{
  "best_metric": 0.7786535620689392,
  "best_model_checkpoint": "./enko_mbartLarge_100p_run1/checkpoint-128123",
  "epoch": 4.999992195000117,
  "eval_steps": 500,
  "global_step": 320307,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.01,
      "learning_rate": 9.940000000000001e-06,
      "loss": 1.6027,
      "step": 500
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.994e-05,
      "loss": 1.3159,
      "step": 1000
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.994e-05,
      "loss": 1.2703,
      "step": 1500
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.994e-05,
      "loss": 1.2587,
      "step": 2000
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.9940000000000006e-05,
      "loss": 1.2526,
      "step": 2500
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.997407177475311e-05,
      "loss": 1.2571,
      "step": 3000
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.994803921057162e-05,
      "loss": 1.2336,
      "step": 3500
    },
    {
      "epoch": 0.06,
      "learning_rate": 4.992195447692284e-05,
      "loss": 1.2062,
      "step": 4000
    },
    {
      "epoch": 0.07,
      "learning_rate": 4.9895869743274054e-05,
      "loss": 1.1793,
      "step": 4500
    },
    {
      "epoch": 0.08,
      "learning_rate": 4.986983717909256e-05,
      "loss": 1.1679,
      "step": 5000
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.984375244544378e-05,
      "loss": 1.148,
      "step": 5500
    },
    {
      "epoch": 0.09,
      "learning_rate": 4.9817667711794995e-05,
      "loss": 1.1374,
      "step": 6000
    },
    {
      "epoch": 0.1,
      "learning_rate": 4.979158297814621e-05,
      "loss": 1.1147,
      "step": 6500
    },
    {
      "epoch": 0.11,
      "learning_rate": 4.976549824449743e-05,
      "loss": 1.1143,
      "step": 7000
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.9739465680315936e-05,
      "loss": 1.1026,
      "step": 7500
    },
    {
      "epoch": 0.12,
      "learning_rate": 4.971338094666715e-05,
      "loss": 1.1118,
      "step": 8000
    },
    {
      "epoch": 0.13,
      "learning_rate": 4.968729621301837e-05,
      "loss": 1.0842,
      "step": 8500
    },
    {
      "epoch": 0.14,
      "learning_rate": 4.9661211479369584e-05,
      "loss": 1.0774,
      "step": 9000
    },
    {
      "epoch": 0.15,
      "learning_rate": 4.96351267457208e-05,
      "loss": 1.0735,
      "step": 9500
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.9609042012072016e-05,
      "loss": 1.0637,
      "step": 10000
    },
    {
      "epoch": 0.16,
      "learning_rate": 4.958300944789053e-05,
      "loss": 1.0474,
      "step": 10500
    },
    {
      "epoch": 0.17,
      "learning_rate": 4.955692471424174e-05,
      "loss": 1.0447,
      "step": 11000
    },
    {
      "epoch": 0.18,
      "learning_rate": 4.953089215006026e-05,
      "loss": 1.048,
      "step": 11500
    },
    {
      "epoch": 0.19,
      "learning_rate": 4.950480741641147e-05,
      "loss": 1.0178,
      "step": 12000
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.947872268276269e-05,
      "loss": 1.0335,
      "step": 12500
    },
    {
      "epoch": 0.2,
      "learning_rate": 4.9452637949113905e-05,
      "loss": 1.031,
      "step": 13000
    },
    {
      "epoch": 0.21,
      "learning_rate": 4.9426605384932414e-05,
      "loss": 1.0121,
      "step": 13500
    },
    {
      "epoch": 0.22,
      "learning_rate": 4.9400624990218224e-05,
      "loss": 1.0188,
      "step": 14000
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.937454025656944e-05,
      "loss": 0.9988,
      "step": 14500
    },
    {
      "epoch": 0.23,
      "learning_rate": 4.9348455522920656e-05,
      "loss": 1.0051,
      "step": 15000
    },
    {
      "epoch": 0.24,
      "learning_rate": 4.932237078927187e-05,
      "loss": 0.9967,
      "step": 15500
    },
    {
      "epoch": 0.25,
      "learning_rate": 4.929628605562309e-05,
      "loss": 0.9931,
      "step": 16000
    },
    {
      "epoch": 0.26,
      "learning_rate": 4.9270201321974304e-05,
      "loss": 0.9833,
      "step": 16500
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.924411658832552e-05,
      "loss": 0.9801,
      "step": 17000
    },
    {
      "epoch": 0.27,
      "learning_rate": 4.9218031854676735e-05,
      "loss": 0.9848,
      "step": 17500
    },
    {
      "epoch": 0.28,
      "learning_rate": 4.919194712102795e-05,
      "loss": 0.9774,
      "step": 18000
    },
    {
      "epoch": 0.29,
      "learning_rate": 4.916586238737916e-05,
      "loss": 0.9867,
      "step": 18500
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.9139777653730376e-05,
      "loss": 0.9738,
      "step": 19000
    },
    {
      "epoch": 0.3,
      "learning_rate": 4.911369292008159e-05,
      "loss": 0.9593,
      "step": 19500
    },
    {
      "epoch": 0.31,
      "learning_rate": 4.908760818643281e-05,
      "loss": 0.9804,
      "step": 20000
    },
    {
      "epoch": 0.32,
      "learning_rate": 4.9061523452784024e-05,
      "loss": 0.9654,
      "step": 20500
    },
    {
      "epoch": 0.33,
      "learning_rate": 4.903543871913524e-05,
      "loss": 0.9641,
      "step": 21000
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.9009353985486456e-05,
      "loss": 0.955,
      "step": 21500
    },
    {
      "epoch": 0.34,
      "learning_rate": 4.8983321421304965e-05,
      "loss": 0.953,
      "step": 22000
    },
    {
      "epoch": 0.35,
      "learning_rate": 4.895723668765618e-05,
      "loss": 0.9585,
      "step": 22500
    },
    {
      "epoch": 0.36,
      "learning_rate": 4.89311519540074e-05,
      "loss": 0.9453,
      "step": 23000
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.890506722035861e-05,
      "loss": 0.9519,
      "step": 23500
    },
    {
      "epoch": 0.37,
      "learning_rate": 4.887898248670983e-05,
      "loss": 0.9512,
      "step": 24000
    },
    {
      "epoch": 0.38,
      "learning_rate": 4.8852897753061045e-05,
      "loss": 0.9388,
      "step": 24500
    },
    {
      "epoch": 0.39,
      "learning_rate": 4.882686518887956e-05,
      "loss": 0.9257,
      "step": 25000
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.880078045523077e-05,
      "loss": 0.9346,
      "step": 25500
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.8774695721581986e-05,
      "loss": 0.9399,
      "step": 26000
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.87486109879332e-05,
      "loss": 0.9269,
      "step": 26500
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.872252625428442e-05,
      "loss": 0.9239,
      "step": 27000
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.8696441520635634e-05,
      "loss": 0.9355,
      "step": 27500
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.867035678698685e-05,
      "loss": 0.9256,
      "step": 28000
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.864437639227266e-05,
      "loss": 0.9278,
      "step": 28500
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.8618343828091176e-05,
      "loss": 0.9259,
      "step": 29000
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.859225909444239e-05,
      "loss": 0.9181,
      "step": 29500
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.85661743607936e-05,
      "loss": 0.9189,
      "step": 30000
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.854008962714482e-05,
      "loss": 0.9094,
      "step": 30500
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.851400489349603e-05,
      "loss": 0.9121,
      "step": 31000
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.848792015984725e-05,
      "loss": 0.8966,
      "step": 31500
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.8461835426198464e-05,
      "loss": 0.9144,
      "step": 32000
    },
    {
      "epoch": 0.51,
      "learning_rate": 4.843575069254968e-05,
      "loss": 0.9067,
      "step": 32500
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.8409665958900896e-05,
      "loss": 0.9144,
      "step": 33000
    },
    {
      "epoch": 0.52,
      "learning_rate": 4.838358122525211e-05,
      "loss": 0.9011,
      "step": 33500
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.835749649160333e-05,
      "loss": 0.8886,
      "step": 34000
    },
    {
      "epoch": 0.54,
      "learning_rate": 4.8331411757954544e-05,
      "loss": 0.9009,
      "step": 34500
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.830532702430576e-05,
      "loss": 0.8945,
      "step": 35000
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.8279242290656976e-05,
      "loss": 0.893,
      "step": 35500
    },
    {
      "epoch": 0.56,
      "learning_rate": 4.825315755700819e-05,
      "loss": 0.9099,
      "step": 36000
    },
    {
      "epoch": 0.57,
      "learning_rate": 4.82270728233594e-05,
      "loss": 0.8861,
      "step": 36500
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.820098808971062e-05,
      "loss": 0.8945,
      "step": 37000
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.817490335606183e-05,
      "loss": 0.8851,
      "step": 37500
    },
    {
      "epoch": 0.59,
      "learning_rate": 4.814881862241305e-05,
      "loss": 0.8827,
      "step": 38000
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.8122733888764264e-05,
      "loss": 0.8887,
      "step": 38500
    },
    {
      "epoch": 0.61,
      "learning_rate": 4.809664915511548e-05,
      "loss": 0.8785,
      "step": 39000
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.8070564421466696e-05,
      "loss": 0.8792,
      "step": 39500
    },
    {
      "epoch": 0.62,
      "learning_rate": 4.804447968781791e-05,
      "loss": 0.8724,
      "step": 40000
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.801839495416913e-05,
      "loss": 0.8725,
      "step": 40500
    },
    {
      "epoch": 0.64,
      "learning_rate": 4.799236238998764e-05,
      "loss": 0.8745,
      "step": 41000
    },
    {
      "epoch": 0.65,
      "learning_rate": 4.7966277656338853e-05,
      "loss": 0.8696,
      "step": 41500
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.794029726162466e-05,
      "loss": 0.8795,
      "step": 42000
    },
    {
      "epoch": 0.66,
      "learning_rate": 4.791421252797588e-05,
      "loss": 0.8744,
      "step": 42500
    },
    {
      "epoch": 0.67,
      "learning_rate": 4.7888127794327095e-05,
      "loss": 0.8728,
      "step": 43000
    },
    {
      "epoch": 0.68,
      "learning_rate": 4.786204306067831e-05,
      "loss": 0.8772,
      "step": 43500
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.783595832702952e-05,
      "loss": 0.8685,
      "step": 44000
    },
    {
      "epoch": 0.69,
      "learning_rate": 4.7809873593380736e-05,
      "loss": 0.8698,
      "step": 44500
    },
    {
      "epoch": 0.7,
      "learning_rate": 4.778378885973195e-05,
      "loss": 0.8696,
      "step": 45000
    },
    {
      "epoch": 0.71,
      "learning_rate": 4.775770412608317e-05,
      "loss": 0.8573,
      "step": 45500
    },
    {
      "epoch": 0.72,
      "learning_rate": 4.7731619392434384e-05,
      "loss": 0.8623,
      "step": 46000
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.77055346587856e-05,
      "loss": 0.8529,
      "step": 46500
    },
    {
      "epoch": 0.73,
      "learning_rate": 4.7679449925136815e-05,
      "loss": 0.8704,
      "step": 47000
    },
    {
      "epoch": 0.74,
      "learning_rate": 4.7653417360955325e-05,
      "loss": 0.8549,
      "step": 47500
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.762733262730654e-05,
      "loss": 0.8648,
      "step": 48000
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.760124789365776e-05,
      "loss": 0.8603,
      "step": 48500
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.757516316000897e-05,
      "loss": 0.876,
      "step": 49000
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.754913059582749e-05,
      "loss": 0.8657,
      "step": 49500
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.7523045862178705e-05,
      "loss": 0.8569,
      "step": 50000
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.749696112852992e-05,
      "loss": 0.8561,
      "step": 50500
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.747092856434843e-05,
      "loss": 0.8585,
      "step": 51000
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.7444843830699646e-05,
      "loss": 0.8592,
      "step": 51500
    },
    {
      "epoch": 0.81,
      "learning_rate": 4.741875909705086e-05,
      "loss": 0.859,
      "step": 52000
    },
    {
      "epoch": 0.82,
      "learning_rate": 4.739267436340208e-05,
      "loss": 0.844,
      "step": 52500
    },
    {
      "epoch": 0.83,
      "learning_rate": 4.7366589629753294e-05,
      "loss": 0.8492,
      "step": 53000
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.734050489610451e-05,
      "loss": 0.8473,
      "step": 53500
    },
    {
      "epoch": 0.84,
      "learning_rate": 4.7314420162455726e-05,
      "loss": 0.8435,
      "step": 54000
    },
    {
      "epoch": 0.85,
      "learning_rate": 4.7288387598274235e-05,
      "loss": 0.8472,
      "step": 54500
    },
    {
      "epoch": 0.86,
      "learning_rate": 4.726230286462545e-05,
      "loss": 0.8356,
      "step": 55000
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.723621813097667e-05,
      "loss": 0.8386,
      "step": 55500
    },
    {
      "epoch": 0.87,
      "learning_rate": 4.721013339732788e-05,
      "loss": 0.8414,
      "step": 56000
    },
    {
      "epoch": 0.88,
      "learning_rate": 4.71840486636791e-05,
      "loss": 0.8404,
      "step": 56500
    },
    {
      "epoch": 0.89,
      "learning_rate": 4.7157963930030314e-05,
      "loss": 0.8419,
      "step": 57000
    },
    {
      "epoch": 0.9,
      "learning_rate": 4.7131931365848824e-05,
      "loss": 0.8479,
      "step": 57500
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.710584663220004e-05,
      "loss": 0.8432,
      "step": 58000
    },
    {
      "epoch": 0.91,
      "learning_rate": 4.7079761898551256e-05,
      "loss": 0.8391,
      "step": 58500
    },
    {
      "epoch": 0.92,
      "learning_rate": 4.705367716490247e-05,
      "loss": 0.8262,
      "step": 59000
    },
    {
      "epoch": 0.93,
      "learning_rate": 4.702759243125369e-05,
      "loss": 0.8401,
      "step": 59500
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.7001507697604903e-05,
      "loss": 0.8386,
      "step": 60000
    },
    {
      "epoch": 0.94,
      "learning_rate": 4.697542296395612e-05,
      "loss": 0.831,
      "step": 60500
    },
    {
      "epoch": 0.95,
      "learning_rate": 4.694939039977463e-05,
      "loss": 0.8343,
      "step": 61000
    },
    {
      "epoch": 0.96,
      "learning_rate": 4.6923305666125845e-05,
      "loss": 0.8212,
      "step": 61500
    },
    {
      "epoch": 0.97,
      "learning_rate": 4.689722093247706e-05,
      "loss": 0.826,
      "step": 62000
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.6871136198828277e-05,
      "loss": 0.833,
      "step": 62500
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.684505146517949e-05,
      "loss": 0.8231,
      "step": 63000
    },
    {
      "epoch": 0.99,
      "learning_rate": 4.681896673153071e-05,
      "loss": 0.8402,
      "step": 63500
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.679293416734922e-05,
      "loss": 0.8286,
      "step": 64000
    },
    {
      "epoch": 1.0,
      "eval_bleu": 50.2043,
      "eval_gen_len": 15.0703,
      "eval_loss": 0.8096853494644165,
      "eval_runtime": 9598.1646,
      "eval_samples_per_second": 13.325,
      "eval_steps_per_second": 1.666,
      "step": 64061
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.6766849433700434e-05,
      "loss": 0.8088,
      "step": 64500
    },
    {
      "epoch": 1.01,
      "learning_rate": 4.674076470005165e-05,
      "loss": 0.8023,
      "step": 65000
    },
    {
      "epoch": 1.02,
      "learning_rate": 4.6714679966402866e-05,
      "loss": 0.8022,
      "step": 65500
    },
    {
      "epoch": 1.03,
      "learning_rate": 4.668859523275408e-05,
      "loss": 0.7841,
      "step": 66000
    },
    {
      "epoch": 1.04,
      "learning_rate": 4.66625104991053e-05,
      "loss": 0.7722,
      "step": 66500
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.6636477934923814e-05,
      "loss": 0.7627,
      "step": 67000
    },
    {
      "epoch": 1.05,
      "learning_rate": 4.661039320127503e-05,
      "loss": 0.7635,
      "step": 67500
    },
    {
      "epoch": 1.06,
      "learning_rate": 4.6584308467626245e-05,
      "loss": 0.7612,
      "step": 68000
    },
    {
      "epoch": 1.07,
      "learning_rate": 4.6558223733977454e-05,
      "loss": 0.7563,
      "step": 68500
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.653213900032867e-05,
      "loss": 0.7468,
      "step": 69000
    },
    {
      "epoch": 1.08,
      "learning_rate": 4.650605426667988e-05,
      "loss": 0.7534,
      "step": 69500
    },
    {
      "epoch": 1.09,
      "learning_rate": 4.6479969533031095e-05,
      "loss": 0.7457,
      "step": 70000
    },
    {
      "epoch": 1.1,
      "learning_rate": 4.645388479938231e-05,
      "loss": 0.7376,
      "step": 70500
    },
    {
      "epoch": 1.11,
      "learning_rate": 4.642780006573353e-05,
      "loss": 0.7377,
      "step": 71000
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.640171533208474e-05,
      "loss": 0.7385,
      "step": 71500
    },
    {
      "epoch": 1.12,
      "learning_rate": 4.637563059843596e-05,
      "loss": 0.7483,
      "step": 72000
    },
    {
      "epoch": 1.13,
      "learning_rate": 4.6349545864787175e-05,
      "loss": 0.7356,
      "step": 72500
    },
    {
      "epoch": 1.14,
      "learning_rate": 4.632346113113839e-05,
      "loss": 0.7303,
      "step": 73000
    },
    {
      "epoch": 1.15,
      "learning_rate": 4.629742856695691e-05,
      "loss": 0.7383,
      "step": 73500
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.627134383330812e-05,
      "loss": 0.7295,
      "step": 74000
    },
    {
      "epoch": 1.16,
      "learning_rate": 4.624531126912663e-05,
      "loss": 0.7284,
      "step": 74500
    },
    {
      "epoch": 1.17,
      "learning_rate": 4.621922653547785e-05,
      "loss": 0.72,
      "step": 75000
    },
    {
      "epoch": 1.18,
      "learning_rate": 4.6193141801829064e-05,
      "loss": 0.7246,
      "step": 75500
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.616705706818028e-05,
      "loss": 0.7091,
      "step": 76000
    },
    {
      "epoch": 1.19,
      "learning_rate": 4.6140972334531496e-05,
      "loss": 0.7227,
      "step": 76500
    },
    {
      "epoch": 1.2,
      "learning_rate": 4.611488760088271e-05,
      "loss": 0.722,
      "step": 77000
    },
    {
      "epoch": 1.21,
      "learning_rate": 4.608880286723393e-05,
      "loss": 0.7167,
      "step": 77500
    },
    {
      "epoch": 1.22,
      "learning_rate": 4.6062718133585144e-05,
      "loss": 0.7211,
      "step": 78000
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.603673773887095e-05,
      "loss": 0.7076,
      "step": 78500
    },
    {
      "epoch": 1.23,
      "learning_rate": 4.601065300522216e-05,
      "loss": 0.7132,
      "step": 79000
    },
    {
      "epoch": 1.24,
      "learning_rate": 4.598456827157338e-05,
      "loss": 0.7142,
      "step": 79500
    },
    {
      "epoch": 1.25,
      "learning_rate": 4.5958483537924594e-05,
      "loss": 0.7061,
      "step": 80000
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.593239880427581e-05,
      "loss": 0.704,
      "step": 80500
    },
    {
      "epoch": 1.26,
      "learning_rate": 4.5906314070627026e-05,
      "loss": 0.6912,
      "step": 81000
    },
    {
      "epoch": 1.27,
      "learning_rate": 4.588022933697824e-05,
      "loss": 0.7105,
      "step": 81500
    },
    {
      "epoch": 1.28,
      "learning_rate": 4.585414460332946e-05,
      "loss": 0.7057,
      "step": 82000
    },
    {
      "epoch": 1.29,
      "learning_rate": 4.5828059869680674e-05,
      "loss": 0.7093,
      "step": 82500
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.580197513603189e-05,
      "loss": 0.7089,
      "step": 83000
    },
    {
      "epoch": 1.3,
      "learning_rate": 4.5775890402383106e-05,
      "loss": 0.6892,
      "step": 83500
    },
    {
      "epoch": 1.31,
      "learning_rate": 4.574980566873432e-05,
      "loss": 0.7142,
      "step": 84000
    },
    {
      "epoch": 1.32,
      "learning_rate": 4.572377310455283e-05,
      "loss": 0.7039,
      "step": 84500
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.569768837090405e-05,
      "loss": 0.6999,
      "step": 85000
    },
    {
      "epoch": 1.33,
      "learning_rate": 4.5671655806722557e-05,
      "loss": 0.698,
      "step": 85500
    },
    {
      "epoch": 1.34,
      "learning_rate": 4.564557107307377e-05,
      "loss": 0.6956,
      "step": 86000
    },
    {
      "epoch": 1.35,
      "learning_rate": 4.561948633942499e-05,
      "loss": 0.7012,
      "step": 86500
    },
    {
      "epoch": 1.36,
      "learning_rate": 4.5593401605776204e-05,
      "loss": 0.6933,
      "step": 87000
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.556736904159472e-05,
      "loss": 0.6986,
      "step": 87500
    },
    {
      "epoch": 1.37,
      "learning_rate": 4.5541284307945936e-05,
      "loss": 0.6973,
      "step": 88000
    },
    {
      "epoch": 1.38,
      "learning_rate": 4.5515251743764446e-05,
      "loss": 0.6912,
      "step": 88500
    },
    {
      "epoch": 1.39,
      "learning_rate": 4.548916701011566e-05,
      "loss": 0.6842,
      "step": 89000
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.546313444593418e-05,
      "loss": 0.6836,
      "step": 89500
    },
    {
      "epoch": 1.4,
      "learning_rate": 4.543704971228539e-05,
      "loss": 0.6926,
      "step": 90000
    },
    {
      "epoch": 1.41,
      "learning_rate": 4.54109649786366e-05,
      "loss": 0.6785,
      "step": 90500
    },
    {
      "epoch": 1.42,
      "learning_rate": 4.538488024498782e-05,
      "loss": 0.6801,
      "step": 91000
    },
    {
      "epoch": 1.43,
      "learning_rate": 4.5358795511339035e-05,
      "loss": 0.6837,
      "step": 91500
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.533271077769025e-05,
      "loss": 0.6895,
      "step": 92000
    },
    {
      "epoch": 1.44,
      "learning_rate": 4.5306626044041467e-05,
      "loss": 0.688,
      "step": 92500
    },
    {
      "epoch": 1.45,
      "learning_rate": 4.528054131039268e-05,
      "loss": 0.6849,
      "step": 93000
    },
    {
      "epoch": 1.46,
      "learning_rate": 4.52544565767439e-05,
      "loss": 0.6814,
      "step": 93500
    },
    {
      "epoch": 1.47,
      "learning_rate": 4.5228371843095114e-05,
      "loss": 0.6862,
      "step": 94000
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.5202339278913624e-05,
      "loss": 0.6727,
      "step": 94500
    },
    {
      "epoch": 1.48,
      "learning_rate": 4.517625454526484e-05,
      "loss": 0.6807,
      "step": 95000
    },
    {
      "epoch": 1.49,
      "learning_rate": 4.5150169811616056e-05,
      "loss": 0.6669,
      "step": 95500
    },
    {
      "epoch": 1.5,
      "learning_rate": 4.512408507796727e-05,
      "loss": 0.6765,
      "step": 96000
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.509800034431849e-05,
      "loss": 0.6792,
      "step": 96500
    },
    {
      "epoch": 1.51,
      "learning_rate": 4.50719156106697e-05,
      "loss": 0.6803,
      "step": 97000
    },
    {
      "epoch": 1.52,
      "learning_rate": 4.504583087702092e-05,
      "loss": 0.6692,
      "step": 97500
    },
    {
      "epoch": 1.53,
      "learning_rate": 4.5019746143372135e-05,
      "loss": 0.6671,
      "step": 98000
    },
    {
      "epoch": 1.54,
      "learning_rate": 4.499366140972335e-05,
      "loss": 0.6712,
      "step": 98500
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.496757667607456e-05,
      "loss": 0.6729,
      "step": 99000
    },
    {
      "epoch": 1.55,
      "learning_rate": 4.4941544111893076e-05,
      "loss": 0.6715,
      "step": 99500
    },
    {
      "epoch": 1.56,
      "learning_rate": 4.491545937824429e-05,
      "loss": 0.6821,
      "step": 100000
    },
    {
      "epoch": 1.57,
      "learning_rate": 4.488937464459551e-05,
      "loss": 0.6641,
      "step": 100500
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.4863289910946724e-05,
      "loss": 0.673,
      "step": 101000
    },
    {
      "epoch": 1.58,
      "learning_rate": 4.483720517729793e-05,
      "loss": 0.6686,
      "step": 101500
    },
    {
      "epoch": 1.59,
      "learning_rate": 4.481112044364915e-05,
      "loss": 0.6614,
      "step": 102000
    },
    {
      "epoch": 1.6,
      "learning_rate": 4.4785035710000365e-05,
      "loss": 0.6708,
      "step": 102500
    },
    {
      "epoch": 1.61,
      "learning_rate": 4.475895097635158e-05,
      "loss": 0.6621,
      "step": 103000
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.47328662427028e-05,
      "loss": 0.6618,
      "step": 103500
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.4706833678521306e-05,
      "loss": 0.6647,
      "step": 104000
    },
    {
      "epoch": 1.63,
      "learning_rate": 4.468074894487252e-05,
      "loss": 0.6587,
      "step": 104500
    },
    {
      "epoch": 1.64,
      "learning_rate": 4.465466421122374e-05,
      "loss": 0.6589,
      "step": 105000
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.4628579477574954e-05,
      "loss": 0.654,
      "step": 105500
    },
    {
      "epoch": 1.65,
      "learning_rate": 4.460249474392617e-05,
      "loss": 0.6614,
      "step": 106000
    },
    {
      "epoch": 1.66,
      "learning_rate": 4.4576410010277386e-05,
      "loss": 0.6663,
      "step": 106500
    },
    {
      "epoch": 1.67,
      "learning_rate": 4.45503252766286e-05,
      "loss": 0.6612,
      "step": 107000
    },
    {
      "epoch": 1.68,
      "learning_rate": 4.452424054297982e-05,
      "loss": 0.6594,
      "step": 107500
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.449820797879833e-05,
      "loss": 0.6584,
      "step": 108000
    },
    {
      "epoch": 1.69,
      "learning_rate": 4.447217541461684e-05,
      "loss": 0.6592,
      "step": 108500
    },
    {
      "epoch": 1.7,
      "learning_rate": 4.444609068096806e-05,
      "loss": 0.6595,
      "step": 109000
    },
    {
      "epoch": 1.71,
      "learning_rate": 4.4420005947319275e-05,
      "loss": 0.6545,
      "step": 109500
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.439392121367049e-05,
      "loss": 0.6543,
      "step": 110000
    },
    {
      "epoch": 1.72,
      "learning_rate": 4.4367888649489e-05,
      "loss": 0.6509,
      "step": 110500
    },
    {
      "epoch": 1.73,
      "learning_rate": 4.4341803915840216e-05,
      "loss": 0.6641,
      "step": 111000
    },
    {
      "epoch": 1.74,
      "learning_rate": 4.431577135165873e-05,
      "loss": 0.6514,
      "step": 111500
    },
    {
      "epoch": 1.75,
      "learning_rate": 4.428968661800995e-05,
      "loss": 0.6588,
      "step": 112000
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.4263601884361164e-05,
      "loss": 0.6565,
      "step": 112500
    },
    {
      "epoch": 1.76,
      "learning_rate": 4.4237517150712373e-05,
      "loss": 0.6654,
      "step": 113000
    },
    {
      "epoch": 1.77,
      "learning_rate": 4.421143241706359e-05,
      "loss": 0.6642,
      "step": 113500
    },
    {
      "epoch": 1.78,
      "learning_rate": 4.4185347683414805e-05,
      "loss": 0.6505,
      "step": 114000
    },
    {
      "epoch": 1.79,
      "learning_rate": 4.415926294976602e-05,
      "loss": 0.6561,
      "step": 114500
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.413317821611724e-05,
      "loss": 0.6555,
      "step": 115000
    },
    {
      "epoch": 1.8,
      "learning_rate": 4.410709348246845e-05,
      "loss": 0.6597,
      "step": 115500
    },
    {
      "epoch": 1.81,
      "learning_rate": 4.408100874881967e-05,
      "loss": 0.6635,
      "step": 116000
    },
    {
      "epoch": 1.82,
      "learning_rate": 4.405497618463818e-05,
      "loss": 0.6469,
      "step": 116500
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.4028891450989394e-05,
      "loss": 0.6466,
      "step": 117000
    },
    {
      "epoch": 1.83,
      "learning_rate": 4.400280671734061e-05,
      "loss": 0.6504,
      "step": 117500
    },
    {
      "epoch": 1.84,
      "learning_rate": 4.3976721983691826e-05,
      "loss": 0.6463,
      "step": 118000
    },
    {
      "epoch": 1.85,
      "learning_rate": 4.395063725004304e-05,
      "loss": 0.6505,
      "step": 118500
    },
    {
      "epoch": 1.86,
      "learning_rate": 4.392455251639426e-05,
      "loss": 0.6392,
      "step": 119000
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.3898467782745474e-05,
      "loss": 0.6446,
      "step": 119500
    },
    {
      "epoch": 1.87,
      "learning_rate": 4.387238304909669e-05,
      "loss": 0.6431,
      "step": 120000
    },
    {
      "epoch": 1.88,
      "learning_rate": 4.38463504849152e-05,
      "loss": 0.6452,
      "step": 120500
    },
    {
      "epoch": 1.89,
      "learning_rate": 4.3820265751266415e-05,
      "loss": 0.6505,
      "step": 121000
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.379418101761763e-05,
      "loss": 0.6487,
      "step": 121500
    },
    {
      "epoch": 1.9,
      "learning_rate": 4.376809628396885e-05,
      "loss": 0.6481,
      "step": 122000
    },
    {
      "epoch": 1.91,
      "learning_rate": 4.3742063719787356e-05,
      "loss": 0.648,
      "step": 122500
    },
    {
      "epoch": 1.92,
      "learning_rate": 4.371597898613857e-05,
      "loss": 0.6378,
      "step": 123000
    },
    {
      "epoch": 1.93,
      "learning_rate": 4.368989425248979e-05,
      "loss": 0.6457,
      "step": 123500
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.3663809518841004e-05,
      "loss": 0.6468,
      "step": 124000
    },
    {
      "epoch": 1.94,
      "learning_rate": 4.363772478519222e-05,
      "loss": 0.6456,
      "step": 124500
    },
    {
      "epoch": 1.95,
      "learning_rate": 4.3611692221010736e-05,
      "loss": 0.6392,
      "step": 125000
    },
    {
      "epoch": 1.96,
      "learning_rate": 4.358560748736195e-05,
      "loss": 0.6388,
      "step": 125500
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.355952275371316e-05,
      "loss": 0.6382,
      "step": 126000
    },
    {
      "epoch": 1.97,
      "learning_rate": 4.353343802006438e-05,
      "loss": 0.6397,
      "step": 126500
    },
    {
      "epoch": 1.98,
      "learning_rate": 4.350735328641559e-05,
      "loss": 0.6368,
      "step": 127000
    },
    {
      "epoch": 1.99,
      "learning_rate": 4.348126855276681e-05,
      "loss": 0.6516,
      "step": 127500
    },
    {
      "epoch": 2.0,
      "learning_rate": 4.3455235988585325e-05,
      "loss": 0.6429,
      "step": 128000
    },
    {
      "epoch": 2.0,
      "eval_bleu": 51.8273,
      "eval_gen_len": 15.0439,
      "eval_loss": 0.7786535620689392,
      "eval_runtime": 9521.3959,
      "eval_samples_per_second": 13.433,
      "eval_steps_per_second": 1.679,
      "step": 128123
    },
    {
      "epoch": 2.01,
      "learning_rate": 4.342915125493654e-05,
      "loss": 0.6315,
      "step": 128500
    },
    {
      "epoch": 2.01,
      "learning_rate": 4.340306652128776e-05,
      "loss": 0.6169,
      "step": 129000
    },
    {
      "epoch": 2.02,
      "learning_rate": 4.337698178763897e-05,
      "loss": 0.6218,
      "step": 129500
    },
    {
      "epoch": 2.03,
      "learning_rate": 4.335089705399019e-05,
      "loss": 0.6108,
      "step": 130000
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.33248644898087e-05,
      "loss": 0.6101,
      "step": 130500
    },
    {
      "epoch": 2.04,
      "learning_rate": 4.3298779756159914e-05,
      "loss": 0.5922,
      "step": 131000
    },
    {
      "epoch": 2.05,
      "learning_rate": 4.327269502251113e-05,
      "loss": 0.6026,
      "step": 131500
    },
    {
      "epoch": 2.06,
      "learning_rate": 4.3246610288862346e-05,
      "loss": 0.591,
      "step": 132000
    },
    {
      "epoch": 2.07,
      "learning_rate": 4.322052555521356e-05,
      "loss": 0.5985,
      "step": 132500
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.319444082156478e-05,
      "loss": 0.5816,
      "step": 133000
    },
    {
      "epoch": 2.08,
      "learning_rate": 4.316840825738329e-05,
      "loss": 0.5942,
      "step": 133500
    },
    {
      "epoch": 2.09,
      "learning_rate": 4.31423235237345e-05,
      "loss": 0.5905,
      "step": 134000
    },
    {
      "epoch": 2.1,
      "learning_rate": 4.311623879008572e-05,
      "loss": 0.5845,
      "step": 134500
    },
    {
      "epoch": 2.11,
      "learning_rate": 4.3090154056436935e-05,
      "loss": 0.5799,
      "step": 135000
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.306406932278815e-05,
      "loss": 0.5847,
      "step": 135500
    },
    {
      "epoch": 2.12,
      "learning_rate": 4.303798458913936e-05,
      "loss": 0.5866,
      "step": 136000
    },
    {
      "epoch": 2.13,
      "learning_rate": 4.3011899855490576e-05,
      "loss": 0.5867,
      "step": 136500
    },
    {
      "epoch": 2.14,
      "learning_rate": 4.298581512184179e-05,
      "loss": 0.5739,
      "step": 137000
    },
    {
      "epoch": 2.15,
      "learning_rate": 4.295973038819301e-05,
      "loss": 0.5882,
      "step": 137500
    },
    {
      "epoch": 2.15,
      "learning_rate": 4.2933697824011524e-05,
      "loss": 0.576,
      "step": 138000
    },
    {
      "epoch": 2.16,
      "learning_rate": 4.290761309036273e-05,
      "loss": 0.5654,
      "step": 138500
    },
    {
      "epoch": 2.17,
      "learning_rate": 4.288152835671395e-05,
      "loss": 0.5708,
      "step": 139000
    },
    {
      "epoch": 2.18,
      "learning_rate": 4.2855443623065165e-05,
      "loss": 0.5725,
      "step": 139500
    },
    {
      "epoch": 2.19,
      "learning_rate": 4.282935888941638e-05,
      "loss": 0.5687,
      "step": 140000
    },
    {
      "epoch": 2.19,
      "learning_rate": 4.28033263252349e-05,
      "loss": 0.5627,
      "step": 140500
    },
    {
      "epoch": 2.2,
      "learning_rate": 4.2777293761053406e-05,
      "loss": 0.5745,
      "step": 141000
    },
    {
      "epoch": 2.21,
      "learning_rate": 4.275120902740462e-05,
      "loss": 0.5617,
      "step": 141500
    },
    {
      "epoch": 2.22,
      "learning_rate": 4.272512429375584e-05,
      "loss": 0.5687,
      "step": 142000
    },
    {
      "epoch": 2.22,
      "learning_rate": 4.2699039560107054e-05,
      "loss": 0.5682,
      "step": 142500
    },
    {
      "epoch": 2.23,
      "learning_rate": 4.267295482645827e-05,
      "loss": 0.5573,
      "step": 143000
    },
    {
      "epoch": 2.24,
      "learning_rate": 4.264687009280948e-05,
      "loss": 0.5729,
      "step": 143500
    },
    {
      "epoch": 2.25,
      "learning_rate": 4.2620785359160695e-05,
      "loss": 0.5602,
      "step": 144000
    },
    {
      "epoch": 2.26,
      "learning_rate": 4.259470062551191e-05,
      "loss": 0.553,
      "step": 144500
    },
    {
      "epoch": 2.26,
      "learning_rate": 4.256861589186313e-05,
      "loss": 0.5501,
      "step": 145000
    },
    {
      "epoch": 2.27,
      "learning_rate": 4.254253115821434e-05,
      "loss": 0.5646,
      "step": 145500
    },
    {
      "epoch": 2.28,
      "learning_rate": 4.251644642456556e-05,
      "loss": 0.5609,
      "step": 146000
    },
    {
      "epoch": 2.29,
      "learning_rate": 4.2490361690916775e-05,
      "loss": 0.5575,
      "step": 146500
    },
    {
      "epoch": 2.29,
      "learning_rate": 4.246432912673529e-05,
      "loss": 0.5687,
      "step": 147000
    },
    {
      "epoch": 2.3,
      "learning_rate": 4.24382443930865e-05,
      "loss": 0.5425,
      "step": 147500
    },
    {
      "epoch": 2.31,
      "learning_rate": 4.2412159659437716e-05,
      "loss": 0.5689,
      "step": 148000
    },
    {
      "epoch": 2.32,
      "learning_rate": 4.238607492578893e-05,
      "loss": 0.557,
      "step": 148500
    },
    {
      "epoch": 2.33,
      "learning_rate": 4.235999019214015e-05,
      "loss": 0.5557,
      "step": 149000
    },
    {
      "epoch": 2.33,
      "learning_rate": 4.2333905458491364e-05,
      "loss": 0.5629,
      "step": 149500
    },
    {
      "epoch": 2.34,
      "learning_rate": 4.230787289430988e-05,
      "loss": 0.556,
      "step": 150000
    },
    {
      "epoch": 2.35,
      "learning_rate": 4.228184033012839e-05,
      "loss": 0.5521,
      "step": 150500
    },
    {
      "epoch": 2.36,
      "learning_rate": 4.2255807765946905e-05,
      "loss": 0.5508,
      "step": 151000
    },
    {
      "epoch": 2.36,
      "learning_rate": 4.2229775201765415e-05,
      "loss": 0.5539,
      "step": 151500
    },
    {
      "epoch": 2.37,
      "learning_rate": 4.220369046811663e-05,
      "loss": 0.5601,
      "step": 152000
    },
    {
      "epoch": 2.38,
      "learning_rate": 4.217760573446785e-05,
      "loss": 0.5504,
      "step": 152500
    },
    {
      "epoch": 2.39,
      "learning_rate": 4.215152100081906e-05,
      "loss": 0.5475,
      "step": 153000
    },
    {
      "epoch": 2.4,
      "learning_rate": 4.212543626717028e-05,
      "loss": 0.5477,
      "step": 153500
    },
    {
      "epoch": 2.4,
      "learning_rate": 4.2099351533521494e-05,
      "loss": 0.5421,
      "step": 154000
    },
    {
      "epoch": 2.41,
      "learning_rate": 4.207326679987271e-05,
      "loss": 0.5374,
      "step": 154500
    },
    {
      "epoch": 2.42,
      "learning_rate": 4.204718206622392e-05,
      "loss": 0.5443,
      "step": 155000
    },
    {
      "epoch": 2.43,
      "learning_rate": 4.2021097332575135e-05,
      "loss": 0.547,
      "step": 155500
    },
    {
      "epoch": 2.44,
      "learning_rate": 4.199501259892635e-05,
      "loss": 0.5496,
      "step": 156000
    },
    {
      "epoch": 2.44,
      "learning_rate": 4.196892786527757e-05,
      "loss": 0.5472,
      "step": 156500
    },
    {
      "epoch": 2.45,
      "learning_rate": 4.194284313162878e-05,
      "loss": 0.5458,
      "step": 157000
    },
    {
      "epoch": 2.46,
      "learning_rate": 4.191675839798e-05,
      "loss": 0.5505,
      "step": 157500
    },
    {
      "epoch": 2.47,
      "learning_rate": 4.1890673664331215e-05,
      "loss": 0.5454,
      "step": 158000
    },
    {
      "epoch": 2.47,
      "learning_rate": 4.186458893068243e-05,
      "loss": 0.5341,
      "step": 158500
    },
    {
      "epoch": 2.48,
      "learning_rate": 4.183850419703365e-05,
      "loss": 0.5402,
      "step": 159000
    },
    {
      "epoch": 2.49,
      "learning_rate": 4.1812471632852156e-05,
      "loss": 0.5363,
      "step": 159500
    },
    {
      "epoch": 2.5,
      "learning_rate": 4.178643906867067e-05,
      "loss": 0.5358,
      "step": 160000
    },
    {
      "epoch": 2.51,
      "learning_rate": 4.176035433502189e-05,
      "loss": 0.5388,
      "step": 160500
    },
    {
      "epoch": 2.51,
      "learning_rate": 4.1734269601373104e-05,
      "loss": 0.5427,
      "step": 161000
    },
    {
      "epoch": 2.52,
      "learning_rate": 4.170818486772432e-05,
      "loss": 0.5325,
      "step": 161500
    },
    {
      "epoch": 2.53,
      "learning_rate": 4.1682100134075536e-05,
      "loss": 0.5375,
      "step": 162000
    },
    {
      "epoch": 2.54,
      "learning_rate": 4.1656015400426745e-05,
      "loss": 0.535,
      "step": 162500
    },
    {
      "epoch": 2.54,
      "learning_rate": 4.162993066677796e-05,
      "loss": 0.5287,
      "step": 163000
    },
    {
      "epoch": 2.55,
      "learning_rate": 4.160384593312918e-05,
      "loss": 0.5345,
      "step": 163500
    },
    {
      "epoch": 2.56,
      "learning_rate": 4.157776119948039e-05,
      "loss": 0.5411,
      "step": 164000
    },
    {
      "epoch": 2.57,
      "learning_rate": 4.155167646583161e-05,
      "loss": 0.5366,
      "step": 164500
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.152569607111742e-05,
      "loss": 0.5378,
      "step": 165000
    },
    {
      "epoch": 2.58,
      "learning_rate": 4.1499611337468634e-05,
      "loss": 0.5349,
      "step": 165500
    },
    {
      "epoch": 2.59,
      "learning_rate": 4.147352660381985e-05,
      "loss": 0.5279,
      "step": 166000
    },
    {
      "epoch": 2.6,
      "learning_rate": 4.1447441870171066e-05,
      "loss": 0.5324,
      "step": 166500
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.142135713652228e-05,
      "loss": 0.5326,
      "step": 167000
    },
    {
      "epoch": 2.61,
      "learning_rate": 4.139532457234079e-05,
      "loss": 0.5254,
      "step": 167500
    },
    {
      "epoch": 2.62,
      "learning_rate": 4.136923983869201e-05,
      "loss": 0.5317,
      "step": 168000
    },
    {
      "epoch": 2.63,
      "learning_rate": 4.134315510504322e-05,
      "loss": 0.5278,
      "step": 168500
    },
    {
      "epoch": 2.64,
      "learning_rate": 4.131707037139444e-05,
      "loss": 0.5223,
      "step": 169000
    },
    {
      "epoch": 2.65,
      "learning_rate": 4.129103780721295e-05,
      "loss": 0.5237,
      "step": 169500
    },
    {
      "epoch": 2.65,
      "learning_rate": 4.1265005243031465e-05,
      "loss": 0.5251,
      "step": 170000
    },
    {
      "epoch": 2.66,
      "learning_rate": 4.123892050938268e-05,
      "loss": 0.5345,
      "step": 170500
    },
    {
      "epoch": 2.67,
      "learning_rate": 4.12128357757339e-05,
      "loss": 0.5302,
      "step": 171000
    },
    {
      "epoch": 2.68,
      "learning_rate": 4.1186751042085106e-05,
      "loss": 0.5256,
      "step": 171500
    },
    {
      "epoch": 2.68,
      "learning_rate": 4.116066630843632e-05,
      "loss": 0.5231,
      "step": 172000
    },
    {
      "epoch": 2.69,
      "learning_rate": 4.113458157478754e-05,
      "loss": 0.5322,
      "step": 172500
    },
    {
      "epoch": 2.7,
      "learning_rate": 4.1108549010606054e-05,
      "loss": 0.5268,
      "step": 173000
    },
    {
      "epoch": 2.71,
      "learning_rate": 4.108246427695727e-05,
      "loss": 0.521,
      "step": 173500
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.1056379543308486e-05,
      "loss": 0.5179,
      "step": 174000
    },
    {
      "epoch": 2.72,
      "learning_rate": 4.10302948096597e-05,
      "loss": 0.5249,
      "step": 174500
    },
    {
      "epoch": 2.73,
      "learning_rate": 4.100421007601092e-05,
      "loss": 0.5267,
      "step": 175000
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.0978125342362133e-05,
      "loss": 0.5198,
      "step": 175500
    },
    {
      "epoch": 2.75,
      "learning_rate": 4.095204060871335e-05,
      "loss": 0.525,
      "step": 176000
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.0925955875064565e-05,
      "loss": 0.5236,
      "step": 176500
    },
    {
      "epoch": 2.76,
      "learning_rate": 4.0899871141415774e-05,
      "loss": 0.5345,
      "step": 177000
    },
    {
      "epoch": 2.77,
      "learning_rate": 4.087378640776699e-05,
      "loss": 0.5307,
      "step": 177500
    },
    {
      "epoch": 2.78,
      "learning_rate": 4.0847701674118206e-05,
      "loss": 0.5227,
      "step": 178000
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.082161694046942e-05,
      "loss": 0.523,
      "step": 178500
    },
    {
      "epoch": 2.79,
      "learning_rate": 4.079553220682064e-05,
      "loss": 0.5221,
      "step": 179000
    },
    {
      "epoch": 2.8,
      "learning_rate": 4.0769447473171854e-05,
      "loss": 0.5195,
      "step": 179500
    },
    {
      "epoch": 2.81,
      "learning_rate": 4.074336273952307e-05,
      "loss": 0.5366,
      "step": 180000
    },
    {
      "epoch": 2.82,
      "learning_rate": 4.071727800587428e-05,
      "loss": 0.5173,
      "step": 180500
    },
    {
      "epoch": 2.83,
      "learning_rate": 4.0691193272225495e-05,
      "loss": 0.5168,
      "step": 181000
    },
    {
      "epoch": 2.83,
      "learning_rate": 4.066516070804402e-05,
      "loss": 0.5227,
      "step": 181500
    },
    {
      "epoch": 2.84,
      "learning_rate": 4.063912814386253e-05,
      "loss": 0.5145,
      "step": 182000
    },
    {
      "epoch": 2.85,
      "learning_rate": 4.061304341021374e-05,
      "loss": 0.5162,
      "step": 182500
    },
    {
      "epoch": 2.86,
      "learning_rate": 4.058695867656496e-05,
      "loss": 0.5155,
      "step": 183000
    },
    {
      "epoch": 2.86,
      "learning_rate": 4.0560873942916175e-05,
      "loss": 0.5142,
      "step": 183500
    },
    {
      "epoch": 2.87,
      "learning_rate": 4.053478920926739e-05,
      "loss": 0.5168,
      "step": 184000
    },
    {
      "epoch": 2.88,
      "learning_rate": 4.05087044756186e-05,
      "loss": 0.5152,
      "step": 184500
    },
    {
      "epoch": 2.89,
      "learning_rate": 4.0482619741969816e-05,
      "loss": 0.5196,
      "step": 185000
    },
    {
      "epoch": 2.9,
      "learning_rate": 4.045653500832103e-05,
      "loss": 0.5173,
      "step": 185500
    },
    {
      "epoch": 2.9,
      "learning_rate": 4.043045027467225e-05,
      "loss": 0.5178,
      "step": 186000
    },
    {
      "epoch": 2.91,
      "learning_rate": 4.040446987995806e-05,
      "loss": 0.5186,
      "step": 186500
    },
    {
      "epoch": 2.92,
      "learning_rate": 4.037838514630927e-05,
      "loss": 0.5108,
      "step": 187000
    },
    {
      "epoch": 2.93,
      "learning_rate": 4.035230041266049e-05,
      "loss": 0.5198,
      "step": 187500
    },
    {
      "epoch": 2.93,
      "learning_rate": 4.0326215679011705e-05,
      "loss": 0.5184,
      "step": 188000
    },
    {
      "epoch": 2.94,
      "learning_rate": 4.030013094536292e-05,
      "loss": 0.5184,
      "step": 188500
    },
    {
      "epoch": 2.95,
      "learning_rate": 4.027409838118143e-05,
      "loss": 0.5123,
      "step": 189000
    },
    {
      "epoch": 2.96,
      "learning_rate": 4.0248013647532646e-05,
      "loss": 0.5031,
      "step": 189500
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.022192891388386e-05,
      "loss": 0.5148,
      "step": 190000
    },
    {
      "epoch": 2.97,
      "learning_rate": 4.019584418023508e-05,
      "loss": 0.5079,
      "step": 190500
    },
    {
      "epoch": 2.98,
      "learning_rate": 4.0169759446586294e-05,
      "loss": 0.5117,
      "step": 191000
    },
    {
      "epoch": 2.99,
      "learning_rate": 4.014367471293751e-05,
      "loss": 0.523,
      "step": 191500
    },
    {
      "epoch": 3.0,
      "learning_rate": 4.011758997928872e-05,
      "loss": 0.517,
      "step": 192000
    },
    {
      "epoch": 3.0,
      "eval_bleu": 52.0502,
      "eval_gen_len": 14.9976,
      "eval_loss": 0.8042312264442444,
      "eval_runtime": 9500.4244,
      "eval_samples_per_second": 13.462,
      "eval_steps_per_second": 1.683,
      "step": 192184
    },
    {
      "epoch": 3.0,
      "learning_rate": 4.0091557415107235e-05,
      "loss": 0.5007,
      "step": 192500
    },
    {
      "epoch": 3.01,
      "learning_rate": 4.006547268145845e-05,
      "loss": 0.4984,
      "step": 193000
    },
    {
      "epoch": 3.02,
      "learning_rate": 4.003938794780967e-05,
      "loss": 0.4934,
      "step": 193500
    },
    {
      "epoch": 3.03,
      "learning_rate": 4.001330321416088e-05,
      "loss": 0.4892,
      "step": 194000
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.998721848051209e-05,
      "loss": 0.4925,
      "step": 194500
    },
    {
      "epoch": 3.04,
      "learning_rate": 3.9961185916330615e-05,
      "loss": 0.4777,
      "step": 195000
    },
    {
      "epoch": 3.05,
      "learning_rate": 3.9935153352149125e-05,
      "loss": 0.4794,
      "step": 195500
    },
    {
      "epoch": 3.06,
      "learning_rate": 3.990906861850034e-05,
      "loss": 0.4773,
      "step": 196000
    },
    {
      "epoch": 3.07,
      "learning_rate": 3.9882983884851556e-05,
      "loss": 0.4829,
      "step": 196500
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.9856951320670066e-05,
      "loss": 0.4692,
      "step": 197000
    },
    {
      "epoch": 3.08,
      "learning_rate": 3.983086658702128e-05,
      "loss": 0.4776,
      "step": 197500
    },
    {
      "epoch": 3.09,
      "learning_rate": 3.98047818533725e-05,
      "loss": 0.472,
      "step": 198000
    },
    {
      "epoch": 3.1,
      "learning_rate": 3.9778697119723714e-05,
      "loss": 0.4704,
      "step": 198500
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.975261238607493e-05,
      "loss": 0.4599,
      "step": 199000
    },
    {
      "epoch": 3.11,
      "learning_rate": 3.9726527652426145e-05,
      "loss": 0.4698,
      "step": 199500
    },
    {
      "epoch": 3.12,
      "learning_rate": 3.970044291877736e-05,
      "loss": 0.4742,
      "step": 200000
    },
    {
      "epoch": 3.13,
      "learning_rate": 3.967435818512858e-05,
      "loss": 0.4715,
      "step": 200500
    },
    {
      "epoch": 3.14,
      "learning_rate": 3.9648273451479786e-05,
      "loss": 0.4599,
      "step": 201000
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.9622188717831e-05,
      "loss": 0.4671,
      "step": 201500
    },
    {
      "epoch": 3.15,
      "learning_rate": 3.959610398418222e-05,
      "loss": 0.4655,
      "step": 202000
    },
    {
      "epoch": 3.16,
      "learning_rate": 3.9570019250533434e-05,
      "loss": 0.4529,
      "step": 202500
    },
    {
      "epoch": 3.17,
      "learning_rate": 3.954393451688465e-05,
      "loss": 0.4639,
      "step": 203000
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.9517849783235866e-05,
      "loss": 0.451,
      "step": 203500
    },
    {
      "epoch": 3.18,
      "learning_rate": 3.949176504958708e-05,
      "loss": 0.4584,
      "step": 204000
    },
    {
      "epoch": 3.19,
      "learning_rate": 3.946573248540559e-05,
      "loss": 0.4503,
      "step": 204500
    },
    {
      "epoch": 3.2,
      "learning_rate": 3.943964775175681e-05,
      "loss": 0.4578,
      "step": 205000
    },
    {
      "epoch": 3.21,
      "learning_rate": 3.9413615187575323e-05,
      "loss": 0.4563,
      "step": 205500
    },
    {
      "epoch": 3.22,
      "learning_rate": 3.938753045392653e-05,
      "loss": 0.455,
      "step": 206000
    },
    {
      "epoch": 3.22,
      "learning_rate": 3.936144572027775e-05,
      "loss": 0.4595,
      "step": 206500
    },
    {
      "epoch": 3.23,
      "learning_rate": 3.9335360986628964e-05,
      "loss": 0.4473,
      "step": 207000
    },
    {
      "epoch": 3.24,
      "learning_rate": 3.930927625298018e-05,
      "loss": 0.4613,
      "step": 207500
    },
    {
      "epoch": 3.25,
      "learning_rate": 3.9283191519331396e-05,
      "loss": 0.4529,
      "step": 208000
    },
    {
      "epoch": 3.25,
      "learning_rate": 3.9257158955149906e-05,
      "loss": 0.4433,
      "step": 208500
    },
    {
      "epoch": 3.26,
      "learning_rate": 3.923107422150112e-05,
      "loss": 0.4402,
      "step": 209000
    },
    {
      "epoch": 3.27,
      "learning_rate": 3.920498948785234e-05,
      "loss": 0.4506,
      "step": 209500
    },
    {
      "epoch": 3.28,
      "learning_rate": 3.917890475420355e-05,
      "loss": 0.4544,
      "step": 210000
    },
    {
      "epoch": 3.29,
      "learning_rate": 3.915282002055477e-05,
      "loss": 0.4503,
      "step": 210500
    },
    {
      "epoch": 3.29,
      "learning_rate": 3.9126787456373285e-05,
      "loss": 0.4585,
      "step": 211000
    },
    {
      "epoch": 3.3,
      "learning_rate": 3.91007027227245e-05,
      "loss": 0.432,
      "step": 211500
    },
    {
      "epoch": 3.31,
      "learning_rate": 3.907461798907572e-05,
      "loss": 0.4579,
      "step": 212000
    },
    {
      "epoch": 3.32,
      "learning_rate": 3.904853325542693e-05,
      "loss": 0.4488,
      "step": 212500
    },
    {
      "epoch": 3.32,
      "learning_rate": 3.902244852177814e-05,
      "loss": 0.4514,
      "step": 213000
    },
    {
      "epoch": 3.33,
      "learning_rate": 3.899641595759666e-05,
      "loss": 0.4534,
      "step": 213500
    },
    {
      "epoch": 3.34,
      "learning_rate": 3.8970331223947874e-05,
      "loss": 0.4451,
      "step": 214000
    },
    {
      "epoch": 3.35,
      "learning_rate": 3.894424649029909e-05,
      "loss": 0.4453,
      "step": 214500
    },
    {
      "epoch": 3.36,
      "learning_rate": 3.8918161756650306e-05,
      "loss": 0.4422,
      "step": 215000
    },
    {
      "epoch": 3.36,
      "learning_rate": 3.8892129192468816e-05,
      "loss": 0.447,
      "step": 215500
    },
    {
      "epoch": 3.37,
      "learning_rate": 3.886604445882003e-05,
      "loss": 0.4494,
      "step": 216000
    },
    {
      "epoch": 3.38,
      "learning_rate": 3.883995972517125e-05,
      "loss": 0.4442,
      "step": 216500
    },
    {
      "epoch": 3.39,
      "learning_rate": 3.8813874991522463e-05,
      "loss": 0.4367,
      "step": 217000
    },
    {
      "epoch": 3.4,
      "learning_rate": 3.878779025787368e-05,
      "loss": 0.4418,
      "step": 217500
    },
    {
      "epoch": 3.4,
      "learning_rate": 3.8761705524224895e-05,
      "loss": 0.4374,
      "step": 218000
    },
    {
      "epoch": 3.41,
      "learning_rate": 3.873562079057611e-05,
      "loss": 0.4372,
      "step": 218500
    },
    {
      "epoch": 3.42,
      "learning_rate": 3.870953605692733e-05,
      "loss": 0.434,
      "step": 219000
    },
    {
      "epoch": 3.43,
      "learning_rate": 3.868345132327854e-05,
      "loss": 0.4363,
      "step": 219500
    },
    {
      "epoch": 3.43,
      "learning_rate": 3.865736658962976e-05,
      "loss": 0.4446,
      "step": 220000
    },
    {
      "epoch": 3.44,
      "learning_rate": 3.8631281855980975e-05,
      "loss": 0.4362,
      "step": 220500
    },
    {
      "epoch": 3.45,
      "learning_rate": 3.860519712233219e-05,
      "loss": 0.4402,
      "step": 221000
    },
    {
      "epoch": 3.46,
      "learning_rate": 3.8579216727617994e-05,
      "loss": 0.4426,
      "step": 221500
    },
    {
      "epoch": 3.47,
      "learning_rate": 3.855313199396921e-05,
      "loss": 0.4387,
      "step": 222000
    },
    {
      "epoch": 3.47,
      "learning_rate": 3.8527047260320425e-05,
      "loss": 0.4312,
      "step": 222500
    },
    {
      "epoch": 3.48,
      "learning_rate": 3.850096252667164e-05,
      "loss": 0.4336,
      "step": 223000
    },
    {
      "epoch": 3.49,
      "learning_rate": 3.847487779302286e-05,
      "loss": 0.4339,
      "step": 223500
    },
    {
      "epoch": 3.5,
      "learning_rate": 3.844879305937407e-05,
      "loss": 0.4244,
      "step": 224000
    },
    {
      "epoch": 3.5,
      "learning_rate": 3.842270832572529e-05,
      "loss": 0.4393,
      "step": 224500
    },
    {
      "epoch": 3.51,
      "learning_rate": 3.83966757615438e-05,
      "loss": 0.4311,
      "step": 225000
    },
    {
      "epoch": 3.52,
      "learning_rate": 3.8370591027895014e-05,
      "loss": 0.434,
      "step": 225500
    },
    {
      "epoch": 3.53,
      "learning_rate": 3.834450629424623e-05,
      "loss": 0.4313,
      "step": 226000
    },
    {
      "epoch": 3.54,
      "learning_rate": 3.8318421560597446e-05,
      "loss": 0.4307,
      "step": 226500
    },
    {
      "epoch": 3.54,
      "learning_rate": 3.829233682694866e-05,
      "loss": 0.4251,
      "step": 227000
    },
    {
      "epoch": 3.55,
      "learning_rate": 3.826625209329988e-05,
      "loss": 0.4315,
      "step": 227500
    },
    {
      "epoch": 3.56,
      "learning_rate": 3.8240167359651094e-05,
      "loss": 0.4372,
      "step": 228000
    },
    {
      "epoch": 3.57,
      "learning_rate": 3.821408262600231e-05,
      "loss": 0.4343,
      "step": 228500
    },
    {
      "epoch": 3.57,
      "learning_rate": 3.818799789235352e-05,
      "loss": 0.4331,
      "step": 229000
    },
    {
      "epoch": 3.58,
      "learning_rate": 3.8161913158704735e-05,
      "loss": 0.4299,
      "step": 229500
    },
    {
      "epoch": 3.59,
      "learning_rate": 3.813582842505595e-05,
      "loss": 0.4238,
      "step": 230000
    },
    {
      "epoch": 3.6,
      "learning_rate": 3.810974369140717e-05,
      "loss": 0.4228,
      "step": 230500
    },
    {
      "epoch": 3.61,
      "learning_rate": 3.808371112722568e-05,
      "loss": 0.425,
      "step": 231000
    },
    {
      "epoch": 3.61,
      "learning_rate": 3.805762639357689e-05,
      "loss": 0.4243,
      "step": 231500
    },
    {
      "epoch": 3.62,
      "learning_rate": 3.803154165992811e-05,
      "loss": 0.4273,
      "step": 232000
    },
    {
      "epoch": 3.63,
      "learning_rate": 3.8005509095746624e-05,
      "loss": 0.4274,
      "step": 232500
    },
    {
      "epoch": 3.64,
      "learning_rate": 3.797942436209784e-05,
      "loss": 0.4191,
      "step": 233000
    },
    {
      "epoch": 3.64,
      "learning_rate": 3.7953339628449056e-05,
      "loss": 0.4225,
      "step": 233500
    },
    {
      "epoch": 3.65,
      "learning_rate": 3.792725489480027e-05,
      "loss": 0.4218,
      "step": 234000
    },
    {
      "epoch": 3.66,
      "learning_rate": 3.790117016115148e-05,
      "loss": 0.4279,
      "step": 234500
    },
    {
      "epoch": 3.67,
      "learning_rate": 3.78750854275027e-05,
      "loss": 0.4233,
      "step": 235000
    },
    {
      "epoch": 3.68,
      "learning_rate": 3.784900069385391e-05,
      "loss": 0.4232,
      "step": 235500
    },
    {
      "epoch": 3.68,
      "learning_rate": 3.782291596020513e-05,
      "loss": 0.4216,
      "step": 236000
    },
    {
      "epoch": 3.69,
      "learning_rate": 3.7796831226556345e-05,
      "loss": 0.4251,
      "step": 236500
    },
    {
      "epoch": 3.7,
      "learning_rate": 3.777074649290756e-05,
      "loss": 0.4321,
      "step": 237000
    },
    {
      "epoch": 3.71,
      "learning_rate": 3.7744661759258777e-05,
      "loss": 0.4179,
      "step": 237500
    },
    {
      "epoch": 3.72,
      "learning_rate": 3.771857702560999e-05,
      "loss": 0.4137,
      "step": 238000
    },
    {
      "epoch": 3.72,
      "learning_rate": 3.769254446142851e-05,
      "loss": 0.4162,
      "step": 238500
    },
    {
      "epoch": 3.73,
      "learning_rate": 3.7666459727779725e-05,
      "loss": 0.431,
      "step": 239000
    },
    {
      "epoch": 3.74,
      "learning_rate": 3.7640427163598234e-05,
      "loss": 0.4126,
      "step": 239500
    },
    {
      "epoch": 3.75,
      "learning_rate": 3.761439459941675e-05,
      "loss": 0.4238,
      "step": 240000
    },
    {
      "epoch": 3.75,
      "learning_rate": 3.758836203523526e-05,
      "loss": 0.4239,
      "step": 240500
    },
    {
      "epoch": 3.76,
      "learning_rate": 3.7562277301586476e-05,
      "loss": 0.4297,
      "step": 241000
    },
    {
      "epoch": 3.77,
      "learning_rate": 3.753619256793769e-05,
      "loss": 0.4225,
      "step": 241500
    },
    {
      "epoch": 3.78,
      "learning_rate": 3.751010783428891e-05,
      "loss": 0.4201,
      "step": 242000
    },
    {
      "epoch": 3.79,
      "learning_rate": 3.748402310064012e-05,
      "loss": 0.4215,
      "step": 242500
    },
    {
      "epoch": 3.79,
      "learning_rate": 3.745793836699133e-05,
      "loss": 0.4182,
      "step": 243000
    },
    {
      "epoch": 3.8,
      "learning_rate": 3.743185363334255e-05,
      "loss": 0.4184,
      "step": 243500
    },
    {
      "epoch": 3.81,
      "learning_rate": 3.7405768899693764e-05,
      "loss": 0.4304,
      "step": 244000
    },
    {
      "epoch": 3.82,
      "learning_rate": 3.737968416604498e-05,
      "loss": 0.4175,
      "step": 244500
    },
    {
      "epoch": 3.82,
      "learning_rate": 3.7353599432396196e-05,
      "loss": 0.4144,
      "step": 245000
    },
    {
      "epoch": 3.83,
      "learning_rate": 3.732751469874741e-05,
      "loss": 0.4197,
      "step": 245500
    },
    {
      "epoch": 3.84,
      "learning_rate": 3.730142996509863e-05,
      "loss": 0.4157,
      "step": 246000
    },
    {
      "epoch": 3.85,
      "learning_rate": 3.7275345231449844e-05,
      "loss": 0.4127,
      "step": 246500
    },
    {
      "epoch": 3.86,
      "learning_rate": 3.724926049780106e-05,
      "loss": 0.4139,
      "step": 247000
    },
    {
      "epoch": 3.86,
      "learning_rate": 3.722322793361957e-05,
      "loss": 0.4154,
      "step": 247500
    },
    {
      "epoch": 3.87,
      "learning_rate": 3.7197143199970785e-05,
      "loss": 0.4172,
      "step": 248000
    },
    {
      "epoch": 3.88,
      "learning_rate": 3.7171058466322e-05,
      "loss": 0.4129,
      "step": 248500
    },
    {
      "epoch": 3.89,
      "learning_rate": 3.714497373267322e-05,
      "loss": 0.4162,
      "step": 249000
    },
    {
      "epoch": 3.89,
      "learning_rate": 3.7118941168491726e-05,
      "loss": 0.42,
      "step": 249500
    },
    {
      "epoch": 3.9,
      "learning_rate": 3.709285643484294e-05,
      "loss": 0.4146,
      "step": 250000
    },
    {
      "epoch": 3.91,
      "learning_rate": 3.706687604012876e-05,
      "loss": 0.4211,
      "step": 250500
    },
    {
      "epoch": 3.92,
      "learning_rate": 3.7040791306479975e-05,
      "loss": 0.4134,
      "step": 251000
    },
    {
      "epoch": 3.93,
      "learning_rate": 3.7014706572831184e-05,
      "loss": 0.4171,
      "step": 251500
    },
    {
      "epoch": 3.93,
      "learning_rate": 3.69886218391824e-05,
      "loss": 0.42,
      "step": 252000
    },
    {
      "epoch": 3.94,
      "learning_rate": 3.6962537105533615e-05,
      "loss": 0.4074,
      "step": 252500
    },
    {
      "epoch": 3.95,
      "learning_rate": 3.693650454135213e-05,
      "loss": 0.4119,
      "step": 253000
    },
    {
      "epoch": 3.96,
      "learning_rate": 3.691047197717064e-05,
      "loss": 0.4126,
      "step": 253500
    },
    {
      "epoch": 3.96,
      "learning_rate": 3.688438724352186e-05,
      "loss": 0.4086,
      "step": 254000
    },
    {
      "epoch": 3.97,
      "learning_rate": 3.685830250987307e-05,
      "loss": 0.4074,
      "step": 254500
    },
    {
      "epoch": 3.98,
      "learning_rate": 3.683221777622429e-05,
      "loss": 0.4154,
      "step": 255000
    },
    {
      "epoch": 3.99,
      "learning_rate": 3.6806133042575505e-05,
      "loss": 0.4223,
      "step": 255500
    },
    {
      "epoch": 4.0,
      "learning_rate": 3.678004830892672e-05,
      "loss": 0.415,
      "step": 256000
    },
    {
      "epoch": 4.0,
      "eval_bleu": 52.0616,
      "eval_gen_len": 15.0215,
      "eval_loss": 0.8553618788719177,
      "eval_runtime": 9536.6894,
      "eval_samples_per_second": 13.411,
      "eval_steps_per_second": 1.676,
      "step": 256246
    },
    {
      "epoch": 4.0,
      "learning_rate": 3.6753963575277937e-05,
      "loss": 0.3994,
      "step": 256500
    },
    {
      "epoch": 4.01,
      "learning_rate": 3.6727878841629146e-05,
      "loss": 0.4028,
      "step": 257000
    },
    {
      "epoch": 4.02,
      "learning_rate": 3.670184627744766e-05,
      "loss": 0.3918,
      "step": 257500
    },
    {
      "epoch": 4.03,
      "learning_rate": 3.667576154379888e-05,
      "loss": 0.3958,
      "step": 258000
    },
    {
      "epoch": 4.04,
      "learning_rate": 3.6649676810150094e-05,
      "loss": 0.396,
      "step": 258500
    },
    {
      "epoch": 4.04,
      "learning_rate": 3.662359207650131e-05,
      "loss": 0.3874,
      "step": 259000
    },
    {
      "epoch": 4.05,
      "learning_rate": 3.659750734285252e-05,
      "loss": 0.3863,
      "step": 259500
    },
    {
      "epoch": 4.06,
      "learning_rate": 3.657147477867104e-05,
      "loss": 0.3821,
      "step": 260000
    },
    {
      "epoch": 4.07,
      "learning_rate": 3.654544221448955e-05,
      "loss": 0.3895,
      "step": 260500
    },
    {
      "epoch": 4.07,
      "learning_rate": 3.651935748084077e-05,
      "loss": 0.3819,
      "step": 261000
    },
    {
      "epoch": 4.08,
      "learning_rate": 3.649327274719198e-05,
      "loss": 0.3798,
      "step": 261500
    },
    {
      "epoch": 4.09,
      "learning_rate": 3.646724018301049e-05,
      "loss": 0.3812,
      "step": 262000
    },
    {
      "epoch": 4.1,
      "learning_rate": 3.644115544936171e-05,
      "loss": 0.3796,
      "step": 262500
    },
    {
      "epoch": 4.11,
      "learning_rate": 3.6415070715712924e-05,
      "loss": 0.3743,
      "step": 263000
    },
    {
      "epoch": 4.11,
      "learning_rate": 3.638898598206414e-05,
      "loss": 0.374,
      "step": 263500
    },
    {
      "epoch": 4.12,
      "learning_rate": 3.6362901248415356e-05,
      "loss": 0.3849,
      "step": 264000
    },
    {
      "epoch": 4.13,
      "learning_rate": 3.633681651476657e-05,
      "loss": 0.3815,
      "step": 264500
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.631073178111779e-05,
      "loss": 0.3717,
      "step": 265000
    },
    {
      "epoch": 4.14,
      "learning_rate": 3.6284647047469e-05,
      "loss": 0.3727,
      "step": 265500
    },
    {
      "epoch": 4.15,
      "learning_rate": 3.625856231382021e-05,
      "loss": 0.3789,
      "step": 266000
    },
    {
      "epoch": 4.16,
      "learning_rate": 3.623247758017143e-05,
      "loss": 0.3658,
      "step": 266500
    },
    {
      "epoch": 4.17,
      "learning_rate": 3.6206392846522645e-05,
      "loss": 0.3717,
      "step": 267000
    },
    {
      "epoch": 4.18,
      "learning_rate": 3.618030811287386e-05,
      "loss": 0.3661,
      "step": 267500
    },
    {
      "epoch": 4.18,
      "learning_rate": 3.6154223379225077e-05,
      "loss": 0.3732,
      "step": 268000
    },
    {
      "epoch": 4.19,
      "learning_rate": 3.612813864557629e-05,
      "loss": 0.361,
      "step": 268500
    },
    {
      "epoch": 4.2,
      "learning_rate": 3.610205391192751e-05,
      "loss": 0.3657,
      "step": 269000
    },
    {
      "epoch": 4.21,
      "learning_rate": 3.6075969178278724e-05,
      "loss": 0.37,
      "step": 269500
    },
    {
      "epoch": 4.21,
      "learning_rate": 3.604988444462994e-05,
      "loss": 0.3647,
      "step": 270000
    },
    {
      "epoch": 4.22,
      "learning_rate": 3.6023799710981156e-05,
      "loss": 0.3665,
      "step": 270500
    },
    {
      "epoch": 4.23,
      "learning_rate": 3.599781931626696e-05,
      "loss": 0.3597,
      "step": 271000
    },
    {
      "epoch": 4.24,
      "learning_rate": 3.5971734582618175e-05,
      "loss": 0.3711,
      "step": 271500
    },
    {
      "epoch": 4.25,
      "learning_rate": 3.594564984896939e-05,
      "loss": 0.3582,
      "step": 272000
    },
    {
      "epoch": 4.25,
      "learning_rate": 3.591956511532061e-05,
      "loss": 0.358,
      "step": 272500
    },
    {
      "epoch": 4.26,
      "learning_rate": 3.589348038167182e-05,
      "loss": 0.3548,
      "step": 273000
    },
    {
      "epoch": 4.27,
      "learning_rate": 3.586744781749033e-05,
      "loss": 0.3592,
      "step": 273500
    },
    {
      "epoch": 4.28,
      "learning_rate": 3.584136308384155e-05,
      "loss": 0.3683,
      "step": 274000
    },
    {
      "epoch": 4.28,
      "learning_rate": 3.5815278350192764e-05,
      "loss": 0.3615,
      "step": 274500
    },
    {
      "epoch": 4.29,
      "learning_rate": 3.578919361654398e-05,
      "loss": 0.3691,
      "step": 275000
    },
    {
      "epoch": 4.3,
      "learning_rate": 3.5763161052362496e-05,
      "loss": 0.3503,
      "step": 275500
    },
    {
      "epoch": 4.31,
      "learning_rate": 3.573707631871371e-05,
      "loss": 0.3686,
      "step": 276000
    },
    {
      "epoch": 4.32,
      "learning_rate": 3.571099158506493e-05,
      "loss": 0.3617,
      "step": 276500
    },
    {
      "epoch": 4.32,
      "learning_rate": 3.5684906851416144e-05,
      "loss": 0.3621,
      "step": 277000
    },
    {
      "epoch": 4.33,
      "learning_rate": 3.565882211776736e-05,
      "loss": 0.3653,
      "step": 277500
    },
    {
      "epoch": 4.34,
      "learning_rate": 3.5632737384118576e-05,
      "loss": 0.3594,
      "step": 278000
    },
    {
      "epoch": 4.35,
      "learning_rate": 3.5606652650469785e-05,
      "loss": 0.358,
      "step": 278500
    },
    {
      "epoch": 4.36,
      "learning_rate": 3.55806200862883e-05,
      "loss": 0.3565,
      "step": 279000
    },
    {
      "epoch": 4.36,
      "learning_rate": 3.555453535263952e-05,
      "loss": 0.3576,
      "step": 279500
    },
    {
      "epoch": 4.37,
      "learning_rate": 3.552845061899073e-05,
      "loss": 0.3646,
      "step": 280000
    },
    {
      "epoch": 4.38,
      "learning_rate": 3.550236588534195e-05,
      "loss": 0.3541,
      "step": 280500
    },
    {
      "epoch": 4.39,
      "learning_rate": 3.5476281151693165e-05,
      "loss": 0.3531,
      "step": 281000
    },
    {
      "epoch": 4.39,
      "learning_rate": 3.545019641804438e-05,
      "loss": 0.3547,
      "step": 281500
    },
    {
      "epoch": 4.4,
      "learning_rate": 3.5424111684395596e-05,
      "loss": 0.3496,
      "step": 282000
    },
    {
      "epoch": 4.41,
      "learning_rate": 3.539802695074681e-05,
      "loss": 0.3525,
      "step": 282500
    },
    {
      "epoch": 4.42,
      "learning_rate": 3.537199438656532e-05,
      "loss": 0.3485,
      "step": 283000
    },
    {
      "epoch": 4.43,
      "learning_rate": 3.534590965291654e-05,
      "loss": 0.352,
      "step": 283500
    },
    {
      "epoch": 4.43,
      "learning_rate": 3.5319824919267754e-05,
      "loss": 0.3585,
      "step": 284000
    },
    {
      "epoch": 4.44,
      "learning_rate": 3.529374018561897e-05,
      "loss": 0.3549,
      "step": 284500
    },
    {
      "epoch": 4.45,
      "learning_rate": 3.5267655451970185e-05,
      "loss": 0.3503,
      "step": 285000
    },
    {
      "epoch": 4.46,
      "learning_rate": 3.52415707183214e-05,
      "loss": 0.3567,
      "step": 285500
    },
    {
      "epoch": 4.46,
      "learning_rate": 3.521548598467261e-05,
      "loss": 0.3526,
      "step": 286000
    },
    {
      "epoch": 4.47,
      "learning_rate": 3.5189401251023826e-05,
      "loss": 0.3458,
      "step": 286500
    },
    {
      "epoch": 4.48,
      "learning_rate": 3.5163420856309636e-05,
      "loss": 0.3486,
      "step": 287000
    },
    {
      "epoch": 4.49,
      "learning_rate": 3.513733612266085e-05,
      "loss": 0.3476,
      "step": 287500
    },
    {
      "epoch": 4.5,
      "learning_rate": 3.511125138901207e-05,
      "loss": 0.346,
      "step": 288000
    },
    {
      "epoch": 4.5,
      "learning_rate": 3.508521882483058e-05,
      "loss": 0.3547,
      "step": 288500
    },
    {
      "epoch": 4.51,
      "learning_rate": 3.505913409118179e-05,
      "loss": 0.3442,
      "step": 289000
    },
    {
      "epoch": 4.52,
      "learning_rate": 3.503304935753301e-05,
      "loss": 0.3488,
      "step": 289500
    },
    {
      "epoch": 4.53,
      "learning_rate": 3.5006964623884225e-05,
      "loss": 0.3464,
      "step": 290000
    },
    {
      "epoch": 4.53,
      "learning_rate": 3.498087989023544e-05,
      "loss": 0.3482,
      "step": 290500
    },
    {
      "epoch": 4.54,
      "learning_rate": 3.495479515658666e-05,
      "loss": 0.342,
      "step": 291000
    },
    {
      "epoch": 4.55,
      "learning_rate": 3.492871042293787e-05,
      "loss": 0.3462,
      "step": 291500
    },
    {
      "epoch": 4.56,
      "learning_rate": 3.490262568928909e-05,
      "loss": 0.3514,
      "step": 292000
    },
    {
      "epoch": 4.57,
      "learning_rate": 3.4876540955640305e-05,
      "loss": 0.352,
      "step": 292500
    },
    {
      "epoch": 4.57,
      "learning_rate": 3.485045622199152e-05,
      "loss": 0.3479,
      "step": 293000
    },
    {
      "epoch": 4.58,
      "learning_rate": 3.4824371488342736e-05,
      "loss": 0.3489,
      "step": 293500
    },
    {
      "epoch": 4.59,
      "learning_rate": 3.4798286754693946e-05,
      "loss": 0.3403,
      "step": 294000
    },
    {
      "epoch": 4.6,
      "learning_rate": 3.477220202104516e-05,
      "loss": 0.337,
      "step": 294500
    },
    {
      "epoch": 4.6,
      "learning_rate": 3.474611728739638e-05,
      "loss": 0.346,
      "step": 295000
    },
    {
      "epoch": 4.61,
      "learning_rate": 3.472003255374759e-05,
      "loss": 0.3395,
      "step": 295500
    },
    {
      "epoch": 4.62,
      "learning_rate": 3.469394782009881e-05,
      "loss": 0.3379,
      "step": 296000
    },
    {
      "epoch": 4.63,
      "learning_rate": 3.4667863086450025e-05,
      "loss": 0.3447,
      "step": 296500
    },
    {
      "epoch": 4.64,
      "learning_rate": 3.464177835280124e-05,
      "loss": 0.3392,
      "step": 297000
    },
    {
      "epoch": 4.64,
      "learning_rate": 3.461569361915246e-05,
      "loss": 0.3439,
      "step": 297500
    },
    {
      "epoch": 4.65,
      "learning_rate": 3.458960888550367e-05,
      "loss": 0.3341,
      "step": 298000
    },
    {
      "epoch": 4.66,
      "learning_rate": 3.456352415185489e-05,
      "loss": 0.3435,
      "step": 298500
    },
    {
      "epoch": 4.67,
      "learning_rate": 3.4537439418206105e-05,
      "loss": 0.3459,
      "step": 299000
    },
    {
      "epoch": 4.68,
      "learning_rate": 3.4511406854024614e-05,
      "loss": 0.3394,
      "step": 299500
    },
    {
      "epoch": 4.68,
      "learning_rate": 3.448532212037583e-05,
      "loss": 0.3338,
      "step": 300000
    },
    {
      "epoch": 4.69,
      "learning_rate": 3.4459237386727046e-05,
      "loss": 0.3435,
      "step": 300500
    },
    {
      "epoch": 4.7,
      "learning_rate": 3.443315265307826e-05,
      "loss": 0.3512,
      "step": 301000
    },
    {
      "epoch": 4.71,
      "learning_rate": 3.440706791942948e-05,
      "loss": 0.3327,
      "step": 301500
    },
    {
      "epoch": 4.71,
      "learning_rate": 3.438113969418259e-05,
      "loss": 0.3351,
      "step": 302000
    },
    {
      "epoch": 4.72,
      "learning_rate": 3.43550549605338e-05,
      "loss": 0.3341,
      "step": 302500
    },
    {
      "epoch": 4.73,
      "learning_rate": 3.432897022688501e-05,
      "loss": 0.3392,
      "step": 303000
    },
    {
      "epoch": 4.74,
      "learning_rate": 3.430288549323623e-05,
      "loss": 0.3324,
      "step": 303500
    },
    {
      "epoch": 4.75,
      "learning_rate": 3.4276800759587445e-05,
      "loss": 0.3373,
      "step": 304000
    },
    {
      "epoch": 4.75,
      "learning_rate": 3.425071602593866e-05,
      "loss": 0.3402,
      "step": 304500
    },
    {
      "epoch": 4.76,
      "learning_rate": 3.4224631292289876e-05,
      "loss": 0.3451,
      "step": 305000
    },
    {
      "epoch": 4.77,
      "learning_rate": 3.419854655864109e-05,
      "loss": 0.3417,
      "step": 305500
    },
    {
      "epoch": 4.78,
      "learning_rate": 3.417246182499231e-05,
      "loss": 0.3386,
      "step": 306000
    },
    {
      "epoch": 4.78,
      "learning_rate": 3.4146377091343524e-05,
      "loss": 0.3384,
      "step": 306500
    },
    {
      "epoch": 4.79,
      "learning_rate": 3.412029235769474e-05,
      "loss": 0.3388,
      "step": 307000
    },
    {
      "epoch": 4.8,
      "learning_rate": 3.4094207624045956e-05,
      "loss": 0.3357,
      "step": 307500
    },
    {
      "epoch": 4.81,
      "learning_rate": 3.4068175059864465e-05,
      "loss": 0.3482,
      "step": 308000
    },
    {
      "epoch": 4.82,
      "learning_rate": 3.404209032621568e-05,
      "loss": 0.334,
      "step": 308500
    },
    {
      "epoch": 4.82,
      "learning_rate": 3.40160055925669e-05,
      "loss": 0.3334,
      "step": 309000
    },
    {
      "epoch": 4.83,
      "learning_rate": 3.398992085891811e-05,
      "loss": 0.3379,
      "step": 309500
    },
    {
      "epoch": 4.84,
      "learning_rate": 3.396383612526933e-05,
      "loss": 0.3321,
      "step": 310000
    },
    {
      "epoch": 4.85,
      "learning_rate": 3.393780356108784e-05,
      "loss": 0.3291,
      "step": 310500
    },
    {
      "epoch": 4.85,
      "learning_rate": 3.391177099690635e-05,
      "loss": 0.3314,
      "step": 311000
    },
    {
      "epoch": 4.86,
      "learning_rate": 3.3885686263257564e-05,
      "loss": 0.3365,
      "step": 311500
    },
    {
      "epoch": 4.87,
      "learning_rate": 3.385960152960878e-05,
      "loss": 0.3352,
      "step": 312000
    },
    {
      "epoch": 4.88,
      "learning_rate": 3.3833516795959996e-05,
      "loss": 0.3276,
      "step": 312500
    },
    {
      "epoch": 4.89,
      "learning_rate": 3.380743206231121e-05,
      "loss": 0.3353,
      "step": 313000
    },
    {
      "epoch": 4.89,
      "learning_rate": 3.378134732866243e-05,
      "loss": 0.3404,
      "step": 313500
    },
    {
      "epoch": 4.9,
      "learning_rate": 3.375526259501364e-05,
      "loss": 0.331,
      "step": 314000
    },
    {
      "epoch": 4.91,
      "learning_rate": 3.372923003083215e-05,
      "loss": 0.3357,
      "step": 314500
    },
    {
      "epoch": 4.92,
      "learning_rate": 3.370314529718337e-05,
      "loss": 0.3313,
      "step": 315000
    },
    {
      "epoch": 4.92,
      "learning_rate": 3.3677060563534585e-05,
      "loss": 0.3392,
      "step": 315500
    },
    {
      "epoch": 4.93,
      "learning_rate": 3.36509758298858e-05,
      "loss": 0.3325,
      "step": 316000
    },
    {
      "epoch": 4.94,
      "learning_rate": 3.3624891096237016e-05,
      "loss": 0.3276,
      "step": 316500
    },
    {
      "epoch": 4.95,
      "learning_rate": 3.359880636258823e-05,
      "loss": 0.3323,
      "step": 317000
    },
    {
      "epoch": 4.96,
      "learning_rate": 3.357272162893945e-05,
      "loss": 0.3269,
      "step": 317500
    },
    {
      "epoch": 4.96,
      "learning_rate": 3.3546636895290664e-05,
      "loss": 0.3339,
      "step": 318000
    },
    {
      "epoch": 4.97,
      "learning_rate": 3.352055216164188e-05,
      "loss": 0.3211,
      "step": 318500
    },
    {
      "epoch": 4.98,
      "learning_rate": 3.3494467427993096e-05,
      "loss": 0.3357,
      "step": 319000
    },
    {
      "epoch": 4.99,
      "learning_rate": 3.3468382694344305e-05,
      "loss": 0.3379,
      "step": 319500
    },
    {
      "epoch": 5.0,
      "learning_rate": 3.344229796069552e-05,
      "loss": 0.3348,
      "step": 320000
    },
    {
      "epoch": 5.0,
      "eval_bleu": 51.995,
      "eval_gen_len": 14.9859,
      "eval_loss": 0.9135012626647949,
      "eval_runtime": 9511.5501,
      "eval_samples_per_second": 13.446,
      "eval_steps_per_second": 1.681,
      "step": 320307
    },
    {
      "epoch": 5.0,
      "step": 320307,
      "total_flos": 1.1106366691632742e+19,
      "train_loss": 0.5943015630864049,
      "train_runtime": 296139.8154,
      "train_samples_per_second": 51.917,
      "train_steps_per_second": 3.245
    }
  ],
  "logging_steps": 500,
  "max_steps": 960915,
  "num_train_epochs": 15,
  "save_steps": 500,
  "total_flos": 1.1106366691632742e+19,
  "trial_name": null,
  "trial_params": null
}