{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.000483392030937,
  "global_step": 14484,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 4.9982733614199875e-05,
      "loss": 3.3505,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.994820084259963e-05,
      "loss": 1.7948,
      "step": 40
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.991366807099938e-05,
      "loss": 1.5872,
      "step": 60
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.9879135299399134e-05,
      "loss": 1.5076,
      "step": 80
    },
    {
      "epoch": 0.01,
      "learning_rate": 4.984460252779889e-05,
      "loss": 1.4722,
      "step": 100
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.981006975619863e-05,
      "loss": 1.4505,
      "step": 120
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.9775536984598386e-05,
      "loss": 1.44,
      "step": 140
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.974100421299814e-05,
      "loss": 1.407,
      "step": 160
    },
    {
      "epoch": 0.02,
      "learning_rate": 4.970647144139789e-05,
      "loss": 1.4086,
      "step": 180
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.9671938669797644e-05,
      "loss": 1.3679,
      "step": 200
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.963740589819739e-05,
      "loss": 1.3867,
      "step": 220
    },
    {
      "epoch": 0.03,
      "learning_rate": 4.960287312659714e-05,
      "loss": 1.3467,
      "step": 240
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9568340354996896e-05,
      "loss": 1.3652,
      "step": 260
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.953380758339665e-05,
      "loss": 1.3433,
      "step": 280
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9499274811796395e-05,
      "loss": 1.3221,
      "step": 300
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.946474204019615e-05,
      "loss": 1.3264,
      "step": 320
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.94302092685959e-05,
      "loss": 1.3122,
      "step": 340
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.939567649699565e-05,
      "loss": 1.3089,
      "step": 360
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.9361143725395406e-05,
      "loss": 1.3078,
      "step": 380
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.932661095379515e-05,
      "loss": 1.2902,
      "step": 400
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.9292078182194905e-05,
      "loss": 1.2991,
      "step": 420
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.925754541059466e-05,
      "loss": 1.2941,
      "step": 440
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.922301263899441e-05,
      "loss": 1.2855,
      "step": 460
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9188479867394163e-05,
      "loss": 1.3001,
      "step": 480
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.915394709579391e-05,
      "loss": 1.2868,
      "step": 500
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.911941432419366e-05,
      "loss": 1.2751,
      "step": 520
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9084881552593415e-05,
      "loss": 1.2751,
      "step": 540
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.905034878099317e-05,
      "loss": 1.2725,
      "step": 560
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.901581600939292e-05,
      "loss": 1.2476,
      "step": 580
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.898128323779267e-05,
      "loss": 1.2625,
      "step": 600
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.894675046619242e-05,
      "loss": 1.2567,
      "step": 620
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.891221769459217e-05,
      "loss": 1.2474,
      "step": 640
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.8877684922991925e-05,
      "loss": 1.2491,
      "step": 660
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.884315215139167e-05,
      "loss": 1.2476,
      "step": 680
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.8808619379791424e-05,
      "loss": 1.2328,
      "step": 700
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.877408660819118e-05,
      "loss": 1.2464,
      "step": 720
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.873955383659093e-05,
      "loss": 1.23,
      "step": 740
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.870502106499068e-05,
      "loss": 1.2462,
      "step": 760
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.867048829339043e-05,
      "loss": 1.2256,
      "step": 780
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.863595552179018e-05,
      "loss": 1.2234,
      "step": 800
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.8601422750189934e-05,
      "loss": 1.2223,
      "step": 820
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.856688997858969e-05,
      "loss": 1.2349,
      "step": 840
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.853235720698944e-05,
      "loss": 1.2273,
      "step": 860
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.8497824435389186e-05,
      "loss": 1.2351,
      "step": 880
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.846329166378894e-05,
      "loss": 1.2239,
      "step": 900
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.842875889218869e-05,
      "loss": 1.2147,
      "step": 920
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.8394226120588445e-05,
      "loss": 1.2129,
      "step": 940
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.83596933489882e-05,
      "loss": 1.2149,
      "step": 960
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.8325160577387943e-05,
      "loss": 1.2097,
      "step": 980
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.8290627805787696e-05,
      "loss": 1.214,
      "step": 1000
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.825609503418745e-05,
      "loss": 1.2035,
      "step": 1020
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.82215622625872e-05,
      "loss": 1.2006,
      "step": 1040
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.818702949098695e-05,
      "loss": 1.1965,
      "step": 1060
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.81524967193867e-05,
      "loss": 1.1995,
      "step": 1080
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.8117963947786454e-05,
      "loss": 1.1923,
      "step": 1100
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.8083431176186207e-05,
      "loss": 1.1872,
      "step": 1120
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.804889840458596e-05,
      "loss": 1.1965,
      "step": 1140
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.8014365632985705e-05,
      "loss": 1.1886,
      "step": 1160
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.797983286138546e-05,
      "loss": 1.1855,
      "step": 1180
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.794530008978521e-05,
      "loss": 1.1894,
      "step": 1200
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.7910767318184964e-05,
      "loss": 1.1936,
      "step": 1220
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.787623454658471e-05,
      "loss": 1.1828,
      "step": 1240
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.784170177498446e-05,
      "loss": 1.2018,
      "step": 1260
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.780716900338421e-05,
      "loss": 1.1821,
      "step": 1280
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.777263623178396e-05,
      "loss": 1.1762,
      "step": 1300
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.7738103460183714e-05,
      "loss": 1.1855,
      "step": 1320
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.770357068858347e-05,
      "loss": 1.1738,
      "step": 1340
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.766903791698322e-05,
      "loss": 1.1739,
      "step": 1360
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.7634505145382966e-05,
      "loss": 1.1756,
      "step": 1380
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.759997237378272e-05,
      "loss": 1.1687,
      "step": 1400
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.756543960218247e-05,
      "loss": 1.16,
      "step": 1420
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.7530906830582225e-05,
      "loss": 1.1658,
      "step": 1440
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.749637405898197e-05,
      "loss": 1.1757,
      "step": 1460
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.7461841287381724e-05,
      "loss": 1.1742,
      "step": 1480
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.7427308515781476e-05,
      "loss": 1.1681,
      "step": 1500
    },
    {
      "epoch": 0.21,
      "eval_validation_accuracy": 0.05685549584758183,
      "eval_validation_loss": 1.171875,
      "eval_validation_runtime": 3753.295,
      "eval_validation_samples_per_second": 0.666,
      "eval_validation_steps_per_second": 0.021,
      "step": 1500
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.739277574418123e-05,
      "loss": 1.1749,
      "step": 1520
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.735824297258098e-05,
      "loss": 1.1617,
      "step": 1540
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.732371020098073e-05,
      "loss": 1.1696,
      "step": 1560
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.728917742938048e-05,
      "loss": 1.1638,
      "step": 1580
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.7254644657780234e-05,
      "loss": 1.1634,
      "step": 1600
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.7220111886179987e-05,
      "loss": 1.1545,
      "step": 1620
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.718557911457974e-05,
      "loss": 1.1667,
      "step": 1640
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.7151046342979485e-05,
      "loss": 1.1503,
      "step": 1660
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.711651357137924e-05,
      "loss": 1.1633,
      "step": 1680
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.708198079977899e-05,
      "loss": 1.1653,
      "step": 1700
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.7047448028178744e-05,
      "loss": 1.1642,
      "step": 1720
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.70129152565785e-05,
      "loss": 1.15,
      "step": 1740
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.697838248497824e-05,
      "loss": 1.1667,
      "step": 1760
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.6943849713377996e-05,
      "loss": 1.1532,
      "step": 1780
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.690931694177775e-05,
      "loss": 1.1621,
      "step": 1800
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.68747841701775e-05,
      "loss": 1.1509,
      "step": 1820
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.684025139857725e-05,
      "loss": 1.1515,
      "step": 1840
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.6805718626977e-05,
      "loss": 1.1513,
      "step": 1860
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.677118585537675e-05,
      "loss": 1.1538,
      "step": 1880
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.6736653083776506e-05,
      "loss": 1.146,
      "step": 1900
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.670212031217626e-05,
      "loss": 1.1537,
      "step": 1920
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.6667587540576005e-05,
      "loss": 1.1524,
      "step": 1940
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.663305476897576e-05,
      "loss": 1.1523,
      "step": 1960
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.659852199737551e-05,
      "loss": 1.15,
      "step": 1980
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.656398922577526e-05,
      "loss": 1.1419,
      "step": 2000
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.6529456454175016e-05,
      "loss": 1.1525,
      "step": 2020
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.649492368257476e-05,
      "loss": 1.1457,
      "step": 2040
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.6460390910974515e-05,
      "loss": 1.1454,
      "step": 2060
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.642585813937427e-05,
      "loss": 1.1284,
      "step": 2080
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.639132536777402e-05,
      "loss": 1.1481,
      "step": 2100
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.6356792596173773e-05,
      "loss": 1.1484,
      "step": 2120
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.632225982457352e-05,
      "loss": 1.1332,
      "step": 2140
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.628772705297327e-05,
      "loss": 1.138,
      "step": 2160
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.6253194281373025e-05,
      "loss": 1.1369,
      "step": 2180
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.621866150977278e-05,
      "loss": 1.1399,
      "step": 2200
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.6184128738172524e-05,
      "loss": 1.1221,
      "step": 2220
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.614959596657228e-05,
      "loss": 1.1233,
      "step": 2240
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.611506319497203e-05,
      "loss": 1.1248,
      "step": 2260
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.608053042337178e-05,
      "loss": 1.1333,
      "step": 2280
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.6045997651771535e-05,
      "loss": 1.1266,
      "step": 2300
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.601146488017128e-05,
      "loss": 1.1368,
      "step": 2320
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.5976932108571034e-05,
      "loss": 1.1296,
      "step": 2340
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.594239933697079e-05,
      "loss": 1.1318,
      "step": 2360
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.590786656537054e-05,
      "loss": 1.1272,
      "step": 2380
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.587333379377029e-05,
      "loss": 1.1197,
      "step": 2400
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.583880102217004e-05,
      "loss": 1.1271,
      "step": 2420
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.580426825056979e-05,
      "loss": 1.1193,
      "step": 2440
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.5769735478969544e-05,
      "loss": 1.1251,
      "step": 2460
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.57352027073693e-05,
      "loss": 1.123,
      "step": 2480
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.570066993576905e-05,
      "loss": 1.1141,
      "step": 2500
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.5666137164168796e-05,
      "loss": 1.1301,
      "step": 2520
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.563160439256855e-05,
      "loss": 1.1106,
      "step": 2540
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.55970716209683e-05,
      "loss": 1.1258,
      "step": 2560
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.5562538849368055e-05,
      "loss": 1.1198,
      "step": 2580
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.552800607776781e-05,
      "loss": 1.1157,
      "step": 2600
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.5493473306167554e-05,
      "loss": 1.1277,
      "step": 2620
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.5458940534567306e-05,
      "loss": 1.1169,
      "step": 2640
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.542440776296706e-05,
      "loss": 1.1062,
      "step": 2660
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.538987499136681e-05,
      "loss": 1.1187,
      "step": 2680
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.535534221976656e-05,
      "loss": 1.1095,
      "step": 2700
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.532080944816631e-05,
      "loss": 1.1235,
      "step": 2720
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.5286276676566064e-05,
      "loss": 1.1165,
      "step": 2740
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.5251743904965817e-05,
      "loss": 1.117,
      "step": 2760
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.521721113336557e-05,
      "loss": 1.1073,
      "step": 2780
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.5182678361765315e-05,
      "loss": 1.1156,
      "step": 2800
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.514814559016507e-05,
      "loss": 1.1034,
      "step": 2820
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.511361281856482e-05,
      "loss": 1.1127,
      "step": 2840
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.5079080046964574e-05,
      "loss": 1.1214,
      "step": 2860
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.504454727536433e-05,
      "loss": 1.116,
      "step": 2880
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.501001450376407e-05,
      "loss": 1.1178,
      "step": 2900
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.4975481732163826e-05,
      "loss": 1.1048,
      "step": 2920
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.494094896056358e-05,
      "loss": 1.1026,
      "step": 2940
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.490641618896333e-05,
      "loss": 1.1151,
      "step": 2960
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.4871883417363084e-05,
      "loss": 1.0999,
      "step": 2980
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.483735064576283e-05,
      "loss": 1.1257,
      "step": 3000
    },
    {
      "epoch": 0.41,
      "eval_validation_accuracy": 0.0578276502198339,
      "eval_validation_loss": 1.109375,
      "eval_validation_runtime": 3693.3851,
      "eval_validation_samples_per_second": 0.677,
      "eval_validation_steps_per_second": 0.021,
      "step": 3000
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.480281787416258e-05,
      "loss": 1.1124,
      "step": 3020
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.4768285102562336e-05,
      "loss": 1.1108,
      "step": 3040
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.473375233096209e-05,
      "loss": 1.0994,
      "step": 3060
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.4699219559361835e-05,
      "loss": 1.11,
      "step": 3080
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.466468678776159e-05,
      "loss": 1.101,
      "step": 3100
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.463015401616134e-05,
      "loss": 1.0904,
      "step": 3120
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.459562124456109e-05,
      "loss": 1.1135,
      "step": 3140
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.4561088472960846e-05,
      "loss": 1.0917,
      "step": 3160
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.452655570136059e-05,
      "loss": 1.109,
      "step": 3180
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.4492022929760345e-05,
      "loss": 1.0901,
      "step": 3200
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.44574901581601e-05,
      "loss": 1.1037,
      "step": 3220
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.442295738655985e-05,
      "loss": 1.1057,
      "step": 3240
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.4388424614959603e-05,
      "loss": 1.1004,
      "step": 3260
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.435389184335935e-05,
      "loss": 1.1008,
      "step": 3280
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.43193590717591e-05,
      "loss": 1.1036,
      "step": 3300
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.4284826300158855e-05,
      "loss": 1.1124,
      "step": 3320
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.425029352855861e-05,
      "loss": 1.1096,
      "step": 3340
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.421576075695836e-05,
      "loss": 1.1069,
      "step": 3360
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.418122798535811e-05,
      "loss": 1.0886,
      "step": 3380
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.414669521375786e-05,
      "loss": 1.0998,
      "step": 3400
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.411216244215761e-05,
      "loss": 1.0858,
      "step": 3420
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.4077629670557365e-05,
      "loss": 1.0987,
      "step": 3440
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.404309689895711e-05,
      "loss": 1.0952,
      "step": 3460
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.4010290765936875e-05,
      "loss": 1.1043,
      "step": 3480
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.397575799433663e-05,
      "loss": 1.0918,
      "step": 3500
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.3941225222736374e-05,
      "loss": 1.0952,
      "step": 3520
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.3906692451136127e-05,
      "loss": 1.0997,
      "step": 3540
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.387215967953588e-05,
      "loss": 1.1022,
      "step": 3560
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.383762690793563e-05,
      "loss": 1.0985,
      "step": 3580
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.3803094136335385e-05,
      "loss": 1.0928,
      "step": 3600
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.376856136473513e-05,
      "loss": 1.0793,
      "step": 3620
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.3734028593134884e-05,
      "loss": 1.0927,
      "step": 3640
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.369949582153464e-05,
      "loss": 1.0959,
      "step": 3660
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.366496304993439e-05,
      "loss": 1.0914,
      "step": 3680
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.3630430278334136e-05,
      "loss": 1.0874,
      "step": 3700
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.359589750673389e-05,
      "loss": 1.0954,
      "step": 3720
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.356136473513364e-05,
      "loss": 1.0906,
      "step": 3740
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.3526831963533394e-05,
      "loss": 1.0896,
      "step": 3760
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.349229919193315e-05,
      "loss": 1.096,
      "step": 3780
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.345776642033289e-05,
      "loss": 1.0946,
      "step": 3800
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.3423233648732646e-05,
      "loss": 1.0868,
      "step": 3820
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.33887008771324e-05,
      "loss": 1.0914,
      "step": 3840
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.335416810553215e-05,
      "loss": 1.0836,
      "step": 3860
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.3319635333931904e-05,
      "loss": 1.0865,
      "step": 3880
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.328510256233165e-05,
      "loss": 1.0922,
      "step": 3900
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.32505697907314e-05,
      "loss": 1.0891,
      "step": 3920
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.3216037019131156e-05,
      "loss": 1.0836,
      "step": 3940
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.318150424753091e-05,
      "loss": 1.0914,
      "step": 3960
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.314697147593066e-05,
      "loss": 1.0772,
      "step": 3980
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.311243870433041e-05,
      "loss": 1.0849,
      "step": 4000
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.307790593273016e-05,
      "loss": 1.0747,
      "step": 4020
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.3043373161129913e-05,
      "loss": 1.0904,
      "step": 4040
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.3008840389529666e-05,
      "loss": 1.0839,
      "step": 4060
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.297430761792942e-05,
      "loss": 1.0799,
      "step": 4080
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.2939774846329165e-05,
      "loss": 1.0787,
      "step": 4100
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.290524207472892e-05,
      "loss": 1.0659,
      "step": 4120
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.287070930312867e-05,
      "loss": 1.0821,
      "step": 4140
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.2836176531528424e-05,
      "loss": 1.0857,
      "step": 4160
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.280164375992817e-05,
      "loss": 1.0834,
      "step": 4180
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.276711098832792e-05,
      "loss": 1.0793,
      "step": 4200
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.2732578216727675e-05,
      "loss": 1.0692,
      "step": 4220
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.269804544512743e-05,
      "loss": 1.0762,
      "step": 4240
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.266351267352718e-05,
      "loss": 1.0859,
      "step": 4260
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.262897990192693e-05,
      "loss": 1.0827,
      "step": 4280
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.259444713032668e-05,
      "loss": 1.0787,
      "step": 4300
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.255991435872643e-05,
      "loss": 1.0711,
      "step": 4320
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.2525381587126186e-05,
      "loss": 1.0852,
      "step": 4340
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.249084881552594e-05,
      "loss": 1.0719,
      "step": 4360
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.2456316043925684e-05,
      "loss": 1.0877,
      "step": 4380
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.242178327232544e-05,
      "loss": 1.0856,
      "step": 4400
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.238725050072519e-05,
      "loss": 1.0704,
      "step": 4420
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.235271772912494e-05,
      "loss": 1.0776,
      "step": 4440
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.2318184957524696e-05,
      "loss": 1.0751,
      "step": 4460
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.228365218592444e-05,
      "loss": 1.0801,
      "step": 4480
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.2249119414324195e-05,
      "loss": 1.0891,
      "step": 4500
    },
    {
      "epoch": 0.62,
      "eval_validation_accuracy": 0.0582704445530044,
      "eval_validation_loss": 1.0771484375,
      "eval_validation_runtime": 3796.035,
      "eval_validation_samples_per_second": 0.659,
      "eval_validation_steps_per_second": 0.021,
      "step": 4500
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.221458664272395e-05,
      "loss": 1.0832,
      "step": 4520
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.21800538711237e-05,
      "loss": 1.076,
      "step": 4540
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.2145521099523446e-05,
      "loss": 1.0764,
      "step": 4560
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.21109883279232e-05,
      "loss": 1.0848,
      "step": 4580
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.207645555632295e-05,
      "loss": 1.0813,
      "step": 4600
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.2041922784722705e-05,
      "loss": 1.081,
      "step": 4620
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.200739001312246e-05,
      "loss": 1.0664,
      "step": 4640
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.1972857241522204e-05,
      "loss": 1.074,
      "step": 4660
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.1938324469921957e-05,
      "loss": 1.069,
      "step": 4680
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.190379169832171e-05,
      "loss": 1.0788,
      "step": 4700
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.186925892672146e-05,
      "loss": 1.0631,
      "step": 4720
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.1836452793701226e-05,
      "loss": 1.0728,
      "step": 4740
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.180192002210098e-05,
      "loss": 1.0665,
      "step": 4760
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.176911388908074e-05,
      "loss": 1.0753,
      "step": 4780
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.1734581117480495e-05,
      "loss": 1.0741,
      "step": 4800
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.170004834588024e-05,
      "loss": 1.071,
      "step": 4820
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.1665515574279994e-05,
      "loss": 1.0826,
      "step": 4840
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.1630982802679747e-05,
      "loss": 1.067,
      "step": 4860
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.15964500310795e-05,
      "loss": 1.0671,
      "step": 4880
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.156191725947925e-05,
      "loss": 1.0666,
      "step": 4900
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.1527384487879e-05,
      "loss": 1.0749,
      "step": 4920
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.149285171627875e-05,
      "loss": 1.0772,
      "step": 4940
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.1458318944678504e-05,
      "loss": 1.0762,
      "step": 4960
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.142378617307826e-05,
      "loss": 1.0741,
      "step": 4980
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.138925340147801e-05,
      "loss": 1.0701,
      "step": 5000
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.1354720629877756e-05,
      "loss": 1.0724,
      "step": 5020
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.132018785827751e-05,
      "loss": 1.0651,
      "step": 5040
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.128565508667726e-05,
      "loss": 1.0653,
      "step": 5060
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.1251122315077014e-05,
      "loss": 1.0854,
      "step": 5080
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.121658954347676e-05,
      "loss": 1.064,
      "step": 5100
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.118205677187651e-05,
      "loss": 1.0818,
      "step": 5120
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.1147524000276266e-05,
      "loss": 1.0686,
      "step": 5140
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.111299122867602e-05,
      "loss": 1.0708,
      "step": 5160
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.107845845707577e-05,
      "loss": 1.0551,
      "step": 5180
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.104392568547552e-05,
      "loss": 1.0667,
      "step": 5200
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.100939291387527e-05,
      "loss": 1.0588,
      "step": 5220
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.097486014227502e-05,
      "loss": 1.0591,
      "step": 5240
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.0940327370674776e-05,
      "loss": 1.085,
      "step": 5260
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.090579459907453e-05,
      "loss": 1.0772,
      "step": 5280
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.0871261827474275e-05,
      "loss": 1.0565,
      "step": 5300
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.083672905587403e-05,
      "loss": 1.0626,
      "step": 5320
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.080219628427378e-05,
      "loss": 1.0589,
      "step": 5340
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.0767663512673533e-05,
      "loss": 1.0693,
      "step": 5360
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.0733130741073286e-05,
      "loss": 1.0834,
      "step": 5380
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.069859796947303e-05,
      "loss": 1.0668,
      "step": 5400
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.066406519787278e-05,
      "loss": 1.0722,
      "step": 5420
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.062953242627253e-05,
      "loss": 1.0591,
      "step": 5440
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.0594999654672284e-05,
      "loss": 1.0579,
      "step": 5460
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.056046688307204e-05,
      "loss": 1.0488,
      "step": 5480
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.052593411147178e-05,
      "loss": 1.0575,
      "step": 5500
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.0491401339871536e-05,
      "loss": 1.0654,
      "step": 5520
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.045686856827129e-05,
      "loss": 1.0793,
      "step": 5540
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.042233579667104e-05,
      "loss": 1.0717,
      "step": 5560
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.0387803025070794e-05,
      "loss": 1.0772,
      "step": 5580
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.035327025347054e-05,
      "loss": 1.0641,
      "step": 5600
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.031873748187029e-05,
      "loss": 1.0649,
      "step": 5620
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.0284204710270046e-05,
      "loss": 1.0728,
      "step": 5640
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.02496719386698e-05,
      "loss": 1.0697,
      "step": 5660
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.021513916706955e-05,
      "loss": 1.0575,
      "step": 5680
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.01806063954693e-05,
      "loss": 1.0569,
      "step": 5700
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.014607362386905e-05,
      "loss": 1.0655,
      "step": 5720
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.01115408522688e-05,
      "loss": 1.0669,
      "step": 5740
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0077008080668556e-05,
      "loss": 1.0599,
      "step": 5760
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.004247530906831e-05,
      "loss": 1.0731,
      "step": 5780
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0007942537468055e-05,
      "loss": 1.0596,
      "step": 5800
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.997340976586781e-05,
      "loss": 1.0625,
      "step": 5820
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.993887699426756e-05,
      "loss": 1.0601,
      "step": 5840
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.9904344222667314e-05,
      "loss": 1.0605,
      "step": 5860
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.986981145106706e-05,
      "loss": 1.0535,
      "step": 5880
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.983527867946681e-05,
      "loss": 1.0612,
      "step": 5900
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.9800745907866565e-05,
      "loss": 1.0652,
      "step": 5920
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.976621313626632e-05,
      "loss": 1.0616,
      "step": 5940
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.973168036466607e-05,
      "loss": 1.0546,
      "step": 5960
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.969714759306582e-05,
      "loss": 1.0518,
      "step": 5980
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.966434146004559e-05,
      "loss": 1.0609,
      "step": 6000
    },
    {
      "epoch": 0.83,
      "eval_validation_accuracy": 0.05839198827552516,
      "eval_validation_loss": 1.0556640625,
      "eval_validation_runtime": 3758.5842,
      "eval_validation_samples_per_second": 0.665,
      "eval_validation_steps_per_second": 0.021,
      "step": 6000
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.962980868844533e-05,
      "loss": 1.0554,
      "step": 6020
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.9595275916845086e-05,
      "loss": 1.0507,
      "step": 6040
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.956074314524484e-05,
      "loss": 1.0517,
      "step": 6060
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.952621037364459e-05,
      "loss": 1.0482,
      "step": 6080
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.9491677602044345e-05,
      "loss": 1.0634,
      "step": 6100
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.945714483044409e-05,
      "loss": 1.0629,
      "step": 6120
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.9422612058843843e-05,
      "loss": 1.0474,
      "step": 6140
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.9388079287243596e-05,
      "loss": 1.0513,
      "step": 6160
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.935527315422336e-05,
      "loss": 1.0527,
      "step": 6180
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.932074038262311e-05,
      "loss": 1.0482,
      "step": 6200
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.9286207611022865e-05,
      "loss": 1.069,
      "step": 6220
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.925167483942261e-05,
      "loss": 1.0485,
      "step": 6240
    },
    {
      "epoch": 0.86,
      "learning_rate": 3.9217142067822364e-05,
      "loss": 1.0525,
      "step": 6260
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.918260929622212e-05,
      "loss": 1.0424,
      "step": 6280
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.914807652462187e-05,
      "loss": 1.0501,
      "step": 6300
    },
    {
      "epoch": 0.87,
      "learning_rate": 3.911354375302162e-05,
      "loss": 1.048,
      "step": 6320
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.907901098142137e-05,
      "loss": 1.0454,
      "step": 6340
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.904447820982112e-05,
      "loss": 1.0691,
      "step": 6360
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.9009945438220875e-05,
      "loss": 1.0534,
      "step": 6380
    },
    {
      "epoch": 0.88,
      "learning_rate": 3.897541266662063e-05,
      "loss": 1.0413,
      "step": 6400
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.8940879895020373e-05,
      "loss": 1.0564,
      "step": 6420
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.8906347123420126e-05,
      "loss": 1.0589,
      "step": 6440
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.887181435181988e-05,
      "loss": 1.0495,
      "step": 6460
    },
    {
      "epoch": 0.89,
      "learning_rate": 3.883900821879964e-05,
      "loss": 1.0499,
      "step": 6480
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.8804475447199395e-05,
      "loss": 1.0551,
      "step": 6500
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.876994267559915e-05,
      "loss": 1.0399,
      "step": 6520
    },
    {
      "epoch": 0.9,
      "learning_rate": 3.87354099039989e-05,
      "loss": 1.0541,
      "step": 6540
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.870087713239865e-05,
      "loss": 1.0507,
      "step": 6560
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.86663443607984e-05,
      "loss": 1.0489,
      "step": 6580
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.863181158919815e-05,
      "loss": 1.0456,
      "step": 6600
    },
    {
      "epoch": 0.91,
      "learning_rate": 3.8597278817597906e-05,
      "loss": 1.0476,
      "step": 6620
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.856274604599766e-05,
      "loss": 1.0498,
      "step": 6640
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.8528213274397404e-05,
      "loss": 1.0444,
      "step": 6660
    },
    {
      "epoch": 0.92,
      "learning_rate": 3.849368050279716e-05,
      "loss": 1.059,
      "step": 6680
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.845914773119691e-05,
      "loss": 1.0575,
      "step": 6700
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.842461495959666e-05,
      "loss": 1.0364,
      "step": 6720
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.839008218799641e-05,
      "loss": 1.0455,
      "step": 6740
    },
    {
      "epoch": 0.93,
      "learning_rate": 3.835554941639616e-05,
      "loss": 1.0484,
      "step": 6760
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.8321016644795915e-05,
      "loss": 1.0263,
      "step": 6780
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.828648387319567e-05,
      "loss": 1.048,
      "step": 6800
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.825195110159542e-05,
      "loss": 1.0526,
      "step": 6820
    },
    {
      "epoch": 0.94,
      "learning_rate": 3.8217418329995166e-05,
      "loss": 1.0433,
      "step": 6840
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.818288555839492e-05,
      "loss": 1.0491,
      "step": 6860
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.814835278679467e-05,
      "loss": 1.0437,
      "step": 6880
    },
    {
      "epoch": 0.95,
      "learning_rate": 3.8113820015194425e-05,
      "loss": 1.0498,
      "step": 6900
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.807928724359418e-05,
      "loss": 1.0457,
      "step": 6920
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.8044754471993924e-05,
      "loss": 1.0481,
      "step": 6940
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.8010221700393677e-05,
      "loss": 1.0372,
      "step": 6960
    },
    {
      "epoch": 0.96,
      "learning_rate": 3.797568892879343e-05,
      "loss": 1.056,
      "step": 6980
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.794115615719318e-05,
      "loss": 1.0413,
      "step": 7000
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.7906623385592935e-05,
      "loss": 1.0426,
      "step": 7020
    },
    {
      "epoch": 0.97,
      "learning_rate": 3.787209061399268e-05,
      "loss": 1.043,
      "step": 7040
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.7837557842392434e-05,
      "loss": 1.0504,
      "step": 7060
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.780302507079219e-05,
      "loss": 1.0467,
      "step": 7080
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.776849229919194e-05,
      "loss": 1.047,
      "step": 7100
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.7733959527591686e-05,
      "loss": 1.0384,
      "step": 7120
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.769942675599143e-05,
      "loss": 1.0451,
      "step": 7140
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.7664893984391185e-05,
      "loss": 1.0366,
      "step": 7160
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.763036121279094e-05,
      "loss": 1.0403,
      "step": 7180
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.759582844119069e-05,
      "loss": 1.0315,
      "step": 7200
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.756129566959044e-05,
      "loss": 1.0303,
      "step": 7220
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.752676289799019e-05,
      "loss": 1.0395,
      "step": 7240
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.749223012638994e-05,
      "loss": 0.985,
      "step": 7260
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.7457697354789695e-05,
      "loss": 0.9691,
      "step": 7280
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.742316458318945e-05,
      "loss": 0.9904,
      "step": 7300
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.73886318115892e-05,
      "loss": 0.973,
      "step": 7320
    },
    {
      "epoch": 1.01,
      "learning_rate": 3.7354099039988946e-05,
      "loss": 0.9727,
      "step": 7340
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.73195662683887e-05,
      "loss": 0.9801,
      "step": 7360
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.728503349678845e-05,
      "loss": 0.9896,
      "step": 7380
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.7250500725188205e-05,
      "loss": 0.9718,
      "step": 7400
    },
    {
      "epoch": 1.02,
      "learning_rate": 3.721596795358796e-05,
      "loss": 0.9824,
      "step": 7420
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.7181435181987704e-05,
      "loss": 0.9742,
      "step": 7440
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.714690241038746e-05,
      "loss": 0.9716,
      "step": 7460
    },
    {
      "epoch": 1.03,
      "learning_rate": 3.711236963878721e-05,
      "loss": 0.9822,
      "step": 7480
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.707783686718696e-05,
      "loss": 0.9863,
      "step": 7500
    },
    {
      "epoch": 1.04,
      "eval_validation_accuracy": 0.05860850024425989,
      "eval_validation_loss": 1.0419921875,
      "eval_validation_runtime": 3723.0597,
      "eval_validation_samples_per_second": 0.671,
      "eval_validation_steps_per_second": 0.021,
      "step": 7500
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.704330409558671e-05,
      "loss": 0.979,
      "step": 7520
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.700877132398646e-05,
      "loss": 0.9744,
      "step": 7540
    },
    {
      "epoch": 1.04,
      "learning_rate": 3.6974238552386214e-05,
      "loss": 0.9763,
      "step": 7560
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.693970578078597e-05,
      "loss": 0.9725,
      "step": 7580
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.690517300918572e-05,
      "loss": 0.9691,
      "step": 7600
    },
    {
      "epoch": 1.05,
      "learning_rate": 3.6870640237585466e-05,
      "loss": 0.9741,
      "step": 7620
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.683610746598522e-05,
      "loss": 0.9836,
      "step": 7640
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.680157469438497e-05,
      "loss": 0.9823,
      "step": 7660
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.6767041922784724e-05,
      "loss": 0.9781,
      "step": 7680
    },
    {
      "epoch": 1.06,
      "learning_rate": 3.673250915118448e-05,
      "loss": 0.9754,
      "step": 7700
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.669797637958422e-05,
      "loss": 0.9769,
      "step": 7720
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.6663443607983976e-05,
      "loss": 0.9831,
      "step": 7740
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.662891083638373e-05,
      "loss": 0.9825,
      "step": 7760
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.659437806478348e-05,
      "loss": 0.9838,
      "step": 7780
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.6559845293183234e-05,
      "loss": 0.97,
      "step": 7800
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.652531252158298e-05,
      "loss": 0.9849,
      "step": 7820
    },
    {
      "epoch": 1.08,
      "learning_rate": 3.649077974998273e-05,
      "loss": 0.9854,
      "step": 7840
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.6456246978382486e-05,
      "loss": 0.9765,
      "step": 7860
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.642171420678224e-05,
      "loss": 0.9758,
      "step": 7880
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.6387181435181985e-05,
      "loss": 0.9696,
      "step": 7900
    },
    {
      "epoch": 1.09,
      "learning_rate": 3.635264866358174e-05,
      "loss": 0.9751,
      "step": 7920
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.631811589198149e-05,
      "loss": 0.979,
      "step": 7940
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.6283583120381244e-05,
      "loss": 0.9609,
      "step": 7960
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.6249050348780996e-05,
      "loss": 0.9798,
      "step": 7980
    },
    {
      "epoch": 1.1,
      "learning_rate": 3.621451757718074e-05,
      "loss": 0.9627,
      "step": 8000
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.6179984805580495e-05,
      "loss": 0.981,
      "step": 8020
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.614545203398025e-05,
      "loss": 0.9759,
      "step": 8040
    },
    {
      "epoch": 1.11,
      "learning_rate": 3.611091926238e-05,
      "loss": 0.9782,
      "step": 8060
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.6076386490779754e-05,
      "loss": 0.9778,
      "step": 8080
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.60418537191795e-05,
      "loss": 0.9752,
      "step": 8100
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.600732094757925e-05,
      "loss": 0.9852,
      "step": 8120
    },
    {
      "epoch": 1.12,
      "learning_rate": 3.5972788175979005e-05,
      "loss": 0.9776,
      "step": 8140
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.593825540437876e-05,
      "loss": 0.9742,
      "step": 8160
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.590372263277851e-05,
      "loss": 0.9716,
      "step": 8180
    },
    {
      "epoch": 1.13,
      "learning_rate": 3.586918986117826e-05,
      "loss": 0.9744,
      "step": 8200
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.583465708957801e-05,
      "loss": 0.9738,
      "step": 8220
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.580012431797776e-05,
      "loss": 0.9851,
      "step": 8240
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.5765591546377516e-05,
      "loss": 0.9794,
      "step": 8260
    },
    {
      "epoch": 1.14,
      "learning_rate": 3.573105877477727e-05,
      "loss": 0.9785,
      "step": 8280
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.5696526003177015e-05,
      "loss": 0.978,
      "step": 8300
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.566199323157677e-05,
      "loss": 0.9681,
      "step": 8320
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.562746045997652e-05,
      "loss": 0.9849,
      "step": 8340
    },
    {
      "epoch": 1.15,
      "learning_rate": 3.559292768837627e-05,
      "loss": 0.9799,
      "step": 8360
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.555839491677602e-05,
      "loss": 0.9694,
      "step": 8380
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.552386214517577e-05,
      "loss": 0.9724,
      "step": 8400
    },
    {
      "epoch": 1.16,
      "learning_rate": 3.5489329373575525e-05,
      "loss": 0.9705,
      "step": 8420
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.545479660197528e-05,
      "loss": 0.9846,
      "step": 8440
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.542026383037503e-05,
      "loss": 0.9729,
      "step": 8460
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.5385731058774776e-05,
      "loss": 0.9741,
      "step": 8480
    },
    {
      "epoch": 1.17,
      "learning_rate": 3.535119828717453e-05,
      "loss": 0.9801,
      "step": 8500
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.531666551557428e-05,
      "loss": 0.9863,
      "step": 8520
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.5282132743974035e-05,
      "loss": 0.9856,
      "step": 8540
    },
    {
      "epoch": 1.18,
      "learning_rate": 3.524759997237379e-05,
      "loss": 0.9845,
      "step": 8560
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.5213067200773534e-05,
      "loss": 0.9791,
      "step": 8580
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.517853442917329e-05,
      "loss": 0.9719,
      "step": 8600
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.514400165757304e-05,
      "loss": 0.9714,
      "step": 8620
    },
    {
      "epoch": 1.19,
      "learning_rate": 3.510946888597279e-05,
      "loss": 0.9802,
      "step": 8640
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.5074936114372545e-05,
      "loss": 0.9823,
      "step": 8660
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.504040334277229e-05,
      "loss": 0.9808,
      "step": 8680
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.5005870571172044e-05,
      "loss": 0.9788,
      "step": 8700
    },
    {
      "epoch": 1.2,
      "learning_rate": 3.49713377995718e-05,
      "loss": 0.9805,
      "step": 8720
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.493680502797155e-05,
      "loss": 0.9724,
      "step": 8740
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.4902272256371296e-05,
      "loss": 0.9887,
      "step": 8760
    },
    {
      "epoch": 1.21,
      "learning_rate": 3.486773948477105e-05,
      "loss": 0.9863,
      "step": 8780
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.48332067131708e-05,
      "loss": 0.9789,
      "step": 8800
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.4798673941570554e-05,
      "loss": 0.9738,
      "step": 8820
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.476414116997031e-05,
      "loss": 0.9881,
      "step": 8840
    },
    {
      "epoch": 1.22,
      "learning_rate": 3.472960839837005e-05,
      "loss": 0.9718,
      "step": 8860
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.4695075626769806e-05,
      "loss": 0.9842,
      "step": 8880
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.466054285516956e-05,
      "loss": 0.9775,
      "step": 8900
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.462601008356931e-05,
      "loss": 0.9701,
      "step": 8920
    },
    {
      "epoch": 1.23,
      "learning_rate": 3.4591477311969064e-05,
      "loss": 0.9826,
      "step": 8940
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.455694454036881e-05,
      "loss": 0.9734,
      "step": 8960
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.452241176876856e-05,
      "loss": 0.9713,
      "step": 8980
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.4487878997168316e-05,
      "loss": 0.9813,
      "step": 9000
    },
    {
      "epoch": 1.24,
      "eval_validation_accuracy": 0.05870815828041036,
      "eval_validation_loss": 1.03515625,
      "eval_validation_runtime": 3747.9661,
      "eval_validation_samples_per_second": 0.667,
      "eval_validation_steps_per_second": 0.021,
      "step": 9000
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.445334622556807e-05,
      "loss": 0.9817,
      "step": 9020
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.441881345396782e-05,
      "loss": 0.9692,
      "step": 9040
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.438428068236757e-05,
      "loss": 0.9694,
      "step": 9060
    },
    {
      "epoch": 1.25,
      "learning_rate": 3.434974791076732e-05,
      "loss": 0.9659,
      "step": 9080
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.4315215139167074e-05,
      "loss": 0.9697,
      "step": 9100
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.4280682367566826e-05,
      "loss": 0.9747,
      "step": 9120
    },
    {
      "epoch": 1.26,
      "learning_rate": 3.424614959596657e-05,
      "loss": 0.98,
      "step": 9140
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.4211616824366325e-05,
      "loss": 0.9731,
      "step": 9160
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.417708405276608e-05,
      "loss": 0.9645,
      "step": 9180
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.414255128116583e-05,
      "loss": 0.986,
      "step": 9200
    },
    {
      "epoch": 1.27,
      "learning_rate": 3.4108018509565584e-05,
      "loss": 0.9826,
      "step": 9220
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.407348573796533e-05,
      "loss": 0.9805,
      "step": 9240
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.403895296636508e-05,
      "loss": 0.9782,
      "step": 9260
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.4004420194764835e-05,
      "loss": 0.9863,
      "step": 9280
    },
    {
      "epoch": 1.28,
      "learning_rate": 3.396988742316459e-05,
      "loss": 0.9724,
      "step": 9300
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.393535465156434e-05,
      "loss": 0.9727,
      "step": 9320
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.390082187996409e-05,
      "loss": 0.9743,
      "step": 9340
    },
    {
      "epoch": 1.29,
      "learning_rate": 3.386628910836384e-05,
      "loss": 0.9782,
      "step": 9360
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.383175633676359e-05,
      "loss": 0.9744,
      "step": 9380
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.3797223565163346e-05,
      "loss": 0.9681,
      "step": 9400
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.37626907935631e-05,
      "loss": 0.9878,
      "step": 9420
    },
    {
      "epoch": 1.3,
      "learning_rate": 3.3728158021962845e-05,
      "loss": 0.9784,
      "step": 9440
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.36936252503626e-05,
      "loss": 0.9665,
      "step": 9460
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.365909247876235e-05,
      "loss": 0.9742,
      "step": 9480
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.36245597071621e-05,
      "loss": 0.965,
      "step": 9500
    },
    {
      "epoch": 1.31,
      "learning_rate": 3.359002693556185e-05,
      "loss": 0.9742,
      "step": 9520
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.35554941639616e-05,
      "loss": 0.9744,
      "step": 9540
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.3520961392361355e-05,
      "loss": 0.9817,
      "step": 9560
    },
    {
      "epoch": 1.32,
      "learning_rate": 3.348642862076111e-05,
      "loss": 0.9752,
      "step": 9580
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.3451895849160854e-05,
      "loss": 0.9851,
      "step": 9600
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.3417363077560606e-05,
      "loss": 0.9791,
      "step": 9620
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.338283030596035e-05,
      "loss": 0.9719,
      "step": 9640
    },
    {
      "epoch": 1.33,
      "learning_rate": 3.3348297534360105e-05,
      "loss": 0.9627,
      "step": 9660
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.331376476275986e-05,
      "loss": 0.9719,
      "step": 9680
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.327923199115961e-05,
      "loss": 0.9627,
      "step": 9700
    },
    {
      "epoch": 1.34,
      "learning_rate": 3.3244699219559364e-05,
      "loss": 0.9758,
      "step": 9720
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.321016644795911e-05,
      "loss": 0.9776,
      "step": 9740
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.317563367635886e-05,
      "loss": 0.9581,
      "step": 9760
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.3141100904758616e-05,
      "loss": 0.9667,
      "step": 9780
    },
    {
      "epoch": 1.35,
      "learning_rate": 3.310656813315837e-05,
      "loss": 0.9544,
      "step": 9800
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.307203536155812e-05,
      "loss": 0.9753,
      "step": 9820
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.303750258995787e-05,
      "loss": 0.9687,
      "step": 9840
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.300296981835762e-05,
      "loss": 0.9813,
      "step": 9860
    },
    {
      "epoch": 1.36,
      "learning_rate": 3.296843704675737e-05,
      "loss": 0.9627,
      "step": 9880
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.2933904275157126e-05,
      "loss": 0.9754,
      "step": 9900
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.289937150355687e-05,
      "loss": 0.9763,
      "step": 9920
    },
    {
      "epoch": 1.37,
      "learning_rate": 3.2864838731956625e-05,
      "loss": 0.9742,
      "step": 9940
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.283030596035638e-05,
      "loss": 0.9606,
      "step": 9960
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.279577318875613e-05,
      "loss": 0.9692,
      "step": 9980
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.276124041715588e-05,
      "loss": 0.9624,
      "step": 10000
    },
    {
      "epoch": 1.38,
      "learning_rate": 3.272670764555563e-05,
      "loss": 0.9659,
      "step": 10020
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.269217487395538e-05,
      "loss": 0.9746,
      "step": 10040
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.2657642102355135e-05,
      "loss": 0.967,
      "step": 10060
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.262310933075489e-05,
      "loss": 0.974,
      "step": 10080
    },
    {
      "epoch": 1.39,
      "learning_rate": 3.258857655915464e-05,
      "loss": 0.9745,
      "step": 10100
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.2554043787554387e-05,
      "loss": 0.9803,
      "step": 10120
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.251951101595414e-05,
      "loss": 0.9752,
      "step": 10140
    },
    {
      "epoch": 1.4,
      "learning_rate": 3.248497824435389e-05,
      "loss": 0.9692,
      "step": 10160
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.2450445472753645e-05,
      "loss": 0.972,
      "step": 10180
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.24159127011534e-05,
      "loss": 0.9833,
      "step": 10200
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.2381379929553144e-05,
      "loss": 0.9744,
      "step": 10220
    },
    {
      "epoch": 1.41,
      "learning_rate": 3.23468471579529e-05,
      "loss": 0.9632,
      "step": 10240
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.231231438635265e-05,
      "loss": 0.9722,
      "step": 10260
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.22777816147524e-05,
      "loss": 0.9773,
      "step": 10280
    },
    {
      "epoch": 1.42,
      "learning_rate": 3.224324884315215e-05,
      "loss": 0.9657,
      "step": 10300
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.22087160715519e-05,
      "loss": 0.9763,
      "step": 10320
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.2174183299951654e-05,
      "loss": 0.9798,
      "step": 10340
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.213965052835141e-05,
      "loss": 0.9653,
      "step": 10360
    },
    {
      "epoch": 1.43,
      "learning_rate": 3.210511775675116e-05,
      "loss": 0.9777,
      "step": 10380
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.2070584985150906e-05,
      "loss": 0.9742,
      "step": 10400
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.203605221355066e-05,
      "loss": 0.9651,
      "step": 10420
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.200151944195041e-05,
      "loss": 0.9809,
      "step": 10440
    },
    {
      "epoch": 1.44,
      "learning_rate": 3.1966986670350164e-05,
      "loss": 0.977,
      "step": 10460
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.193245389874992e-05,
      "loss": 0.9675,
      "step": 10480
    },
    {
      "epoch": 1.45,
      "learning_rate": 3.189792112714966e-05,
      "loss": 0.968,
      "step": 10500
    },
    {
      "epoch": 1.45,
      "eval_validation_accuracy": 0.058771861260381045,
      "eval_validation_loss": 1.0263671875,
      "eval_validation_runtime": 3706.0967,
      "eval_validation_samples_per_second": 0.675,
      "eval_validation_steps_per_second": 0.021,
      "step": 10500
    },
    {
      "epoch": 1.45,
      "learning_rate": 1.3734290843806105e-05,
      "loss": 0.97,
      "step": 10520
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.366523960778898e-05,
      "loss": 0.9648,
      "step": 10540
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.3596188371771857e-05,
      "loss": 0.9569,
      "step": 10560
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.352713713575473e-05,
      "loss": 0.9731,
      "step": 10580
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.3458085899737605e-05,
      "loss": 0.9581,
      "step": 10600
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.3392487225521338e-05,
      "loss": 0.9711,
      "step": 10620
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.3323435989504214e-05,
      "loss": 0.9612,
      "step": 10640
    },
    {
      "epoch": 1.47,
      "learning_rate": 1.3254384753487087e-05,
      "loss": 0.9696,
      "step": 10660
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.3185333517469964e-05,
      "loss": 0.9738,
      "step": 10680
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.3116282281452839e-05,
      "loss": 0.9559,
      "step": 10700
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.3047231045435716e-05,
      "loss": 0.962,
      "step": 10720
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.2978179809418589e-05,
      "loss": 0.9499,
      "step": 10740
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.291258113520232e-05,
      "loss": 0.9599,
      "step": 10760
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.2843529899185197e-05,
      "loss": 0.9667,
      "step": 10780
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.2777931224968928e-05,
      "loss": 0.9513,
      "step": 10800
    },
    {
      "epoch": 1.49,
      "learning_rate": 1.2708879988951802e-05,
      "loss": 0.9456,
      "step": 10820
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.2639828752934679e-05,
      "loss": 0.9569,
      "step": 10840
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.2570777516917554e-05,
      "loss": 0.9531,
      "step": 10860
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.2501726280900427e-05,
      "loss": 0.9626,
      "step": 10880
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.2432675044883304e-05,
      "loss": 0.9502,
      "step": 10900
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.2363623808866179e-05,
      "loss": 0.9624,
      "step": 10920
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.2294572572849054e-05,
      "loss": 0.9558,
      "step": 10940
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.2225521336831929e-05,
      "loss": 0.9466,
      "step": 10960
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.2156470100814806e-05,
      "loss": 0.9497,
      "step": 10980
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.2087418864797681e-05,
      "loss": 0.9523,
      "step": 11000
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.2018367628780556e-05,
      "loss": 0.951,
      "step": 11020
    },
    {
      "epoch": 1.52,
      "learning_rate": 1.1949316392763431e-05,
      "loss": 0.9636,
      "step": 11040
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.1880265156746306e-05,
      "loss": 0.9631,
      "step": 11060
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.1811213920729181e-05,
      "loss": 0.9474,
      "step": 11080
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.1742162684712058e-05,
      "loss": 0.9698,
      "step": 11100
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.1673111448694931e-05,
      "loss": 0.9616,
      "step": 11120
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.1604060212677806e-05,
      "loss": 0.96,
      "step": 11140
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.1535008976660683e-05,
      "loss": 0.9632,
      "step": 11160
    },
    {
      "epoch": 1.54,
      "learning_rate": 1.1465957740643558e-05,
      "loss": 0.9508,
      "step": 11180
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.1396906504626433e-05,
      "loss": 0.9587,
      "step": 11200
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.1327855268609308e-05,
      "loss": 0.9466,
      "step": 11220
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.1258804032592184e-05,
      "loss": 0.9518,
      "step": 11240
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.1189752796575059e-05,
      "loss": 0.95,
      "step": 11260
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.1120701560557935e-05,
      "loss": 0.9504,
      "step": 11280
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.105165032454081e-05,
      "loss": 0.958,
      "step": 11300
    },
    {
      "epoch": 1.56,
      "learning_rate": 1.0982599088523684e-05,
      "loss": 0.9506,
      "step": 11320
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.091354785250656e-05,
      "loss": 0.9422,
      "step": 11340
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.0844496616489436e-05,
      "loss": 0.9462,
      "step": 11360
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.077544538047231e-05,
      "loss": 0.9432,
      "step": 11380
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.0706394144455187e-05,
      "loss": 0.9396,
      "step": 11400
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.0637342908438061e-05,
      "loss": 0.9434,
      "step": 11420
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.0568291672420936e-05,
      "loss": 0.954,
      "step": 11440
    },
    {
      "epoch": 1.58,
      "learning_rate": 1.0499240436403813e-05,
      "loss": 0.9513,
      "step": 11460
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.0430189200386688e-05,
      "loss": 0.9538,
      "step": 11480
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.0361137964369563e-05,
      "loss": 0.9518,
      "step": 11500
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.0292086728352438e-05,
      "loss": 0.9603,
      "step": 11520
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.0223035492335313e-05,
      "loss": 0.9567,
      "step": 11540
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.0153984256318188e-05,
      "loss": 0.9569,
      "step": 11560
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.0084933020301065e-05,
      "loss": 0.9496,
      "step": 11580
    },
    {
      "epoch": 1.6,
      "learning_rate": 1.001588178428394e-05,
      "loss": 0.9549,
      "step": 11600
    },
    {
      "epoch": 1.6,
      "learning_rate": 9.946830548266813e-06,
      "loss": 0.9547,
      "step": 11620
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.87777931224969e-06,
      "loss": 0.9597,
      "step": 11640
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.808728076232565e-06,
      "loss": 0.9566,
      "step": 11660
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.73967684021544e-06,
      "loss": 0.9474,
      "step": 11680
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.670625604198317e-06,
      "loss": 0.9548,
      "step": 11700
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.60157436818119e-06,
      "loss": 0.9529,
      "step": 11720
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.532523132164065e-06,
      "loss": 0.9418,
      "step": 11740
    },
    {
      "epoch": 1.62,
      "learning_rate": 9.463471896146942e-06,
      "loss": 0.9463,
      "step": 11760
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.394420660129817e-06,
      "loss": 0.9557,
      "step": 11780
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.325369424112692e-06,
      "loss": 0.9495,
      "step": 11800
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.256318188095567e-06,
      "loss": 0.9496,
      "step": 11820
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.187266952078443e-06,
      "loss": 0.955,
      "step": 11840
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.118215716061318e-06,
      "loss": 0.9505,
      "step": 11860
    },
    {
      "epoch": 1.64,
      "learning_rate": 9.049164480044194e-06,
      "loss": 0.9545,
      "step": 11880
    },
    {
      "epoch": 1.64,
      "learning_rate": 8.98011324402707e-06,
      "loss": 0.9602,
      "step": 11900
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.911062008009943e-06,
      "loss": 0.9466,
      "step": 11920
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.84201077199282e-06,
      "loss": 0.9423,
      "step": 11940
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.772959535975695e-06,
      "loss": 0.9469,
      "step": 11960
    },
    {
      "epoch": 1.65,
      "learning_rate": 8.70390829995857e-06,
      "loss": 0.9347,
      "step": 11980
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.634857063941445e-06,
      "loss": 0.9444,
      "step": 12000
    },
    {
      "epoch": 1.66,
      "eval_validation_accuracy": 0.059099560332193456,
      "eval_validation_loss": 1.0048828125,
      "eval_validation_runtime": 3761.2942,
      "eval_validation_samples_per_second": 0.665,
      "eval_validation_steps_per_second": 0.021,
      "step": 12000
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.56580582792432e-06,
      "loss": 0.9547,
      "step": 12020
    },
    {
      "epoch": 1.66,
      "learning_rate": 8.496754591907195e-06,
      "loss": 0.9516,
      "step": 12040
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.42770335589007e-06,
      "loss": 0.9476,
      "step": 12060
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.358652119872947e-06,
      "loss": 0.9455,
      "step": 12080
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.289600883855822e-06,
      "loss": 0.9453,
      "step": 12100
    },
    {
      "epoch": 1.67,
      "learning_rate": 8.220549647838697e-06,
      "loss": 0.9336,
      "step": 12120
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.151498411821572e-06,
      "loss": 0.9553,
      "step": 12140
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.082447175804447e-06,
      "loss": 0.9419,
      "step": 12160
    },
    {
      "epoch": 1.68,
      "learning_rate": 8.013395939787322e-06,
      "loss": 0.938,
      "step": 12180
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.944344703770199e-06,
      "loss": 0.9498,
      "step": 12200
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.875293467753072e-06,
      "loss": 0.9559,
      "step": 12220
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.806242231735947e-06,
      "loss": 0.9426,
      "step": 12240
    },
    {
      "epoch": 1.69,
      "learning_rate": 7.737190995718824e-06,
      "loss": 0.9384,
      "step": 12260
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.6681397597017e-06,
      "loss": 0.9393,
      "step": 12280
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.599088523684575e-06,
      "loss": 0.947,
      "step": 12300
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.530037287667449e-06,
      "loss": 0.9624,
      "step": 12320
    },
    {
      "epoch": 1.7,
      "learning_rate": 7.4609860516503245e-06,
      "loss": 0.9476,
      "step": 12340
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.3919348156332e-06,
      "loss": 0.9529,
      "step": 12360
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.3228835796160755e-06,
      "loss": 0.9414,
      "step": 12380
    },
    {
      "epoch": 1.71,
      "learning_rate": 7.253832343598951e-06,
      "loss": 0.9393,
      "step": 12400
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.184781107581826e-06,
      "loss": 0.9482,
      "step": 12420
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.115729871564701e-06,
      "loss": 0.9416,
      "step": 12440
    },
    {
      "epoch": 1.72,
      "learning_rate": 7.046678635547577e-06,
      "loss": 0.9353,
      "step": 12460
    },
    {
      "epoch": 1.72,
      "learning_rate": 6.9776273995304526e-06,
      "loss": 0.941,
      "step": 12480
    },
    {
      "epoch": 1.73,
      "learning_rate": 6.908576163513328e-06,
      "loss": 0.9452,
      "step": 12500
    },
    {
      "epoch": 1.73,
      "learning_rate": 6.839524927496202e-06,
      "loss": 0.9494,
      "step": 12520
    },
    {
      "epoch": 1.73,
      "learning_rate": 6.770473691479078e-06,
      "loss": 0.9493,
      "step": 12540
    },
    {
      "epoch": 1.73,
      "learning_rate": 6.701422455461953e-06,
      "loss": 0.943,
      "step": 12560
    },
    {
      "epoch": 1.74,
      "learning_rate": 6.632371219444829e-06,
      "loss": 0.9486,
      "step": 12580
    },
    {
      "epoch": 1.74,
      "learning_rate": 6.563319983427704e-06,
      "loss": 0.9506,
      "step": 12600
    },
    {
      "epoch": 1.74,
      "learning_rate": 6.494268747410578e-06,
      "loss": 0.9474,
      "step": 12620
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.425217511393454e-06,
      "loss": 0.9442,
      "step": 12640
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.35616627537633e-06,
      "loss": 0.9527,
      "step": 12660
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.287115039359205e-06,
      "loss": 0.9332,
      "step": 12680
    },
    {
      "epoch": 1.75,
      "learning_rate": 6.21806380334208e-06,
      "loss": 0.943,
      "step": 12700
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.149012567324956e-06,
      "loss": 0.9441,
      "step": 12720
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.07996133130783e-06,
      "loss": 0.9359,
      "step": 12740
    },
    {
      "epoch": 1.76,
      "learning_rate": 6.010910095290706e-06,
      "loss": 0.9368,
      "step": 12760
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.941858859273581e-06,
      "loss": 0.9466,
      "step": 12780
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.872807623256456e-06,
      "loss": 0.9447,
      "step": 12800
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.803756387239332e-06,
      "loss": 0.9455,
      "step": 12820
    },
    {
      "epoch": 1.77,
      "learning_rate": 5.734705151222207e-06,
      "loss": 0.9436,
      "step": 12840
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.665653915205082e-06,
      "loss": 0.9425,
      "step": 12860
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.596602679187958e-06,
      "loss": 0.9377,
      "step": 12880
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.5275514431708325e-06,
      "loss": 0.9464,
      "step": 12900
    },
    {
      "epoch": 1.78,
      "learning_rate": 5.4585002071537085e-06,
      "loss": 0.9432,
      "step": 12920
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.3894489711365835e-06,
      "loss": 0.9326,
      "step": 12940
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.320397735119459e-06,
      "loss": 0.9413,
      "step": 12960
    },
    {
      "epoch": 1.79,
      "learning_rate": 5.2513464991023345e-06,
      "loss": 0.9353,
      "step": 12980
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.18229526308521e-06,
      "loss": 0.9483,
      "step": 13000
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.116696588868941e-06,
      "loss": 0.934,
      "step": 13020
    },
    {
      "epoch": 1.8,
      "learning_rate": 5.047645352851816e-06,
      "loss": 0.9493,
      "step": 13040
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.978594116834691e-06,
      "loss": 0.946,
      "step": 13060
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.909542880817567e-06,
      "loss": 0.9425,
      "step": 13080
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.840491644800442e-06,
      "loss": 0.9464,
      "step": 13100
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.771440408783317e-06,
      "loss": 0.9347,
      "step": 13120
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.702389172766192e-06,
      "loss": 0.9393,
      "step": 13140
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.633337936749068e-06,
      "loss": 0.9484,
      "step": 13160
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.564286700731943e-06,
      "loss": 0.9284,
      "step": 13180
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.495235464714818e-06,
      "loss": 0.9433,
      "step": 13200
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.426184228697694e-06,
      "loss": 0.9414,
      "step": 13220
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.3571329926805685e-06,
      "loss": 0.9464,
      "step": 13240
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.2880817566634444e-06,
      "loss": 0.9395,
      "step": 13260
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.21903052064632e-06,
      "loss": 0.9437,
      "step": 13280
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.149979284629195e-06,
      "loss": 0.9348,
      "step": 13300
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.084380610412927e-06,
      "loss": 0.9384,
      "step": 13320
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.015329374395802e-06,
      "loss": 0.9448,
      "step": 13340
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.946278138378677e-06,
      "loss": 0.9334,
      "step": 13360
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.877226902361553e-06,
      "loss": 0.9362,
      "step": 13380
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.8081756663444276e-06,
      "loss": 0.9433,
      "step": 13400
    },
    {
      "epoch": 1.85,
      "learning_rate": 3.739124430327303e-06,
      "loss": 0.9558,
      "step": 13420
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.6700731943101786e-06,
      "loss": 0.9451,
      "step": 13440
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.6044745200939096e-06,
      "loss": 0.9362,
      "step": 13460
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.535423284076785e-06,
      "loss": 0.9374,
      "step": 13480
    },
    {
      "epoch": 1.86,
      "learning_rate": 3.46637204805966e-06,
      "loss": 0.9326,
      "step": 13500
    },
    {
      "epoch": 1.86,
      "eval_validation_accuracy": 0.05914255007327797,
      "eval_validation_loss": 0.9990234375,
      "eval_validation_runtime": 3709.9254,
      "eval_validation_samples_per_second": 0.674,
      "eval_validation_steps_per_second": 0.021,
      "step": 13500
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.3973208120425356e-06,
      "loss": 0.9416,
      "step": 13520
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.328269576025411e-06,
      "loss": 0.9395,
      "step": 13540
    },
    {
      "epoch": 1.87,
      "learning_rate": 3.259218340008286e-06,
      "loss": 0.9436,
      "step": 13560
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.1901671039911617e-06,
      "loss": 0.9387,
      "step": 13580
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.1211158679740368e-06,
      "loss": 0.9475,
      "step": 13600
    },
    {
      "epoch": 1.88,
      "learning_rate": 3.0520646319569123e-06,
      "loss": 0.9339,
      "step": 13620
    },
    {
      "epoch": 1.88,
      "learning_rate": 2.9830133959397874e-06,
      "loss": 0.9387,
      "step": 13640
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.913962159922663e-06,
      "loss": 0.94,
      "step": 13660
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.844910923905538e-06,
      "loss": 0.9323,
      "step": 13680
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.7758596878884134e-06,
      "loss": 0.9402,
      "step": 13700
    },
    {
      "epoch": 1.89,
      "learning_rate": 2.7068084518712885e-06,
      "loss": 0.9363,
      "step": 13720
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.6377572158541636e-06,
      "loss": 0.9501,
      "step": 13740
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.5687059798370395e-06,
      "loss": 0.9503,
      "step": 13760
    },
    {
      "epoch": 1.9,
      "learning_rate": 2.4996547438199146e-06,
      "loss": 0.938,
      "step": 13780
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.4306035078027896e-06,
      "loss": 0.9322,
      "step": 13800
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.361552271785665e-06,
      "loss": 0.945,
      "step": 13820
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.2925010357685402e-06,
      "loss": 0.9415,
      "step": 13840
    },
    {
      "epoch": 1.91,
      "learning_rate": 2.2234497997514157e-06,
      "loss": 0.9299,
      "step": 13860
    },
    {
      "epoch": 1.92,
      "learning_rate": 2.1543985637342912e-06,
      "loss": 0.9413,
      "step": 13880
    },
    {
      "epoch": 1.92,
      "learning_rate": 2.0853473277171663e-06,
      "loss": 0.9463,
      "step": 13900
    },
    {
      "epoch": 1.92,
      "learning_rate": 2.0162960917000414e-06,
      "loss": 0.9393,
      "step": 13920
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.947244855682917e-06,
      "loss": 0.9352,
      "step": 13940
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.8781936196657922e-06,
      "loss": 0.9319,
      "step": 13960
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.8091423836486674e-06,
      "loss": 0.9436,
      "step": 13980
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.7400911476315427e-06,
      "loss": 0.9393,
      "step": 14000
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.671039911614418e-06,
      "loss": 0.9338,
      "step": 14020
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.601988675597293e-06,
      "loss": 0.9431,
      "step": 14040
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.5329374395801686e-06,
      "loss": 0.9462,
      "step": 14060
    },
    {
      "epoch": 1.94,
      "learning_rate": 1.4638862035630439e-06,
      "loss": 0.9487,
      "step": 14080
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.3948349675459192e-06,
      "loss": 0.9335,
      "step": 14100
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.3257837315287944e-06,
      "loss": 0.9268,
      "step": 14120
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.2567324955116697e-06,
      "loss": 0.9389,
      "step": 14140
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.187681259494545e-06,
      "loss": 0.9378,
      "step": 14160
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.1186300234774203e-06,
      "loss": 0.9277,
      "step": 14180
    },
    {
      "epoch": 1.96,
      "learning_rate": 1.0495787874602956e-06,
      "loss": 0.9347,
      "step": 14200
    },
    {
      "epoch": 1.96,
      "learning_rate": 9.805275514431709e-07,
      "loss": 0.9425,
      "step": 14220
    },
    {
      "epoch": 1.97,
      "learning_rate": 9.114763154260462e-07,
      "loss": 0.9396,
      "step": 14240
    },
    {
      "epoch": 1.97,
      "learning_rate": 8.424250794089215e-07,
      "loss": 0.939,
      "step": 14260
    },
    {
      "epoch": 1.97,
      "learning_rate": 7.733738433917967e-07,
      "loss": 0.9398,
      "step": 14280
    },
    {
      "epoch": 1.98,
      "learning_rate": 7.04322607374672e-07,
      "loss": 0.9339,
      "step": 14300
    },
    {
      "epoch": 1.98,
      "learning_rate": 6.352713713575473e-07,
      "loss": 0.932,
      "step": 14320
    },
    {
      "epoch": 1.98,
      "learning_rate": 5.662201353404226e-07,
      "loss": 0.9349,
      "step": 14340
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.971688993232979e-07,
      "loss": 0.9355,
      "step": 14360
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.281176633061732e-07,
      "loss": 0.9358,
      "step": 14380
    },
    {
      "epoch": 1.99,
      "learning_rate": 3.590664272890485e-07,
      "loss": 0.9347,
      "step": 14400
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.9001519127192375e-07,
      "loss": 0.9311,
      "step": 14420
    },
    {
      "epoch": 1.99,
      "learning_rate": 2.2096395525479907e-07,
      "loss": 0.9445,
      "step": 14440
    },
    {
      "epoch": 2.0,
      "learning_rate": 1.5191271923767435e-07,
      "loss": 0.9501,
      "step": 14460
    },
    {
      "epoch": 2.0,
      "learning_rate": 8.286148322054965e-08,
      "loss": 0.9385,
      "step": 14480
    },
    {
      "epoch": 2.0,
      "step": 14484,
      "total_flos": 3.0144649130395304e+19,
      "train_loss": 0.26028791997551887,
      "train_runtime": 64104.6589,
      "train_samples_per_second": 14.46,
      "train_steps_per_second": 0.226
    }
  ],
  "max_steps": 14484,
  "num_train_epochs": 3,
  "total_flos": 3.0144649130395304e+19,
  "trial_name": null,
  "trial_params": null
}