{
  "best_metric": 0.09185712039470673,
  "best_model_checkpoint": "checkpoints_corpus_v2/checkpoint-211352",
  "epoch": 40.0,
  "eval_steps": 500,
  "global_step": 1056760,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.02,
      "learning_rate": 1.9990537113441085e-05,
      "loss": 1.5304,
      "step": 500
    },
    {
      "epoch": 0.04,
      "learning_rate": 1.998107422688217e-05,
      "loss": 0.8661,
      "step": 1000
    },
    {
      "epoch": 0.06,
      "learning_rate": 1.9971611340323253e-05,
      "loss": 0.6917,
      "step": 1500
    },
    {
      "epoch": 0.08,
      "learning_rate": 1.996214845376434e-05,
      "loss": 0.573,
      "step": 2000
    },
    {
      "epoch": 0.09,
      "learning_rate": 1.995268556720542e-05,
      "loss": 0.5146,
      "step": 2500
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.9943222680646507e-05,
      "loss": 0.4646,
      "step": 3000
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.993375979408759e-05,
      "loss": 0.4205,
      "step": 3500
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.9924315833301794e-05,
      "loss": 0.3912,
      "step": 4000
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.9914852946742874e-05,
      "loss": 0.3759,
      "step": 4500
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.990539006018396e-05,
      "loss": 0.3575,
      "step": 5000
    },
    {
      "epoch": 0.21,
      "learning_rate": 1.9895927173625045e-05,
      "loss": 0.3415,
      "step": 5500
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.988646428706613e-05,
      "loss": 0.3225,
      "step": 6000
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9877001400507212e-05,
      "loss": 0.3104,
      "step": 6500
    },
    {
      "epoch": 0.26,
      "learning_rate": 1.9867557439721415e-05,
      "loss": 0.2976,
      "step": 7000
    },
    {
      "epoch": 0.28,
      "learning_rate": 1.98580945531625e-05,
      "loss": 0.2939,
      "step": 7500
    },
    {
      "epoch": 0.3,
      "learning_rate": 1.9848631666603582e-05,
      "loss": 0.2803,
      "step": 8000
    },
    {
      "epoch": 0.32,
      "learning_rate": 1.9839168780044666e-05,
      "loss": 0.2759,
      "step": 8500
    },
    {
      "epoch": 0.34,
      "learning_rate": 1.982970589348575e-05,
      "loss": 0.2741,
      "step": 9000
    },
    {
      "epoch": 0.36,
      "learning_rate": 1.9820261932699953e-05,
      "loss": 0.2661,
      "step": 9500
    },
    {
      "epoch": 0.38,
      "learning_rate": 1.9810817971914152e-05,
      "loss": 0.261,
      "step": 10000
    },
    {
      "epoch": 0.4,
      "learning_rate": 1.980135508535524e-05,
      "loss": 0.2514,
      "step": 10500
    },
    {
      "epoch": 0.42,
      "learning_rate": 1.9791892198796323e-05,
      "loss": 0.2514,
      "step": 11000
    },
    {
      "epoch": 0.44,
      "learning_rate": 1.9782429312237407e-05,
      "loss": 0.2437,
      "step": 11500
    },
    {
      "epoch": 0.45,
      "learning_rate": 1.977296642567849e-05,
      "loss": 0.2403,
      "step": 12000
    },
    {
      "epoch": 0.47,
      "learning_rate": 1.9763503539119574e-05,
      "loss": 0.2282,
      "step": 12500
    },
    {
      "epoch": 0.49,
      "learning_rate": 1.9754040652560658e-05,
      "loss": 0.2308,
      "step": 13000
    },
    {
      "epoch": 0.51,
      "learning_rate": 1.974457776600174e-05,
      "loss": 0.2303,
      "step": 13500
    },
    {
      "epoch": 0.53,
      "learning_rate": 1.973511487944283e-05,
      "loss": 0.2212,
      "step": 14000
    },
    {
      "epoch": 0.55,
      "learning_rate": 1.9725651992883912e-05,
      "loss": 0.2171,
      "step": 14500
    },
    {
      "epoch": 0.57,
      "learning_rate": 1.9716189106324996e-05,
      "loss": 0.2156,
      "step": 15000
    },
    {
      "epoch": 0.59,
      "learning_rate": 1.9706745145539195e-05,
      "loss": 0.2159,
      "step": 15500
    },
    {
      "epoch": 0.61,
      "learning_rate": 1.9697282258980282e-05,
      "loss": 0.2095,
      "step": 16000
    },
    {
      "epoch": 0.62,
      "learning_rate": 1.9687819372421363e-05,
      "loss": 0.209,
      "step": 16500
    },
    {
      "epoch": 0.64,
      "learning_rate": 1.967835648586245e-05,
      "loss": 0.2079,
      "step": 17000
    },
    {
      "epoch": 0.66,
      "learning_rate": 1.9668893599303533e-05,
      "loss": 0.1994,
      "step": 17500
    },
    {
      "epoch": 0.68,
      "learning_rate": 1.9659449638517736e-05,
      "loss": 0.2032,
      "step": 18000
    },
    {
      "epoch": 0.7,
      "learning_rate": 1.964998675195882e-05,
      "loss": 0.1983,
      "step": 18500
    },
    {
      "epoch": 0.72,
      "learning_rate": 1.9640523865399904e-05,
      "loss": 0.1989,
      "step": 19000
    },
    {
      "epoch": 0.74,
      "learning_rate": 1.9631060978840987e-05,
      "loss": 0.1938,
      "step": 19500
    },
    {
      "epoch": 0.76,
      "learning_rate": 1.962161701805519e-05,
      "loss": 0.2004,
      "step": 20000
    },
    {
      "epoch": 0.78,
      "learning_rate": 1.9612154131496274e-05,
      "loss": 0.1946,
      "step": 20500
    },
    {
      "epoch": 0.79,
      "learning_rate": 1.9602691244937358e-05,
      "loss": 0.1849,
      "step": 21000
    },
    {
      "epoch": 0.81,
      "learning_rate": 1.959322835837844e-05,
      "loss": 0.1874,
      "step": 21500
    },
    {
      "epoch": 0.83,
      "learning_rate": 1.9583765471819525e-05,
      "loss": 0.1896,
      "step": 22000
    },
    {
      "epoch": 0.85,
      "learning_rate": 1.957430258526061e-05,
      "loss": 0.1865,
      "step": 22500
    },
    {
      "epoch": 0.87,
      "learning_rate": 1.9564839698701696e-05,
      "loss": 0.1815,
      "step": 23000
    },
    {
      "epoch": 0.89,
      "learning_rate": 1.9555376812142776e-05,
      "loss": 0.1819,
      "step": 23500
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9545913925583863e-05,
      "loss": 0.187,
      "step": 24000
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.9536451039024947e-05,
      "loss": 0.179,
      "step": 24500
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.952700707823915e-05,
      "loss": 0.1775,
      "step": 25000
    },
    {
      "epoch": 0.97,
      "learning_rate": 1.951754419168023e-05,
      "loss": 0.1751,
      "step": 25500
    },
    {
      "epoch": 0.98,
      "learning_rate": 1.9508100230894433e-05,
      "loss": 0.1748,
      "step": 26000
    },
    {
      "epoch": 1.0,
      "eval_bleu": 88.9021,
      "eval_gen_len": 16.3286,
      "eval_loss": 0.14062626659870148,
      "eval_runtime": 1229.1102,
      "eval_samples_per_second": 37.155,
      "eval_steps_per_second": 1.162,
      "step": 26419
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.949863734433552e-05,
      "loss": 0.1731,
      "step": 26500
    },
    {
      "epoch": 1.02,
      "learning_rate": 1.94891744577766e-05,
      "loss": 0.163,
      "step": 27000
    },
    {
      "epoch": 1.04,
      "learning_rate": 1.9479711571217684e-05,
      "loss": 0.1602,
      "step": 27500
    },
    {
      "epoch": 1.06,
      "learning_rate": 1.947024868465877e-05,
      "loss": 0.1604,
      "step": 28000
    },
    {
      "epoch": 1.08,
      "learning_rate": 1.9460785798099855e-05,
      "loss": 0.1592,
      "step": 28500
    },
    {
      "epoch": 1.1,
      "learning_rate": 1.945132291154094e-05,
      "loss": 0.1611,
      "step": 29000
    },
    {
      "epoch": 1.12,
      "learning_rate": 1.9441860024982022e-05,
      "loss": 0.1591,
      "step": 29500
    },
    {
      "epoch": 1.14,
      "learning_rate": 1.9432397138423106e-05,
      "loss": 0.1564,
      "step": 30000
    },
    {
      "epoch": 1.15,
      "learning_rate": 1.942293425186419e-05,
      "loss": 0.1589,
      "step": 30500
    },
    {
      "epoch": 1.17,
      "learning_rate": 1.9413471365305273e-05,
      "loss": 0.1566,
      "step": 31000
    },
    {
      "epoch": 1.19,
      "learning_rate": 1.9404027404519476e-05,
      "loss": 0.1536,
      "step": 31500
    },
    {
      "epoch": 1.21,
      "learning_rate": 1.939456451796056e-05,
      "loss": 0.1566,
      "step": 32000
    },
    {
      "epoch": 1.23,
      "learning_rate": 1.9385101631401643e-05,
      "loss": 0.155,
      "step": 32500
    },
    {
      "epoch": 1.25,
      "learning_rate": 1.9375638744842727e-05,
      "loss": 0.1528,
      "step": 33000
    },
    {
      "epoch": 1.27,
      "learning_rate": 1.9366175858283814e-05,
      "loss": 0.153,
      "step": 33500
    },
    {
      "epoch": 1.29,
      "learning_rate": 1.9356712971724894e-05,
      "loss": 0.1523,
      "step": 34000
    },
    {
      "epoch": 1.31,
      "learning_rate": 1.934725008516598e-05,
      "loss": 0.1526,
      "step": 34500
    },
    {
      "epoch": 1.32,
      "learning_rate": 1.9337787198607065e-05,
      "loss": 0.1489,
      "step": 35000
    },
    {
      "epoch": 1.34,
      "learning_rate": 1.9328362163594384e-05,
      "loss": 0.152,
      "step": 35500
    },
    {
      "epoch": 1.36,
      "learning_rate": 1.9318899277035467e-05,
      "loss": 0.1506,
      "step": 36000
    },
    {
      "epoch": 1.38,
      "learning_rate": 1.930943639047655e-05,
      "loss": 0.1498,
      "step": 36500
    },
    {
      "epoch": 1.4,
      "learning_rate": 1.9299973503917638e-05,
      "loss": 0.1467,
      "step": 37000
    },
    {
      "epoch": 1.42,
      "learning_rate": 1.929052954313184e-05,
      "loss": 0.148,
      "step": 37500
    },
    {
      "epoch": 1.44,
      "learning_rate": 1.928106665657292e-05,
      "loss": 0.1436,
      "step": 38000
    },
    {
      "epoch": 1.46,
      "learning_rate": 1.927160377001401e-05,
      "loss": 0.1457,
      "step": 38500
    },
    {
      "epoch": 1.48,
      "learning_rate": 1.9262140883455092e-05,
      "loss": 0.1443,
      "step": 39000
    },
    {
      "epoch": 1.5,
      "learning_rate": 1.9252677996896172e-05,
      "loss": 0.1442,
      "step": 39500
    },
    {
      "epoch": 1.51,
      "learning_rate": 1.9243234036110375e-05,
      "loss": 0.1435,
      "step": 40000
    },
    {
      "epoch": 1.53,
      "learning_rate": 1.9233771149551462e-05,
      "loss": 0.1467,
      "step": 40500
    },
    {
      "epoch": 1.55,
      "learning_rate": 1.9224308262992543e-05,
      "loss": 0.1455,
      "step": 41000
    },
    {
      "epoch": 1.57,
      "learning_rate": 1.921484537643363e-05,
      "loss": 0.1437,
      "step": 41500
    },
    {
      "epoch": 1.59,
      "learning_rate": 1.920540141564783e-05,
      "loss": 0.142,
      "step": 42000
    },
    {
      "epoch": 1.61,
      "learning_rate": 1.9195938529088916e-05,
      "loss": 0.1417,
      "step": 42500
    },
    {
      "epoch": 1.63,
      "learning_rate": 1.9186475642529997e-05,
      "loss": 0.1408,
      "step": 43000
    },
    {
      "epoch": 1.65,
      "learning_rate": 1.91770316817442e-05,
      "loss": 0.1438,
      "step": 43500
    },
    {
      "epoch": 1.67,
      "learning_rate": 1.9167568795185286e-05,
      "loss": 0.1432,
      "step": 44000
    },
    {
      "epoch": 1.68,
      "learning_rate": 1.915810590862637e-05,
      "loss": 0.1396,
      "step": 44500
    },
    {
      "epoch": 1.7,
      "learning_rate": 1.9148643022067454e-05,
      "loss": 0.14,
      "step": 45000
    },
    {
      "epoch": 1.72,
      "learning_rate": 1.9139180135508538e-05,
      "loss": 0.1384,
      "step": 45500
    },
    {
      "epoch": 1.74,
      "learning_rate": 1.912971724894962e-05,
      "loss": 0.1402,
      "step": 46000
    },
    {
      "epoch": 1.76,
      "learning_rate": 1.9120254362390705e-05,
      "loss": 0.1393,
      "step": 46500
    },
    {
      "epoch": 1.78,
      "learning_rate": 1.911079147583179e-05,
      "loss": 0.1385,
      "step": 47000
    },
    {
      "epoch": 1.8,
      "learning_rate": 1.9101328589272872e-05,
      "loss": 0.1379,
      "step": 47500
    },
    {
      "epoch": 1.82,
      "learning_rate": 1.9091865702713956e-05,
      "loss": 0.1391,
      "step": 48000
    },
    {
      "epoch": 1.84,
      "learning_rate": 1.908242174192816e-05,
      "loss": 0.1362,
      "step": 48500
    },
    {
      "epoch": 1.85,
      "learning_rate": 1.9072958855369243e-05,
      "loss": 0.1332,
      "step": 49000
    },
    {
      "epoch": 1.87,
      "learning_rate": 1.906349596881033e-05,
      "loss": 0.1368,
      "step": 49500
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.905403308225141e-05,
      "loss": 0.1369,
      "step": 50000
    },
    {
      "epoch": 1.91,
      "learning_rate": 1.9044570195692497e-05,
      "loss": 0.1368,
      "step": 50500
    },
    {
      "epoch": 1.93,
      "learning_rate": 1.9035145160679816e-05,
      "loss": 0.1361,
      "step": 51000
    },
    {
      "epoch": 1.95,
      "learning_rate": 1.90256822741209e-05,
      "loss": 0.1365,
      "step": 51500
    },
    {
      "epoch": 1.97,
      "learning_rate": 1.9016219387561983e-05,
      "loss": 0.1338,
      "step": 52000
    },
    {
      "epoch": 1.99,
      "learning_rate": 1.9006756501003067e-05,
      "loss": 0.1344,
      "step": 52500
    },
    {
      "epoch": 2.0,
      "eval_bleu": 90.5089,
      "eval_gen_len": 16.3365,
      "eval_loss": 0.11337216943502426,
      "eval_runtime": 1219.6509,
      "eval_samples_per_second": 37.443,
      "eval_steps_per_second": 1.171,
      "step": 52838
    },
    {
      "epoch": 2.01,
      "learning_rate": 1.8997293614444154e-05,
      "loss": 0.1299,
      "step": 53000
    },
    {
      "epoch": 2.03,
      "learning_rate": 1.8987830727885234e-05,
      "loss": 0.1238,
      "step": 53500
    },
    {
      "epoch": 2.04,
      "learning_rate": 1.897836784132632e-05,
      "loss": 0.1217,
      "step": 54000
    },
    {
      "epoch": 2.06,
      "learning_rate": 1.8968904954767405e-05,
      "loss": 0.1209,
      "step": 54500
    },
    {
      "epoch": 2.08,
      "learning_rate": 1.8959442068208485e-05,
      "loss": 0.1217,
      "step": 55000
    },
    {
      "epoch": 2.1,
      "learning_rate": 1.8949979181649572e-05,
      "loss": 0.1217,
      "step": 55500
    },
    {
      "epoch": 2.12,
      "learning_rate": 1.8940535220863775e-05,
      "loss": 0.1196,
      "step": 56000
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.893107233430486e-05,
      "loss": 0.123,
      "step": 56500
    },
    {
      "epoch": 2.16,
      "learning_rate": 1.8921609447745942e-05,
      "loss": 0.1229,
      "step": 57000
    },
    {
      "epoch": 2.18,
      "learning_rate": 1.8912146561187026e-05,
      "loss": 0.1264,
      "step": 57500
    },
    {
      "epoch": 2.2,
      "learning_rate": 1.890268367462811e-05,
      "loss": 0.1187,
      "step": 58000
    },
    {
      "epoch": 2.21,
      "learning_rate": 1.8893220788069194e-05,
      "loss": 0.1226,
      "step": 58500
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.8883757901510277e-05,
      "loss": 0.1213,
      "step": 59000
    },
    {
      "epoch": 2.25,
      "learning_rate": 1.8874295014951364e-05,
      "loss": 0.1207,
      "step": 59500
    },
    {
      "epoch": 2.27,
      "learning_rate": 1.8864851054165564e-05,
      "loss": 0.1208,
      "step": 60000
    },
    {
      "epoch": 2.29,
      "learning_rate": 1.8855407093379767e-05,
      "loss": 0.1219,
      "step": 60500
    },
    {
      "epoch": 2.31,
      "learning_rate": 1.884594420682085e-05,
      "loss": 0.1216,
      "step": 61000
    },
    {
      "epoch": 2.33,
      "learning_rate": 1.8836481320261934e-05,
      "loss": 0.1184,
      "step": 61500
    },
    {
      "epoch": 2.35,
      "learning_rate": 1.8827018433703018e-05,
      "loss": 0.1187,
      "step": 62000
    },
    {
      "epoch": 2.37,
      "learning_rate": 1.88175555471441e-05,
      "loss": 0.1188,
      "step": 62500
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.8808092660585185e-05,
      "loss": 0.1187,
      "step": 63000
    },
    {
      "epoch": 2.4,
      "learning_rate": 1.8798629774026272e-05,
      "loss": 0.1207,
      "step": 63500
    },
    {
      "epoch": 2.42,
      "learning_rate": 1.878918581324047e-05,
      "loss": 0.1218,
      "step": 64000
    },
    {
      "epoch": 2.44,
      "learning_rate": 1.8779722926681555e-05,
      "loss": 0.1173,
      "step": 64500
    },
    {
      "epoch": 2.46,
      "learning_rate": 1.8770260040122642e-05,
      "loss": 0.1181,
      "step": 65000
    },
    {
      "epoch": 2.48,
      "learning_rate": 1.8760797153563726e-05,
      "loss": 0.1202,
      "step": 65500
    },
    {
      "epoch": 2.5,
      "learning_rate": 1.875133426700481e-05,
      "loss": 0.1181,
      "step": 66000
    },
    {
      "epoch": 2.52,
      "learning_rate": 1.8741871380445893e-05,
      "loss": 0.1178,
      "step": 66500
    },
    {
      "epoch": 2.54,
      "learning_rate": 1.8732408493886977e-05,
      "loss": 0.1189,
      "step": 67000
    },
    {
      "epoch": 2.55,
      "learning_rate": 1.872294560732806e-05,
      "loss": 0.1163,
      "step": 67500
    },
    {
      "epoch": 2.57,
      "learning_rate": 1.8713501646542264e-05,
      "loss": 0.1179,
      "step": 68000
    },
    {
      "epoch": 2.59,
      "learning_rate": 1.8704057685756466e-05,
      "loss": 0.1161,
      "step": 68500
    },
    {
      "epoch": 2.61,
      "learning_rate": 1.869459479919755e-05,
      "loss": 0.1185,
      "step": 69000
    },
    {
      "epoch": 2.63,
      "learning_rate": 1.8685131912638634e-05,
      "loss": 0.1172,
      "step": 69500
    },
    {
      "epoch": 2.65,
      "learning_rate": 1.8675669026079718e-05,
      "loss": 0.1191,
      "step": 70000
    },
    {
      "epoch": 2.67,
      "learning_rate": 1.86662061395208e-05,
      "loss": 0.1174,
      "step": 70500
    },
    {
      "epoch": 2.69,
      "learning_rate": 1.8656762178735004e-05,
      "loss": 0.1159,
      "step": 71000
    },
    {
      "epoch": 2.71,
      "learning_rate": 1.8647299292176088e-05,
      "loss": 0.1195,
      "step": 71500
    },
    {
      "epoch": 2.73,
      "learning_rate": 1.863783640561717e-05,
      "loss": 0.1144,
      "step": 72000
    },
    {
      "epoch": 2.74,
      "learning_rate": 1.8628392444831374e-05,
      "loss": 0.1165,
      "step": 72500
    },
    {
      "epoch": 2.76,
      "learning_rate": 1.8618929558272455e-05,
      "loss": 0.1155,
      "step": 73000
    },
    {
      "epoch": 2.78,
      "learning_rate": 1.860946667171354e-05,
      "loss": 0.1155,
      "step": 73500
    },
    {
      "epoch": 2.8,
      "learning_rate": 1.8600003785154625e-05,
      "loss": 0.1164,
      "step": 74000
    },
    {
      "epoch": 2.82,
      "learning_rate": 1.859054089859571e-05,
      "loss": 0.1129,
      "step": 74500
    },
    {
      "epoch": 2.84,
      "learning_rate": 1.8581096937809912e-05,
      "loss": 0.115,
      "step": 75000
    },
    {
      "epoch": 2.86,
      "learning_rate": 1.8571634051250996e-05,
      "loss": 0.1158,
      "step": 75500
    },
    {
      "epoch": 2.88,
      "learning_rate": 1.856217116469208e-05,
      "loss": 0.1136,
      "step": 76000
    },
    {
      "epoch": 2.9,
      "learning_rate": 1.8552708278133163e-05,
      "loss": 0.1163,
      "step": 76500
    },
    {
      "epoch": 2.91,
      "learning_rate": 1.8543245391574247e-05,
      "loss": 0.1145,
      "step": 77000
    },
    {
      "epoch": 2.93,
      "learning_rate": 1.853378250501533e-05,
      "loss": 0.1156,
      "step": 77500
    },
    {
      "epoch": 2.95,
      "learning_rate": 1.8524319618456414e-05,
      "loss": 0.1122,
      "step": 78000
    },
    {
      "epoch": 2.97,
      "learning_rate": 1.8514856731897498e-05,
      "loss": 0.1143,
      "step": 78500
    },
    {
      "epoch": 2.99,
      "learning_rate": 1.8505393845338585e-05,
      "loss": 0.1143,
      "step": 79000
    },
    {
      "epoch": 3.0,
      "eval_bleu": 91.1209,
      "eval_gen_len": 16.3757,
      "eval_loss": 0.10402025282382965,
      "eval_runtime": 1204.974,
      "eval_samples_per_second": 37.899,
      "eval_steps_per_second": 1.185,
      "step": 79257
    },
    {
      "epoch": 3.01,
      "learning_rate": 1.8495949884552788e-05,
      "loss": 0.1113,
      "step": 79500
    },
    {
      "epoch": 3.03,
      "learning_rate": 1.8486486997993868e-05,
      "loss": 0.1047,
      "step": 80000
    },
    {
      "epoch": 3.05,
      "learning_rate": 1.8477024111434955e-05,
      "loss": 0.1023,
      "step": 80500
    },
    {
      "epoch": 3.07,
      "learning_rate": 1.846756122487604e-05,
      "loss": 0.1044,
      "step": 81000
    },
    {
      "epoch": 3.08,
      "learning_rate": 1.8458098338317122e-05,
      "loss": 0.1057,
      "step": 81500
    },
    {
      "epoch": 3.1,
      "learning_rate": 1.8448635451758206e-05,
      "loss": 0.1026,
      "step": 82000
    },
    {
      "epoch": 3.12,
      "learning_rate": 1.843917256519929e-05,
      "loss": 0.1019,
      "step": 82500
    },
    {
      "epoch": 3.14,
      "learning_rate": 1.8429709678640374e-05,
      "loss": 0.1034,
      "step": 83000
    },
    {
      "epoch": 3.16,
      "learning_rate": 1.8420246792081457e-05,
      "loss": 0.1037,
      "step": 83500
    },
    {
      "epoch": 3.18,
      "learning_rate": 1.841080283129566e-05,
      "loss": 0.107,
      "step": 84000
    },
    {
      "epoch": 3.2,
      "learning_rate": 1.8401358870509863e-05,
      "loss": 0.1029,
      "step": 84500
    },
    {
      "epoch": 3.22,
      "learning_rate": 1.8391895983950947e-05,
      "loss": 0.1027,
      "step": 85000
    },
    {
      "epoch": 3.24,
      "learning_rate": 1.838243309739203e-05,
      "loss": 0.1055,
      "step": 85500
    },
    {
      "epoch": 3.26,
      "learning_rate": 1.8372970210833114e-05,
      "loss": 0.1031,
      "step": 86000
    },
    {
      "epoch": 3.27,
      "learning_rate": 1.8363526250047317e-05,
      "loss": 0.1044,
      "step": 86500
    },
    {
      "epoch": 3.29,
      "learning_rate": 1.83540633634884e-05,
      "loss": 0.1044,
      "step": 87000
    },
    {
      "epoch": 3.31,
      "learning_rate": 1.8344600476929484e-05,
      "loss": 0.1037,
      "step": 87500
    },
    {
      "epoch": 3.33,
      "learning_rate": 1.8335137590370568e-05,
      "loss": 0.103,
      "step": 88000
    },
    {
      "epoch": 3.35,
      "learning_rate": 1.832567470381165e-05,
      "loss": 0.1032,
      "step": 88500
    },
    {
      "epoch": 3.37,
      "learning_rate": 1.8316211817252735e-05,
      "loss": 0.1029,
      "step": 89000
    },
    {
      "epoch": 3.39,
      "learning_rate": 1.8306748930693822e-05,
      "loss": 0.1002,
      "step": 89500
    },
    {
      "epoch": 3.41,
      "learning_rate": 1.8297286044134906e-05,
      "loss": 0.1044,
      "step": 90000
    },
    {
      "epoch": 3.43,
      "learning_rate": 1.8287861009122224e-05,
      "loss": 0.1038,
      "step": 90500
    },
    {
      "epoch": 3.44,
      "learning_rate": 1.8278398122563308e-05,
      "loss": 0.1039,
      "step": 91000
    },
    {
      "epoch": 3.46,
      "learning_rate": 1.8268935236004392e-05,
      "loss": 0.1026,
      "step": 91500
    },
    {
      "epoch": 3.48,
      "learning_rate": 1.825947234944548e-05,
      "loss": 0.102,
      "step": 92000
    },
    {
      "epoch": 3.5,
      "learning_rate": 1.825000946288656e-05,
      "loss": 0.1036,
      "step": 92500
    },
    {
      "epoch": 3.52,
      "learning_rate": 1.8240546576327643e-05,
      "loss": 0.1024,
      "step": 93000
    },
    {
      "epoch": 3.54,
      "learning_rate": 1.823108368976873e-05,
      "loss": 0.1041,
      "step": 93500
    },
    {
      "epoch": 3.56,
      "learning_rate": 1.822162080320981e-05,
      "loss": 0.102,
      "step": 94000
    },
    {
      "epoch": 3.58,
      "learning_rate": 1.8212157916650897e-05,
      "loss": 0.1049,
      "step": 94500
    },
    {
      "epoch": 3.6,
      "learning_rate": 1.820269503009198e-05,
      "loss": 0.0992,
      "step": 95000
    },
    {
      "epoch": 3.61,
      "learning_rate": 1.8193232143533065e-05,
      "loss": 0.104,
      "step": 95500
    },
    {
      "epoch": 3.63,
      "learning_rate": 1.818376925697415e-05,
      "loss": 0.1007,
      "step": 96000
    },
    {
      "epoch": 3.65,
      "learning_rate": 1.817432529618835e-05,
      "loss": 0.1037,
      "step": 96500
    },
    {
      "epoch": 3.67,
      "learning_rate": 1.8164862409629435e-05,
      "loss": 0.1036,
      "step": 97000
    },
    {
      "epoch": 3.69,
      "learning_rate": 1.8155418448843635e-05,
      "loss": 0.1012,
      "step": 97500
    },
    {
      "epoch": 3.71,
      "learning_rate": 1.814595556228472e-05,
      "loss": 0.1041,
      "step": 98000
    },
    {
      "epoch": 3.73,
      "learning_rate": 1.8136492675725805e-05,
      "loss": 0.1011,
      "step": 98500
    },
    {
      "epoch": 3.75,
      "learning_rate": 1.812702978916689e-05,
      "loss": 0.1025,
      "step": 99000
    },
    {
      "epoch": 3.77,
      "learning_rate": 1.8117566902607973e-05,
      "loss": 0.1012,
      "step": 99500
    },
    {
      "epoch": 3.79,
      "learning_rate": 1.8108104016049056e-05,
      "loss": 0.1014,
      "step": 100000
    },
    {
      "epoch": 3.8,
      "learning_rate": 1.8098641129490143e-05,
      "loss": 0.1011,
      "step": 100500
    },
    {
      "epoch": 3.82,
      "learning_rate": 1.8089178242931224e-05,
      "loss": 0.1008,
      "step": 101000
    },
    {
      "epoch": 3.84,
      "learning_rate": 1.8079734282145427e-05,
      "loss": 0.1009,
      "step": 101500
    },
    {
      "epoch": 3.86,
      "learning_rate": 1.807027139558651e-05,
      "loss": 0.1014,
      "step": 102000
    },
    {
      "epoch": 3.88,
      "learning_rate": 1.8060808509027594e-05,
      "loss": 0.099,
      "step": 102500
    },
    {
      "epoch": 3.9,
      "learning_rate": 1.8051345622468678e-05,
      "loss": 0.1014,
      "step": 103000
    },
    {
      "epoch": 3.92,
      "learning_rate": 1.8041882735909765e-05,
      "loss": 0.1012,
      "step": 103500
    },
    {
      "epoch": 3.94,
      "learning_rate": 1.803241984935085e-05,
      "loss": 0.1008,
      "step": 104000
    },
    {
      "epoch": 3.96,
      "learning_rate": 1.8022975888565048e-05,
      "loss": 0.1014,
      "step": 104500
    },
    {
      "epoch": 3.97,
      "learning_rate": 1.801351300200613e-05,
      "loss": 0.1021,
      "step": 105000
    },
    {
      "epoch": 3.99,
      "learning_rate": 1.800405011544722e-05,
      "loss": 0.0991,
      "step": 105500
    },
    {
      "epoch": 4.0,
      "eval_bleu": 91.4665,
      "eval_gen_len": 16.423,
      "eval_loss": 0.09734214097261429,
      "eval_runtime": 1213.4766,
      "eval_samples_per_second": 37.633,
      "eval_steps_per_second": 1.177,
      "step": 105676
    },
    {
      "epoch": 4.01,
      "learning_rate": 1.79945872288883e-05,
      "loss": 0.0962,
      "step": 106000
    },
    {
      "epoch": 4.03,
      "learning_rate": 1.7985124342329386e-05,
      "loss": 0.0908,
      "step": 106500
    },
    {
      "epoch": 4.05,
      "learning_rate": 1.797568038154359e-05,
      "loss": 0.0924,
      "step": 107000
    },
    {
      "epoch": 4.07,
      "learning_rate": 1.7966217494984673e-05,
      "loss": 0.0894,
      "step": 107500
    },
    {
      "epoch": 4.09,
      "learning_rate": 1.7956754608425756e-05,
      "loss": 0.0908,
      "step": 108000
    },
    {
      "epoch": 4.11,
      "learning_rate": 1.794729172186684e-05,
      "loss": 0.094,
      "step": 108500
    },
    {
      "epoch": 4.13,
      "learning_rate": 1.7937828835307924e-05,
      "loss": 0.0921,
      "step": 109000
    },
    {
      "epoch": 4.14,
      "learning_rate": 1.7928365948749007e-05,
      "loss": 0.0891,
      "step": 109500
    },
    {
      "epoch": 4.16,
      "learning_rate": 1.791890306219009e-05,
      "loss": 0.0895,
      "step": 110000
    },
    {
      "epoch": 4.18,
      "learning_rate": 1.7909459101404294e-05,
      "loss": 0.0905,
      "step": 110500
    },
    {
      "epoch": 4.2,
      "learning_rate": 1.7899996214845378e-05,
      "loss": 0.091,
      "step": 111000
    },
    {
      "epoch": 4.22,
      "learning_rate": 1.789053332828646e-05,
      "loss": 0.0949,
      "step": 111500
    },
    {
      "epoch": 4.24,
      "learning_rate": 1.7881070441727545e-05,
      "loss": 0.0922,
      "step": 112000
    },
    {
      "epoch": 4.26,
      "learning_rate": 1.7871607555168632e-05,
      "loss": 0.0929,
      "step": 112500
    },
    {
      "epoch": 4.28,
      "learning_rate": 1.7862144668609712e-05,
      "loss": 0.0907,
      "step": 113000
    },
    {
      "epoch": 4.3,
      "learning_rate": 1.78526817820508e-05,
      "loss": 0.0932,
      "step": 113500
    },
    {
      "epoch": 4.32,
      "learning_rate": 1.7843218895491883e-05,
      "loss": 0.093,
      "step": 114000
    },
    {
      "epoch": 4.33,
      "learning_rate": 1.7833774934706086e-05,
      "loss": 0.0921,
      "step": 114500
    },
    {
      "epoch": 4.35,
      "learning_rate": 1.7824330973920285e-05,
      "loss": 0.0902,
      "step": 115000
    },
    {
      "epoch": 4.37,
      "learning_rate": 1.781486808736137e-05,
      "loss": 0.0911,
      "step": 115500
    },
    {
      "epoch": 4.39,
      "learning_rate": 1.7805405200802456e-05,
      "loss": 0.0903,
      "step": 116000
    },
    {
      "epoch": 4.41,
      "learning_rate": 1.7795942314243536e-05,
      "loss": 0.094,
      "step": 116500
    },
    {
      "epoch": 4.43,
      "learning_rate": 1.778649835345774e-05,
      "loss": 0.0912,
      "step": 117000
    },
    {
      "epoch": 4.45,
      "learning_rate": 1.7777035466898823e-05,
      "loss": 0.0913,
      "step": 117500
    },
    {
      "epoch": 4.47,
      "learning_rate": 1.776757258033991e-05,
      "loss": 0.0938,
      "step": 118000
    },
    {
      "epoch": 4.49,
      "learning_rate": 1.775810969378099e-05,
      "loss": 0.0903,
      "step": 118500
    },
    {
      "epoch": 4.5,
      "learning_rate": 1.7748646807222077e-05,
      "loss": 0.0917,
      "step": 119000
    },
    {
      "epoch": 4.52,
      "learning_rate": 1.773920284643628e-05,
      "loss": 0.0906,
      "step": 119500
    },
    {
      "epoch": 4.54,
      "learning_rate": 1.7729739959877364e-05,
      "loss": 0.0928,
      "step": 120000
    },
    {
      "epoch": 4.56,
      "learning_rate": 1.7720277073318444e-05,
      "loss": 0.0919,
      "step": 120500
    },
    {
      "epoch": 4.58,
      "learning_rate": 1.771081418675953e-05,
      "loss": 0.0933,
      "step": 121000
    },
    {
      "epoch": 4.6,
      "learning_rate": 1.7701370225973734e-05,
      "loss": 0.0925,
      "step": 121500
    },
    {
      "epoch": 4.62,
      "learning_rate": 1.7691907339414814e-05,
      "loss": 0.0909,
      "step": 122000
    },
    {
      "epoch": 4.64,
      "learning_rate": 1.76824444528559e-05,
      "loss": 0.0923,
      "step": 122500
    },
    {
      "epoch": 4.66,
      "learning_rate": 1.7672981566296985e-05,
      "loss": 0.0898,
      "step": 123000
    },
    {
      "epoch": 4.67,
      "learning_rate": 1.766351867973807e-05,
      "loss": 0.0906,
      "step": 123500
    },
    {
      "epoch": 4.69,
      "learning_rate": 1.7654055793179153e-05,
      "loss": 0.0921,
      "step": 124000
    },
    {
      "epoch": 4.71,
      "learning_rate": 1.7644592906620236e-05,
      "loss": 0.09,
      "step": 124500
    },
    {
      "epoch": 4.73,
      "learning_rate": 1.763514894583444e-05,
      "loss": 0.0898,
      "step": 125000
    },
    {
      "epoch": 4.75,
      "learning_rate": 1.7625686059275523e-05,
      "loss": 0.091,
      "step": 125500
    },
    {
      "epoch": 4.77,
      "learning_rate": 1.7616223172716607e-05,
      "loss": 0.09,
      "step": 126000
    },
    {
      "epoch": 4.79,
      "learning_rate": 1.760676028615769e-05,
      "loss": 0.0943,
      "step": 126500
    },
    {
      "epoch": 4.81,
      "learning_rate": 1.7597297399598777e-05,
      "loss": 0.0904,
      "step": 127000
    },
    {
      "epoch": 4.83,
      "learning_rate": 1.7587834513039858e-05,
      "loss": 0.0919,
      "step": 127500
    },
    {
      "epoch": 4.84,
      "learning_rate": 1.7578371626480945e-05,
      "loss": 0.0904,
      "step": 128000
    },
    {
      "epoch": 4.86,
      "learning_rate": 1.7568927665695144e-05,
      "loss": 0.0935,
      "step": 128500
    },
    {
      "epoch": 4.88,
      "learning_rate": 1.7559464779136228e-05,
      "loss": 0.0916,
      "step": 129000
    },
    {
      "epoch": 4.9,
      "learning_rate": 1.755000189257731e-05,
      "loss": 0.0886,
      "step": 129500
    },
    {
      "epoch": 4.92,
      "learning_rate": 1.75405390060184e-05,
      "loss": 0.0943,
      "step": 130000
    },
    {
      "epoch": 4.94,
      "learning_rate": 1.75310950452326e-05,
      "loss": 0.0911,
      "step": 130500
    },
    {
      "epoch": 4.96,
      "learning_rate": 1.7521632158673682e-05,
      "loss": 0.0917,
      "step": 131000
    },
    {
      "epoch": 4.98,
      "learning_rate": 1.751216927211477e-05,
      "loss": 0.0925,
      "step": 131500
    },
    {
      "epoch": 5.0,
      "learning_rate": 1.7502706385555853e-05,
      "loss": 0.0917,
      "step": 132000
    },
    {
      "epoch": 5.0,
      "eval_bleu": 91.7718,
      "eval_gen_len": 16.394,
      "eval_loss": 0.09411130845546722,
      "eval_runtime": 1205.451,
      "eval_samples_per_second": 37.884,
      "eval_steps_per_second": 1.185,
      "step": 132095
    },
    {
      "epoch": 5.02,
      "learning_rate": 1.7493243498996936e-05,
      "loss": 0.0822,
      "step": 132500
    },
    {
      "epoch": 5.03,
      "learning_rate": 1.748378061243802e-05,
      "loss": 0.0819,
      "step": 133000
    },
    {
      "epoch": 5.05,
      "learning_rate": 1.7474317725879104e-05,
      "loss": 0.0835,
      "step": 133500
    },
    {
      "epoch": 5.07,
      "learning_rate": 1.7464854839320187e-05,
      "loss": 0.0805,
      "step": 134000
    },
    {
      "epoch": 5.09,
      "learning_rate": 1.745539195276127e-05,
      "loss": 0.0806,
      "step": 134500
    },
    {
      "epoch": 5.11,
      "learning_rate": 1.7445947991975474e-05,
      "loss": 0.085,
      "step": 135000
    },
    {
      "epoch": 5.13,
      "learning_rate": 1.7436485105416558e-05,
      "loss": 0.0823,
      "step": 135500
    },
    {
      "epoch": 5.15,
      "learning_rate": 1.742702221885764e-05,
      "loss": 0.0832,
      "step": 136000
    },
    {
      "epoch": 5.17,
      "learning_rate": 1.7417559332298725e-05,
      "loss": 0.085,
      "step": 136500
    },
    {
      "epoch": 5.19,
      "learning_rate": 1.7408096445739812e-05,
      "loss": 0.0823,
      "step": 137000
    },
    {
      "epoch": 5.2,
      "learning_rate": 1.7398633559180892e-05,
      "loss": 0.0837,
      "step": 137500
    },
    {
      "epoch": 5.22,
      "learning_rate": 1.7389170672621976e-05,
      "loss": 0.0835,
      "step": 138000
    },
    {
      "epoch": 5.24,
      "learning_rate": 1.7379707786063063e-05,
      "loss": 0.0808,
      "step": 138500
    },
    {
      "epoch": 5.26,
      "learning_rate": 1.7370244899504147e-05,
      "loss": 0.0815,
      "step": 139000
    },
    {
      "epoch": 5.28,
      "learning_rate": 1.7360819864491465e-05,
      "loss": 0.0857,
      "step": 139500
    },
    {
      "epoch": 5.3,
      "learning_rate": 1.735135697793255e-05,
      "loss": 0.0824,
      "step": 140000
    },
    {
      "epoch": 5.32,
      "learning_rate": 1.7341894091373633e-05,
      "loss": 0.0833,
      "step": 140500
    },
    {
      "epoch": 5.34,
      "learning_rate": 1.733243120481472e-05,
      "loss": 0.0828,
      "step": 141000
    },
    {
      "epoch": 5.36,
      "learning_rate": 1.73229683182558e-05,
      "loss": 0.085,
      "step": 141500
    },
    {
      "epoch": 5.37,
      "learning_rate": 1.7313505431696887e-05,
      "loss": 0.0817,
      "step": 142000
    },
    {
      "epoch": 5.39,
      "learning_rate": 1.730406147091109e-05,
      "loss": 0.0822,
      "step": 142500
    },
    {
      "epoch": 5.41,
      "learning_rate": 1.729459858435217e-05,
      "loss": 0.0851,
      "step": 143000
    },
    {
      "epoch": 5.43,
      "learning_rate": 1.7285135697793257e-05,
      "loss": 0.0818,
      "step": 143500
    },
    {
      "epoch": 5.45,
      "learning_rate": 1.727567281123434e-05,
      "loss": 0.083,
      "step": 144000
    },
    {
      "epoch": 5.47,
      "learning_rate": 1.7266228850448544e-05,
      "loss": 0.0833,
      "step": 144500
    },
    {
      "epoch": 5.49,
      "learning_rate": 1.7256765963889624e-05,
      "loss": 0.0818,
      "step": 145000
    },
    {
      "epoch": 5.51,
      "learning_rate": 1.724730307733071e-05,
      "loss": 0.0833,
      "step": 145500
    },
    {
      "epoch": 5.53,
      "learning_rate": 1.7237840190771795e-05,
      "loss": 0.0842,
      "step": 146000
    },
    {
      "epoch": 5.55,
      "learning_rate": 1.722837730421288e-05,
      "loss": 0.0814,
      "step": 146500
    },
    {
      "epoch": 5.56,
      "learning_rate": 1.7218914417653962e-05,
      "loss": 0.0806,
      "step": 147000
    },
    {
      "epoch": 5.58,
      "learning_rate": 1.7209470456868165e-05,
      "loss": 0.0837,
      "step": 147500
    },
    {
      "epoch": 5.6,
      "learning_rate": 1.720000757030925e-05,
      "loss": 0.0842,
      "step": 148000
    },
    {
      "epoch": 5.62,
      "learning_rate": 1.7190544683750333e-05,
      "loss": 0.0822,
      "step": 148500
    },
    {
      "epoch": 5.64,
      "learning_rate": 1.7181081797191416e-05,
      "loss": 0.0823,
      "step": 149000
    },
    {
      "epoch": 5.66,
      "learning_rate": 1.71716189106325e-05,
      "loss": 0.0874,
      "step": 149500
    },
    {
      "epoch": 5.68,
      "learning_rate": 1.7162156024073584e-05,
      "loss": 0.0832,
      "step": 150000
    },
    {
      "epoch": 5.7,
      "learning_rate": 1.7152712063287787e-05,
      "loss": 0.084,
      "step": 150500
    },
    {
      "epoch": 5.72,
      "learning_rate": 1.714324917672887e-05,
      "loss": 0.083,
      "step": 151000
    },
    {
      "epoch": 5.73,
      "learning_rate": 1.7133805215943073e-05,
      "loss": 0.0825,
      "step": 151500
    },
    {
      "epoch": 5.75,
      "learning_rate": 1.7124342329384157e-05,
      "loss": 0.0823,
      "step": 152000
    },
    {
      "epoch": 5.77,
      "learning_rate": 1.711487944282524e-05,
      "loss": 0.0829,
      "step": 152500
    },
    {
      "epoch": 5.79,
      "learning_rate": 1.7105416556266324e-05,
      "loss": 0.0833,
      "step": 153000
    },
    {
      "epoch": 5.81,
      "learning_rate": 1.7095953669707408e-05,
      "loss": 0.0828,
      "step": 153500
    },
    {
      "epoch": 5.83,
      "learning_rate": 1.708649078314849e-05,
      "loss": 0.081,
      "step": 154000
    },
    {
      "epoch": 5.85,
      "learning_rate": 1.707702789658958e-05,
      "loss": 0.0834,
      "step": 154500
    },
    {
      "epoch": 5.87,
      "learning_rate": 1.7067565010030662e-05,
      "loss": 0.0823,
      "step": 155000
    },
    {
      "epoch": 5.89,
      "learning_rate": 1.7058102123471746e-05,
      "loss": 0.0833,
      "step": 155500
    },
    {
      "epoch": 5.9,
      "learning_rate": 1.704863923691283e-05,
      "loss": 0.0836,
      "step": 156000
    },
    {
      "epoch": 5.92,
      "learning_rate": 1.7039176350353913e-05,
      "loss": 0.0836,
      "step": 156500
    },
    {
      "epoch": 5.94,
      "learning_rate": 1.7029713463794997e-05,
      "loss": 0.0829,
      "step": 157000
    },
    {
      "epoch": 5.96,
      "learning_rate": 1.702025057723608e-05,
      "loss": 0.0835,
      "step": 157500
    },
    {
      "epoch": 5.98,
      "learning_rate": 1.7010806616450284e-05,
      "loss": 0.0841,
      "step": 158000
    },
    {
      "epoch": 6.0,
      "learning_rate": 1.7001343729891367e-05,
      "loss": 0.0822,
      "step": 158500
    },
    {
      "epoch": 6.0,
      "eval_bleu": 91.8571,
      "eval_gen_len": 16.3677,
      "eval_loss": 0.0922790989279747,
      "eval_runtime": 1204.098,
      "eval_samples_per_second": 37.926,
      "eval_steps_per_second": 1.186,
      "step": 158514
    },
    {
      "epoch": 6.02,
      "learning_rate": 1.699188084333245e-05,
      "loss": 0.0756,
      "step": 159000
    },
    {
      "epoch": 6.04,
      "learning_rate": 1.6982417956773535e-05,
      "loss": 0.0749,
      "step": 159500
    },
    {
      "epoch": 6.06,
      "learning_rate": 1.6972955070214622e-05,
      "loss": 0.0736,
      "step": 160000
    },
    {
      "epoch": 6.08,
      "learning_rate": 1.6963492183655702e-05,
      "loss": 0.0743,
      "step": 160500
    },
    {
      "epoch": 6.09,
      "learning_rate": 1.6954048222869905e-05,
      "loss": 0.0743,
      "step": 161000
    },
    {
      "epoch": 6.11,
      "learning_rate": 1.6944604262084108e-05,
      "loss": 0.0748,
      "step": 161500
    },
    {
      "epoch": 6.13,
      "learning_rate": 1.693514137552519e-05,
      "loss": 0.0728,
      "step": 162000
    },
    {
      "epoch": 6.15,
      "learning_rate": 1.6925678488966275e-05,
      "loss": 0.0759,
      "step": 162500
    },
    {
      "epoch": 6.17,
      "learning_rate": 1.691621560240736e-05,
      "loss": 0.076,
      "step": 163000
    },
    {
      "epoch": 6.19,
      "learning_rate": 1.6906752715848446e-05,
      "loss": 0.0752,
      "step": 163500
    },
    {
      "epoch": 6.21,
      "learning_rate": 1.6897289829289526e-05,
      "loss": 0.0754,
      "step": 164000
    },
    {
      "epoch": 6.23,
      "learning_rate": 1.6887826942730613e-05,
      "loss": 0.0741,
      "step": 164500
    },
    {
      "epoch": 6.25,
      "learning_rate": 1.6878364056171697e-05,
      "loss": 0.0786,
      "step": 165000
    },
    {
      "epoch": 6.26,
      "learning_rate": 1.686890116961278e-05,
      "loss": 0.076,
      "step": 165500
    },
    {
      "epoch": 6.28,
      "learning_rate": 1.685945720882698e-05,
      "loss": 0.075,
      "step": 166000
    },
    {
      "epoch": 6.3,
      "learning_rate": 1.6850013248041183e-05,
      "loss": 0.0744,
      "step": 166500
    },
    {
      "epoch": 6.32,
      "learning_rate": 1.6840569287255386e-05,
      "loss": 0.0758,
      "step": 167000
    },
    {
      "epoch": 6.34,
      "learning_rate": 1.683110640069647e-05,
      "loss": 0.0756,
      "step": 167500
    },
    {
      "epoch": 6.36,
      "learning_rate": 1.6821643514137553e-05,
      "loss": 0.0748,
      "step": 168000
    },
    {
      "epoch": 6.38,
      "learning_rate": 1.6812180627578637e-05,
      "loss": 0.0776,
      "step": 168500
    },
    {
      "epoch": 6.4,
      "learning_rate": 1.6802717741019724e-05,
      "loss": 0.076,
      "step": 169000
    },
    {
      "epoch": 6.42,
      "learning_rate": 1.6793254854460804e-05,
      "loss": 0.0744,
      "step": 169500
    },
    {
      "epoch": 6.43,
      "learning_rate": 1.678379196790189e-05,
      "loss": 0.0757,
      "step": 170000
    },
    {
      "epoch": 6.45,
      "learning_rate": 1.6774329081342975e-05,
      "loss": 0.078,
      "step": 170500
    },
    {
      "epoch": 6.47,
      "learning_rate": 1.676486619478406e-05,
      "loss": 0.0772,
      "step": 171000
    },
    {
      "epoch": 6.49,
      "learning_rate": 1.6755403308225142e-05,
      "loss": 0.0763,
      "step": 171500
    },
    {
      "epoch": 6.51,
      "learning_rate": 1.6745940421666226e-05,
      "loss": 0.0734,
      "step": 172000
    },
    {
      "epoch": 6.53,
      "learning_rate": 1.673649646088043e-05,
      "loss": 0.0758,
      "step": 172500
    },
    {
      "epoch": 6.55,
      "learning_rate": 1.6727033574321513e-05,
      "loss": 0.0768,
      "step": 173000
    },
    {
      "epoch": 6.57,
      "learning_rate": 1.6717570687762596e-05,
      "loss": 0.0762,
      "step": 173500
    },
    {
      "epoch": 6.59,
      "learning_rate": 1.670810780120368e-05,
      "loss": 0.0774,
      "step": 174000
    },
    {
      "epoch": 6.61,
      "learning_rate": 1.6698644914644764e-05,
      "loss": 0.0744,
      "step": 174500
    },
    {
      "epoch": 6.62,
      "learning_rate": 1.6689200953858967e-05,
      "loss": 0.0775,
      "step": 175000
    },
    {
      "epoch": 6.64,
      "learning_rate": 1.667973806730005e-05,
      "loss": 0.0749,
      "step": 175500
    },
    {
      "epoch": 6.66,
      "learning_rate": 1.6670275180741134e-05,
      "loss": 0.0778,
      "step": 176000
    },
    {
      "epoch": 6.68,
      "learning_rate": 1.6660812294182218e-05,
      "loss": 0.0757,
      "step": 176500
    },
    {
      "epoch": 6.7,
      "learning_rate": 1.66513494076233e-05,
      "loss": 0.0751,
      "step": 177000
    },
    {
      "epoch": 6.72,
      "learning_rate": 1.664188652106439e-05,
      "loss": 0.0756,
      "step": 177500
    },
    {
      "epoch": 6.74,
      "learning_rate": 1.6632442560278588e-05,
      "loss": 0.0734,
      "step": 178000
    },
    {
      "epoch": 6.76,
      "learning_rate": 1.662297967371967e-05,
      "loss": 0.076,
      "step": 178500
    },
    {
      "epoch": 6.78,
      "learning_rate": 1.661351678716076e-05,
      "loss": 0.0754,
      "step": 179000
    },
    {
      "epoch": 6.79,
      "learning_rate": 1.6604053900601842e-05,
      "loss": 0.076,
      "step": 179500
    },
    {
      "epoch": 6.81,
      "learning_rate": 1.6594591014042926e-05,
      "loss": 0.0739,
      "step": 180000
    },
    {
      "epoch": 6.83,
      "learning_rate": 1.6585165979030245e-05,
      "loss": 0.0749,
      "step": 180500
    },
    {
      "epoch": 6.85,
      "learning_rate": 1.6575703092471328e-05,
      "loss": 0.0747,
      "step": 181000
    },
    {
      "epoch": 6.87,
      "learning_rate": 1.6566240205912415e-05,
      "loss": 0.0777,
      "step": 181500
    },
    {
      "epoch": 6.89,
      "learning_rate": 1.6556777319353496e-05,
      "loss": 0.0768,
      "step": 182000
    },
    {
      "epoch": 6.91,
      "learning_rate": 1.6547314432794583e-05,
      "loss": 0.0777,
      "step": 182500
    },
    {
      "epoch": 6.93,
      "learning_rate": 1.6537851546235666e-05,
      "loss": 0.0762,
      "step": 183000
    },
    {
      "epoch": 6.95,
      "learning_rate": 1.6528388659676747e-05,
      "loss": 0.0766,
      "step": 183500
    },
    {
      "epoch": 6.96,
      "learning_rate": 1.6518925773117834e-05,
      "loss": 0.0757,
      "step": 184000
    },
    {
      "epoch": 6.98,
      "learning_rate": 1.6509481812332037e-05,
      "loss": 0.0773,
      "step": 184500
    },
    {
      "epoch": 7.0,
      "eval_bleu": 92.0146,
      "eval_gen_len": 16.3778,
      "eval_loss": 0.0919041708111763,
      "eval_runtime": 1212.2042,
      "eval_samples_per_second": 37.673,
      "eval_steps_per_second": 1.178,
      "step": 184933
    },
    {
      "epoch": 7.0,
      "learning_rate": 1.650001892577312e-05,
      "loss": 0.0745,
      "step": 185000
    },
    {
      "epoch": 7.02,
      "learning_rate": 1.6490556039214204e-05,
      "loss": 0.0657,
      "step": 185500
    },
    {
      "epoch": 7.04,
      "learning_rate": 1.6481093152655288e-05,
      "loss": 0.0673,
      "step": 186000
    },
    {
      "epoch": 7.06,
      "learning_rate": 1.647163026609637e-05,
      "loss": 0.0688,
      "step": 186500
    },
    {
      "epoch": 7.08,
      "learning_rate": 1.6462167379537455e-05,
      "loss": 0.0674,
      "step": 187000
    },
    {
      "epoch": 7.1,
      "learning_rate": 1.645270449297854e-05,
      "loss": 0.0669,
      "step": 187500
    },
    {
      "epoch": 7.12,
      "learning_rate": 1.6443241606419626e-05,
      "loss": 0.07,
      "step": 188000
    },
    {
      "epoch": 7.14,
      "learning_rate": 1.6433778719860706e-05,
      "loss": 0.0683,
      "step": 188500
    },
    {
      "epoch": 7.15,
      "learning_rate": 1.642433475907491e-05,
      "loss": 0.0672,
      "step": 189000
    },
    {
      "epoch": 7.17,
      "learning_rate": 1.6414871872515993e-05,
      "loss": 0.0675,
      "step": 189500
    },
    {
      "epoch": 7.19,
      "learning_rate": 1.640540898595708e-05,
      "loss": 0.0688,
      "step": 190000
    },
    {
      "epoch": 7.21,
      "learning_rate": 1.639594609939816e-05,
      "loss": 0.0691,
      "step": 190500
    },
    {
      "epoch": 7.23,
      "learning_rate": 1.6386483212839247e-05,
      "loss": 0.0676,
      "step": 191000
    },
    {
      "epoch": 7.25,
      "learning_rate": 1.6377039252053447e-05,
      "loss": 0.0706,
      "step": 191500
    },
    {
      "epoch": 7.27,
      "learning_rate": 1.636757636549453e-05,
      "loss": 0.0701,
      "step": 192000
    },
    {
      "epoch": 7.29,
      "learning_rate": 1.6358113478935614e-05,
      "loss": 0.07,
      "step": 192500
    },
    {
      "epoch": 7.31,
      "learning_rate": 1.63486505923767e-05,
      "loss": 0.069,
      "step": 193000
    },
    {
      "epoch": 7.32,
      "learning_rate": 1.6339187705817785e-05,
      "loss": 0.068,
      "step": 193500
    },
    {
      "epoch": 7.34,
      "learning_rate": 1.632972481925887e-05,
      "loss": 0.0678,
      "step": 194000
    },
    {
      "epoch": 7.36,
      "learning_rate": 1.632028085847307e-05,
      "loss": 0.0701,
      "step": 194500
    },
    {
      "epoch": 7.38,
      "learning_rate": 1.6310817971914155e-05,
      "loss": 0.0702,
      "step": 195000
    },
    {
      "epoch": 7.4,
      "learning_rate": 1.630135508535524e-05,
      "loss": 0.0688,
      "step": 195500
    },
    {
      "epoch": 7.42,
      "learning_rate": 1.6291911124569438e-05,
      "loss": 0.0687,
      "step": 196000
    },
    {
      "epoch": 7.44,
      "learning_rate": 1.6282448238010525e-05,
      "loss": 0.0676,
      "step": 196500
    },
    {
      "epoch": 7.46,
      "learning_rate": 1.627298535145161e-05,
      "loss": 0.0697,
      "step": 197000
    },
    {
      "epoch": 7.48,
      "learning_rate": 1.6263522464892693e-05,
      "loss": 0.0686,
      "step": 197500
    },
    {
      "epoch": 7.49,
      "learning_rate": 1.6254059578333776e-05,
      "loss": 0.0683,
      "step": 198000
    },
    {
      "epoch": 7.51,
      "learning_rate": 1.624461561754798e-05,
      "loss": 0.0684,
      "step": 198500
    },
    {
      "epoch": 7.53,
      "learning_rate": 1.6235152730989063e-05,
      "loss": 0.0687,
      "step": 199000
    },
    {
      "epoch": 7.55,
      "learning_rate": 1.6225689844430147e-05,
      "loss": 0.0698,
      "step": 199500
    },
    {
      "epoch": 7.57,
      "learning_rate": 1.621622695787123e-05,
      "loss": 0.0689,
      "step": 200000
    },
    {
      "epoch": 7.59,
      "learning_rate": 1.6206764071312314e-05,
      "loss": 0.0717,
      "step": 200500
    },
    {
      "epoch": 7.61,
      "learning_rate": 1.6197301184753398e-05,
      "loss": 0.0706,
      "step": 201000
    },
    {
      "epoch": 7.63,
      "learning_rate": 1.61878572239676e-05,
      "loss": 0.0686,
      "step": 201500
    },
    {
      "epoch": 7.65,
      "learning_rate": 1.6178394337408684e-05,
      "loss": 0.0704,
      "step": 202000
    },
    {
      "epoch": 7.66,
      "learning_rate": 1.6168931450849768e-05,
      "loss": 0.0701,
      "step": 202500
    },
    {
      "epoch": 7.68,
      "learning_rate": 1.615946856429085e-05,
      "loss": 0.0711,
      "step": 203000
    },
    {
      "epoch": 7.7,
      "learning_rate": 1.6150005677731935e-05,
      "loss": 0.0697,
      "step": 203500
    },
    {
      "epoch": 7.72,
      "learning_rate": 1.6140542791173022e-05,
      "loss": 0.0698,
      "step": 204000
    },
    {
      "epoch": 7.74,
      "learning_rate": 1.6131079904614103e-05,
      "loss": 0.0731,
      "step": 204500
    },
    {
      "epoch": 7.76,
      "learning_rate": 1.612161701805519e-05,
      "loss": 0.0692,
      "step": 205000
    },
    {
      "epoch": 7.78,
      "learning_rate": 1.6112173057269393e-05,
      "loss": 0.0725,
      "step": 205500
    },
    {
      "epoch": 7.8,
      "learning_rate": 1.6102710170710473e-05,
      "loss": 0.0698,
      "step": 206000
    },
    {
      "epoch": 7.82,
      "learning_rate": 1.609324728415156e-05,
      "loss": 0.071,
      "step": 206500
    },
    {
      "epoch": 7.84,
      "learning_rate": 1.6083784397592644e-05,
      "loss": 0.0715,
      "step": 207000
    },
    {
      "epoch": 7.85,
      "learning_rate": 1.6074321511033727e-05,
      "loss": 0.0717,
      "step": 207500
    },
    {
      "epoch": 7.87,
      "learning_rate": 1.606485862447481e-05,
      "loss": 0.07,
      "step": 208000
    },
    {
      "epoch": 7.89,
      "learning_rate": 1.6055414663689014e-05,
      "loss": 0.0694,
      "step": 208500
    },
    {
      "epoch": 7.91,
      "learning_rate": 1.6045951777130098e-05,
      "loss": 0.0693,
      "step": 209000
    },
    {
      "epoch": 7.93,
      "learning_rate": 1.603648889057118e-05,
      "loss": 0.071,
      "step": 209500
    },
    {
      "epoch": 7.95,
      "learning_rate": 1.6027026004012265e-05,
      "loss": 0.0713,
      "step": 210000
    },
    {
      "epoch": 7.97,
      "learning_rate": 1.601756311745335e-05,
      "loss": 0.0701,
      "step": 210500
    },
    {
      "epoch": 7.99,
      "learning_rate": 1.6008100230894436e-05,
      "loss": 0.0705,
      "step": 211000
    },
    {
      "epoch": 8.0,
      "eval_bleu": 92.0134,
      "eval_gen_len": 16.388,
      "eval_loss": 0.09185712039470673,
      "eval_runtime": 1202.772,
      "eval_samples_per_second": 37.968,
      "eval_steps_per_second": 1.187,
      "step": 211352
    },
    {
      "epoch": 8.01,
      "learning_rate": 1.5998656270108635e-05,
      "loss": 0.0665,
      "step": 211500
    },
    {
      "epoch": 8.02,
      "learning_rate": 1.598919338354972e-05,
      "loss": 0.0607,
      "step": 212000
    },
    {
      "epoch": 8.04,
      "learning_rate": 1.5979730496990803e-05,
      "loss": 0.0613,
      "step": 212500
    },
    {
      "epoch": 8.06,
      "learning_rate": 1.5970267610431886e-05,
      "loss": 0.0611,
      "step": 213000
    },
    {
      "epoch": 8.08,
      "learning_rate": 1.596080472387297e-05,
      "loss": 0.0616,
      "step": 213500
    },
    {
      "epoch": 8.1,
      "learning_rate": 1.5951360763087173e-05,
      "loss": 0.0644,
      "step": 214000
    },
    {
      "epoch": 8.12,
      "learning_rate": 1.594189787652826e-05,
      "loss": 0.0605,
      "step": 214500
    },
    {
      "epoch": 8.14,
      "learning_rate": 1.593243498996934e-05,
      "loss": 0.063,
      "step": 215000
    },
    {
      "epoch": 8.16,
      "learning_rate": 1.5922972103410427e-05,
      "loss": 0.0635,
      "step": 215500
    },
    {
      "epoch": 8.18,
      "learning_rate": 1.591350921685151e-05,
      "loss": 0.0635,
      "step": 216000
    },
    {
      "epoch": 8.19,
      "learning_rate": 1.590406525606571e-05,
      "loss": 0.0634,
      "step": 216500
    },
    {
      "epoch": 8.21,
      "learning_rate": 1.5894602369506794e-05,
      "loss": 0.0641,
      "step": 217000
    },
    {
      "epoch": 8.23,
      "learning_rate": 1.588513948294788e-05,
      "loss": 0.0626,
      "step": 217500
    },
    {
      "epoch": 8.25,
      "learning_rate": 1.5875676596388965e-05,
      "loss": 0.0629,
      "step": 218000
    },
    {
      "epoch": 8.27,
      "learning_rate": 1.586621370983005e-05,
      "loss": 0.0646,
      "step": 218500
    },
    {
      "epoch": 8.29,
      "learning_rate": 1.5856750823271132e-05,
      "loss": 0.0638,
      "step": 219000
    },
    {
      "epoch": 8.31,
      "learning_rate": 1.5847287936712216e-05,
      "loss": 0.0637,
      "step": 219500
    },
    {
      "epoch": 8.33,
      "learning_rate": 1.5837843975926415e-05,
      "loss": 0.064,
      "step": 220000
    },
    {
      "epoch": 8.35,
      "learning_rate": 1.5828381089367502e-05,
      "loss": 0.0643,
      "step": 220500
    },
    {
      "epoch": 8.37,
      "learning_rate": 1.5818918202808586e-05,
      "loss": 0.0616,
      "step": 221000
    },
    {
      "epoch": 8.38,
      "learning_rate": 1.580945531624967e-05,
      "loss": 0.0633,
      "step": 221500
    },
    {
      "epoch": 8.4,
      "learning_rate": 1.5799992429690754e-05,
      "loss": 0.0625,
      "step": 222000
    },
    {
      "epoch": 8.42,
      "learning_rate": 1.5790529543131837e-05,
      "loss": 0.0636,
      "step": 222500
    },
    {
      "epoch": 8.44,
      "learning_rate": 1.578108558234604e-05,
      "loss": 0.0638,
      "step": 223000
    },
    {
      "epoch": 8.46,
      "learning_rate": 1.5771622695787124e-05,
      "loss": 0.0643,
      "step": 223500
    },
    {
      "epoch": 8.48,
      "learning_rate": 1.5762159809228207e-05,
      "loss": 0.0625,
      "step": 224000
    },
    {
      "epoch": 8.5,
      "learning_rate": 1.575269692266929e-05,
      "loss": 0.0624,
      "step": 224500
    },
    {
      "epoch": 8.52,
      "learning_rate": 1.5743252961883494e-05,
      "loss": 0.0623,
      "step": 225000
    },
    {
      "epoch": 8.54,
      "learning_rate": 1.5733790075324578e-05,
      "loss": 0.0649,
      "step": 225500
    },
    {
      "epoch": 8.55,
      "learning_rate": 1.572432718876566e-05,
      "loss": 0.0627,
      "step": 226000
    },
    {
      "epoch": 8.57,
      "learning_rate": 1.571486430220675e-05,
      "loss": 0.0639,
      "step": 226500
    },
    {
      "epoch": 8.59,
      "learning_rate": 1.570540141564783e-05,
      "loss": 0.0635,
      "step": 227000
    },
    {
      "epoch": 8.61,
      "learning_rate": 1.5695938529088916e-05,
      "loss": 0.062,
      "step": 227500
    },
    {
      "epoch": 8.63,
      "learning_rate": 1.568647564253e-05,
      "loss": 0.0683,
      "step": 228000
    },
    {
      "epoch": 8.65,
      "learning_rate": 1.5677031681744202e-05,
      "loss": 0.0632,
      "step": 228500
    },
    {
      "epoch": 8.67,
      "learning_rate": 1.5667568795185283e-05,
      "loss": 0.0648,
      "step": 229000
    },
    {
      "epoch": 8.69,
      "learning_rate": 1.565810590862637e-05,
      "loss": 0.0652,
      "step": 229500
    },
    {
      "epoch": 8.71,
      "learning_rate": 1.5648643022067453e-05,
      "loss": 0.0653,
      "step": 230000
    },
    {
      "epoch": 8.72,
      "learning_rate": 1.5639199061281653e-05,
      "loss": 0.0638,
      "step": 230500
    },
    {
      "epoch": 8.74,
      "learning_rate": 1.562973617472274e-05,
      "loss": 0.0655,
      "step": 231000
    },
    {
      "epoch": 8.76,
      "learning_rate": 1.5620273288163824e-05,
      "loss": 0.063,
      "step": 231500
    },
    {
      "epoch": 8.78,
      "learning_rate": 1.5610810401604907e-05,
      "loss": 0.0647,
      "step": 232000
    },
    {
      "epoch": 8.8,
      "learning_rate": 1.560134751504599e-05,
      "loss": 0.0629,
      "step": 232500
    },
    {
      "epoch": 8.82,
      "learning_rate": 1.5591884628487075e-05,
      "loss": 0.063,
      "step": 233000
    },
    {
      "epoch": 8.84,
      "learning_rate": 1.558242174192816e-05,
      "loss": 0.0634,
      "step": 233500
    },
    {
      "epoch": 8.86,
      "learning_rate": 1.5572958855369242e-05,
      "loss": 0.0642,
      "step": 234000
    },
    {
      "epoch": 8.88,
      "learning_rate": 1.5563495968810326e-05,
      "loss": 0.0637,
      "step": 234500
    },
    {
      "epoch": 8.9,
      "learning_rate": 1.5554033082251413e-05,
      "loss": 0.0648,
      "step": 235000
    },
    {
      "epoch": 8.91,
      "learning_rate": 1.5544589121465616e-05,
      "loss": 0.0647,
      "step": 235500
    },
    {
      "epoch": 8.93,
      "learning_rate": 1.5535126234906696e-05,
      "loss": 0.0636,
      "step": 236000
    },
    {
      "epoch": 8.95,
      "learning_rate": 1.552566334834778e-05,
      "loss": 0.0648,
      "step": 236500
    },
    {
      "epoch": 8.97,
      "learning_rate": 1.5516200461788867e-05,
      "loss": 0.0654,
      "step": 237000
    },
    {
      "epoch": 8.99,
      "learning_rate": 1.5506737575229947e-05,
      "loss": 0.0662,
      "step": 237500
    },
    {
      "epoch": 9.0,
      "eval_bleu": 92.1314,
      "eval_gen_len": 16.3862,
      "eval_loss": 0.09200727194547653,
      "eval_runtime": 1201.1498,
      "eval_samples_per_second": 38.019,
      "eval_steps_per_second": 1.189,
      "step": 237771
    },
    {
      "epoch": 9.01,
      "learning_rate": 1.5497274688671034e-05,
      "loss": 0.0541,
      "step": 238000
    },
    {
      "epoch": 9.03,
      "learning_rate": 1.5487811802112118e-05,
      "loss": 0.0553,
      "step": 238500
    },
    {
      "epoch": 9.05,
      "learning_rate": 1.54783489155532e-05,
      "loss": 0.0573,
      "step": 239000
    },
    {
      "epoch": 9.07,
      "learning_rate": 1.5468886028994285e-05,
      "loss": 0.0581,
      "step": 239500
    },
    {
      "epoch": 9.08,
      "learning_rate": 1.545942314243537e-05,
      "loss": 0.0559,
      "step": 240000
    },
    {
      "epoch": 9.1,
      "learning_rate": 1.5449979181649572e-05,
      "loss": 0.0567,
      "step": 240500
    },
    {
      "epoch": 9.12,
      "learning_rate": 1.5440516295090656e-05,
      "loss": 0.0572,
      "step": 241000
    },
    {
      "epoch": 9.14,
      "learning_rate": 1.543105340853174e-05,
      "loss": 0.0561,
      "step": 241500
    },
    {
      "epoch": 9.16,
      "learning_rate": 1.5421590521972823e-05,
      "loss": 0.0564,
      "step": 242000
    },
    {
      "epoch": 9.18,
      "learning_rate": 1.5412127635413907e-05,
      "loss": 0.0562,
      "step": 242500
    },
    {
      "epoch": 9.2,
      "learning_rate": 1.540266474885499e-05,
      "loss": 0.0568,
      "step": 243000
    },
    {
      "epoch": 9.22,
      "learning_rate": 1.5393201862296077e-05,
      "loss": 0.0591,
      "step": 243500
    },
    {
      "epoch": 9.24,
      "learning_rate": 1.538375790151028e-05,
      "loss": 0.0578,
      "step": 244000
    },
    {
      "epoch": 9.25,
      "learning_rate": 1.537429501495136e-05,
      "loss": 0.0567,
      "step": 244500
    },
    {
      "epoch": 9.27,
      "learning_rate": 1.5364832128392448e-05,
      "loss": 0.058,
      "step": 245000
    },
    {
      "epoch": 9.29,
      "learning_rate": 1.535536924183353e-05,
      "loss": 0.0561,
      "step": 245500
    },
    {
      "epoch": 9.31,
      "learning_rate": 1.534592528104773e-05,
      "loss": 0.0589,
      "step": 246000
    },
    {
      "epoch": 9.33,
      "learning_rate": 1.5336462394488814e-05,
      "loss": 0.057,
      "step": 246500
    },
    {
      "epoch": 9.35,
      "learning_rate": 1.53269995079299e-05,
      "loss": 0.0577,
      "step": 247000
    },
    {
      "epoch": 9.37,
      "learning_rate": 1.5317536621370985e-05,
      "loss": 0.0557,
      "step": 247500
    },
    {
      "epoch": 9.39,
      "learning_rate": 1.530807373481207e-05,
      "loss": 0.0586,
      "step": 248000
    },
    {
      "epoch": 9.41,
      "learning_rate": 1.5298610848253153e-05,
      "loss": 0.059,
      "step": 248500
    },
    {
      "epoch": 9.43,
      "learning_rate": 1.5289166887467355e-05,
      "loss": 0.0598,
      "step": 249000
    },
    {
      "epoch": 9.44,
      "learning_rate": 1.5279722926681558e-05,
      "loss": 0.0585,
      "step": 249500
    },
    {
      "epoch": 9.46,
      "learning_rate": 1.527026004012264e-05,
      "loss": 0.0589,
      "step": 250000
    },
    {
      "epoch": 9.48,
      "learning_rate": 1.5260797153563726e-05,
      "loss": 0.0579,
      "step": 250500
    },
    {
      "epoch": 9.5,
      "learning_rate": 1.525133426700481e-05,
      "loss": 0.0577,
      "step": 251000
    },
    {
      "epoch": 9.52,
      "learning_rate": 1.5241871380445891e-05,
      "loss": 0.0577,
      "step": 251500
    },
    {
      "epoch": 9.54,
      "learning_rate": 1.5232408493886977e-05,
      "loss": 0.0587,
      "step": 252000
    },
    {
      "epoch": 9.56,
      "learning_rate": 1.522294560732806e-05,
      "loss": 0.0596,
      "step": 252500
    },
    {
      "epoch": 9.58,
      "learning_rate": 1.5213501646542263e-05,
      "loss": 0.0609,
      "step": 253000
    },
    {
      "epoch": 9.6,
      "learning_rate": 1.5204038759983347e-05,
      "loss": 0.0572,
      "step": 253500
    },
    {
      "epoch": 9.61,
      "learning_rate": 1.519457587342443e-05,
      "loss": 0.0581,
      "step": 254000
    },
    {
      "epoch": 9.63,
      "learning_rate": 1.5185112986865516e-05,
      "loss": 0.0591,
      "step": 254500
    },
    {
      "epoch": 9.65,
      "learning_rate": 1.5175650100306598e-05,
      "loss": 0.0606,
      "step": 255000
    },
    {
      "epoch": 9.67,
      "learning_rate": 1.5166187213747683e-05,
      "loss": 0.0594,
      "step": 255500
    },
    {
      "epoch": 9.69,
      "learning_rate": 1.5156724327188767e-05,
      "loss": 0.0603,
      "step": 256000
    },
    {
      "epoch": 9.71,
      "learning_rate": 1.514726144062985e-05,
      "loss": 0.06,
      "step": 256500
    },
    {
      "epoch": 9.73,
      "learning_rate": 1.5137798554070935e-05,
      "loss": 0.0585,
      "step": 257000
    },
    {
      "epoch": 9.75,
      "learning_rate": 1.512833566751202e-05,
      "loss": 0.058,
      "step": 257500
    },
    {
      "epoch": 9.77,
      "learning_rate": 1.5118891706726223e-05,
      "loss": 0.0595,
      "step": 258000
    },
    {
      "epoch": 9.78,
      "learning_rate": 1.5109428820167305e-05,
      "loss": 0.061,
      "step": 258500
    },
    {
      "epoch": 9.8,
      "learning_rate": 1.5099965933608388e-05,
      "loss": 0.0589,
      "step": 259000
    },
    {
      "epoch": 9.82,
      "learning_rate": 1.5090503047049474e-05,
      "loss": 0.0592,
      "step": 259500
    },
    {
      "epoch": 9.84,
      "learning_rate": 1.5081040160490556e-05,
      "loss": 0.0596,
      "step": 260000
    },
    {
      "epoch": 9.86,
      "learning_rate": 1.5071577273931641e-05,
      "loss": 0.0595,
      "step": 260500
    },
    {
      "epoch": 9.88,
      "learning_rate": 1.5062114387372727e-05,
      "loss": 0.0594,
      "step": 261000
    },
    {
      "epoch": 9.9,
      "learning_rate": 1.5052670426586928e-05,
      "loss": 0.0579,
      "step": 261500
    },
    {
      "epoch": 9.92,
      "learning_rate": 1.5043207540028011e-05,
      "loss": 0.0598,
      "step": 262000
    },
    {
      "epoch": 9.94,
      "learning_rate": 1.5033744653469095e-05,
      "loss": 0.06,
      "step": 262500
    },
    {
      "epoch": 9.95,
      "learning_rate": 1.502428176691018e-05,
      "loss": 0.0574,
      "step": 263000
    },
    {
      "epoch": 9.97,
      "learning_rate": 1.5014818880351263e-05,
      "loss": 0.0582,
      "step": 263500
    },
    {
      "epoch": 9.99,
      "learning_rate": 1.5005374919565465e-05,
      "loss": 0.0588,
      "step": 264000
    },
    {
      "epoch": 10.0,
      "eval_bleu": 92.2075,
      "eval_gen_len": 16.4288,
      "eval_loss": 0.09403391927480698,
      "eval_runtime": 1204.1417,
      "eval_samples_per_second": 37.925,
      "eval_steps_per_second": 1.186,
      "step": 264190
    },
    {
      "epoch": 10.01,
      "learning_rate": 1.499591203300655e-05,
      "loss": 0.0548,
      "step": 264500
    },
    {
      "epoch": 10.03,
      "learning_rate": 1.4986449146447634e-05,
      "loss": 0.0504,
      "step": 265000
    },
    {
      "epoch": 10.05,
      "learning_rate": 1.4976986259888716e-05,
      "loss": 0.0518,
      "step": 265500
    },
    {
      "epoch": 10.07,
      "learning_rate": 1.4967523373329802e-05,
      "loss": 0.052,
      "step": 266000
    },
    {
      "epoch": 10.09,
      "learning_rate": 1.4958060486770887e-05,
      "loss": 0.0529,
      "step": 266500
    },
    {
      "epoch": 10.11,
      "learning_rate": 1.494859760021197e-05,
      "loss": 0.0519,
      "step": 267000
    },
    {
      "epoch": 10.13,
      "learning_rate": 1.493917256519929e-05,
      "loss": 0.0516,
      "step": 267500
    },
    {
      "epoch": 10.14,
      "learning_rate": 1.4929709678640373e-05,
      "loss": 0.0532,
      "step": 268000
    },
    {
      "epoch": 10.16,
      "learning_rate": 1.4920246792081459e-05,
      "loss": 0.0513,
      "step": 268500
    },
    {
      "epoch": 10.18,
      "learning_rate": 1.491078390552254e-05,
      "loss": 0.053,
      "step": 269000
    },
    {
      "epoch": 10.2,
      "learning_rate": 1.4901321018963626e-05,
      "loss": 0.0528,
      "step": 269500
    },
    {
      "epoch": 10.22,
      "learning_rate": 1.4891858132404711e-05,
      "loss": 0.0504,
      "step": 270000
    },
    {
      "epoch": 10.24,
      "learning_rate": 1.488241417161891e-05,
      "loss": 0.0539,
      "step": 270500
    },
    {
      "epoch": 10.26,
      "learning_rate": 1.4872951285059996e-05,
      "loss": 0.0541,
      "step": 271000
    },
    {
      "epoch": 10.28,
      "learning_rate": 1.486348839850108e-05,
      "loss": 0.0516,
      "step": 271500
    },
    {
      "epoch": 10.3,
      "learning_rate": 1.4854025511942165e-05,
      "loss": 0.0532,
      "step": 272000
    },
    {
      "epoch": 10.31,
      "learning_rate": 1.4844562625383247e-05,
      "loss": 0.0511,
      "step": 272500
    },
    {
      "epoch": 10.33,
      "learning_rate": 1.4835099738824333e-05,
      "loss": 0.0525,
      "step": 273000
    },
    {
      "epoch": 10.35,
      "learning_rate": 1.4825636852265416e-05,
      "loss": 0.0539,
      "step": 273500
    },
    {
      "epoch": 10.37,
      "learning_rate": 1.48161739657065e-05,
      "loss": 0.0522,
      "step": 274000
    },
    {
      "epoch": 10.39,
      "learning_rate": 1.4806711079147584e-05,
      "loss": 0.0536,
      "step": 274500
    },
    {
      "epoch": 10.41,
      "learning_rate": 1.4797248192588669e-05,
      "loss": 0.0536,
      "step": 275000
    },
    {
      "epoch": 10.43,
      "learning_rate": 1.4787785306029751e-05,
      "loss": 0.0527,
      "step": 275500
    },
    {
      "epoch": 10.45,
      "learning_rate": 1.4778322419470837e-05,
      "loss": 0.053,
      "step": 276000
    },
    {
      "epoch": 10.47,
      "learning_rate": 1.4768859532911922e-05,
      "loss": 0.0527,
      "step": 276500
    },
    {
      "epoch": 10.48,
      "learning_rate": 1.4759415572126123e-05,
      "loss": 0.0547,
      "step": 277000
    },
    {
      "epoch": 10.5,
      "learning_rate": 1.4749952685567207e-05,
      "loss": 0.054,
      "step": 277500
    },
    {
      "epoch": 10.52,
      "learning_rate": 1.474048979900829e-05,
      "loss": 0.0541,
      "step": 278000
    },
    {
      "epoch": 10.54,
      "learning_rate": 1.473106476399561e-05,
      "loss": 0.055,
      "step": 278500
    },
    {
      "epoch": 10.56,
      "learning_rate": 1.4721601877436696e-05,
      "loss": 0.0546,
      "step": 279000
    },
    {
      "epoch": 10.58,
      "learning_rate": 1.4712138990877778e-05,
      "loss": 0.0522,
      "step": 279500
    },
    {
      "epoch": 10.6,
      "learning_rate": 1.4702676104318863e-05,
      "loss": 0.0536,
      "step": 280000
    },
    {
      "epoch": 10.62,
      "learning_rate": 1.4693232143533065e-05,
      "loss": 0.0531,
      "step": 280500
    },
    {
      "epoch": 10.64,
      "learning_rate": 1.468376925697415e-05,
      "loss": 0.0557,
      "step": 281000
    },
    {
      "epoch": 10.66,
      "learning_rate": 1.4674306370415232e-05,
      "loss": 0.0557,
      "step": 281500
    },
    {
      "epoch": 10.67,
      "learning_rate": 1.4664843483856317e-05,
      "loss": 0.0535,
      "step": 282000
    },
    {
      "epoch": 10.69,
      "learning_rate": 1.4655380597297401e-05,
      "loss": 0.0535,
      "step": 282500
    },
    {
      "epoch": 10.71,
      "learning_rate": 1.4645917710738485e-05,
      "loss": 0.0532,
      "step": 283000
    },
    {
      "epoch": 10.73,
      "learning_rate": 1.4636473749952686e-05,
      "loss": 0.0541,
      "step": 283500
    },
    {
      "epoch": 10.75,
      "learning_rate": 1.4627010863393771e-05,
      "loss": 0.0542,
      "step": 284000
    },
    {
      "epoch": 10.77,
      "learning_rate": 1.4617547976834853e-05,
      "loss": 0.054,
      "step": 284500
    },
    {
      "epoch": 10.79,
      "learning_rate": 1.4608085090275939e-05,
      "loss": 0.052,
      "step": 285000
    },
    {
      "epoch": 10.81,
      "learning_rate": 1.4598622203717024e-05,
      "loss": 0.0537,
      "step": 285500
    },
    {
      "epoch": 10.83,
      "learning_rate": 1.4589159317158108e-05,
      "loss": 0.0558,
      "step": 286000
    },
    {
      "epoch": 10.84,
      "learning_rate": 1.4579696430599191e-05,
      "loss": 0.0558,
      "step": 286500
    },
    {
      "epoch": 10.86,
      "learning_rate": 1.4570233544040275e-05,
      "loss": 0.0545,
      "step": 287000
    },
    {
      "epoch": 10.88,
      "learning_rate": 1.4560789583254478e-05,
      "loss": 0.0544,
      "step": 287500
    },
    {
      "epoch": 10.9,
      "learning_rate": 1.455134562246868e-05,
      "loss": 0.0539,
      "step": 288000
    },
    {
      "epoch": 10.92,
      "learning_rate": 1.4541882735909763e-05,
      "loss": 0.0548,
      "step": 288500
    },
    {
      "epoch": 10.94,
      "learning_rate": 1.4532419849350846e-05,
      "loss": 0.0524,
      "step": 289000
    },
    {
      "epoch": 10.96,
      "learning_rate": 1.4522956962791932e-05,
      "loss": 0.0546,
      "step": 289500
    },
    {
      "epoch": 10.98,
      "learning_rate": 1.4513494076233014e-05,
      "loss": 0.0552,
      "step": 290000
    },
    {
      "epoch": 11.0,
      "learning_rate": 1.45040311896741e-05,
      "loss": 0.0556,
      "step": 290500
    },
    {
      "epoch": 11.0,
      "eval_bleu": 92.2181,
      "eval_gen_len": 16.4826,
      "eval_loss": 0.09591283649206161,
      "eval_runtime": 1213.601,
      "eval_samples_per_second": 37.629,
      "eval_steps_per_second": 1.177,
      "step": 290609
    },
    {
      "epoch": 11.01,
      "learning_rate": 1.4494568303115185e-05,
      "loss": 0.0494,
      "step": 291000
    },
    {
      "epoch": 11.03,
      "learning_rate": 1.4485105416556267e-05,
      "loss": 0.0457,
      "step": 291500
    },
    {
      "epoch": 11.05,
      "learning_rate": 1.447566145577047e-05,
      "loss": 0.0474,
      "step": 292000
    },
    {
      "epoch": 11.07,
      "learning_rate": 1.4466198569211553e-05,
      "loss": 0.0492,
      "step": 292500
    },
    {
      "epoch": 11.09,
      "learning_rate": 1.4456735682652639e-05,
      "loss": 0.0467,
      "step": 293000
    },
    {
      "epoch": 11.11,
      "learning_rate": 1.444727279609372e-05,
      "loss": 0.0489,
      "step": 293500
    },
    {
      "epoch": 11.13,
      "learning_rate": 1.4437809909534806e-05,
      "loss": 0.0478,
      "step": 294000
    },
    {
      "epoch": 11.15,
      "learning_rate": 1.442834702297589e-05,
      "loss": 0.0472,
      "step": 294500
    },
    {
      "epoch": 11.17,
      "learning_rate": 1.4418903062190092e-05,
      "loss": 0.0475,
      "step": 295000
    },
    {
      "epoch": 11.19,
      "learning_rate": 1.4409459101404294e-05,
      "loss": 0.0484,
      "step": 295500
    },
    {
      "epoch": 11.2,
      "learning_rate": 1.4399996214845377e-05,
      "loss": 0.0487,
      "step": 296000
    },
    {
      "epoch": 11.22,
      "learning_rate": 1.4390533328286463e-05,
      "loss": 0.0472,
      "step": 296500
    },
    {
      "epoch": 11.24,
      "learning_rate": 1.4381070441727545e-05,
      "loss": 0.0473,
      "step": 297000
    },
    {
      "epoch": 11.26,
      "learning_rate": 1.437160755516863e-05,
      "loss": 0.0465,
      "step": 297500
    },
    {
      "epoch": 11.28,
      "learning_rate": 1.4362144668609714e-05,
      "loss": 0.0477,
      "step": 298000
    },
    {
      "epoch": 11.3,
      "learning_rate": 1.4352681782050797e-05,
      "loss": 0.0478,
      "step": 298500
    },
    {
      "epoch": 11.32,
      "learning_rate": 1.4343218895491881e-05,
      "loss": 0.0478,
      "step": 299000
    },
    {
      "epoch": 11.34,
      "learning_rate": 1.4333756008932967e-05,
      "loss": 0.0489,
      "step": 299500
    },
    {
      "epoch": 11.36,
      "learning_rate": 1.4324293122374052e-05,
      "loss": 0.0479,
      "step": 300000
    },
    {
      "epoch": 11.37,
      "learning_rate": 1.4314849161588251e-05,
      "loss": 0.0482,
      "step": 300500
    },
    {
      "epoch": 11.39,
      "learning_rate": 1.4305386275029337e-05,
      "loss": 0.0478,
      "step": 301000
    },
    {
      "epoch": 11.41,
      "learning_rate": 1.429592338847042e-05,
      "loss": 0.0488,
      "step": 301500
    },
    {
      "epoch": 11.43,
      "learning_rate": 1.4286479427684623e-05,
      "loss": 0.0487,
      "step": 302000
    },
    {
      "epoch": 11.45,
      "learning_rate": 1.4277016541125705e-05,
      "loss": 0.0488,
      "step": 302500
    },
    {
      "epoch": 11.47,
      "learning_rate": 1.426755365456679e-05,
      "loss": 0.0496,
      "step": 303000
    },
    {
      "epoch": 11.49,
      "learning_rate": 1.4258090768007874e-05,
      "loss": 0.0492,
      "step": 303500
    },
    {
      "epoch": 11.51,
      "learning_rate": 1.4248627881448958e-05,
      "loss": 0.0494,
      "step": 304000
    },
    {
      "epoch": 11.53,
      "learning_rate": 1.4239164994890042e-05,
      "loss": 0.0499,
      "step": 304500
    },
    {
      "epoch": 11.54,
      "learning_rate": 1.4229702108331127e-05,
      "loss": 0.0492,
      "step": 305000
    },
    {
      "epoch": 11.56,
      "learning_rate": 1.4220239221772209e-05,
      "loss": 0.0485,
      "step": 305500
    },
    {
      "epoch": 11.58,
      "learning_rate": 1.4210776335213295e-05,
      "loss": 0.0492,
      "step": 306000
    },
    {
      "epoch": 11.6,
      "learning_rate": 1.4201332374427497e-05,
      "loss": 0.0494,
      "step": 306500
    },
    {
      "epoch": 11.62,
      "learning_rate": 1.4191869487868581e-05,
      "loss": 0.0501,
      "step": 307000
    },
    {
      "epoch": 11.64,
      "learning_rate": 1.4182406601309665e-05,
      "loss": 0.0499,
      "step": 307500
    },
    {
      "epoch": 11.66,
      "learning_rate": 1.4172943714750748e-05,
      "loss": 0.0507,
      "step": 308000
    },
    {
      "epoch": 11.68,
      "learning_rate": 1.4163480828191834e-05,
      "loss": 0.0493,
      "step": 308500
    },
    {
      "epoch": 11.7,
      "learning_rate": 1.4154017941632916e-05,
      "loss": 0.0499,
      "step": 309000
    },
    {
      "epoch": 11.72,
      "learning_rate": 1.4144555055074001e-05,
      "loss": 0.0506,
      "step": 309500
    },
    {
      "epoch": 11.73,
      "learning_rate": 1.4135092168515085e-05,
      "loss": 0.0506,
      "step": 310000
    },
    {
      "epoch": 11.75,
      "learning_rate": 1.4125648207729288e-05,
      "loss": 0.0492,
      "step": 310500
    },
    {
      "epoch": 11.77,
      "learning_rate": 1.4116204246943487e-05,
      "loss": 0.0485,
      "step": 311000
    },
    {
      "epoch": 11.79,
      "learning_rate": 1.4106741360384573e-05,
      "loss": 0.0509,
      "step": 311500
    },
    {
      "epoch": 11.81,
      "learning_rate": 1.4097297399598775e-05,
      "loss": 0.0497,
      "step": 312000
    },
    {
      "epoch": 11.83,
      "learning_rate": 1.4087834513039859e-05,
      "loss": 0.0492,
      "step": 312500
    },
    {
      "epoch": 11.85,
      "learning_rate": 1.4078371626480943e-05,
      "loss": 0.05,
      "step": 313000
    },
    {
      "epoch": 11.87,
      "learning_rate": 1.4068908739922026e-05,
      "loss": 0.0507,
      "step": 313500
    },
    {
      "epoch": 11.89,
      "learning_rate": 1.4059445853363112e-05,
      "loss": 0.0508,
      "step": 314000
    },
    {
      "epoch": 11.9,
      "learning_rate": 1.4049982966804194e-05,
      "loss": 0.0506,
      "step": 314500
    },
    {
      "epoch": 11.92,
      "learning_rate": 1.404052008024528e-05,
      "loss": 0.0488,
      "step": 315000
    },
    {
      "epoch": 11.94,
      "learning_rate": 1.4031057193686363e-05,
      "loss": 0.0501,
      "step": 315500
    },
    {
      "epoch": 11.96,
      "learning_rate": 1.4021594307127447e-05,
      "loss": 0.048,
      "step": 316000
    },
    {
      "epoch": 11.98,
      "learning_rate": 1.401215034634165e-05,
      "loss": 0.0519,
      "step": 316500
    },
    {
      "epoch": 12.0,
      "learning_rate": 1.4002687459782733e-05,
      "loss": 0.049,
      "step": 317000
    },
    {
      "epoch": 12.0,
      "eval_bleu": 92.2414,
      "eval_gen_len": 16.4298,
      "eval_loss": 0.09784110635519028,
      "eval_runtime": 1192.5959,
      "eval_samples_per_second": 38.292,
      "eval_steps_per_second": 1.197,
      "step": 317028
    },
    {
      "epoch": 12.02,
      "learning_rate": 1.3993224573223819e-05,
      "loss": 0.0424,
      "step": 317500
    },
    {
      "epoch": 12.04,
      "learning_rate": 1.3983780612438018e-05,
      "loss": 0.0414,
      "step": 318000
    },
    {
      "epoch": 12.06,
      "learning_rate": 1.3974317725879103e-05,
      "loss": 0.0424,
      "step": 318500
    },
    {
      "epoch": 12.07,
      "learning_rate": 1.3964854839320187e-05,
      "loss": 0.0422,
      "step": 319000
    },
    {
      "epoch": 12.09,
      "learning_rate": 1.3955391952761272e-05,
      "loss": 0.0433,
      "step": 319500
    },
    {
      "epoch": 12.11,
      "learning_rate": 1.3945929066202354e-05,
      "loss": 0.0429,
      "step": 320000
    },
    {
      "epoch": 12.13,
      "learning_rate": 1.393646617964344e-05,
      "loss": 0.0433,
      "step": 320500
    },
    {
      "epoch": 12.15,
      "learning_rate": 1.3927003293084525e-05,
      "loss": 0.044,
      "step": 321000
    },
    {
      "epoch": 12.17,
      "learning_rate": 1.3917540406525607e-05,
      "loss": 0.0443,
      "step": 321500
    },
    {
      "epoch": 12.19,
      "learning_rate": 1.3908077519966691e-05,
      "loss": 0.0436,
      "step": 322000
    },
    {
      "epoch": 12.21,
      "learning_rate": 1.3898614633407776e-05,
      "loss": 0.0436,
      "step": 322500
    },
    {
      "epoch": 12.23,
      "learning_rate": 1.3889151746848858e-05,
      "loss": 0.0435,
      "step": 323000
    },
    {
      "epoch": 12.24,
      "learning_rate": 1.3879688860289944e-05,
      "loss": 0.0437,
      "step": 323500
    },
    {
      "epoch": 12.26,
      "learning_rate": 1.3870244899504147e-05,
      "loss": 0.0461,
      "step": 324000
    },
    {
      "epoch": 12.28,
      "learning_rate": 1.386078201294523e-05,
      "loss": 0.0442,
      "step": 324500
    },
    {
      "epoch": 12.3,
      "learning_rate": 1.3851319126386314e-05,
      "loss": 0.0447,
      "step": 325000
    },
    {
      "epoch": 12.32,
      "learning_rate": 1.3841856239827398e-05,
      "loss": 0.0457,
      "step": 325500
    },
    {
      "epoch": 12.34,
      "learning_rate": 1.38324122790416e-05,
      "loss": 0.0436,
      "step": 326000
    },
    {
      "epoch": 12.36,
      "learning_rate": 1.3822949392482682e-05,
      "loss": 0.0445,
      "step": 326500
    },
    {
      "epoch": 12.38,
      "learning_rate": 1.3813486505923768e-05,
      "loss": 0.0456,
      "step": 327000
    },
    {
      "epoch": 12.4,
      "learning_rate": 1.3804023619364853e-05,
      "loss": 0.0452,
      "step": 327500
    },
    {
      "epoch": 12.42,
      "learning_rate": 1.3794560732805937e-05,
      "loss": 0.0456,
      "step": 328000
    },
    {
      "epoch": 12.43,
      "learning_rate": 1.3785116772020138e-05,
      "loss": 0.0446,
      "step": 328500
    },
    {
      "epoch": 12.45,
      "learning_rate": 1.3775653885461222e-05,
      "loss": 0.0451,
      "step": 329000
    },
    {
      "epoch": 12.47,
      "learning_rate": 1.3766190998902307e-05,
      "loss": 0.0457,
      "step": 329500
    },
    {
      "epoch": 12.49,
      "learning_rate": 1.3756728112343389e-05,
      "loss": 0.0448,
      "step": 330000
    },
    {
      "epoch": 12.51,
      "learning_rate": 1.3747265225784475e-05,
      "loss": 0.0446,
      "step": 330500
    },
    {
      "epoch": 12.53,
      "learning_rate": 1.3737840190771795e-05,
      "loss": 0.0455,
      "step": 331000
    },
    {
      "epoch": 12.55,
      "learning_rate": 1.3728377304212878e-05,
      "loss": 0.0449,
      "step": 331500
    },
    {
      "epoch": 12.57,
      "learning_rate": 1.371891441765396e-05,
      "loss": 0.044,
      "step": 332000
    },
    {
      "epoch": 12.59,
      "learning_rate": 1.3709451531095046e-05,
      "loss": 0.0445,
      "step": 332500
    },
    {
      "epoch": 12.6,
      "learning_rate": 1.3699988644536131e-05,
      "loss": 0.0454,
      "step": 333000
    },
    {
      "epoch": 12.62,
      "learning_rate": 1.3690525757977215e-05,
      "loss": 0.0455,
      "step": 333500
    },
    {
      "epoch": 12.64,
      "learning_rate": 1.3681062871418299e-05,
      "loss": 0.0453,
      "step": 334000
    },
    {
      "epoch": 12.66,
      "learning_rate": 1.36716189106325e-05,
      "loss": 0.0456,
      "step": 334500
    },
    {
      "epoch": 12.68,
      "learning_rate": 1.3662156024073585e-05,
      "loss": 0.0447,
      "step": 335000
    },
    {
      "epoch": 12.7,
      "learning_rate": 1.3652693137514667e-05,
      "loss": 0.0442,
      "step": 335500
    },
    {
      "epoch": 12.72,
      "learning_rate": 1.3643230250955753e-05,
      "loss": 0.0455,
      "step": 336000
    },
    {
      "epoch": 12.74,
      "learning_rate": 1.3633786290169955e-05,
      "loss": 0.0449,
      "step": 336500
    },
    {
      "epoch": 12.76,
      "learning_rate": 1.3624323403611039e-05,
      "loss": 0.0446,
      "step": 337000
    },
    {
      "epoch": 12.77,
      "learning_rate": 1.3614860517052123e-05,
      "loss": 0.0451,
      "step": 337500
    },
    {
      "epoch": 12.79,
      "learning_rate": 1.3605397630493206e-05,
      "loss": 0.0448,
      "step": 338000
    },
    {
      "epoch": 12.81,
      "learning_rate": 1.3595934743934292e-05,
      "loss": 0.0454,
      "step": 338500
    },
    {
      "epoch": 12.83,
      "learning_rate": 1.3586490783148495e-05,
      "loss": 0.0464,
      "step": 339000
    },
    {
      "epoch": 12.85,
      "learning_rate": 1.3577027896589577e-05,
      "loss": 0.0458,
      "step": 339500
    },
    {
      "epoch": 12.87,
      "learning_rate": 1.356756501003066e-05,
      "loss": 0.0449,
      "step": 340000
    },
    {
      "epoch": 12.89,
      "learning_rate": 1.3558102123471746e-05,
      "loss": 0.0459,
      "step": 340500
    },
    {
      "epoch": 12.91,
      "learning_rate": 1.3548639236912828e-05,
      "loss": 0.0443,
      "step": 341000
    },
    {
      "epoch": 12.93,
      "learning_rate": 1.3539176350353913e-05,
      "loss": 0.0458,
      "step": 341500
    },
    {
      "epoch": 12.95,
      "learning_rate": 1.3529713463794999e-05,
      "loss": 0.0463,
      "step": 342000
    },
    {
      "epoch": 12.96,
      "learning_rate": 1.352025057723608e-05,
      "loss": 0.0468,
      "step": 342500
    },
    {
      "epoch": 12.98,
      "learning_rate": 1.3510806616450283e-05,
      "loss": 0.045,
      "step": 343000
    },
    {
      "epoch": 13.0,
      "eval_bleu": 92.2161,
      "eval_gen_len": 16.443,
      "eval_loss": 0.09937400370836258,
      "eval_runtime": 1185.3348,
      "eval_samples_per_second": 38.527,
      "eval_steps_per_second": 1.205,
      "step": 343447
    },
    {
      "epoch": 13.0,
      "learning_rate": 1.3501343729891367e-05,
      "loss": 0.0445,
      "step": 343500
    },
    {
      "epoch": 13.02,
      "learning_rate": 1.3491880843332452e-05,
      "loss": 0.0381,
      "step": 344000
    },
    {
      "epoch": 13.04,
      "learning_rate": 1.3482417956773534e-05,
      "loss": 0.0397,
      "step": 344500
    },
    {
      "epoch": 13.06,
      "learning_rate": 1.3472973995987737e-05,
      "loss": 0.0385,
      "step": 345000
    },
    {
      "epoch": 13.08,
      "learning_rate": 1.3463511109428823e-05,
      "loss": 0.0393,
      "step": 345500
    },
    {
      "epoch": 13.1,
      "learning_rate": 1.3454067148643024e-05,
      "loss": 0.0392,
      "step": 346000
    },
    {
      "epoch": 13.12,
      "learning_rate": 1.3444604262084107e-05,
      "loss": 0.0385,
      "step": 346500
    },
    {
      "epoch": 13.13,
      "learning_rate": 1.3435141375525191e-05,
      "loss": 0.0403,
      "step": 347000
    },
    {
      "epoch": 13.15,
      "learning_rate": 1.3425697414739394e-05,
      "loss": 0.039,
      "step": 347500
    },
    {
      "epoch": 13.17,
      "learning_rate": 1.3416234528180476e-05,
      "loss": 0.0402,
      "step": 348000
    },
    {
      "epoch": 13.19,
      "learning_rate": 1.3406771641621561e-05,
      "loss": 0.0397,
      "step": 348500
    },
    {
      "epoch": 13.21,
      "learning_rate": 1.3397308755062645e-05,
      "loss": 0.0405,
      "step": 349000
    },
    {
      "epoch": 13.23,
      "learning_rate": 1.338784586850373e-05,
      "loss": 0.0407,
      "step": 349500
    },
    {
      "epoch": 13.25,
      "learning_rate": 1.3378382981944812e-05,
      "loss": 0.04,
      "step": 350000
    },
    {
      "epoch": 13.27,
      "learning_rate": 1.3368920095385898e-05,
      "loss": 0.0395,
      "step": 350500
    },
    {
      "epoch": 13.29,
      "learning_rate": 1.3359457208826983e-05,
      "loss": 0.0393,
      "step": 351000
    },
    {
      "epoch": 13.3,
      "learning_rate": 1.3349994322268065e-05,
      "loss": 0.0404,
      "step": 351500
    },
    {
      "epoch": 13.32,
      "learning_rate": 1.334053143570915e-05,
      "loss": 0.0401,
      "step": 352000
    },
    {
      "epoch": 13.34,
      "learning_rate": 1.3331106400696469e-05,
      "loss": 0.0404,
      "step": 352500
    },
    {
      "epoch": 13.36,
      "learning_rate": 1.3321643514137554e-05,
      "loss": 0.041,
      "step": 353000
    },
    {
      "epoch": 13.38,
      "learning_rate": 1.3312180627578637e-05,
      "loss": 0.0418,
      "step": 353500
    },
    {
      "epoch": 13.4,
      "learning_rate": 1.3302717741019722e-05,
      "loss": 0.0402,
      "step": 354000
    },
    {
      "epoch": 13.42,
      "learning_rate": 1.3293254854460806e-05,
      "loss": 0.0405,
      "step": 354500
    },
    {
      "epoch": 13.44,
      "learning_rate": 1.328379196790189e-05,
      "loss": 0.0415,
      "step": 355000
    },
    {
      "epoch": 13.46,
      "learning_rate": 1.3274348007116092e-05,
      "loss": 0.0415,
      "step": 355500
    },
    {
      "epoch": 13.48,
      "learning_rate": 1.3264885120557176e-05,
      "loss": 0.0406,
      "step": 356000
    },
    {
      "epoch": 13.49,
      "learning_rate": 1.3255422233998261e-05,
      "loss": 0.0409,
      "step": 356500
    },
    {
      "epoch": 13.51,
      "learning_rate": 1.3245959347439343e-05,
      "loss": 0.0404,
      "step": 357000
    },
    {
      "epoch": 13.53,
      "learning_rate": 1.3236496460880429e-05,
      "loss": 0.0405,
      "step": 357500
    },
    {
      "epoch": 13.55,
      "learning_rate": 1.3227033574321512e-05,
      "loss": 0.0419,
      "step": 358000
    },
    {
      "epoch": 13.57,
      "learning_rate": 1.3217570687762596e-05,
      "loss": 0.0396,
      "step": 358500
    },
    {
      "epoch": 13.59,
      "learning_rate": 1.320810780120368e-05,
      "loss": 0.0409,
      "step": 359000
    },
    {
      "epoch": 13.61,
      "learning_rate": 1.3198644914644765e-05,
      "loss": 0.0419,
      "step": 359500
    },
    {
      "epoch": 13.63,
      "learning_rate": 1.3189182028085847e-05,
      "loss": 0.0407,
      "step": 360000
    },
    {
      "epoch": 13.65,
      "learning_rate": 1.3179719141526932e-05,
      "loss": 0.0413,
      "step": 360500
    },
    {
      "epoch": 13.66,
      "learning_rate": 1.3170256254968016e-05,
      "loss": 0.041,
      "step": 361000
    },
    {
      "epoch": 13.68,
      "learning_rate": 1.3160812294182219e-05,
      "loss": 0.0416,
      "step": 361500
    },
    {
      "epoch": 13.7,
      "learning_rate": 1.3151349407623301e-05,
      "loss": 0.0415,
      "step": 362000
    },
    {
      "epoch": 13.72,
      "learning_rate": 1.3141905446837504e-05,
      "loss": 0.0426,
      "step": 362500
    },
    {
      "epoch": 13.74,
      "learning_rate": 1.313244256027859e-05,
      "loss": 0.0418,
      "step": 363000
    },
    {
      "epoch": 13.76,
      "learning_rate": 1.3122979673719673e-05,
      "loss": 0.0414,
      "step": 363500
    },
    {
      "epoch": 13.78,
      "learning_rate": 1.3113535712933874e-05,
      "loss": 0.0431,
      "step": 364000
    },
    {
      "epoch": 13.8,
      "learning_rate": 1.3104072826374958e-05,
      "loss": 0.0402,
      "step": 364500
    },
    {
      "epoch": 13.82,
      "learning_rate": 1.3094609939816043e-05,
      "loss": 0.0417,
      "step": 365000
    },
    {
      "epoch": 13.83,
      "learning_rate": 1.3085147053257125e-05,
      "loss": 0.0405,
      "step": 365500
    },
    {
      "epoch": 13.85,
      "learning_rate": 1.307568416669821e-05,
      "loss": 0.0416,
      "step": 366000
    },
    {
      "epoch": 13.87,
      "learning_rate": 1.3066221280139296e-05,
      "loss": 0.0429,
      "step": 366500
    },
    {
      "epoch": 13.89,
      "learning_rate": 1.305675839358038e-05,
      "loss": 0.0408,
      "step": 367000
    },
    {
      "epoch": 13.91,
      "learning_rate": 1.3047295507021462e-05,
      "loss": 0.0425,
      "step": 367500
    },
    {
      "epoch": 13.93,
      "learning_rate": 1.3037832620462547e-05,
      "loss": 0.0414,
      "step": 368000
    },
    {
      "epoch": 13.95,
      "learning_rate": 1.3028369733903632e-05,
      "loss": 0.0427,
      "step": 368500
    },
    {
      "epoch": 13.97,
      "learning_rate": 1.3018906847344714e-05,
      "loss": 0.0413,
      "step": 369000
    },
    {
      "epoch": 13.99,
      "learning_rate": 1.30094439607858e-05,
      "loss": 0.0418,
      "step": 369500
    },
    {
      "epoch": 14.0,
      "eval_bleu": 92.2123,
      "eval_gen_len": 16.4615,
      "eval_loss": 0.10214965045452118,
      "eval_runtime": 1189.3745,
      "eval_samples_per_second": 38.396,
      "eval_steps_per_second": 1.201,
      "step": 369866
    },
    {
      "epoch": 14.01,
      "learning_rate": 1.2999981074226883e-05,
      "loss": 0.0396,
      "step": 370000
    },
    {
      "epoch": 14.02,
      "learning_rate": 1.2990537113441085e-05,
      "loss": 0.0336,
      "step": 370500
    },
    {
      "epoch": 14.04,
      "learning_rate": 1.2981074226882168e-05,
      "loss": 0.0348,
      "step": 371000
    },
    {
      "epoch": 14.06,
      "learning_rate": 1.2971630266096371e-05,
      "loss": 0.0358,
      "step": 371500
    },
    {
      "epoch": 14.08,
      "learning_rate": 1.2962167379537456e-05,
      "loss": 0.0359,
      "step": 372000
    },
    {
      "epoch": 14.1,
      "learning_rate": 1.2952704492978538e-05,
      "loss": 0.0358,
      "step": 372500
    },
    {
      "epoch": 14.12,
      "learning_rate": 1.2943241606419624e-05,
      "loss": 0.0351,
      "step": 373000
    },
    {
      "epoch": 14.14,
      "learning_rate": 1.2933778719860708e-05,
      "loss": 0.0362,
      "step": 373500
    },
    {
      "epoch": 14.16,
      "learning_rate": 1.292431583330179e-05,
      "loss": 0.0356,
      "step": 374000
    },
    {
      "epoch": 14.18,
      "learning_rate": 1.2914871872515992e-05,
      "loss": 0.0359,
      "step": 374500
    },
    {
      "epoch": 14.19,
      "learning_rate": 1.2905408985957078e-05,
      "loss": 0.0366,
      "step": 375000
    },
    {
      "epoch": 14.21,
      "learning_rate": 1.2895946099398161e-05,
      "loss": 0.035,
      "step": 375500
    },
    {
      "epoch": 14.23,
      "learning_rate": 1.2886483212839245e-05,
      "loss": 0.0363,
      "step": 376000
    },
    {
      "epoch": 14.25,
      "learning_rate": 1.2877020326280329e-05,
      "loss": 0.0374,
      "step": 376500
    },
    {
      "epoch": 14.27,
      "learning_rate": 1.2867576365494532e-05,
      "loss": 0.0369,
      "step": 377000
    },
    {
      "epoch": 14.29,
      "learning_rate": 1.2858113478935617e-05,
      "loss": 0.0353,
      "step": 377500
    },
    {
      "epoch": 14.31,
      "learning_rate": 1.2848650592376699e-05,
      "loss": 0.0366,
      "step": 378000
    },
    {
      "epoch": 14.33,
      "learning_rate": 1.2839187705817784e-05,
      "loss": 0.0355,
      "step": 378500
    },
    {
      "epoch": 14.35,
      "learning_rate": 1.2829724819258868e-05,
      "loss": 0.0365,
      "step": 379000
    },
    {
      "epoch": 14.36,
      "learning_rate": 1.2820261932699952e-05,
      "loss": 0.0371,
      "step": 379500
    },
    {
      "epoch": 14.38,
      "learning_rate": 1.2810799046141036e-05,
      "loss": 0.0371,
      "step": 380000
    },
    {
      "epoch": 14.4,
      "learning_rate": 1.2801336159582121e-05,
      "loss": 0.0366,
      "step": 380500
    },
    {
      "epoch": 14.42,
      "learning_rate": 1.2791873273023203e-05,
      "loss": 0.0371,
      "step": 381000
    },
    {
      "epoch": 14.44,
      "learning_rate": 1.2782429312237406e-05,
      "loss": 0.0371,
      "step": 381500
    },
    {
      "epoch": 14.46,
      "learning_rate": 1.277296642567849e-05,
      "loss": 0.0362,
      "step": 382000
    },
    {
      "epoch": 14.48,
      "learning_rate": 1.2763503539119575e-05,
      "loss": 0.0365,
      "step": 382500
    },
    {
      "epoch": 14.5,
      "learning_rate": 1.2754059578333774e-05,
      "loss": 0.037,
      "step": 383000
    },
    {
      "epoch": 14.52,
      "learning_rate": 1.274459669177486e-05,
      "loss": 0.0375,
      "step": 383500
    },
    {
      "epoch": 14.53,
      "learning_rate": 1.2735133805215945e-05,
      "loss": 0.0372,
      "step": 384000
    },
    {
      "epoch": 14.55,
      "learning_rate": 1.2725670918657027e-05,
      "loss": 0.0373,
      "step": 384500
    },
    {
      "epoch": 14.57,
      "learning_rate": 1.2716208032098112e-05,
      "loss": 0.0363,
      "step": 385000
    },
    {
      "epoch": 14.59,
      "learning_rate": 1.2706745145539196e-05,
      "loss": 0.0374,
      "step": 385500
    },
    {
      "epoch": 14.61,
      "learning_rate": 1.2697282258980282e-05,
      "loss": 0.0373,
      "step": 386000
    },
    {
      "epoch": 14.63,
      "learning_rate": 1.2687838298194481e-05,
      "loss": 0.0381,
      "step": 386500
    },
    {
      "epoch": 14.65,
      "learning_rate": 1.2678375411635566e-05,
      "loss": 0.0383,
      "step": 387000
    },
    {
      "epoch": 14.67,
      "learning_rate": 1.266891252507665e-05,
      "loss": 0.0369,
      "step": 387500
    },
    {
      "epoch": 14.69,
      "learning_rate": 1.2659449638517734e-05,
      "loss": 0.0385,
      "step": 388000
    },
    {
      "epoch": 14.71,
      "learning_rate": 1.2649986751958817e-05,
      "loss": 0.0384,
      "step": 388500
    },
    {
      "epoch": 14.72,
      "learning_rate": 1.2640523865399903e-05,
      "loss": 0.0391,
      "step": 389000
    },
    {
      "epoch": 14.74,
      "learning_rate": 1.2631060978840988e-05,
      "loss": 0.0379,
      "step": 389500
    },
    {
      "epoch": 14.76,
      "learning_rate": 1.262159809228207e-05,
      "loss": 0.0376,
      "step": 390000
    },
    {
      "epoch": 14.78,
      "learning_rate": 1.2612135205723156e-05,
      "loss": 0.0383,
      "step": 390500
    },
    {
      "epoch": 14.8,
      "learning_rate": 1.260267231916424e-05,
      "loss": 0.0378,
      "step": 391000
    },
    {
      "epoch": 14.82,
      "learning_rate": 1.2593209432605323e-05,
      "loss": 0.0372,
      "step": 391500
    },
    {
      "epoch": 14.84,
      "learning_rate": 1.2583746546046407e-05,
      "loss": 0.0375,
      "step": 392000
    },
    {
      "epoch": 14.86,
      "learning_rate": 1.2574283659487492e-05,
      "loss": 0.0375,
      "step": 392500
    },
    {
      "epoch": 14.88,
      "learning_rate": 1.2564839698701693e-05,
      "loss": 0.0379,
      "step": 393000
    },
    {
      "epoch": 14.89,
      "learning_rate": 1.2555395737915894e-05,
      "loss": 0.0393,
      "step": 393500
    },
    {
      "epoch": 14.91,
      "learning_rate": 1.2545932851356978e-05,
      "loss": 0.0381,
      "step": 394000
    },
    {
      "epoch": 14.93,
      "learning_rate": 1.2536488890571181e-05,
      "loss": 0.0382,
      "step": 394500
    },
    {
      "epoch": 14.95,
      "learning_rate": 1.2527026004012266e-05,
      "loss": 0.0392,
      "step": 395000
    },
    {
      "epoch": 14.97,
      "learning_rate": 1.2517563117453348e-05,
      "loss": 0.0375,
      "step": 395500
    },
    {
      "epoch": 14.99,
      "learning_rate": 1.2508100230894434e-05,
      "loss": 0.0389,
      "step": 396000
    },
    {
      "epoch": 15.0,
      "eval_bleu": 92.1099,
      "eval_gen_len": 16.4498,
      "eval_loss": 0.10496072471141815,
      "eval_runtime": 1187.4741,
      "eval_samples_per_second": 38.457,
      "eval_steps_per_second": 1.203,
      "step": 396285
    },
    {
      "epoch": 15.01,
      "learning_rate": 1.2498637344335517e-05,
      "loss": 0.0357,
      "step": 396500
    },
    {
      "epoch": 15.03,
      "learning_rate": 1.2489174457776601e-05,
      "loss": 0.0318,
      "step": 397000
    },
    {
      "epoch": 15.05,
      "learning_rate": 1.2479711571217685e-05,
      "loss": 0.0329,
      "step": 397500
    },
    {
      "epoch": 15.06,
      "learning_rate": 1.247024868465877e-05,
      "loss": 0.0329,
      "step": 398000
    },
    {
      "epoch": 15.08,
      "learning_rate": 1.2460785798099852e-05,
      "loss": 0.0331,
      "step": 398500
    },
    {
      "epoch": 15.1,
      "learning_rate": 1.2451322911540938e-05,
      "loss": 0.0338,
      "step": 399000
    },
    {
      "epoch": 15.12,
      "learning_rate": 1.2441860024982021e-05,
      "loss": 0.0319,
      "step": 399500
    },
    {
      "epoch": 15.14,
      "learning_rate": 1.2432397138423105e-05,
      "loss": 0.0337,
      "step": 400000
    },
    {
      "epoch": 15.16,
      "learning_rate": 1.2422934251864189e-05,
      "loss": 0.0329,
      "step": 400500
    },
    {
      "epoch": 15.18,
      "learning_rate": 1.2413490291078391e-05,
      "loss": 0.0326,
      "step": 401000
    },
    {
      "epoch": 15.2,
      "learning_rate": 1.2404046330292594e-05,
      "loss": 0.0325,
      "step": 401500
    },
    {
      "epoch": 15.22,
      "learning_rate": 1.2394583443733676e-05,
      "loss": 0.0332,
      "step": 402000
    },
    {
      "epoch": 15.24,
      "learning_rate": 1.2385120557174762e-05,
      "loss": 0.0343,
      "step": 402500
    },
    {
      "epoch": 15.25,
      "learning_rate": 1.2375657670615845e-05,
      "loss": 0.0324,
      "step": 403000
    },
    {
      "epoch": 15.27,
      "learning_rate": 1.236619478405693e-05,
      "loss": 0.0331,
      "step": 403500
    },
    {
      "epoch": 15.29,
      "learning_rate": 1.2356731897498013e-05,
      "loss": 0.0318,
      "step": 404000
    },
    {
      "epoch": 15.31,
      "learning_rate": 1.2347269010939098e-05,
      "loss": 0.0317,
      "step": 404500
    },
    {
      "epoch": 15.33,
      "learning_rate": 1.2337806124380184e-05,
      "loss": 0.0333,
      "step": 405000
    },
    {
      "epoch": 15.35,
      "learning_rate": 1.2328343237821266e-05,
      "loss": 0.0346,
      "step": 405500
    },
    {
      "epoch": 15.37,
      "learning_rate": 1.231888035126235e-05,
      "loss": 0.0334,
      "step": 406000
    },
    {
      "epoch": 15.39,
      "learning_rate": 1.2309436390476552e-05,
      "loss": 0.0331,
      "step": 406500
    },
    {
      "epoch": 15.41,
      "learning_rate": 1.2299973503917637e-05,
      "loss": 0.0342,
      "step": 407000
    },
    {
      "epoch": 15.42,
      "learning_rate": 1.229051061735872e-05,
      "loss": 0.0338,
      "step": 407500
    },
    {
      "epoch": 15.44,
      "learning_rate": 1.2281047730799805e-05,
      "loss": 0.0331,
      "step": 408000
    },
    {
      "epoch": 15.46,
      "learning_rate": 1.2271603770014006e-05,
      "loss": 0.0343,
      "step": 408500
    },
    {
      "epoch": 15.48,
      "learning_rate": 1.226214088345509e-05,
      "loss": 0.0345,
      "step": 409000
    },
    {
      "epoch": 15.5,
      "learning_rate": 1.2252677996896173e-05,
      "loss": 0.0337,
      "step": 409500
    },
    {
      "epoch": 15.52,
      "learning_rate": 1.2243215110337259e-05,
      "loss": 0.0343,
      "step": 410000
    },
    {
      "epoch": 15.54,
      "learning_rate": 1.223375222377834e-05,
      "loss": 0.0335,
      "step": 410500
    },
    {
      "epoch": 15.56,
      "learning_rate": 1.2224289337219426e-05,
      "loss": 0.0335,
      "step": 411000
    },
    {
      "epoch": 15.58,
      "learning_rate": 1.2214845376433629e-05,
      "loss": 0.0345,
      "step": 411500
    },
    {
      "epoch": 15.59,
      "learning_rate": 1.220540141564783e-05,
      "loss": 0.0337,
      "step": 412000
    },
    {
      "epoch": 15.61,
      "learning_rate": 1.2195938529088914e-05,
      "loss": 0.0339,
      "step": 412500
    },
    {
      "epoch": 15.63,
      "learning_rate": 1.2186475642529997e-05,
      "loss": 0.0359,
      "step": 413000
    },
    {
      "epoch": 15.65,
      "learning_rate": 1.2177012755971083e-05,
      "loss": 0.0329,
      "step": 413500
    },
    {
      "epoch": 15.67,
      "learning_rate": 1.2167549869412167e-05,
      "loss": 0.0349,
      "step": 414000
    },
    {
      "epoch": 15.69,
      "learning_rate": 1.215808698285325e-05,
      "loss": 0.0341,
      "step": 414500
    },
    {
      "epoch": 15.71,
      "learning_rate": 1.2148624096294334e-05,
      "loss": 0.0343,
      "step": 415000
    },
    {
      "epoch": 15.73,
      "learning_rate": 1.213916120973542e-05,
      "loss": 0.0343,
      "step": 415500
    },
    {
      "epoch": 15.75,
      "learning_rate": 1.2129698323176501e-05,
      "loss": 0.0337,
      "step": 416000
    },
    {
      "epoch": 15.77,
      "learning_rate": 1.2120235436617587e-05,
      "loss": 0.0344,
      "step": 416500
    },
    {
      "epoch": 15.78,
      "learning_rate": 1.2110772550058672e-05,
      "loss": 0.0341,
      "step": 417000
    },
    {
      "epoch": 15.8,
      "learning_rate": 1.2101309663499754e-05,
      "loss": 0.0341,
      "step": 417500
    },
    {
      "epoch": 15.82,
      "learning_rate": 1.2091865702713957e-05,
      "loss": 0.0341,
      "step": 418000
    },
    {
      "epoch": 15.84,
      "learning_rate": 1.208240281615504e-05,
      "loss": 0.0351,
      "step": 418500
    },
    {
      "epoch": 15.86,
      "learning_rate": 1.2072939929596126e-05,
      "loss": 0.0339,
      "step": 419000
    },
    {
      "epoch": 15.88,
      "learning_rate": 1.2063477043037208e-05,
      "loss": 0.035,
      "step": 419500
    },
    {
      "epoch": 15.9,
      "learning_rate": 1.2054014156478293e-05,
      "loss": 0.0343,
      "step": 420000
    },
    {
      "epoch": 15.92,
      "learning_rate": 1.2044551269919377e-05,
      "loss": 0.034,
      "step": 420500
    },
    {
      "epoch": 15.94,
      "learning_rate": 1.203510730913358e-05,
      "loss": 0.0346,
      "step": 421000
    },
    {
      "epoch": 15.95,
      "learning_rate": 1.2025644422574662e-05,
      "loss": 0.0352,
      "step": 421500
    },
    {
      "epoch": 15.97,
      "learning_rate": 1.2016200461788865e-05,
      "loss": 0.0341,
      "step": 422000
    },
    {
      "epoch": 15.99,
      "learning_rate": 1.200673757522995e-05,
      "loss": 0.0337,
      "step": 422500
    },
    {
      "epoch": 16.0,
      "eval_bleu": 92.2303,
      "eval_gen_len": 16.4514,
      "eval_loss": 0.10698259621858597,
      "eval_runtime": 1187.5586,
      "eval_samples_per_second": 38.455,
      "eval_steps_per_second": 1.202,
      "step": 422704
    },
    {
      "epoch": 16.01,
      "learning_rate": 1.1997274688671032e-05,
      "loss": 0.0312,
      "step": 423000
    },
    {
      "epoch": 16.03,
      "learning_rate": 1.1987811802112118e-05,
      "loss": 0.0283,
      "step": 423500
    },
    {
      "epoch": 16.05,
      "learning_rate": 1.1978348915553201e-05,
      "loss": 0.0278,
      "step": 424000
    },
    {
      "epoch": 16.07,
      "learning_rate": 1.1968886028994285e-05,
      "loss": 0.0298,
      "step": 424500
    },
    {
      "epoch": 16.09,
      "learning_rate": 1.1959423142435369e-05,
      "loss": 0.0289,
      "step": 425000
    },
    {
      "epoch": 16.11,
      "learning_rate": 1.1949960255876454e-05,
      "loss": 0.0286,
      "step": 425500
    },
    {
      "epoch": 16.12,
      "learning_rate": 1.1940497369317538e-05,
      "loss": 0.0287,
      "step": 426000
    },
    {
      "epoch": 16.14,
      "learning_rate": 1.1931053408531739e-05,
      "loss": 0.0298,
      "step": 426500
    },
    {
      "epoch": 16.16,
      "learning_rate": 1.1921590521972823e-05,
      "loss": 0.0292,
      "step": 427000
    },
    {
      "epoch": 16.18,
      "learning_rate": 1.1912127635413908e-05,
      "loss": 0.0295,
      "step": 427500
    },
    {
      "epoch": 16.2,
      "learning_rate": 1.190266474885499e-05,
      "loss": 0.0307,
      "step": 428000
    },
    {
      "epoch": 16.22,
      "learning_rate": 1.1893201862296075e-05,
      "loss": 0.0291,
      "step": 428500
    },
    {
      "epoch": 16.24,
      "learning_rate": 1.1883757901510278e-05,
      "loss": 0.0309,
      "step": 429000
    },
    {
      "epoch": 16.26,
      "learning_rate": 1.1874295014951362e-05,
      "loss": 0.0301,
      "step": 429500
    },
    {
      "epoch": 16.28,
      "learning_rate": 1.1864832128392446e-05,
      "loss": 0.0298,
      "step": 430000
    },
    {
      "epoch": 16.3,
      "learning_rate": 1.185536924183353e-05,
      "loss": 0.0302,
      "step": 430500
    },
    {
      "epoch": 16.31,
      "learning_rate": 1.1845925281047732e-05,
      "loss": 0.0303,
      "step": 431000
    },
    {
      "epoch": 16.33,
      "learning_rate": 1.1836462394488817e-05,
      "loss": 0.0309,
      "step": 431500
    },
    {
      "epoch": 16.35,
      "learning_rate": 1.18269995079299e-05,
      "loss": 0.0295,
      "step": 432000
    },
    {
      "epoch": 16.37,
      "learning_rate": 1.1817536621370985e-05,
      "loss": 0.0311,
      "step": 432500
    },
    {
      "epoch": 16.39,
      "learning_rate": 1.1808073734812069e-05,
      "loss": 0.0306,
      "step": 433000
    },
    {
      "epoch": 16.41,
      "learning_rate": 1.179861084825315e-05,
      "loss": 0.0305,
      "step": 433500
    },
    {
      "epoch": 16.43,
      "learning_rate": 1.1789147961694236e-05,
      "loss": 0.0307,
      "step": 434000
    },
    {
      "epoch": 16.45,
      "learning_rate": 1.1779704000908439e-05,
      "loss": 0.0307,
      "step": 434500
    },
    {
      "epoch": 16.47,
      "learning_rate": 1.1770241114349522e-05,
      "loss": 0.0315,
      "step": 435000
    },
    {
      "epoch": 16.48,
      "learning_rate": 1.1760778227790606e-05,
      "loss": 0.0305,
      "step": 435500
    },
    {
      "epoch": 16.5,
      "learning_rate": 1.175131534123169e-05,
      "loss": 0.0312,
      "step": 436000
    },
    {
      "epoch": 16.52,
      "learning_rate": 1.1741852454672775e-05,
      "loss": 0.0311,
      "step": 436500
    },
    {
      "epoch": 16.54,
      "learning_rate": 1.1732389568113857e-05,
      "loss": 0.0313,
      "step": 437000
    },
    {
      "epoch": 16.56,
      "learning_rate": 1.172294560732806e-05,
      "loss": 0.0311,
      "step": 437500
    },
    {
      "epoch": 16.58,
      "learning_rate": 1.1713501646542263e-05,
      "loss": 0.0317,
      "step": 438000
    },
    {
      "epoch": 16.6,
      "learning_rate": 1.1704038759983347e-05,
      "loss": 0.031,
      "step": 438500
    },
    {
      "epoch": 16.62,
      "learning_rate": 1.169457587342443e-05,
      "loss": 0.0298,
      "step": 439000
    },
    {
      "epoch": 16.64,
      "learning_rate": 1.1685112986865514e-05,
      "loss": 0.0304,
      "step": 439500
    },
    {
      "epoch": 16.65,
      "learning_rate": 1.16756501003066e-05,
      "loss": 0.0311,
      "step": 440000
    },
    {
      "epoch": 16.67,
      "learning_rate": 1.1666187213747681e-05,
      "loss": 0.0305,
      "step": 440500
    },
    {
      "epoch": 16.69,
      "learning_rate": 1.1656724327188767e-05,
      "loss": 0.0312,
      "step": 441000
    },
    {
      "epoch": 16.71,
      "learning_rate": 1.164726144062985e-05,
      "loss": 0.031,
      "step": 441500
    },
    {
      "epoch": 16.73,
      "learning_rate": 1.1637798554070934e-05,
      "loss": 0.0301,
      "step": 442000
    },
    {
      "epoch": 16.75,
      "learning_rate": 1.1628335667512018e-05,
      "loss": 0.0325,
      "step": 442500
    },
    {
      "epoch": 16.77,
      "learning_rate": 1.1618872780953103e-05,
      "loss": 0.0305,
      "step": 443000
    },
    {
      "epoch": 16.79,
      "learning_rate": 1.1609409894394189e-05,
      "loss": 0.0315,
      "step": 443500
    },
    {
      "epoch": 16.81,
      "learning_rate": 1.159994700783527e-05,
      "loss": 0.0308,
      "step": 444000
    },
    {
      "epoch": 16.83,
      "learning_rate": 1.1590503047049473e-05,
      "loss": 0.0315,
      "step": 444500
    },
    {
      "epoch": 16.84,
      "learning_rate": 1.1581059086263675e-05,
      "loss": 0.0315,
      "step": 445000
    },
    {
      "epoch": 16.86,
      "learning_rate": 1.1571615125477877e-05,
      "loss": 0.032,
      "step": 445500
    },
    {
      "epoch": 16.88,
      "learning_rate": 1.156215223891896e-05,
      "loss": 0.0313,
      "step": 446000
    },
    {
      "epoch": 16.9,
      "learning_rate": 1.1552689352360045e-05,
      "loss": 0.0303,
      "step": 446500
    },
    {
      "epoch": 16.92,
      "learning_rate": 1.154322646580113e-05,
      "loss": 0.031,
      "step": 447000
    },
    {
      "epoch": 16.94,
      "learning_rate": 1.1533763579242212e-05,
      "loss": 0.0317,
      "step": 447500
    },
    {
      "epoch": 16.96,
      "learning_rate": 1.1524300692683298e-05,
      "loss": 0.0325,
      "step": 448000
    },
    {
      "epoch": 16.98,
      "learning_rate": 1.1514837806124381e-05,
      "loss": 0.0315,
      "step": 448500
    },
    {
      "epoch": 17.0,
      "learning_rate": 1.1505374919565467e-05,
      "loss": 0.032,
      "step": 449000
    },
    {
      "epoch": 17.0,
      "eval_bleu": 92.1868,
      "eval_gen_len": 16.4864,
      "eval_loss": 0.10941721498966217,
      "eval_runtime": 1187.2938,
      "eval_samples_per_second": 38.463,
      "eval_steps_per_second": 1.203,
      "step": 449123
    },
    {
      "epoch": 17.01,
      "learning_rate": 1.1495912033006549e-05,
      "loss": 0.027,
      "step": 449500
    },
    {
      "epoch": 17.03,
      "learning_rate": 1.1486468072220751e-05,
      "loss": 0.0272,
      "step": 450000
    },
    {
      "epoch": 17.05,
      "learning_rate": 1.1477005185661835e-05,
      "loss": 0.0265,
      "step": 450500
    },
    {
      "epoch": 17.07,
      "learning_rate": 1.1467542299102919e-05,
      "loss": 0.0257,
      "step": 451000
    },
    {
      "epoch": 17.09,
      "learning_rate": 1.1458079412544003e-05,
      "loss": 0.0264,
      "step": 451500
    },
    {
      "epoch": 17.11,
      "learning_rate": 1.1448616525985088e-05,
      "loss": 0.0268,
      "step": 452000
    },
    {
      "epoch": 17.13,
      "learning_rate": 1.143915363942617e-05,
      "loss": 0.027,
      "step": 452500
    },
    {
      "epoch": 17.15,
      "learning_rate": 1.1429690752867255e-05,
      "loss": 0.0268,
      "step": 453000
    },
    {
      "epoch": 17.17,
      "learning_rate": 1.1420246792081458e-05,
      "loss": 0.0271,
      "step": 453500
    },
    {
      "epoch": 17.18,
      "learning_rate": 1.1410783905522542e-05,
      "loss": 0.0275,
      "step": 454000
    },
    {
      "epoch": 17.2,
      "learning_rate": 1.1401321018963626e-05,
      "loss": 0.0263,
      "step": 454500
    },
    {
      "epoch": 17.22,
      "learning_rate": 1.139185813240471e-05,
      "loss": 0.0266,
      "step": 455000
    },
    {
      "epoch": 17.24,
      "learning_rate": 1.1382395245845795e-05,
      "loss": 0.0279,
      "step": 455500
    },
    {
      "epoch": 17.26,
      "learning_rate": 1.1372932359286877e-05,
      "loss": 0.0266,
      "step": 456000
    },
    {
      "epoch": 17.28,
      "learning_rate": 1.1363469472727962e-05,
      "loss": 0.027,
      "step": 456500
    },
    {
      "epoch": 17.3,
      "learning_rate": 1.1354006586169046e-05,
      "loss": 0.0273,
      "step": 457000
    },
    {
      "epoch": 17.32,
      "learning_rate": 1.1344562625383249e-05,
      "loss": 0.0283,
      "step": 457500
    },
    {
      "epoch": 17.34,
      "learning_rate": 1.133509973882433e-05,
      "loss": 0.0267,
      "step": 458000
    },
    {
      "epoch": 17.35,
      "learning_rate": 1.1325655778038533e-05,
      "loss": 0.0272,
      "step": 458500
    },
    {
      "epoch": 17.37,
      "learning_rate": 1.1316192891479619e-05,
      "loss": 0.028,
      "step": 459000
    },
    {
      "epoch": 17.39,
      "learning_rate": 1.130674893069382e-05,
      "loss": 0.0273,
      "step": 459500
    },
    {
      "epoch": 17.41,
      "learning_rate": 1.1297286044134904e-05,
      "loss": 0.0268,
      "step": 460000
    },
    {
      "epoch": 17.43,
      "learning_rate": 1.1287823157575987e-05,
      "loss": 0.0276,
      "step": 460500
    },
    {
      "epoch": 17.45,
      "learning_rate": 1.1278360271017073e-05,
      "loss": 0.0272,
      "step": 461000
    },
    {
      "epoch": 17.47,
      "learning_rate": 1.1268897384458155e-05,
      "loss": 0.0271,
      "step": 461500
    },
    {
      "epoch": 17.49,
      "learning_rate": 1.125943449789924e-05,
      "loss": 0.029,
      "step": 462000
    },
    {
      "epoch": 17.51,
      "learning_rate": 1.1249971611340324e-05,
      "loss": 0.0279,
      "step": 462500
    },
    {
      "epoch": 17.53,
      "learning_rate": 1.124050872478141e-05,
      "loss": 0.0288,
      "step": 463000
    },
    {
      "epoch": 17.54,
      "learning_rate": 1.1231045838222491e-05,
      "loss": 0.0286,
      "step": 463500
    },
    {
      "epoch": 17.56,
      "learning_rate": 1.1221601877436694e-05,
      "loss": 0.0284,
      "step": 464000
    },
    {
      "epoch": 17.58,
      "learning_rate": 1.121213899087778e-05,
      "loss": 0.0274,
      "step": 464500
    },
    {
      "epoch": 17.6,
      "learning_rate": 1.1202676104318861e-05,
      "loss": 0.0278,
      "step": 465000
    },
    {
      "epoch": 17.62,
      "learning_rate": 1.1193213217759947e-05,
      "loss": 0.0285,
      "step": 465500
    },
    {
      "epoch": 17.64,
      "learning_rate": 1.118375033120103e-05,
      "loss": 0.0281,
      "step": 466000
    },
    {
      "epoch": 17.66,
      "learning_rate": 1.1174287444642114e-05,
      "loss": 0.0277,
      "step": 466500
    },
    {
      "epoch": 17.68,
      "learning_rate": 1.1164843483856315e-05,
      "loss": 0.028,
      "step": 467000
    },
    {
      "epoch": 17.7,
      "learning_rate": 1.11553805972974e-05,
      "loss": 0.0275,
      "step": 467500
    },
    {
      "epoch": 17.71,
      "learning_rate": 1.1145936636511603e-05,
      "loss": 0.0278,
      "step": 468000
    },
    {
      "epoch": 17.73,
      "learning_rate": 1.1136473749952685e-05,
      "loss": 0.0287,
      "step": 468500
    },
    {
      "epoch": 17.75,
      "learning_rate": 1.1127010863393771e-05,
      "loss": 0.0289,
      "step": 469000
    },
    {
      "epoch": 17.77,
      "learning_rate": 1.1117547976834855e-05,
      "loss": 0.0279,
      "step": 469500
    },
    {
      "epoch": 17.79,
      "learning_rate": 1.110808509027594e-05,
      "loss": 0.0279,
      "step": 470000
    },
    {
      "epoch": 17.81,
      "learning_rate": 1.1098622203717022e-05,
      "loss": 0.0296,
      "step": 470500
    },
    {
      "epoch": 17.83,
      "learning_rate": 1.1089159317158107e-05,
      "loss": 0.029,
      "step": 471000
    },
    {
      "epoch": 17.85,
      "learning_rate": 1.1079715356372308e-05,
      "loss": 0.0293,
      "step": 471500
    },
    {
      "epoch": 17.87,
      "learning_rate": 1.1070252469813392e-05,
      "loss": 0.0275,
      "step": 472000
    },
    {
      "epoch": 17.88,
      "learning_rate": 1.1060789583254476e-05,
      "loss": 0.0286,
      "step": 472500
    },
    {
      "epoch": 17.9,
      "learning_rate": 1.1051326696695561e-05,
      "loss": 0.0285,
      "step": 473000
    },
    {
      "epoch": 17.92,
      "learning_rate": 1.1041863810136647e-05,
      "loss": 0.0285,
      "step": 473500
    },
    {
      "epoch": 17.94,
      "learning_rate": 1.1032400923577729e-05,
      "loss": 0.0288,
      "step": 474000
    },
    {
      "epoch": 17.96,
      "learning_rate": 1.1022938037018814e-05,
      "loss": 0.0291,
      "step": 474500
    },
    {
      "epoch": 17.98,
      "learning_rate": 1.1013475150459898e-05,
      "loss": 0.0291,
      "step": 475000
    },
    {
      "epoch": 18.0,
      "learning_rate": 1.1004031189674099e-05,
      "loss": 0.0302,
      "step": 475500
    },
    {
      "epoch": 18.0,
      "eval_bleu": 92.1969,
      "eval_gen_len": 16.4745,
      "eval_loss": 0.11121129244565964,
      "eval_runtime": 1185.9896,
      "eval_samples_per_second": 38.505,
      "eval_steps_per_second": 1.204,
      "step": 475542
    },
    {
      "epoch": 18.02,
      "learning_rate": 1.0994568303115183e-05,
      "loss": 0.0234,
      "step": 476000
    },
    {
      "epoch": 18.04,
      "learning_rate": 1.0985105416556268e-05,
      "loss": 0.0245,
      "step": 476500
    },
    {
      "epoch": 18.06,
      "learning_rate": 1.0975642529997352e-05,
      "loss": 0.0241,
      "step": 477000
    },
    {
      "epoch": 18.07,
      "learning_rate": 1.0966179643438435e-05,
      "loss": 0.0248,
      "step": 477500
    },
    {
      "epoch": 18.09,
      "learning_rate": 1.0956716756879519e-05,
      "loss": 0.0242,
      "step": 478000
    },
    {
      "epoch": 18.11,
      "learning_rate": 1.0947272796093722e-05,
      "loss": 0.0247,
      "step": 478500
    },
    {
      "epoch": 18.13,
      "learning_rate": 1.0937809909534804e-05,
      "loss": 0.0243,
      "step": 479000
    },
    {
      "epoch": 18.15,
      "learning_rate": 1.092834702297589e-05,
      "loss": 0.0238,
      "step": 479500
    },
    {
      "epoch": 18.17,
      "learning_rate": 1.0918884136416975e-05,
      "loss": 0.0249,
      "step": 480000
    },
    {
      "epoch": 18.19,
      "learning_rate": 1.0909440175631176e-05,
      "loss": 0.0251,
      "step": 480500
    },
    {
      "epoch": 18.21,
      "learning_rate": 1.089997728907226e-05,
      "loss": 0.0243,
      "step": 481000
    },
    {
      "epoch": 18.23,
      "learning_rate": 1.0890514402513343e-05,
      "loss": 0.0245,
      "step": 481500
    },
    {
      "epoch": 18.24,
      "learning_rate": 1.0881051515954429e-05,
      "loss": 0.0248,
      "step": 482000
    },
    {
      "epoch": 18.26,
      "learning_rate": 1.0871607555168628e-05,
      "loss": 0.0241,
      "step": 482500
    },
    {
      "epoch": 18.28,
      "learning_rate": 1.0862144668609713e-05,
      "loss": 0.0242,
      "step": 483000
    },
    {
      "epoch": 18.3,
      "learning_rate": 1.0852681782050799e-05,
      "loss": 0.0251,
      "step": 483500
    },
    {
      "epoch": 18.32,
      "learning_rate": 1.0843218895491882e-05,
      "loss": 0.025,
      "step": 484000
    },
    {
      "epoch": 18.34,
      "learning_rate": 1.0833756008932964e-05,
      "loss": 0.0246,
      "step": 484500
    },
    {
      "epoch": 18.36,
      "learning_rate": 1.082429312237405e-05,
      "loss": 0.0255,
      "step": 485000
    },
    {
      "epoch": 18.38,
      "learning_rate": 1.0814830235815135e-05,
      "loss": 0.0252,
      "step": 485500
    },
    {
      "epoch": 18.4,
      "learning_rate": 1.0805367349256217e-05,
      "loss": 0.0257,
      "step": 486000
    },
    {
      "epoch": 18.41,
      "learning_rate": 1.0795904462697303e-05,
      "loss": 0.0249,
      "step": 486500
    },
    {
      "epoch": 18.43,
      "learning_rate": 1.0786460501911504e-05,
      "loss": 0.0248,
      "step": 487000
    },
    {
      "epoch": 18.45,
      "learning_rate": 1.0776997615352589e-05,
      "loss": 0.025,
      "step": 487500
    },
    {
      "epoch": 18.47,
      "learning_rate": 1.0767553654566789e-05,
      "loss": 0.0258,
      "step": 488000
    },
    {
      "epoch": 18.49,
      "learning_rate": 1.0758090768007874e-05,
      "loss": 0.0249,
      "step": 488500
    },
    {
      "epoch": 18.51,
      "learning_rate": 1.074862788144896e-05,
      "loss": 0.0242,
      "step": 489000
    },
    {
      "epoch": 18.53,
      "learning_rate": 1.0739164994890041e-05,
      "loss": 0.0248,
      "step": 489500
    },
    {
      "epoch": 18.55,
      "learning_rate": 1.0729702108331127e-05,
      "loss": 0.026,
      "step": 490000
    },
    {
      "epoch": 18.57,
      "learning_rate": 1.0720258147545328e-05,
      "loss": 0.0254,
      "step": 490500
    },
    {
      "epoch": 18.59,
      "learning_rate": 1.0710795260986413e-05,
      "loss": 0.025,
      "step": 491000
    },
    {
      "epoch": 18.6,
      "learning_rate": 1.0701332374427495e-05,
      "loss": 0.0253,
      "step": 491500
    },
    {
      "epoch": 18.62,
      "learning_rate": 1.069186948786858e-05,
      "loss": 0.0255,
      "step": 492000
    },
    {
      "epoch": 18.64,
      "learning_rate": 1.0682406601309664e-05,
      "loss": 0.0263,
      "step": 492500
    },
    {
      "epoch": 18.66,
      "learning_rate": 1.0672962640523867e-05,
      "loss": 0.0254,
      "step": 493000
    },
    {
      "epoch": 18.68,
      "learning_rate": 1.0663499753964949e-05,
      "loss": 0.0255,
      "step": 493500
    },
    {
      "epoch": 18.7,
      "learning_rate": 1.0654055793179152e-05,
      "loss": 0.0255,
      "step": 494000
    },
    {
      "epoch": 18.72,
      "learning_rate": 1.0644592906620237e-05,
      "loss": 0.0266,
      "step": 494500
    },
    {
      "epoch": 18.74,
      "learning_rate": 1.063513002006132e-05,
      "loss": 0.0256,
      "step": 495000
    },
    {
      "epoch": 18.76,
      "learning_rate": 1.0625667133502405e-05,
      "loss": 0.0244,
      "step": 495500
    },
    {
      "epoch": 18.77,
      "learning_rate": 1.0616204246943488e-05,
      "loss": 0.0262,
      "step": 496000
    },
    {
      "epoch": 18.79,
      "learning_rate": 1.0606741360384572e-05,
      "loss": 0.0252,
      "step": 496500
    },
    {
      "epoch": 18.81,
      "learning_rate": 1.0597278473825656e-05,
      "loss": 0.0266,
      "step": 497000
    },
    {
      "epoch": 18.83,
      "learning_rate": 1.0587815587266741e-05,
      "loss": 0.0251,
      "step": 497500
    },
    {
      "epoch": 18.85,
      "learning_rate": 1.0578352700707825e-05,
      "loss": 0.0259,
      "step": 498000
    },
    {
      "epoch": 18.87,
      "learning_rate": 1.0568889814148909e-05,
      "loss": 0.0267,
      "step": 498500
    },
    {
      "epoch": 18.89,
      "learning_rate": 1.055944585336311e-05,
      "loss": 0.0254,
      "step": 499000
    },
    {
      "epoch": 18.91,
      "learning_rate": 1.0549982966804195e-05,
      "loss": 0.0254,
      "step": 499500
    },
    {
      "epoch": 18.93,
      "learning_rate": 1.0540539006018398e-05,
      "loss": 0.0262,
      "step": 500000
    },
    {
      "epoch": 18.94,
      "learning_rate": 1.053107611945948e-05,
      "loss": 0.0266,
      "step": 500500
    },
    {
      "epoch": 18.96,
      "learning_rate": 1.0521613232900565e-05,
      "loss": 0.0263,
      "step": 501000
    },
    {
      "epoch": 18.98,
      "learning_rate": 1.0512150346341649e-05,
      "loss": 0.0263,
      "step": 501500
    },
    {
      "epoch": 19.0,
      "eval_bleu": 92.283,
      "eval_gen_len": 16.4824,
      "eval_loss": 0.11453192681074142,
      "eval_runtime": 1203.1736,
      "eval_samples_per_second": 37.955,
      "eval_steps_per_second": 1.187,
      "step": 501961
    },
    {
      "epoch": 19.0,
      "learning_rate": 1.0502687459782733e-05,
      "loss": 0.0262,
      "step": 502000
    },
    {
      "epoch": 19.02,
      "learning_rate": 1.0493224573223816e-05,
      "loss": 0.0214,
      "step": 502500
    },
    {
      "epoch": 19.04,
      "learning_rate": 1.0483761686664902e-05,
      "loss": 0.0221,
      "step": 503000
    },
    {
      "epoch": 19.06,
      "learning_rate": 1.0474298800105984e-05,
      "loss": 0.0222,
      "step": 503500
    },
    {
      "epoch": 19.08,
      "learning_rate": 1.046483591354707e-05,
      "loss": 0.022,
      "step": 504000
    },
    {
      "epoch": 19.1,
      "learning_rate": 1.0455391952761272e-05,
      "loss": 0.0223,
      "step": 504500
    },
    {
      "epoch": 19.12,
      "learning_rate": 1.0445929066202356e-05,
      "loss": 0.0215,
      "step": 505000
    },
    {
      "epoch": 19.13,
      "learning_rate": 1.0436466179643438e-05,
      "loss": 0.0219,
      "step": 505500
    },
    {
      "epoch": 19.15,
      "learning_rate": 1.042702221885764e-05,
      "loss": 0.0223,
      "step": 506000
    },
    {
      "epoch": 19.17,
      "learning_rate": 1.0417559332298726e-05,
      "loss": 0.0217,
      "step": 506500
    },
    {
      "epoch": 19.19,
      "learning_rate": 1.040809644573981e-05,
      "loss": 0.0221,
      "step": 507000
    },
    {
      "epoch": 19.21,
      "learning_rate": 1.0398633559180893e-05,
      "loss": 0.0216,
      "step": 507500
    },
    {
      "epoch": 19.23,
      "learning_rate": 1.0389170672621977e-05,
      "loss": 0.0226,
      "step": 508000
    },
    {
      "epoch": 19.25,
      "learning_rate": 1.037972671183618e-05,
      "loss": 0.0228,
      "step": 508500
    },
    {
      "epoch": 19.27,
      "learning_rate": 1.0370263825277262e-05,
      "loss": 0.0226,
      "step": 509000
    },
    {
      "epoch": 19.29,
      "learning_rate": 1.0360800938718347e-05,
      "loss": 0.0228,
      "step": 509500
    },
    {
      "epoch": 19.3,
      "learning_rate": 1.0351338052159433e-05,
      "loss": 0.0223,
      "step": 510000
    },
    {
      "epoch": 19.32,
      "learning_rate": 1.0341875165600515e-05,
      "loss": 0.0222,
      "step": 510500
    },
    {
      "epoch": 19.34,
      "learning_rate": 1.0332431204814717e-05,
      "loss": 0.0222,
      "step": 511000
    },
    {
      "epoch": 19.36,
      "learning_rate": 1.0322968318255801e-05,
      "loss": 0.0224,
      "step": 511500
    },
    {
      "epoch": 19.38,
      "learning_rate": 1.0313505431696887e-05,
      "loss": 0.0225,
      "step": 512000
    },
    {
      "epoch": 19.4,
      "learning_rate": 1.0304042545137969e-05,
      "loss": 0.0226,
      "step": 512500
    },
    {
      "epoch": 19.42,
      "learning_rate": 1.0294579658579054e-05,
      "loss": 0.0235,
      "step": 513000
    },
    {
      "epoch": 19.44,
      "learning_rate": 1.0285116772020138e-05,
      "loss": 0.0229,
      "step": 513500
    },
    {
      "epoch": 19.46,
      "learning_rate": 1.0275653885461221e-05,
      "loss": 0.0232,
      "step": 514000
    },
    {
      "epoch": 19.47,
      "learning_rate": 1.0266190998902305e-05,
      "loss": 0.0226,
      "step": 514500
    },
    {
      "epoch": 19.49,
      "learning_rate": 1.025672811234339e-05,
      "loss": 0.023,
      "step": 515000
    },
    {
      "epoch": 19.51,
      "learning_rate": 1.0247265225784476e-05,
      "loss": 0.0227,
      "step": 515500
    },
    {
      "epoch": 19.53,
      "learning_rate": 1.0237802339225558e-05,
      "loss": 0.0234,
      "step": 516000
    },
    {
      "epoch": 19.55,
      "learning_rate": 1.022835837843976e-05,
      "loss": 0.0247,
      "step": 516500
    },
    {
      "epoch": 19.57,
      "learning_rate": 1.0218895491880844e-05,
      "loss": 0.0227,
      "step": 517000
    },
    {
      "epoch": 19.59,
      "learning_rate": 1.0209432605321928e-05,
      "loss": 0.0248,
      "step": 517500
    },
    {
      "epoch": 19.61,
      "learning_rate": 1.0199969718763012e-05,
      "loss": 0.0222,
      "step": 518000
    },
    {
      "epoch": 19.63,
      "learning_rate": 1.0190506832204097e-05,
      "loss": 0.0237,
      "step": 518500
    },
    {
      "epoch": 19.64,
      "learning_rate": 1.018104394564518e-05,
      "loss": 0.0232,
      "step": 519000
    },
    {
      "epoch": 19.66,
      "learning_rate": 1.0171581059086265e-05,
      "loss": 0.0232,
      "step": 519500
    },
    {
      "epoch": 19.68,
      "learning_rate": 1.0162118172527348e-05,
      "loss": 0.0227,
      "step": 520000
    },
    {
      "epoch": 19.7,
      "learning_rate": 1.0152655285968434e-05,
      "loss": 0.0237,
      "step": 520500
    },
    {
      "epoch": 19.72,
      "learning_rate": 1.0143230250955754e-05,
      "loss": 0.0232,
      "step": 521000
    },
    {
      "epoch": 19.74,
      "learning_rate": 1.0133767364396836e-05,
      "loss": 0.0228,
      "step": 521500
    },
    {
      "epoch": 19.76,
      "learning_rate": 1.0124304477837921e-05,
      "loss": 0.0241,
      "step": 522000
    },
    {
      "epoch": 19.78,
      "learning_rate": 1.0114841591279005e-05,
      "loss": 0.0236,
      "step": 522500
    },
    {
      "epoch": 19.8,
      "learning_rate": 1.0105378704720089e-05,
      "loss": 0.0234,
      "step": 523000
    },
    {
      "epoch": 19.82,
      "learning_rate": 1.009593474393429e-05,
      "loss": 0.0245,
      "step": 523500
    },
    {
      "epoch": 19.83,
      "learning_rate": 1.0086471857375375e-05,
      "loss": 0.0234,
      "step": 524000
    },
    {
      "epoch": 19.85,
      "learning_rate": 1.0077008970816457e-05,
      "loss": 0.0234,
      "step": 524500
    },
    {
      "epoch": 19.87,
      "learning_rate": 1.0067546084257543e-05,
      "loss": 0.0231,
      "step": 525000
    },
    {
      "epoch": 19.89,
      "learning_rate": 1.0058083197698626e-05,
      "loss": 0.0222,
      "step": 525500
    },
    {
      "epoch": 19.91,
      "learning_rate": 1.0048620311139712e-05,
      "loss": 0.0233,
      "step": 526000
    },
    {
      "epoch": 19.93,
      "learning_rate": 1.0039157424580794e-05,
      "loss": 0.0238,
      "step": 526500
    },
    {
      "epoch": 19.95,
      "learning_rate": 1.0029694538021879e-05,
      "loss": 0.0238,
      "step": 527000
    },
    {
      "epoch": 19.97,
      "learning_rate": 1.0020231651462964e-05,
      "loss": 0.0237,
      "step": 527500
    },
    {
      "epoch": 19.99,
      "learning_rate": 1.0010787690677164e-05,
      "loss": 0.0233,
      "step": 528000
    },
    {
      "epoch": 20.0,
      "eval_bleu": 92.2202,
      "eval_gen_len": 16.5063,
      "eval_loss": 0.11753202229738235,
      "eval_runtime": 1202.9336,
      "eval_samples_per_second": 37.963,
      "eval_steps_per_second": 1.187,
      "step": 528380
    },
    {
      "epoch": 20.0,
      "learning_rate": 1.000132480411825e-05,
      "loss": 0.022,
      "step": 528500
    },
    {
      "epoch": 20.02,
      "learning_rate": 9.991861917559333e-06,
      "loss": 0.0201,
      "step": 529000
    },
    {
      "epoch": 20.04,
      "learning_rate": 9.982399031000417e-06,
      "loss": 0.0198,
      "step": 529500
    },
    {
      "epoch": 20.06,
      "learning_rate": 9.97295507021462e-06,
      "loss": 0.0198,
      "step": 530000
    },
    {
      "epoch": 20.08,
      "learning_rate": 9.96351110942882e-06,
      "loss": 0.0202,
      "step": 530500
    },
    {
      "epoch": 20.1,
      "learning_rate": 9.954048222869906e-06,
      "loss": 0.0198,
      "step": 531000
    },
    {
      "epoch": 20.12,
      "learning_rate": 9.94458533631099e-06,
      "loss": 0.02,
      "step": 531500
    },
    {
      "epoch": 20.14,
      "learning_rate": 9.93514137552519e-06,
      "loss": 0.0201,
      "step": 532000
    },
    {
      "epoch": 20.16,
      "learning_rate": 9.925678488966274e-06,
      "loss": 0.0201,
      "step": 532500
    },
    {
      "epoch": 20.17,
      "learning_rate": 9.91621560240736e-06,
      "loss": 0.0203,
      "step": 533000
    },
    {
      "epoch": 20.19,
      "learning_rate": 9.906752715848444e-06,
      "loss": 0.021,
      "step": 533500
    },
    {
      "epoch": 20.21,
      "learning_rate": 9.897289829289527e-06,
      "loss": 0.0207,
      "step": 534000
    },
    {
      "epoch": 20.23,
      "learning_rate": 9.887826942730611e-06,
      "loss": 0.0206,
      "step": 534500
    },
    {
      "epoch": 20.25,
      "learning_rate": 9.878364056171695e-06,
      "loss": 0.0205,
      "step": 535000
    },
    {
      "epoch": 20.27,
      "learning_rate": 9.868901169612778e-06,
      "loss": 0.0203,
      "step": 535500
    },
    {
      "epoch": 20.29,
      "learning_rate": 9.859438283053864e-06,
      "loss": 0.02,
      "step": 536000
    },
    {
      "epoch": 20.31,
      "learning_rate": 9.849975396494947e-06,
      "loss": 0.0201,
      "step": 536500
    },
    {
      "epoch": 20.33,
      "learning_rate": 9.840512509936033e-06,
      "loss": 0.0209,
      "step": 537000
    },
    {
      "epoch": 20.35,
      "learning_rate": 9.831049623377117e-06,
      "loss": 0.0202,
      "step": 537500
    },
    {
      "epoch": 20.36,
      "learning_rate": 9.821605662591318e-06,
      "loss": 0.0196,
      "step": 538000
    },
    {
      "epoch": 20.38,
      "learning_rate": 9.812161701805519e-06,
      "loss": 0.0206,
      "step": 538500
    },
    {
      "epoch": 20.4,
      "learning_rate": 9.802698815246604e-06,
      "loss": 0.0212,
      "step": 539000
    },
    {
      "epoch": 20.42,
      "learning_rate": 9.793235928687688e-06,
      "loss": 0.0204,
      "step": 539500
    },
    {
      "epoch": 20.44,
      "learning_rate": 9.783773042128772e-06,
      "loss": 0.0199,
      "step": 540000
    },
    {
      "epoch": 20.46,
      "learning_rate": 9.774310155569855e-06,
      "loss": 0.0213,
      "step": 540500
    },
    {
      "epoch": 20.48,
      "learning_rate": 9.764866194784058e-06,
      "loss": 0.0206,
      "step": 541000
    },
    {
      "epoch": 20.5,
      "learning_rate": 9.755403308225142e-06,
      "loss": 0.0205,
      "step": 541500
    },
    {
      "epoch": 20.52,
      "learning_rate": 9.745940421666225e-06,
      "loss": 0.0208,
      "step": 542000
    },
    {
      "epoch": 20.53,
      "learning_rate": 9.73647753510731e-06,
      "loss": 0.0212,
      "step": 542500
    },
    {
      "epoch": 20.55,
      "learning_rate": 9.727014648548395e-06,
      "loss": 0.0206,
      "step": 543000
    },
    {
      "epoch": 20.57,
      "learning_rate": 9.717570687762596e-06,
      "loss": 0.0214,
      "step": 543500
    },
    {
      "epoch": 20.59,
      "learning_rate": 9.70810780120368e-06,
      "loss": 0.0207,
      "step": 544000
    },
    {
      "epoch": 20.61,
      "learning_rate": 9.698644914644763e-06,
      "loss": 0.021,
      "step": 544500
    },
    {
      "epoch": 20.63,
      "learning_rate": 9.689182028085848e-06,
      "loss": 0.0215,
      "step": 545000
    },
    {
      "epoch": 20.65,
      "learning_rate": 9.679719141526932e-06,
      "loss": 0.0214,
      "step": 545500
    },
    {
      "epoch": 20.67,
      "learning_rate": 9.670275180741135e-06,
      "loss": 0.0206,
      "step": 546000
    },
    {
      "epoch": 20.69,
      "learning_rate": 9.660812294182219e-06,
      "loss": 0.0208,
      "step": 546500
    },
    {
      "epoch": 20.7,
      "learning_rate": 9.65136833339642e-06,
      "loss": 0.0213,
      "step": 547000
    },
    {
      "epoch": 20.72,
      "learning_rate": 9.641905446837503e-06,
      "loss": 0.021,
      "step": 547500
    },
    {
      "epoch": 20.74,
      "learning_rate": 9.632442560278589e-06,
      "loss": 0.0213,
      "step": 548000
    },
    {
      "epoch": 20.76,
      "learning_rate": 9.622979673719673e-06,
      "loss": 0.0206,
      "step": 548500
    },
    {
      "epoch": 20.78,
      "learning_rate": 9.613516787160756e-06,
      "loss": 0.0215,
      "step": 549000
    },
    {
      "epoch": 20.8,
      "learning_rate": 9.60405390060184e-06,
      "loss": 0.0217,
      "step": 549500
    },
    {
      "epoch": 20.82,
      "learning_rate": 9.594591014042924e-06,
      "loss": 0.0216,
      "step": 550000
    },
    {
      "epoch": 20.84,
      "learning_rate": 9.585128127484007e-06,
      "loss": 0.021,
      "step": 550500
    },
    {
      "epoch": 20.86,
      "learning_rate": 9.575665240925093e-06,
      "loss": 0.0216,
      "step": 551000
    },
    {
      "epoch": 20.88,
      "learning_rate": 9.566202354366176e-06,
      "loss": 0.0216,
      "step": 551500
    },
    {
      "epoch": 20.89,
      "learning_rate": 9.556739467807262e-06,
      "loss": 0.0214,
      "step": 552000
    },
    {
      "epoch": 20.91,
      "learning_rate": 9.547295507021463e-06,
      "loss": 0.0214,
      "step": 552500
    },
    {
      "epoch": 20.93,
      "learning_rate": 9.537832620462547e-06,
      "loss": 0.0219,
      "step": 553000
    },
    {
      "epoch": 20.95,
      "learning_rate": 9.528388659676748e-06,
      "loss": 0.0223,
      "step": 553500
    },
    {
      "epoch": 20.97,
      "learning_rate": 9.518925773117833e-06,
      "loss": 0.0208,
      "step": 554000
    },
    {
      "epoch": 20.99,
      "learning_rate": 9.509462886558917e-06,
      "loss": 0.0219,
      "step": 554500
    },
    {
      "epoch": 21.0,
      "eval_bleu": 92.2963,
      "eval_gen_len": 16.4984,
      "eval_loss": 0.11960500478744507,
      "eval_runtime": 1195.7203,
      "eval_samples_per_second": 38.192,
      "eval_steps_per_second": 1.194,
      "step": 554799
    },
    {
      "epoch": 21.01,
      "learning_rate": 9.5e-06,
      "loss": 0.0197,
      "step": 555000
    },
    {
      "epoch": 21.03,
      "learning_rate": 9.490537113441086e-06,
      "loss": 0.0184,
      "step": 555500
    },
    {
      "epoch": 21.05,
      "learning_rate": 9.481074226882168e-06,
      "loss": 0.0185,
      "step": 556000
    },
    {
      "epoch": 21.06,
      "learning_rate": 9.471611340323253e-06,
      "loss": 0.0183,
      "step": 556500
    },
    {
      "epoch": 21.08,
      "learning_rate": 9.462148453764337e-06,
      "loss": 0.0184,
      "step": 557000
    },
    {
      "epoch": 21.1,
      "learning_rate": 9.45268556720542e-06,
      "loss": 0.0183,
      "step": 557500
    },
    {
      "epoch": 21.12,
      "learning_rate": 9.443222680646506e-06,
      "loss": 0.0179,
      "step": 558000
    },
    {
      "epoch": 21.14,
      "learning_rate": 9.43375979408759e-06,
      "loss": 0.0186,
      "step": 558500
    },
    {
      "epoch": 21.16,
      "learning_rate": 9.424315833301791e-06,
      "loss": 0.0184,
      "step": 559000
    },
    {
      "epoch": 21.18,
      "learning_rate": 9.414871872515992e-06,
      "loss": 0.0184,
      "step": 559500
    },
    {
      "epoch": 21.2,
      "learning_rate": 9.405408985957077e-06,
      "loss": 0.0184,
      "step": 560000
    },
    {
      "epoch": 21.22,
      "learning_rate": 9.395946099398161e-06,
      "loss": 0.0185,
      "step": 560500
    },
    {
      "epoch": 21.23,
      "learning_rate": 9.386483212839247e-06,
      "loss": 0.0184,
      "step": 561000
    },
    {
      "epoch": 21.25,
      "learning_rate": 9.37702032628033e-06,
      "loss": 0.0178,
      "step": 561500
    },
    {
      "epoch": 21.27,
      "learning_rate": 9.367557439721412e-06,
      "loss": 0.0189,
      "step": 562000
    },
    {
      "epoch": 21.29,
      "learning_rate": 9.358094553162498e-06,
      "loss": 0.0189,
      "step": 562500
    },
    {
      "epoch": 21.31,
      "learning_rate": 9.348631666603581e-06,
      "loss": 0.0185,
      "step": 563000
    },
    {
      "epoch": 21.33,
      "learning_rate": 9.339168780044665e-06,
      "loss": 0.0189,
      "step": 563500
    },
    {
      "epoch": 21.35,
      "learning_rate": 9.329724819258868e-06,
      "loss": 0.018,
      "step": 564000
    },
    {
      "epoch": 21.37,
      "learning_rate": 9.320261932699952e-06,
      "loss": 0.0183,
      "step": 564500
    },
    {
      "epoch": 21.39,
      "learning_rate": 9.310817971914153e-06,
      "loss": 0.0188,
      "step": 565000
    },
    {
      "epoch": 21.41,
      "learning_rate": 9.301355085355236e-06,
      "loss": 0.0193,
      "step": 565500
    },
    {
      "epoch": 21.42,
      "learning_rate": 9.291892198796322e-06,
      "loss": 0.0187,
      "step": 566000
    },
    {
      "epoch": 21.44,
      "learning_rate": 9.282429312237405e-06,
      "loss": 0.0195,
      "step": 566500
    },
    {
      "epoch": 21.46,
      "learning_rate": 9.27296642567849e-06,
      "loss": 0.0191,
      "step": 567000
    },
    {
      "epoch": 21.48,
      "learning_rate": 9.263503539119575e-06,
      "loss": 0.0182,
      "step": 567500
    },
    {
      "epoch": 21.5,
      "learning_rate": 9.254040652560658e-06,
      "loss": 0.0196,
      "step": 568000
    },
    {
      "epoch": 21.52,
      "learning_rate": 9.244577766001742e-06,
      "loss": 0.0189,
      "step": 568500
    },
    {
      "epoch": 21.54,
      "learning_rate": 9.235114879442826e-06,
      "loss": 0.019,
      "step": 569000
    },
    {
      "epoch": 21.56,
      "learning_rate": 9.225651992883911e-06,
      "loss": 0.0191,
      "step": 569500
    },
    {
      "epoch": 21.58,
      "learning_rate": 9.216189106324995e-06,
      "loss": 0.0192,
      "step": 570000
    },
    {
      "epoch": 21.59,
      "learning_rate": 9.206745145539196e-06,
      "loss": 0.0197,
      "step": 570500
    },
    {
      "epoch": 21.61,
      "learning_rate": 9.19728225898028e-06,
      "loss": 0.0184,
      "step": 571000
    },
    {
      "epoch": 21.63,
      "learning_rate": 9.187819372421363e-06,
      "loss": 0.019,
      "step": 571500
    },
    {
      "epoch": 21.65,
      "learning_rate": 9.178375411635566e-06,
      "loss": 0.0197,
      "step": 572000
    },
    {
      "epoch": 21.67,
      "learning_rate": 9.16891252507665e-06,
      "loss": 0.019,
      "step": 572500
    },
    {
      "epoch": 21.69,
      "learning_rate": 9.159449638517735e-06,
      "loss": 0.0186,
      "step": 573000
    },
    {
      "epoch": 21.71,
      "learning_rate": 9.149986751958819e-06,
      "loss": 0.0186,
      "step": 573500
    },
    {
      "epoch": 21.73,
      "learning_rate": 9.140523865399903e-06,
      "loss": 0.0184,
      "step": 574000
    },
    {
      "epoch": 21.75,
      "learning_rate": 9.131060978840986e-06,
      "loss": 0.0196,
      "step": 574500
    },
    {
      "epoch": 21.76,
      "learning_rate": 9.121617018055189e-06,
      "loss": 0.0197,
      "step": 575000
    },
    {
      "epoch": 21.78,
      "learning_rate": 9.112154131496273e-06,
      "loss": 0.0188,
      "step": 575500
    },
    {
      "epoch": 21.8,
      "learning_rate": 9.102691244937356e-06,
      "loss": 0.0196,
      "step": 576000
    },
    {
      "epoch": 21.82,
      "learning_rate": 9.09322835837844e-06,
      "loss": 0.0188,
      "step": 576500
    },
    {
      "epoch": 21.84,
      "learning_rate": 9.083765471819524e-06,
      "loss": 0.0194,
      "step": 577000
    },
    {
      "epoch": 21.86,
      "learning_rate": 9.074302585260608e-06,
      "loss": 0.0195,
      "step": 577500
    },
    {
      "epoch": 21.88,
      "learning_rate": 9.064839698701693e-06,
      "loss": 0.0199,
      "step": 578000
    },
    {
      "epoch": 21.9,
      "learning_rate": 9.055376812142777e-06,
      "loss": 0.0188,
      "step": 578500
    },
    {
      "epoch": 21.92,
      "learning_rate": 9.045913925583862e-06,
      "loss": 0.0196,
      "step": 579000
    },
    {
      "epoch": 21.93,
      "learning_rate": 9.036451039024946e-06,
      "loss": 0.0195,
      "step": 579500
    },
    {
      "epoch": 21.95,
      "learning_rate": 9.026988152466028e-06,
      "loss": 0.0202,
      "step": 580000
    },
    {
      "epoch": 21.97,
      "learning_rate": 9.017525265907113e-06,
      "loss": 0.0189,
      "step": 580500
    },
    {
      "epoch": 21.99,
      "learning_rate": 9.008062379348197e-06,
      "loss": 0.0196,
      "step": 581000
    },
    {
      "epoch": 22.0,
      "eval_bleu": 92.2659,
      "eval_gen_len": 16.477,
      "eval_loss": 0.12201742082834244,
      "eval_runtime": 1201.2438,
      "eval_samples_per_second": 38.016,
      "eval_steps_per_second": 1.189,
      "step": 581218
    },
    {
      "epoch": 22.01,
      "learning_rate": 8.998637344335517e-06,
      "loss": 0.0182,
      "step": 581500
    },
    {
      "epoch": 22.03,
      "learning_rate": 8.9891744577766e-06,
      "loss": 0.0163,
      "step": 582000
    },
    {
      "epoch": 22.05,
      "learning_rate": 8.979730496990804e-06,
      "loss": 0.0169,
      "step": 582500
    },
    {
      "epoch": 22.07,
      "learning_rate": 8.970267610431887e-06,
      "loss": 0.0162,
      "step": 583000
    },
    {
      "epoch": 22.09,
      "learning_rate": 8.960804723872971e-06,
      "loss": 0.0167,
      "step": 583500
    },
    {
      "epoch": 22.11,
      "learning_rate": 8.951341837314055e-06,
      "loss": 0.016,
      "step": 584000
    },
    {
      "epoch": 22.12,
      "learning_rate": 8.94187895075514e-06,
      "loss": 0.0163,
      "step": 584500
    },
    {
      "epoch": 22.14,
      "learning_rate": 8.932416064196224e-06,
      "loss": 0.0165,
      "step": 585000
    },
    {
      "epoch": 22.16,
      "learning_rate": 8.922953177637307e-06,
      "loss": 0.0173,
      "step": 585500
    },
    {
      "epoch": 22.18,
      "learning_rate": 8.913490291078391e-06,
      "loss": 0.0166,
      "step": 586000
    },
    {
      "epoch": 22.2,
      "learning_rate": 8.904027404519475e-06,
      "loss": 0.0172,
      "step": 586500
    },
    {
      "epoch": 22.22,
      "learning_rate": 8.894564517960559e-06,
      "loss": 0.0171,
      "step": 587000
    },
    {
      "epoch": 22.24,
      "learning_rate": 8.885101631401644e-06,
      "loss": 0.0167,
      "step": 587500
    },
    {
      "epoch": 22.26,
      "learning_rate": 8.875638744842728e-06,
      "loss": 0.0169,
      "step": 588000
    },
    {
      "epoch": 22.28,
      "learning_rate": 8.86619478405693e-06,
      "loss": 0.0175,
      "step": 588500
    },
    {
      "epoch": 22.29,
      "learning_rate": 8.856731897498012e-06,
      "loss": 0.0161,
      "step": 589000
    },
    {
      "epoch": 22.31,
      "learning_rate": 8.847269010939098e-06,
      "loss": 0.0167,
      "step": 589500
    },
    {
      "epoch": 22.33,
      "learning_rate": 8.837806124380182e-06,
      "loss": 0.0171,
      "step": 590000
    },
    {
      "epoch": 22.35,
      "learning_rate": 8.828362163594384e-06,
      "loss": 0.0175,
      "step": 590500
    },
    {
      "epoch": 22.37,
      "learning_rate": 8.818899277035468e-06,
      "loss": 0.0166,
      "step": 591000
    },
    {
      "epoch": 22.39,
      "learning_rate": 8.809455316249669e-06,
      "loss": 0.0164,
      "step": 591500
    },
    {
      "epoch": 22.41,
      "learning_rate": 8.799992429690753e-06,
      "loss": 0.0172,
      "step": 592000
    },
    {
      "epoch": 22.43,
      "learning_rate": 8.790529543131837e-06,
      "loss": 0.0176,
      "step": 592500
    },
    {
      "epoch": 22.45,
      "learning_rate": 8.781066656572922e-06,
      "loss": 0.0175,
      "step": 593000
    },
    {
      "epoch": 22.46,
      "learning_rate": 8.771622695787123e-06,
      "loss": 0.0172,
      "step": 593500
    },
    {
      "epoch": 22.48,
      "learning_rate": 8.762159809228208e-06,
      "loss": 0.0177,
      "step": 594000
    },
    {
      "epoch": 22.5,
      "learning_rate": 8.752696922669292e-06,
      "loss": 0.0173,
      "step": 594500
    },
    {
      "epoch": 22.52,
      "learning_rate": 8.743234036110376e-06,
      "loss": 0.0167,
      "step": 595000
    },
    {
      "epoch": 22.54,
      "learning_rate": 8.73377114955146e-06,
      "loss": 0.0173,
      "step": 595500
    },
    {
      "epoch": 22.56,
      "learning_rate": 8.72434611453878e-06,
      "loss": 0.0173,
      "step": 596000
    },
    {
      "epoch": 22.58,
      "learning_rate": 8.714883227979863e-06,
      "loss": 0.0178,
      "step": 596500
    },
    {
      "epoch": 22.6,
      "learning_rate": 8.705420341420949e-06,
      "loss": 0.0176,
      "step": 597000
    },
    {
      "epoch": 22.62,
      "learning_rate": 8.695957454862033e-06,
      "loss": 0.017,
      "step": 597500
    },
    {
      "epoch": 22.64,
      "learning_rate": 8.686494568303116e-06,
      "loss": 0.0175,
      "step": 598000
    },
    {
      "epoch": 22.65,
      "learning_rate": 8.6770316817442e-06,
      "loss": 0.0171,
      "step": 598500
    },
    {
      "epoch": 22.67,
      "learning_rate": 8.667568795185284e-06,
      "loss": 0.0181,
      "step": 599000
    },
    {
      "epoch": 22.69,
      "learning_rate": 8.658105908626369e-06,
      "loss": 0.0175,
      "step": 599500
    },
    {
      "epoch": 22.71,
      "learning_rate": 8.648643022067453e-06,
      "loss": 0.0175,
      "step": 600000
    },
    {
      "epoch": 22.73,
      "learning_rate": 8.639180135508536e-06,
      "loss": 0.0185,
      "step": 600500
    },
    {
      "epoch": 22.75,
      "learning_rate": 8.629736174722738e-06,
      "loss": 0.0179,
      "step": 601000
    },
    {
      "epoch": 22.77,
      "learning_rate": 8.620273288163821e-06,
      "loss": 0.018,
      "step": 601500
    },
    {
      "epoch": 22.79,
      "learning_rate": 8.610810401604907e-06,
      "loss": 0.0172,
      "step": 602000
    },
    {
      "epoch": 22.81,
      "learning_rate": 8.60134751504599e-06,
      "loss": 0.0176,
      "step": 602500
    },
    {
      "epoch": 22.82,
      "learning_rate": 8.591903554260193e-06,
      "loss": 0.0177,
      "step": 603000
    },
    {
      "epoch": 22.84,
      "learning_rate": 8.582440667701277e-06,
      "loss": 0.0171,
      "step": 603500
    },
    {
      "epoch": 22.86,
      "learning_rate": 8.57297778114236e-06,
      "loss": 0.0176,
      "step": 604000
    },
    {
      "epoch": 22.88,
      "learning_rate": 8.563514894583444e-06,
      "loss": 0.0176,
      "step": 604500
    },
    {
      "epoch": 22.9,
      "learning_rate": 8.554052008024528e-06,
      "loss": 0.0177,
      "step": 605000
    },
    {
      "epoch": 22.92,
      "learning_rate": 8.544589121465613e-06,
      "loss": 0.0176,
      "step": 605500
    },
    {
      "epoch": 22.94,
      "learning_rate": 8.535126234906697e-06,
      "loss": 0.018,
      "step": 606000
    },
    {
      "epoch": 22.96,
      "learning_rate": 8.52566334834778e-06,
      "loss": 0.0179,
      "step": 606500
    },
    {
      "epoch": 22.98,
      "learning_rate": 8.516200461788864e-06,
      "loss": 0.0173,
      "step": 607000
    },
    {
      "epoch": 22.99,
      "learning_rate": 8.506737575229948e-06,
      "loss": 0.0181,
      "step": 607500
    },
    {
      "epoch": 23.0,
      "eval_bleu": 92.2975,
      "eval_gen_len": 16.4935,
      "eval_loss": 0.12451058626174927,
      "eval_runtime": 1203.5654,
      "eval_samples_per_second": 37.943,
      "eval_steps_per_second": 1.186,
      "step": 607637
    },
    {
      "epoch": 23.01,
      "learning_rate": 8.497312540217268e-06,
      "loss": 0.0158,
      "step": 608000
    },
    {
      "epoch": 23.03,
      "learning_rate": 8.487849653658352e-06,
      "loss": 0.0151,
      "step": 608500
    },
    {
      "epoch": 23.05,
      "learning_rate": 8.478386767099437e-06,
      "loss": 0.0156,
      "step": 609000
    },
    {
      "epoch": 23.07,
      "learning_rate": 8.468923880540521e-06,
      "loss": 0.0154,
      "step": 609500
    },
    {
      "epoch": 23.09,
      "learning_rate": 8.459460993981605e-06,
      "loss": 0.0149,
      "step": 610000
    },
    {
      "epoch": 23.11,
      "learning_rate": 8.449998107422689e-06,
      "loss": 0.0154,
      "step": 610500
    },
    {
      "epoch": 23.13,
      "learning_rate": 8.440535220863772e-06,
      "loss": 0.0157,
      "step": 611000
    },
    {
      "epoch": 23.15,
      "learning_rate": 8.431072334304858e-06,
      "loss": 0.0157,
      "step": 611500
    },
    {
      "epoch": 23.17,
      "learning_rate": 8.421609447745941e-06,
      "loss": 0.0151,
      "step": 612000
    },
    {
      "epoch": 23.18,
      "learning_rate": 8.412146561187025e-06,
      "loss": 0.0149,
      "step": 612500
    },
    {
      "epoch": 23.2,
      "learning_rate": 8.402683674628109e-06,
      "loss": 0.0151,
      "step": 613000
    },
    {
      "epoch": 23.22,
      "learning_rate": 8.393220788069192e-06,
      "loss": 0.0151,
      "step": 613500
    },
    {
      "epoch": 23.24,
      "learning_rate": 8.383757901510278e-06,
      "loss": 0.0154,
      "step": 614000
    },
    {
      "epoch": 23.26,
      "learning_rate": 8.374332866497598e-06,
      "loss": 0.0156,
      "step": 614500
    },
    {
      "epoch": 23.28,
      "learning_rate": 8.364869979938682e-06,
      "loss": 0.0157,
      "step": 615000
    },
    {
      "epoch": 23.3,
      "learning_rate": 8.355407093379765e-06,
      "loss": 0.0158,
      "step": 615500
    },
    {
      "epoch": 23.32,
      "learning_rate": 8.345944206820849e-06,
      "loss": 0.0156,
      "step": 616000
    },
    {
      "epoch": 23.34,
      "learning_rate": 8.336481320261933e-06,
      "loss": 0.0152,
      "step": 616500
    },
    {
      "epoch": 23.35,
      "learning_rate": 8.327037359476136e-06,
      "loss": 0.0153,
      "step": 617000
    },
    {
      "epoch": 23.37,
      "learning_rate": 8.31757447291722e-06,
      "loss": 0.0154,
      "step": 617500
    },
    {
      "epoch": 23.39,
      "learning_rate": 8.308111586358305e-06,
      "loss": 0.0159,
      "step": 618000
    },
    {
      "epoch": 23.41,
      "learning_rate": 8.298648699799388e-06,
      "loss": 0.0154,
      "step": 618500
    },
    {
      "epoch": 23.43,
      "learning_rate": 8.28918581324047e-06,
      "loss": 0.0156,
      "step": 619000
    },
    {
      "epoch": 23.45,
      "learning_rate": 8.279722926681556e-06,
      "loss": 0.0155,
      "step": 619500
    },
    {
      "epoch": 23.47,
      "learning_rate": 8.27026004012264e-06,
      "loss": 0.0157,
      "step": 620000
    },
    {
      "epoch": 23.49,
      "learning_rate": 8.260797153563723e-06,
      "loss": 0.016,
      "step": 620500
    },
    {
      "epoch": 23.51,
      "learning_rate": 8.251353192777926e-06,
      "loss": 0.016,
      "step": 621000
    },
    {
      "epoch": 23.52,
      "learning_rate": 8.24189030621901e-06,
      "loss": 0.0161,
      "step": 621500
    },
    {
      "epoch": 23.54,
      "learning_rate": 8.232427419660093e-06,
      "loss": 0.0158,
      "step": 622000
    },
    {
      "epoch": 23.56,
      "learning_rate": 8.222964533101177e-06,
      "loss": 0.0157,
      "step": 622500
    },
    {
      "epoch": 23.58,
      "learning_rate": 8.213501646542262e-06,
      "loss": 0.017,
      "step": 623000
    },
    {
      "epoch": 23.6,
      "learning_rate": 8.204057685756464e-06,
      "loss": 0.0157,
      "step": 623500
    },
    {
      "epoch": 23.62,
      "learning_rate": 8.194594799197549e-06,
      "loss": 0.0167,
      "step": 624000
    },
    {
      "epoch": 23.64,
      "learning_rate": 8.185131912638633e-06,
      "loss": 0.0156,
      "step": 624500
    },
    {
      "epoch": 23.66,
      "learning_rate": 8.175669026079716e-06,
      "loss": 0.0162,
      "step": 625000
    },
    {
      "epoch": 23.68,
      "learning_rate": 8.1662061395208e-06,
      "loss": 0.0164,
      "step": 625500
    },
    {
      "epoch": 23.7,
      "learning_rate": 8.156762178735001e-06,
      "loss": 0.0165,
      "step": 626000
    },
    {
      "epoch": 23.71,
      "learning_rate": 8.147299292176087e-06,
      "loss": 0.0157,
      "step": 626500
    },
    {
      "epoch": 23.73,
      "learning_rate": 8.13783640561717e-06,
      "loss": 0.0164,
      "step": 627000
    },
    {
      "epoch": 23.75,
      "learning_rate": 8.128373519058254e-06,
      "loss": 0.0164,
      "step": 627500
    },
    {
      "epoch": 23.77,
      "learning_rate": 8.118910632499338e-06,
      "loss": 0.0158,
      "step": 628000
    },
    {
      "epoch": 23.79,
      "learning_rate": 8.10946667171354e-06,
      "loss": 0.0157,
      "step": 628500
    },
    {
      "epoch": 23.81,
      "learning_rate": 8.100022710927742e-06,
      "loss": 0.0161,
      "step": 629000
    },
    {
      "epoch": 23.83,
      "learning_rate": 8.090578750141944e-06,
      "loss": 0.0168,
      "step": 629500
    },
    {
      "epoch": 23.85,
      "learning_rate": 8.081115863583028e-06,
      "loss": 0.0152,
      "step": 630000
    },
    {
      "epoch": 23.87,
      "learning_rate": 8.071652977024112e-06,
      "loss": 0.0162,
      "step": 630500
    },
    {
      "epoch": 23.88,
      "learning_rate": 8.062190090465195e-06,
      "loss": 0.0158,
      "step": 631000
    },
    {
      "epoch": 23.9,
      "learning_rate": 8.05272720390628e-06,
      "loss": 0.0164,
      "step": 631500
    },
    {
      "epoch": 23.92,
      "learning_rate": 8.043264317347365e-06,
      "loss": 0.0166,
      "step": 632000
    },
    {
      "epoch": 23.94,
      "learning_rate": 8.033801430788448e-06,
      "loss": 0.0152,
      "step": 632500
    },
    {
      "epoch": 23.96,
      "learning_rate": 8.024338544229534e-06,
      "loss": 0.0164,
      "step": 633000
    },
    {
      "epoch": 23.98,
      "learning_rate": 8.014875657670617e-06,
      "loss": 0.0166,
      "step": 633500
    },
    {
      "epoch": 24.0,
      "learning_rate": 8.0054127711117e-06,
      "loss": 0.0161,
      "step": 634000
    },
    {
      "epoch": 24.0,
      "eval_bleu": 92.2619,
      "eval_gen_len": 16.5368,
      "eval_loss": 0.12691813707351685,
      "eval_runtime": 1199.0212,
      "eval_samples_per_second": 38.087,
      "eval_steps_per_second": 1.191,
      "step": 634056
    },
    {
      "epoch": 24.02,
      "learning_rate": 7.995949884552785e-06,
      "loss": 0.0141,
      "step": 634500
    },
    {
      "epoch": 24.04,
      "learning_rate": 7.986486997993868e-06,
      "loss": 0.013,
      "step": 635000
    },
    {
      "epoch": 24.05,
      "learning_rate": 7.977024111434952e-06,
      "loss": 0.0136,
      "step": 635500
    },
    {
      "epoch": 24.07,
      "learning_rate": 7.967580150649155e-06,
      "loss": 0.0136,
      "step": 636000
    },
    {
      "epoch": 24.09,
      "learning_rate": 7.958117264090239e-06,
      "loss": 0.014,
      "step": 636500
    },
    {
      "epoch": 24.11,
      "learning_rate": 7.948654377531322e-06,
      "loss": 0.0136,
      "step": 637000
    },
    {
      "epoch": 24.13,
      "learning_rate": 7.939191490972406e-06,
      "loss": 0.0139,
      "step": 637500
    },
    {
      "epoch": 24.15,
      "learning_rate": 7.929747530186609e-06,
      "loss": 0.0146,
      "step": 638000
    },
    {
      "epoch": 24.17,
      "learning_rate": 7.92030356940081e-06,
      "loss": 0.0142,
      "step": 638500
    },
    {
      "epoch": 24.19,
      "learning_rate": 7.910840682841895e-06,
      "loss": 0.0135,
      "step": 639000
    },
    {
      "epoch": 24.21,
      "learning_rate": 7.901377796282979e-06,
      "loss": 0.0137,
      "step": 639500
    },
    {
      "epoch": 24.22,
      "learning_rate": 7.891914909724063e-06,
      "loss": 0.0145,
      "step": 640000
    },
    {
      "epoch": 24.24,
      "learning_rate": 7.882452023165146e-06,
      "loss": 0.0142,
      "step": 640500
    },
    {
      "epoch": 24.26,
      "learning_rate": 7.87300806237935e-06,
      "loss": 0.0146,
      "step": 641000
    },
    {
      "epoch": 24.28,
      "learning_rate": 7.863545175820433e-06,
      "loss": 0.0136,
      "step": 641500
    },
    {
      "epoch": 24.3,
      "learning_rate": 7.854082289261517e-06,
      "loss": 0.0141,
      "step": 642000
    },
    {
      "epoch": 24.32,
      "learning_rate": 7.844619402702602e-06,
      "loss": 0.0142,
      "step": 642500
    },
    {
      "epoch": 24.34,
      "learning_rate": 7.835156516143684e-06,
      "loss": 0.0149,
      "step": 643000
    },
    {
      "epoch": 24.36,
      "learning_rate": 7.82569362958477e-06,
      "loss": 0.0149,
      "step": 643500
    },
    {
      "epoch": 24.38,
      "learning_rate": 7.816230743025853e-06,
      "loss": 0.0141,
      "step": 644000
    },
    {
      "epoch": 24.4,
      "learning_rate": 7.806767856466937e-06,
      "loss": 0.0148,
      "step": 644500
    },
    {
      "epoch": 24.41,
      "learning_rate": 7.797304969908022e-06,
      "loss": 0.0143,
      "step": 645000
    },
    {
      "epoch": 24.43,
      "learning_rate": 7.787842083349106e-06,
      "loss": 0.0139,
      "step": 645500
    },
    {
      "epoch": 24.45,
      "learning_rate": 7.77837919679019e-06,
      "loss": 0.0143,
      "step": 646000
    },
    {
      "epoch": 24.47,
      "learning_rate": 7.76893523600439e-06,
      "loss": 0.0139,
      "step": 646500
    },
    {
      "epoch": 24.49,
      "learning_rate": 7.759472349445476e-06,
      "loss": 0.0147,
      "step": 647000
    },
    {
      "epoch": 24.51,
      "learning_rate": 7.75000946288656e-06,
      "loss": 0.0148,
      "step": 647500
    },
    {
      "epoch": 24.53,
      "learning_rate": 7.740546576327644e-06,
      "loss": 0.0156,
      "step": 648000
    },
    {
      "epoch": 24.55,
      "learning_rate": 7.731083689768727e-06,
      "loss": 0.0146,
      "step": 648500
    },
    {
      "epoch": 24.57,
      "learning_rate": 7.721658654756047e-06,
      "loss": 0.0145,
      "step": 649000
    },
    {
      "epoch": 24.58,
      "learning_rate": 7.712195768197131e-06,
      "loss": 0.0147,
      "step": 649500
    },
    {
      "epoch": 24.6,
      "learning_rate": 7.702732881638215e-06,
      "loss": 0.0144,
      "step": 650000
    },
    {
      "epoch": 24.62,
      "learning_rate": 7.6932699950793e-06,
      "loss": 0.0145,
      "step": 650500
    },
    {
      "epoch": 24.64,
      "learning_rate": 7.683807108520384e-06,
      "loss": 0.0144,
      "step": 651000
    },
    {
      "epoch": 24.66,
      "learning_rate": 7.674363147734585e-06,
      "loss": 0.0145,
      "step": 651500
    },
    {
      "epoch": 24.68,
      "learning_rate": 7.664900261175669e-06,
      "loss": 0.0152,
      "step": 652000
    },
    {
      "epoch": 24.7,
      "learning_rate": 7.655437374616752e-06,
      "loss": 0.0146,
      "step": 652500
    },
    {
      "epoch": 24.72,
      "learning_rate": 7.645974488057838e-06,
      "loss": 0.014,
      "step": 653000
    },
    {
      "epoch": 24.74,
      "learning_rate": 7.636511601498922e-06,
      "loss": 0.0151,
      "step": 653500
    },
    {
      "epoch": 24.75,
      "learning_rate": 7.6270676407131235e-06,
      "loss": 0.0144,
      "step": 654000
    },
    {
      "epoch": 24.77,
      "learning_rate": 7.617604754154207e-06,
      "loss": 0.0149,
      "step": 654500
    },
    {
      "epoch": 24.79,
      "learning_rate": 7.608141867595293e-06,
      "loss": 0.0152,
      "step": 655000
    },
    {
      "epoch": 24.81,
      "learning_rate": 7.598678981036376e-06,
      "loss": 0.0147,
      "step": 655500
    },
    {
      "epoch": 24.83,
      "learning_rate": 7.589216094477459e-06,
      "loss": 0.0136,
      "step": 656000
    },
    {
      "epoch": 24.85,
      "learning_rate": 7.579753207918545e-06,
      "loss": 0.0145,
      "step": 656500
    },
    {
      "epoch": 24.87,
      "learning_rate": 7.570309247132746e-06,
      "loss": 0.0149,
      "step": 657000
    },
    {
      "epoch": 24.89,
      "learning_rate": 7.56084636057383e-06,
      "loss": 0.0157,
      "step": 657500
    },
    {
      "epoch": 24.91,
      "learning_rate": 7.551383474014914e-06,
      "loss": 0.0146,
      "step": 658000
    },
    {
      "epoch": 24.93,
      "learning_rate": 7.5419205874559985e-06,
      "loss": 0.0147,
      "step": 658500
    },
    {
      "epoch": 24.94,
      "learning_rate": 7.5324766266702e-06,
      "loss": 0.0153,
      "step": 659000
    },
    {
      "epoch": 24.96,
      "learning_rate": 7.523013740111285e-06,
      "loss": 0.0149,
      "step": 659500
    },
    {
      "epoch": 24.98,
      "learning_rate": 7.513550853552368e-06,
      "loss": 0.0153,
      "step": 660000
    },
    {
      "epoch": 25.0,
      "eval_bleu": 92.2799,
      "eval_gen_len": 16.4882,
      "eval_loss": 0.1280273050069809,
      "eval_runtime": 1203.4771,
      "eval_samples_per_second": 37.946,
      "eval_steps_per_second": 1.187,
      "step": 660475
    },
    {
      "epoch": 25.0,
      "learning_rate": 7.5040879669934515e-06,
      "loss": 0.0147,
      "step": 660500
    },
    {
      "epoch": 25.02,
      "learning_rate": 7.494625080434537e-06,
      "loss": 0.0129,
      "step": 661000
    },
    {
      "epoch": 25.04,
      "learning_rate": 7.485162193875621e-06,
      "loss": 0.0131,
      "step": 661500
    },
    {
      "epoch": 25.06,
      "learning_rate": 7.475699307316705e-06,
      "loss": 0.0123,
      "step": 662000
    },
    {
      "epoch": 25.08,
      "learning_rate": 7.466236420757789e-06,
      "loss": 0.0127,
      "step": 662500
    },
    {
      "epoch": 25.1,
      "learning_rate": 7.456792459971991e-06,
      "loss": 0.0124,
      "step": 663000
    },
    {
      "epoch": 25.11,
      "learning_rate": 7.447348499186192e-06,
      "loss": 0.0122,
      "step": 663500
    },
    {
      "epoch": 25.13,
      "learning_rate": 7.437885612627277e-06,
      "loss": 0.0133,
      "step": 664000
    },
    {
      "epoch": 25.15,
      "learning_rate": 7.42842272606836e-06,
      "loss": 0.0125,
      "step": 664500
    },
    {
      "epoch": 25.17,
      "learning_rate": 7.418959839509444e-06,
      "loss": 0.0132,
      "step": 665000
    },
    {
      "epoch": 25.19,
      "learning_rate": 7.409496952950529e-06,
      "loss": 0.013,
      "step": 665500
    },
    {
      "epoch": 25.21,
      "learning_rate": 7.40005299216473e-06,
      "loss": 0.0126,
      "step": 666000
    },
    {
      "epoch": 25.23,
      "learning_rate": 7.390590105605815e-06,
      "loss": 0.0133,
      "step": 666500
    },
    {
      "epoch": 25.25,
      "learning_rate": 7.381127219046899e-06,
      "loss": 0.0134,
      "step": 667000
    },
    {
      "epoch": 25.27,
      "learning_rate": 7.371664332487982e-06,
      "loss": 0.0129,
      "step": 667500
    },
    {
      "epoch": 25.28,
      "learning_rate": 7.362201445929067e-06,
      "loss": 0.0129,
      "step": 668000
    },
    {
      "epoch": 25.3,
      "learning_rate": 7.3527385593701506e-06,
      "loss": 0.0132,
      "step": 668500
    },
    {
      "epoch": 25.32,
      "learning_rate": 7.343275672811235e-06,
      "loss": 0.0131,
      "step": 669000
    },
    {
      "epoch": 25.34,
      "learning_rate": 7.333812786252319e-06,
      "loss": 0.0133,
      "step": 669500
    },
    {
      "epoch": 25.36,
      "learning_rate": 7.3243498996934025e-06,
      "loss": 0.013,
      "step": 670000
    },
    {
      "epoch": 25.38,
      "learning_rate": 7.314905938907605e-06,
      "loss": 0.0139,
      "step": 670500
    },
    {
      "epoch": 25.4,
      "learning_rate": 7.305461978121807e-06,
      "loss": 0.0131,
      "step": 671000
    },
    {
      "epoch": 25.42,
      "learning_rate": 7.295999091562891e-06,
      "loss": 0.0132,
      "step": 671500
    },
    {
      "epoch": 25.44,
      "learning_rate": 7.286536205003975e-06,
      "loss": 0.013,
      "step": 672000
    },
    {
      "epoch": 25.46,
      "learning_rate": 7.277073318445059e-06,
      "loss": 0.0136,
      "step": 672500
    },
    {
      "epoch": 25.47,
      "learning_rate": 7.267610431886143e-06,
      "loss": 0.0134,
      "step": 673000
    },
    {
      "epoch": 25.49,
      "learning_rate": 7.2581475453272275e-06,
      "loss": 0.013,
      "step": 673500
    },
    {
      "epoch": 25.51,
      "learning_rate": 7.2487035845414286e-06,
      "loss": 0.013,
      "step": 674000
    },
    {
      "epoch": 25.53,
      "learning_rate": 7.239240697982514e-06,
      "loss": 0.0139,
      "step": 674500
    },
    {
      "epoch": 25.55,
      "learning_rate": 7.229777811423598e-06,
      "loss": 0.0128,
      "step": 675000
    },
    {
      "epoch": 25.57,
      "learning_rate": 7.2203149248646805e-06,
      "loss": 0.0137,
      "step": 675500
    },
    {
      "epoch": 25.59,
      "learning_rate": 7.210852038305766e-06,
      "loss": 0.013,
      "step": 676000
    },
    {
      "epoch": 25.61,
      "learning_rate": 7.20138915174685e-06,
      "loss": 0.0127,
      "step": 676500
    },
    {
      "epoch": 25.63,
      "learning_rate": 7.191926265187934e-06,
      "loss": 0.0126,
      "step": 677000
    },
    {
      "epoch": 25.64,
      "learning_rate": 7.182463378629018e-06,
      "loss": 0.014,
      "step": 677500
    },
    {
      "epoch": 25.66,
      "learning_rate": 7.1730004920701016e-06,
      "loss": 0.0133,
      "step": 678000
    },
    {
      "epoch": 25.68,
      "learning_rate": 7.163537605511186e-06,
      "loss": 0.0135,
      "step": 678500
    },
    {
      "epoch": 25.7,
      "learning_rate": 7.154112570498506e-06,
      "loss": 0.014,
      "step": 679000
    },
    {
      "epoch": 25.72,
      "learning_rate": 7.144649683939589e-06,
      "loss": 0.0137,
      "step": 679500
    },
    {
      "epoch": 25.74,
      "learning_rate": 7.135186797380673e-06,
      "loss": 0.0136,
      "step": 680000
    },
    {
      "epoch": 25.76,
      "learning_rate": 7.125723910821758e-06,
      "loss": 0.0135,
      "step": 680500
    },
    {
      "epoch": 25.78,
      "learning_rate": 7.116261024262842e-06,
      "loss": 0.0138,
      "step": 681000
    },
    {
      "epoch": 25.8,
      "learning_rate": 7.106798137703925e-06,
      "loss": 0.0135,
      "step": 681500
    },
    {
      "epoch": 25.81,
      "learning_rate": 7.09733525114501e-06,
      "loss": 0.0133,
      "step": 682000
    },
    {
      "epoch": 25.83,
      "learning_rate": 7.087872364586094e-06,
      "loss": 0.0133,
      "step": 682500
    },
    {
      "epoch": 25.85,
      "learning_rate": 7.0784094780271785e-06,
      "loss": 0.0135,
      "step": 683000
    },
    {
      "epoch": 25.87,
      "learning_rate": 7.068946591468262e-06,
      "loss": 0.013,
      "step": 683500
    },
    {
      "epoch": 25.89,
      "learning_rate": 7.059502630682464e-06,
      "loss": 0.0134,
      "step": 684000
    },
    {
      "epoch": 25.91,
      "learning_rate": 7.050039744123548e-06,
      "loss": 0.0142,
      "step": 684500
    },
    {
      "epoch": 25.93,
      "learning_rate": 7.0405768575646315e-06,
      "loss": 0.0135,
      "step": 685000
    },
    {
      "epoch": 25.95,
      "learning_rate": 7.031113971005716e-06,
      "loss": 0.0135,
      "step": 685500
    },
    {
      "epoch": 25.97,
      "learning_rate": 7.0216510844468e-06,
      "loss": 0.0139,
      "step": 686000
    },
    {
      "epoch": 25.99,
      "learning_rate": 7.012188197887884e-06,
      "loss": 0.0142,
      "step": 686500
    },
    {
      "epoch": 26.0,
      "eval_bleu": 92.3179,
      "eval_gen_len": 16.4738,
      "eval_loss": 0.13104189932346344,
      "eval_runtime": 1208.6929,
      "eval_samples_per_second": 37.782,
      "eval_steps_per_second": 1.181,
      "step": 686894
    },
    {
      "epoch": 26.0,
      "learning_rate": 7.002725311328968e-06,
      "loss": 0.0136,
      "step": 687000
    },
    {
      "epoch": 26.02,
      "learning_rate": 6.993262424770052e-06,
      "loss": 0.0117,
      "step": 687500
    },
    {
      "epoch": 26.04,
      "learning_rate": 6.9838184639842545e-06,
      "loss": 0.0114,
      "step": 688000
    },
    {
      "epoch": 26.06,
      "learning_rate": 6.974355577425338e-06,
      "loss": 0.012,
      "step": 688500
    },
    {
      "epoch": 26.08,
      "learning_rate": 6.964892690866423e-06,
      "loss": 0.0118,
      "step": 689000
    },
    {
      "epoch": 26.1,
      "learning_rate": 6.955448730080624e-06,
      "loss": 0.0118,
      "step": 689500
    },
    {
      "epoch": 26.12,
      "learning_rate": 6.945985843521708e-06,
      "loss": 0.012,
      "step": 690000
    },
    {
      "epoch": 26.14,
      "learning_rate": 6.936522956962792e-06,
      "loss": 0.012,
      "step": 690500
    },
    {
      "epoch": 26.16,
      "learning_rate": 6.927060070403877e-06,
      "loss": 0.0116,
      "step": 691000
    },
    {
      "epoch": 26.17,
      "learning_rate": 6.91759718384496e-06,
      "loss": 0.0122,
      "step": 691500
    },
    {
      "epoch": 26.19,
      "learning_rate": 6.908134297286044e-06,
      "loss": 0.0127,
      "step": 692000
    },
    {
      "epoch": 26.21,
      "learning_rate": 6.8986714107271295e-06,
      "loss": 0.0115,
      "step": 692500
    },
    {
      "epoch": 26.23,
      "learning_rate": 6.889208524168212e-06,
      "loss": 0.0122,
      "step": 693000
    },
    {
      "epoch": 26.25,
      "learning_rate": 6.879764563382415e-06,
      "loss": 0.0125,
      "step": 693500
    },
    {
      "epoch": 26.27,
      "learning_rate": 6.870301676823499e-06,
      "loss": 0.012,
      "step": 694000
    },
    {
      "epoch": 26.29,
      "learning_rate": 6.8608387902645825e-06,
      "loss": 0.0121,
      "step": 694500
    },
    {
      "epoch": 26.31,
      "learning_rate": 6.851375903705667e-06,
      "loss": 0.0113,
      "step": 695000
    },
    {
      "epoch": 26.33,
      "learning_rate": 6.841931942919868e-06,
      "loss": 0.013,
      "step": 695500
    },
    {
      "epoch": 26.34,
      "learning_rate": 6.832469056360953e-06,
      "loss": 0.012,
      "step": 696000
    },
    {
      "epoch": 26.36,
      "learning_rate": 6.823006169802036e-06,
      "loss": 0.0124,
      "step": 696500
    },
    {
      "epoch": 26.38,
      "learning_rate": 6.813543283243122e-06,
      "loss": 0.0126,
      "step": 697000
    },
    {
      "epoch": 26.4,
      "learning_rate": 6.804080396684205e-06,
      "loss": 0.0116,
      "step": 697500
    },
    {
      "epoch": 26.42,
      "learning_rate": 6.794617510125288e-06,
      "loss": 0.0121,
      "step": 698000
    },
    {
      "epoch": 26.44,
      "learning_rate": 6.785154623566374e-06,
      "loss": 0.0126,
      "step": 698500
    },
    {
      "epoch": 26.46,
      "learning_rate": 6.7756917370074575e-06,
      "loss": 0.0114,
      "step": 699000
    },
    {
      "epoch": 26.48,
      "learning_rate": 6.766247776221659e-06,
      "loss": 0.012,
      "step": 699500
    },
    {
      "epoch": 26.5,
      "learning_rate": 6.756784889662743e-06,
      "loss": 0.0122,
      "step": 700000
    },
    {
      "epoch": 26.52,
      "learning_rate": 6.747322003103828e-06,
      "loss": 0.0124,
      "step": 700500
    },
    {
      "epoch": 26.53,
      "learning_rate": 6.737859116544911e-06,
      "loss": 0.012,
      "step": 701000
    },
    {
      "epoch": 26.55,
      "learning_rate": 6.728415155759114e-06,
      "loss": 0.0123,
      "step": 701500
    },
    {
      "epoch": 26.57,
      "learning_rate": 6.718952269200197e-06,
      "loss": 0.0128,
      "step": 702000
    },
    {
      "epoch": 26.59,
      "learning_rate": 6.709489382641281e-06,
      "loss": 0.0129,
      "step": 702500
    },
    {
      "epoch": 26.61,
      "learning_rate": 6.7000454218554835e-06,
      "loss": 0.0123,
      "step": 703000
    },
    {
      "epoch": 26.63,
      "learning_rate": 6.690582535296567e-06,
      "loss": 0.0122,
      "step": 703500
    },
    {
      "epoch": 26.65,
      "learning_rate": 6.681119648737652e-06,
      "loss": 0.0126,
      "step": 704000
    },
    {
      "epoch": 26.67,
      "learning_rate": 6.6716567621787355e-06,
      "loss": 0.0119,
      "step": 704500
    },
    {
      "epoch": 26.69,
      "learning_rate": 6.66219387561982e-06,
      "loss": 0.0123,
      "step": 705000
    },
    {
      "epoch": 26.7,
      "learning_rate": 6.652730989060904e-06,
      "loss": 0.0125,
      "step": 705500
    },
    {
      "epoch": 26.72,
      "learning_rate": 6.643268102501987e-06,
      "loss": 0.0124,
      "step": 706000
    },
    {
      "epoch": 26.74,
      "learning_rate": 6.633824141716189e-06,
      "loss": 0.0124,
      "step": 706500
    },
    {
      "epoch": 26.76,
      "learning_rate": 6.624361255157273e-06,
      "loss": 0.0121,
      "step": 707000
    },
    {
      "epoch": 26.78,
      "learning_rate": 6.6148983685983585e-06,
      "loss": 0.0129,
      "step": 707500
    },
    {
      "epoch": 26.8,
      "learning_rate": 6.605435482039442e-06,
      "loss": 0.0126,
      "step": 708000
    },
    {
      "epoch": 26.82,
      "learning_rate": 6.595991521253644e-06,
      "loss": 0.0127,
      "step": 708500
    },
    {
      "epoch": 26.84,
      "learning_rate": 6.586528634694728e-06,
      "loss": 0.0126,
      "step": 709000
    },
    {
      "epoch": 26.86,
      "learning_rate": 6.5770657481358115e-06,
      "loss": 0.0125,
      "step": 709500
    },
    {
      "epoch": 26.87,
      "learning_rate": 6.567602861576896e-06,
      "loss": 0.0129,
      "step": 710000
    },
    {
      "epoch": 26.89,
      "learning_rate": 6.55813997501798e-06,
      "loss": 0.012,
      "step": 710500
    },
    {
      "epoch": 26.91,
      "learning_rate": 6.548696014232182e-06,
      "loss": 0.0126,
      "step": 711000
    },
    {
      "epoch": 26.93,
      "learning_rate": 6.539233127673265e-06,
      "loss": 0.0121,
      "step": 711500
    },
    {
      "epoch": 26.95,
      "learning_rate": 6.529770241114351e-06,
      "loss": 0.0124,
      "step": 712000
    },
    {
      "epoch": 26.97,
      "learning_rate": 6.520307354555434e-06,
      "loss": 0.0127,
      "step": 712500
    },
    {
      "epoch": 26.99,
      "learning_rate": 6.510844467996517e-06,
      "loss": 0.0126,
      "step": 713000
    },
    {
      "epoch": 27.0,
      "eval_bleu": 92.2702,
      "eval_gen_len": 16.4554,
      "eval_loss": 0.13306637108325958,
      "eval_runtime": 1206.8893,
      "eval_samples_per_second": 37.839,
      "eval_steps_per_second": 1.183,
      "step": 713313
    },
    {
      "epoch": 27.01,
      "learning_rate": 6.501381581437603e-06,
      "loss": 0.0107,
      "step": 713500
    },
    {
      "epoch": 27.03,
      "learning_rate": 6.4919186948786865e-06,
      "loss": 0.0108,
      "step": 714000
    },
    {
      "epoch": 27.04,
      "learning_rate": 6.482455808319771e-06,
      "loss": 0.0112,
      "step": 714500
    },
    {
      "epoch": 27.06,
      "learning_rate": 6.472992921760855e-06,
      "loss": 0.0114,
      "step": 715000
    },
    {
      "epoch": 27.08,
      "learning_rate": 6.463530035201938e-06,
      "loss": 0.011,
      "step": 715500
    },
    {
      "epoch": 27.1,
      "learning_rate": 6.45408607441614e-06,
      "loss": 0.0113,
      "step": 716000
    },
    {
      "epoch": 27.12,
      "learning_rate": 6.444623187857224e-06,
      "loss": 0.0114,
      "step": 716500
    },
    {
      "epoch": 27.14,
      "learning_rate": 6.435179227071426e-06,
      "loss": 0.011,
      "step": 717000
    },
    {
      "epoch": 27.16,
      "learning_rate": 6.42571634051251e-06,
      "loss": 0.0111,
      "step": 717500
    },
    {
      "epoch": 27.18,
      "learning_rate": 6.416253453953595e-06,
      "loss": 0.0115,
      "step": 718000
    },
    {
      "epoch": 27.2,
      "learning_rate": 6.406790567394679e-06,
      "loss": 0.0111,
      "step": 718500
    },
    {
      "epoch": 27.22,
      "learning_rate": 6.397327680835763e-06,
      "loss": 0.0112,
      "step": 719000
    },
    {
      "epoch": 27.23,
      "learning_rate": 6.387864794276847e-06,
      "loss": 0.0109,
      "step": 719500
    },
    {
      "epoch": 27.25,
      "learning_rate": 6.378401907717931e-06,
      "loss": 0.0106,
      "step": 720000
    },
    {
      "epoch": 27.27,
      "learning_rate": 6.368939021159015e-06,
      "loss": 0.0114,
      "step": 720500
    },
    {
      "epoch": 27.29,
      "learning_rate": 6.359476134600099e-06,
      "loss": 0.0113,
      "step": 721000
    },
    {
      "epoch": 27.31,
      "learning_rate": 6.350013248041183e-06,
      "loss": 0.011,
      "step": 721500
    },
    {
      "epoch": 27.33,
      "learning_rate": 6.340550361482267e-06,
      "loss": 0.0114,
      "step": 722000
    },
    {
      "epoch": 27.35,
      "learning_rate": 6.331106400696468e-06,
      "loss": 0.0112,
      "step": 722500
    },
    {
      "epoch": 27.37,
      "learning_rate": 6.321643514137553e-06,
      "loss": 0.0111,
      "step": 723000
    },
    {
      "epoch": 27.39,
      "learning_rate": 6.312180627578637e-06,
      "loss": 0.0113,
      "step": 723500
    },
    {
      "epoch": 27.4,
      "learning_rate": 6.302736666792839e-06,
      "loss": 0.0112,
      "step": 724000
    },
    {
      "epoch": 27.42,
      "learning_rate": 6.293273780233923e-06,
      "loss": 0.0113,
      "step": 724500
    },
    {
      "epoch": 27.44,
      "learning_rate": 6.283810893675008e-06,
      "loss": 0.0117,
      "step": 725000
    },
    {
      "epoch": 27.46,
      "learning_rate": 6.274348007116091e-06,
      "loss": 0.0112,
      "step": 725500
    },
    {
      "epoch": 27.48,
      "learning_rate": 6.264885120557175e-06,
      "loss": 0.0108,
      "step": 726000
    },
    {
      "epoch": 27.5,
      "learning_rate": 6.25542223399826e-06,
      "loss": 0.0117,
      "step": 726500
    },
    {
      "epoch": 27.52,
      "learning_rate": 6.245959347439343e-06,
      "loss": 0.0112,
      "step": 727000
    },
    {
      "epoch": 27.54,
      "learning_rate": 6.236496460880428e-06,
      "loss": 0.0114,
      "step": 727500
    },
    {
      "epoch": 27.56,
      "learning_rate": 6.2270335743215116e-06,
      "loss": 0.0115,
      "step": 728000
    },
    {
      "epoch": 27.57,
      "learning_rate": 6.217570687762595e-06,
      "loss": 0.0113,
      "step": 728500
    },
    {
      "epoch": 27.59,
      "learning_rate": 6.20810780120368e-06,
      "loss": 0.012,
      "step": 729000
    },
    {
      "epoch": 27.61,
      "learning_rate": 6.1986449146447635e-06,
      "loss": 0.0105,
      "step": 729500
    },
    {
      "epoch": 27.63,
      "learning_rate": 6.189200953858966e-06,
      "loss": 0.0126,
      "step": 730000
    },
    {
      "epoch": 27.65,
      "learning_rate": 6.179756993073167e-06,
      "loss": 0.0116,
      "step": 730500
    },
    {
      "epoch": 27.67,
      "learning_rate": 6.170313032287369e-06,
      "loss": 0.0114,
      "step": 731000
    },
    {
      "epoch": 27.69,
      "learning_rate": 6.160850145728453e-06,
      "loss": 0.0113,
      "step": 731500
    },
    {
      "epoch": 27.71,
      "learning_rate": 6.151387259169538e-06,
      "loss": 0.012,
      "step": 732000
    },
    {
      "epoch": 27.73,
      "learning_rate": 6.141924372610621e-06,
      "loss": 0.011,
      "step": 732500
    },
    {
      "epoch": 27.75,
      "learning_rate": 6.132461486051706e-06,
      "loss": 0.012,
      "step": 733000
    },
    {
      "epoch": 27.76,
      "learning_rate": 6.1229985994927896e-06,
      "loss": 0.0115,
      "step": 733500
    },
    {
      "epoch": 27.78,
      "learning_rate": 6.113535712933873e-06,
      "loss": 0.0113,
      "step": 734000
    },
    {
      "epoch": 27.8,
      "learning_rate": 6.104072826374959e-06,
      "loss": 0.0114,
      "step": 734500
    },
    {
      "epoch": 27.82,
      "learning_rate": 6.0946099398160415e-06,
      "loss": 0.0117,
      "step": 735000
    },
    {
      "epoch": 27.84,
      "learning_rate": 6.085165979030244e-06,
      "loss": 0.0117,
      "step": 735500
    },
    {
      "epoch": 27.86,
      "learning_rate": 6.075703092471328e-06,
      "loss": 0.0112,
      "step": 736000
    },
    {
      "epoch": 27.88,
      "learning_rate": 6.066240205912412e-06,
      "loss": 0.0121,
      "step": 736500
    },
    {
      "epoch": 27.9,
      "learning_rate": 6.056777319353496e-06,
      "loss": 0.0115,
      "step": 737000
    },
    {
      "epoch": 27.92,
      "learning_rate": 6.04731443279458e-06,
      "loss": 0.0118,
      "step": 737500
    },
    {
      "epoch": 27.93,
      "learning_rate": 6.0378515462356645e-06,
      "loss": 0.0118,
      "step": 738000
    },
    {
      "epoch": 27.95,
      "learning_rate": 6.028407585449866e-06,
      "loss": 0.0117,
      "step": 738500
    },
    {
      "epoch": 27.97,
      "learning_rate": 6.01894469889095e-06,
      "loss": 0.0116,
      "step": 739000
    },
    {
      "epoch": 27.99,
      "learning_rate": 6.009481812332034e-06,
      "loss": 0.012,
      "step": 739500
    },
    {
      "epoch": 28.0,
      "eval_bleu": 92.2664,
      "eval_gen_len": 16.4506,
      "eval_loss": 0.1342269480228424,
      "eval_runtime": 1200.216,
      "eval_samples_per_second": 38.049,
      "eval_steps_per_second": 1.19,
      "step": 739732
    },
    {
      "epoch": 28.01,
      "learning_rate": 6.0000189257731176e-06,
      "loss": 0.0107,
      "step": 740000
    },
    {
      "epoch": 28.03,
      "learning_rate": 5.990556039214203e-06,
      "loss": 0.0101,
      "step": 740500
    },
    {
      "epoch": 28.05,
      "learning_rate": 5.981112078428404e-06,
      "loss": 0.0104,
      "step": 741000
    },
    {
      "epoch": 28.07,
      "learning_rate": 5.971649191869489e-06,
      "loss": 0.0101,
      "step": 741500
    },
    {
      "epoch": 28.09,
      "learning_rate": 5.962186305310572e-06,
      "loss": 0.0102,
      "step": 742000
    },
    {
      "epoch": 28.1,
      "learning_rate": 5.952723418751657e-06,
      "loss": 0.0107,
      "step": 742500
    },
    {
      "epoch": 28.12,
      "learning_rate": 5.9432605321927405e-06,
      "loss": 0.0103,
      "step": 743000
    },
    {
      "epoch": 28.14,
      "learning_rate": 5.933797645633824e-06,
      "loss": 0.0102,
      "step": 743500
    },
    {
      "epoch": 28.16,
      "learning_rate": 5.924334759074909e-06,
      "loss": 0.0104,
      "step": 744000
    },
    {
      "epoch": 28.18,
      "learning_rate": 5.9148718725159925e-06,
      "loss": 0.0101,
      "step": 744500
    },
    {
      "epoch": 28.2,
      "learning_rate": 5.905408985957077e-06,
      "loss": 0.0102,
      "step": 745000
    },
    {
      "epoch": 28.22,
      "learning_rate": 5.895946099398161e-06,
      "loss": 0.0101,
      "step": 745500
    },
    {
      "epoch": 28.24,
      "learning_rate": 5.8865021386123635e-06,
      "loss": 0.0103,
      "step": 746000
    },
    {
      "epoch": 28.26,
      "learning_rate": 5.877039252053447e-06,
      "loss": 0.0105,
      "step": 746500
    },
    {
      "epoch": 28.28,
      "learning_rate": 5.867576365494531e-06,
      "loss": 0.0104,
      "step": 747000
    },
    {
      "epoch": 28.29,
      "learning_rate": 5.8581134789356155e-06,
      "loss": 0.0102,
      "step": 747500
    },
    {
      "epoch": 28.31,
      "learning_rate": 5.848650592376699e-06,
      "loss": 0.0099,
      "step": 748000
    },
    {
      "epoch": 28.33,
      "learning_rate": 5.839187705817783e-06,
      "loss": 0.0107,
      "step": 748500
    },
    {
      "epoch": 28.35,
      "learning_rate": 5.829743745031985e-06,
      "loss": 0.0108,
      "step": 749000
    },
    {
      "epoch": 28.37,
      "learning_rate": 5.8202808584730685e-06,
      "loss": 0.0106,
      "step": 749500
    },
    {
      "epoch": 28.39,
      "learning_rate": 5.810855823460389e-06,
      "loss": 0.01,
      "step": 750000
    },
    {
      "epoch": 28.41,
      "learning_rate": 5.801392936901473e-06,
      "loss": 0.0105,
      "step": 750500
    },
    {
      "epoch": 28.43,
      "learning_rate": 5.791930050342557e-06,
      "loss": 0.0107,
      "step": 751000
    },
    {
      "epoch": 28.45,
      "learning_rate": 5.782467163783641e-06,
      "loss": 0.0108,
      "step": 751500
    },
    {
      "epoch": 28.46,
      "learning_rate": 5.773004277224725e-06,
      "loss": 0.0106,
      "step": 752000
    },
    {
      "epoch": 28.48,
      "learning_rate": 5.763541390665809e-06,
      "loss": 0.0107,
      "step": 752500
    },
    {
      "epoch": 28.5,
      "learning_rate": 5.7540785041068935e-06,
      "loss": 0.0107,
      "step": 753000
    },
    {
      "epoch": 28.52,
      "learning_rate": 5.744615617547977e-06,
      "loss": 0.0106,
      "step": 753500
    },
    {
      "epoch": 28.54,
      "learning_rate": 5.73517165676218e-06,
      "loss": 0.0109,
      "step": 754000
    },
    {
      "epoch": 28.56,
      "learning_rate": 5.725708770203263e-06,
      "loss": 0.0106,
      "step": 754500
    },
    {
      "epoch": 28.58,
      "learning_rate": 5.7162458836443465e-06,
      "loss": 0.0109,
      "step": 755000
    },
    {
      "epoch": 28.6,
      "learning_rate": 5.706782997085432e-06,
      "loss": 0.0103,
      "step": 755500
    },
    {
      "epoch": 28.62,
      "learning_rate": 5.697320110526516e-06,
      "loss": 0.0109,
      "step": 756000
    },
    {
      "epoch": 28.63,
      "learning_rate": 5.6878572239676e-06,
      "loss": 0.0102,
      "step": 756500
    },
    {
      "epoch": 28.65,
      "learning_rate": 5.678394337408684e-06,
      "loss": 0.0107,
      "step": 757000
    },
    {
      "epoch": 28.67,
      "learning_rate": 5.668931450849768e-06,
      "loss": 0.0113,
      "step": 757500
    },
    {
      "epoch": 28.69,
      "learning_rate": 5.659468564290852e-06,
      "loss": 0.011,
      "step": 758000
    },
    {
      "epoch": 28.71,
      "learning_rate": 5.650005677731936e-06,
      "loss": 0.0106,
      "step": 758500
    },
    {
      "epoch": 28.73,
      "learning_rate": 5.640561716946138e-06,
      "loss": 0.0105,
      "step": 759000
    },
    {
      "epoch": 28.75,
      "learning_rate": 5.6310988303872215e-06,
      "loss": 0.0106,
      "step": 759500
    },
    {
      "epoch": 28.77,
      "learning_rate": 5.621635943828306e-06,
      "loss": 0.0107,
      "step": 760000
    },
    {
      "epoch": 28.79,
      "learning_rate": 5.612191983042507e-06,
      "loss": 0.0108,
      "step": 760500
    },
    {
      "epoch": 28.81,
      "learning_rate": 5.6027290964835925e-06,
      "loss": 0.0103,
      "step": 761000
    },
    {
      "epoch": 28.82,
      "learning_rate": 5.593266209924676e-06,
      "loss": 0.0102,
      "step": 761500
    },
    {
      "epoch": 28.84,
      "learning_rate": 5.58380332336576e-06,
      "loss": 0.0104,
      "step": 762000
    },
    {
      "epoch": 28.86,
      "learning_rate": 5.5743404368068445e-06,
      "loss": 0.0113,
      "step": 762500
    },
    {
      "epoch": 28.88,
      "learning_rate": 5.564877550247928e-06,
      "loss": 0.011,
      "step": 763000
    },
    {
      "epoch": 28.9,
      "learning_rate": 5.555414663689012e-06,
      "loss": 0.0108,
      "step": 763500
    },
    {
      "epoch": 28.92,
      "learning_rate": 5.5459517771300964e-06,
      "loss": 0.011,
      "step": 764000
    },
    {
      "epoch": 28.94,
      "learning_rate": 5.536526742117417e-06,
      "loss": 0.0106,
      "step": 764500
    },
    {
      "epoch": 28.96,
      "learning_rate": 5.527082781331618e-06,
      "loss": 0.0107,
      "step": 765000
    },
    {
      "epoch": 28.98,
      "learning_rate": 5.517619894772702e-06,
      "loss": 0.011,
      "step": 765500
    },
    {
      "epoch": 28.99,
      "learning_rate": 5.508157008213786e-06,
      "loss": 0.0112,
      "step": 766000
    },
    {
      "epoch": 29.0,
      "eval_bleu": 92.3004,
      "eval_gen_len": 16.4833,
      "eval_loss": 0.13662399351596832,
      "eval_runtime": 1207.6101,
      "eval_samples_per_second": 37.816,
      "eval_steps_per_second": 1.183,
      "step": 766151
    },
    {
      "epoch": 29.01,
      "learning_rate": 5.4986941216548705e-06,
      "loss": 0.0097,
      "step": 766500
    },
    {
      "epoch": 29.03,
      "learning_rate": 5.489231235095954e-06,
      "loss": 0.0093,
      "step": 767000
    },
    {
      "epoch": 29.05,
      "learning_rate": 5.479768348537038e-06,
      "loss": 0.0096,
      "step": 767500
    },
    {
      "epoch": 29.07,
      "learning_rate": 5.4703054619781225e-06,
      "loss": 0.0095,
      "step": 768000
    },
    {
      "epoch": 29.09,
      "learning_rate": 5.460842575419206e-06,
      "loss": 0.0093,
      "step": 768500
    },
    {
      "epoch": 29.11,
      "learning_rate": 5.45137968886029e-06,
      "loss": 0.0098,
      "step": 769000
    },
    {
      "epoch": 29.13,
      "learning_rate": 5.4419168023013744e-06,
      "loss": 0.0094,
      "step": 769500
    },
    {
      "epoch": 29.15,
      "learning_rate": 5.432453915742458e-06,
      "loss": 0.0094,
      "step": 770000
    },
    {
      "epoch": 29.16,
      "learning_rate": 5.422991029183543e-06,
      "loss": 0.0098,
      "step": 770500
    },
    {
      "epoch": 29.18,
      "learning_rate": 5.413547068397745e-06,
      "loss": 0.0098,
      "step": 771000
    },
    {
      "epoch": 29.2,
      "learning_rate": 5.404084181838829e-06,
      "loss": 0.0098,
      "step": 771500
    },
    {
      "epoch": 29.22,
      "learning_rate": 5.394621295279913e-06,
      "loss": 0.0099,
      "step": 772000
    },
    {
      "epoch": 29.24,
      "learning_rate": 5.385158408720997e-06,
      "loss": 0.0094,
      "step": 772500
    },
    {
      "epoch": 29.26,
      "learning_rate": 5.3757144479351985e-06,
      "loss": 0.0097,
      "step": 773000
    },
    {
      "epoch": 29.28,
      "learning_rate": 5.366251561376282e-06,
      "loss": 0.0096,
      "step": 773500
    },
    {
      "epoch": 29.3,
      "learning_rate": 5.356788674817367e-06,
      "loss": 0.0096,
      "step": 774000
    },
    {
      "epoch": 29.32,
      "learning_rate": 5.3473257882584505e-06,
      "loss": 0.0097,
      "step": 774500
    },
    {
      "epoch": 29.33,
      "learning_rate": 5.337881827472653e-06,
      "loss": 0.0101,
      "step": 775000
    },
    {
      "epoch": 29.35,
      "learning_rate": 5.328418940913737e-06,
      "loss": 0.0098,
      "step": 775500
    },
    {
      "epoch": 29.37,
      "learning_rate": 5.3189560543548215e-06,
      "loss": 0.0099,
      "step": 776000
    },
    {
      "epoch": 29.39,
      "learning_rate": 5.309493167795905e-06,
      "loss": 0.0101,
      "step": 776500
    },
    {
      "epoch": 29.41,
      "learning_rate": 5.300030281236989e-06,
      "loss": 0.0099,
      "step": 777000
    },
    {
      "epoch": 29.43,
      "learning_rate": 5.2905673946780735e-06,
      "loss": 0.0099,
      "step": 777500
    },
    {
      "epoch": 29.45,
      "learning_rate": 5.281123433892275e-06,
      "loss": 0.01,
      "step": 778000
    },
    {
      "epoch": 29.47,
      "learning_rate": 5.271660547333359e-06,
      "loss": 0.0099,
      "step": 778500
    },
    {
      "epoch": 29.49,
      "learning_rate": 5.262197660774443e-06,
      "loss": 0.0101,
      "step": 779000
    },
    {
      "epoch": 29.51,
      "learning_rate": 5.252734774215527e-06,
      "loss": 0.0097,
      "step": 779500
    },
    {
      "epoch": 29.52,
      "learning_rate": 5.243271887656611e-06,
      "loss": 0.0097,
      "step": 780000
    },
    {
      "epoch": 29.54,
      "learning_rate": 5.233827926870814e-06,
      "loss": 0.0099,
      "step": 780500
    },
    {
      "epoch": 29.56,
      "learning_rate": 5.224365040311898e-06,
      "loss": 0.0101,
      "step": 781000
    },
    {
      "epoch": 29.58,
      "learning_rate": 5.214902153752981e-06,
      "loss": 0.0098,
      "step": 781500
    },
    {
      "epoch": 29.6,
      "learning_rate": 5.205439267194066e-06,
      "loss": 0.0101,
      "step": 782000
    },
    {
      "epoch": 29.62,
      "learning_rate": 5.1959763806351495e-06,
      "loss": 0.0095,
      "step": 782500
    },
    {
      "epoch": 29.64,
      "learning_rate": 5.186513494076233e-06,
      "loss": 0.0101,
      "step": 783000
    },
    {
      "epoch": 29.66,
      "learning_rate": 5.177050607517318e-06,
      "loss": 0.0097,
      "step": 783500
    },
    {
      "epoch": 29.68,
      "learning_rate": 5.167606646731519e-06,
      "loss": 0.0099,
      "step": 784000
    },
    {
      "epoch": 29.69,
      "learning_rate": 5.1581437601726034e-06,
      "loss": 0.0099,
      "step": 784500
    },
    {
      "epoch": 29.71,
      "learning_rate": 5.148680873613687e-06,
      "loss": 0.0103,
      "step": 785000
    },
    {
      "epoch": 29.73,
      "learning_rate": 5.139217987054772e-06,
      "loss": 0.0101,
      "step": 785500
    },
    {
      "epoch": 29.75,
      "learning_rate": 5.129755100495855e-06,
      "loss": 0.0104,
      "step": 786000
    },
    {
      "epoch": 29.77,
      "learning_rate": 5.120292213936939e-06,
      "loss": 0.0097,
      "step": 786500
    },
    {
      "epoch": 29.79,
      "learning_rate": 5.110848253151142e-06,
      "loss": 0.0097,
      "step": 787000
    },
    {
      "epoch": 29.81,
      "learning_rate": 5.101385366592226e-06,
      "loss": 0.0102,
      "step": 787500
    },
    {
      "epoch": 29.83,
      "learning_rate": 5.09192248003331e-06,
      "loss": 0.0102,
      "step": 788000
    },
    {
      "epoch": 29.85,
      "learning_rate": 5.082459593474394e-06,
      "loss": 0.01,
      "step": 788500
    },
    {
      "epoch": 29.86,
      "learning_rate": 5.072996706915478e-06,
      "loss": 0.0098,
      "step": 789000
    },
    {
      "epoch": 29.88,
      "learning_rate": 5.063533820356562e-06,
      "loss": 0.0104,
      "step": 789500
    },
    {
      "epoch": 29.9,
      "learning_rate": 5.054070933797646e-06,
      "loss": 0.01,
      "step": 790000
    },
    {
      "epoch": 29.92,
      "learning_rate": 5.04460804723873e-06,
      "loss": 0.0103,
      "step": 790500
    },
    {
      "epoch": 29.94,
      "learning_rate": 5.035145160679814e-06,
      "loss": 0.0099,
      "step": 791000
    },
    {
      "epoch": 29.96,
      "learning_rate": 5.025682274120898e-06,
      "loss": 0.0096,
      "step": 791500
    },
    {
      "epoch": 29.98,
      "learning_rate": 5.016219387561982e-06,
      "loss": 0.0095,
      "step": 792000
    },
    {
      "epoch": 30.0,
      "learning_rate": 5.006756501003066e-06,
      "loss": 0.0101,
      "step": 792500
    },
    {
      "epoch": 30.0,
      "eval_bleu": 92.2957,
      "eval_gen_len": 16.4943,
      "eval_loss": 0.13711272180080414,
      "eval_runtime": 1176.4369,
      "eval_samples_per_second": 38.818,
      "eval_steps_per_second": 1.214,
      "step": 792570
    },
    {
      "epoch": 30.02,
      "learning_rate": 4.997331465990386e-06,
      "loss": 0.0092,
      "step": 793000
    },
    {
      "epoch": 30.04,
      "learning_rate": 4.98786857943147e-06,
      "loss": 0.0088,
      "step": 793500
    },
    {
      "epoch": 30.05,
      "learning_rate": 4.9784056928725544e-06,
      "loss": 0.0087,
      "step": 794000
    },
    {
      "epoch": 30.07,
      "learning_rate": 4.968942806313639e-06,
      "loss": 0.0087,
      "step": 794500
    },
    {
      "epoch": 30.09,
      "learning_rate": 4.95949884552784e-06,
      "loss": 0.0088,
      "step": 795000
    },
    {
      "epoch": 30.11,
      "learning_rate": 4.950035958968925e-06,
      "loss": 0.0093,
      "step": 795500
    },
    {
      "epoch": 30.13,
      "learning_rate": 4.9405919981831266e-06,
      "loss": 0.0089,
      "step": 796000
    },
    {
      "epoch": 30.15,
      "learning_rate": 4.93112911162421e-06,
      "loss": 0.0085,
      "step": 796500
    },
    {
      "epoch": 30.17,
      "learning_rate": 4.921666225065294e-06,
      "loss": 0.009,
      "step": 797000
    },
    {
      "epoch": 30.19,
      "learning_rate": 4.912222264279497e-06,
      "loss": 0.0091,
      "step": 797500
    },
    {
      "epoch": 30.21,
      "learning_rate": 4.9027593777205805e-06,
      "loss": 0.0087,
      "step": 798000
    },
    {
      "epoch": 30.22,
      "learning_rate": 4.893296491161664e-06,
      "loss": 0.0093,
      "step": 798500
    },
    {
      "epoch": 30.24,
      "learning_rate": 4.883833604602749e-06,
      "loss": 0.0091,
      "step": 799000
    },
    {
      "epoch": 30.26,
      "learning_rate": 4.874370718043832e-06,
      "loss": 0.0088,
      "step": 799500
    },
    {
      "epoch": 30.28,
      "learning_rate": 4.864907831484916e-06,
      "loss": 0.0086,
      "step": 800000
    },
    {
      "epoch": 30.3,
      "learning_rate": 4.855444944926001e-06,
      "loss": 0.0091,
      "step": 800500
    },
    {
      "epoch": 30.32,
      "learning_rate": 4.846000984140203e-06,
      "loss": 0.0093,
      "step": 801000
    },
    {
      "epoch": 30.34,
      "learning_rate": 4.836538097581286e-06,
      "loss": 0.0093,
      "step": 801500
    },
    {
      "epoch": 30.36,
      "learning_rate": 4.827075211022371e-06,
      "loss": 0.009,
      "step": 802000
    },
    {
      "epoch": 30.38,
      "learning_rate": 4.8176123244634546e-06,
      "loss": 0.0096,
      "step": 802500
    },
    {
      "epoch": 30.39,
      "learning_rate": 4.8081683636776565e-06,
      "loss": 0.0092,
      "step": 803000
    },
    {
      "epoch": 30.41,
      "learning_rate": 4.798705477118741e-06,
      "loss": 0.0091,
      "step": 803500
    },
    {
      "epoch": 30.43,
      "learning_rate": 4.789242590559825e-06,
      "loss": 0.0089,
      "step": 804000
    },
    {
      "epoch": 30.45,
      "learning_rate": 4.7797797040009085e-06,
      "loss": 0.009,
      "step": 804500
    },
    {
      "epoch": 30.47,
      "learning_rate": 4.770316817441993e-06,
      "loss": 0.0093,
      "step": 805000
    },
    {
      "epoch": 30.49,
      "learning_rate": 4.760853930883077e-06,
      "loss": 0.0093,
      "step": 805500
    },
    {
      "epoch": 30.51,
      "learning_rate": 4.751391044324161e-06,
      "loss": 0.0091,
      "step": 806000
    },
    {
      "epoch": 30.53,
      "learning_rate": 4.741928157765245e-06,
      "loss": 0.009,
      "step": 806500
    },
    {
      "epoch": 30.55,
      "learning_rate": 4.732484196979447e-06,
      "loss": 0.0091,
      "step": 807000
    },
    {
      "epoch": 30.57,
      "learning_rate": 4.723021310420531e-06,
      "loss": 0.0094,
      "step": 807500
    },
    {
      "epoch": 30.58,
      "learning_rate": 4.713558423861615e-06,
      "loss": 0.0095,
      "step": 808000
    },
    {
      "epoch": 30.6,
      "learning_rate": 4.704095537302699e-06,
      "loss": 0.0094,
      "step": 808500
    },
    {
      "epoch": 30.62,
      "learning_rate": 4.694632650743783e-06,
      "loss": 0.0092,
      "step": 809000
    },
    {
      "epoch": 30.64,
      "learning_rate": 4.685169764184868e-06,
      "loss": 0.0092,
      "step": 809500
    },
    {
      "epoch": 30.66,
      "learning_rate": 4.675706877625951e-06,
      "loss": 0.0095,
      "step": 810000
    },
    {
      "epoch": 30.68,
      "learning_rate": 4.666243991067035e-06,
      "loss": 0.0096,
      "step": 810500
    },
    {
      "epoch": 30.7,
      "learning_rate": 4.65678110450812e-06,
      "loss": 0.0095,
      "step": 811000
    },
    {
      "epoch": 30.72,
      "learning_rate": 4.647318217949204e-06,
      "loss": 0.0092,
      "step": 811500
    },
    {
      "epoch": 30.74,
      "learning_rate": 4.637855331390287e-06,
      "loss": 0.009,
      "step": 812000
    },
    {
      "epoch": 30.75,
      "learning_rate": 4.628392444831372e-06,
      "loss": 0.0096,
      "step": 812500
    },
    {
      "epoch": 30.77,
      "learning_rate": 4.6189295582724556e-06,
      "loss": 0.009,
      "step": 813000
    },
    {
      "epoch": 30.79,
      "learning_rate": 4.60946667171354e-06,
      "loss": 0.0093,
      "step": 813500
    },
    {
      "epoch": 30.81,
      "learning_rate": 4.600022710927742e-06,
      "loss": 0.0089,
      "step": 814000
    },
    {
      "epoch": 30.83,
      "learning_rate": 4.590578750141943e-06,
      "loss": 0.0096,
      "step": 814500
    },
    {
      "epoch": 30.85,
      "learning_rate": 4.581115863583028e-06,
      "loss": 0.0091,
      "step": 815000
    },
    {
      "epoch": 30.87,
      "learning_rate": 4.57167190279723e-06,
      "loss": 0.0091,
      "step": 815500
    },
    {
      "epoch": 30.89,
      "learning_rate": 4.562209016238313e-06,
      "loss": 0.0093,
      "step": 816000
    },
    {
      "epoch": 30.91,
      "learning_rate": 4.552746129679398e-06,
      "loss": 0.0089,
      "step": 816500
    },
    {
      "epoch": 30.92,
      "learning_rate": 4.5432832431204825e-06,
      "loss": 0.0095,
      "step": 817000
    },
    {
      "epoch": 30.94,
      "learning_rate": 4.533820356561565e-06,
      "loss": 0.0094,
      "step": 817500
    },
    {
      "epoch": 30.96,
      "learning_rate": 4.52435747000265e-06,
      "loss": 0.0096,
      "step": 818000
    },
    {
      "epoch": 30.98,
      "learning_rate": 4.514894583443734e-06,
      "loss": 0.0095,
      "step": 818500
    },
    {
      "epoch": 31.0,
      "eval_bleu": 92.3047,
      "eval_gen_len": 16.4694,
      "eval_loss": 0.13931500911712646,
      "eval_runtime": 1172.6751,
      "eval_samples_per_second": 38.943,
      "eval_steps_per_second": 1.218,
      "step": 818989
    },
    {
      "epoch": 31.0,
      "learning_rate": 4.505431696884818e-06,
      "loss": 0.0097,
      "step": 819000
    },
    {
      "epoch": 31.02,
      "learning_rate": 4.495968810325902e-06,
      "loss": 0.0083,
      "step": 819500
    },
    {
      "epoch": 31.04,
      "learning_rate": 4.486524849540105e-06,
      "loss": 0.0081,
      "step": 820000
    },
    {
      "epoch": 31.06,
      "learning_rate": 4.477061962981188e-06,
      "loss": 0.0083,
      "step": 820500
    },
    {
      "epoch": 31.08,
      "learning_rate": 4.467599076422272e-06,
      "loss": 0.0085,
      "step": 821000
    },
    {
      "epoch": 31.1,
      "learning_rate": 4.4581361898633566e-06,
      "loss": 0.0083,
      "step": 821500
    },
    {
      "epoch": 31.11,
      "learning_rate": 4.44867330330444e-06,
      "loss": 0.0086,
      "step": 822000
    },
    {
      "epoch": 31.13,
      "learning_rate": 4.439210416745525e-06,
      "loss": 0.0084,
      "step": 822500
    },
    {
      "epoch": 31.15,
      "learning_rate": 4.4297475301866085e-06,
      "loss": 0.0083,
      "step": 823000
    },
    {
      "epoch": 31.17,
      "learning_rate": 4.420284643627692e-06,
      "loss": 0.0084,
      "step": 823500
    },
    {
      "epoch": 31.19,
      "learning_rate": 4.410840682841894e-06,
      "loss": 0.0086,
      "step": 824000
    },
    {
      "epoch": 31.21,
      "learning_rate": 4.401377796282979e-06,
      "loss": 0.0085,
      "step": 824500
    },
    {
      "epoch": 31.23,
      "learning_rate": 4.391914909724062e-06,
      "loss": 0.0089,
      "step": 825000
    },
    {
      "epoch": 31.25,
      "learning_rate": 4.382452023165147e-06,
      "loss": 0.0087,
      "step": 825500
    },
    {
      "epoch": 31.27,
      "learning_rate": 4.372989136606231e-06,
      "loss": 0.0083,
      "step": 826000
    },
    {
      "epoch": 31.28,
      "learning_rate": 4.363545175820433e-06,
      "loss": 0.0082,
      "step": 826500
    },
    {
      "epoch": 31.3,
      "learning_rate": 4.354082289261516e-06,
      "loss": 0.0081,
      "step": 827000
    },
    {
      "epoch": 31.32,
      "learning_rate": 4.344619402702601e-06,
      "loss": 0.0087,
      "step": 827500
    },
    {
      "epoch": 31.34,
      "learning_rate": 4.3351565161436846e-06,
      "loss": 0.0081,
      "step": 828000
    },
    {
      "epoch": 31.36,
      "learning_rate": 4.325693629584769e-06,
      "loss": 0.0085,
      "step": 828500
    },
    {
      "epoch": 31.38,
      "learning_rate": 4.316249668798971e-06,
      "loss": 0.0085,
      "step": 829000
    },
    {
      "epoch": 31.4,
      "learning_rate": 4.306786782240055e-06,
      "loss": 0.0085,
      "step": 829500
    },
    {
      "epoch": 31.42,
      "learning_rate": 4.297342821454257e-06,
      "loss": 0.0094,
      "step": 830000
    },
    {
      "epoch": 31.44,
      "learning_rate": 4.287879934895341e-06,
      "loss": 0.0089,
      "step": 830500
    },
    {
      "epoch": 31.45,
      "learning_rate": 4.278417048336425e-06,
      "loss": 0.0089,
      "step": 831000
    },
    {
      "epoch": 31.47,
      "learning_rate": 4.268954161777509e-06,
      "loss": 0.0086,
      "step": 831500
    },
    {
      "epoch": 31.49,
      "learning_rate": 4.259491275218593e-06,
      "loss": 0.0085,
      "step": 832000
    },
    {
      "epoch": 31.51,
      "learning_rate": 4.250047314432794e-06,
      "loss": 0.0083,
      "step": 832500
    },
    {
      "epoch": 31.53,
      "learning_rate": 4.240584427873879e-06,
      "loss": 0.0092,
      "step": 833000
    },
    {
      "epoch": 31.55,
      "learning_rate": 4.231121541314963e-06,
      "loss": 0.0088,
      "step": 833500
    },
    {
      "epoch": 31.57,
      "learning_rate": 4.221658654756047e-06,
      "loss": 0.0086,
      "step": 834000
    },
    {
      "epoch": 31.59,
      "learning_rate": 4.212195768197131e-06,
      "loss": 0.008,
      "step": 834500
    },
    {
      "epoch": 31.61,
      "learning_rate": 4.202732881638215e-06,
      "loss": 0.009,
      "step": 835000
    },
    {
      "epoch": 31.62,
      "learning_rate": 4.193269995079299e-06,
      "loss": 0.0083,
      "step": 835500
    },
    {
      "epoch": 31.64,
      "learning_rate": 4.183844960066619e-06,
      "loss": 0.0088,
      "step": 836000
    },
    {
      "epoch": 31.66,
      "learning_rate": 4.174382073507703e-06,
      "loss": 0.0085,
      "step": 836500
    },
    {
      "epoch": 31.68,
      "learning_rate": 4.164919186948787e-06,
      "loss": 0.0084,
      "step": 837000
    },
    {
      "epoch": 31.7,
      "learning_rate": 4.1554752261629895e-06,
      "loss": 0.0082,
      "step": 837500
    },
    {
      "epoch": 31.72,
      "learning_rate": 4.146012339604073e-06,
      "loss": 0.0087,
      "step": 838000
    },
    {
      "epoch": 31.74,
      "learning_rate": 4.136549453045157e-06,
      "loss": 0.0085,
      "step": 838500
    },
    {
      "epoch": 31.76,
      "learning_rate": 4.127086566486241e-06,
      "loss": 0.0085,
      "step": 839000
    },
    {
      "epoch": 31.78,
      "learning_rate": 4.117623679927326e-06,
      "loss": 0.0089,
      "step": 839500
    },
    {
      "epoch": 31.8,
      "learning_rate": 4.108160793368409e-06,
      "loss": 0.0089,
      "step": 840000
    },
    {
      "epoch": 31.81,
      "learning_rate": 4.098697906809493e-06,
      "loss": 0.0083,
      "step": 840500
    },
    {
      "epoch": 31.83,
      "learning_rate": 4.089235020250578e-06,
      "loss": 0.0087,
      "step": 841000
    },
    {
      "epoch": 31.85,
      "learning_rate": 4.079772133691662e-06,
      "loss": 0.0084,
      "step": 841500
    },
    {
      "epoch": 31.87,
      "learning_rate": 4.070309247132745e-06,
      "loss": 0.009,
      "step": 842000
    },
    {
      "epoch": 31.89,
      "learning_rate": 4.06084636057383e-06,
      "loss": 0.0091,
      "step": 842500
    },
    {
      "epoch": 31.91,
      "learning_rate": 4.0513834740149136e-06,
      "loss": 0.0088,
      "step": 843000
    },
    {
      "epoch": 31.93,
      "learning_rate": 4.041920587455998e-06,
      "loss": 0.0086,
      "step": 843500
    },
    {
      "epoch": 31.95,
      "learning_rate": 4.032457700897082e-06,
      "loss": 0.009,
      "step": 844000
    },
    {
      "epoch": 31.97,
      "learning_rate": 4.0229948143381655e-06,
      "loss": 0.0089,
      "step": 844500
    },
    {
      "epoch": 31.98,
      "learning_rate": 4.01353192777925e-06,
      "loss": 0.0091,
      "step": 845000
    },
    {
      "epoch": 32.0,
      "eval_bleu": 92.3903,
      "eval_gen_len": 16.4963,
      "eval_loss": 0.14042149484157562,
      "eval_runtime": 1210.5356,
      "eval_samples_per_second": 37.725,
      "eval_steps_per_second": 1.18,
      "step": 845408
    },
    {
      "epoch": 32.0,
      "learning_rate": 4.004087966993452e-06,
      "loss": 0.0085,
      "step": 845500
    },
    {
      "epoch": 32.02,
      "learning_rate": 3.994625080434536e-06,
      "loss": 0.0076,
      "step": 846000
    },
    {
      "epoch": 32.04,
      "learning_rate": 3.98516219387562e-06,
      "loss": 0.0079,
      "step": 846500
    },
    {
      "epoch": 32.06,
      "learning_rate": 3.975699307316705e-06,
      "loss": 0.008,
      "step": 847000
    },
    {
      "epoch": 32.08,
      "learning_rate": 3.966236420757788e-06,
      "loss": 0.0077,
      "step": 847500
    },
    {
      "epoch": 32.1,
      "learning_rate": 3.956773534198872e-06,
      "loss": 0.0079,
      "step": 848000
    },
    {
      "epoch": 32.12,
      "learning_rate": 3.947310647639957e-06,
      "loss": 0.0078,
      "step": 848500
    },
    {
      "epoch": 32.14,
      "learning_rate": 3.937866686854158e-06,
      "loss": 0.0078,
      "step": 849000
    },
    {
      "epoch": 32.15,
      "learning_rate": 3.928403800295242e-06,
      "loss": 0.0077,
      "step": 849500
    },
    {
      "epoch": 32.17,
      "learning_rate": 3.918940913736327e-06,
      "loss": 0.0075,
      "step": 850000
    },
    {
      "epoch": 32.19,
      "learning_rate": 3.909478027177411e-06,
      "loss": 0.008,
      "step": 850500
    },
    {
      "epoch": 32.21,
      "learning_rate": 3.900034066391613e-06,
      "loss": 0.0081,
      "step": 851000
    },
    {
      "epoch": 32.23,
      "learning_rate": 3.890571179832696e-06,
      "loss": 0.0083,
      "step": 851500
    },
    {
      "epoch": 32.25,
      "learning_rate": 3.88110829327378e-06,
      "loss": 0.0079,
      "step": 852000
    },
    {
      "epoch": 32.27,
      "learning_rate": 3.8716454067148646e-06,
      "loss": 0.0083,
      "step": 852500
    },
    {
      "epoch": 32.29,
      "learning_rate": 3.862182520155949e-06,
      "loss": 0.0079,
      "step": 853000
    },
    {
      "epoch": 32.31,
      "learning_rate": 3.852719633597033e-06,
      "loss": 0.0077,
      "step": 853500
    },
    {
      "epoch": 32.33,
      "learning_rate": 3.8432567470381165e-06,
      "loss": 0.0081,
      "step": 854000
    },
    {
      "epoch": 32.34,
      "learning_rate": 3.833793860479201e-06,
      "loss": 0.0083,
      "step": 854500
    },
    {
      "epoch": 32.36,
      "learning_rate": 3.82436882546652e-06,
      "loss": 0.0079,
      "step": 855000
    },
    {
      "epoch": 32.38,
      "learning_rate": 3.814905938907605e-06,
      "loss": 0.0082,
      "step": 855500
    },
    {
      "epoch": 32.4,
      "learning_rate": 3.8054430523486887e-06,
      "loss": 0.0076,
      "step": 856000
    },
    {
      "epoch": 32.42,
      "learning_rate": 3.7959801657897728e-06,
      "loss": 0.0082,
      "step": 856500
    },
    {
      "epoch": 32.44,
      "learning_rate": 3.786517279230857e-06,
      "loss": 0.0082,
      "step": 857000
    },
    {
      "epoch": 32.46,
      "learning_rate": 3.777054392671941e-06,
      "loss": 0.0084,
      "step": 857500
    },
    {
      "epoch": 32.48,
      "learning_rate": 3.767610431886143e-06,
      "loss": 0.0083,
      "step": 858000
    },
    {
      "epoch": 32.5,
      "learning_rate": 3.758147545327227e-06,
      "loss": 0.0079,
      "step": 858500
    },
    {
      "epoch": 32.51,
      "learning_rate": 3.7486846587683112e-06,
      "loss": 0.0082,
      "step": 859000
    },
    {
      "epoch": 32.53,
      "learning_rate": 3.739221772209395e-06,
      "loss": 0.0079,
      "step": 859500
    },
    {
      "epoch": 32.55,
      "learning_rate": 3.729758885650479e-06,
      "loss": 0.0078,
      "step": 860000
    },
    {
      "epoch": 32.57,
      "learning_rate": 3.720295999091563e-06,
      "loss": 0.0077,
      "step": 860500
    },
    {
      "epoch": 32.59,
      "learning_rate": 3.7108331125326473e-06,
      "loss": 0.0083,
      "step": 861000
    },
    {
      "epoch": 32.61,
      "learning_rate": 3.701370225973731e-06,
      "loss": 0.0085,
      "step": 861500
    },
    {
      "epoch": 32.63,
      "learning_rate": 3.6919262651879334e-06,
      "loss": 0.008,
      "step": 862000
    },
    {
      "epoch": 32.65,
      "learning_rate": 3.682463378629017e-06,
      "loss": 0.0078,
      "step": 862500
    },
    {
      "epoch": 32.67,
      "learning_rate": 3.673000492070101e-06,
      "loss": 0.0082,
      "step": 863000
    },
    {
      "epoch": 32.68,
      "learning_rate": 3.6635376055111853e-06,
      "loss": 0.0082,
      "step": 863500
    },
    {
      "epoch": 32.7,
      "learning_rate": 3.6540936447253873e-06,
      "loss": 0.008,
      "step": 864000
    },
    {
      "epoch": 32.72,
      "learning_rate": 3.6446307581664714e-06,
      "loss": 0.0085,
      "step": 864500
    },
    {
      "epoch": 32.74,
      "learning_rate": 3.6351867973806733e-06,
      "loss": 0.0084,
      "step": 865000
    },
    {
      "epoch": 32.76,
      "learning_rate": 3.6257239108217575e-06,
      "loss": 0.0081,
      "step": 865500
    },
    {
      "epoch": 32.78,
      "learning_rate": 3.6162610242628416e-06,
      "loss": 0.0081,
      "step": 866000
    },
    {
      "epoch": 32.8,
      "learning_rate": 3.6067981377039257e-06,
      "loss": 0.0085,
      "step": 866500
    },
    {
      "epoch": 32.82,
      "learning_rate": 3.5973352511450094e-06,
      "loss": 0.0084,
      "step": 867000
    },
    {
      "epoch": 32.84,
      "learning_rate": 3.5878723645860935e-06,
      "loss": 0.0079,
      "step": 867500
    },
    {
      "epoch": 32.86,
      "learning_rate": 3.5784094780271777e-06,
      "loss": 0.0083,
      "step": 868000
    },
    {
      "epoch": 32.87,
      "learning_rate": 3.568946591468262e-06,
      "loss": 0.0082,
      "step": 868500
    },
    {
      "epoch": 32.89,
      "learning_rate": 3.5594837049093455e-06,
      "loss": 0.0084,
      "step": 869000
    },
    {
      "epoch": 32.91,
      "learning_rate": 3.5500208183504296e-06,
      "loss": 0.0083,
      "step": 869500
    },
    {
      "epoch": 32.93,
      "learning_rate": 3.540557931791514e-06,
      "loss": 0.0083,
      "step": 870000
    },
    {
      "epoch": 32.95,
      "learning_rate": 3.5310950452325983e-06,
      "loss": 0.0078,
      "step": 870500
    },
    {
      "epoch": 32.97,
      "learning_rate": 3.5216321586736824e-06,
      "loss": 0.0079,
      "step": 871000
    },
    {
      "epoch": 32.99,
      "learning_rate": 3.512188197887884e-06,
      "loss": 0.0083,
      "step": 871500
    },
    {
      "epoch": 33.0,
      "eval_bleu": 92.367,
      "eval_gen_len": 16.4617,
      "eval_loss": 0.14141865074634552,
      "eval_runtime": 1174.1308,
      "eval_samples_per_second": 38.894,
      "eval_steps_per_second": 1.216,
      "step": 871827
    },
    {
      "epoch": 33.01,
      "learning_rate": 3.5027253113289685e-06,
      "loss": 0.008,
      "step": 872000
    },
    {
      "epoch": 33.03,
      "learning_rate": 3.4932624247700518e-06,
      "loss": 0.0078,
      "step": 872500
    },
    {
      "epoch": 33.04,
      "learning_rate": 3.4837995382111363e-06,
      "loss": 0.0073,
      "step": 873000
    },
    {
      "epoch": 33.06,
      "learning_rate": 3.4743366516522204e-06,
      "loss": 0.0072,
      "step": 873500
    },
    {
      "epoch": 33.08,
      "learning_rate": 3.4648737650933046e-06,
      "loss": 0.0077,
      "step": 874000
    },
    {
      "epoch": 33.1,
      "learning_rate": 3.455429804307506e-06,
      "loss": 0.0076,
      "step": 874500
    },
    {
      "epoch": 33.12,
      "learning_rate": 3.445985843521708e-06,
      "loss": 0.0075,
      "step": 875000
    },
    {
      "epoch": 33.14,
      "learning_rate": 3.436522956962792e-06,
      "loss": 0.0071,
      "step": 875500
    },
    {
      "epoch": 33.16,
      "learning_rate": 3.4270600704038763e-06,
      "loss": 0.0073,
      "step": 876000
    },
    {
      "epoch": 33.18,
      "learning_rate": 3.41759718384496e-06,
      "loss": 0.0071,
      "step": 876500
    },
    {
      "epoch": 33.2,
      "learning_rate": 3.408134297286044e-06,
      "loss": 0.0073,
      "step": 877000
    },
    {
      "epoch": 33.21,
      "learning_rate": 3.3986714107271287e-06,
      "loss": 0.0075,
      "step": 877500
    },
    {
      "epoch": 33.23,
      "learning_rate": 3.389208524168213e-06,
      "loss": 0.0076,
      "step": 878000
    },
    {
      "epoch": 33.25,
      "learning_rate": 3.3797645633824143e-06,
      "loss": 0.0072,
      "step": 878500
    },
    {
      "epoch": 33.27,
      "learning_rate": 3.3703016768234984e-06,
      "loss": 0.0075,
      "step": 879000
    },
    {
      "epoch": 33.29,
      "learning_rate": 3.360838790264583e-06,
      "loss": 0.0075,
      "step": 879500
    },
    {
      "epoch": 33.31,
      "learning_rate": 3.3513759037056663e-06,
      "loss": 0.0078,
      "step": 880000
    },
    {
      "epoch": 33.33,
      "learning_rate": 3.341913017146751e-06,
      "loss": 0.0075,
      "step": 880500
    },
    {
      "epoch": 33.35,
      "learning_rate": 3.3324690563609523e-06,
      "loss": 0.0076,
      "step": 881000
    },
    {
      "epoch": 33.37,
      "learning_rate": 3.3230061698020365e-06,
      "loss": 0.0075,
      "step": 881500
    },
    {
      "epoch": 33.39,
      "learning_rate": 3.3135432832431206e-06,
      "loss": 0.0074,
      "step": 882000
    },
    {
      "epoch": 33.4,
      "learning_rate": 3.304080396684205e-06,
      "loss": 0.008,
      "step": 882500
    },
    {
      "epoch": 33.42,
      "learning_rate": 3.2946364358984067e-06,
      "loss": 0.0078,
      "step": 883000
    },
    {
      "epoch": 33.44,
      "learning_rate": 3.285173549339491e-06,
      "loss": 0.0074,
      "step": 883500
    },
    {
      "epoch": 33.46,
      "learning_rate": 3.2757106627805745e-06,
      "loss": 0.0077,
      "step": 884000
    },
    {
      "epoch": 33.48,
      "learning_rate": 3.266266701994777e-06,
      "loss": 0.0076,
      "step": 884500
    },
    {
      "epoch": 33.5,
      "learning_rate": 3.2568038154358606e-06,
      "loss": 0.007,
      "step": 885000
    },
    {
      "epoch": 33.52,
      "learning_rate": 3.2473409288769447e-06,
      "loss": 0.0078,
      "step": 885500
    },
    {
      "epoch": 33.54,
      "learning_rate": 3.237878042318029e-06,
      "loss": 0.0075,
      "step": 886000
    },
    {
      "epoch": 33.56,
      "learning_rate": 3.228415155759113e-06,
      "loss": 0.0075,
      "step": 886500
    },
    {
      "epoch": 33.57,
      "learning_rate": 3.2189522692001975e-06,
      "loss": 0.008,
      "step": 887000
    },
    {
      "epoch": 33.59,
      "learning_rate": 3.2094893826412808e-06,
      "loss": 0.0075,
      "step": 887500
    },
    {
      "epoch": 33.61,
      "learning_rate": 3.2000264960823653e-06,
      "loss": 0.0077,
      "step": 888000
    },
    {
      "epoch": 33.63,
      "learning_rate": 3.1905636095234494e-06,
      "loss": 0.0078,
      "step": 888500
    },
    {
      "epoch": 33.65,
      "learning_rate": 3.1811007229645336e-06,
      "loss": 0.0072,
      "step": 889000
    },
    {
      "epoch": 33.67,
      "learning_rate": 3.1716378364056173e-06,
      "loss": 0.0074,
      "step": 889500
    },
    {
      "epoch": 33.69,
      "learning_rate": 3.1621749498467014e-06,
      "loss": 0.0077,
      "step": 890000
    },
    {
      "epoch": 33.71,
      "learning_rate": 3.1527309890609033e-06,
      "loss": 0.0077,
      "step": 890500
    },
    {
      "epoch": 33.73,
      "learning_rate": 3.1432681025019875e-06,
      "loss": 0.0076,
      "step": 891000
    },
    {
      "epoch": 33.74,
      "learning_rate": 3.1338052159430716e-06,
      "loss": 0.0078,
      "step": 891500
    },
    {
      "epoch": 33.76,
      "learning_rate": 3.124361255157273e-06,
      "loss": 0.0079,
      "step": 892000
    },
    {
      "epoch": 33.78,
      "learning_rate": 3.1148983685983577e-06,
      "loss": 0.0075,
      "step": 892500
    },
    {
      "epoch": 33.8,
      "learning_rate": 3.1054354820394418e-06,
      "loss": 0.0077,
      "step": 893000
    },
    {
      "epoch": 33.82,
      "learning_rate": 3.095972595480526e-06,
      "loss": 0.0075,
      "step": 893500
    },
    {
      "epoch": 33.84,
      "learning_rate": 3.0865286346947274e-06,
      "loss": 0.0077,
      "step": 894000
    },
    {
      "epoch": 33.86,
      "learning_rate": 3.0770846739089294e-06,
      "loss": 0.0077,
      "step": 894500
    },
    {
      "epoch": 33.88,
      "learning_rate": 3.0676217873500135e-06,
      "loss": 0.0076,
      "step": 895000
    },
    {
      "epoch": 33.9,
      "learning_rate": 3.0581589007910976e-06,
      "loss": 0.0077,
      "step": 895500
    },
    {
      "epoch": 33.91,
      "learning_rate": 3.0486960142321813e-06,
      "loss": 0.0082,
      "step": 896000
    },
    {
      "epoch": 33.93,
      "learning_rate": 3.0392331276732655e-06,
      "loss": 0.0078,
      "step": 896500
    },
    {
      "epoch": 33.95,
      "learning_rate": 3.02977024111435e-06,
      "loss": 0.0078,
      "step": 897000
    },
    {
      "epoch": 33.97,
      "learning_rate": 3.020307354555434e-06,
      "loss": 0.0076,
      "step": 897500
    },
    {
      "epoch": 33.99,
      "learning_rate": 3.010844467996518e-06,
      "loss": 0.0079,
      "step": 898000
    },
    {
      "epoch": 34.0,
      "eval_bleu": 92.3356,
      "eval_gen_len": 16.4683,
      "eval_loss": 0.14310364425182343,
      "eval_runtime": 1172.1921,
      "eval_samples_per_second": 38.959,
      "eval_steps_per_second": 1.218,
      "step": 898246
    },
    {
      "epoch": 34.01,
      "learning_rate": 3.001381581437602e-06,
      "loss": 0.0078,
      "step": 898500
    },
    {
      "epoch": 34.03,
      "learning_rate": 2.991918694878686e-06,
      "loss": 0.0074,
      "step": 899000
    },
    {
      "epoch": 34.05,
      "learning_rate": 2.9824747340928876e-06,
      "loss": 0.0069,
      "step": 899500
    },
    {
      "epoch": 34.07,
      "learning_rate": 2.97303077330709e-06,
      "loss": 0.0072,
      "step": 900000
    },
    {
      "epoch": 34.09,
      "learning_rate": 2.9635678867481737e-06,
      "loss": 0.0072,
      "step": 900500
    },
    {
      "epoch": 34.1,
      "learning_rate": 2.954105000189258e-06,
      "loss": 0.0073,
      "step": 901000
    },
    {
      "epoch": 34.12,
      "learning_rate": 2.944642113630342e-06,
      "loss": 0.0071,
      "step": 901500
    },
    {
      "epoch": 34.14,
      "learning_rate": 2.9351792270714265e-06,
      "loss": 0.007,
      "step": 902000
    },
    {
      "epoch": 34.16,
      "learning_rate": 2.9257163405125098e-06,
      "loss": 0.0072,
      "step": 902500
    },
    {
      "epoch": 34.18,
      "learning_rate": 2.916272379726712e-06,
      "loss": 0.0076,
      "step": 903000
    },
    {
      "epoch": 34.2,
      "learning_rate": 2.906809493167796e-06,
      "loss": 0.0071,
      "step": 903500
    },
    {
      "epoch": 34.22,
      "learning_rate": 2.89734660660888e-06,
      "loss": 0.0071,
      "step": 904000
    },
    {
      "epoch": 34.24,
      "learning_rate": 2.8878837200499645e-06,
      "loss": 0.007,
      "step": 904500
    },
    {
      "epoch": 34.26,
      "learning_rate": 2.8784208334910486e-06,
      "loss": 0.0071,
      "step": 905000
    },
    {
      "epoch": 34.27,
      "learning_rate": 2.8689579469321323e-06,
      "loss": 0.0073,
      "step": 905500
    },
    {
      "epoch": 34.29,
      "learning_rate": 2.8594950603732165e-06,
      "loss": 0.0068,
      "step": 906000
    },
    {
      "epoch": 34.31,
      "learning_rate": 2.8500321738143006e-06,
      "loss": 0.0069,
      "step": 906500
    },
    {
      "epoch": 34.33,
      "learning_rate": 2.8405692872553847e-06,
      "loss": 0.0074,
      "step": 907000
    },
    {
      "epoch": 34.35,
      "learning_rate": 2.831106400696469e-06,
      "loss": 0.0072,
      "step": 907500
    },
    {
      "epoch": 34.37,
      "learning_rate": 2.8216624399106708e-06,
      "loss": 0.0072,
      "step": 908000
    },
    {
      "epoch": 34.39,
      "learning_rate": 2.812199553351755e-06,
      "loss": 0.0076,
      "step": 908500
    },
    {
      "epoch": 34.41,
      "learning_rate": 2.8027366667928386e-06,
      "loss": 0.0076,
      "step": 909000
    },
    {
      "epoch": 34.43,
      "learning_rate": 2.7932737802339227e-06,
      "loss": 0.0073,
      "step": 909500
    },
    {
      "epoch": 34.44,
      "learning_rate": 2.783810893675007e-06,
      "loss": 0.0071,
      "step": 910000
    },
    {
      "epoch": 34.46,
      "learning_rate": 2.774348007116091e-06,
      "loss": 0.0075,
      "step": 910500
    },
    {
      "epoch": 34.48,
      "learning_rate": 2.7648851205571747e-06,
      "loss": 0.0072,
      "step": 911000
    },
    {
      "epoch": 34.5,
      "learning_rate": 2.755422233998259e-06,
      "loss": 0.0072,
      "step": 911500
    },
    {
      "epoch": 34.52,
      "learning_rate": 2.745959347439343e-06,
      "loss": 0.0075,
      "step": 912000
    },
    {
      "epoch": 34.54,
      "learning_rate": 2.736515386653545e-06,
      "loss": 0.0075,
      "step": 912500
    },
    {
      "epoch": 34.56,
      "learning_rate": 2.727052500094629e-06,
      "loss": 0.007,
      "step": 913000
    },
    {
      "epoch": 34.58,
      "learning_rate": 2.717589613535713e-06,
      "loss": 0.0072,
      "step": 913500
    },
    {
      "epoch": 34.6,
      "learning_rate": 2.7081267269767973e-06,
      "loss": 0.0069,
      "step": 914000
    },
    {
      "epoch": 34.62,
      "learning_rate": 2.698663840417881e-06,
      "loss": 0.0068,
      "step": 914500
    },
    {
      "epoch": 34.63,
      "learning_rate": 2.6892198796320833e-06,
      "loss": 0.0074,
      "step": 915000
    },
    {
      "epoch": 34.65,
      "learning_rate": 2.679756993073167e-06,
      "loss": 0.0075,
      "step": 915500
    },
    {
      "epoch": 34.67,
      "learning_rate": 2.670294106514251e-06,
      "loss": 0.0072,
      "step": 916000
    },
    {
      "epoch": 34.69,
      "learning_rate": 2.6608312199553353e-06,
      "loss": 0.0072,
      "step": 916500
    },
    {
      "epoch": 34.71,
      "learning_rate": 2.65136833339642e-06,
      "loss": 0.0074,
      "step": 917000
    },
    {
      "epoch": 34.73,
      "learning_rate": 2.6419243726106214e-06,
      "loss": 0.0069,
      "step": 917500
    },
    {
      "epoch": 34.75,
      "learning_rate": 2.6324614860517055e-06,
      "loss": 0.0071,
      "step": 918000
    },
    {
      "epoch": 34.77,
      "learning_rate": 2.6230175252659074e-06,
      "loss": 0.0075,
      "step": 918500
    },
    {
      "epoch": 34.79,
      "learning_rate": 2.6135546387069916e-06,
      "loss": 0.0074,
      "step": 919000
    },
    {
      "epoch": 34.8,
      "learning_rate": 2.6040917521480753e-06,
      "loss": 0.0071,
      "step": 919500
    },
    {
      "epoch": 34.82,
      "learning_rate": 2.5946288655891594e-06,
      "loss": 0.0071,
      "step": 920000
    },
    {
      "epoch": 34.84,
      "learning_rate": 2.5851849048033617e-06,
      "loss": 0.0075,
      "step": 920500
    },
    {
      "epoch": 34.86,
      "learning_rate": 2.5757220182444454e-06,
      "loss": 0.0071,
      "step": 921000
    },
    {
      "epoch": 34.88,
      "learning_rate": 2.5662591316855296e-06,
      "loss": 0.0073,
      "step": 921500
    },
    {
      "epoch": 34.9,
      "learning_rate": 2.5567962451266137e-06,
      "loss": 0.0073,
      "step": 922000
    },
    {
      "epoch": 34.92,
      "learning_rate": 2.547333358567698e-06,
      "loss": 0.0071,
      "step": 922500
    },
    {
      "epoch": 34.94,
      "learning_rate": 2.5378704720087815e-06,
      "loss": 0.0072,
      "step": 923000
    },
    {
      "epoch": 34.96,
      "learning_rate": 2.5284075854498657e-06,
      "loss": 0.0076,
      "step": 923500
    },
    {
      "epoch": 34.97,
      "learning_rate": 2.5189446988909498e-06,
      "loss": 0.0073,
      "step": 924000
    },
    {
      "epoch": 34.99,
      "learning_rate": 2.5095007381051517e-06,
      "loss": 0.0071,
      "step": 924500
    },
    {
      "epoch": 35.0,
      "eval_bleu": 92.3866,
      "eval_gen_len": 16.4826,
      "eval_loss": 0.14408156275749207,
      "eval_runtime": 1173.9215,
      "eval_samples_per_second": 38.901,
      "eval_steps_per_second": 1.216,
      "step": 924665
    },
    {
      "epoch": 35.01,
      "learning_rate": 2.500037851546236e-06,
      "loss": 0.007,
      "step": 925000
    },
    {
      "epoch": 35.03,
      "learning_rate": 2.49057496498732e-06,
      "loss": 0.0068,
      "step": 925500
    },
    {
      "epoch": 35.05,
      "learning_rate": 2.481112078428404e-06,
      "loss": 0.0068,
      "step": 926000
    },
    {
      "epoch": 35.07,
      "learning_rate": 2.4716491918694882e-06,
      "loss": 0.0064,
      "step": 926500
    },
    {
      "epoch": 35.09,
      "learning_rate": 2.462186305310572e-06,
      "loss": 0.0068,
      "step": 927000
    },
    {
      "epoch": 35.11,
      "learning_rate": 2.4527234187516565e-06,
      "loss": 0.0072,
      "step": 927500
    },
    {
      "epoch": 35.13,
      "learning_rate": 2.44326053219274e-06,
      "loss": 0.0068,
      "step": 928000
    },
    {
      "epoch": 35.15,
      "learning_rate": 2.4337976456338243e-06,
      "loss": 0.0066,
      "step": 928500
    },
    {
      "epoch": 35.16,
      "learning_rate": 2.4243347590749084e-06,
      "loss": 0.0068,
      "step": 929000
    },
    {
      "epoch": 35.18,
      "learning_rate": 2.4148718725159926e-06,
      "loss": 0.0065,
      "step": 929500
    },
    {
      "epoch": 35.2,
      "learning_rate": 2.4054279117301945e-06,
      "loss": 0.007,
      "step": 930000
    },
    {
      "epoch": 35.22,
      "learning_rate": 2.3959839509443964e-06,
      "loss": 0.0071,
      "step": 930500
    },
    {
      "epoch": 35.24,
      "learning_rate": 2.38652106438548e-06,
      "loss": 0.0068,
      "step": 931000
    },
    {
      "epoch": 35.26,
      "learning_rate": 2.3770581778265643e-06,
      "loss": 0.0073,
      "step": 931500
    },
    {
      "epoch": 35.28,
      "learning_rate": 2.3675952912676484e-06,
      "loss": 0.0064,
      "step": 932000
    },
    {
      "epoch": 35.3,
      "learning_rate": 2.3581324047087325e-06,
      "loss": 0.0069,
      "step": 932500
    },
    {
      "epoch": 35.32,
      "learning_rate": 2.3486695181498166e-06,
      "loss": 0.007,
      "step": 933000
    },
    {
      "epoch": 35.33,
      "learning_rate": 2.3392066315909008e-06,
      "loss": 0.0064,
      "step": 933500
    },
    {
      "epoch": 35.35,
      "learning_rate": 2.3297626708051027e-06,
      "loss": 0.0068,
      "step": 934000
    },
    {
      "epoch": 35.37,
      "learning_rate": 2.3202997842461864e-06,
      "loss": 0.0067,
      "step": 934500
    },
    {
      "epoch": 35.39,
      "learning_rate": 2.310836897687271e-06,
      "loss": 0.0067,
      "step": 935000
    },
    {
      "epoch": 35.41,
      "learning_rate": 2.3013740111283547e-06,
      "loss": 0.0066,
      "step": 935500
    },
    {
      "epoch": 35.43,
      "learning_rate": 2.291911124569439e-06,
      "loss": 0.007,
      "step": 936000
    },
    {
      "epoch": 35.45,
      "learning_rate": 2.2824671637836407e-06,
      "loss": 0.0068,
      "step": 936500
    },
    {
      "epoch": 35.47,
      "learning_rate": 2.273004277224725e-06,
      "loss": 0.0065,
      "step": 937000
    },
    {
      "epoch": 35.49,
      "learning_rate": 2.263541390665809e-06,
      "loss": 0.0067,
      "step": 937500
    },
    {
      "epoch": 35.5,
      "learning_rate": 2.254078504106893e-06,
      "loss": 0.0071,
      "step": 938000
    },
    {
      "epoch": 35.52,
      "learning_rate": 2.244615617547977e-06,
      "loss": 0.0067,
      "step": 938500
    },
    {
      "epoch": 35.54,
      "learning_rate": 2.235152730989061e-06,
      "loss": 0.007,
      "step": 939000
    },
    {
      "epoch": 35.56,
      "learning_rate": 2.225708770203263e-06,
      "loss": 0.0066,
      "step": 939500
    },
    {
      "epoch": 35.58,
      "learning_rate": 2.216245883644347e-06,
      "loss": 0.0072,
      "step": 940000
    },
    {
      "epoch": 35.6,
      "learning_rate": 2.206782997085431e-06,
      "loss": 0.0067,
      "step": 940500
    },
    {
      "epoch": 35.62,
      "learning_rate": 2.1973201105265153e-06,
      "loss": 0.0072,
      "step": 941000
    },
    {
      "epoch": 35.64,
      "learning_rate": 2.1878572239675994e-06,
      "loss": 0.0065,
      "step": 941500
    },
    {
      "epoch": 35.66,
      "learning_rate": 2.178413263181801e-06,
      "loss": 0.0072,
      "step": 942000
    },
    {
      "epoch": 35.68,
      "learning_rate": 2.1689503766228855e-06,
      "loss": 0.007,
      "step": 942500
    },
    {
      "epoch": 35.69,
      "learning_rate": 2.159487490063969e-06,
      "loss": 0.0066,
      "step": 943000
    },
    {
      "epoch": 35.71,
      "learning_rate": 2.1500246035050533e-06,
      "loss": 0.0071,
      "step": 943500
    },
    {
      "epoch": 35.73,
      "learning_rate": 2.1405806427192552e-06,
      "loss": 0.0068,
      "step": 944000
    },
    {
      "epoch": 35.75,
      "learning_rate": 2.1311177561603394e-06,
      "loss": 0.0065,
      "step": 944500
    },
    {
      "epoch": 35.77,
      "learning_rate": 2.1216548696014235e-06,
      "loss": 0.007,
      "step": 945000
    },
    {
      "epoch": 35.79,
      "learning_rate": 2.1121919830425076e-06,
      "loss": 0.0066,
      "step": 945500
    },
    {
      "epoch": 35.81,
      "learning_rate": 2.1027290964835913e-06,
      "loss": 0.0066,
      "step": 946000
    },
    {
      "epoch": 35.83,
      "learning_rate": 2.0932662099246754e-06,
      "loss": 0.0069,
      "step": 946500
    },
    {
      "epoch": 35.85,
      "learning_rate": 2.08380332336576e-06,
      "loss": 0.0073,
      "step": 947000
    },
    {
      "epoch": 35.86,
      "learning_rate": 2.0743404368068437e-06,
      "loss": 0.0067,
      "step": 947500
    },
    {
      "epoch": 35.88,
      "learning_rate": 2.064877550247928e-06,
      "loss": 0.0075,
      "step": 948000
    },
    {
      "epoch": 35.9,
      "learning_rate": 2.055414663689012e-06,
      "loss": 0.0074,
      "step": 948500
    },
    {
      "epoch": 35.92,
      "learning_rate": 2.045951777130096e-06,
      "loss": 0.0069,
      "step": 949000
    },
    {
      "epoch": 35.94,
      "learning_rate": 2.0364888905711798e-06,
      "loss": 0.0068,
      "step": 949500
    },
    {
      "epoch": 35.96,
      "learning_rate": 2.027044929785382e-06,
      "loss": 0.007,
      "step": 950000
    },
    {
      "epoch": 35.98,
      "learning_rate": 2.017582043226466e-06,
      "loss": 0.0067,
      "step": 950500
    },
    {
      "epoch": 36.0,
      "learning_rate": 2.00811915666755e-06,
      "loss": 0.0068,
      "step": 951000
    },
    {
      "epoch": 36.0,
      "eval_bleu": 92.3662,
      "eval_gen_len": 16.471,
      "eval_loss": 0.14456987380981445,
      "eval_runtime": 1176.3281,
      "eval_samples_per_second": 38.822,
      "eval_steps_per_second": 1.214,
      "step": 951084
    },
    {
      "epoch": 36.02,
      "learning_rate": 1.998656270108634e-06,
      "loss": 0.0065,
      "step": 951500
    },
    {
      "epoch": 36.03,
      "learning_rate": 1.9891933835497182e-06,
      "loss": 0.0065,
      "step": 952000
    },
    {
      "epoch": 36.05,
      "learning_rate": 1.979730496990802e-06,
      "loss": 0.0064,
      "step": 952500
    },
    {
      "epoch": 36.07,
      "learning_rate": 1.9702676104318865e-06,
      "loss": 0.0068,
      "step": 953000
    },
    {
      "epoch": 36.09,
      "learning_rate": 1.9608047238729706e-06,
      "loss": 0.0068,
      "step": 953500
    },
    {
      "epoch": 36.11,
      "learning_rate": 1.9513418373140543e-06,
      "loss": 0.0065,
      "step": 954000
    },
    {
      "epoch": 36.13,
      "learning_rate": 1.9418978765282562e-06,
      "loss": 0.0064,
      "step": 954500
    },
    {
      "epoch": 36.15,
      "learning_rate": 1.9324349899693404e-06,
      "loss": 0.0067,
      "step": 955000
    },
    {
      "epoch": 36.17,
      "learning_rate": 1.9229721034104245e-06,
      "loss": 0.0063,
      "step": 955500
    },
    {
      "epoch": 36.19,
      "learning_rate": 1.9135092168515086e-06,
      "loss": 0.0064,
      "step": 956000
    },
    {
      "epoch": 36.21,
      "learning_rate": 1.9040463302925927e-06,
      "loss": 0.0063,
      "step": 956500
    },
    {
      "epoch": 36.22,
      "learning_rate": 1.8945834437336767e-06,
      "loss": 0.0066,
      "step": 957000
    },
    {
      "epoch": 36.24,
      "learning_rate": 1.8851205571747608e-06,
      "loss": 0.0062,
      "step": 957500
    },
    {
      "epoch": 36.26,
      "learning_rate": 1.8756765963889625e-06,
      "loss": 0.0062,
      "step": 958000
    },
    {
      "epoch": 36.28,
      "learning_rate": 1.8662137098300466e-06,
      "loss": 0.0063,
      "step": 958500
    },
    {
      "epoch": 36.3,
      "learning_rate": 1.8567508232711306e-06,
      "loss": 0.0067,
      "step": 959000
    },
    {
      "epoch": 36.32,
      "learning_rate": 1.8472879367122149e-06,
      "loss": 0.0067,
      "step": 959500
    },
    {
      "epoch": 36.34,
      "learning_rate": 1.837825050153299e-06,
      "loss": 0.0062,
      "step": 960000
    },
    {
      "epoch": 36.36,
      "learning_rate": 1.828362163594383e-06,
      "loss": 0.0072,
      "step": 960500
    },
    {
      "epoch": 36.38,
      "learning_rate": 1.818899277035467e-06,
      "loss": 0.0068,
      "step": 961000
    },
    {
      "epoch": 36.39,
      "learning_rate": 1.809455316249669e-06,
      "loss": 0.0063,
      "step": 961500
    },
    {
      "epoch": 36.41,
      "learning_rate": 1.7999924296907531e-06,
      "loss": 0.0064,
      "step": 962000
    },
    {
      "epoch": 36.43,
      "learning_rate": 1.7905484689049549e-06,
      "loss": 0.0064,
      "step": 962500
    },
    {
      "epoch": 36.45,
      "learning_rate": 1.781085582346039e-06,
      "loss": 0.0066,
      "step": 963000
    },
    {
      "epoch": 36.47,
      "learning_rate": 1.771622695787123e-06,
      "loss": 0.0072,
      "step": 963500
    },
    {
      "epoch": 36.49,
      "learning_rate": 1.7621598092282072e-06,
      "loss": 0.0066,
      "step": 964000
    },
    {
      "epoch": 36.51,
      "learning_rate": 1.7526969226692912e-06,
      "loss": 0.0071,
      "step": 964500
    },
    {
      "epoch": 36.53,
      "learning_rate": 1.7432340361103753e-06,
      "loss": 0.0066,
      "step": 965000
    },
    {
      "epoch": 36.55,
      "learning_rate": 1.7337711495514592e-06,
      "loss": 0.0069,
      "step": 965500
    },
    {
      "epoch": 36.56,
      "learning_rate": 1.7243082629925433e-06,
      "loss": 0.0068,
      "step": 966000
    },
    {
      "epoch": 36.58,
      "learning_rate": 1.7148453764336277e-06,
      "loss": 0.0065,
      "step": 966500
    },
    {
      "epoch": 36.6,
      "learning_rate": 1.7053824898747116e-06,
      "loss": 0.0064,
      "step": 967000
    },
    {
      "epoch": 36.62,
      "learning_rate": 1.6959385290889135e-06,
      "loss": 0.0063,
      "step": 967500
    },
    {
      "epoch": 36.64,
      "learning_rate": 1.6864756425299974e-06,
      "loss": 0.0063,
      "step": 968000
    },
    {
      "epoch": 36.66,
      "learning_rate": 1.6770127559710816e-06,
      "loss": 0.0068,
      "step": 968500
    },
    {
      "epoch": 36.68,
      "learning_rate": 1.6675498694121655e-06,
      "loss": 0.0061,
      "step": 969000
    },
    {
      "epoch": 36.7,
      "learning_rate": 1.6580869828532498e-06,
      "loss": 0.0062,
      "step": 969500
    },
    {
      "epoch": 36.72,
      "learning_rate": 1.6486240962943337e-06,
      "loss": 0.0063,
      "step": 970000
    },
    {
      "epoch": 36.73,
      "learning_rate": 1.6391612097354178e-06,
      "loss": 0.0066,
      "step": 970500
    },
    {
      "epoch": 36.75,
      "learning_rate": 1.6296983231765018e-06,
      "loss": 0.0062,
      "step": 971000
    },
    {
      "epoch": 36.77,
      "learning_rate": 1.6202354366175859e-06,
      "loss": 0.007,
      "step": 971500
    },
    {
      "epoch": 36.79,
      "learning_rate": 1.6108104016049056e-06,
      "loss": 0.0065,
      "step": 972000
    },
    {
      "epoch": 36.81,
      "learning_rate": 1.6013475150459898e-06,
      "loss": 0.0071,
      "step": 972500
    },
    {
      "epoch": 36.83,
      "learning_rate": 1.5918846284870737e-06,
      "loss": 0.0066,
      "step": 973000
    },
    {
      "epoch": 36.85,
      "learning_rate": 1.5824217419281578e-06,
      "loss": 0.0064,
      "step": 973500
    },
    {
      "epoch": 36.87,
      "learning_rate": 1.5729588553692421e-06,
      "loss": 0.0064,
      "step": 974000
    },
    {
      "epoch": 36.89,
      "learning_rate": 1.563495968810326e-06,
      "loss": 0.0065,
      "step": 974500
    },
    {
      "epoch": 36.91,
      "learning_rate": 1.5540330822514102e-06,
      "loss": 0.0065,
      "step": 975000
    },
    {
      "epoch": 36.92,
      "learning_rate": 1.544570195692494e-06,
      "loss": 0.0066,
      "step": 975500
    },
    {
      "epoch": 36.94,
      "learning_rate": 1.5351073091335782e-06,
      "loss": 0.0072,
      "step": 976000
    },
    {
      "epoch": 36.96,
      "learning_rate": 1.5256633483477802e-06,
      "loss": 0.0066,
      "step": 976500
    },
    {
      "epoch": 36.98,
      "learning_rate": 1.5162004617888643e-06,
      "loss": 0.0063,
      "step": 977000
    },
    {
      "epoch": 37.0,
      "learning_rate": 1.5067375752299482e-06,
      "loss": 0.0066,
      "step": 977500
    },
    {
      "epoch": 37.0,
      "eval_bleu": 92.3884,
      "eval_gen_len": 16.4703,
      "eval_loss": 0.145610511302948,
      "eval_runtime": 1252.6827,
      "eval_samples_per_second": 36.455,
      "eval_steps_per_second": 1.14,
      "step": 977503
    },
    {
      "epoch": 37.02,
      "learning_rate": 1.4972746886710323e-06,
      "loss": 0.0061,
      "step": 978000
    },
    {
      "epoch": 37.04,
      "learning_rate": 1.4878118021121163e-06,
      "loss": 0.0061,
      "step": 978500
    },
    {
      "epoch": 37.06,
      "learning_rate": 1.4783678413263184e-06,
      "loss": 0.0061,
      "step": 979000
    },
    {
      "epoch": 37.08,
      "learning_rate": 1.4689049547674023e-06,
      "loss": 0.0062,
      "step": 979500
    },
    {
      "epoch": 37.09,
      "learning_rate": 1.4594420682084864e-06,
      "loss": 0.0063,
      "step": 980000
    },
    {
      "epoch": 37.11,
      "learning_rate": 1.4499791816495706e-06,
      "loss": 0.0063,
      "step": 980500
    },
    {
      "epoch": 37.13,
      "learning_rate": 1.4405162950906545e-06,
      "loss": 0.0063,
      "step": 981000
    },
    {
      "epoch": 37.15,
      "learning_rate": 1.4310534085317388e-06,
      "loss": 0.0065,
      "step": 981500
    },
    {
      "epoch": 37.17,
      "learning_rate": 1.4215905219728227e-06,
      "loss": 0.006,
      "step": 982000
    },
    {
      "epoch": 37.19,
      "learning_rate": 1.4121276354139069e-06,
      "loss": 0.0064,
      "step": 982500
    },
    {
      "epoch": 37.21,
      "learning_rate": 1.4027026004012264e-06,
      "loss": 0.0065,
      "step": 983000
    },
    {
      "epoch": 37.23,
      "learning_rate": 1.3932586396154286e-06,
      "loss": 0.006,
      "step": 983500
    },
    {
      "epoch": 37.25,
      "learning_rate": 1.3837957530565125e-06,
      "loss": 0.0062,
      "step": 984000
    },
    {
      "epoch": 37.26,
      "learning_rate": 1.3743328664975966e-06,
      "loss": 0.0063,
      "step": 984500
    },
    {
      "epoch": 37.28,
      "learning_rate": 1.3648699799386805e-06,
      "loss": 0.0058,
      "step": 985000
    },
    {
      "epoch": 37.3,
      "learning_rate": 1.3554070933797647e-06,
      "loss": 0.006,
      "step": 985500
    },
    {
      "epoch": 37.32,
      "learning_rate": 1.3459442068208486e-06,
      "loss": 0.0065,
      "step": 986000
    },
    {
      "epoch": 37.34,
      "learning_rate": 1.336481320261933e-06,
      "loss": 0.0063,
      "step": 986500
    },
    {
      "epoch": 37.36,
      "learning_rate": 1.3270184337030168e-06,
      "loss": 0.0063,
      "step": 987000
    },
    {
      "epoch": 37.38,
      "learning_rate": 1.317555547144101e-06,
      "loss": 0.0061,
      "step": 987500
    },
    {
      "epoch": 37.4,
      "learning_rate": 1.3081115863583027e-06,
      "loss": 0.0064,
      "step": 988000
    },
    {
      "epoch": 37.42,
      "learning_rate": 1.2986486997993868e-06,
      "loss": 0.0065,
      "step": 988500
    },
    {
      "epoch": 37.44,
      "learning_rate": 1.2891858132404711e-06,
      "loss": 0.0062,
      "step": 989000
    },
    {
      "epoch": 37.45,
      "learning_rate": 1.279722926681555e-06,
      "loss": 0.0061,
      "step": 989500
    },
    {
      "epoch": 37.47,
      "learning_rate": 1.270278965895757e-06,
      "loss": 0.0067,
      "step": 990000
    },
    {
      "epoch": 37.49,
      "learning_rate": 1.260816079336841e-06,
      "loss": 0.0058,
      "step": 990500
    },
    {
      "epoch": 37.51,
      "learning_rate": 1.2513531927779253e-06,
      "loss": 0.007,
      "step": 991000
    },
    {
      "epoch": 37.53,
      "learning_rate": 1.2418903062190092e-06,
      "loss": 0.0066,
      "step": 991500
    },
    {
      "epoch": 37.55,
      "learning_rate": 1.232446345433211e-06,
      "loss": 0.0064,
      "step": 992000
    },
    {
      "epoch": 37.57,
      "learning_rate": 1.2229834588742952e-06,
      "loss": 0.0064,
      "step": 992500
    },
    {
      "epoch": 37.59,
      "learning_rate": 1.2135205723153792e-06,
      "loss": 0.0061,
      "step": 993000
    },
    {
      "epoch": 37.61,
      "learning_rate": 1.204076611529581e-06,
      "loss": 0.0064,
      "step": 993500
    },
    {
      "epoch": 37.62,
      "learning_rate": 1.1946137249706652e-06,
      "loss": 0.0063,
      "step": 994000
    },
    {
      "epoch": 37.64,
      "learning_rate": 1.1851508384117491e-06,
      "loss": 0.0062,
      "step": 994500
    },
    {
      "epoch": 37.66,
      "learning_rate": 1.1756879518528333e-06,
      "loss": 0.006,
      "step": 995000
    },
    {
      "epoch": 37.68,
      "learning_rate": 1.1662250652939174e-06,
      "loss": 0.0058,
      "step": 995500
    },
    {
      "epoch": 37.7,
      "learning_rate": 1.1567621787350013e-06,
      "loss": 0.0062,
      "step": 996000
    },
    {
      "epoch": 37.72,
      "learning_rate": 1.1472992921760854e-06,
      "loss": 0.0063,
      "step": 996500
    },
    {
      "epoch": 37.74,
      "learning_rate": 1.1378364056171696e-06,
      "loss": 0.0061,
      "step": 997000
    },
    {
      "epoch": 37.76,
      "learning_rate": 1.1283735190582535e-06,
      "loss": 0.0065,
      "step": 997500
    },
    {
      "epoch": 37.78,
      "learning_rate": 1.1189106324993376e-06,
      "loss": 0.0063,
      "step": 998000
    },
    {
      "epoch": 37.79,
      "learning_rate": 1.1094477459404217e-06,
      "loss": 0.0062,
      "step": 998500
    },
    {
      "epoch": 37.81,
      "learning_rate": 1.0999848593815058e-06,
      "loss": 0.0062,
      "step": 999000
    },
    {
      "epoch": 37.83,
      "learning_rate": 1.0905408985957078e-06,
      "loss": 0.0062,
      "step": 999500
    },
    {
      "epoch": 37.85,
      "learning_rate": 1.081078012036792e-06,
      "loss": 0.0062,
      "step": 1000000
    },
    {
      "epoch": 37.87,
      "learning_rate": 1.0716151254778758e-06,
      "loss": 0.0062,
      "step": 1000500
    },
    {
      "epoch": 37.89,
      "learning_rate": 1.06215223891896e-06,
      "loss": 0.0063,
      "step": 1001000
    },
    {
      "epoch": 37.91,
      "learning_rate": 1.052689352360044e-06,
      "loss": 0.0062,
      "step": 1001500
    },
    {
      "epoch": 37.93,
      "learning_rate": 1.043226465801128e-06,
      "loss": 0.0066,
      "step": 1002000
    },
    {
      "epoch": 37.95,
      "learning_rate": 1.0337635792422121e-06,
      "loss": 0.0065,
      "step": 1002500
    },
    {
      "epoch": 37.97,
      "learning_rate": 1.0243006926832962e-06,
      "loss": 0.0063,
      "step": 1003000
    },
    {
      "epoch": 37.98,
      "learning_rate": 1.0148378061243802e-06,
      "loss": 0.0065,
      "step": 1003500
    },
    {
      "epoch": 38.0,
      "eval_bleu": 92.4009,
      "eval_gen_len": 16.4752,
      "eval_loss": 0.14630930125713348,
      "eval_runtime": 1235.7047,
      "eval_samples_per_second": 36.956,
      "eval_steps_per_second": 1.156,
      "step": 1003922
    },
    {
      "epoch": 38.0,
      "learning_rate": 1.005393845338582e-06,
      "loss": 0.0061,
      "step": 1004000
    },
    {
      "epoch": 38.02,
      "learning_rate": 9.959309587796662e-07,
      "loss": 0.0061,
      "step": 1004500
    },
    {
      "epoch": 38.04,
      "learning_rate": 9.864680722207504e-07,
      "loss": 0.006,
      "step": 1005000
    },
    {
      "epoch": 38.06,
      "learning_rate": 9.770051856618345e-07,
      "loss": 0.0062,
      "step": 1005500
    },
    {
      "epoch": 38.08,
      "learning_rate": 9.675422991029184e-07,
      "loss": 0.0064,
      "step": 1006000
    },
    {
      "epoch": 38.1,
      "learning_rate": 9.580983383171203e-07,
      "loss": 0.006,
      "step": 1006500
    },
    {
      "epoch": 38.12,
      "learning_rate": 9.486354517582045e-07,
      "loss": 0.0058,
      "step": 1007000
    },
    {
      "epoch": 38.14,
      "learning_rate": 9.391725651992885e-07,
      "loss": 0.0061,
      "step": 1007500
    },
    {
      "epoch": 38.15,
      "learning_rate": 9.297096786403725e-07,
      "loss": 0.0061,
      "step": 1008000
    },
    {
      "epoch": 38.17,
      "learning_rate": 9.202467920814566e-07,
      "loss": 0.0063,
      "step": 1008500
    },
    {
      "epoch": 38.19,
      "learning_rate": 9.107839055225406e-07,
      "loss": 0.0062,
      "step": 1009000
    },
    {
      "epoch": 38.21,
      "learning_rate": 9.013399447367426e-07,
      "loss": 0.0059,
      "step": 1009500
    },
    {
      "epoch": 38.23,
      "learning_rate": 8.918959839509444e-07,
      "loss": 0.0061,
      "step": 1010000
    },
    {
      "epoch": 38.25,
      "learning_rate": 8.824330973920286e-07,
      "loss": 0.0066,
      "step": 1010500
    },
    {
      "epoch": 38.27,
      "learning_rate": 8.729702108331126e-07,
      "loss": 0.0057,
      "step": 1011000
    },
    {
      "epoch": 38.29,
      "learning_rate": 8.635073242741966e-07,
      "loss": 0.0065,
      "step": 1011500
    },
    {
      "epoch": 38.31,
      "learning_rate": 8.540444377152807e-07,
      "loss": 0.0061,
      "step": 1012000
    },
    {
      "epoch": 38.32,
      "learning_rate": 8.445815511563648e-07,
      "loss": 0.0058,
      "step": 1012500
    },
    {
      "epoch": 38.34,
      "learning_rate": 8.351186645974489e-07,
      "loss": 0.0061,
      "step": 1013000
    },
    {
      "epoch": 38.36,
      "learning_rate": 8.25655778038533e-07,
      "loss": 0.006,
      "step": 1013500
    },
    {
      "epoch": 38.38,
      "learning_rate": 8.16192891479617e-07,
      "loss": 0.0059,
      "step": 1014000
    },
    {
      "epoch": 38.4,
      "learning_rate": 8.06730004920701e-07,
      "loss": 0.0061,
      "step": 1014500
    },
    {
      "epoch": 38.42,
      "learning_rate": 7.97286044134903e-07,
      "loss": 0.0058,
      "step": 1015000
    },
    {
      "epoch": 38.44,
      "learning_rate": 7.87823157575987e-07,
      "loss": 0.0059,
      "step": 1015500
    },
    {
      "epoch": 38.46,
      "learning_rate": 7.783602710170711e-07,
      "loss": 0.0056,
      "step": 1016000
    },
    {
      "epoch": 38.48,
      "learning_rate": 7.688973844581551e-07,
      "loss": 0.0059,
      "step": 1016500
    },
    {
      "epoch": 38.5,
      "learning_rate": 7.594344978992392e-07,
      "loss": 0.0057,
      "step": 1017000
    },
    {
      "epoch": 38.51,
      "learning_rate": 7.499905371134411e-07,
      "loss": 0.0064,
      "step": 1017500
    },
    {
      "epoch": 38.53,
      "learning_rate": 7.405276505545251e-07,
      "loss": 0.0064,
      "step": 1018000
    },
    {
      "epoch": 38.55,
      "learning_rate": 7.310836897687271e-07,
      "loss": 0.006,
      "step": 1018500
    },
    {
      "epoch": 38.57,
      "learning_rate": 7.21639728982929e-07,
      "loss": 0.0063,
      "step": 1019000
    },
    {
      "epoch": 38.59,
      "learning_rate": 7.12176842424013e-07,
      "loss": 0.0059,
      "step": 1019500
    },
    {
      "epoch": 38.61,
      "learning_rate": 7.027139558650971e-07,
      "loss": 0.006,
      "step": 1020000
    },
    {
      "epoch": 38.63,
      "learning_rate": 6.932510693061813e-07,
      "loss": 0.006,
      "step": 1020500
    },
    {
      "epoch": 38.65,
      "learning_rate": 6.837881827472653e-07,
      "loss": 0.0062,
      "step": 1021000
    },
    {
      "epoch": 38.67,
      "learning_rate": 6.743442219614673e-07,
      "loss": 0.0064,
      "step": 1021500
    },
    {
      "epoch": 38.68,
      "learning_rate": 6.648813354025513e-07,
      "loss": 0.0065,
      "step": 1022000
    },
    {
      "epoch": 38.7,
      "learning_rate": 6.554184488436353e-07,
      "loss": 0.0061,
      "step": 1022500
    },
    {
      "epoch": 38.72,
      "learning_rate": 6.459555622847194e-07,
      "loss": 0.0063,
      "step": 1023000
    },
    {
      "epoch": 38.74,
      "learning_rate": 6.364926757258034e-07,
      "loss": 0.0061,
      "step": 1023500
    },
    {
      "epoch": 38.76,
      "learning_rate": 6.270297891668875e-07,
      "loss": 0.0059,
      "step": 1024000
    },
    {
      "epoch": 38.78,
      "learning_rate": 6.175669026079716e-07,
      "loss": 0.0063,
      "step": 1024500
    },
    {
      "epoch": 38.8,
      "learning_rate": 6.081040160490557e-07,
      "loss": 0.0065,
      "step": 1025000
    },
    {
      "epoch": 38.82,
      "learning_rate": 5.986411294901397e-07,
      "loss": 0.0058,
      "step": 1025500
    },
    {
      "epoch": 38.84,
      "learning_rate": 5.891971687043417e-07,
      "loss": 0.0063,
      "step": 1026000
    },
    {
      "epoch": 38.85,
      "learning_rate": 5.797342821454257e-07,
      "loss": 0.0062,
      "step": 1026500
    },
    {
      "epoch": 38.87,
      "learning_rate": 5.702713955865097e-07,
      "loss": 0.0061,
      "step": 1027000
    },
    {
      "epoch": 38.89,
      "learning_rate": 5.608085090275938e-07,
      "loss": 0.0063,
      "step": 1027500
    },
    {
      "epoch": 38.91,
      "learning_rate": 5.513456224686779e-07,
      "loss": 0.0061,
      "step": 1028000
    },
    {
      "epoch": 38.93,
      "learning_rate": 5.41882735909762e-07,
      "loss": 0.0058,
      "step": 1028500
    },
    {
      "epoch": 38.95,
      "learning_rate": 5.32419849350846e-07,
      "loss": 0.0063,
      "step": 1029000
    },
    {
      "epoch": 38.97,
      "learning_rate": 5.229569627919301e-07,
      "loss": 0.0058,
      "step": 1029500
    },
    {
      "epoch": 38.99,
      "learning_rate": 5.13513002006132e-07,
      "loss": 0.0063,
      "step": 1030000
    },
    {
      "epoch": 39.0,
      "eval_bleu": 92.4327,
      "eval_gen_len": 16.4778,
      "eval_loss": 0.14681382477283478,
      "eval_runtime": 1461.2179,
      "eval_samples_per_second": 31.253,
      "eval_steps_per_second": 0.977,
      "step": 1030341
    },
    {
      "epoch": 39.01,
      "learning_rate": 5.040690412203339e-07,
      "loss": 0.0059,
      "step": 1030500
    },
    {
      "epoch": 39.02,
      "learning_rate": 4.946061546614179e-07,
      "loss": 0.006,
      "step": 1031000
    },
    {
      "epoch": 39.04,
      "learning_rate": 4.85143268102502e-07,
      "loss": 0.0059,
      "step": 1031500
    },
    {
      "epoch": 39.06,
      "learning_rate": 4.756803815435861e-07,
      "loss": 0.0061,
      "step": 1032000
    },
    {
      "epoch": 39.08,
      "learning_rate": 4.662174949846702e-07,
      "loss": 0.0063,
      "step": 1032500
    },
    {
      "epoch": 39.1,
      "learning_rate": 4.567546084257542e-07,
      "loss": 0.0059,
      "step": 1033000
    },
    {
      "epoch": 39.12,
      "learning_rate": 4.472917218668383e-07,
      "loss": 0.0058,
      "step": 1033500
    },
    {
      "epoch": 39.14,
      "learning_rate": 4.3782883530792237e-07,
      "loss": 0.0058,
      "step": 1034000
    },
    {
      "epoch": 39.16,
      "learning_rate": 4.283659487490064e-07,
      "loss": 0.0057,
      "step": 1034500
    },
    {
      "epoch": 39.18,
      "learning_rate": 4.189030621900905e-07,
      "loss": 0.0056,
      "step": 1035000
    },
    {
      "epoch": 39.2,
      "learning_rate": 4.094591014042924e-07,
      "loss": 0.0057,
      "step": 1035500
    },
    {
      "epoch": 39.21,
      "learning_rate": 3.999962148453765e-07,
      "loss": 0.0062,
      "step": 1036000
    },
    {
      "epoch": 39.23,
      "learning_rate": 3.9053332828646055e-07,
      "loss": 0.0057,
      "step": 1036500
    },
    {
      "epoch": 39.25,
      "learning_rate": 3.8107044172754457e-07,
      "loss": 0.0059,
      "step": 1037000
    },
    {
      "epoch": 39.27,
      "learning_rate": 3.7160755516862864e-07,
      "loss": 0.0062,
      "step": 1037500
    },
    {
      "epoch": 39.29,
      "learning_rate": 3.621446686097127e-07,
      "loss": 0.0056,
      "step": 1038000
    },
    {
      "epoch": 39.31,
      "learning_rate": 3.527007078239146e-07,
      "loss": 0.006,
      "step": 1038500
    },
    {
      "epoch": 39.33,
      "learning_rate": 3.4323782126499874e-07,
      "loss": 0.0058,
      "step": 1039000
    },
    {
      "epoch": 39.35,
      "learning_rate": 3.337749347060828e-07,
      "loss": 0.0056,
      "step": 1039500
    },
    {
      "epoch": 39.37,
      "learning_rate": 3.2431204814716683e-07,
      "loss": 0.0056,
      "step": 1040000
    },
    {
      "epoch": 39.38,
      "learning_rate": 3.148491615882509e-07,
      "loss": 0.006,
      "step": 1040500
    },
    {
      "epoch": 39.4,
      "learning_rate": 3.0538627502933497e-07,
      "loss": 0.006,
      "step": 1041000
    },
    {
      "epoch": 39.42,
      "learning_rate": 2.9592338847041904e-07,
      "loss": 0.0056,
      "step": 1041500
    },
    {
      "epoch": 39.44,
      "learning_rate": 2.8647942768462094e-07,
      "loss": 0.0061,
      "step": 1042000
    },
    {
      "epoch": 39.46,
      "learning_rate": 2.77016541125705e-07,
      "loss": 0.0057,
      "step": 1042500
    },
    {
      "epoch": 39.48,
      "learning_rate": 2.675536545667891e-07,
      "loss": 0.0061,
      "step": 1043000
    },
    {
      "epoch": 39.5,
      "learning_rate": 2.5809076800787315e-07,
      "loss": 0.006,
      "step": 1043500
    },
    {
      "epoch": 39.52,
      "learning_rate": 2.4862788144895723e-07,
      "loss": 0.0055,
      "step": 1044000
    },
    {
      "epoch": 39.54,
      "learning_rate": 2.391649948900413e-07,
      "loss": 0.006,
      "step": 1044500
    },
    {
      "epoch": 39.55,
      "learning_rate": 2.2970210833112534e-07,
      "loss": 0.0062,
      "step": 1045000
    },
    {
      "epoch": 39.57,
      "learning_rate": 2.2023922177220942e-07,
      "loss": 0.0064,
      "step": 1045500
    },
    {
      "epoch": 39.59,
      "learning_rate": 2.107763352132935e-07,
      "loss": 0.0058,
      "step": 1046000
    },
    {
      "epoch": 39.61,
      "learning_rate": 2.0131344865437753e-07,
      "loss": 0.0058,
      "step": 1046500
    },
    {
      "epoch": 39.63,
      "learning_rate": 1.9185056209546163e-07,
      "loss": 0.0058,
      "step": 1047000
    },
    {
      "epoch": 39.65,
      "learning_rate": 1.8242552708278135e-07,
      "loss": 0.0063,
      "step": 1047500
    },
    {
      "epoch": 39.67,
      "learning_rate": 1.7296264052386542e-07,
      "loss": 0.0063,
      "step": 1048000
    },
    {
      "epoch": 39.69,
      "learning_rate": 1.634997539649495e-07,
      "loss": 0.0058,
      "step": 1048500
    },
    {
      "epoch": 39.71,
      "learning_rate": 1.5403686740603357e-07,
      "loss": 0.0055,
      "step": 1049000
    },
    {
      "epoch": 39.73,
      "learning_rate": 1.445739808471176e-07,
      "loss": 0.0056,
      "step": 1049500
    },
    {
      "epoch": 39.74,
      "learning_rate": 1.3511109428820168e-07,
      "loss": 0.0058,
      "step": 1050000
    },
    {
      "epoch": 39.76,
      "learning_rate": 1.2566713350240358e-07,
      "loss": 0.0058,
      "step": 1050500
    },
    {
      "epoch": 39.78,
      "learning_rate": 1.1620424694348766e-07,
      "loss": 0.0061,
      "step": 1051000
    },
    {
      "epoch": 39.8,
      "learning_rate": 1.0674136038457171e-07,
      "loss": 0.0057,
      "step": 1051500
    },
    {
      "epoch": 39.82,
      "learning_rate": 9.727847382565578e-08,
      "loss": 0.006,
      "step": 1052000
    },
    {
      "epoch": 39.84,
      "learning_rate": 8.781558726673985e-08,
      "loss": 0.0057,
      "step": 1052500
    },
    {
      "epoch": 39.86,
      "learning_rate": 7.835270070782393e-08,
      "loss": 0.0059,
      "step": 1053000
    },
    {
      "epoch": 39.88,
      "learning_rate": 6.888981414890799e-08,
      "loss": 0.0058,
      "step": 1053500
    },
    {
      "epoch": 39.9,
      "learning_rate": 5.942692758999205e-08,
      "loss": 0.0057,
      "step": 1054000
    },
    {
      "epoch": 39.91,
      "learning_rate": 4.998296680419396e-08,
      "loss": 0.0059,
      "step": 1054500
    },
    {
      "epoch": 39.93,
      "learning_rate": 4.0520080245278024e-08,
      "loss": 0.0059,
      "step": 1055000
    },
    {
      "epoch": 39.95,
      "learning_rate": 3.105719368636209e-08,
      "loss": 0.0063,
      "step": 1055500
    },
    {
      "epoch": 39.97,
      "learning_rate": 2.1594307127446155e-08,
      "loss": 0.0057,
      "step": 1056000
    },
    {
      "epoch": 39.99,
      "learning_rate": 1.2131420568530226e-08,
      "loss": 0.0057,
      "step": 1056500
    },
    {
      "epoch": 40.0,
      "eval_bleu": 92.4356,
      "eval_gen_len": 16.4742,
      "eval_loss": 0.14701814949512482,
      "eval_runtime": 1304.0934,
      "eval_samples_per_second": 35.018,
      "eval_steps_per_second": 1.095,
      "step": 1056760
    }
  ],
  "logging_steps": 500,
  "max_steps": 1056760,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 40,
  "save_steps": 500,
  "total_flos": 2.5508178908682977e+18,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}