
      "epoch": 22.42,
      "learning_rate": 1.0488343434343435e-05,
      "loss": 1.079,
      "step": 481000
    },
    {
      "epoch": 22.45,
      "learning_rate": 1.0478242424242424e-05,
      "loss": 1.0785,
      "step": 481500
    },
    {
      "epoch": 22.47,
      "learning_rate": 1.0468141414141416e-05,
      "loss": 1.0791,
      "step": 482000
    },
    {
      "epoch": 22.49,
      "learning_rate": 1.0458040404040406e-05,
      "loss": 1.0786,
      "step": 482500
    },
    {
      "epoch": 22.52,
      "learning_rate": 1.0447939393939395e-05,
      "loss": 1.0789,
      "step": 483000
    },
    {
      "epoch": 22.54,
      "learning_rate": 1.0437878787878787e-05,
      "loss": 1.0825,
      "step": 483500
    },
    {
      "epoch": 22.56,
      "learning_rate": 1.042777777777778e-05,
      "loss": 1.0768,
      "step": 484000
    },
    {
      "epoch": 22.59,
      "learning_rate": 1.0417676767676768e-05,
      "loss": 1.0797,
      "step": 484500
    },
    {
      "epoch": 22.61,
      "learning_rate": 1.0407575757575758e-05,
      "loss": 1.0777,
      "step": 485000
    },
    {
      "epoch": 22.63,
      "learning_rate": 1.0397474747474747e-05,
      "loss": 1.0784,
      "step": 485500
    },
    {
      "epoch": 22.66,
      "learning_rate": 1.0387373737373739e-05,
      "loss": 1.0782,
      "step": 486000
    },
    {
      "epoch": 22.68,
      "learning_rate": 1.0377272727272729e-05,
      "loss": 1.0819,
      "step": 486500
    },
    {
      "epoch": 22.7,
      "learning_rate": 1.0367171717171718e-05,
      "loss": 1.0774,
      "step": 487000
    },
    {
      "epoch": 22.73,
      "learning_rate": 1.035709090909091e-05,
      "loss": 1.0811,
      "step": 487500
    },
    {
      "epoch": 22.75,
      "learning_rate": 1.03469898989899e-05,
      "loss": 1.0819,
      "step": 488000
    },
    {
      "epoch": 22.77,
      "learning_rate": 1.033688888888889e-05,
      "loss": 1.0782,
      "step": 488500
    },
    {
      "epoch": 22.8,
      "learning_rate": 1.032678787878788e-05,
      "loss": 1.0781,
      "step": 489000
    },
    {
      "epoch": 22.82,
      "learning_rate": 1.031668686868687e-05,
      "loss": 1.0796,
      "step": 489500
    },
    {
      "epoch": 22.84,
      "learning_rate": 1.0306626262626262e-05,
      "loss": 1.0802,
      "step": 490000
    },
    {
      "epoch": 22.87,
      "learning_rate": 1.0296525252525254e-05,
      "loss": 1.0791,
      "step": 490500
    },
    {
      "epoch": 22.89,
      "learning_rate": 1.0286424242424242e-05,
      "loss": 1.0793,
      "step": 491000
    },
    {
      "epoch": 22.91,
      "learning_rate": 1.0276323232323233e-05,
      "loss": 1.0821,
      "step": 491500
    },
    {
      "epoch": 22.94,
      "learning_rate": 1.0266222222222225e-05,
      "loss": 1.0789,
      "step": 492000
    },
    {
      "epoch": 22.96,
      "learning_rate": 1.0256121212121213e-05,
      "loss": 1.0791,
      "step": 492500
    },
    {
      "epoch": 22.98,
      "learning_rate": 1.0246020202020203e-05,
      "loss": 1.0795,
      "step": 493000
    },
    {
      "epoch": 23.0,
      "eval_accuracy": 0.7739334417842054,
      "eval_loss": 0.9446934461593628,
      "eval_runtime": 52.0679,
      "eval_samples_per_second": 897.059,
      "eval_steps_per_second": 14.02,
      "step": 493373
    },
    {
      "epoch": 23.01,
      "learning_rate": 1.0235939393939396e-05,
      "loss": 1.0805,
      "step": 493500
    },
    {
      "epoch": 23.03,
      "learning_rate": 1.0225838383838384e-05,
      "loss": 1.0775,
      "step": 494000
    },
    {
      "epoch": 23.05,
      "learning_rate": 1.0215737373737376e-05,
      "loss": 1.0787,
      "step": 494500
    },
    {
      "epoch": 23.08,
      "learning_rate": 1.0205636363636365e-05,
      "loss": 1.0776,
      "step": 495000
    },
    {
      "epoch": 23.1,
      "learning_rate": 1.0195535353535355e-05,
      "loss": 1.0775,
      "step": 495500
    },
    {
      "epoch": 23.12,
      "learning_rate": 1.0185434343434344e-05,
      "loss": 1.0777,
      "step": 496000
    },
    {
      "epoch": 23.15,
      "learning_rate": 1.0175353535353536e-05,
      "loss": 1.0779,
      "step": 496500
    },
    {
      "epoch": 23.17,
      "learning_rate": 1.0165252525252526e-05,
      "loss": 1.0787,
      "step": 497000
    },
    {
      "epoch": 23.19,
      "learning_rate": 1.0155151515151515e-05,
      "loss": 1.0765,
      "step": 497500
    },
    {
      "epoch": 23.22,
      "learning_rate": 1.0145050505050507e-05,
      "loss": 1.0748,
      "step": 498000
    },
    {
      "epoch": 23.24,
      "learning_rate": 1.0134949494949495e-05,
      "loss": 1.0788,
      "step": 498500
    },
    {
      "epoch": 23.26,
      "learning_rate": 1.0124868686868688e-05,
      "loss": 1.078,
      "step": 499000
    },
    {
      "epoch": 23.29,
      "learning_rate": 1.0114767676767678e-05,
      "loss": 1.0789,
      "step": 499500
    },
    {
      "epoch": 23.31,
      "learning_rate": 1.0104666666666667e-05,
      "loss": 1.0803,
      "step": 500000
    },
    {
      "epoch": 23.33,
      "learning_rate": 1.0094565656565659e-05,
      "loss": 1.0782,
      "step": 500500
    },
    {
      "epoch": 23.36,
      "learning_rate": 1.0084464646464647e-05,
      "loss": 1.0768,
      "step": 501000
    },
    {
      "epoch": 23.38,
      "learning_rate": 1.0074363636363637e-05,
      "loss": 1.0764,
      "step": 501500
    },
    {
      "epoch": 23.4,
      "learning_rate": 1.0064262626262626e-05,
      "loss": 1.0782,
      "step": 502000
    },
    {
      "epoch": 23.43,
      "learning_rate": 1.0054202020202022e-05,
      "loss": 1.0759,
      "step": 502500
    },
    {
      "epoch": 23.45,
      "learning_rate": 1.004410101010101e-05,
      "loss": 1.0782,
      "step": 503000
    },
    {
      "epoch": 23.47,
      "learning_rate": 1.0034e-05,
      "loss": 1.0797,
      "step": 503500
    },
    {
      "epoch": 23.5,
      "learning_rate": 1.002389898989899e-05,
      "loss": 1.0763,
      "step": 504000
    },
    {
      "epoch": 23.52,
      "learning_rate": 1.0013797979797981e-05,
      "loss": 1.0785,
      "step": 504500
    },
    {
      "epoch": 23.54,
      "learning_rate": 1.000369696969697e-05,
      "loss": 1.0785,
      "step": 505000
    },
    {
      "epoch": 23.57,
      "learning_rate": 9.993616161616162e-06,
      "loss": 1.0783,
      "step": 505500
    },
    {
      "epoch": 23.59,
      "learning_rate": 9.983515151515153e-06,
      "loss": 1.0745,
      "step": 506000
    },
    {
      "epoch": 23.61,
      "learning_rate": 9.973414141414143e-06,
      "loss": 1.0753,
      "step": 506500
    },
    {
      "epoch": 23.64,
      "learning_rate": 9.963313131313133e-06,
      "loss": 1.0739,
      "step": 507000
    },
    {
      "epoch": 23.66,
      "learning_rate": 9.953212121212122e-06,
      "loss": 1.0781,
      "step": 507500
    },
    {
      "epoch": 23.68,
      "learning_rate": 9.943111111111112e-06,
      "loss": 1.0776,
      "step": 508000
    },
    {
      "epoch": 23.71,
      "learning_rate": 9.933010101010102e-06,
      "loss": 1.0767,
      "step": 508500
    },
    {
      "epoch": 23.73,
      "learning_rate": 9.922909090909093e-06,
      "loss": 1.0754,
      "step": 509000
    },
    {
      "epoch": 23.75,
      "learning_rate": 9.912828282828285e-06,
      "loss": 1.0785,
      "step": 509500
    },
    {
      "epoch": 23.78,
      "learning_rate": 9.902727272727273e-06,
      "loss": 1.0745,
      "step": 510000
    },
    {
      "epoch": 23.8,
      "learning_rate": 9.892626262626264e-06,
      "loss": 1.0779,
      "step": 510500
    },
    {
      "epoch": 23.82,
      "learning_rate": 9.882525252525254e-06,
      "loss": 1.0754,
      "step": 511000
    },
    {
      "epoch": 23.85,
      "learning_rate": 9.872424242424243e-06,
      "loss": 1.0792,
      "step": 511500
    },
    {
      "epoch": 23.87,
      "learning_rate": 9.862323232323233e-06,
      "loss": 1.0761,
      "step": 512000
    },
    {
      "epoch": 23.89,
      "learning_rate": 9.852242424242425e-06,
      "loss": 1.0762,
      "step": 512500
    },
    {
      "epoch": 23.91,
      "learning_rate": 9.842141414141415e-06,
      "loss": 1.0769,
      "step": 513000
    },
    {
      "epoch": 23.94,
      "learning_rate": 9.832040404040406e-06,
      "loss": 1.0776,
      "step": 513500
    },
    {
      "epoch": 23.96,
      "learning_rate": 9.821939393939394e-06,
      "loss": 1.076,
      "step": 514000
    },
    {
      "epoch": 23.98,
      "learning_rate": 9.811838383838385e-06,
      "loss": 1.077,
      "step": 514500
    },
    {
      "epoch": 24.0,
      "eval_accuracy": 0.7741739686131365,
      "eval_loss": 0.9426098465919495,
      "eval_runtime": 51.6543,
      "eval_samples_per_second": 904.242,
      "eval_steps_per_second": 14.132,
      "step": 514824
    },
    {
      "epoch": 24.01,
      "learning_rate": 9.801757575757577e-06,
      "loss": 1.0759,
      "step": 515000
    },
    {
      "epoch": 24.03,
      "learning_rate": 9.791676767676769e-06,
      "loss": 1.0751,
      "step": 515500
    },
    {
      "epoch": 24.05,
      "learning_rate": 9.78157575757576e-06,
      "loss": 1.0786,
      "step": 516000
    },
    {
      "epoch": 24.08,
      "learning_rate": 9.771474747474748e-06,
      "loss": 1.0742,
      "step": 516500
    },
    {
      "epoch": 24.1,
      "learning_rate": 9.761373737373738e-06,
      "loss": 1.0758,
      "step": 517000
    },
    {
      "epoch": 24.12,
      "learning_rate": 9.751272727272729e-06,
      "loss": 1.0779,
      "step": 517500
    },
    {
      "epoch": 24.15,
      "learning_rate": 9.741171717171719e-06,
      "loss": 1.0775,
      "step": 518000
    },
    {
      "epoch": 24.17,
      "learning_rate": 9.731070707070707e-06,
      "loss": 1.076,
      "step": 518500
    },
    {
      "epoch": 24.19,
      "learning_rate": 9.720969696969698e-06,
      "loss": 1.0765,
      "step": 519000
    },
    {
      "epoch": 24.22,
      "learning_rate": 9.710868686868688e-06,
      "loss": 1.0722,
      "step": 519500
    },
    {
      "epoch": 24.24,
      "learning_rate": 9.700767676767677e-06,
      "loss": 1.077,
      "step": 520000
    },
    {
      "epoch": 24.26,
      "learning_rate": 9.690666666666667e-06,
      "loss": 1.0774,
      "step": 520500
    },
    {
      "epoch": 24.29,
      "learning_rate": 9.680565656565657e-06,
      "loss": 1.0749,
      "step": 521000
    },
    {
      "epoch": 24.31,
      "learning_rate": 9.67048484848485e-06,
      "loss": 1.0729,
      "step": 521500
    },
    {
      "epoch": 24.33,
      "learning_rate": 9.66038383838384e-06,
      "loss": 1.0778,
      "step": 522000
    },
    {
      "epoch": 24.36,
      "learning_rate": 9.65030303030303e-06,
      "loss": 1.0749,
      "step": 522500
    },
    {
      "epoch": 24.38,
      "learning_rate": 9.64020202020202e-06,
      "loss": 1.0782,
      "step": 523000
    },
    {
      "epoch": 24.4,
      "learning_rate": 9.63010101010101e-06,
      "loss": 1.0761,
      "step": 523500
    },
    {
      "epoch": 24.43,
      "learning_rate": 9.620000000000001e-06,
      "loss": 1.0745,
      "step": 524000
    },
    {
      "epoch": 24.45,
      "learning_rate": 9.60989898989899e-06,
      "loss": 1.0722,
      "step": 524500
    },
    {
      "epoch": 24.47,
      "learning_rate": 9.59979797979798e-06,
      "loss": 1.0739,
      "step": 525000
    },
    {
      "epoch": 24.5,
      "learning_rate": 9.58969696969697e-06,
      "loss": 1.0768,
      "step": 525500
    },
    {
      "epoch": 24.52,
      "learning_rate": 9.57959595959596e-06,
      "loss": 1.0743,
      "step": 526000
    },
    {
      "epoch": 24.54,
      "learning_rate": 9.56949494949495e-06,
      "loss": 1.0739,
      "step": 526500
    },
    {
      "epoch": 24.57,
      "learning_rate": 9.559393939393941e-06,
      "loss": 1.0734,
      "step": 527000
    },
    {
      "epoch": 24.59,
      "learning_rate": 9.549313131313133e-06,
      "loss": 1.0764,
      "step": 527500
    },
    {
      "epoch": 24.61,
      "learning_rate": 9.539212121212122e-06,
      "loss": 1.0758,
      "step": 528000
    },
    {
      "epoch": 24.64,
      "learning_rate": 9.529111111111112e-06,
      "loss": 1.0737,
      "step": 528500
    },
    {
      "epoch": 24.66,
      "learning_rate": 9.519010101010103e-06,
      "loss": 1.076,
      "step": 529000
    },
    {
      "epoch": 24.68,
      "learning_rate": 9.508909090909093e-06,
      "loss": 1.0761,
      "step": 529500
    },
    {
      "epoch": 24.71,
      "learning_rate": 9.498828282828283e-06,
      "loss": 1.075,
      "step": 530000
    },
    {
      "epoch": 24.73,
      "learning_rate": 9.488727272727274e-06,
      "loss": 1.0748,
      "step": 530500
    },
    {
      "epoch": 24.75,
      "learning_rate": 9.478626262626264e-06,
      "loss": 1.0733,
      "step": 531000
    },
    {
      "epoch": 24.78,
      "learning_rate": 9.468545454545456e-06,
      "loss": 1.0753,
      "step": 531500
    },
    {
      "epoch": 24.8,
      "learning_rate": 9.458444444444446e-06,
      "loss": 1.0737,
      "step": 532000
    },
    {
      "epoch": 24.82,
      "learning_rate": 9.448343434343435e-06,
      "loss": 1.0771,
      "step": 532500
    },
    {
      "epoch": 24.85,
      "learning_rate": 9.438242424242425e-06,
      "loss": 1.0781,
      "step": 533000
    },
    {
      "epoch": 24.87,
      "learning_rate": 9.428141414141416e-06,
      "loss": 1.0739,
      "step": 533500
    },
    {
      "epoch": 24.89,
      "learning_rate": 9.418040404040404e-06,
      "loss": 1.0732,
      "step": 534000
    },
    {
      "epoch": 24.92,
      "learning_rate": 9.407939393939395e-06,
      "loss": 1.0733,
      "step": 534500
    },
    {
      "epoch": 24.94,
      "learning_rate": 9.397838383838385e-06,
      "loss": 1.0758,
      "step": 535000
    },
    {
      "epoch": 24.96,
      "learning_rate": 9.387737373737375e-06,
      "loss": 1.0757,
      "step": 535500
    },
    {
      "epoch": 24.99,
      "learning_rate": 9.377636363636364e-06,
      "loss": 1.0751,
      "step": 536000
    },
    {
      "epoch": 25.0,
      "eval_accuracy": 0.7740775950467353,
      "eval_loss": 0.9411866068840027,
      "eval_runtime": 50.6282,
      "eval_samples_per_second": 922.569,
      "eval_steps_per_second": 14.419,
      "step": 536275
    },
    {
      "epoch": 25.01,
      "learning_rate": 9.367535353535354e-06,
      "loss": 1.0742,
      "step": 536500
    },
    {
      "epoch": 25.03,
      "learning_rate": 9.357454545454546e-06,
      "loss": 1.0757,
      "step": 537000
    },
    {
      "epoch": 25.06,
      "learning_rate": 9.347353535353536e-06,
      "loss": 1.0751,
      "step": 537500
    },
    {
      "epoch": 25.08,
      "learning_rate": 9.337252525252525e-06,
      "loss": 1.0737,
      "step": 538000
    },
    {
      "epoch": 25.1,
      "learning_rate": 9.327151515151515e-06,
      "loss": 1.0737,
      "step": 538500
    },
    {
      "epoch": 25.13,
      "learning_rate": 9.317050505050506e-06,
      "loss": 1.0746,
      "step": 539000
    },
    {
      "epoch": 25.15,
      "learning_rate": 9.306949494949496e-06,
      "loss": 1.073,
      "step": 539500
    },
    {
      "epoch": 25.17,
      "learning_rate": 9.296868686868688e-06,
      "loss": 1.0743,
      "step": 540000
    },
    {
      "epoch": 25.2,
      "learning_rate": 9.286767676767677e-06,
      "loss": 1.0746,
      "step": 540500
    },
    {
      "epoch": 25.22,
      "learning_rate": 9.276666666666667e-06,
      "loss": 1.0726,
      "step": 541000
    },
    {
      "epoch": 25.24,
      "learning_rate": 9.266565656565657e-06,
      "loss": 1.0757,
      "step": 541500
    },
    {
      "epoch": 25.27,
      "learning_rate": 9.256464646464648e-06,
      "loss": 1.0744,
      "step": 542000
    },
    {
      "epoch": 25.29,
      "learning_rate": 9.246363636363636e-06,
      "loss": 1.0729,
      "step": 542500
    },
    {
      "epoch": 25.31,
      "learning_rate": 9.236282828282828e-06,
      "loss": 1.0749,
      "step": 543000
    },
    {
      "epoch": 25.34,
      "learning_rate": 9.226181818181819e-06,
      "loss": 1.0721,
      "step": 543500
    },
    {
      "epoch": 25.36,
      "learning_rate": 9.216080808080809e-06,
      "loss": 1.0729,
      "step": 544000
    },
    {
      "epoch": 25.38,
      "learning_rate": 9.205979797979798e-06,
      "loss": 1.0724,
      "step": 544500
    },
    {
      "epoch": 25.41,
      "learning_rate": 9.19587878787879e-06,
      "loss": 1.0733,
      "step": 545000
    },
    {
      "epoch": 25.43,
      "learning_rate": 9.18579797979798e-06,
      "loss": 1.0702,
      "step": 545500
    },
    {
      "epoch": 25.45,
      "learning_rate": 9.17569696969697e-06,
      "loss": 1.0755,
      "step": 546000
    },
    {
      "epoch": 25.48,
      "learning_rate": 9.165595959595959e-06,
      "loss": 1.0738,
      "step": 546500
    },
    {
      "epoch": 25.5,
      "learning_rate": 9.155494949494951e-06,
      "loss": 1.0743,
      "step": 547000
    },
    {
      "epoch": 25.52,
      "learning_rate": 9.145393939393941e-06,
      "loss": 1.074,
      "step": 547500
    },
    {
      "epoch": 25.55,
      "learning_rate": 9.13529292929293e-06,
      "loss": 1.0731,
      "step": 548000
    },
    {
      "epoch": 25.57,
      "learning_rate": 9.12519191919192e-06,
      "loss": 1.0747,
      "step": 548500
    },
    {
      "epoch": 25.59,
      "learning_rate": 9.115111111111112e-06,
      "loss": 1.0726,
      "step": 549000
    },
    {
      "epoch": 25.62,
      "learning_rate": 9.105010101010103e-06,
      "loss": 1.0704,
      "step": 549500
    },
    {
      "epoch": 25.64,
      "learning_rate": 9.094909090909091e-06,
      "loss": 1.0732,
      "step": 550000
    },
    {
      "epoch": 25.66,
      "learning_rate": 9.084808080808082e-06,
      "loss": 1.0745,
      "step": 550500
    },
    {
      "epoch": 25.69,
      "learning_rate": 9.074707070707072e-06,
      "loss": 1.0731,
      "step": 551000
    },
    {
      "epoch": 25.71,
      "learning_rate": 9.064606060606062e-06,
      "loss": 1.0719,
      "step": 551500
    },
    {
      "epoch": 25.73,
      "learning_rate": 9.05450505050505e-06,
      "loss": 1.0721,
      "step": 552000
    },
    {
      "epoch": 25.76,
      "learning_rate": 9.044404040404041e-06,
      "loss": 1.0748,
      "step": 552500
    },
    {
      "epoch": 25.78,
      "learning_rate": 9.034303030303031e-06,
      "loss": 1.0726,
      "step": 553000
    },
    {
      "epoch": 25.8,
      "learning_rate": 9.024222222222224e-06,
      "loss": 1.0716,
      "step": 553500
    },
    {
      "epoch": 25.83,
      "learning_rate": 9.014121212121212e-06,
      "loss": 1.0725,
      "step": 554000
    },
    {
      "epoch": 25.85,
      "learning_rate": 9.004020202020202e-06,
      "loss": 1.0751,
      "step": 554500
    },
    {
      "epoch": 25.87,
      "learning_rate": 8.993939393939395e-06,
      "loss": 1.0723,
      "step": 555000
    },
    {
      "epoch": 25.9,
      "learning_rate": 8.983838383838385e-06,
      "loss": 1.0735,
      "step": 555500
    },
    {
      "epoch": 25.92,
      "learning_rate": 8.973757575757577e-06,
      "loss": 1.0678,
      "step": 556000
    },
    {
      "epoch": 25.94,
      "learning_rate": 8.963656565656566e-06,
      "loss": 1.0729,
      "step": 556500
    },
    {
      "epoch": 25.97,
      "learning_rate": 8.953555555555556e-06,
      "loss": 1.074,
      "step": 557000
    },
    {
      "epoch": 25.99,
      "learning_rate": 8.943454545454546e-06,
      "loss": 1.0736,
      "step": 557500
    },
    {
      "epoch": 26.0,
      "eval_accuracy": 0.7744943101620074,
      "eval_loss": 0.9402647018432617,
      "eval_runtime": 51.303,
      "eval_samples_per_second": 910.434,
      "eval_steps_per_second": 14.229,
      "step": 557726
    },
    {
      "epoch": 26.01,
      "learning_rate": 8.933353535353537e-06,
      "loss": 1.0712,
      "step": 558000
    },
    {
      "epoch": 26.04,
      "learning_rate": 8.923272727272729e-06,
      "loss": 1.0712,
      "step": 558500
    },
    {
      "epoch": 26.06,
      "learning_rate": 8.913171717171718e-06,
      "loss": 1.0729,
      "step": 559000
    },
    {
      "epoch": 26.08,
      "learning_rate": 8.903070707070708e-06,
      "loss": 1.0721,
      "step": 559500
    },
    {
      "epoch": 26.11,
      "learning_rate": 8.892969696969698e-06,
      "loss": 1.0726,
      "step": 560000
    },
    {
      "epoch": 26.13,
      "learning_rate": 8.882868686868687e-06,
      "loss": 1.0743,
      "step": 560500
    },
    {
      "epoch": 26.15,
      "learning_rate": 8.872767676767677e-06,
      "loss": 1.0709,
      "step": 561000
    },
    {
      "epoch": 26.18,
      "learning_rate": 8.862666666666667e-06,
      "loss": 1.0711,
      "step": 561500
    },
    {
      "epoch": 26.2,
      "learning_rate": 8.852565656565658e-06,
      "loss": 1.0696,
      "step": 562000
    },
    {
      "epoch": 26.22,
      "learning_rate": 8.842464646464646e-06,
      "loss": 1.0733,
      "step": 562500
    },
    {
      "epoch": 26.25,
      "learning_rate": 8.832363636363636e-06,
      "loss": 1.073,
      "step": 563000
    },
    {
      "epoch": 26.27,
      "learning_rate": 8.822262626262627e-06,
      "loss": 1.0715,
      "step": 563500
    },
    {
      "epoch": 26.29,
      "learning_rate": 8.812161616161617e-06,
      "loss": 1.0725,
      "step": 564000
    },
    {
      "epoch": 26.32,
      "learning_rate": 8.80208080808081e-06,
      "loss": 1.0708,
      "step": 564500
    },
    {
      "epoch": 26.34,
      "learning_rate": 8.791979797979798e-06,
      "loss": 1.0712,
      "step": 565000
    },
    {
      "epoch": 26.36,
      "learning_rate": 8.781878787878788e-06,
      "loss": 1.0692,
      "step": 565500
    },
    {
      "epoch": 26.39,
      "learning_rate": 8.771777777777778e-06,
      "loss": 1.0713,
      "step": 566000
    },
    {
      "epoch": 26.41,
      "learning_rate": 8.761676767676769e-06,
      "loss": 1.0711,
      "step": 566500
    },
    {
      "epoch": 26.43,
      "learning_rate": 8.75159595959596e-06,
      "loss": 1.0686,
      "step": 567000
    },
    {
      "epoch": 26.46,
      "learning_rate": 8.74149494949495e-06,
      "loss": 1.0729,
      "step": 567500
    },
    {
      "epoch": 26.48,
      "learning_rate": 8.73139393939394e-06,
      "loss": 1.0696,
      "step": 568000
    },
    {
      "epoch": 26.5,
      "learning_rate": 8.72129292929293e-06,
      "loss": 1.0715,
      "step": 568500
    },
    {
      "epoch": 26.53,
      "learning_rate": 8.71119191919192e-06,
      "loss": 1.0731,
      "step": 569000
    },
    {
      "epoch": 26.55,
      "learning_rate": 8.70109090909091e-06,
      "loss": 1.07,
      "step": 569500
    },
    {
      "epoch": 26.57,
      "learning_rate": 8.691010101010103e-06,
      "loss": 1.073,
      "step": 570000
    },
    {
      "epoch": 26.6,
      "learning_rate": 8.680909090909092e-06,
      "loss": 1.0716,
      "step": 570500
    },
    {
      "epoch": 26.62,
      "learning_rate": 8.670808080808082e-06,
      "loss": 1.0712,
      "step": 571000
    },
    {
      "epoch": 26.64,
      "learning_rate": 8.660707070707072e-06,
      "loss": 1.073,
      "step": 571500
    },
    {
      "epoch": 26.67,
      "learning_rate": 8.65060606060606e-06,
      "loss": 1.0715,
      "step": 572000
    },
    {
      "epoch": 26.69,
      "learning_rate": 8.640525252525253e-06,
      "loss": 1.0699,
      "step": 572500
    },
    {
      "epoch": 26.71,
      "learning_rate": 8.630424242424243e-06,
      "loss": 1.0715,
      "step": 573000
    },
    {
      "epoch": 26.74,
      "learning_rate": 8.620323232323234e-06,
      "loss": 1.0722,
      "step": 573500
    },
    {
      "epoch": 26.76,
      "learning_rate": 8.610242424242426e-06,
      "loss": 1.0721,
      "step": 574000
    },
    {
      "epoch": 26.78,
      "learning_rate": 8.600141414141414e-06,
      "loss": 1.0706,
      "step": 574500
    },
    {
      "epoch": 26.81,
      "learning_rate": 8.590040404040405e-06,
      "loss": 1.0687,
      "step": 575000
    },
    {
      "epoch": 26.83,
      "learning_rate": 8.579939393939395e-06,
      "loss": 1.0727,
      "step": 575500
    },
    {
      "epoch": 26.85,
      "learning_rate": 8.569838383838385e-06,
      "loss": 1.0722,
      "step": 576000
    },
    {
      "epoch": 26.88,
      "learning_rate": 8.559737373737374e-06,
      "loss": 1.0712,
      "step": 576500
    },
    {
      "epoch": 26.9,
      "learning_rate": 8.549636363636364e-06,
      "loss": 1.0692,
      "step": 577000
    },
    {
      "epoch": 26.92,
      "learning_rate": 8.539535353535354e-06,
      "loss": 1.071,
      "step": 577500
    },
    {
      "epoch": 26.95,
      "learning_rate": 8.529434343434345e-06,
      "loss": 1.0699,
      "step": 578000
    },
    {
      "epoch": 26.97,
      "learning_rate": 8.519333333333333e-06,
      "loss": 1.0701,
      "step": 578500
    },
    {
      "epoch": 26.99,
      "learning_rate": 8.509232323232324e-06,
      "loss": 1.0723,
      "step": 579000
    },
    {
      "epoch": 27.0,
      "eval_accuracy": 0.7750111375101019,
      "eval_loss": 0.9380190968513489,
      "eval_runtime": 51.1897,
      "eval_samples_per_second": 912.449,
      "eval_steps_per_second": 14.261,
      "step": 579177
    },
    {
      "epoch": 27.02,
      "learning_rate": 8.499131313131314e-06,
      "loss": 1.0718,
      "step": 579500
    },
    {
      "epoch": 27.04,
      "learning_rate": 8.489050505050506e-06,
      "loss": 1.0705,
      "step": 580000
    },
    {
      "epoch": 27.06,
      "learning_rate": 8.478949494949495e-06,
      "loss": 1.0731,
      "step": 580500
    },
    {
      "epoch": 27.08,
      "learning_rate": 8.468848484848485e-06,
      "loss": 1.0687,
      "step": 581000
    },
    {
      "epoch": 27.11,
      "learning_rate": 8.458747474747475e-06,
      "loss": 1.0681,
      "step": 581500
    },
    {
      "epoch": 27.13,
      "learning_rate": 8.448666666666668e-06,
      "loss": 1.0669,
      "step": 582000
    },
    {
      "epoch": 27.15,
      "learning_rate": 8.438565656565658e-06,
      "loss": 1.0682,
      "step": 582500
    },
    {
      "epoch": 27.18,
      "learning_rate": 8.428464646464646e-06,
      "loss": 1.0701,
      "step": 583000
    },
    {
      "epoch": 27.2,
      "learning_rate": 8.418383838383839e-06,
      "loss": 1.0691,
      "step": 583500
    },
    {
      "epoch": 27.22,
      "learning_rate": 8.408282828282829e-06,
      "loss": 1.07,
      "step": 584000
    },
    {
      "epoch": 27.25,
      "learning_rate": 8.39818181818182e-06,
      "loss": 1.0705,
      "step": 584500
    },
    {
      "epoch": 27.27,
      "learning_rate": 8.388080808080808e-06,
      "loss": 1.0701,
      "step": 585000
    },
    {
      "epoch": 27.29,
      "learning_rate": 8.377979797979798e-06,
      "loss": 1.0693,
      "step": 585500
    },
    {
      "epoch": 27.32,
      "learning_rate": 8.367878787878788e-06,
      "loss": 1.0692,
      "step": 586000
    },
    {
      "epoch": 27.34,
      "learning_rate": 8.357777777777779e-06,
      "loss": 1.0685,
      "step": 586500
    },
    {
      "epoch": 27.36,
      "learning_rate": 8.347676767676767e-06,
      "loss": 1.069,
      "step": 587000
    },
    {
      "epoch": 27.39,
      "learning_rate": 8.33757575757576e-06,
      "loss": 1.0697,
      "step": 587500
    },
    {
      "epoch": 27.41,
      "learning_rate": 8.327474747474748e-06,
      "loss": 1.0697,
      "step": 588000
    },
    {
      "epoch": 27.43,
      "learning_rate": 8.317373737373738e-06,
      "loss": 1.07,
      "step": 588500
    },
    {
      "epoch": 27.46,
      "learning_rate": 8.307272727272728e-06,
      "loss": 1.0686,
      "step": 589000
    },
    {
      "epoch": 27.48,
      "learning_rate": 8.297171717171719e-06,
      "loss": 1.0697,
      "step": 589500
    },
    {
      "epoch": 27.5,
      "learning_rate": 8.287070707070707e-06,
      "loss": 1.0719,
      "step": 590000
    },
    {
      "epoch": 27.53,
      "learning_rate": 8.276969696969698e-06,
      "loss": 1.0691,
      "step": 590500
    },
    {
      "epoch": 27.55,
      "learning_rate": 8.26688888888889e-06,
      "loss": 1.0678,
      "step": 591000
    },
    {
      "epoch": 27.57,
      "learning_rate": 8.25678787878788e-06,
      "loss": 1.0706,
      "step": 591500
    },
    {
      "epoch": 27.6,
      "learning_rate": 8.246686868686869e-06,
      "loss": 1.0685,
      "step": 592000
    },
    {
      "epoch": 27.62,
      "learning_rate": 8.236585858585859e-06,
      "loss": 1.0681,
      "step": 592500
    },
    {
      "epoch": 27.64,
      "learning_rate": 8.226525252525253e-06,
      "loss": 1.0698,
      "step": 593000
    },
    {
      "epoch": 27.67,
      "learning_rate": 8.216424242424244e-06,
      "loss": 1.0673,
      "step": 593500
    },
    {
      "epoch": 27.69,
      "learning_rate": 8.206323232323234e-06,
      "loss": 1.0696,
      "step": 594000
    },
    {
      "epoch": 27.71,
      "learning_rate": 8.196222222222222e-06,
      "loss": 1.0701,
      "step": 594500
    },
    {
      "epoch": 27.74,
      "learning_rate": 8.186121212121213e-06,
      "loss": 1.0675,
      "step": 595000
    },
    {
      "epoch": 27.76,
      "learning_rate": 8.176020202020203e-06,
      "loss": 1.0692,
      "step": 595500
    },
    {
      "epoch": 27.78,
      "learning_rate": 8.165919191919193e-06,
      "loss": 1.0687,
      "step": 596000
    },
    {
      "epoch": 27.81,
      "learning_rate": 8.155818181818182e-06,
      "loss": 1.0696,
      "step": 596500
    },
    {
      "epoch": 27.83,
      "learning_rate": 8.145717171717172e-06,
      "loss": 1.0727,
      "step": 597000
    },
    {
      "epoch": 27.85,
      "learning_rate": 8.135636363636364e-06,
      "loss": 1.0679,
      "step": 597500
    },
    {
      "epoch": 27.88,
      "learning_rate": 8.125535353535355e-06,
      "loss": 1.0695,
      "step": 598000
    },
    {
      "epoch": 27.9,
      "learning_rate": 8.115434343434345e-06,
      "loss": 1.0691,
      "step": 598500
    },
    {
      "epoch": 27.92,
      "learning_rate": 8.105333333333334e-06,
      "loss": 1.0699,
      "step": 599000
    },
    {
      "epoch": 27.95,
      "learning_rate": 8.095232323232324e-06,
      "loss": 1.0701,
      "step": 599500
    },
    {
      "epoch": 27.97,
      "learning_rate": 8.085131313131314e-06,
      "loss": 1.071,
      "step": 600000
    },
    {
      "epoch": 27.99,
      "learning_rate": 8.075050505050506e-06,
      "loss": 1.0698,
      "step": 600500
    },
    {
      "epoch": 28.0,
      "eval_accuracy": 0.7756976814239462,
      "eval_loss": 0.9341722130775452,
      "eval_runtime": 51.3454,
      "eval_samples_per_second": 909.683,
      "eval_steps_per_second": 14.217,
      "step": 600628
    },
    {
      "epoch": 28.02,
      "learning_rate": 8.064969696969699e-06,
      "loss": 1.07,
      "step": 601000
    },
    {
      "epoch": 28.04,
      "learning_rate": 8.054868686868687e-06,
      "loss": 1.069,
      "step": 601500
    },
    {
      "epoch": 28.06,
      "learning_rate": 8.044767676767677e-06,
      "loss": 1.0696,
      "step": 602000
    },
    {
      "epoch": 28.09,
      "learning_rate": 8.034666666666668e-06,
      "loss": 1.0681,
      "step": 602500
    },
    {
      "epoch": 28.11,
      "learning_rate": 8.024565656565656e-06,
      "loss": 1.0654,
      "step": 603000
    },
    {
      "epoch": 28.13,
      "learning_rate": 8.014484848484849e-06,
      "loss": 1.0688,
      "step": 603500
    },
    {
      "epoch": 28.16,
      "learning_rate": 8.004383838383839e-06,
      "loss": 1.0679,
      "step": 604000
    },
    {
      "epoch": 28.18,
      "learning_rate": 7.99428282828283e-06,
      "loss": 1.0689,
      "step": 604500
    },
    {
      "epoch": 28.2,
      "learning_rate": 7.98418181818182e-06,
      "loss": 1.0697,
      "step": 605000
    },
    {
      "epoch": 28.23,
      "learning_rate": 7.974080808080808e-06,
      "loss": 1.0681,
      "step": 605500
    },
    {
      "epoch": 28.25,
      "learning_rate": 7.963979797979798e-06,
      "loss": 1.0699,
      "step": 606000
    },
    {
      "epoch": 28.27,
      "learning_rate": 7.953878787878789e-06,
      "loss": 1.0674,
      "step": 606500
    },
    {
      "epoch": 28.3,
      "learning_rate": 7.943797979797981e-06,
      "loss": 1.0694,
      "step": 607000
    },
    {
      "epoch": 28.32,
      "learning_rate": 7.93369696969697e-06,
      "loss": 1.0692,
      "step": 607500
    },
    {
      "epoch": 28.34,
      "learning_rate": 7.92359595959596e-06,
      "loss": 1.0686,
      "step": 608000
    },
    {
      "epoch": 28.37,
      "learning_rate": 7.91349494949495e-06,
      "loss": 1.0676,
      "step": 608500
    },
    {
      "epoch": 28.39,
      "learning_rate": 7.90339393939394e-06,
      "loss": 1.068,
      "step": 609000
    },
    {
      "epoch": 28.41,
      "learning_rate": 7.893292929292929e-06,
      "loss": 1.0663,
      "step": 609500
    },
    {
      "epoch": 28.44,
      "learning_rate": 7.88319191919192e-06,
      "loss": 1.0699,
      "step": 610000
    },
    {
      "epoch": 28.46,
      "learning_rate": 7.873111111111111e-06,
      "loss": 1.0702,
      "step": 610500
    },
    {
      "epoch": 28.48,
      "learning_rate": 7.863010101010102e-06,
      "loss": 1.0689,
      "step": 611000
    },
    {
      "epoch": 28.51,
      "learning_rate": 7.85290909090909e-06,
      "loss": 1.067,
      "step": 611500
    },
    {
      "epoch": 28.53,
      "learning_rate": 7.842808080808082e-06,
      "loss": 1.0672,
      "step": 612000
    },
    {
      "epoch": 28.55,
      "learning_rate": 7.832707070707073e-06,
      "loss": 1.0675,
      "step": 612500
    },
    {
      "epoch": 28.58,
      "learning_rate": 7.822606060606061e-06,
      "loss": 1.0684,
      "step": 613000
    },
    {
      "epoch": 28.6,
      "learning_rate": 7.812505050505052e-06,
      "loss": 1.0683,
      "step": 613500
    },
    {
      "epoch": 28.62,
      "learning_rate": 7.802404040404042e-06,
      "loss": 1.0701,
      "step": 614000
    },
    {
      "epoch": 28.65,
      "learning_rate": 7.79230303030303e-06,
      "loss": 1.0658,
      "step": 614500
    },
    {
      "epoch": 28.67,
      "learning_rate": 7.78220202020202e-06,
      "loss": 1.0666,
      "step": 615000
    },
    {
      "epoch": 28.69,
      "learning_rate": 7.772101010101011e-06,
      "loss": 1.0687,
      "step": 615500
    },
    {
      "epoch": 28.72,
      "learning_rate": 7.762020202020203e-06,
      "loss": 1.068,
      "step": 616000
    },
    {
      "epoch": 28.74,
      "learning_rate": 7.751919191919193e-06,
      "loss": 1.0688,
      "step": 616500
    },
    {
      "epoch": 28.76,
      "learning_rate": 7.741818181818182e-06,
      "loss": 1.069,
      "step": 617000
    },
    {
      "epoch": 28.79,
      "learning_rate": 7.731737373737374e-06,
      "loss": 1.0679,
      "step": 617500
    },
    {
      "epoch": 28.81,
      "learning_rate": 7.721636363636365e-06,
      "loss": 1.0652,
      "step": 618000
    },
    {
      "epoch": 28.83,
      "learning_rate": 7.711535353535355e-06,
      "loss": 1.0687,
      "step": 618500
    },
    {
      "epoch": 28.86,
      "learning_rate": 7.701434343434343e-06,
      "loss": 1.0686,
      "step": 619000
    },
    {
      "epoch": 28.88,
      "learning_rate": 7.691333333333334e-06,
      "loss": 1.0682,
      "step": 619500
    },
    {
      "epoch": 28.9,
      "learning_rate": 7.681232323232324e-06,
      "loss": 1.067,
      "step": 620000
    },
    {
      "epoch": 28.93,
      "learning_rate": 7.671151515151516e-06,
      "loss": 1.0663,
      "step": 620500
    },
    {
      "epoch": 28.95,
      "learning_rate": 7.661050505050505e-06,
      "loss": 1.0691,
      "step": 621000
    },
    {
      "epoch": 28.97,
      "learning_rate": 7.650969696969697e-06,
      "loss": 1.0673,
      "step": 621500
    },
    {
      "epoch": 29.0,
      "learning_rate": 7.640868686868687e-06,
      "loss": 1.0671,
      "step": 622000
    },
    {
      "epoch": 29.0,
      "eval_accuracy": 0.7755197536958046,
      "eval_loss": 0.9362251162528992,
      "eval_runtime": 51.4398,
      "eval_samples_per_second": 908.012,
      "eval_steps_per_second": 14.191,
      "step": 622079
    },
    {
      "epoch": 29.02,
      "learning_rate": 7.630767676767678e-06,
      "loss": 1.0657,
      "step": 622500
    },
    {
      "epoch": 29.04,
      "learning_rate": 7.620666666666667e-06,
      "loss": 1.0646,
      "step": 623000
    },
    {
      "epoch": 29.07,
      "learning_rate": 7.6105656565656575e-06,
      "loss": 1.0662,
      "step": 623500
    },
    {
      "epoch": 29.09,
      "learning_rate": 7.600464646464647e-06,
      "loss": 1.0658,
      "step": 624000
    },
    {
      "epoch": 29.11,
      "learning_rate": 7.590363636363637e-06,
      "loss": 1.0681,
      "step": 624500
    },
    {
      "epoch": 29.14,
      "learning_rate": 7.580262626262627e-06,
      "loss": 1.0661,
      "step": 625000
    },
    {
      "epoch": 29.16,
      "learning_rate": 7.570181818181819e-06,
      "loss": 1.0658,
      "step": 625500
    },
    {
      "epoch": 29.18,
      "learning_rate": 7.560080808080808e-06,
      "loss": 1.0671,
      "step": 626000
    },
    {
      "epoch": 29.21,
      "learning_rate": 7.549979797979799e-06,
      "loss": 1.0663,
      "step": 626500
    },
    {
      "epoch": 29.23,
      "learning_rate": 7.539878787878788e-06,
      "loss": 1.0676,
      "step": 627000
    },
    {
      "epoch": 29.25,
      "learning_rate": 7.529777777777778e-06,
      "loss": 1.068,
      "step": 627500
    },
    {
      "epoch": 29.28,
      "learning_rate": 7.519676767676768e-06,
      "loss": 1.0653,
      "step": 628000
    },
    {
      "epoch": 29.3,
      "learning_rate": 7.509575757575758e-06,
      "loss": 1.0652,
      "step": 628500
    },
    {
      "epoch": 29.32,
      "learning_rate": 7.4994747474747475e-06,
      "loss": 1.0669,
      "step": 629000
    },
    {
      "epoch": 29.35,
      "learning_rate": 7.489373737373739e-06,
      "loss": 1.0688,
      "step": 629500
    },
    {
      "epoch": 29.37,
      "learning_rate": 7.479272727272728e-06,
      "loss": 1.0692,
      "step": 630000
    },
    {
      "epoch": 29.39,
      "learning_rate": 7.4691919191919195e-06,
      "loss": 1.0686,
      "step": 630500
    },
    {
      "epoch": 29.42,
      "learning_rate": 7.459131313131314e-06,
      "loss": 1.0673,
      "step": 631000
    },
    {
      "epoch": 29.44,
      "learning_rate": 7.449030303030303e-06,
      "loss": 1.0651,
      "step": 631500
    },
    {
      "epoch": 29.46,
      "learning_rate": 7.438929292929293e-06,
      "loss": 1.0665,
      "step": 632000
    },
    {
      "epoch": 29.49,
      "learning_rate": 7.428828282828283e-06,
      "loss": 1.0698,
      "step": 632500
    },
    {
      "epoch": 29.51,
      "learning_rate": 7.418727272727273e-06,
      "loss": 1.0684,
      "step": 633000
    },
    {
      "epoch": 29.53,
      "learning_rate": 7.4086262626262626e-06,
      "loss": 1.0655,
      "step": 633500
    },
    {
      "epoch": 29.56,
      "learning_rate": 7.398545454545455e-06,
      "loss": 1.0675,
      "step": 634000
    },
    {
      "epoch": 29.58,
      "learning_rate": 7.388444444444445e-06,
      "loss": 1.0641,
      "step": 634500
    },
    {
      "epoch": 29.6,
      "learning_rate": 7.3783434343434345e-06,
      "loss": 1.0644,
      "step": 635000
    },
    {
      "epoch": 29.63,
      "learning_rate": 7.368242424242425e-06,
      "loss": 1.065,
      "step": 635500
    },
    {
      "epoch": 29.65,
      "learning_rate": 7.358141414141414e-06,
      "loss": 1.068,
      "step": 636000
    },
    {
      "epoch": 29.67,
      "learning_rate": 7.348040404040405e-06,
      "loss": 1.0653,
      "step": 636500
    },
    {
      "epoch": 29.7,
      "learning_rate": 7.337939393939395e-06,
      "loss": 1.0639,
      "step": 637000
    },
    {
      "epoch": 29.72,
      "learning_rate": 7.327838383838385e-06,
      "loss": 1.0684,
      "step": 637500
    },
    {
      "epoch": 29.74,
      "learning_rate": 7.3177373737373746e-06,
      "loss": 1.0658,
      "step": 638000
    },
    {
      "epoch": 29.77,
      "learning_rate": 7.307636363636365e-06,
      "loss": 1.0665,
      "step": 638500
    },
    {
      "epoch": 29.79,
      "learning_rate": 7.297535353535354e-06,
      "loss": 1.0667,
      "step": 639000
    },
    {
      "epoch": 29.81,
      "learning_rate": 7.287434343434345e-06,
      "loss": 1.0671,
      "step": 639500
    },
    {
      "epoch": 29.84,
      "learning_rate": 7.277333333333334e-06,
      "loss": 1.0681,
      "step": 640000
    },
    {
      "epoch": 29.86,
      "learning_rate": 7.2672323232323235e-06,
      "loss": 1.0632,
      "step": 640500
    },
    {
      "epoch": 29.88,
      "learning_rate": 7.257151515151516e-06,
      "loss": 1.0653,
      "step": 641000
    },
    {
      "epoch": 29.91,
      "learning_rate": 7.247070707070708e-06,
      "loss": 1.0657,
      "step": 641500
    },
    {
      "epoch": 29.93,
      "learning_rate": 7.236969696969698e-06,
      "loss": 1.0669,
      "step": 642000
    },
    {
      "epoch": 29.95,
      "learning_rate": 7.226868686868688e-06,
      "loss": 1.0672,
      "step": 642500
    },
    {
      "epoch": 29.98,
      "learning_rate": 7.216767676767677e-06,
      "loss": 1.0664,
      "step": 643000
    },
    {
      "epoch": 30.0,
      "learning_rate": 7.206666666666667e-06,
      "loss": 1.0661,
      "step": 643500
    },
    {
      "epoch": 30.0,
      "eval_accuracy": 0.7755370444239342,
      "eval_loss": 0.9345055818557739,
      "eval_runtime": 50.9635,
      "eval_samples_per_second": 916.499,
      "eval_steps_per_second": 14.324,
      "step": 643530
    },
    {
      "epoch": 30.02,
      "learning_rate": 7.196565656565657e-06,
      "loss": 1.0653,
      "step": 644000
    },
    {
      "epoch": 30.05,
      "learning_rate": 7.186464646464647e-06,
      "loss": 1.0651,
      "step": 644500
    },
    {
      "epoch": 30.07,
      "learning_rate": 7.1763636363636366e-06,
      "loss": 1.064,
      "step": 645000
    },
    {
      "epoch": 30.09,
      "learning_rate": 7.166282828282829e-06,
      "loss": 1.063,
      "step": 645500
    },
    {
      "epoch": 30.12,
      "learning_rate": 7.156181818181819e-06,
      "loss": 1.0644,
      "step": 646000
    },
    {
      "epoch": 30.14,
      "learning_rate": 7.1461010101010105e-06,
      "loss": 1.0645,
      "step": 646500
    },
    {
      "epoch": 30.16,
      "learning_rate": 7.136000000000001e-06,
      "loss": 1.067,
      "step": 647000
    },
    {
      "epoch": 30.19,
      "learning_rate": 7.12589898989899e-06,
      "loss": 1.0669,
      "step": 647500
    },
    {
      "epoch": 30.21,
      "learning_rate": 7.1158181818181825e-06,
      "loss": 1.0643,
      "step": 648000
    },
    {
      "epoch": 30.23,
      "learning_rate": 7.105717171717173e-06,
      "loss": 1.0688,
      "step": 648500
    },
    {
      "epoch": 30.25,
      "learning_rate": 7.095616161616162e-06,
      "loss": 1.0665,
      "step": 649000
    },
    {
      "epoch": 30.28,
      "learning_rate": 7.0855151515151525e-06,
      "loss": 1.0667,
      "step": 649500
    },
    {
      "epoch": 30.3,
      "learning_rate": 7.075414141414142e-06,
      "loss": 1.0654,
      "step": 650000
    },
    {
      "epoch": 30.32,
      "learning_rate": 7.065313131313131e-06,
      "loss": 1.064,
      "step": 650500
    },
    {
      "epoch": 30.35,
      "learning_rate": 7.055212121212122e-06,
      "loss": 1.065,
      "step": 651000
    },
    {
      "epoch": 30.37,
      "learning_rate": 7.045111111111111e-06,
      "loss": 1.0645,
      "step": 651500
    },
    {
      "epoch": 30.39,
      "learning_rate": 7.035030303030303e-06,
      "loss": 1.0621,
      "step": 652000
    },
    {
      "epoch": 30.42,
      "learning_rate": 7.024929292929294e-06,
      "loss": 1.062,
      "step": 652500
    },
    {
      "epoch": 30.44,
      "learning_rate": 7.014828282828283e-06,
      "loss": 1.0636,
      "step": 653000
    },
    {
      "epoch": 30.46,
      "learning_rate": 7.004727272727273e-06,
      "loss": 1.0639,
      "step": 653500
    },
    {
      "epoch": 30.49,
      "learning_rate": 6.994626262626263e-06,
      "loss": 1.0661,
      "step": 654000
    },
    {
      "epoch": 30.51,
      "learning_rate": 6.984525252525253e-06,
      "loss": 1.0629,
      "step": 654500
    },
    {
      "epoch": 30.53,
      "learning_rate": 6.9744242424242425e-06,
      "loss": 1.0637,
      "step": 655000
    },
    {
      "epoch": 30.56,
      "learning_rate": 6.964323232323232e-06,
      "loss": 1.0641,
      "step": 655500
    },
    {
      "epoch": 30.58,
      "learning_rate": 6.954222222222223e-06,
      "loss": 1.0659,
      "step": 656000
    },
    {
      "epoch": 30.6,
      "learning_rate": 6.9441414141414145e-06,
      "loss": 1.0637,
      "step": 656500
    },
    {
      "epoch": 30.63,
      "learning_rate": 6.934040404040404e-06,
      "loss": 1.0644,
      "step": 657000
    },
    {
      "epoch": 30.65,
      "learning_rate": 6.923939393939394e-06,
      "loss": 1.0644,
      "step": 657500
    },
    {
      "epoch": 30.67,
      "learning_rate": 6.9138383838383845e-06,
      "loss": 1.0632,
      "step": 658000
    },
    {
      "epoch": 30.7,
      "learning_rate": 6.903737373737375e-06,
      "loss": 1.0649,
      "step": 658500
    },
    {
      "epoch": 30.72,
      "learning_rate": 6.893656565656565e-06,
      "loss": 1.0661,
      "step": 659000
    },
    {
      "epoch": 30.74,
      "learning_rate": 6.883575757575758e-06,
      "loss": 1.0654,
      "step": 659500
    },
    {
      "epoch": 30.77,
      "learning_rate": 6.873474747474748e-06,
      "loss": 1.0679,
      "step": 660000
    },
    {
      "epoch": 30.79,
      "learning_rate": 6.863373737373737e-06,
      "loss": 1.0651,
      "step": 660500
    },
    {
      "epoch": 30.81,
      "learning_rate": 6.853272727272728e-06,
      "loss": 1.0638,
      "step": 661000
    },
    {
      "epoch": 30.84,
      "learning_rate": 6.843171717171718e-06,
      "loss": 1.0636,
      "step": 661500
    },
    {
      "epoch": 30.86,
      "learning_rate": 6.833070707070708e-06,
      "loss": 1.0627,
      "step": 662000
    },
    {
      "epoch": 30.88,
      "learning_rate": 6.822969696969698e-06,
      "loss": 1.0652,
      "step": 662500
    },
    {
      "epoch": 30.91,
      "learning_rate": 6.812868686868688e-06,
      "loss": 1.0645,
      "step": 663000
    },
    {
      "epoch": 30.93,
      "learning_rate": 6.802808080808081e-06,
      "loss": 1.0643,
      "step": 663500
    },
    {
      "epoch": 30.95,
      "learning_rate": 6.792707070707071e-06,
      "loss": 1.0649,
      "step": 664000
    },
    {
      "epoch": 30.98,
      "learning_rate": 6.782606060606061e-06,
      "loss": 1.0665,
      "step": 664500
    },
    {
      "epoch": 31.0,
      "eval_accuracy": 0.7756377564557672,
      "eval_loss": 0.9325026273727417,
      "eval_runtime": 51.8873,
      "eval_samples_per_second": 900.182,
      "eval_steps_per_second": 14.069,
      "step": 664981
    },
    {
      "epoch": 31.0,
      "learning_rate": 6.772505050505051e-06,
      "loss": 1.0664,
      "step": 665000
    },
    {
      "epoch": 31.02,
      "learning_rate": 6.762404040404042e-06,
      "loss": 1.064,
      "step": 665500
    },
    {
      "epoch": 31.05,
      "learning_rate": 6.752323232323232e-06,
      "loss": 1.0651,
      "step": 666000
    },
    {
      "epoch": 31.07,
      "learning_rate": 6.742222222222222e-06,
      "loss": 1.0635,
      "step": 666500
    },
    {
      "epoch": 31.09,
      "learning_rate": 6.7321212121212136e-06,
      "loss": 1.0643,
      "step": 667000
    },
    {
      "epoch": 31.12,
      "learning_rate": 6.722040404040404e-06,
      "loss": 1.0651,
      "step": 667500
    },
    {
      "epoch": 31.14,
      "learning_rate": 6.711939393939394e-06,
      "loss": 1.0636,
      "step": 668000
    },
    {
      "epoch": 31.16,
      "learning_rate": 6.701838383838385e-06,
      "loss": 1.0636,
      "step": 668500
    },
    {
      "epoch": 31.19,
      "learning_rate": 6.691737373737375e-06,
      "loss": 1.0647,
      "step": 669000
    },
    {
      "epoch": 31.21,
      "learning_rate": 6.681636363636364e-06,
      "loss": 1.0603,
      "step": 669500
    },
    {
      "epoch": 31.23,
      "learning_rate": 6.671535353535355e-06,
      "loss": 1.0653,
      "step": 670000
    },
    {
      "epoch": 31.26,
      "learning_rate": 6.661434343434344e-06,
      "loss": 1.0622,
      "step": 670500
    },
    {
      "epoch": 31.28,
      "learning_rate": 6.651333333333334e-06,
      "loss": 1.0618,
      "step": 671000
    },
    {
      "epoch": 31.3,
      "learning_rate": 6.641232323232324e-06,
      "loss": 1.0629,
      "step": 671500
    },
    {
      "epoch": 31.33,
      "learning_rate": 6.631131313131314e-06,
      "loss": 1.0626,
      "step": 672000
    },
    {
      "epoch": 31.35,
      "learning_rate": 6.621030303030304e-06,
      "loss": 1.0637,
      "step": 672500
    },
    {
      "epoch": 31.37,
      "learning_rate": 6.610929292929293e-06,
      "loss": 1.0606,
      "step": 673000
    },
    {
      "epoch": 31.4,
      "learning_rate": 6.600848484848485e-06,
      "loss": 1.0631,
      "step": 673500
    },
    {
      "epoch": 31.42,
      "learning_rate": 6.5907676767676775e-06,
      "loss": 1.0657,
      "step": 674000
    },
    {
      "epoch": 31.44,
      "learning_rate": 6.580666666666668e-06,
      "loss": 1.0653,
      "step": 674500
    },
    {
      "epoch": 31.47,
      "learning_rate": 6.570565656565657e-06,
      "loss": 1.0638,
      "step": 675000
    },
    {
      "epoch": 31.49,
      "learning_rate": 6.560464646464647e-06,
      "loss": 1.06,
      "step": 675500
    },
    {
      "epoch": 31.51,
      "learning_rate": 6.550363636363637e-06,
      "loss": 1.0648,
      "step": 676000
    },
    {
      "epoch": 31.54,
      "learning_rate": 6.540262626262626e-06,
      "loss": 1.0593,
      "step": 676500
    },
    {
      "epoch": 31.56,
      "learning_rate": 6.530161616161617e-06,
      "loss": 1.0615,
      "step": 677000
    },
    {
      "epoch": 31.58,
      "learning_rate": 6.520060606060606e-06,
      "loss": 1.0651,
      "step": 677500
    },
    {
      "epoch": 31.61,
      "learning_rate": 6.5099595959595964e-06,
      "loss": 1.0632,
      "step": 678000
    },
    {
      "epoch": 31.63,
      "learning_rate": 6.499858585858586e-06,
      "loss": 1.0639,
      "step": 678500
    },
    {
      "epoch": 31.65,
      "learning_rate": 6.489757575757576e-06,
      "loss": 1.0623,
      "step": 679000
    },
    {
      "epoch": 31.68,
      "learning_rate": 6.479656565656566e-06,
      "loss": 1.0648,
      "step": 679500
    },
    {
      "epoch": 31.7,
      "learning_rate": 6.469575757575758e-06,
      "loss": 1.0643,
      "step": 680000
    },
    {
      "epoch": 31.72,
      "learning_rate": 6.459474747474747e-06,
      "loss": 1.0636,
      "step": 680500
    },
    {
      "epoch": 31.75,
      "learning_rate": 6.4493939393939395e-06,
      "loss": 1.0634,
      "step": 681000
    },
    {
      "epoch": 31.77,
      "learning_rate": 6.43929292929293e-06,
      "loss": 1.0624,
      "step": 681500
    },
    {
      "epoch": 31.79,
      "learning_rate": 6.429191919191919e-06,
      "loss": 1.0629,
      "step": 682000
    },
    {
      "epoch": 31.82,
      "learning_rate": 6.4190909090909095e-06,
      "loss": 1.0632,
      "step": 682500
    },
    {
      "epoch": 31.84,
      "learning_rate": 6.408989898989899e-06,
      "loss": 1.0627,
      "step": 683000
    },
    {
      "epoch": 31.86,
      "learning_rate": 6.398888888888889e-06,
      "loss": 1.0643,
      "step": 683500
    },
    {
      "epoch": 31.89,
      "learning_rate": 6.388787878787879e-06,
      "loss": 1.0622,
      "step": 684000
    },
    {
      "epoch": 31.91,
      "learning_rate": 6.37868686868687e-06,
      "loss": 1.0626,
      "step": 684500
    },
    {
      "epoch": 31.93,
      "learning_rate": 6.368585858585859e-06,
      "loss": 1.064,
      "step": 685000
    },
    {
      "epoch": 31.96,
      "learning_rate": 6.358525252525253e-06,
      "loss": 1.0613,
      "step": 685500
    },
    {
      "epoch": 31.98,
      "learning_rate": 6.348424242424243e-06,
      "loss": 1.0651,
      "step": 686000
    },
    {
      "epoch": 32.0,
      "eval_accuracy": 0.7763840093516086,
      "eval_loss": 0.9292020201683044,
      "eval_runtime": 51.8283,
      "eval_samples_per_second": 901.207,
      "eval_steps_per_second": 14.085,
      "step": 686432
    },
    {
      "epoch": 32.0,
      "learning_rate": 6.338323232323232e-06,
      "loss": 1.0618,
      "step": 686500
    },
    {
      "epoch": 32.03,
      "learning_rate": 6.328222222222223e-06,
      "loss": 1.0623,
      "step": 687000
    },
    {
      "epoch": 32.05,
      "learning_rate": 6.318121212121212e-06,
      "loss": 1.0617,
      "step": 687500
    },
    {
      "epoch": 32.07,
      "learning_rate": 6.308020202020203e-06,
      "loss": 1.0606,
      "step": 688000
    },
    {
      "epoch": 32.1,
      "learning_rate": 6.297919191919193e-06,
      "loss": 1.0635,
      "step": 688500
    },
    {
      "epoch": 32.12,
      "learning_rate": 6.287818181818183e-06,
      "loss": 1.0609,
      "step": 689000
    },
    {
      "epoch": 32.14,
      "learning_rate": 6.277737373737375e-06,
      "loss": 1.0654,
      "step": 689500
    },
    {
      "epoch": 32.17,
      "learning_rate": 6.267656565656566e-06,
      "loss": 1.0638,
      "step": 690000
    },
    {
      "epoch": 32.19,
      "learning_rate": 6.257555555555555e-06,
      "loss": 1.0616,
      "step": 690500
    },
    {
      "epoch": 32.21,
      "learning_rate": 6.2474545454545455e-06,
      "loss": 1.063,
      "step": 691000
    },
    {
      "epoch": 32.24,
      "learning_rate": 6.237353535353537e-06,
      "loss": 1.0609,
      "step": 691500
    },
    {
      "epoch": 32.26,
      "learning_rate": 6.227252525252526e-06,
      "loss": 1.0604,
      "step": 692000
    },
    {
      "epoch": 32.28,
      "learning_rate": 6.217151515151516e-06,
      "loss": 1.0625,
      "step": 692500
    },
    {
      "epoch": 32.31,
      "learning_rate": 6.207050505050506e-06,
      "loss": 1.0641,
      "step": 693000
    },
    {
      "epoch": 32.33,
      "learning_rate": 6.196949494949496e-06,
      "loss": 1.0635,
      "step": 693500
    },
    {
      "epoch": 32.35,
      "learning_rate": 6.1868484848484855e-06,
      "loss": 1.0607,
      "step": 694000
    },
    {
      "epoch": 32.38,
      "learning_rate": 6.176767676767678e-06,
      "loss": 1.0643,
      "step": 694500
    },
    {
      "epoch": 32.4,
      "learning_rate": 6.166666666666667e-06,
      "loss": 1.0627,
      "step": 695000
    },
    {
      "epoch": 32.42,
      "learning_rate": 6.1565656565656575e-06,
      "loss": 1.0634,
      "step": 695500
    },
    {
      "epoch": 32.45,
      "learning_rate": 6.146464646464647e-06,
      "loss": 1.0627,
      "step": 696000
    },
    {
      "epoch": 32.47,
      "learning_rate": 6.136363636363637e-06,
      "loss": 1.0619,
      "step": 696500
    },
    {
      "epoch": 32.49,
      "learning_rate": 6.126262626262627e-06,
      "loss": 1.061,
      "step": 697000
    },
    {
      "epoch": 32.52,
      "learning_rate": 6.116161616161617e-06,
      "loss": 1.0631,
      "step": 697500
    },
    {
      "epoch": 32.54,
      "learning_rate": 6.106080808080808e-06,
      "loss": 1.0624,
      "step": 698000
    },
    {
      "epoch": 32.56,
      "learning_rate": 6.095979797979799e-06,
      "loss": 1.0607,
      "step": 698500
    },
    {
      "epoch": 32.59,
      "learning_rate": 6.085878787878788e-06,
      "loss": 1.0626,
      "step": 699000
    },
    {
      "epoch": 32.61,
      "learning_rate": 6.075777777777778e-06,
      "loss": 1.0624,
      "step": 699500
    },
    {
      "epoch": 32.63,
      "learning_rate": 6.065676767676768e-06,
      "loss": 1.0627,
      "step": 700000
    },
    {
      "epoch": 32.66,
      "learning_rate": 6.055575757575758e-06,
      "loss": 1.0628,
      "step": 700500
    },
    {
      "epoch": 32.68,
      "learning_rate": 6.04549494949495e-06,
      "loss": 1.0617,
      "step": 701000
    },
    {
      "epoch": 32.7,
      "learning_rate": 6.03539393939394e-06,
      "loss": 1.0632,
      "step": 701500
    },
    {
      "epoch": 32.73,
      "learning_rate": 6.02529292929293e-06,
      "loss": 1.0616,
      "step": 702000
    },
    {
      "epoch": 32.75,
      "learning_rate": 6.0151919191919195e-06,
      "loss": 1.0619,
      "step": 702500
    },
    {
      "epoch": 32.77,
      "learning_rate": 6.005090909090909e-06,
      "loss": 1.0621,
      "step": 703000
    },
    {
      "epoch": 32.8,
      "learning_rate": 5.994989898989899e-06,
      "loss": 1.0611,
      "step": 703500
    },
    {
      "epoch": 32.82,
      "learning_rate": 5.984888888888889e-06,
      "loss": 1.0634,
      "step": 704000
    },
    {
      "epoch": 32.84,
      "learning_rate": 5.974808080808081e-06,
      "loss": 1.0608,
      "step": 704500
    },
    {
      "epoch": 32.87,
      "learning_rate": 5.964707070707071e-06,
      "loss": 1.0625,
      "step": 705000
    },
    {
      "epoch": 32.89,
      "learning_rate": 5.954606060606061e-06,
      "loss": 1.0622,
      "step": 705500
    },
    {
      "epoch": 32.91,
      "learning_rate": 5.944505050505051e-06,
      "loss": 1.0609,
      "step": 706000
    },
    {
      "epoch": 32.94,
      "learning_rate": 5.934424242424242e-06,
      "loss": 1.061,
      "step": 706500
    },
    {
      "epoch": 32.96,
      "learning_rate": 5.924323232323233e-06,
      "loss": 1.0592,
      "step": 707000
    },
    {
      "epoch": 32.98,
      "learning_rate": 5.914222222222222e-06,
      "loss": 1.0621,
      "step": 707500
    },
    {
      "epoch": 33.0,
      "eval_accuracy": 0.7764886248284107,
      "eval_loss": 0.9295682311058044,
      "eval_runtime": 51.7308,
      "eval_samples_per_second": 902.905,
      "eval_steps_per_second": 14.112,
      "step": 707883
    },
    {
      "epoch": 33.01,
      "learning_rate": 5.904121212121212e-06,
      "loss": 1.0626,
      "step": 708000
    },
    {
      "epoch": 33.03,
      "learning_rate": 5.894020202020202e-06,
      "loss": 1.0598,
      "step": 708500
    },
    {
      "epoch": 33.05,
      "learning_rate": 5.883919191919193e-06,
      "loss": 1.0624,
      "step": 709000
    },
    {
      "epoch": 33.08,
      "learning_rate": 5.873818181818182e-06,
      "loss": 1.0616,
      "step": 709500
    },
    {
      "epoch": 33.1,
      "learning_rate": 5.863717171717173e-06,
      "loss": 1.0617,
      "step": 710000
    },
    {
      "epoch": 33.12,
      "learning_rate": 5.853616161616162e-06,
      "loss": 1.0611,
      "step": 710500
    },
    {
      "epoch": 33.15,
      "learning_rate": 5.843535353535354e-06,
      "loss": 1.06,
      "step": 711000
    },
    {
      "epoch": 33.17,
      "learning_rate": 5.833454545454546e-06,
      "loss": 1.0615,
      "step": 711500
    },
    {
      "epoch": 33.19,
      "learning_rate": 5.823353535353535e-06,
      "loss": 1.0602,
      "step": 712000
    },
    {
      "epoch": 33.22,
      "learning_rate": 5.813252525252526e-06,
      "loss": 1.06,
      "step": 712500
    },
    {
      "epoch": 33.24,
      "learning_rate": 5.803151515151516e-06,
      "loss": 1.0606,
      "step": 713000
    },
    {
      "epoch": 33.26,
      "learning_rate": 5.793050505050506e-06,
      "loss": 1.0624,
      "step": 713500
    },
    {
      "epoch": 33.29,
      "learning_rate": 5.7829696969696966e-06,
      "loss": 1.0603,
      "step": 714000
    },
    {
      "epoch": 33.31,
      "learning_rate": 5.772868686868688e-06,
      "loss": 1.0606,
      "step": 714500
    },
    {
      "epoch": 33.33,
      "learning_rate": 5.762787878787879e-06,
      "loss": 1.0599,
      "step": 715000
    },
    {
      "epoch": 33.36,
      "learning_rate": 5.7526868686868686e-06,
      "loss": 1.0618,
      "step": 715500
    },
    {
      "epoch": 33.38,
      "learning_rate": 5.74258585858586e-06,
      "loss": 1.0602,
      "step": 716000
    },
    {
      "epoch": 33.4,
      "learning_rate": 5.732484848484849e-06,
      "loss": 1.0607,
      "step": 716500
    },
    {
      "epoch": 33.43,
      "learning_rate": 5.722383838383839e-06,
      "loss": 1.0583,
      "step": 717000
    },
    {
      "epoch": 33.45,
      "learning_rate": 5.712282828282829e-06,
      "loss": 1.0595,
      "step": 717500
    },
    {
      "epoch": 33.47,
      "learning_rate": 5.702181818181819e-06,
      "loss": 1.0612,
      "step": 718000
    },
    {
      "epoch": 33.49,
      "learning_rate": 5.692101010101011e-06,
      "loss": 1.0612,
      "step": 718500
    },
    {
      "epoch": 33.52,
      "learning_rate": 5.682000000000001e-06,
      "loss": 1.0611,
      "step": 719000
    },
    {
      "epoch": 33.54,
      "learning_rate": 5.67189898989899e-06,
      "loss": 1.0593,
      "step": 719500
    },
    {
      "epoch": 33.56,
      "learning_rate": 5.6617979797979806e-06,
      "loss": 1.0601,
      "step": 720000
    },
    {
      "epoch": 33.59,
      "learning_rate": 5.65169696969697e-06,
      "loss": 1.0619,
      "step": 720500
    },
    {
      "epoch": 33.61,
      "learning_rate": 5.64159595959596e-06,
      "loss": 1.0618,
      "step": 721000
    },
    {
      "epoch": 33.63,
      "learning_rate": 5.6315151515151525e-06,
      "loss": 1.061,
      "step": 721500
    },
    {
      "epoch": 33.66,
      "learning_rate": 5.621414141414142e-06,
      "loss": 1.064,
      "step": 722000
    },
    {
      "epoch": 33.68,
      "learning_rate": 5.611313131313132e-06,
      "loss": 1.0603,
      "step": 722500
    },
    {
      "epoch": 33.7,
      "learning_rate": 5.601212121212122e-06,
      "loss": 1.0592,
      "step": 723000
    },
    {
      "epoch": 33.73,
      "learning_rate": 5.591131313131314e-06,
      "loss": 1.061,
      "step": 723500
    },
    {
      "epoch": 33.75,
      "learning_rate": 5.581030303030303e-06,
      "loss": 1.0585,
      "step": 724000
    },
    {
      "epoch": 33.77,
      "learning_rate": 5.570929292929294e-06,
      "loss": 1.0612,
      "step": 724500
    },
    {
      "epoch": 33.8,
      "learning_rate": 5.560828282828283e-06,
      "loss": 1.0613,
      "step": 725000
    },
    {
      "epoch": 33.82,
      "learning_rate": 5.550727272727273e-06,
      "loss": 1.0607,
      "step": 725500
    },
    {
      "epoch": 33.84,
      "learning_rate": 5.540626262626263e-06,
      "loss": 1.0601,
      "step": 726000
    },
    {
      "epoch": 33.87,
      "learning_rate": 5.530525252525253e-06,
      "loss": 1.0604,
      "step": 726500
    },
    {
      "epoch": 33.89,
      "learning_rate": 5.5204242424242426e-06,
      "loss": 1.0601,
      "step": 727000
    },
    {
      "epoch": 33.91,
      "learning_rate": 5.510323232323233e-06,
      "loss": 1.0587,
      "step": 727500
    },
    {
      "epoch": 33.94,
      "learning_rate": 5.500222222222222e-06,
      "loss": 1.0625,
      "step": 728000
    },
    {
      "epoch": 33.96,
      "learning_rate": 5.4901212121212126e-06,
      "loss": 1.0623,
      "step": 728500
    },
    {
      "epoch": 33.98,
      "learning_rate": 5.480020202020202e-06,
      "loss": 1.0604,
      "step": 729000
    },
    {
      "epoch": 34.0,
      "eval_accuracy": 0.7766372529019103,
      "eval_loss": 0.9284267425537109,
      "eval_runtime": 51.875,
      "eval_samples_per_second": 900.394,
      "eval_steps_per_second": 14.072,
      "step": 729334
    },
    {
      "epoch": 34.01,
      "learning_rate": 5.4699191919191914e-06,
      "loss": 1.0579,
      "step": 729500
    },
    {
      "epoch": 34.03,
      "learning_rate": 5.459818181818183e-06,
      "loss": 1.0565,
      "step": 730000
    },
    {
      "epoch": 34.05,
      "learning_rate": 5.449757575757576e-06,
      "loss": 1.0576,
      "step": 730500
    },
    {
      "epoch": 34.08,
      "learning_rate": 5.439656565656566e-06,
      "loss": 1.0601,
      "step": 731000
    },
    {
      "epoch": 34.1,
      "learning_rate": 5.429575757575758e-06,
      "loss": 1.0591,
      "step": 731500
    },
    {
      "epoch": 34.12,
      "learning_rate": 5.419474747474748e-06,
      "loss": 1.0599,
      "step": 732000
    },
    {
      "epoch": 34.15,
      "learning_rate": 5.409373737373737e-06,
      "loss": 1.0621,
      "step": 732500
    },
    {
      "epoch": 34.17,
      "learning_rate": 5.399272727272728e-06,
      "loss": 1.0605,
      "step": 733000
    },
    {
      "epoch": 34.19,
      "learning_rate": 5.389171717171717e-06,
      "loss": 1.0603,
      "step": 733500
    },
    {
      "epoch": 34.22,
      "learning_rate": 5.379070707070707e-06,
      "loss": 1.0597,
      "step": 734000
    },
    {
      "epoch": 34.24,
      "learning_rate": 5.368969696969697e-06,
      "loss": 1.059,
      "step": 734500
    },
    {
      "epoch": 34.26,
      "learning_rate": 5.358868686868687e-06,
      "loss": 1.0595,
      "step": 735000
    },
    {
      "epoch": 34.29,
      "learning_rate": 5.348767676767677e-06,
      "loss": 1.0603,
      "step": 735500
    },
    {
      "epoch": 34.31,
      "learning_rate": 5.338666666666668e-06,
      "loss": 1.0595,
      "step": 736000
    },
    {
      "epoch": 34.33,
      "learning_rate": 5.328565656565657e-06,
      "loss": 1.0598,
      "step": 736500
    },
    {
      "epoch": 34.36,
      "learning_rate": 5.318464646464647e-06,
      "loss": 1.0589,
      "step": 737000
    },
    {
      "epoch": 34.38,
      "learning_rate": 5.308363636363637e-06,
      "loss": 1.0576,
      "step": 737500
    },
    {
      "epoch": 34.4,
      "learning_rate": 5.298262626262627e-06,
      "loss": 1.0613,
      "step": 738000
    },
    {
      "epoch": 34.43,
      "learning_rate": 5.2881616161616166e-06,
      "loss": 1.0579,
      "step": 738500
    },
    {
      "epoch": 34.45,
      "learning_rate": 5.278060606060607e-06,
      "loss": 1.0601,
      "step": 739000
    },
    {
      "epoch": 34.47,
      "learning_rate": 5.267959595959596e-06,
      "loss": 1.0595,
      "step": 739500
    },
    {
      "epoch": 34.5,
      "learning_rate": 5.257858585858587e-06,
      "loss": 1.0597,
      "step": 740000
    },
    {
      "epoch": 34.52,
      "learning_rate": 5.247777777777778e-06,
      "loss": 1.0619,
      "step": 740500
    },
    {
      "epoch": 34.54,
      "learning_rate": 5.237676767676768e-06,
      "loss": 1.0589,
      "step": 741000
    },
    {
      "epoch": 34.57,
      "learning_rate": 5.227575757575758e-06,
      "loss": 1.0594,
      "step": 741500
    },
    {
      "epoch": 34.59,
      "learning_rate": 5.21749494949495e-06,
      "loss": 1.0583,
      "step": 742000
    },
    {
      "epoch": 34.61,
      "learning_rate": 5.20739393939394e-06,
      "loss": 1.0591,
      "step": 742500
    },
    {
      "epoch": 34.64,
      "learning_rate": 5.19729292929293e-06,
      "loss": 1.0572,
      "step": 743000
    },
    {
      "epoch": 34.66,
      "learning_rate": 5.18719191919192e-06,
      "loss": 1.0585,
      "step": 743500
    },
    {
      "epoch": 34.68,
      "learning_rate": 5.177090909090909e-06,
      "loss": 1.0599,
      "step": 744000
    },
    {
      "epoch": 34.71,
      "learning_rate": 5.166989898989899e-06,
      "loss": 1.0598,
      "step": 744500
    },
    {
      "epoch": 34.73,
      "learning_rate": 5.156888888888889e-06,
      "loss": 1.059,
      "step": 745000
    },
    {
      "epoch": 34.75,
      "learning_rate": 5.1467878787878786e-06,
      "loss": 1.0594,
      "step": 745500
    },
    {
      "epoch": 34.78,
      "learning_rate": 5.136686868686869e-06,
      "loss": 1.0597,
      "step": 746000
    },
    {
      "epoch": 34.8,
      "learning_rate": 5.126585858585858e-06,
      "loss": 1.0589,
      "step": 746500
    },
    {
      "epoch": 34.82,
      "learning_rate": 5.116484848484849e-06,
      "loss": 1.0587,
      "step": 747000
    },
    {
      "epoch": 34.85,
      "learning_rate": 5.10638383838384e-06,
      "loss": 1.0577,
      "step": 747500
    },
    {
      "epoch": 34.87,
      "learning_rate": 5.096323232323232e-06,
      "loss": 1.0583,
      "step": 748000
    },
    {
      "epoch": 34.89,
      "learning_rate": 5.0862222222222225e-06,
      "loss": 1.0585,
      "step": 748500
    },
    {
      "epoch": 34.92,
      "learning_rate": 5.076121212121212e-06,
      "loss": 1.0634,
      "step": 749000
    },
    {
      "epoch": 34.94,
      "learning_rate": 5.066020202020202e-06,
      "loss": 1.0587,
      "step": 749500
    },
    {
      "epoch": 34.96,
      "learning_rate": 5.0559393939393945e-06,
      "loss": 1.0594,
      "step": 750000
    },
    {
      "epoch": 34.99,
      "learning_rate": 5.045858585858586e-06,
      "loss": 1.0589,
      "step": 750500
    },
    {
      "epoch": 35.0,
      "eval_accuracy": 0.7765094261586242,
      "eval_loss": 0.9280404448509216,
      "eval_runtime": 50.983,
      "eval_samples_per_second": 916.149,
      "eval_steps_per_second": 14.319,
      "step": 750785
    },
    {
      "epoch": 35.01,
      "learning_rate": 5.035757575757576e-06,
      "loss": 1.0576,
      "step": 751000
    },
    {
      "epoch": 35.03,
      "learning_rate": 5.025656565656566e-06,
      "loss": 1.0591,
      "step": 751500
    },
    {
      "epoch": 35.06,
      "learning_rate": 5.015555555555556e-06,
      "loss": 1.0618,
      "step": 752000
    },
    {
      "epoch": 35.08,
      "learning_rate": 5.005474747474748e-06,
      "loss": 1.0564,
      "step": 752500
    },
    {
      "epoch": 35.1,
      "learning_rate": 4.995373737373738e-06,
      "loss": 1.0606,
      "step": 753000
    },
    {
      "epoch": 35.13,
      "learning_rate": 4.985272727272728e-06,
      "loss": 1.0591,
      "step": 753500
    },
    {
      "epoch": 35.15,
      "learning_rate": 4.975171717171717e-06,
      "loss": 1.0584,
      "step": 754000
    },
    {
      "epoch": 35.17,
      "learning_rate": 4.965070707070708e-06,
      "loss": 1.0595,
      "step": 754500
    },
    {
      "epoch": 35.2,
      "learning_rate": 4.954989898989899e-06,
      "loss": 1.0593,
      "step": 755000
    },
    {
      "epoch": 35.22,
      "learning_rate": 4.944888888888889e-06,
      "loss": 1.0568,
      "step": 755500
    },
    {
      "epoch": 35.24,
      "learning_rate": 4.93478787878788e-06,
      "loss": 1.0614,
      "step": 756000
    },
    {
      "epoch": 35.27,
      "learning_rate": 4.924686868686869e-06,
      "loss": 1.0599,
      "step": 756500
    },
    {
      "epoch": 35.29,
      "learning_rate": 4.914585858585859e-06,
      "loss": 1.058,
      "step": 757000
    },
    {
      "epoch": 35.31,
      "learning_rate": 4.904484848484849e-06,
      "loss": 1.0598,
      "step": 757500
    },
    {
      "epoch": 35.34,
      "learning_rate": 4.894404040404041e-06,
      "loss": 1.0578,
      "step": 758000
    },
    {
      "epoch": 35.36,
      "learning_rate": 4.884303030303031e-06,
      "loss": 1.0564,
      "step": 758500
    },
    {
      "epoch": 35.38,
      "learning_rate": 4.874202020202021e-06,
      "loss": 1.058,
      "step": 759000
    },
    {
      "epoch": 35.41,
      "learning_rate": 4.86410101010101e-06,
      "loss": 1.0571,
      "step": 759500
    },
    {
      "epoch": 35.43,
      "learning_rate": 4.8540000000000005e-06,
      "loss": 1.0576,
      "step": 760000
    },
    {
      "epoch": 35.45,
      "learning_rate": 4.84389898989899e-06,
      "loss": 1.0581,
      "step": 760500
    },
    {
      "epoch": 35.48,
      "learning_rate": 4.833818181818182e-06,
      "loss": 1.0579,
      "step": 761000
    },
    {
      "epoch": 35.5,
      "learning_rate": 4.8237171717171724e-06,
      "loss": 1.0596,
      "step": 761500
    },
    {
      "epoch": 35.52,
      "learning_rate": 4.813616161616162e-06,
      "loss": 1.0605,
      "step": 762000
    },
    {
      "epoch": 35.55,
      "learning_rate": 4.803515151515152e-06,
      "loss": 1.0559,
      "step": 762500
    },
    {
      "epoch": 35.57,
      "learning_rate": 4.793414141414142e-06,
      "loss": 1.0593,
      "step": 763000
    },
    {
      "epoch": 35.59,
      "learning_rate": 4.783313131313132e-06,
      "loss": 1.0561,
      "step": 763500
    },
    {
      "epoch": 35.62,
      "learning_rate": 4.773212121212122e-06,
      "loss": 1.0598,
      "step": 764000
    },
    {
      "epoch": 35.64,
      "learning_rate": 4.7631313131313136e-06,
      "loss": 1.0588,
      "step": 764500
    },
    {
      "epoch": 35.66,
      "learning_rate": 4.753030303030303e-06,
      "loss": 1.0566,
      "step": 765000
    },
    {
      "epoch": 35.69,
      "learning_rate": 4.742929292929293e-06,
      "loss": 1.0571,
      "step": 765500
    },
    {
      "epoch": 35.71,
      "learning_rate": 4.732828282828284e-06,
      "loss": 1.057,
      "step": 766000
    },
    {
      "epoch": 35.73,
      "learning_rate": 4.722727272727273e-06,
      "loss": 1.0578,
      "step": 766500
    },
    {
      "epoch": 35.76,
      "learning_rate": 4.712626262626263e-06,
      "loss": 1.0581,
      "step": 767000
    },
    {
      "epoch": 35.78,
      "learning_rate": 4.702525252525253e-06,
      "loss": 1.0581,
      "step": 767500
    },
    {
      "epoch": 35.8,
      "learning_rate": 4.692424242424243e-06,
      "loss": 1.0609,
      "step": 768000
    },
    {
      "epoch": 35.83,
      "learning_rate": 4.6823232323232325e-06,
      "loss": 1.0565,
      "step": 768500
    },
    {
      "epoch": 35.85,
      "learning_rate": 4.672222222222223e-06,
      "loss": 1.0566,
      "step": 769000
    },
    {
      "epoch": 35.87,
      "learning_rate": 4.662121212121212e-06,
      "loss": 1.0553,
      "step": 769500
    },
    {
      "epoch": 35.9,
      "learning_rate": 4.6520404040404044e-06,
      "loss": 1.057,
      "step": 770000
    },
    {
      "epoch": 35.92,
      "learning_rate": 4.641939393939394e-06,
      "loss": 1.0573,
      "step": 770500
    },
    {
      "epoch": 35.94,
      "learning_rate": 4.631838383838384e-06,
      "loss": 1.0575,
      "step": 771000
    },
    {
      "epoch": 35.97,
      "learning_rate": 4.621737373737374e-06,
      "loss": 1.0588,
      "step": 771500
    },
    {
      "epoch": 35.99,
      "learning_rate": 4.611656565656566e-06,
      "loss": 1.058,
      "step": 772000
    },
    {
      "epoch": 36.0,
      "eval_accuracy": 0.7773394538349336,
      "eval_loss": 0.9250308871269226,
      "eval_runtime": 51.3526,
      "eval_samples_per_second": 909.555,
      "eval_steps_per_second": 14.215,
      "step": 772236
    },
    {
      "epoch": 36.01,
      "learning_rate": 4.601555555555556e-06,
      "loss": 1.0569,
      "step": 772500
    },
    {
      "epoch": 36.04,
      "learning_rate": 4.591454545454546e-06,
      "loss": 1.0574,
      "step": 773000
    },
    {
      "epoch": 36.06,
      "learning_rate": 4.581353535353536e-06,
      "loss": 1.0582,
      "step": 773500
    },
    {
      "epoch": 36.08,
      "learning_rate": 4.571252525252526e-06,
      "loss": 1.0566,
      "step": 774000
    },
    {
      "epoch": 36.11,
      "learning_rate": 4.561151515151516e-06,
      "loss": 1.0566,
      "step": 774500
    },
    {
      "epoch": 36.13,
      "learning_rate": 4.551050505050506e-06,
      "loss": 1.0554,
      "step": 775000
    },
    {
      "epoch": 36.15,
      "learning_rate": 4.540969696969697e-06,
      "loss": 1.0608,
      "step": 775500
    },
    {
      "epoch": 36.18,
      "learning_rate": 4.5308686868686876e-06,
      "loss": 1.0568,
      "step": 776000
    },
    {
      "epoch": 36.2,
      "learning_rate": 4.520767676767677e-06,
      "loss": 1.0549,
      "step": 776500
    },
    {
      "epoch": 36.22,
      "learning_rate": 4.510666666666667e-06,
      "loss": 1.0576,
      "step": 777000
    },
    {
      "epoch": 36.25,
      "learning_rate": 4.5005858585858595e-06,
      "loss": 1.0565,
      "step": 777500
    },
    {
      "epoch": 36.27,
      "learning_rate": 4.490484848484849e-06,
      "loss": 1.0579,
      "step": 778000
    },
    {
      "epoch": 36.29,
      "learning_rate": 4.48040404040404e-06,
      "loss": 1.0575,
      "step": 778500
    },
    {
      "epoch": 36.32,
      "learning_rate": 4.470303030303031e-06,
      "loss": 1.0569,
      "step": 779000
    },
    {
      "epoch": 36.34,
      "learning_rate": 4.460202020202021e-06,
      "loss": 1.0572,
      "step": 779500
    },
    {
      "epoch": 36.36,
      "learning_rate": 4.45010101010101e-06,
      "loss": 1.0574,
      "step": 780000
    },
    {
      "epoch": 36.39,
      "learning_rate": 4.440000000000001e-06,
      "loss": 1.0554,
      "step": 780500
    },
    {
      "epoch": 36.41,
      "learning_rate": 4.42989898989899e-06,
      "loss": 1.054,
      "step": 781000
    },
    {
      "epoch": 36.43,
      "learning_rate": 4.41979797979798e-06,
      "loss": 1.0564,
      "step": 781500
    },
    {
      "epoch": 36.46,
      "learning_rate": 4.40969696969697e-06,
      "loss": 1.0593,
      "step": 782000
    },
    {
      "epoch": 36.48,
      "learning_rate": 4.39959595959596e-06,
      "loss": 1.0573,
      "step": 782500
    },
    {
      "epoch": 36.5,
      "learning_rate": 4.3895151515151515e-06,
      "loss": 1.0589,
      "step": 783000
    },
    {
      "epoch": 36.53,
      "learning_rate": 4.379414141414142e-06,
      "loss": 1.0608,
      "step": 783500
    },
    {
      "epoch": 36.55,
      "learning_rate": 4.369313131313131e-06,
      "loss": 1.0582,
      "step": 784000
    },
    {
      "epoch": 36.57,
      "learning_rate": 4.3592121212121216e-06,
      "loss": 1.0573,
      "step": 784500
    },
    {
      "epoch": 36.6,
      "learning_rate": 4.349131313131314e-06,
      "loss": 1.0569,
      "step": 785000
    },
    {
      "epoch": 36.62,
      "learning_rate": 4.339030303030303e-06,
      "loss": 1.0589,
      "step": 785500
    },
    {
      "epoch": 36.64,
      "learning_rate": 4.328929292929293e-06,
      "loss": 1.0564,
      "step": 786000
    },
    {
      "epoch": 36.66,
      "learning_rate": 4.318828282828283e-06,
      "loss": 1.0554,
      "step": 786500
    },
    {
      "epoch": 36.69,
      "learning_rate": 4.308747474747475e-06,
      "loss": 1.0592,
      "step": 787000
    },
    {
      "epoch": 36.71,
      "learning_rate": 4.298646464646465e-06,
      "loss": 1.0543,
      "step": 787500
    },
    {
      "epoch": 36.73,
      "learning_rate": 4.288545454545455e-06,
      "loss": 1.057,
      "step": 788000
    },
    {
      "epoch": 36.76,
      "learning_rate": 4.278444444444445e-06,
      "loss": 1.0583,
      "step": 788500
    },
    {
      "epoch": 36.78,
      "learning_rate": 4.268343434343435e-06,
      "loss": 1.059,
      "step": 789000
    },
    {
      "epoch": 36.8,
      "learning_rate": 4.258242424242425e-06,
      "loss": 1.0569,
      "step": 789500
    },
    {
      "epoch": 36.83,
      "learning_rate": 4.248161616161616e-06,
      "loss": 1.0612,
      "step": 790000
    },
    {
      "epoch": 36.85,
      "learning_rate": 4.238060606060607e-06,
      "loss": 1.0581,
      "step": 790500
    },
    {
      "epoch": 36.87,
      "learning_rate": 4.227959595959596e-06,
      "loss": 1.0578,
      "step": 791000
    },
    {
      "epoch": 36.9,
      "learning_rate": 4.217858585858586e-06,
      "loss": 1.0566,
      "step": 791500
    },
    {
      "epoch": 36.92,
      "learning_rate": 4.207757575757576e-06,
      "loss": 1.0572,
      "step": 792000
    },
    {
      "epoch": 36.94,
      "learning_rate": 4.197676767676768e-06,
      "loss": 1.0579,
      "step": 792500
    },
    {
      "epoch": 36.97,
      "learning_rate": 4.187575757575758e-06,
      "loss": 1.0543,
      "step": 793000
    },
    {
      "epoch": 36.99,
      "learning_rate": 4.177474747474748e-06,
      "loss": 1.058,
      "step": 793500
    },
    {
      "epoch": 37.0,
      "eval_accuracy": 0.7769176344438438,
      "eval_loss": 0.9259941577911377,
      "eval_runtime": 52.0304,
      "eval_samples_per_second": 897.706,
      "eval_steps_per_second": 14.03,
      "step": 793687
    },
    {
      "epoch": 37.01,
      "learning_rate": 4.167373737373738e-06,
      "loss": 1.0546,
      "step": 794000
    },
    {
      "epoch": 37.04,
      "learning_rate": 4.1572929292929295e-06,
      "loss": 1.0565,
      "step": 794500
    },
    {
      "epoch": 37.06,
      "learning_rate": 4.147212121212122e-06,
      "loss": 1.0587,
      "step": 795000
    },
    {
      "epoch": 37.08,
      "learning_rate": 4.137111111111111e-06,
      "loss": 1.0576,
      "step": 795500
    },
    {
      "epoch": 37.11,
      "learning_rate": 4.1270101010101014e-06,
      "loss": 1.0536,
      "step": 796000
    },
    {
      "epoch": 37.13,
      "learning_rate": 4.116909090909092e-06,
      "loss": 1.0566,
      "step": 796500
    },
    {
      "epoch": 37.15,
      "learning_rate": 4.106808080808081e-06,
      "loss": 1.057,
      "step": 797000
    },
    {
      "epoch": 37.18,
      "learning_rate": 4.0967070707070715e-06,
      "loss": 1.0548,
      "step": 797500
    },
    {
      "epoch": 37.2,
      "learning_rate": 4.086606060606061e-06,
      "loss": 1.0556,
      "step": 798000
    },
    {
      "epoch": 37.22,
      "learning_rate": 4.07650505050505e-06,
      "loss": 1.0591,
      "step": 798500
    },
    {
      "epoch": 37.25,
      "learning_rate": 4.066424242424243e-06,
      "loss": 1.0588,
      "step": 799000
    },
    {
      "epoch": 37.27,
      "learning_rate": 4.056323232323233e-06,
      "loss": 1.0566,
      "step": 799500
    },
    {
      "epoch": 37.29,
      "learning_rate": 4.046242424242425e-06,
      "loss": 1.0549,
      "step": 800000
    },
    {
      "epoch": 37.32,
      "learning_rate": 4.0361414141414146e-06,
      "loss": 1.054,
      "step": 800500
    },
    {
      "epoch": 37.34,
      "learning_rate": 4.026040404040404e-06,
      "loss": 1.0546,
      "step": 801000
    },
    {
      "epoch": 37.36,
      "learning_rate": 4.015939393939394e-06,
      "loss": 1.0538,
      "step": 801500
    },
    {
      "epoch": 37.39,
      "learning_rate": 4.005838383838384e-06,
      "loss": 1.0563,
      "step": 802000
    },
    {
      "epoch": 37.41,
      "learning_rate": 3.995737373737374e-06,
      "loss": 1.0574,
      "step": 802500
    },
    {
      "epoch": 37.43,
      "learning_rate": 3.9856363636363635e-06,
      "loss": 1.0586,
      "step": 803000
    },
    {
      "epoch": 37.46,
      "learning_rate": 3.975535353535354e-06,
      "loss": 1.0552,
      "step": 803500
    },
    {
      "epoch": 37.48,
      "learning_rate": 3.965434343434344e-06,
      "loss": 1.0569,
      "step": 804000
    },
    {
      "epoch": 37.5,
      "learning_rate": 3.9553535353535354e-06,
      "loss": 1.0557,
      "step": 804500
    },
    {
      "epoch": 37.53,
      "learning_rate": 3.945252525252526e-06,
      "loss": 1.0559,
      "step": 805000
    },
    {
      "epoch": 37.55,
      "learning_rate": 3.935151515151516e-06,
      "loss": 1.0558,
      "step": 805500
    },
    {
      "epoch": 37.57,
      "learning_rate": 3.9250505050505054e-06,
      "loss": 1.0555,
      "step": 806000
    },
    {
      "epoch": 37.6,
      "learning_rate": 3.914949494949496e-06,
      "loss": 1.0566,
      "step": 806500
    },
    {
      "epoch": 37.62,
      "learning_rate": 3.904848484848485e-06,
      "loss": 1.0561,
      "step": 807000
    },
    {
      "epoch": 37.64,
      "learning_rate": 3.894767676767677e-06,
      "loss": 1.0562,
      "step": 807500
    },
    {
      "epoch": 37.67,
      "learning_rate": 3.884686868686869e-06,
      "loss": 1.0571,
      "step": 808000
    },
    {
      "epoch": 37.69,
      "learning_rate": 3.874585858585858e-06,
      "loss": 1.0566,
      "step": 808500
    },
    {
      "epoch": 37.71,
      "learning_rate": 3.864484848484849e-06,
      "loss": 1.054,
      "step": 809000
    },
    {
      "epoch": 37.74,
      "learning_rate": 3.854383838383839e-06,
      "loss": 1.0557,
      "step": 809500
    },
    {
      "epoch": 37.76,
      "learning_rate": 3.844282828282829e-06,
      "loss": 1.0557,
      "step": 810000
    },
    {
      "epoch": 37.78,
      "learning_rate": 3.8341818181818186e-06,
      "loss": 1.0569,
      "step": 810500
    },
    {
      "epoch": 37.81,
      "learning_rate": 3.824080808080808e-06,
      "loss": 1.0567,
      "step": 811000
    },
    {
      "epoch": 37.83,
      "learning_rate": 3.8139797979797983e-06,
      "loss": 1.0541,
      "step": 811500
    },
    {
      "epoch": 37.85,
      "learning_rate": 3.803878787878788e-06,
      "loss": 1.0569,
      "step": 812000
    },
    {
      "epoch": 37.88,
      "learning_rate": 3.793777777777778e-06,
      "loss": 1.0555,
      "step": 812500
    },
    {
      "epoch": 37.9,
      "learning_rate": 3.783676767676768e-06,
      "loss": 1.0562,
      "step": 813000
    },
    {
      "epoch": 37.92,
      "learning_rate": 3.773575757575758e-06,
      "loss": 1.0557,
      "step": 813500
    },
    {
      "epoch": 37.95,
      "learning_rate": 3.7634949494949496e-06,
      "loss": 1.0566,
      "step": 814000
    },
    {
      "epoch": 37.97,
      "learning_rate": 3.753434343434344e-06,
      "loss": 1.0568,
      "step": 814500
    },
    {
      "epoch": 37.99,
      "learning_rate": 3.7433333333333336e-06,
      "loss": 1.056,
      "step": 815000
    },
    {
      "epoch": 38.0,
      "eval_accuracy": 0.7771948308796734,
      "eval_loss": 0.9245167374610901,
      "eval_runtime": 51.5115,
      "eval_samples_per_second": 906.748,
      "eval_steps_per_second": 14.172,
      "step": 815138
    },
    {
      "epoch": 38.02,
      "learning_rate": 3.7332323232323235e-06,
      "loss": 1.0552,
      "step": 815500
    },
    {
      "epoch": 38.04,
      "learning_rate": 3.7231313131313134e-06,
      "loss": 1.0566,
      "step": 816000
    },
    {
      "epoch": 38.06,
      "learning_rate": 3.7130303030303032e-06,
      "loss": 1.0568,
      "step": 816500
    },
    {
      "epoch": 38.09,
      "learning_rate": 3.702929292929293e-06,
      "loss": 1.0555,
      "step": 817000
    },
    {
      "epoch": 38.11,
      "learning_rate": 3.692828282828283e-06,
      "loss": 1.0559,
      "step": 817500
    },
    {
      "epoch": 38.13,
      "learning_rate": 3.682727272727273e-06,
      "loss": 1.053,
      "step": 818000
    },
    {
      "epoch": 38.16,
      "learning_rate": 3.6726262626262627e-06,
      "loss": 1.0558,
      "step": 818500
    },
    {
      "epoch": 38.18,
      "learning_rate": 3.662525252525253e-06,
      "loss": 1.0543,
      "step": 819000
    },
    {
      "epoch": 38.2,
      "learning_rate": 3.6524444444444444e-06,
      "loss": 1.058,
      "step": 819500
    },
    {
      "epoch": 38.23,
      "learning_rate": 3.6423434343434342e-06,
      "loss": 1.0564,
      "step": 820000
    },
    {
      "epoch": 38.25,
      "learning_rate": 3.632242424242425e-06,
      "loss": 1.0557,
      "step": 820500
    },
    {
      "epoch": 38.27,
      "learning_rate": 3.622141414141415e-06,
      "loss": 1.0566,
      "step": 821000
    },
    {
      "epoch": 38.3,
      "learning_rate": 3.612060606060606e-06,
      "loss": 1.0561,
      "step": 821500
    },
    {
      "epoch": 38.32,
      "learning_rate": 3.601959595959596e-06,
      "loss": 1.0563,
      "step": 822000
    },
    {
      "epoch": 38.34,
      "learning_rate": 3.5918585858585864e-06,
      "loss": 1.0554,
      "step": 822500
    },
    {
      "epoch": 38.37,
      "learning_rate": 3.5817575757575762e-06,
      "loss": 1.0562,
      "step": 823000
    },
    {
      "epoch": 38.39,
      "learning_rate": 3.571656565656566e-06,
      "loss": 1.0551,
      "step": 823500
    },
    {
      "epoch": 38.41,
      "learning_rate": 3.561555555555556e-06,
      "loss": 1.0544,
      "step": 824000
    },
    {
      "epoch": 38.44,
      "learning_rate": 3.5514747474747478e-06,
      "loss": 1.0565,
      "step": 824500
    },
    {
      "epoch": 38.46,
      "learning_rate": 3.5413737373737376e-06,
      "loss": 1.0554,
      "step": 825000
    },
    {
      "epoch": 38.48,
      "learning_rate": 3.5312727272727275e-06,
      "loss": 1.0549,
      "step": 825500
    },
    {
      "epoch": 38.51,
      "learning_rate": 3.5211717171717174e-06,
      "loss": 1.0545,
      "step": 826000
    },
    {
      "epoch": 38.53,
      "learning_rate": 3.5110707070707072e-06,
      "loss": 1.0564,
      "step": 826500
    },
    {
      "epoch": 38.55,
      "learning_rate": 3.500969696969697e-06,
      "loss": 1.0569,
      "step": 827000
    },
    {
      "epoch": 38.58,
      "learning_rate": 3.490868686868687e-06,
      "loss": 1.0557,
      "step": 827500
    },
    {
      "epoch": 38.6,
      "learning_rate": 3.480767676767677e-06,
      "loss": 1.0535,
      "step": 828000
    },
    {
      "epoch": 38.62,
      "learning_rate": 3.470686868686869e-06,
      "loss": 1.0551,
      "step": 828500
    },
    {
      "epoch": 38.65,
      "learning_rate": 3.460585858585859e-06,
      "loss": 1.0548,
      "step": 829000
    },
    {
      "epoch": 38.67,
      "learning_rate": 3.4505050505050507e-06,
      "loss": 1.0548,
      "step": 829500
    },
    {
      "epoch": 38.69,
      "learning_rate": 3.4404040404040406e-06,
      "loss": 1.0567,
      "step": 830000
    },
    {
      "epoch": 38.72,
      "learning_rate": 3.4303030303030305e-06,
      "loss": 1.0544,
      "step": 830500
    },
    {
      "epoch": 38.74,
      "learning_rate": 3.4202020202020203e-06,
      "loss": 1.0537,
      "step": 831000
    },
    {
      "epoch": 38.76,
      "learning_rate": 3.4101212121212126e-06,
      "loss": 1.0543,
      "step": 831500
    },
    {
      "epoch": 38.79,
      "learning_rate": 3.400020202020202e-06,
      "loss": 1.0531,
      "step": 832000
    },
    {
      "epoch": 38.81,
      "learning_rate": 3.389919191919192e-06,
      "loss": 1.0566,
      "step": 832500
    },
    {
      "epoch": 38.83,
      "learning_rate": 3.3798181818181817e-06,
      "loss": 1.0547,
      "step": 833000
    },
    {
      "epoch": 38.86,
      "learning_rate": 3.3697171717171716e-06,
      "loss": 1.0557,
      "step": 833500
    },
    {
      "epoch": 38.88,
      "learning_rate": 3.3596161616161623e-06,
      "loss": 1.0565,
      "step": 834000
    },
    {
      "epoch": 38.9,
      "learning_rate": 3.3495151515151518e-06,
      "loss": 1.0536,
      "step": 834500
    },
    {
      "epoch": 38.93,
      "learning_rate": 3.3394141414141416e-06,
      "loss": 1.0534,
      "step": 835000
    },
    {
      "epoch": 38.95,
      "learning_rate": 3.3293535353535354e-06,
      "loss": 1.0537,
      "step": 835500
    },
    {
      "epoch": 38.97,
      "learning_rate": 3.3192525252525253e-06,
      "loss": 1.0561,
      "step": 836000
    },
    {
      "epoch": 39.0,
      "learning_rate": 3.309151515151515e-06,
      "loss": 1.0531,
      "step": 836500
    },
    {
      "epoch": 39.0,
      "eval_accuracy": 0.777148177284278,
      "eval_loss": 0.9247538447380066,
      "eval_runtime": 51.8853,
      "eval_samples_per_second": 900.216,
      "eval_steps_per_second": 14.069,
      "step": 836589
    },
    {
      "epoch": 39.02,
      "learning_rate": 3.299050505050505e-06,
      "loss": 1.0549,
      "step": 837000
    },
    {
      "epoch": 39.04,
      "learning_rate": 3.2889494949494953e-06,
      "loss": 1.054,
      "step": 837500
    },
    {
      "epoch": 39.07,
      "learning_rate": 3.278848484848485e-06,
      "loss": 1.0564,
      "step": 838000
    },
    {
      "epoch": 39.09,
      "learning_rate": 3.268747474747475e-06,
      "loss": 1.0541,
      "step": 838500
    },
    {
      "epoch": 39.11,
      "learning_rate": 3.258646464646465e-06,
      "loss": 1.0528,
      "step": 839000
    },
    {
      "epoch": 39.14,
      "learning_rate": 3.2485454545454547e-06,
      "loss": 1.0545,
      "step": 839500
    },
    {
      "epoch": 39.16,
      "learning_rate": 3.2384444444444446e-06,
      "loss": 1.0534,
      "step": 840000
    },
    {
      "epoch": 39.18,
      "learning_rate": 3.2283434343434345e-06,
      "loss": 1.0537,
      "step": 840500
    },
    {
      "epoch": 39.21,
      "learning_rate": 3.2182424242424243e-06,
      "loss": 1.0554,
      "step": 841000
    },
    {
      "epoch": 39.23,
      "learning_rate": 3.2081414141414146e-06,
      "loss": 1.0547,
      "step": 841500
    },
    {
      "epoch": 39.25,
      "learning_rate": 3.198060606060606e-06,
      "loss": 1.0558,
      "step": 842000
    },
    {
      "epoch": 39.28,
      "learning_rate": 3.187959595959596e-06,
      "loss": 1.0535,
      "step": 842500
    },
    {
      "epoch": 39.3,
      "learning_rate": 3.177878787878788e-06,
      "loss": 1.056,
      "step": 843000
    },
    {
      "epoch": 39.32,
      "learning_rate": 3.167777777777778e-06,
      "loss": 1.0539,
      "step": 843500
    },
    {
      "epoch": 39.35,
      "learning_rate": 3.1576969696969702e-06,
      "loss": 1.0559,
      "step": 844000
    },
    {
      "epoch": 39.37,
      "learning_rate": 3.1475959595959597e-06,
      "loss": 1.056,
      "step": 844500
    },
    {
      "epoch": 39.39,
      "learning_rate": 3.1374949494949495e-06,
      "loss": 1.0553,
      "step": 845000
    },
    {
      "epoch": 39.42,
      "learning_rate": 3.1273939393939394e-06,
      "loss": 1.0539,
      "step": 845500
    },
    {
      "epoch": 39.44,
      "learning_rate": 3.1172929292929293e-06,
      "loss": 1.0546,
      "step": 846000
    },
    {
      "epoch": 39.46,
      "learning_rate": 3.107191919191919e-06,
      "loss": 1.0561,
      "step": 846500
    },
    {
      "epoch": 39.49,
      "learning_rate": 3.0970909090909094e-06,
      "loss": 1.0545,
      "step": 847000
    },
    {
      "epoch": 39.51,
      "learning_rate": 3.0869898989898993e-06,
      "loss": 1.0531,
      "step": 847500
    },
    {
      "epoch": 39.53,
      "learning_rate": 3.076909090909091e-06,
      "loss": 1.0543,
      "step": 848000
    },
    {
      "epoch": 39.56,
      "learning_rate": 3.066808080808081e-06,
      "loss": 1.0539,
      "step": 848500
    },
    {
      "epoch": 39.58,
      "learning_rate": 3.0567070707070713e-06,
      "loss": 1.0524,
      "step": 849000
    },
    {
      "epoch": 39.6,
      "learning_rate": 3.046606060606061e-06,
      "loss": 1.0571,
      "step": 849500
    },
    {
      "epoch": 39.63,
      "learning_rate": 3.036505050505051e-06,
      "loss": 1.0537,
      "step": 850000
    },
    {
      "epoch": 39.65,
      "learning_rate": 3.026404040404041e-06,
      "loss": 1.0545,
      "step": 850500
    },
    {
      "epoch": 39.67,
      "learning_rate": 3.0163030303030307e-06,
      "loss": 1.0547,
      "step": 851000
    },
    {
      "epoch": 39.7,
      "learning_rate": 3.00620202020202e-06,
      "loss": 1.0543,
      "step": 851500
    },
    {
      "epoch": 39.72,
      "learning_rate": 2.99610101010101e-06,
      "loss": 1.0551,
      "step": 852000
    },
    {
      "epoch": 39.74,
      "learning_rate": 2.9860404040404046e-06,
      "loss": 1.053,
      "step": 852500
    },
    {
      "epoch": 39.77,
      "learning_rate": 2.9759393939393945e-06,
      "loss": 1.0533,
      "step": 853000
    },
    {
      "epoch": 39.79,
      "learning_rate": 2.9658383838383844e-06,
      "loss": 1.0528,
      "step": 853500
    },
    {
      "epoch": 39.81,
      "learning_rate": 2.955737373737374e-06,
      "loss": 1.0552,
      "step": 854000
    },
    {
      "epoch": 39.83,
      "learning_rate": 2.9456363636363637e-06,
      "loss": 1.0507,
      "step": 854500
    },
    {
      "epoch": 39.86,
      "learning_rate": 2.9355353535353535e-06,
      "loss": 1.0534,
      "step": 855000
    },
    {
      "epoch": 39.88,
      "learning_rate": 2.9254343434343434e-06,
      "loss": 1.0518,
      "step": 855500
    },
    {
      "epoch": 39.9,
      "learning_rate": 2.9153333333333333e-06,
      "loss": 1.0555,
      "step": 856000
    },
    {
      "epoch": 39.93,
      "learning_rate": 2.9052525252525255e-06,
      "loss": 1.0551,
      "step": 856500
    },
    {
      "epoch": 39.95,
      "learning_rate": 2.8951515151515154e-06,
      "loss": 1.0567,
      "step": 857000
    },
    {
      "epoch": 39.97,
      "learning_rate": 2.8850505050505052e-06,
      "loss": 1.0522,
      "step": 857500
    },
    {
      "epoch": 40.0,
      "learning_rate": 2.874949494949495e-06,
      "loss": 1.0548,
      "step": 858000
    },
    {
      "epoch": 40.0,
      "eval_accuracy": 0.7775596755808146,
      "eval_loss": 0.9227122068405151,
      "eval_runtime": 51.8063,
      "eval_samples_per_second": 901.588,
      "eval_steps_per_second": 14.091,
      "step": 858040
    },
    {
      "epoch": 40.02,
      "learning_rate": 2.8648484848484854e-06,
      "loss": 1.055,
      "step": 858500
    },
    {
      "epoch": 40.04,
      "learning_rate": 2.8547474747474752e-06,
      "loss": 1.052,
      "step": 859000
    },
    {
      "epoch": 40.07,
      "learning_rate": 2.8446666666666666e-06,
      "loss": 1.053,
      "step": 859500
    },
    {
      "epoch": 40.09,
      "learning_rate": 2.8345656565656565e-06,
      "loss": 1.0515,
      "step": 860000
    },
    {
      "epoch": 40.11,
      "learning_rate": 2.824464646464647e-06,
      "loss": 1.0554,
      "step": 860500
    },
    {
      "epoch": 40.14,
      "learning_rate": 2.8143636363636367e-06,
      "loss": 1.0546,
      "step": 861000
    },
    {
      "epoch": 40.16,
      "learning_rate": 2.804282828282828e-06,
      "loss": 1.0546,
      "step": 861500
    },
    {
      "epoch": 40.18,
      "learning_rate": 2.7941818181818188e-06,
      "loss": 1.0529,
      "step": 862000
    },
    {
      "epoch": 40.21,
      "learning_rate": 2.7840808080808086e-06,
      "loss": 1.0541,
      "step": 862500
    },
    {
      "epoch": 40.23,
      "learning_rate": 2.7740202020202024e-06,
      "loss": 1.0534,
      "step": 863000
    },
    {
      "epoch": 40.25,
      "learning_rate": 2.7639191919191923e-06,
      "loss": 1.0538,
      "step": 863500
    },
    {
      "epoch": 40.28,
      "learning_rate": 2.753818181818182e-06,
      "loss": 1.0541,
      "step": 864000
    },
    {
      "epoch": 40.3,
      "learning_rate": 2.7437171717171716e-06,
      "loss": 1.0537,
      "step": 864500
    },
    {
      "epoch": 40.32,
      "learning_rate": 2.7336161616161615e-06,
      "loss": 1.0534,
      "step": 865000
    },
    {
      "epoch": 40.35,
      "learning_rate": 2.723515151515152e-06,
      "loss": 1.0536,
      "step": 865500
    },
    {
      "epoch": 40.37,
      "learning_rate": 2.713414141414142e-06,
      "loss": 1.054,
      "step": 866000
    },
    {
      "epoch": 40.39,
      "learning_rate": 2.7033131313131315e-06,
      "loss": 1.0518,
      "step": 866500
    },
    {
      "epoch": 40.42,
      "learning_rate": 2.6932121212121213e-06,
      "loss": 1.0525,
      "step": 867000
    },
    {
      "epoch": 40.44,
      "learning_rate": 2.683111111111111e-06,
      "loss": 1.0556,
      "step": 867500
    },
    {
      "epoch": 40.46,
      "learning_rate": 2.673010101010101e-06,
      "loss": 1.0547,
      "step": 868000
    },
    {
      "epoch": 40.49,
      "learning_rate": 2.662909090909091e-06,
      "loss": 1.0535,
      "step": 868500
    },
    {
      "epoch": 40.51,
      "learning_rate": 2.6528080808080808e-06,
      "loss": 1.0514,
      "step": 869000
    },
    {
      "epoch": 40.53,
      "learning_rate": 2.6427070707070706e-06,
      "loss": 1.0535,
      "step": 869500
    },
    {
      "epoch": 40.56,
      "learning_rate": 2.632606060606061e-06,
      "loss": 1.0525,
      "step": 870000
    },
    {
      "epoch": 40.58,
      "learning_rate": 2.6225454545454547e-06,
      "loss": 1.0549,
      "step": 870500
    },
    {
      "epoch": 40.6,
      "learning_rate": 2.6124444444444446e-06,
      "loss": 1.0543,
      "step": 871000
    },
    {
      "epoch": 40.63,
      "learning_rate": 2.6023434343434344e-06,
      "loss": 1.0541,
      "step": 871500
    },
    {
      "epoch": 40.65,
      "learning_rate": 2.5922424242424243e-06,
      "loss": 1.0567,
      "step": 872000
    },
    {
      "epoch": 40.67,
      "learning_rate": 2.582141414141414e-06,
      "loss": 1.0542,
      "step": 872500
    },
    {
      "epoch": 40.7,
      "learning_rate": 2.572040404040404e-06,
      "loss": 1.0535,
      "step": 873000
    },
    {
      "epoch": 40.72,
      "learning_rate": 2.5619393939393943e-06,
      "loss": 1.0551,
      "step": 873500
    },
    {
      "epoch": 40.74,
      "learning_rate": 2.551838383838384e-06,
      "loss": 1.0527,
      "step": 874000
    },
    {
      "epoch": 40.77,
      "learning_rate": 2.541737373737374e-06,
      "loss": 1.0532,
      "step": 874500
    },
    {
      "epoch": 40.79,
      "learning_rate": 2.5316565656565654e-06,
      "loss": 1.0529,
      "step": 875000
    },
    {
      "epoch": 40.81,
      "learning_rate": 2.521555555555556e-06,
      "loss": 1.0536,
      "step": 875500
    },
    {
      "epoch": 40.84,
      "learning_rate": 2.511454545454546e-06,
      "loss": 1.0537,
      "step": 876000
    },
    {
      "epoch": 40.86,
      "learning_rate": 2.5013535353535355e-06,
      "loss": 1.0544,
      "step": 876500
    },
    {
      "epoch": 40.88,
      "learning_rate": 2.4912727272727273e-06,
      "loss": 1.0535,
      "step": 877000
    },
    {
      "epoch": 40.91,
      "learning_rate": 2.481191919191919e-06,
      "loss": 1.0529,
      "step": 877500
    },
    {
      "epoch": 40.93,
      "learning_rate": 2.4710909090909094e-06,
      "loss": 1.0524,
      "step": 878000
    },
    {
      "epoch": 40.95,
      "learning_rate": 2.4610101010101012e-06,
      "loss": 1.052,
      "step": 878500
    },
    {
      "epoch": 40.98,
      "learning_rate": 2.450909090909091e-06,
      "loss": 1.0504,
      "step": 879000
    },
    {
      "epoch": 41.0,
      "eval_accuracy": 0.7770689971307361,
      "eval_loss": 0.923800528049469,
      "eval_runtime": 53.0209,
      "eval_samples_per_second": 880.936,
      "eval_steps_per_second": 13.768,
      "step": 879491
    },
    {
      "epoch": 41.0,
      "learning_rate": 2.440808080808081e-06,
      "loss": 1.0506,
      "step": 879500
    },
    {
      "epoch": 41.02,
      "learning_rate": 2.430707070707071e-06,
      "loss": 1.0517,
      "step": 880000
    },
    {
      "epoch": 41.05,
      "learning_rate": 2.4206060606060607e-06,
      "loss": 1.0525,
      "step": 880500
    },
    {
      "epoch": 41.07,
      "learning_rate": 2.4105050505050505e-06,
      "loss": 1.0533,
      "step": 881000
    },
    {
      "epoch": 41.09,
      "learning_rate": 2.400404040404041e-06,
      "loss": 1.0498,
      "step": 881500
    },
    {
      "epoch": 41.12,
      "learning_rate": 2.3903030303030307e-06,
      "loss": 1.0525,
      "step": 882000
    },
    {
      "epoch": 41.14,
      "learning_rate": 2.3802020202020205e-06,
      "loss": 1.0535,
      "step": 882500
    },
    {
      "epoch": 41.16,
      "learning_rate": 2.3701212121212124e-06,
      "loss": 1.0518,
      "step": 883000
    },
    {
      "epoch": 41.19,
      "learning_rate": 2.3600202020202022e-06,
      "loss": 1.0538,
      "step": 883500
    },
    {
      "epoch": 41.21,
      "learning_rate": 2.349919191919192e-06,
      "loss": 1.0524,
      "step": 884000
    },
    {
      "epoch": 41.23,
      "learning_rate": 2.339818181818182e-06,
      "loss": 1.0531,
      "step": 884500
    },
    {
      "epoch": 41.26,
      "learning_rate": 2.329717171717172e-06,
      "loss": 1.0506,
      "step": 885000
    },
    {
      "epoch": 41.28,
      "learning_rate": 2.319636363636364e-06,
      "loss": 1.0551,
      "step": 885500
    },
    {
      "epoch": 41.3,
      "learning_rate": 2.309535353535354e-06,
      "loss": 1.0517,
      "step": 886000
    },
    {
      "epoch": 41.33,
      "learning_rate": 2.2994343434343434e-06,
      "loss": 1.0519,
      "step": 886500
    },
    {
      "epoch": 41.35,
      "learning_rate": 2.2893333333333332e-06,
      "loss": 1.0558,
      "step": 887000
    },
    {
      "epoch": 41.37,
      "learning_rate": 2.2792323232323235e-06,
      "loss": 1.0541,
      "step": 887500
    },
    {
      "epoch": 41.4,
      "learning_rate": 2.2691313131313134e-06,
      "loss": 1.0513,
      "step": 888000
    },
    {
      "epoch": 41.42,
      "learning_rate": 2.2590303030303033e-06,
      "loss": 1.0521,
      "step": 888500
    },
    {
      "epoch": 41.44,
      "learning_rate": 2.248949494949495e-06,
      "loss": 1.0536,
      "step": 889000
    },
    {
      "epoch": 41.47,
      "learning_rate": 2.238848484848485e-06,
      "loss": 1.0525,
      "step": 889500
    },
    {
      "epoch": 41.49,
      "learning_rate": 2.228747474747475e-06,
      "loss": 1.0526,
      "step": 890000
    },
    {
      "epoch": 41.51,
      "learning_rate": 2.2186464646464647e-06,
      "loss": 1.0553,
      "step": 890500
    },
    {
      "epoch": 41.54,
      "learning_rate": 2.2085454545454545e-06,
      "loss": 1.0519,
      "step": 891000
    },
    {
      "epoch": 41.56,
      "learning_rate": 2.198444444444445e-06,
      "loss": 1.0524,
      "step": 891500
    },
    {
      "epoch": 41.58,
      "learning_rate": 2.1883434343434347e-06,
      "loss": 1.0522,
      "step": 892000
    },
    {
      "epoch": 41.61,
      "learning_rate": 2.1782626262626265e-06,
      "loss": 1.0539,
      "step": 892500
    },
    {
      "epoch": 41.63,
      "learning_rate": 2.1681616161616164e-06,
      "loss": 1.0507,
      "step": 893000
    },
    {
      "epoch": 41.65,
      "learning_rate": 2.1580606060606062e-06,
      "loss": 1.0545,
      "step": 893500
    },
    {
      "epoch": 41.68,
      "learning_rate": 2.147959595959596e-06,
      "loss": 1.0524,
      "step": 894000
    },
    {
      "epoch": 41.7,
      "learning_rate": 2.137858585858586e-06,
      "loss": 1.0545,
      "step": 894500
    },
    {
      "epoch": 41.72,
      "learning_rate": 2.127757575757576e-06,
      "loss": 1.0534,
      "step": 895000
    },
    {
      "epoch": 41.75,
      "learning_rate": 2.117676767676768e-06,
      "loss": 1.0538,
      "step": 895500
    },
    {
      "epoch": 41.77,
      "learning_rate": 2.1075757575757575e-06,
      "loss": 1.0527,
      "step": 896000
    },
    {
      "epoch": 41.79,
      "learning_rate": 2.0974747474747474e-06,
      "loss": 1.0526,
      "step": 896500
    },
    {
      "epoch": 41.82,
      "learning_rate": 2.0873737373737377e-06,
      "loss": 1.0524,
      "step": 897000
    },
    {
      "epoch": 41.84,
      "learning_rate": 2.0772727272727275e-06,
      "loss": 1.0536,
      "step": 897500
    },
    {
      "epoch": 41.86,
      "learning_rate": 2.0671717171717174e-06,
      "loss": 1.0537,
      "step": 898000
    },
    {
      "epoch": 41.89,
      "learning_rate": 2.0570707070707072e-06,
      "loss": 1.0518,
      "step": 898500
    },
    {
      "epoch": 41.91,
      "learning_rate": 2.046969696969697e-06,
      "loss": 1.0517,
      "step": 899000
    },
    {
      "epoch": 41.93,
      "learning_rate": 2.036888888888889e-06,
      "loss": 1.0544,
      "step": 899500
    },
    {
      "epoch": 41.96,
      "learning_rate": 2.026787878787879e-06,
      "loss": 1.0518,
      "step": 900000
    },
    {
      "epoch": 41.98,
      "learning_rate": 2.016707070707071e-06,
      "loss": 1.0524,
      "step": 900500
    },
    {
      "epoch": 42.0,
      "eval_accuracy": 0.7778208287599419,
      "eval_loss": 0.9227592945098877,
      "eval_runtime": 52.9756,
      "eval_samples_per_second": 881.688,
      "eval_steps_per_second": 13.78,
      "step": 900942
    },
    {
      "epoch": 42.0,
      "learning_rate": 2.006606060606061e-06,
      "loss": 1.0538,
      "step": 901000
    },
    {
      "epoch": 42.03,
      "learning_rate": 1.9965050505050508e-06,
      "loss": 1.0554,
      "step": 901500
    },
    {
      "epoch": 42.05,
      "learning_rate": 1.9864040404040406e-06,
      "loss": 1.0539,
      "step": 902000
    },
    {
      "epoch": 42.07,
      "learning_rate": 1.9763030303030305e-06,
      "loss": 1.0509,
      "step": 902500
    },
    {
      "epoch": 42.1,
      "learning_rate": 1.9662222222222223e-06,
      "loss": 1.0529,
      "step": 903000
    },
    {
      "epoch": 42.12,
      "learning_rate": 1.956121212121212e-06,
      "loss": 1.0517,
      "step": 903500
    },
    {
      "epoch": 42.14,
      "learning_rate": 1.946020202020202e-06,
      "loss": 1.0515,
      "step": 904000
    },
    {
      "epoch": 42.17,
      "learning_rate": 1.9359191919191923e-06,
      "loss": 1.0525,
      "step": 904500
    },
    {
      "epoch": 42.19,
      "learning_rate": 1.925818181818182e-06,
      "loss": 1.0527,
      "step": 905000
    },
    {
      "epoch": 42.21,
      "learning_rate": 1.915717171717172e-06,
      "loss": 1.0517,
      "step": 905500
    },
    {
      "epoch": 42.24,
      "learning_rate": 1.9056161616161617e-06,
      "loss": 1.0522,
      "step": 906000
    },
    {
      "epoch": 42.26,
      "learning_rate": 1.8955151515151518e-06,
      "loss": 1.0522,
      "step": 906500
    },
    {
      "epoch": 42.28,
      "learning_rate": 1.8854343434343436e-06,
      "loss": 1.0534,
      "step": 907000
    },
    {
      "epoch": 42.31,
      "learning_rate": 1.8753333333333335e-06,
      "loss": 1.0521,
      "step": 907500
    },
    {
      "epoch": 42.33,
      "learning_rate": 1.8652525252525255e-06,
      "loss": 1.0523,
      "step": 908000
    },
    {
      "epoch": 42.35,
      "learning_rate": 1.8551515151515154e-06,
      "loss": 1.0523,
      "step": 908500
    },
    {
      "epoch": 42.38,
      "learning_rate": 1.8450505050505052e-06,
      "loss": 1.0533,
      "step": 909000
    },
    {
      "epoch": 42.4,
      "learning_rate": 1.834949494949495e-06,
      "loss": 1.0531,
      "step": 909500
    },
    {
      "epoch": 42.42,
      "learning_rate": 1.824848484848485e-06,
      "loss": 1.0522,
      "step": 910000
    },
    {
      "epoch": 42.45,
      "learning_rate": 1.814747474747475e-06,
      "loss": 1.0521,
      "step": 910500
    },
    {
      "epoch": 42.47,
      "learning_rate": 1.804646464646465e-06,
      "loss": 1.0534,
      "step": 911000
    },
    {
      "epoch": 42.49,
      "learning_rate": 1.7945454545454546e-06,
      "loss": 1.0522,
      "step": 911500
    },
    {
      "epoch": 42.52,
      "learning_rate": 1.7844444444444444e-06,
      "loss": 1.0532,
      "step": 912000
    },
    {
      "epoch": 42.54,
      "learning_rate": 1.7743838383838386e-06,
      "loss": 1.0524,
      "step": 912500
    },
    {
      "epoch": 42.56,
      "learning_rate": 1.7642828282828283e-06,
      "loss": 1.0501,
      "step": 913000
    },
    {
      "epoch": 42.59,
      "learning_rate": 1.7541818181818181e-06,
      "loss": 1.0522,
      "step": 913500
    },
    {
      "epoch": 42.61,
      "learning_rate": 1.7440808080808082e-06,
      "loss": 1.053,
      "step": 914000
    },
    {
      "epoch": 42.63,
      "learning_rate": 1.733979797979798e-06,
      "loss": 1.0513,
      "step": 914500
    },
    {
      "epoch": 42.66,
      "learning_rate": 1.72389898989899e-06,
      "loss": 1.0517,
      "step": 915000
    },
    {
      "epoch": 42.68,
      "learning_rate": 1.71379797979798e-06,
      "loss": 1.0497,
      "step": 915500
    },
    {
      "epoch": 42.7,
      "learning_rate": 1.7036969696969698e-06,
      "loss": 1.0513,
      "step": 916000
    },
    {
      "epoch": 42.73,
      "learning_rate": 1.6935959595959597e-06,
      "loss": 1.0525,
      "step": 916500
    },
    {
      "epoch": 42.75,
      "learning_rate": 1.6834949494949496e-06,
      "loss": 1.0511,
      "step": 917000
    },
    {
      "epoch": 42.77,
      "learning_rate": 1.6733939393939394e-06,
      "loss": 1.0519,
      "step": 917500
    },
    {
      "epoch": 42.8,
      "learning_rate": 1.6632929292929295e-06,
      "loss": 1.0507,
      "step": 918000
    },
    {
      "epoch": 42.82,
      "learning_rate": 1.6531919191919194e-06,
      "loss": 1.0532,
      "step": 918500
    },
    {
      "epoch": 42.84,
      "learning_rate": 1.6431111111111112e-06,
      "loss": 1.0518,
      "step": 919000
    },
    {
      "epoch": 42.87,
      "learning_rate": 1.6330101010101013e-06,
      "loss": 1.0524,
      "step": 919500
    },
    {
      "epoch": 42.89,
      "learning_rate": 1.622929292929293e-06,
      "loss": 1.0517,
      "step": 920000
    },
    {
      "epoch": 42.91,
      "learning_rate": 1.612828282828283e-06,
      "loss": 1.0505,
      "step": 920500
    },
    {
      "epoch": 42.94,
      "learning_rate": 1.6027272727272728e-06,
      "loss": 1.0518,
      "step": 921000
    },
    {
      "epoch": 42.96,
      "learning_rate": 1.592626262626263e-06,
      "loss": 1.0515,
      "step": 921500
    },
    {
      "epoch": 42.98,
      "learning_rate": 1.5825454545454547e-06,
      "loss": 1.0507,
      "step": 922000
    },
    {
      "epoch": 43.0,
      "eval_accuracy": 0.7778332572233402,
      "eval_loss": 0.9218084216117859,
      "eval_runtime": 53.348,
      "eval_samples_per_second": 875.533,
      "eval_steps_per_second": 13.684,
      "step": 922393
    },
    {
      "epoch": 43.0,
      "learning_rate": 1.5724444444444446e-06,
      "loss": 1.0526,
      "step": 922500
    },
    {
      "epoch": 43.03,
      "learning_rate": 1.5623434343434347e-06,
      "loss": 1.053,
      "step": 923000
    },
    {
      "epoch": 43.05,
      "learning_rate": 1.5522424242424245e-06,
      "loss": 1.0543,
      "step": 923500
    },
    {
      "epoch": 43.07,
      "learning_rate": 1.5421414141414142e-06,
      "loss": 1.0523,
      "step": 924000
    },
    {
      "epoch": 43.1,
      "learning_rate": 1.532040404040404e-06,
      "loss": 1.0543,
      "step": 924500
    },
    {
      "epoch": 43.12,
      "learning_rate": 1.521939393939394e-06,
      "loss": 1.0528,
      "step": 925000
    },
    {
      "epoch": 43.14,
      "learning_rate": 1.511838383838384e-06,
      "loss": 1.0511,
      "step": 925500
    },
    {
      "epoch": 43.17,
      "learning_rate": 1.5017373737373738e-06,
      "loss": 1.0521,
      "step": 926000
    },
    {
      "epoch": 43.19,
      "learning_rate": 1.4916363636363637e-06,
      "loss": 1.0523,
      "step": 926500
    },
    {
      "epoch": 43.21,
      "learning_rate": 1.4815353535353536e-06,
      "loss": 1.0508,
      "step": 927000
    },
    {
      "epoch": 43.24,
      "learning_rate": 1.4714343434343436e-06,
      "loss": 1.048,
      "step": 927500
    },
    {
      "epoch": 43.26,
      "learning_rate": 1.4613535353535355e-06,
      "loss": 1.0505,
      "step": 928000
    },
    {
      "epoch": 43.28,
      "learning_rate": 1.4512525252525253e-06,
      "loss": 1.0509,
      "step": 928500
    },
    {
      "epoch": 43.31,
      "learning_rate": 1.4411717171717174e-06,
      "loss": 1.0496,
      "step": 929000
    },
    {
      "epoch": 43.33,
      "learning_rate": 1.4310707070707072e-06,
      "loss": 1.0516,
      "step": 929500
    },
    {
      "epoch": 43.35,
      "learning_rate": 1.420969696969697e-06,
      "loss": 1.0497,
      "step": 930000
    },
    {
      "epoch": 43.38,
      "learning_rate": 1.410868686868687e-06,
      "loss": 1.0499,
      "step": 930500
    },
    {
      "epoch": 43.4,
      "learning_rate": 1.400787878787879e-06,
      "loss": 1.0517,
      "step": 931000
    },
    {
      "epoch": 43.42,
      "learning_rate": 1.3906868686868689e-06,
      "loss": 1.0515,
      "step": 931500
    },
    {
      "epoch": 43.45,
      "learning_rate": 1.3805858585858587e-06,
      "loss": 1.0497,
      "step": 932000
    },
    {
      "epoch": 43.47,
      "learning_rate": 1.3704848484848486e-06,
      "loss": 1.0522,
      "step": 932500
    },
    {
      "epoch": 43.49,
      "learning_rate": 1.3603838383838387e-06,
      "loss": 1.0537,
      "step": 933000
    },
    {
      "epoch": 43.52,
      "learning_rate": 1.3502828282828285e-06,
      "loss": 1.0518,
      "step": 933500
    },
    {
      "epoch": 43.54,
      "learning_rate": 1.3401818181818182e-06,
      "loss": 1.0517,
      "step": 934000
    },
    {
      "epoch": 43.56,
      "learning_rate": 1.330080808080808e-06,
      "loss": 1.0526,
      "step": 934500
    },
    {
      "epoch": 43.59,
      "learning_rate": 1.319979797979798e-06,
      "loss": 1.0506,
      "step": 935000
    },
    {
      "epoch": 43.61,
      "learning_rate": 1.309878787878788e-06,
      "loss": 1.0493,
      "step": 935500
    },
    {
      "epoch": 43.63,
      "learning_rate": 1.2997979797979798e-06,
      "loss": 1.0526,
      "step": 936000
    },
    {
      "epoch": 43.66,
      "learning_rate": 1.2896969696969697e-06,
      "loss": 1.0489,
      "step": 936500
    },
    {
      "epoch": 43.68,
      "learning_rate": 1.2795959595959597e-06,
      "loss": 1.0528,
      "step": 937000
    },
    {
      "epoch": 43.7,
      "learning_rate": 1.2694949494949496e-06,
      "loss": 1.0516,
      "step": 937500
    },
    {
      "epoch": 43.73,
      "learning_rate": 1.2594343434343436e-06,
      "loss": 1.0501,
      "step": 938000
    },
    {
      "epoch": 43.75,
      "learning_rate": 1.2493333333333335e-06,
      "loss": 1.0501,
      "step": 938500
    },
    {
      "epoch": 43.77,
      "learning_rate": 1.2392323232323233e-06,
      "loss": 1.0495,
      "step": 939000
    },
    {
      "epoch": 43.8,
      "learning_rate": 1.2291313131313132e-06,
      "loss": 1.0528,
      "step": 939500
    },
    {
      "epoch": 43.82,
      "learning_rate": 1.219030303030303e-06,
      "loss": 1.0477,
      "step": 940000
    },
    {
      "epoch": 43.84,
      "learning_rate": 1.208929292929293e-06,
      "loss": 1.0499,
      "step": 940500
    },
    {
      "epoch": 43.87,
      "learning_rate": 1.198848484848485e-06,
      "loss": 1.0534,
      "step": 941000
    },
    {
      "epoch": 43.89,
      "learning_rate": 1.1887474747474748e-06,
      "loss": 1.0507,
      "step": 941500
    },
    {
      "epoch": 43.91,
      "learning_rate": 1.1786464646464647e-06,
      "loss": 1.0482,
      "step": 942000
    },
    {
      "epoch": 43.94,
      "learning_rate": 1.1685454545454547e-06,
      "loss": 1.0484,
      "step": 942500
    },
    {
      "epoch": 43.96,
      "learning_rate": 1.1584444444444446e-06,
      "loss": 1.048,
      "step": 943000
    },
    {
      "epoch": 43.98,
      "learning_rate": 1.1483434343434345e-06,
      "loss": 1.0495,
      "step": 943500
    },
    {
      "epoch": 44.0,
      "eval_accuracy": 0.7776427747542983,
      "eval_loss": 0.9226658344268799,
      "eval_runtime": 52.4909,
      "eval_samples_per_second": 889.83,
      "eval_steps_per_second": 13.907,
      "step": 943844
    },
    {
      "epoch": 44.01,
      "learning_rate": 1.1382424242424243e-06,
      "loss": 1.0515,
      "step": 944000
    },
    {
      "epoch": 44.03,
      "learning_rate": 1.1281414141414142e-06,
      "loss": 1.0501,
      "step": 944500
    },
    {
      "epoch": 44.05,
      "learning_rate": 1.118040404040404e-06,
      "loss": 1.0534,
      "step": 945000
    },
    {
      "epoch": 44.08,
      "learning_rate": 1.107979797979798e-06,
      "loss": 1.0491,
      "step": 945500
    },
    {
      "epoch": 44.1,
      "learning_rate": 1.097878787878788e-06,
      "loss": 1.0513,
      "step": 946000
    },
    {
      "epoch": 44.12,
      "learning_rate": 1.0877777777777778e-06,
      "loss": 1.0497,
      "step": 946500
    },
    {
      "epoch": 44.15,
      "learning_rate": 1.0776767676767676e-06,
      "loss": 1.0504,
      "step": 947000
    },
    {
      "epoch": 44.17,
      "learning_rate": 1.0675757575757577e-06,
      "loss": 1.0508,
      "step": 947500
    },
    {
      "epoch": 44.19,
      "learning_rate": 1.0574747474747476e-06,
      "loss": 1.0507,
      "step": 948000
    },
    {
      "epoch": 44.22,
      "learning_rate": 1.0473737373737375e-06,
      "loss": 1.0514,
      "step": 948500
    },
    {
      "epoch": 44.24,
      "learning_rate": 1.0372727272727273e-06,
      "loss": 1.0497,
      "step": 949000
    },
    {
      "epoch": 44.26,
      "learning_rate": 1.0271717171717174e-06,
      "loss": 1.0528,
      "step": 949500
    },
    {
      "epoch": 44.29,
      "learning_rate": 1.0170909090909092e-06,
      "loss": 1.0502,
      "step": 950000
    },
    {
      "epoch": 44.31,
      "learning_rate": 1.006989898989899e-06,
      "loss": 1.0499,
      "step": 950500
    },
    {
      "epoch": 44.33,
      "learning_rate": 9.96888888888889e-07,
      "loss": 1.0517,
      "step": 951000
    },
    {
      "epoch": 44.36,
      "learning_rate": 9.867878787878788e-07,
      "loss": 1.0501,
      "step": 951500
    },
    {
      "epoch": 44.38,
      "learning_rate": 9.766868686868687e-07,
      "loss": 1.0493,
      "step": 952000
    },
    {
      "epoch": 44.4,
      "learning_rate": 9.665858585858587e-07,
      "loss": 1.0513,
      "step": 952500
    },
    {
      "epoch": 44.43,
      "learning_rate": 9.565050505050506e-07,
      "loss": 1.0513,
      "step": 953000
    },
    {
      "epoch": 44.45,
      "learning_rate": 9.464242424242424e-07,
      "loss": 1.0536,
      "step": 953500
    },
    {
      "epoch": 44.47,
      "learning_rate": 9.363232323232324e-07,
      "loss": 1.0508,
      "step": 954000
    },
    {
      "epoch": 44.5,
      "learning_rate": 9.262222222222222e-07,
      "loss": 1.0498,
      "step": 954500
    },
    {
      "epoch": 44.52,
      "learning_rate": 9.161212121212122e-07,
      "loss": 1.0506,
      "step": 955000
    },
    {
      "epoch": 44.54,
      "learning_rate": 9.060202020202021e-07,
      "loss": 1.0487,
      "step": 955500
    },
    {
      "epoch": 44.57,
      "learning_rate": 8.95919191919192e-07,
      "loss": 1.0513,
      "step": 956000
    },
    {
      "epoch": 44.59,
      "learning_rate": 8.858181818181819e-07,
      "loss": 1.0517,
      "step": 956500
    },
    {
      "epoch": 44.61,
      "learning_rate": 8.757171717171719e-07,
      "loss": 1.0493,
      "step": 957000
    },
    {
      "epoch": 44.64,
      "learning_rate": 8.656363636363638e-07,
      "loss": 1.0525,
      "step": 957500
    },
    {
      "epoch": 44.66,
      "learning_rate": 8.555353535353536e-07,
      "loss": 1.0497,
      "step": 958000
    },
    {
      "epoch": 44.68,
      "learning_rate": 8.454343434343434e-07,
      "loss": 1.0504,
      "step": 958500
    },
    {
      "epoch": 44.71,
      "learning_rate": 8.353535353535353e-07,
      "loss": 1.0504,
      "step": 959000
    },
    {
      "epoch": 44.73,
      "learning_rate": 8.252525252525254e-07,
      "loss": 1.0514,
      "step": 959500
    },
    {
      "epoch": 44.75,
      "learning_rate": 8.151515151515152e-07,
      "loss": 1.0512,
      "step": 960000
    },
    {
      "epoch": 44.78,
      "learning_rate": 8.050505050505051e-07,
      "loss": 1.0506,
      "step": 960500
    },
    {
      "epoch": 44.8,
      "learning_rate": 7.94949494949495e-07,
      "loss": 1.0515,
      "step": 961000
    },
    {
      "epoch": 44.82,
      "learning_rate": 7.84848484848485e-07,
      "loss": 1.0529,
      "step": 961500
    },
    {
      "epoch": 44.85,
      "learning_rate": 7.747474747474748e-07,
      "loss": 1.0502,
      "step": 962000
    },
    {
      "epoch": 44.87,
      "learning_rate": 7.646464646464648e-07,
      "loss": 1.05,
      "step": 962500
    },
    {
      "epoch": 44.89,
      "learning_rate": 7.545454545454546e-07,
      "loss": 1.0518,
      "step": 963000
    },
    {
      "epoch": 44.92,
      "learning_rate": 7.444646464646466e-07,
      "loss": 1.0504,
      "step": 963500
    },
    {
      "epoch": 44.94,
      "learning_rate": 7.343636363636364e-07,
      "loss": 1.051,
      "step": 964000
    },
    {
      "epoch": 44.96,
      "learning_rate": 7.242626262626263e-07,
      "loss": 1.0505,
      "step": 964500
    },
    {
      "epoch": 44.99,
      "learning_rate": 7.141616161616162e-07,
      "loss": 1.051,
      "step": 965000
    },
    {
      "epoch": 45.0,
      "eval_accuracy": 0.7787597321854502,
      "eval_loss": 0.917751669883728,
      "eval_runtime": 53.4964,
      "eval_samples_per_second": 873.106,
      "eval_steps_per_second": 13.646,
      "step": 965295
    },
    {
      "epoch": 45.01,
      "learning_rate": 7.040606060606062e-07,
      "loss": 1.0517,
      "step": 965500
    },
    {
      "epoch": 45.03,
      "learning_rate": 6.93979797979798e-07,
      "loss": 1.0497,
      "step": 966000
    },
    {
      "epoch": 45.06,
      "learning_rate": 6.83878787878788e-07,
      "loss": 1.0517,
      "step": 966500
    },
    {
      "epoch": 45.08,
      "learning_rate": 6.737777777777778e-07,
      "loss": 1.0502,
      "step": 967000
    },
    {
      "epoch": 45.1,
      "learning_rate": 6.636767676767678e-07,
      "loss": 1.0509,
      "step": 967500
    },
    {
      "epoch": 45.13,
      "learning_rate": 6.535757575757575e-07,
      "loss": 1.0487,
      "step": 968000
    },
    {
      "epoch": 45.15,
      "learning_rate": 6.434949494949496e-07,
      "loss": 1.0493,
      "step": 968500
    },
    {
      "epoch": 45.17,
      "learning_rate": 6.334141414141415e-07,
      "loss": 1.0508,
      "step": 969000
    },
    {
      "epoch": 45.2,
      "learning_rate": 6.233131313131314e-07,
      "loss": 1.0482,
      "step": 969500
    },
    {
      "epoch": 45.22,
      "learning_rate": 6.132121212121213e-07,
      "loss": 1.0505,
      "step": 970000
    },
    {
      "epoch": 45.24,
      "learning_rate": 6.031111111111112e-07,
      "loss": 1.0481,
      "step": 970500
    },
    {
      "epoch": 45.27,
      "learning_rate": 5.930101010101011e-07,
      "loss": 1.053,
      "step": 971000
    },
    {
      "epoch": 45.29,
      "learning_rate": 5.829090909090909e-07,
      "loss": 1.0512,
      "step": 971500
    },
    {
      "epoch": 45.31,
      "learning_rate": 5.728080808080808e-07,
      "loss": 1.0506,
      "step": 972000
    },
    {
      "epoch": 45.34,
      "learning_rate": 5.627474747474748e-07,
      "loss": 1.05,
      "step": 972500
    },
    {
      "epoch": 45.36,
      "learning_rate": 5.526464646464648e-07,
      "loss": 1.0504,
      "step": 973000
    },
    {
      "epoch": 45.38,
      "learning_rate": 5.425454545454545e-07,
      "loss": 1.0508,
      "step": 973500
    },
    {
      "epoch": 45.41,
      "learning_rate": 5.324444444444445e-07,
      "loss": 1.05,
      "step": 974000
    },
    {
      "epoch": 45.43,
      "learning_rate": 5.223434343434343e-07,
      "loss": 1.0507,
      "step": 974500
    },
    {
      "epoch": 45.45,
      "learning_rate": 5.122424242424243e-07,
      "loss": 1.0496,
      "step": 975000
    },
    {
      "epoch": 45.48,
      "learning_rate": 5.021414141414142e-07,
      "loss": 1.0501,
      "step": 975500
    },
    {
      "epoch": 45.5,
      "learning_rate": 4.92040404040404e-07,
      "loss": 1.0508,
      "step": 976000
    },
    {
      "epoch": 45.52,
      "learning_rate": 4.81939393939394e-07,
      "loss": 1.0515,
      "step": 976500
    },
    {
      "epoch": 45.55,
      "learning_rate": 4.7183838383838387e-07,
      "loss": 1.0507,
      "step": 977000
    },
    {
      "epoch": 45.57,
      "learning_rate": 4.617575757575758e-07,
      "loss": 1.0506,
      "step": 977500
    },
    {
      "epoch": 45.59,
      "learning_rate": 4.516565656565657e-07,
      "loss": 1.0488,
      "step": 978000
    },
    {
      "epoch": 45.62,
      "learning_rate": 4.4155555555555564e-07,
      "loss": 1.0509,
      "step": 978500
    },
    {
      "epoch": 45.64,
      "learning_rate": 4.3145454545454545e-07,
      "loss": 1.048,
      "step": 979000
    },
    {
      "epoch": 45.66,
      "learning_rate": 4.2135353535353536e-07,
      "loss": 1.0484,
      "step": 979500
    },
    {
      "epoch": 45.69,
      "learning_rate": 4.112727272727273e-07,
      "loss": 1.0526,
      "step": 980000
    },
    {
      "epoch": 45.71,
      "learning_rate": 4.011919191919193e-07,
      "loss": 1.0492,
      "step": 980500
    },
    {
      "epoch": 45.73,
      "learning_rate": 3.910909090909091e-07,
      "loss": 1.0508,
      "step": 981000
    },
    {
      "epoch": 45.76,
      "learning_rate": 3.80989898989899e-07,
      "loss": 1.0503,
      "step": 981500
    },
    {
      "epoch": 45.78,
      "learning_rate": 3.708888888888889e-07,
      "loss": 1.046,
      "step": 982000
    },
    {
      "epoch": 45.8,
      "learning_rate": 3.607878787878788e-07,
      "loss": 1.051,
      "step": 982500
    },
    {
      "epoch": 45.83,
      "learning_rate": 3.506868686868687e-07,
      "loss": 1.0494,
      "step": 983000
    },
    {
      "epoch": 45.85,
      "learning_rate": 3.405858585858586e-07,
      "loss": 1.0477,
      "step": 983500
    },
    {
      "epoch": 45.87,
      "learning_rate": 3.3048484848484853e-07,
      "loss": 1.0506,
      "step": 984000
    },
    {
      "epoch": 45.9,
      "learning_rate": 3.2040404040404046e-07,
      "loss": 1.0502,
      "step": 984500
    },
    {
      "epoch": 45.92,
      "learning_rate": 3.103030303030303e-07,
      "loss": 1.0507,
      "step": 985000
    },
    {
      "epoch": 45.94,
      "learning_rate": 3.0020202020202024e-07,
      "loss": 1.0485,
      "step": 985500
    },
    {
      "epoch": 45.97,
      "learning_rate": 2.901010101010101e-07,
      "loss": 1.0525,
      "step": 986000
    },
    {
      "epoch": 45.99,
      "learning_rate": 2.8e-07,
      "loss": 1.0477,
      "step": 986500
    },
    {
      "epoch": 46.0,
      "eval_accuracy": 0.7781940216531738,
      "eval_loss": 0.9196534156799316,
      "eval_runtime": 51.1139,
      "eval_samples_per_second": 913.803,
      "eval_steps_per_second": 14.282,
      "step": 986746
    },
    {
      "epoch": 46.01,
      "learning_rate": 2.6991919191919195e-07,
      "loss": 1.0487,
      "step": 987000
    },
    {
      "epoch": 46.04,
      "learning_rate": 2.5981818181818187e-07,
      "loss": 1.0516,
      "step": 987500
    },
    {
      "epoch": 46.06,
      "learning_rate": 2.4971717171717173e-07,
      "loss": 1.0494,
      "step": 988000
    },
    {
      "epoch": 46.08,
      "learning_rate": 2.3961616161616165e-07,
      "loss": 1.0497,
      "step": 988500
    },
    {
      "epoch": 46.11,
      "learning_rate": 2.295151515151515e-07,
      "loss": 1.0525,
      "step": 989000
    },
    {
      "epoch": 46.13,
      "learning_rate": 2.1941414141414142e-07,
      "loss": 1.0477,
      "step": 989500
    },
    {
      "epoch": 46.15,
      "learning_rate": 2.0931313131313131e-07,
      "loss": 1.049,
      "step": 990000
    },
    {
      "epoch": 46.18,
      "learning_rate": 1.9921212121212123e-07,
      "loss": 1.0503,
      "step": 990500
    },
    {
      "epoch": 46.2,
      "learning_rate": 1.8913131313131313e-07,
      "loss": 1.0496,
      "step": 991000
    },
    {
      "epoch": 46.22,
      "learning_rate": 1.7903030303030305e-07,
      "loss": 1.0485,
      "step": 991500
    },
    {
      "epoch": 46.24,
      "learning_rate": 1.6894949494949495e-07,
      "loss": 1.0523,
      "step": 992000
    },
    {
      "epoch": 46.27,
      "learning_rate": 1.5884848484848484e-07,
      "loss": 1.049,
      "step": 992500
    },
    {
      "epoch": 46.29,
      "learning_rate": 1.4874747474747476e-07,
      "loss": 1.049,
      "step": 993000
    },
    {
      "epoch": 46.31,
      "learning_rate": 1.3864646464646465e-07,
      "loss": 1.0488,
      "step": 993500
    },
    {
      "epoch": 46.34,
      "learning_rate": 1.2854545454545457e-07,
      "loss": 1.0493,
      "step": 994000
    },
    {
      "epoch": 46.36,
      "learning_rate": 1.1844444444444446e-07,
      "loss": 1.0514,
      "step": 994500
    },
    {
      "epoch": 46.38,
      "learning_rate": 1.0834343434343434e-07,
      "loss": 1.0492,
      "step": 995000
    },
    {
      "epoch": 46.41,
      "learning_rate": 9.824242424242425e-08,
      "loss": 1.0479,
      "step": 995500
    },
    {
      "epoch": 46.43,
      "learning_rate": 8.814141414141415e-08,
      "loss": 1.0518,
      "step": 996000
    },
    {
      "epoch": 46.45,
      "learning_rate": 7.806060606060607e-08,
      "loss": 1.0477,
      "step": 996500
    },
    {
      "epoch": 46.48,
      "learning_rate": 6.795959595959596e-08,
      "loss": 1.0474,
      "step": 997000
    },
    {
      "epoch": 46.5,
      "learning_rate": 5.785858585858587e-08,
      "loss": 1.0521,
      "step": 997500
    },
    {
      "epoch": 46.52,
      "learning_rate": 4.777777777777778e-08,
      "loss": 1.05,
      "step": 998000
    },
    {
      "epoch": 46.55,
      "learning_rate": 3.767676767676768e-08,
      "loss": 1.0509,
      "step": 998500
    },
    {
      "epoch": 46.57,
      "learning_rate": 2.757575757575758e-08,
      "loss": 1.051,
      "step": 999000
    },
    {
      "epoch": 46.59,
      "learning_rate": 1.7474747474747476e-08,
      "loss": 1.0523,
      "step": 999500
    },
    {
      "epoch": 46.62,
      "learning_rate": 7.373737373737375e-09,
      "loss": 1.0505,
      "step": 1000000
    },
    {
      "epoch": 46.62,
      "eval_accuracy": 0.7779879254303171,
      "eval_loss": 0.9209505319595337,
      "eval_runtime": 51.1248,
      "eval_samples_per_second": 913.607,
      "eval_steps_per_second": 14.279,
      "step": 1000000
    },
    {
      "epoch": 46.62,
      "step": 1000000,
      "total_flos": 6.7380511206774145e+19,
      "train_loss": 0.031499467041015625,
      "train_runtime": 14331.573,
      "train_samples_per_second": 17862.659,
      "train_steps_per_second": 69.776
    }
  ],
  "max_steps": 1000000,
  "num_train_epochs": 47,
  "total_flos": 6.7380511206774145e+19,
  "trial_name": null,
  "trial_params": null
}