{
  "best_metric": 0.8906223955728906,
  "best_model_checkpoint": "save_model/flan-t5-xl-twitter-sentiment-analysis-lora/checkpoint-149990",
  "epoch": 10.0,
  "global_step": 149990,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.03,
      "learning_rate": 4.983332222148143e-05,
      "loss": 0.179,
      "step": 500
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.966664444296287e-05,
      "loss": 0.1691,
      "step": 1000
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.94999666644443e-05,
      "loss": 0.1768,
      "step": 1500
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.933328888592573e-05,
      "loss": 0.1731,
      "step": 2000
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.916661110740716e-05,
      "loss": 0.1688,
      "step": 2500
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.89999333288886e-05,
      "loss": 0.1695,
      "step": 3000
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.883325555037003e-05,
      "loss": 0.1619,
      "step": 3500
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.866657777185146e-05,
      "loss": 0.1634,
      "step": 4000
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.8499899993332895e-05,
      "loss": 0.1649,
      "step": 4500
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.8333222214814325e-05,
      "loss": 0.164,
      "step": 5000
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.8166544436295755e-05,
      "loss": 0.1624,
      "step": 5500
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.7999866657777186e-05,
      "loss": 0.1577,
      "step": 6000
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.7833188879258616e-05,
      "loss": 0.166,
      "step": 6500
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.766651110074005e-05,
      "loss": 0.1524,
      "step": 7000
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.749983332222148e-05,
      "loss": 0.1654,
      "step": 7500
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.733315554370292e-05,
      "loss": 0.1548,
      "step": 8000
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.716647776518435e-05,
      "loss": 0.1584,
      "step": 8500
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.699979998666578e-05,
      "loss": 0.1615,
      "step": 9000
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.683312220814721e-05,
      "loss": 0.1555,
      "step": 9500
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.666644442962864e-05,
      "loss": 0.161,
      "step": 10000
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.649976665111008e-05,
      "loss": 0.156,
      "step": 10500
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.633308887259151e-05,
      "loss": 0.1506,
      "step": 11000
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6166411094072945e-05,
      "loss": 0.1476,
      "step": 11500
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.599973331555437e-05,
      "loss": 0.1466,
      "step": 12000
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.5833055537035806e-05,
      "loss": 0.1497,
      "step": 12500
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.5666377758517236e-05,
      "loss": 0.158,
      "step": 13000
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.5499699979998666e-05,
      "loss": 0.1489,
      "step": 13500
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.53330222014801e-05,
      "loss": 0.155,
      "step": 14000
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.5166344422961533e-05,
      "loss": 0.1546,
      "step": 14500
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.8832883288328833,
      "eval_loss": 0.14019545912742615,
      "eval_runtime": 1811.1495,
      "eval_samples_per_second": 16.562,
      "eval_steps_per_second": 2.071,
      "step": 14999
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.499966664444297e-05,
      "loss": 0.1556,
      "step": 15000
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.4832988865924394e-05,
      "loss": 0.1564,
      "step": 15500
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.4666311087405824e-05,
      "loss": 0.1462,
      "step": 16000
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.449963330888726e-05,
      "loss": 0.1475,
      "step": 16500
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.433295553036869e-05,
      "loss": 0.1514,
      "step": 17000
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.416627775185013e-05,
      "loss": 0.1506,
      "step": 17500
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.399959997333156e-05,
      "loss": 0.1478,
      "step": 18000
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.383292219481299e-05,
      "loss": 0.144,
      "step": 18500
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.366624441629442e-05,
      "loss": 0.1485,
      "step": 19000
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.349956663777585e-05,
      "loss": 0.1449,
      "step": 19500
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.3332888859257286e-05,
      "loss": 0.1421,
      "step": 20000
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.316621108073872e-05,
      "loss": 0.1452,
      "step": 20500
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.2999533302220154e-05,
      "loss": 0.1384,
      "step": 21000
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.2832855523701584e-05,
      "loss": 0.1489,
      "step": 21500
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.2666177745183014e-05,
      "loss": 0.1486,
      "step": 22000
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.2499499966664444e-05,
      "loss": 0.1498,
      "step": 22500
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.2332822188145875e-05,
      "loss": 0.1478,
      "step": 23000
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.216614440962731e-05,
      "loss": 0.1425,
      "step": 23500
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.199946663110874e-05,
      "loss": 0.1518,
      "step": 24000
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.183278885259018e-05,
      "loss": 0.1474,
      "step": 24500
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.166611107407161e-05,
      "loss": 0.1462,
      "step": 25000
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.149943329555304e-05,
      "loss": 0.1477,
      "step": 25500
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.133275551703447e-05,
      "loss": 0.1361,
      "step": 26000
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.11660777385159e-05,
      "loss": 0.1429,
      "step": 26500
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.099939995999734e-05,
      "loss": 0.1388,
      "step": 27000
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.083272218147877e-05,
      "loss": 0.1416,
      "step": 27500
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.0666044402960204e-05,
      "loss": 0.1378,
      "step": 28000
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.0499366624441634e-05,
      "loss": 0.1511,
      "step": 28500
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.0332688845923065e-05,
      "loss": 0.1408,
      "step": 29000
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.0166011067404495e-05,
      "loss": 0.1539,
      "step": 29500
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.8872220555388872,
      "eval_loss": 0.13637296855449677,
      "eval_runtime": 1812.4601,
      "eval_samples_per_second": 16.55,
      "eval_steps_per_second": 2.069,
      "step": 29998
    },
    {
      "epoch": 2.0,
      "learning_rate": 3.9999333288885925e-05,
      "loss": 0.1394,
      "step": 30000
    },
    {
      "epoch": 2.03,
      "learning_rate": 3.983265551036736e-05,
      "loss": 0.1395,
      "step": 30500
    },
    {
      "epoch": 2.07,
      "learning_rate": 3.966597773184879e-05,
      "loss": 0.1401,
      "step": 31000
    },
    {
      "epoch": 2.1,
      "learning_rate": 3.949929995333023e-05,
      "loss": 0.1335,
      "step": 31500
    },
    {
      "epoch": 2.13,
      "learning_rate": 3.933262217481166e-05,
      "loss": 0.1439,
      "step": 32000
    },
    {
      "epoch": 2.17,
      "learning_rate": 3.916594439629308e-05,
      "loss": 0.1369,
      "step": 32500
    },
    {
      "epoch": 2.2,
      "learning_rate": 3.899926661777452e-05,
      "loss": 0.1318,
      "step": 33000
    },
    {
      "epoch": 2.23,
      "learning_rate": 3.883258883925595e-05,
      "loss": 0.1441,
      "step": 33500
    },
    {
      "epoch": 2.27,
      "learning_rate": 3.866591106073739e-05,
      "loss": 0.1398,
      "step": 34000
    },
    {
      "epoch": 2.3,
      "learning_rate": 3.849923328221882e-05,
      "loss": 0.1422,
      "step": 34500
    },
    {
      "epoch": 2.33,
      "learning_rate": 3.833255550370025e-05,
      "loss": 0.1418,
      "step": 35000
    },
    {
      "epoch": 2.37,
      "learning_rate": 3.816587772518168e-05,
      "loss": 0.1371,
      "step": 35500
    },
    {
      "epoch": 2.4,
      "learning_rate": 3.799919994666311e-05,
      "loss": 0.1417,
      "step": 36000
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.7832522168144545e-05,
      "loss": 0.14,
      "step": 36500
    },
    {
      "epoch": 2.47,
      "learning_rate": 3.7665844389625975e-05,
      "loss": 0.1411,
      "step": 37000
    },
    {
      "epoch": 2.5,
      "learning_rate": 3.749916661110741e-05,
      "loss": 0.1406,
      "step": 37500
    },
    {
      "epoch": 2.53,
      "learning_rate": 3.733248883258884e-05,
      "loss": 0.1437,
      "step": 38000
    },
    {
      "epoch": 2.57,
      "learning_rate": 3.716581105407027e-05,
      "loss": 0.1378,
      "step": 38500
    },
    {
      "epoch": 2.6,
      "learning_rate": 3.69991332755517e-05,
      "loss": 0.1432,
      "step": 39000
    },
    {
      "epoch": 2.63,
      "learning_rate": 3.683245549703313e-05,
      "loss": 0.1383,
      "step": 39500
    },
    {
      "epoch": 2.67,
      "learning_rate": 3.666577771851457e-05,
      "loss": 0.1385,
      "step": 40000
    },
    {
      "epoch": 2.7,
      "learning_rate": 3.6499099939996e-05,
      "loss": 0.1402,
      "step": 40500
    },
    {
      "epoch": 2.73,
      "learning_rate": 3.633242216147744e-05,
      "loss": 0.1272,
      "step": 41000
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.616574438295887e-05,
      "loss": 0.1457,
      "step": 41500
    },
    {
      "epoch": 2.8,
      "learning_rate": 3.59990666044403e-05,
      "loss": 0.1306,
      "step": 42000
    },
    {
      "epoch": 2.83,
      "learning_rate": 3.583238882592173e-05,
      "loss": 0.1335,
      "step": 42500
    },
    {
      "epoch": 2.87,
      "learning_rate": 3.566571104740316e-05,
      "loss": 0.1422,
      "step": 43000
    },
    {
      "epoch": 2.9,
      "learning_rate": 3.5499033268884596e-05,
      "loss": 0.1322,
      "step": 43500
    },
    {
      "epoch": 2.93,
      "learning_rate": 3.5332355490366026e-05,
      "loss": 0.1317,
      "step": 44000
    },
    {
      "epoch": 2.97,
      "learning_rate": 3.5165677711847456e-05,
      "loss": 0.1393,
      "step": 44500
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.8885555222188886,
      "eval_loss": 0.13739028573036194,
      "eval_runtime": 1811.3641,
      "eval_samples_per_second": 16.56,
      "eval_steps_per_second": 2.07,
      "step": 44997
    },
    {
      "epoch": 3.0,
      "learning_rate": 3.499899993332889e-05,
      "loss": 0.1306,
      "step": 45000
    },
    {
      "epoch": 3.03,
      "learning_rate": 3.483232215481032e-05,
      "loss": 0.131,
      "step": 45500
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.4665644376291754e-05,
      "loss": 0.1323,
      "step": 46000
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.4498966597773184e-05,
      "loss": 0.1308,
      "step": 46500
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.433228881925462e-05,
      "loss": 0.1376,
      "step": 47000
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.416561104073605e-05,
      "loss": 0.1229,
      "step": 47500
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.399893326221748e-05,
      "loss": 0.1384,
      "step": 48000
    },
    {
      "epoch": 3.23,
      "learning_rate": 3.383225548369892e-05,
      "loss": 0.1338,
      "step": 48500
    },
    {
      "epoch": 3.27,
      "learning_rate": 3.366557770518035e-05,
      "loss": 0.1255,
      "step": 49000
    },
    {
      "epoch": 3.3,
      "learning_rate": 3.349889992666178e-05,
      "loss": 0.1317,
      "step": 49500
    },
    {
      "epoch": 3.33,
      "learning_rate": 3.333222214814321e-05,
      "loss": 0.1327,
      "step": 50000
    },
    {
      "epoch": 3.37,
      "learning_rate": 3.3165544369624646e-05,
      "loss": 0.1308,
      "step": 50500
    },
    {
      "epoch": 3.4,
      "learning_rate": 3.2998866591106076e-05,
      "loss": 0.1252,
      "step": 51000
    },
    {
      "epoch": 3.43,
      "learning_rate": 3.2832188812587506e-05,
      "loss": 0.1356,
      "step": 51500
    },
    {
      "epoch": 3.47,
      "learning_rate": 3.2665511034068943e-05,
      "loss": 0.1408,
      "step": 52000
    },
    {
      "epoch": 3.5,
      "learning_rate": 3.2498833255550374e-05,
      "loss": 0.1322,
      "step": 52500
    },
    {
      "epoch": 3.53,
      "learning_rate": 3.2332155477031804e-05,
      "loss": 0.1362,
      "step": 53000
    },
    {
      "epoch": 3.57,
      "learning_rate": 3.2165477698513234e-05,
      "loss": 0.1269,
      "step": 53500
    },
    {
      "epoch": 3.6,
      "learning_rate": 3.1998799919994664e-05,
      "loss": 0.1357,
      "step": 54000
    },
    {
      "epoch": 3.63,
      "learning_rate": 3.18321221414761e-05,
      "loss": 0.133,
      "step": 54500
    },
    {
      "epoch": 3.67,
      "learning_rate": 3.166544436295753e-05,
      "loss": 0.1348,
      "step": 55000
    },
    {
      "epoch": 3.7,
      "learning_rate": 3.149876658443897e-05,
      "loss": 0.1322,
      "step": 55500
    },
    {
      "epoch": 3.73,
      "learning_rate": 3.133208880592039e-05,
      "loss": 0.1349,
      "step": 56000
    },
    {
      "epoch": 3.77,
      "learning_rate": 3.116541102740183e-05,
      "loss": 0.1398,
      "step": 56500
    },
    {
      "epoch": 3.8,
      "learning_rate": 3.099873324888326e-05,
      "loss": 0.1331,
      "step": 57000
    },
    {
      "epoch": 3.83,
      "learning_rate": 3.083205547036469e-05,
      "loss": 0.135,
      "step": 57500
    },
    {
      "epoch": 3.87,
      "learning_rate": 3.066537769184613e-05,
      "loss": 0.1346,
      "step": 58000
    },
    {
      "epoch": 3.9,
      "learning_rate": 3.0498699913327557e-05,
      "loss": 0.13,
      "step": 58500
    },
    {
      "epoch": 3.93,
      "learning_rate": 3.033202213480899e-05,
      "loss": 0.1283,
      "step": 59000
    },
    {
      "epoch": 3.97,
      "learning_rate": 3.016534435629042e-05,
      "loss": 0.1392,
      "step": 59500
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.8873554022068874,
      "eval_loss": 0.1396227777004242,
      "eval_runtime": 1812.658,
      "eval_samples_per_second": 16.549,
      "eval_steps_per_second": 2.069,
      "step": 59996
    },
    {
      "epoch": 4.0,
      "learning_rate": 2.9998666577771854e-05,
      "loss": 0.1335,
      "step": 60000
    },
    {
      "epoch": 4.03,
      "learning_rate": 2.9831988799253285e-05,
      "loss": 0.1241,
      "step": 60500
    },
    {
      "epoch": 4.07,
      "learning_rate": 2.9665311020734715e-05,
      "loss": 0.1267,
      "step": 61000
    },
    {
      "epoch": 4.1,
      "learning_rate": 2.9498633242216152e-05,
      "loss": 0.1263,
      "step": 61500
    },
    {
      "epoch": 4.13,
      "learning_rate": 2.933195546369758e-05,
      "loss": 0.1183,
      "step": 62000
    },
    {
      "epoch": 4.17,
      "learning_rate": 2.9165277685179016e-05,
      "loss": 0.1317,
      "step": 62500
    },
    {
      "epoch": 4.2,
      "learning_rate": 2.8998599906660446e-05,
      "loss": 0.1298,
      "step": 63000
    },
    {
      "epoch": 4.23,
      "learning_rate": 2.8831922128141876e-05,
      "loss": 0.1302,
      "step": 63500
    },
    {
      "epoch": 4.27,
      "learning_rate": 2.866524434962331e-05,
      "loss": 0.1264,
      "step": 64000
    },
    {
      "epoch": 4.3,
      "learning_rate": 2.849856657110474e-05,
      "loss": 0.1328,
      "step": 64500
    },
    {
      "epoch": 4.33,
      "learning_rate": 2.8331888792586177e-05,
      "loss": 0.1376,
      "step": 65000
    },
    {
      "epoch": 4.37,
      "learning_rate": 2.8165211014067604e-05,
      "loss": 0.1289,
      "step": 65500
    },
    {
      "epoch": 4.4,
      "learning_rate": 2.799853323554904e-05,
      "loss": 0.1291,
      "step": 66000
    },
    {
      "epoch": 4.43,
      "learning_rate": 2.783185545703047e-05,
      "loss": 0.1299,
      "step": 66500
    },
    {
      "epoch": 4.47,
      "learning_rate": 2.76651776785119e-05,
      "loss": 0.1305,
      "step": 67000
    },
    {
      "epoch": 4.5,
      "learning_rate": 2.7498499899993335e-05,
      "loss": 0.1282,
      "step": 67500
    },
    {
      "epoch": 4.53,
      "learning_rate": 2.7331822121474765e-05,
      "loss": 0.13,
      "step": 68000
    },
    {
      "epoch": 4.57,
      "learning_rate": 2.71651443429562e-05,
      "loss": 0.1259,
      "step": 68500
    },
    {
      "epoch": 4.6,
      "learning_rate": 2.699846656443763e-05,
      "loss": 0.1131,
      "step": 69000
    },
    {
      "epoch": 4.63,
      "learning_rate": 2.6831788785919066e-05,
      "loss": 0.1252,
      "step": 69500
    },
    {
      "epoch": 4.67,
      "learning_rate": 2.6665111007400496e-05,
      "loss": 0.124,
      "step": 70000
    },
    {
      "epoch": 4.7,
      "learning_rate": 2.6498433228881923e-05,
      "loss": 0.1325,
      "step": 70500
    },
    {
      "epoch": 4.73,
      "learning_rate": 2.633175545036336e-05,
      "loss": 0.1387,
      "step": 71000
    },
    {
      "epoch": 4.77,
      "learning_rate": 2.616507767184479e-05,
      "loss": 0.1316,
      "step": 71500
    },
    {
      "epoch": 4.8,
      "learning_rate": 2.5998399893326224e-05,
      "loss": 0.1245,
      "step": 72000
    },
    {
      "epoch": 4.83,
      "learning_rate": 2.5831722114807654e-05,
      "loss": 0.1329,
      "step": 72500
    },
    {
      "epoch": 4.87,
      "learning_rate": 2.5665044336289085e-05,
      "loss": 0.1286,
      "step": 73000
    },
    {
      "epoch": 4.9,
      "learning_rate": 2.549836655777052e-05,
      "loss": 0.1311,
      "step": 73500
    },
    {
      "epoch": 4.93,
      "learning_rate": 2.533168877925195e-05,
      "loss": 0.1322,
      "step": 74000
    },
    {
      "epoch": 4.97,
      "learning_rate": 2.5165011000733385e-05,
      "loss": 0.1299,
      "step": 74500
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.8897223055638898,
      "eval_loss": 0.1437413990497589,
      "eval_runtime": 1810.9994,
      "eval_samples_per_second": 16.564,
      "eval_steps_per_second": 2.071,
      "step": 74995
    },
    {
      "epoch": 5.0,
      "learning_rate": 2.4998333222214816e-05,
      "loss": 0.1272,
      "step": 75000
    },
    {
      "epoch": 5.03,
      "learning_rate": 2.4831655443696246e-05,
      "loss": 0.1287,
      "step": 75500
    },
    {
      "epoch": 5.07,
      "learning_rate": 2.466497766517768e-05,
      "loss": 0.1251,
      "step": 76000
    },
    {
      "epoch": 5.1,
      "learning_rate": 2.4498299886659113e-05,
      "loss": 0.1224,
      "step": 76500
    },
    {
      "epoch": 5.13,
      "learning_rate": 2.4331622108140543e-05,
      "loss": 0.1238,
      "step": 77000
    },
    {
      "epoch": 5.17,
      "learning_rate": 2.4164944329621974e-05,
      "loss": 0.1245,
      "step": 77500
    },
    {
      "epoch": 5.2,
      "learning_rate": 2.3998266551103407e-05,
      "loss": 0.1138,
      "step": 78000
    },
    {
      "epoch": 5.23,
      "learning_rate": 2.383158877258484e-05,
      "loss": 0.1248,
      "step": 78500
    },
    {
      "epoch": 5.27,
      "learning_rate": 2.366491099406627e-05,
      "loss": 0.1156,
      "step": 79000
    },
    {
      "epoch": 5.3,
      "learning_rate": 2.3498233215547705e-05,
      "loss": 0.1253,
      "step": 79500
    },
    {
      "epoch": 5.33,
      "learning_rate": 2.333155543702914e-05,
      "loss": 0.1274,
      "step": 80000
    },
    {
      "epoch": 5.37,
      "learning_rate": 2.316487765851057e-05,
      "loss": 0.1289,
      "step": 80500
    },
    {
      "epoch": 5.4,
      "learning_rate": 2.2998199879992e-05,
      "loss": 0.1285,
      "step": 81000
    },
    {
      "epoch": 5.43,
      "learning_rate": 2.2831522101473432e-05,
      "loss": 0.1245,
      "step": 81500
    },
    {
      "epoch": 5.47,
      "learning_rate": 2.2664844322954866e-05,
      "loss": 0.1286,
      "step": 82000
    },
    {
      "epoch": 5.5,
      "learning_rate": 2.2498166544436296e-05,
      "loss": 0.1255,
      "step": 82500
    },
    {
      "epoch": 5.53,
      "learning_rate": 2.233148876591773e-05,
      "loss": 0.128,
      "step": 83000
    },
    {
      "epoch": 5.57,
      "learning_rate": 2.2164810987399164e-05,
      "loss": 0.1326,
      "step": 83500
    },
    {
      "epoch": 5.6,
      "learning_rate": 2.199813320888059e-05,
      "loss": 0.1191,
      "step": 84000
    },
    {
      "epoch": 5.63,
      "learning_rate": 2.1831455430362024e-05,
      "loss": 0.129,
      "step": 84500
    },
    {
      "epoch": 5.67,
      "learning_rate": 2.1664777651843458e-05,
      "loss": 0.1262,
      "step": 85000
    },
    {
      "epoch": 5.7,
      "learning_rate": 2.1498099873324888e-05,
      "loss": 0.1252,
      "step": 85500
    },
    {
      "epoch": 5.73,
      "learning_rate": 2.133142209480632e-05,
      "loss": 0.1279,
      "step": 86000
    },
    {
      "epoch": 5.77,
      "learning_rate": 2.1164744316287755e-05,
      "loss": 0.1271,
      "step": 86500
    },
    {
      "epoch": 5.8,
      "learning_rate": 2.0998066537769185e-05,
      "loss": 0.1198,
      "step": 87000
    },
    {
      "epoch": 5.83,
      "learning_rate": 2.0831388759250616e-05,
      "loss": 0.1267,
      "step": 87500
    },
    {
      "epoch": 5.87,
      "learning_rate": 2.066471098073205e-05,
      "loss": 0.1184,
      "step": 88000
    },
    {
      "epoch": 5.9,
      "learning_rate": 2.0498033202213483e-05,
      "loss": 0.1201,
      "step": 88500
    },
    {
      "epoch": 5.93,
      "learning_rate": 2.0331355423694913e-05,
      "loss": 0.1186,
      "step": 89000
    },
    {
      "epoch": 5.97,
      "learning_rate": 2.0164677645176347e-05,
      "loss": 0.1315,
      "step": 89500
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.8873887388738874,
      "eval_loss": 0.14759084582328796,
      "eval_runtime": 1810.7596,
      "eval_samples_per_second": 16.566,
      "eval_steps_per_second": 2.071,
      "step": 89994
    },
    {
      "epoch": 6.0,
      "learning_rate": 1.999799986665778e-05,
      "loss": 0.1245,
      "step": 90000
    },
    {
      "epoch": 6.03,
      "learning_rate": 1.983132208813921e-05,
      "loss": 0.1145,
      "step": 90500
    },
    {
      "epoch": 6.07,
      "learning_rate": 1.966464430962064e-05,
      "loss": 0.1239,
      "step": 91000
    },
    {
      "epoch": 6.1,
      "learning_rate": 1.9497966531102074e-05,
      "loss": 0.1159,
      "step": 91500
    },
    {
      "epoch": 6.13,
      "learning_rate": 1.9331288752583508e-05,
      "loss": 0.1243,
      "step": 92000
    },
    {
      "epoch": 6.17,
      "learning_rate": 1.9164610974064938e-05,
      "loss": 0.1178,
      "step": 92500
    },
    {
      "epoch": 6.2,
      "learning_rate": 1.8997933195546372e-05,
      "loss": 0.1214,
      "step": 93000
    },
    {
      "epoch": 6.23,
      "learning_rate": 1.8831255417027802e-05,
      "loss": 0.124,
      "step": 93500
    },
    {
      "epoch": 6.27,
      "learning_rate": 1.8664577638509236e-05,
      "loss": 0.119,
      "step": 94000
    },
    {
      "epoch": 6.3,
      "learning_rate": 1.8497899859990666e-05,
      "loss": 0.119,
      "step": 94500
    },
    {
      "epoch": 6.33,
      "learning_rate": 1.83312220814721e-05,
      "loss": 0.1235,
      "step": 95000
    },
    {
      "epoch": 6.37,
      "learning_rate": 1.8164544302953533e-05,
      "loss": 0.1246,
      "step": 95500
    },
    {
      "epoch": 6.4,
      "learning_rate": 1.7997866524434963e-05,
      "loss": 0.1273,
      "step": 96000
    },
    {
      "epoch": 6.43,
      "learning_rate": 1.7831188745916394e-05,
      "loss": 0.1211,
      "step": 96500
    },
    {
      "epoch": 6.47,
      "learning_rate": 1.7664510967397827e-05,
      "loss": 0.1185,
      "step": 97000
    },
    {
      "epoch": 6.5,
      "learning_rate": 1.7497833188879258e-05,
      "loss": 0.1255,
      "step": 97500
    },
    {
      "epoch": 6.53,
      "learning_rate": 1.733115541036069e-05,
      "loss": 0.1229,
      "step": 98000
    },
    {
      "epoch": 6.57,
      "learning_rate": 1.7164477631842125e-05,
      "loss": 0.1182,
      "step": 98500
    },
    {
      "epoch": 6.6,
      "learning_rate": 1.6997799853323555e-05,
      "loss": 0.1245,
      "step": 99000
    },
    {
      "epoch": 6.63,
      "learning_rate": 1.683112207480499e-05,
      "loss": 0.1246,
      "step": 99500
    },
    {
      "epoch": 6.67,
      "learning_rate": 1.666444429628642e-05,
      "loss": 0.1231,
      "step": 100000
    },
    {
      "epoch": 6.7,
      "learning_rate": 1.6497766517767853e-05,
      "loss": 0.1214,
      "step": 100500
    },
    {
      "epoch": 6.73,
      "learning_rate": 1.6331088739249283e-05,
      "loss": 0.112,
      "step": 101000
    },
    {
      "epoch": 6.77,
      "learning_rate": 1.6164410960730716e-05,
      "loss": 0.1234,
      "step": 101500
    },
    {
      "epoch": 6.8,
      "learning_rate": 1.599773318221215e-05,
      "loss": 0.1294,
      "step": 102000
    },
    {
      "epoch": 6.83,
      "learning_rate": 1.583105540369358e-05,
      "loss": 0.1232,
      "step": 102500
    },
    {
      "epoch": 6.87,
      "learning_rate": 1.566437762517501e-05,
      "loss": 0.1241,
      "step": 103000
    },
    {
      "epoch": 6.9,
      "learning_rate": 1.5497699846656444e-05,
      "loss": 0.1151,
      "step": 103500
    },
    {
      "epoch": 6.93,
      "learning_rate": 1.5331022068137878e-05,
      "loss": 0.121,
      "step": 104000
    },
    {
      "epoch": 6.97,
      "learning_rate": 1.516434428961931e-05,
      "loss": 0.1287,
      "step": 104500
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.8897223055638898,
      "eval_loss": 0.1499263346195221,
      "eval_runtime": 1810.3934,
      "eval_samples_per_second": 16.569,
      "eval_steps_per_second": 2.071,
      "step": 104993
    },
    {
      "epoch": 7.0,
      "learning_rate": 1.4997666511100742e-05,
      "loss": 0.1175,
      "step": 105000
    },
    {
      "epoch": 7.03,
      "learning_rate": 1.4830988732582174e-05,
      "loss": 0.1159,
      "step": 105500
    },
    {
      "epoch": 7.07,
      "learning_rate": 1.4664310954063604e-05,
      "loss": 0.1194,
      "step": 106000
    },
    {
      "epoch": 7.1,
      "learning_rate": 1.4497633175545036e-05,
      "loss": 0.1195,
      "step": 106500
    },
    {
      "epoch": 7.13,
      "learning_rate": 1.433095539702647e-05,
      "loss": 0.1212,
      "step": 107000
    },
    {
      "epoch": 7.17,
      "learning_rate": 1.4164277618507901e-05,
      "loss": 0.1206,
      "step": 107500
    },
    {
      "epoch": 7.2,
      "learning_rate": 1.3997599839989333e-05,
      "loss": 0.1196,
      "step": 108000
    },
    {
      "epoch": 7.23,
      "learning_rate": 1.3830922061470767e-05,
      "loss": 0.1201,
      "step": 108500
    },
    {
      "epoch": 7.27,
      "learning_rate": 1.3664244282952199e-05,
      "loss": 0.1237,
      "step": 109000
    },
    {
      "epoch": 7.3,
      "learning_rate": 1.3497566504433629e-05,
      "loss": 0.1181,
      "step": 109500
    },
    {
      "epoch": 7.33,
      "learning_rate": 1.3330888725915061e-05,
      "loss": 0.1197,
      "step": 110000
    },
    {
      "epoch": 7.37,
      "learning_rate": 1.3164210947396493e-05,
      "loss": 0.1189,
      "step": 110500
    },
    {
      "epoch": 7.4,
      "learning_rate": 1.2997533168877926e-05,
      "loss": 0.1218,
      "step": 111000
    },
    {
      "epoch": 7.43,
      "learning_rate": 1.2830855390359358e-05,
      "loss": 0.1209,
      "step": 111500
    },
    {
      "epoch": 7.47,
      "learning_rate": 1.2664177611840792e-05,
      "loss": 0.1238,
      "step": 112000
    },
    {
      "epoch": 7.5,
      "learning_rate": 1.2497499833322222e-05,
      "loss": 0.1071,
      "step": 112500
    },
    {
      "epoch": 7.53,
      "learning_rate": 1.2330822054803654e-05,
      "loss": 0.1065,
      "step": 113000
    },
    {
      "epoch": 7.57,
      "learning_rate": 1.2164144276285086e-05,
      "loss": 0.1201,
      "step": 113500
    },
    {
      "epoch": 7.6,
      "learning_rate": 1.1997466497766518e-05,
      "loss": 0.1179,
      "step": 114000
    },
    {
      "epoch": 7.63,
      "learning_rate": 1.1830788719247952e-05,
      "loss": 0.1154,
      "step": 114500
    },
    {
      "epoch": 7.67,
      "learning_rate": 1.1664110940729382e-05,
      "loss": 0.1229,
      "step": 115000
    },
    {
      "epoch": 7.7,
      "learning_rate": 1.1497433162210816e-05,
      "loss": 0.12,
      "step": 115500
    },
    {
      "epoch": 7.73,
      "learning_rate": 1.1330755383692247e-05,
      "loss": 0.1234,
      "step": 116000
    },
    {
      "epoch": 7.77,
      "learning_rate": 1.1164077605173678e-05,
      "loss": 0.1246,
      "step": 116500
    },
    {
      "epoch": 7.8,
      "learning_rate": 1.0997399826655111e-05,
      "loss": 0.1153,
      "step": 117000
    },
    {
      "epoch": 7.83,
      "learning_rate": 1.0830722048136543e-05,
      "loss": 0.118,
      "step": 117500
    },
    {
      "epoch": 7.87,
      "learning_rate": 1.0664044269617975e-05,
      "loss": 0.1249,
      "step": 118000
    },
    {
      "epoch": 7.9,
      "learning_rate": 1.0497366491099407e-05,
      "loss": 0.1104,
      "step": 118500
    },
    {
      "epoch": 7.93,
      "learning_rate": 1.0330688712580839e-05,
      "loss": 0.1152,
      "step": 119000
    },
    {
      "epoch": 7.97,
      "learning_rate": 1.0164010934062271e-05,
      "loss": 0.1188,
      "step": 119500
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.8884221755508884,
      "eval_loss": 0.15246762335300446,
      "eval_runtime": 1810.1751,
      "eval_samples_per_second": 16.571,
      "eval_steps_per_second": 2.072,
      "step": 119992
    },
    {
      "epoch": 8.0,
      "learning_rate": 9.997333155543703e-06,
      "loss": 0.1162,
      "step": 120000
    },
    {
      "epoch": 8.03,
      "learning_rate": 9.830655377025137e-06,
      "loss": 0.1135,
      "step": 120500
    },
    {
      "epoch": 8.07,
      "learning_rate": 9.663977598506567e-06,
      "loss": 0.1179,
      "step": 121000
    },
    {
      "epoch": 8.1,
      "learning_rate": 9.497299819987999e-06,
      "loss": 0.1231,
      "step": 121500
    },
    {
      "epoch": 8.13,
      "learning_rate": 9.330622041469432e-06,
      "loss": 0.1154,
      "step": 122000
    },
    {
      "epoch": 8.17,
      "learning_rate": 9.163944262950863e-06,
      "loss": 0.1241,
      "step": 122500
    },
    {
      "epoch": 8.2,
      "learning_rate": 8.997266484432296e-06,
      "loss": 0.1228,
      "step": 123000
    },
    {
      "epoch": 8.23,
      "learning_rate": 8.830588705913728e-06,
      "loss": 0.1154,
      "step": 123500
    },
    {
      "epoch": 8.27,
      "learning_rate": 8.66391092739516e-06,
      "loss": 0.1099,
      "step": 124000
    },
    {
      "epoch": 8.3,
      "learning_rate": 8.497233148876592e-06,
      "loss": 0.1159,
      "step": 124500
    },
    {
      "epoch": 8.33,
      "learning_rate": 8.330555370358024e-06,
      "loss": 0.1147,
      "step": 125000
    },
    {
      "epoch": 8.37,
      "learning_rate": 8.163877591839458e-06,
      "loss": 0.1191,
      "step": 125500
    },
    {
      "epoch": 8.4,
      "learning_rate": 7.997199813320888e-06,
      "loss": 0.1175,
      "step": 126000
    },
    {
      "epoch": 8.43,
      "learning_rate": 7.830522034802321e-06,
      "loss": 0.1215,
      "step": 126500
    },
    {
      "epoch": 8.47,
      "learning_rate": 7.663844256283753e-06,
      "loss": 0.121,
      "step": 127000
    },
    {
      "epoch": 8.5,
      "learning_rate": 7.497166477765184e-06,
      "loss": 0.1139,
      "step": 127500
    },
    {
      "epoch": 8.53,
      "learning_rate": 7.330488699246617e-06,
      "loss": 0.1189,
      "step": 128000
    },
    {
      "epoch": 8.57,
      "learning_rate": 7.163810920728049e-06,
      "loss": 0.1206,
      "step": 128500
    },
    {
      "epoch": 8.6,
      "learning_rate": 6.99713314220948e-06,
      "loss": 0.1171,
      "step": 129000
    },
    {
      "epoch": 8.63,
      "learning_rate": 6.830455363690913e-06,
      "loss": 0.1197,
      "step": 129500
    },
    {
      "epoch": 8.67,
      "learning_rate": 6.663777585172346e-06,
      "loss": 0.1196,
      "step": 130000
    },
    {
      "epoch": 8.7,
      "learning_rate": 6.497099806653777e-06,
      "loss": 0.1271,
      "step": 130500
    },
    {
      "epoch": 8.73,
      "learning_rate": 6.33042202813521e-06,
      "loss": 0.1056,
      "step": 131000
    },
    {
      "epoch": 8.77,
      "learning_rate": 6.1637442496166415e-06,
      "loss": 0.1219,
      "step": 131500
    },
    {
      "epoch": 8.8,
      "learning_rate": 5.9970664710980735e-06,
      "loss": 0.1188,
      "step": 132000
    },
    {
      "epoch": 8.83,
      "learning_rate": 5.830388692579505e-06,
      "loss": 0.1088,
      "step": 132500
    },
    {
      "epoch": 8.87,
      "learning_rate": 5.663710914060938e-06,
      "loss": 0.1185,
      "step": 133000
    },
    {
      "epoch": 8.9,
      "learning_rate": 5.49703313554237e-06,
      "loss": 0.1126,
      "step": 133500
    },
    {
      "epoch": 8.93,
      "learning_rate": 5.330355357023801e-06,
      "loss": 0.1141,
      "step": 134000
    },
    {
      "epoch": 8.97,
      "learning_rate": 5.163677578505234e-06,
      "loss": 0.1172,
      "step": 134500
    },
    {
      "epoch": 9.0,
      "eval_accuracy": 0.8899556622328899,
      "eval_loss": 0.1522732973098755,
      "eval_runtime": 1802.2047,
      "eval_samples_per_second": 16.645,
      "eval_steps_per_second": 2.081,
      "step": 134991
    },
    {
      "epoch": 9.0,
      "learning_rate": 4.996999799986666e-06,
      "loss": 0.1153,
      "step": 135000
    },
    {
      "epoch": 9.03,
      "learning_rate": 4.830322021468098e-06,
      "loss": 0.1109,
      "step": 135500
    },
    {
      "epoch": 9.07,
      "learning_rate": 4.663644242949531e-06,
      "loss": 0.1137,
      "step": 136000
    },
    {
      "epoch": 9.1,
      "learning_rate": 4.4969664644309625e-06,
      "loss": 0.1216,
      "step": 136500
    },
    {
      "epoch": 9.13,
      "learning_rate": 4.3302886859123945e-06,
      "loss": 0.111,
      "step": 137000
    },
    {
      "epoch": 9.17,
      "learning_rate": 4.163610907393826e-06,
      "loss": 0.1228,
      "step": 137500
    },
    {
      "epoch": 9.2,
      "learning_rate": 3.996933128875258e-06,
      "loss": 0.1061,
      "step": 138000
    },
    {
      "epoch": 9.23,
      "learning_rate": 3.830255350356691e-06,
      "loss": 0.1127,
      "step": 138500
    },
    {
      "epoch": 9.27,
      "learning_rate": 3.6635775718381226e-06,
      "loss": 0.107,
      "step": 139000
    },
    {
      "epoch": 9.3,
      "learning_rate": 3.4968997933195545e-06,
      "loss": 0.1254,
      "step": 139500
    },
    {
      "epoch": 9.33,
      "learning_rate": 3.330222014800987e-06,
      "loss": 0.1122,
      "step": 140000
    },
    {
      "epoch": 9.37,
      "learning_rate": 3.163544236282419e-06,
      "loss": 0.1149,
      "step": 140500
    },
    {
      "epoch": 9.4,
      "learning_rate": 2.996866457763851e-06,
      "loss": 0.1193,
      "step": 141000
    },
    {
      "epoch": 9.43,
      "learning_rate": 2.830188679245283e-06,
      "loss": 0.1154,
      "step": 141500
    },
    {
      "epoch": 9.47,
      "learning_rate": 2.6635109007267155e-06,
      "loss": 0.1213,
      "step": 142000
    },
    {
      "epoch": 9.5,
      "learning_rate": 2.4968331222081474e-06,
      "loss": 0.1203,
      "step": 142500
    },
    {
      "epoch": 9.53,
      "learning_rate": 2.3301553436895793e-06,
      "loss": 0.1131,
      "step": 143000
    },
    {
      "epoch": 9.57,
      "learning_rate": 2.1634775651710117e-06,
      "loss": 0.1082,
      "step": 143500
    },
    {
      "epoch": 9.6,
      "learning_rate": 1.9967997866524436e-06,
      "loss": 0.113,
      "step": 144000
    },
    {
      "epoch": 9.63,
      "learning_rate": 1.8301220081338755e-06,
      "loss": 0.1103,
      "step": 144500
    },
    {
      "epoch": 9.67,
      "learning_rate": 1.6634442296153079e-06,
      "loss": 0.1042,
      "step": 145000
    },
    {
      "epoch": 9.7,
      "learning_rate": 1.4967664510967398e-06,
      "loss": 0.1218,
      "step": 145500
    },
    {
      "epoch": 9.73,
      "learning_rate": 1.330088672578172e-06,
      "loss": 0.1093,
      "step": 146000
    },
    {
      "epoch": 9.77,
      "learning_rate": 1.1634108940596039e-06,
      "loss": 0.1206,
      "step": 146500
    },
    {
      "epoch": 9.8,
      "learning_rate": 9.967331155410362e-07,
      "loss": 0.1209,
      "step": 147000
    },
    {
      "epoch": 9.83,
      "learning_rate": 8.300553370224682e-07,
      "loss": 0.1161,
      "step": 147500
    },
    {
      "epoch": 9.87,
      "learning_rate": 6.633775585039002e-07,
      "loss": 0.1109,
      "step": 148000
    },
    {
      "epoch": 9.9,
      "learning_rate": 4.966997799853323e-07,
      "loss": 0.1268,
      "step": 148500
    },
    {
      "epoch": 9.93,
      "learning_rate": 3.300220014667645e-07,
      "loss": 0.1131,
      "step": 149000
    },
    {
      "epoch": 9.97,
      "learning_rate": 1.6334422294819655e-07,
      "loss": 0.1131,
      "step": 149500
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.8906223955728906,
      "eval_loss": 0.155410036444664,
      "eval_runtime": 1805.48,
      "eval_samples_per_second": 16.614,
      "eval_steps_per_second": 2.077,
      "step": 149990
    },
    {
      "epoch": 10.0,
      "step": 149990,
      "total_flos": 2.5697864138725786e+18,
      "train_loss": 0.1303185607474743,
      "train_runtime": 92608.1075,
      "train_samples_per_second": 12.957,
      "train_steps_per_second": 1.62
    }
  ],
  "max_steps": 149990,
  "num_train_epochs": 10,
  "total_flos": 2.5697864138725786e+18,
  "trial_name": null,
  "trial_params": null
}