diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5416 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 13.79615349165873,
+  "eval_steps": 1000000,
+  "global_step": 449051,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 5.138,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 5.1341,
+      "step": 1000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5e-06,
+      "loss": 5.1358,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 5.1302,
+      "step": 2000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.5e-06,
+      "loss": 5.1358,
+      "step": 2500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3e-06,
+      "loss": 5.1287,
+      "step": 3000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.5e-06,
+      "loss": 5.1274,
+      "step": 3500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 5.1241,
+      "step": 4000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5e-06,
+      "loss": 5.1179,
+      "step": 4500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 5e-06,
+      "loss": 5.1171,
+      "step": 5000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 5.114,
+      "step": 5500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 6e-06,
+      "loss": 5.1133,
+      "step": 6000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 5.108,
+      "step": 6500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7e-06,
+      "loss": 5.1033,
+      "step": 7000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 5.1003,
+      "step": 7500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 5.0936,
+      "step": 8000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.5e-06,
+      "loss": 5.0883,
+      "step": 8500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9e-06,
+      "loss": 5.0792,
+      "step": 9000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.5e-06,
+      "loss": 5.0737,
+      "step": 9500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1e-05,
+      "loss": 5.0679,
+      "step": 10000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.998459120465963e-06,
+      "loss": 5.0633,
+      "step": 10500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.996918240931923e-06,
+      "loss": 5.0573,
+      "step": 11000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.995377361397887e-06,
+      "loss": 5.0444,
+      "step": 11500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.99383648186385e-06,
+      "loss": 5.0405,
+      "step": 12000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.992295602329812e-06,
+      "loss": 5.0347,
+      "step": 12500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.990754722795772e-06,
+      "loss": 5.0315,
+      "step": 13000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.989213843261734e-06,
+      "loss": 5.0211,
+      "step": 13500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.987672963727697e-06,
+      "loss": 5.0205,
+      "step": 14000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.986132084193659e-06,
+      "loss": 5.0089,
+      "step": 14500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.984591204659621e-06,
+      "loss": 5.0051,
+      "step": 15000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.983050325125583e-06,
+      "loss": 4.9992,
+      "step": 15500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.981509445591544e-06,
+      "loss": 4.991,
+      "step": 16000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.979968566057506e-06,
+      "loss": 4.9866,
+      "step": 16500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.978427686523468e-06,
+      "loss": 4.9848,
+      "step": 17000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.97688680698943e-06,
+      "loss": 4.9763,
+      "step": 17500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.975345927455392e-06,
+      "loss": 4.9712,
+      "step": 18000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.973805047921354e-06,
+      "loss": 4.9634,
+      "step": 18500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.972264168387317e-06,
+      "loss": 4.9602,
+      "step": 19000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.970723288853277e-06,
+      "loss": 4.9553,
+      "step": 19500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.969182409319241e-06,
+      "loss": 4.9473,
+      "step": 20000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.967641529785203e-06,
+      "loss": 4.9456,
+      "step": 20500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.966100650251164e-06,
+      "loss": 4.9359,
+      "step": 21000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.964559770717126e-06,
+      "loss": 4.9346,
+      "step": 21500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.963018891183088e-06,
+      "loss": 4.9281,
+      "step": 22000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.96147801164905e-06,
+      "loss": 4.9256,
+      "step": 22500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.959937132115012e-06,
+      "loss": 4.917,
+      "step": 23000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.958396252580975e-06,
+      "loss": 4.9096,
+      "step": 23500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.956855373046937e-06,
+      "loss": 4.9054,
+      "step": 24000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.955314493512897e-06,
+      "loss": 4.9027,
+      "step": 24500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.95377361397886e-06,
+      "loss": 4.896,
+      "step": 25000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.952232734444822e-06,
+      "loss": 4.8931,
+      "step": 25500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.950691854910784e-06,
+      "loss": 4.8873,
+      "step": 26000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.949150975376746e-06,
+      "loss": 4.8854,
+      "step": 26500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.947610095842708e-06,
+      "loss": 4.8803,
+      "step": 27000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.94606921630867e-06,
+      "loss": 4.8755,
+      "step": 27500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.94452833677463e-06,
+      "loss": 4.874,
+      "step": 28000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.942987457240593e-06,
+      "loss": 4.8675,
+      "step": 28500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.941446577706557e-06,
+      "loss": 4.8636,
+      "step": 29000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.939905698172517e-06,
+      "loss": 4.8586,
+      "step": 29500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.93836481863848e-06,
+      "loss": 4.8531,
+      "step": 30000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.936823939104442e-06,
+      "loss": 4.8509,
+      "step": 30500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.935283059570404e-06,
+      "loss": 4.8428,
+      "step": 31000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.933742180036366e-06,
+      "loss": 4.8429,
+      "step": 31500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.932201300502328e-06,
+      "loss": 4.8359,
+      "step": 32000
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.93066042096829e-06,
+      "loss": 4.8354,
+      "step": 32500
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.929119541434251e-06,
+      "loss": 4.8288,
+      "step": 33000
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.927578661900213e-06,
+      "loss": 4.8239,
+      "step": 33500
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.926037782366175e-06,
+      "loss": 4.8201,
+      "step": 34000
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.924496902832137e-06,
+      "loss": 4.8127,
+      "step": 34500
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.9229560232981e-06,
+      "loss": 4.8121,
+      "step": 35000
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.921415143764062e-06,
+      "loss": 4.8096,
+      "step": 35500
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.919874264230022e-06,
+      "loss": 4.8057,
+      "step": 36000
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.918333384695984e-06,
+      "loss": 4.8004,
+      "step": 36500
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 9.916792505161947e-06,
+      "loss": 4.7978,
+      "step": 37000
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 9.915251625627909e-06,
+      "loss": 4.7921,
+      "step": 37500
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 9.913710746093871e-06,
+      "loss": 4.7939,
+      "step": 38000
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 9.912169866559833e-06,
+      "loss": 4.7875,
+      "step": 38500
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 9.910628987025795e-06,
+      "loss": 4.7821,
+      "step": 39000
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.909088107491756e-06,
+      "loss": 4.7785,
+      "step": 39500
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.90754722795772e-06,
+      "loss": 4.7738,
+      "step": 40000
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 9.906006348423682e-06,
+      "loss": 4.7733,
+      "step": 40500
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 9.904465468889644e-06,
+      "loss": 4.7685,
+      "step": 41000
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 9.902924589355605e-06,
+      "loss": 4.7634,
+      "step": 41500
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 9.901383709821567e-06,
+      "loss": 4.7619,
+      "step": 42000
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 9.899842830287529e-06,
+      "loss": 4.7592,
+      "step": 42500
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 9.898301950753491e-06,
+      "loss": 4.7535,
+      "step": 43000
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 9.896761071219453e-06,
+      "loss": 4.7541,
+      "step": 43500
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 9.895220191685415e-06,
+      "loss": 4.7484,
+      "step": 44000
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 9.893679312151376e-06,
+      "loss": 4.7479,
+      "step": 44500
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 9.892138432617338e-06,
+      "loss": 4.7394,
+      "step": 45000
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 9.8905975530833e-06,
+      "loss": 4.7363,
+      "step": 45500
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 9.889056673549262e-06,
+      "loss": 4.735,
+      "step": 46000
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 9.887515794015225e-06,
+      "loss": 4.7274,
+      "step": 46500
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 9.885974914481187e-06,
+      "loss": 4.7262,
+      "step": 47000
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 9.884434034947149e-06,
+      "loss": 4.7214,
+      "step": 47500
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.88289315541311e-06,
+      "loss": 4.7176,
+      "step": 48000
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 9.881352275879072e-06,
+      "loss": 4.714,
+      "step": 48500
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.879811396345036e-06,
+      "loss": 4.7147,
+      "step": 49000
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.878270516810996e-06,
+      "loss": 4.713,
+      "step": 49500
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.876729637276958e-06,
+      "loss": 4.7094,
+      "step": 50000
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.87518875774292e-06,
+      "loss": 4.7003,
+      "step": 50500
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.873647878208883e-06,
+      "loss": 4.6968,
+      "step": 51000
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.872106998674845e-06,
+      "loss": 4.6938,
+      "step": 51500
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.870566119140807e-06,
+      "loss": 4.688,
+      "step": 52000
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.869025239606769e-06,
+      "loss": 4.6805,
+      "step": 52500
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.86748436007273e-06,
+      "loss": 4.6743,
+      "step": 53000
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.865943480538692e-06,
+      "loss": 4.6706,
+      "step": 53500
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 9.864402601004654e-06,
+      "loss": 4.6585,
+      "step": 54000
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 9.862861721470616e-06,
+      "loss": 4.6524,
+      "step": 54500
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 9.861320841936578e-06,
+      "loss": 4.6466,
+      "step": 55000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 9.85977996240254e-06,
+      "loss": 4.6417,
+      "step": 55500
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.858239082868503e-06,
+      "loss": 4.6315,
+      "step": 56000
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 9.856698203334463e-06,
+      "loss": 4.6215,
+      "step": 56500
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 9.855157323800425e-06,
+      "loss": 4.6082,
+      "step": 57000
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 9.85361644426639e-06,
+      "loss": 4.6014,
+      "step": 57500
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 9.85207556473235e-06,
+      "loss": 4.5951,
+      "step": 58000
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 9.850534685198312e-06,
+      "loss": 4.5844,
+      "step": 58500
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.848993805664274e-06,
+      "loss": 4.5778,
+      "step": 59000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 9.847452926130235e-06,
+      "loss": 4.5775,
+      "step": 59500
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 9.845912046596198e-06,
+      "loss": 4.5698,
+      "step": 60000
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 9.84437116706216e-06,
+      "loss": 4.565,
+      "step": 60500
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 9.842830287528123e-06,
+      "loss": 4.5615,
+      "step": 61000
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 9.841289407994083e-06,
+      "loss": 4.5467,
+      "step": 61500
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.839748528460045e-06,
+      "loss": 4.5423,
+      "step": 62000
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.838207648926008e-06,
+      "loss": 4.5394,
+      "step": 62500
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.83666676939197e-06,
+      "loss": 4.5346,
+      "step": 63000
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 9.835125889857932e-06,
+      "loss": 4.5251,
+      "step": 63500
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.833585010323894e-06,
+      "loss": 4.5222,
+      "step": 64000
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 9.832044130789856e-06,
+      "loss": 4.5139,
+      "step": 64500
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.830503251255817e-06,
+      "loss": 4.5106,
+      "step": 65000
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 9.828962371721779e-06,
+      "loss": 4.5031,
+      "step": 65500
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.827421492187741e-06,
+      "loss": 4.4957,
+      "step": 66000
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 9.825880612653703e-06,
+      "loss": 4.4983,
+      "step": 66500
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.824339733119666e-06,
+      "loss": 4.489,
+      "step": 67000
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 9.822798853585628e-06,
+      "loss": 4.4857,
+      "step": 67500
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 9.821257974051588e-06,
+      "loss": 4.4822,
+      "step": 68000
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 9.819717094517552e-06,
+      "loss": 4.478,
+      "step": 68500
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 9.818176214983514e-06,
+      "loss": 4.4677,
+      "step": 69000
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.816635335449476e-06,
+      "loss": 4.4682,
+      "step": 69500
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 9.815094455915437e-06,
+      "loss": 4.4624,
+      "step": 70000
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.813553576381399e-06,
+      "loss": 4.4567,
+      "step": 70500
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 9.812012696847361e-06,
+      "loss": 4.4561,
+      "step": 71000
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.810471817313323e-06,
+      "loss": 4.4505,
+      "step": 71500
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 9.808930937779286e-06,
+      "loss": 4.4458,
+      "step": 72000
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 9.807390058245248e-06,
+      "loss": 4.4399,
+      "step": 72500
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 9.805849178711208e-06,
+      "loss": 4.439,
+      "step": 73000
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 9.80430829917717e-06,
+      "loss": 4.435,
+      "step": 73500
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 9.802767419643133e-06,
+      "loss": 4.4333,
+      "step": 74000
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 9.801226540109095e-06,
+      "loss": 4.4246,
+      "step": 74500
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 9.799685660575057e-06,
+      "loss": 4.4268,
+      "step": 75000
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 9.79814478104102e-06,
+      "loss": 4.418,
+      "step": 75500
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 9.796603901506981e-06,
+      "loss": 4.4201,
+      "step": 76000
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 9.795063021972942e-06,
+      "loss": 4.4084,
+      "step": 76500
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 9.793522142438904e-06,
+      "loss": 4.4033,
+      "step": 77000
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 9.791981262904868e-06,
+      "loss": 4.4051,
+      "step": 77500
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 9.790440383370828e-06,
+      "loss": 4.4024,
+      "step": 78000
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 9.78889950383679e-06,
+      "loss": 4.3974,
+      "step": 78500
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 9.787358624302753e-06,
+      "loss": 4.3934,
+      "step": 79000
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 9.785817744768715e-06,
+      "loss": 4.3886,
+      "step": 79500
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 9.784276865234677e-06,
+      "loss": 4.3898,
+      "step": 80000
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 9.78273598570064e-06,
+      "loss": 4.3795,
+      "step": 80500
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 9.781195106166602e-06,
+      "loss": 4.3793,
+      "step": 81000
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 9.779654226632562e-06,
+      "loss": 4.377,
+      "step": 81500
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 9.778113347098524e-06,
+      "loss": 4.3777,
+      "step": 82000
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 9.776572467564486e-06,
+      "loss": 4.3732,
+      "step": 82500
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 9.775031588030449e-06,
+      "loss": 4.3656,
+      "step": 83000
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.77349070849641e-06,
+      "loss": 4.364,
+      "step": 83500
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.771949828962373e-06,
+      "loss": 4.3612,
+      "step": 84000
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 9.770408949428335e-06,
+      "loss": 4.3586,
+      "step": 84500
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 9.768868069894296e-06,
+      "loss": 4.3569,
+      "step": 85000
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 9.767327190360258e-06,
+      "loss": 4.3502,
+      "step": 85500
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 9.76578631082622e-06,
+      "loss": 4.3459,
+      "step": 86000
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 9.764245431292182e-06,
+      "loss": 4.3434,
+      "step": 86500
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 9.762704551758144e-06,
+      "loss": 4.3405,
+      "step": 87000
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 9.761163672224106e-06,
+      "loss": 4.3387,
+      "step": 87500
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 9.759622792690067e-06,
+      "loss": 4.3361,
+      "step": 88000
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 9.75808191315603e-06,
+      "loss": 4.3365,
+      "step": 88500
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 9.756541033621993e-06,
+      "loss": 4.3365,
+      "step": 89000
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 9.755000154087955e-06,
+      "loss": 4.3265,
+      "step": 89500
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 9.753459274553916e-06,
+      "loss": 4.327,
+      "step": 90000
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 9.751918395019878e-06,
+      "loss": 4.323,
+      "step": 90500
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.75037751548584e-06,
+      "loss": 4.3216,
+      "step": 91000
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.748836635951802e-06,
+      "loss": 4.3187,
+      "step": 91500
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 9.747295756417764e-06,
+      "loss": 4.3144,
+      "step": 92000
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 9.745754876883727e-06,
+      "loss": 4.3151,
+      "step": 92500
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 9.744213997349689e-06,
+      "loss": 4.3112,
+      "step": 93000
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 9.74267311781565e-06,
+      "loss": 4.3078,
+      "step": 93500
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 9.741132238281611e-06,
+      "loss": 4.307,
+      "step": 94000
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 9.739591358747574e-06,
+      "loss": 4.301,
+      "step": 94500
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 9.738050479213536e-06,
+      "loss": 4.3001,
+      "step": 95000
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 9.736509599679498e-06,
+      "loss": 4.2995,
+      "step": 95500
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 9.73496872014546e-06,
+      "loss": 4.2944,
+      "step": 96000
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.73342784061142e-06,
+      "loss": 4.2941,
+      "step": 96500
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 9.731886961077383e-06,
+      "loss": 4.2897,
+      "step": 97000
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 9.730346081543347e-06,
+      "loss": 4.2923,
+      "step": 97500
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 9.728805202009309e-06,
+      "loss": 4.2865,
+      "step": 98000
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 9.72726432247527e-06,
+      "loss": 4.2828,
+      "step": 98500
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 9.725723442941232e-06,
+      "loss": 4.2766,
+      "step": 99000
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 9.724182563407194e-06,
+      "loss": 4.2835,
+      "step": 99500
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 9.722641683873156e-06,
+      "loss": 4.2822,
+      "step": 100000
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 9.721100804339118e-06,
+      "loss": 4.2787,
+      "step": 100500
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 9.71955992480508e-06,
+      "loss": 4.2751,
+      "step": 101000
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 9.71801904527104e-06,
+      "loss": 4.2737,
+      "step": 101500
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 9.716478165737003e-06,
+      "loss": 4.2673,
+      "step": 102000
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 9.714937286202965e-06,
+      "loss": 4.2658,
+      "step": 102500
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 9.713396406668927e-06,
+      "loss": 4.2643,
+      "step": 103000
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 9.71185552713489e-06,
+      "loss": 4.2611,
+      "step": 103500
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 9.710314647600852e-06,
+      "loss": 4.2649,
+      "step": 104000
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 9.708773768066814e-06,
+      "loss": 4.2618,
+      "step": 104500
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 9.707232888532774e-06,
+      "loss": 4.256,
+      "step": 105000
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 9.705692008998736e-06,
+      "loss": 4.2549,
+      "step": 105500
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 9.7041511294647e-06,
+      "loss": 4.2552,
+      "step": 106000
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 9.70261024993066e-06,
+      "loss": 4.249,
+      "step": 106500
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 9.701069370396623e-06,
+      "loss": 4.2496,
+      "step": 107000
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 9.699528490862585e-06,
+      "loss": 4.2496,
+      "step": 107500
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 9.697987611328547e-06,
+      "loss": 4.2504,
+      "step": 108000
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 9.69644673179451e-06,
+      "loss": 4.2393,
+      "step": 108500
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 9.694905852260472e-06,
+      "loss": 4.2437,
+      "step": 109000
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 9.693364972726434e-06,
+      "loss": 4.2431,
+      "step": 109500
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 9.691824093192394e-06,
+      "loss": 4.2427,
+      "step": 110000
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 9.690283213658357e-06,
+      "loss": 4.2366,
+      "step": 110500
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 9.688742334124319e-06,
+      "loss": 4.2381,
+      "step": 111000
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 9.687201454590281e-06,
+      "loss": 4.2349,
+      "step": 111500
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 9.685660575056243e-06,
+      "loss": 4.2354,
+      "step": 112000
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 9.684119695522205e-06,
+      "loss": 4.2302,
+      "step": 112500
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 9.682578815988167e-06,
+      "loss": 4.2239,
+      "step": 113000
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 9.681037936454128e-06,
+      "loss": 4.2288,
+      "step": 113500
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 9.67949705692009e-06,
+      "loss": 4.2229,
+      "step": 114000
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 9.677956177386052e-06,
+      "loss": 4.2231,
+      "step": 114500
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 9.676415297852014e-06,
+      "loss": 4.2227,
+      "step": 115000
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 9.674874418317977e-06,
+      "loss": 4.2212,
+      "step": 115500
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 9.673333538783939e-06,
+      "loss": 4.2175,
+      "step": 116000
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 9.671792659249901e-06,
+      "loss": 4.2193,
+      "step": 116500
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 9.670251779715863e-06,
+      "loss": 4.2193,
+      "step": 117000
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 9.668710900181825e-06,
+      "loss": 4.2179,
+      "step": 117500
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 9.667170020647788e-06,
+      "loss": 4.2126,
+      "step": 118000
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 9.665629141113748e-06,
+      "loss": 4.215,
+      "step": 118500
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 9.66408826157971e-06,
+      "loss": 4.2082,
+      "step": 119000
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 9.662547382045672e-06,
+      "loss": 4.2118,
+      "step": 119500
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 9.661006502511635e-06,
+      "loss": 4.2114,
+      "step": 120000
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 9.659465622977597e-06,
+      "loss": 4.2077,
+      "step": 120500
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 9.657924743443559e-06,
+      "loss": 4.2069,
+      "step": 121000
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 9.656383863909521e-06,
+      "loss": 4.203,
+      "step": 121500
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 9.654842984375482e-06,
+      "loss": 4.2035,
+      "step": 122000
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 9.653302104841444e-06,
+      "loss": 4.2051,
+      "step": 122500
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 9.651761225307406e-06,
+      "loss": 4.2018,
+      "step": 123000
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 9.650220345773368e-06,
+      "loss": 4.1997,
+      "step": 123500
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 9.64867946623933e-06,
+      "loss": 4.1917,
+      "step": 124000
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 9.647138586705292e-06,
+      "loss": 4.1984,
+      "step": 124500
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 9.645597707171253e-06,
+      "loss": 4.1963,
+      "step": 125000
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 9.644056827637215e-06,
+      "loss": 4.1966,
+      "step": 125500
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 9.642515948103179e-06,
+      "loss": 4.1933,
+      "step": 126000
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 9.640975068569141e-06,
+      "loss": 4.1864,
+      "step": 126500
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 9.639434189035102e-06,
+      "loss": 4.191,
+      "step": 127000
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 9.637893309501064e-06,
+      "loss": 4.1877,
+      "step": 127500
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 9.636352429967026e-06,
+      "loss": 4.1884,
+      "step": 128000
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 9.634811550432988e-06,
+      "loss": 4.1836,
+      "step": 128500
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 9.63327067089895e-06,
+      "loss": 4.1841,
+      "step": 129000
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 9.631729791364913e-06,
+      "loss": 4.1845,
+      "step": 129500
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 9.630188911830873e-06,
+      "loss": 4.184,
+      "step": 130000
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 9.628648032296835e-06,
+      "loss": 4.1809,
+      "step": 130500
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 9.627107152762797e-06,
+      "loss": 4.1775,
+      "step": 131000
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 9.62556627322876e-06,
+      "loss": 4.1794,
+      "step": 131500
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 9.624025393694722e-06,
+      "loss": 4.1816,
+      "step": 132000
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 9.622484514160684e-06,
+      "loss": 4.1774,
+      "step": 132500
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 9.620943634626646e-06,
+      "loss": 4.1768,
+      "step": 133000
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 9.619402755092607e-06,
+      "loss": 4.1693,
+      "step": 133500
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 9.617861875558569e-06,
+      "loss": 4.1768,
+      "step": 134000
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 9.616320996024531e-06,
+      "loss": 4.1723,
+      "step": 134500
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 9.614780116490493e-06,
+      "loss": 4.1722,
+      "step": 135000
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 9.613239236956455e-06,
+      "loss": 4.1702,
+      "step": 135500
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 9.611698357422418e-06,
+      "loss": 4.1662,
+      "step": 136000
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 9.61015747788838e-06,
+      "loss": 4.1664,
+      "step": 136500
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 9.608616598354342e-06,
+      "loss": 4.1633,
+      "step": 137000
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 9.607075718820304e-06,
+      "loss": 4.1651,
+      "step": 137500
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 9.605534839286266e-06,
+      "loss": 4.1622,
+      "step": 138000
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 9.603993959752227e-06,
+      "loss": 4.1558,
+      "step": 138500
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 9.602453080218189e-06,
+      "loss": 4.1611,
+      "step": 139000
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 9.600912200684151e-06,
+      "loss": 4.1569,
+      "step": 139500
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 9.599371321150113e-06,
+      "loss": 4.1593,
+      "step": 140000
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 9.597830441616075e-06,
+      "loss": 4.1582,
+      "step": 140500
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 9.596289562082038e-06,
+      "loss": 4.1618,
+      "step": 141000
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 9.594748682548e-06,
+      "loss": 4.1558,
+      "step": 141500
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 9.59320780301396e-06,
+      "loss": 4.1572,
+      "step": 142000
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 9.591666923479922e-06,
+      "loss": 4.1524,
+      "step": 142500
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 9.590126043945885e-06,
+      "loss": 4.1534,
+      "step": 143000
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 9.588585164411847e-06,
+      "loss": 4.1519,
+      "step": 143500
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 9.587044284877809e-06,
+      "loss": 4.15,
+      "step": 144000
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 9.585503405343771e-06,
+      "loss": 4.1493,
+      "step": 144500
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 9.583962525809733e-06,
+      "loss": 4.1481,
+      "step": 145000
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 9.582421646275694e-06,
+      "loss": 4.1491,
+      "step": 145500
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 9.580880766741658e-06,
+      "loss": 4.1463,
+      "step": 146000
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 9.57933988720762e-06,
+      "loss": 4.1439,
+      "step": 146500
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 9.57779900767358e-06,
+      "loss": 4.1429,
+      "step": 147000
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 9.576258128139543e-06,
+      "loss": 4.1426,
+      "step": 147500
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 9.574717248605505e-06,
+      "loss": 4.1426,
+      "step": 148000
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 9.573176369071467e-06,
+      "loss": 4.1425,
+      "step": 148500
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 9.571635489537429e-06,
+      "loss": 4.1411,
+      "step": 149000
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 9.570094610003391e-06,
+      "loss": 4.1404,
+      "step": 149500
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 9.568553730469353e-06,
+      "loss": 4.1386,
+      "step": 150000
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 9.567012850935314e-06,
+      "loss": 4.1382,
+      "step": 150500
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 9.565471971401276e-06,
+      "loss": 4.1366,
+      "step": 151000
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 9.563931091867238e-06,
+      "loss": 4.1353,
+      "step": 151500
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 9.5623902123332e-06,
+      "loss": 4.134,
+      "step": 152000
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 9.560849332799163e-06,
+      "loss": 4.134,
+      "step": 152500
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 9.559308453265125e-06,
+      "loss": 4.1355,
+      "step": 153000
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 9.557767573731085e-06,
+      "loss": 4.1331,
+      "step": 153500
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 9.556226694197048e-06,
+      "loss": 4.1326,
+      "step": 154000
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 9.554685814663011e-06,
+      "loss": 4.1292,
+      "step": 154500
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 9.553144935128974e-06,
+      "loss": 4.1324,
+      "step": 155000
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 9.551604055594934e-06,
+      "loss": 4.1314,
+      "step": 155500
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 9.550063176060896e-06,
+      "loss": 4.1295,
+      "step": 156000
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 9.548522296526858e-06,
+      "loss": 4.1288,
+      "step": 156500
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 9.54698141699282e-06,
+      "loss": 4.1271,
+      "step": 157000
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 9.545440537458783e-06,
+      "loss": 4.1264,
+      "step": 157500
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 9.543899657924745e-06,
+      "loss": 4.1236,
+      "step": 158000
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 9.542358778390705e-06,
+      "loss": 4.1247,
+      "step": 158500
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 9.540817898856668e-06,
+      "loss": 4.1251,
+      "step": 159000
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 9.53927701932263e-06,
+      "loss": 4.1222,
+      "step": 159500
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 9.537736139788592e-06,
+      "loss": 4.1187,
+      "step": 160000
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 9.536195260254554e-06,
+      "loss": 4.1205,
+      "step": 160500
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 9.534654380720516e-06,
+      "loss": 4.1181,
+      "step": 161000
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 9.533113501186479e-06,
+      "loss": 4.1203,
+      "step": 161500
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 9.531572621652439e-06,
+      "loss": 4.1197,
+      "step": 162000
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 9.530031742118401e-06,
+      "loss": 4.1159,
+      "step": 162500
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 9.528490862584363e-06,
+      "loss": 4.1151,
+      "step": 163000
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 9.526949983050326e-06,
+      "loss": 4.1164,
+      "step": 163500
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 9.525409103516288e-06,
+      "loss": 4.1162,
+      "step": 164000
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 9.52386822398225e-06,
+      "loss": 4.1144,
+      "step": 164500
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 9.522327344448212e-06,
+      "loss": 4.1176,
+      "step": 165000
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 9.520786464914174e-06,
+      "loss": 4.1118,
+      "step": 165500
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 9.519245585380136e-06,
+      "loss": 4.112,
+      "step": 166000
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 9.517704705846099e-06,
+      "loss": 4.1125,
+      "step": 166500
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 9.516163826312059e-06,
+      "loss": 4.114,
+      "step": 167000
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 9.514622946778021e-06,
+      "loss": 4.1119,
+      "step": 167500
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 9.513082067243983e-06,
+      "loss": 4.1107,
+      "step": 168000
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 9.511541187709946e-06,
+      "loss": 4.1102,
+      "step": 168500
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 9.510000308175908e-06,
+      "loss": 4.1088,
+      "step": 169000
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 9.50845942864187e-06,
+      "loss": 4.109,
+      "step": 169500
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 9.506918549107832e-06,
+      "loss": 4.1072,
+      "step": 170000
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 9.505377669573793e-06,
+      "loss": 4.1102,
+      "step": 170500
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 9.503836790039755e-06,
+      "loss": 4.1066,
+      "step": 171000
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 9.502295910505717e-06,
+      "loss": 4.1067,
+      "step": 171500
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 9.50075503097168e-06,
+      "loss": 4.1056,
+      "step": 172000
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 9.499214151437641e-06,
+      "loss": 4.1018,
+      "step": 172500
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 9.497673271903604e-06,
+      "loss": 4.1047,
+      "step": 173000
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 9.496132392369566e-06,
+      "loss": 4.1014,
+      "step": 173500
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 9.494591512835526e-06,
+      "loss": 4.1063,
+      "step": 174000
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 9.49305063330149e-06,
+      "loss": 4.1036,
+      "step": 174500
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 9.491509753767452e-06,
+      "loss": 4.102,
+      "step": 175000
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 9.489968874233413e-06,
+      "loss": 4.1008,
+      "step": 175500
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 9.488427994699375e-06,
+      "loss": 4.1017,
+      "step": 176000
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 9.486887115165337e-06,
+      "loss": 4.0991,
+      "step": 176500
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 9.4853462356313e-06,
+      "loss": 4.0965,
+      "step": 177000
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 9.483805356097262e-06,
+      "loss": 4.0978,
+      "step": 177500
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 9.482264476563224e-06,
+      "loss": 4.0973,
+      "step": 178000
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 9.480723597029186e-06,
+      "loss": 4.094,
+      "step": 178500
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 9.479182717495146e-06,
+      "loss": 4.0947,
+      "step": 179000
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 9.477641837961109e-06,
+      "loss": 4.0945,
+      "step": 179500
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 9.47610095842707e-06,
+      "loss": 4.0949,
+      "step": 180000
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 9.474560078893033e-06,
+      "loss": 4.0943,
+      "step": 180500
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 9.473019199358995e-06,
+      "loss": 4.0936,
+      "step": 181000
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 9.471478319824957e-06,
+      "loss": 4.0883,
+      "step": 181500
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 9.469937440290918e-06,
+      "loss": 4.0892,
+      "step": 182000
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 9.46839656075688e-06,
+      "loss": 4.0883,
+      "step": 182500
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 9.466855681222842e-06,
+      "loss": 4.0895,
+      "step": 183000
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 9.465314801688806e-06,
+      "loss": 4.0914,
+      "step": 183500
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 9.463773922154766e-06,
+      "loss": 4.0909,
+      "step": 184000
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 9.462233042620729e-06,
+      "loss": 4.0862,
+      "step": 184500
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 9.46069216308669e-06,
+      "loss": 4.0922,
+      "step": 185000
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 9.459151283552653e-06,
+      "loss": 4.082,
+      "step": 185500
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 9.457610404018615e-06,
+      "loss": 4.0908,
+      "step": 186000
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 9.456069524484577e-06,
+      "loss": 4.0842,
+      "step": 186500
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 9.454528644950538e-06,
+      "loss": 4.089,
+      "step": 187000
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 9.4529877654165e-06,
+      "loss": 4.0857,
+      "step": 187500
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 9.451446885882462e-06,
+      "loss": 4.0842,
+      "step": 188000
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 9.449906006348424e-06,
+      "loss": 4.0828,
+      "step": 188500
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 9.448365126814387e-06,
+      "loss": 4.0804,
+      "step": 189000
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 9.446824247280349e-06,
+      "loss": 4.0828,
+      "step": 189500
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 9.445283367746311e-06,
+      "loss": 4.0833,
+      "step": 190000
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 9.443742488212271e-06,
+      "loss": 4.0845,
+      "step": 190500
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 9.442201608678234e-06,
+      "loss": 4.0826,
+      "step": 191000
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 9.440660729144196e-06,
+      "loss": 4.0787,
+      "step": 191500
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 9.439119849610158e-06,
+      "loss": 4.0775,
+      "step": 192000
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 9.43757897007612e-06,
+      "loss": 4.081,
+      "step": 192500
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 9.436038090542082e-06,
+      "loss": 4.0827,
+      "step": 193000
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 9.434497211008044e-06,
+      "loss": 4.077,
+      "step": 193500
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 9.432956331474005e-06,
+      "loss": 4.0814,
+      "step": 194000
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 9.431415451939969e-06,
+      "loss": 4.076,
+      "step": 194500
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 9.429874572405931e-06,
+      "loss": 4.0776,
+      "step": 195000
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 9.428333692871891e-06,
+      "loss": 4.0754,
+      "step": 195500
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 9.426792813337854e-06,
+      "loss": 4.077,
+      "step": 196000
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 9.425251933803816e-06,
+      "loss": 4.0764,
+      "step": 196500
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 9.423711054269778e-06,
+      "loss": 4.0774,
+      "step": 197000
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 9.42217017473574e-06,
+      "loss": 4.0724,
+      "step": 197500
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 9.420629295201702e-06,
+      "loss": 4.0728,
+      "step": 198000
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 9.419088415667665e-06,
+      "loss": 4.0731,
+      "step": 198500
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 9.417547536133625e-06,
+      "loss": 4.0692,
+      "step": 199000
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 9.416006656599587e-06,
+      "loss": 4.0773,
+      "step": 199500
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 9.41446577706555e-06,
+      "loss": 4.0725,
+      "step": 200000
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 9.412924897531512e-06,
+      "loss": 4.0695,
+      "step": 200500
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 9.411384017997474e-06,
+      "loss": 4.0717,
+      "step": 201000
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 9.409843138463436e-06,
+      "loss": 4.069,
+      "step": 201500
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 9.408302258929398e-06,
+      "loss": 4.0679,
+      "step": 202000
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 9.406761379395359e-06,
+      "loss": 4.0745,
+      "step": 202500
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 9.405220499861322e-06,
+      "loss": 4.0704,
+      "step": 203000
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 9.403679620327285e-06,
+      "loss": 4.0672,
+      "step": 203500
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 9.402138740793245e-06,
+      "loss": 4.0663,
+      "step": 204000
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 9.400597861259207e-06,
+      "loss": 4.0629,
+      "step": 204500
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 9.39905698172517e-06,
+      "loss": 4.0618,
+      "step": 205000
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 9.397516102191132e-06,
+      "loss": 4.0631,
+      "step": 205500
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 9.395975222657094e-06,
+      "loss": 4.0663,
+      "step": 206000
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 9.394434343123056e-06,
+      "loss": 4.0622,
+      "step": 206500
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 9.392893463589018e-06,
+      "loss": 4.0652,
+      "step": 207000
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 9.391352584054979e-06,
+      "loss": 4.0643,
+      "step": 207500
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 9.389811704520941e-06,
+      "loss": 4.0631,
+      "step": 208000
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 9.388270824986903e-06,
+      "loss": 4.0681,
+      "step": 208500
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 9.386729945452865e-06,
+      "loss": 4.0625,
+      "step": 209000
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 9.385189065918827e-06,
+      "loss": 4.0649,
+      "step": 209500
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 9.38364818638479e-06,
+      "loss": 4.0613,
+      "step": 210000
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 9.38210730685075e-06,
+      "loss": 4.0579,
+      "step": 210500
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 9.380566427316712e-06,
+      "loss": 4.0612,
+      "step": 211000
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 9.379025547782674e-06,
+      "loss": 4.0576,
+      "step": 211500
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 9.377484668248638e-06,
+      "loss": 4.0562,
+      "step": 212000
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 9.375943788714599e-06,
+      "loss": 4.0584,
+      "step": 212500
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 9.374402909180561e-06,
+      "loss": 4.0612,
+      "step": 213000
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 9.372862029646523e-06,
+      "loss": 4.0633,
+      "step": 213500
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 9.371321150112485e-06,
+      "loss": 4.0587,
+      "step": 214000
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 9.369780270578448e-06,
+      "loss": 4.0531,
+      "step": 214500
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 9.36823939104441e-06,
+      "loss": 4.0639,
+      "step": 215000
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 9.36669851151037e-06,
+      "loss": 4.0564,
+      "step": 215500
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 9.365157631976332e-06,
+      "loss": 4.0534,
+      "step": 216000
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 9.363616752442295e-06,
+      "loss": 4.0581,
+      "step": 216500
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.362075872908257e-06,
+      "loss": 4.0585,
+      "step": 217000
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 9.360534993374219e-06,
+      "loss": 4.0582,
+      "step": 217500
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.358994113840181e-06,
+      "loss": 4.054,
+      "step": 218000
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 9.357453234306143e-06,
+      "loss": 4.0513,
+      "step": 218500
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 9.355912354772104e-06,
+      "loss": 4.0521,
+      "step": 219000
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 9.354371475238066e-06,
+      "loss": 4.055,
+      "step": 219500
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 9.352830595704028e-06,
+      "loss": 4.0521,
+      "step": 220000
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.35128971616999e-06,
+      "loss": 4.0513,
+      "step": 220500
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 9.349748836635952e-06,
+      "loss": 4.0543,
+      "step": 221000
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.348207957101915e-06,
+      "loss": 4.0512,
+      "step": 221500
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 9.346667077567877e-06,
+      "loss": 4.0516,
+      "step": 222000
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.345126198033837e-06,
+      "loss": 4.0509,
+      "step": 222500
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.343585318499801e-06,
+      "loss": 4.0535,
+      "step": 223000
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 9.342044438965763e-06,
+      "loss": 4.05,
+      "step": 223500
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.340503559431724e-06,
+      "loss": 4.0504,
+      "step": 224000
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 9.338962679897686e-06,
+      "loss": 4.0519,
+      "step": 224500
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.337421800363648e-06,
+      "loss": 4.0497,
+      "step": 225000
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.33588092082961e-06,
+      "loss": 4.048,
+      "step": 225500
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 9.334340041295573e-06,
+      "loss": 4.0442,
+      "step": 226000
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.332799161761535e-06,
+      "loss": 4.047,
+      "step": 226500
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.331258282227497e-06,
+      "loss": 4.049,
+      "step": 227000
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.329717402693457e-06,
+      "loss": 4.0486,
+      "step": 227500
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.32817652315942e-06,
+      "loss": 4.0465,
+      "step": 228000
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 9.326635643625382e-06,
+      "loss": 4.0475,
+      "step": 228500
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 9.325094764091344e-06,
+      "loss": 4.0447,
+      "step": 229000
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 9.323553884557306e-06,
+      "loss": 4.0457,
+      "step": 229500
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 9.322013005023268e-06,
+      "loss": 4.0458,
+      "step": 230000
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 9.32047212548923e-06,
+      "loss": 4.0455,
+      "step": 230500
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 9.318931245955191e-06,
+      "loss": 4.0446,
+      "step": 231000
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 9.317390366421153e-06,
+      "loss": 4.0432,
+      "step": 231500
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 9.315849486887117e-06,
+      "loss": 4.0428,
+      "step": 232000
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 9.314308607353078e-06,
+      "loss": 4.0421,
+      "step": 232500
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 9.31276772781904e-06,
+      "loss": 4.038,
+      "step": 233000
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 9.311226848285002e-06,
+      "loss": 4.04,
+      "step": 233500
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 9.309685968750964e-06,
+      "loss": 4.0387,
+      "step": 234000
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 9.308145089216926e-06,
+      "loss": 4.0426,
+      "step": 234500
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 9.306604209682888e-06,
+      "loss": 4.0376,
+      "step": 235000
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 9.30506333014885e-06,
+      "loss": 4.0393,
+      "step": 235500
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 9.303522450614811e-06,
+      "loss": 4.0375,
+      "step": 236000
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 9.301981571080773e-06,
+      "loss": 4.0395,
+      "step": 236500
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 9.300440691546735e-06,
+      "loss": 4.0416,
+      "step": 237000
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 9.298899812012698e-06,
+      "loss": 4.0358,
+      "step": 237500
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 9.29735893247866e-06,
+      "loss": 4.0349,
+      "step": 238000
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 9.295818052944622e-06,
+      "loss": 4.0391,
+      "step": 238500
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 9.294277173410582e-06,
+      "loss": 4.035,
+      "step": 239000
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 9.292736293876545e-06,
+      "loss": 4.0387,
+      "step": 239500
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 9.291195414342507e-06,
+      "loss": 4.038,
+      "step": 240000
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 9.28965453480847e-06,
+      "loss": 4.039,
+      "step": 240500
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 9.288113655274431e-06,
+      "loss": 4.0359,
+      "step": 241000
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 9.286572775740393e-06,
+      "loss": 4.0368,
+      "step": 241500
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 9.285031896206356e-06,
+      "loss": 4.0367,
+      "step": 242000
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 9.283491016672316e-06,
+      "loss": 4.0345,
+      "step": 242500
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 9.28195013713828e-06,
+      "loss": 4.0366,
+      "step": 243000
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 9.280409257604242e-06,
+      "loss": 4.0354,
+      "step": 243500
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 9.278868378070203e-06,
+      "loss": 4.0381,
+      "step": 244000
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 9.277327498536165e-06,
+      "loss": 4.0354,
+      "step": 244500
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 9.275786619002127e-06,
+      "loss": 4.0357,
+      "step": 245000
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 9.274245739468089e-06,
+      "loss": 4.034,
+      "step": 245500
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 9.272704859934051e-06,
+      "loss": 4.0291,
+      "step": 246000
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 9.271163980400013e-06,
+      "loss": 4.0316,
+      "step": 246500
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 9.269623100865976e-06,
+      "loss": 4.0297,
+      "step": 247000
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 9.268082221331936e-06,
+      "loss": 4.0296,
+      "step": 247500
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 9.266541341797898e-06,
+      "loss": 4.0372,
+      "step": 248000
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 9.26500046226386e-06,
+      "loss": 4.0309,
+      "step": 248500
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 9.263459582729823e-06,
+      "loss": 4.0301,
+      "step": 249000
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 9.261918703195785e-06,
+      "loss": 4.0273,
+      "step": 249500
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 9.260377823661747e-06,
+      "loss": 4.0253,
+      "step": 250000
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 9.25883694412771e-06,
+      "loss": 4.0248,
+      "step": 250500
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 9.25729606459367e-06,
+      "loss": 4.0245,
+      "step": 251000
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 9.255755185059634e-06,
+      "loss": 4.0291,
+      "step": 251500
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 9.254214305525596e-06,
+      "loss": 4.0287,
+      "step": 252000
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 9.252673425991556e-06,
+      "loss": 4.0248,
+      "step": 252500
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 9.251132546457518e-06,
+      "loss": 4.0297,
+      "step": 253000
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 9.24959166692348e-06,
+      "loss": 4.0288,
+      "step": 253500
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 9.248050787389443e-06,
+      "loss": 4.0263,
+      "step": 254000
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 9.246509907855405e-06,
+      "loss": 4.0249,
+      "step": 254500
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 9.244969028321367e-06,
+      "loss": 4.0247,
+      "step": 255000
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 9.24342814878733e-06,
+      "loss": 4.0307,
+      "step": 255500
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 9.24188726925329e-06,
+      "loss": 4.0248,
+      "step": 256000
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 9.240346389719252e-06,
+      "loss": 4.0278,
+      "step": 256500
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 9.238805510185214e-06,
+      "loss": 4.0209,
+      "step": 257000
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 9.237264630651176e-06,
+      "loss": 4.0271,
+      "step": 257500
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 9.235723751117139e-06,
+      "loss": 4.0218,
+      "step": 258000
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 9.2341828715831e-06,
+      "loss": 4.0259,
+      "step": 258500
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 9.232641992049063e-06,
+      "loss": 4.0258,
+      "step": 259000
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 9.231101112515023e-06,
+      "loss": 4.0226,
+      "step": 259500
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 9.229560232980986e-06,
+      "loss": 4.0201,
+      "step": 260000
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 9.22801935344695e-06,
+      "loss": 4.0244,
+      "step": 260500
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 9.22647847391291e-06,
+      "loss": 4.0192,
+      "step": 261000
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 9.224937594378872e-06,
+      "loss": 4.0226,
+      "step": 261500
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 9.223396714844834e-06,
+      "loss": 4.0193,
+      "step": 262000
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 9.221855835310796e-06,
+      "loss": 4.0165,
+      "step": 262500
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 9.220314955776759e-06,
+      "loss": 4.0207,
+      "step": 263000
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 9.21877407624272e-06,
+      "loss": 4.0236,
+      "step": 263500
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 9.217233196708683e-06,
+      "loss": 4.0198,
+      "step": 264000
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 9.215692317174643e-06,
+      "loss": 4.0217,
+      "step": 264500
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 9.214151437640606e-06,
+      "loss": 4.0214,
+      "step": 265000
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 9.212610558106568e-06,
+      "loss": 4.0196,
+      "step": 265500
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 9.21106967857253e-06,
+      "loss": 4.0192,
+      "step": 266000
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 9.209528799038492e-06,
+      "loss": 4.0189,
+      "step": 266500
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 9.207987919504454e-06,
+      "loss": 4.0159,
+      "step": 267000
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 9.206447039970415e-06,
+      "loss": 4.0187,
+      "step": 267500
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 9.204906160436377e-06,
+      "loss": 4.014,
+      "step": 268000
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 9.20336528090234e-06,
+      "loss": 4.0172,
+      "step": 268500
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 9.201824401368301e-06,
+      "loss": 4.0149,
+      "step": 269000
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 9.200283521834264e-06,
+      "loss": 4.0177,
+      "step": 269500
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 9.198742642300226e-06,
+      "loss": 4.0129,
+      "step": 270000
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 9.197201762766188e-06,
+      "loss": 4.0135,
+      "step": 270500
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 9.195660883232148e-06,
+      "loss": 4.0193,
+      "step": 271000
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 9.194120003698112e-06,
+      "loss": 4.0156,
+      "step": 271500
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 9.192579124164074e-06,
+      "loss": 4.0158,
+      "step": 272000
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 9.191038244630035e-06,
+      "loss": 4.0138,
+      "step": 272500
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 9.189497365095997e-06,
+      "loss": 4.0131,
+      "step": 273000
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 9.18795648556196e-06,
+      "loss": 4.0167,
+      "step": 273500
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 9.186415606027921e-06,
+      "loss": 4.0106,
+      "step": 274000
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 9.184874726493884e-06,
+      "loss": 4.0168,
+      "step": 274500
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 9.183333846959846e-06,
+      "loss": 4.0136,
+      "step": 275000
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 9.181792967425808e-06,
+      "loss": 4.0113,
+      "step": 275500
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 9.180252087891769e-06,
+      "loss": 4.0164,
+      "step": 276000
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 9.17871120835773e-06,
+      "loss": 4.0089,
+      "step": 276500
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 9.177170328823693e-06,
+      "loss": 4.0121,
+      "step": 277000
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 9.175629449289655e-06,
+      "loss": 4.013,
+      "step": 277500
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 9.174088569755617e-06,
+      "loss": 4.0093,
+      "step": 278000
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 9.17254769022158e-06,
+      "loss": 4.0108,
+      "step": 278500
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 9.171006810687542e-06,
+      "loss": 4.0134,
+      "step": 279000
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 9.169465931153502e-06,
+      "loss": 4.011,
+      "step": 279500
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 9.167925051619464e-06,
+      "loss": 4.0094,
+      "step": 280000
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 9.166384172085428e-06,
+      "loss": 4.0132,
+      "step": 280500
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 9.164843292551389e-06,
+      "loss": 4.0123,
+      "step": 281000
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 9.16330241301735e-06,
+      "loss": 4.0112,
+      "step": 281500
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 9.161761533483313e-06,
+      "loss": 4.0078,
+      "step": 282000
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 9.160220653949275e-06,
+      "loss": 4.0056,
+      "step": 282500
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 9.158679774415237e-06,
+      "loss": 4.0058,
+      "step": 283000
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 9.1571388948812e-06,
+      "loss": 4.0075,
+      "step": 283500
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 9.155598015347162e-06,
+      "loss": 4.0087,
+      "step": 284000
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 9.154057135813122e-06,
+      "loss": 4.0092,
+      "step": 284500
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 9.152516256279084e-06,
+      "loss": 4.0081,
+      "step": 285000
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 9.150975376745047e-06,
+      "loss": 4.0094,
+      "step": 285500
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 9.149434497211009e-06,
+      "loss": 4.0043,
+      "step": 286000
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 9.147893617676971e-06,
+      "loss": 4.0108,
+      "step": 286500
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 9.146352738142933e-06,
+      "loss": 4.0075,
+      "step": 287000
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 9.144811858608895e-06,
+      "loss": 4.0067,
+      "step": 287500
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 9.143270979074856e-06,
+      "loss": 4.0067,
+      "step": 288000
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 9.141730099540818e-06,
+      "loss": 4.008,
+      "step": 288500
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 9.140189220006782e-06,
+      "loss": 4.0047,
+      "step": 289000
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 9.138648340472742e-06,
+      "loss": 4.0038,
+      "step": 289500
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 9.137107460938704e-06,
+      "loss": 4.0052,
+      "step": 290000
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 9.135566581404667e-06,
+      "loss": 4.0056,
+      "step": 290500
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 9.134025701870629e-06,
+      "loss": 4.0029,
+      "step": 291000
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 9.132484822336591e-06,
+      "loss": 4.0047,
+      "step": 291500
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 9.130943942802553e-06,
+      "loss": 4.0048,
+      "step": 292000
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 9.129403063268515e-06,
+      "loss": 4.0019,
+      "step": 292500
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 9.127862183734476e-06,
+      "loss": 4.0006,
+      "step": 293000
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 9.126321304200438e-06,
+      "loss": 4.0025,
+      "step": 293500
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 9.1247804246664e-06,
+      "loss": 4.0029,
+      "step": 294000
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 9.123239545132362e-06,
+      "loss": 3.999,
+      "step": 294500
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 9.121698665598325e-06,
+      "loss": 4.0019,
+      "step": 295000
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 9.120157786064287e-06,
+      "loss": 4.0034,
+      "step": 295500
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 9.118616906530247e-06,
+      "loss": 4.0018,
+      "step": 296000
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 9.11707602699621e-06,
+      "loss": 4.0,
+      "step": 296500
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 9.115535147462172e-06,
+      "loss": 4.002,
+      "step": 297000
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 9.113994267928134e-06,
+      "loss": 4.0011,
+      "step": 297500
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 9.112453388394096e-06,
+      "loss": 4.0008,
+      "step": 298000
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 9.110912508860058e-06,
+      "loss": 4.0031,
+      "step": 298500
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 9.10937162932602e-06,
+      "loss": 4.0001,
+      "step": 299000
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 9.10783074979198e-06,
+      "loss": 4.0013,
+      "step": 299500
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 9.106289870257945e-06,
+      "loss": 4.0036,
+      "step": 300000
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 9.104748990723907e-06,
+      "loss": 4.0063,
+      "step": 300500
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 9.103208111189869e-06,
+      "loss": 4.005,
+      "step": 301000
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 9.10166723165583e-06,
+      "loss": 3.9976,
+      "step": 301500
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 9.100126352121792e-06,
+      "loss": 3.9937,
+      "step": 302000
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 9.098585472587754e-06,
+      "loss": 3.996,
+      "step": 302500
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 9.097044593053716e-06,
+      "loss": 3.9968,
+      "step": 303000
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 9.095503713519678e-06,
+      "loss": 4.0,
+      "step": 303500
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 9.09396283398564e-06,
+      "loss": 3.996,
+      "step": 304000
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 9.092421954451601e-06,
+      "loss": 3.996,
+      "step": 304500
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 9.090881074917563e-06,
+      "loss": 3.9979,
+      "step": 305000
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 9.089340195383525e-06,
+      "loss": 3.9961,
+      "step": 305500
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 9.087799315849487e-06,
+      "loss": 4.0,
+      "step": 306000
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 9.08625843631545e-06,
+      "loss": 3.9977,
+      "step": 306500
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 9.084717556781412e-06,
+      "loss": 3.9959,
+      "step": 307000
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 9.083176677247374e-06,
+      "loss": 3.9947,
+      "step": 307500
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 9.081635797713334e-06,
+      "loss": 3.9969,
+      "step": 308000
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 9.080094918179297e-06,
+      "loss": 3.9959,
+      "step": 308500
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 9.07855403864526e-06,
+      "loss": 3.9938,
+      "step": 309000
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 9.077013159111221e-06,
+      "loss": 3.9935,
+      "step": 309500
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 9.075472279577183e-06,
+      "loss": 3.9913,
+      "step": 310000
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 9.073931400043145e-06,
+      "loss": 3.9944,
+      "step": 310500
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 9.072390520509108e-06,
+      "loss": 3.9934,
+      "step": 311000
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 9.07084964097507e-06,
+      "loss": 3.9964,
+      "step": 311500
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 9.069308761441032e-06,
+      "loss": 3.9915,
+      "step": 312000
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 9.067767881906994e-06,
+      "loss": 3.9896,
+      "step": 312500
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 9.066227002372955e-06,
+      "loss": 3.9931,
+      "step": 313000
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 9.064686122838917e-06,
+      "loss": 3.9906,
+      "step": 313500
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 9.063145243304879e-06,
+      "loss": 3.9976,
+      "step": 314000
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 9.061604363770841e-06,
+      "loss": 3.9922,
+      "step": 314500
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.060063484236803e-06,
+      "loss": 3.9915,
+      "step": 315000
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 9.058522604702765e-06,
+      "loss": 3.992,
+      "step": 315500
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 9.056981725168728e-06,
+      "loss": 3.9922,
+      "step": 316000
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 9.055440845634688e-06,
+      "loss": 3.9927,
+      "step": 316500
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 9.05389996610065e-06,
+      "loss": 3.9902,
+      "step": 317000
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 9.052359086566614e-06,
+      "loss": 3.9911,
+      "step": 317500
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 9.050818207032575e-06,
+      "loss": 3.9891,
+      "step": 318000
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 9.049277327498537e-06,
+      "loss": 3.9908,
+      "step": 318500
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 9.047736447964499e-06,
+      "loss": 3.9914,
+      "step": 319000
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 9.04619556843046e-06,
+      "loss": 3.9886,
+      "step": 319500
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 9.044654688896423e-06,
+      "loss": 3.9923,
+      "step": 320000
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 9.043113809362386e-06,
+      "loss": 3.9884,
+      "step": 320500
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 9.041572929828348e-06,
+      "loss": 3.9919,
+      "step": 321000
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 9.040032050294308e-06,
+      "loss": 3.9878,
+      "step": 321500
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 9.03849117076027e-06,
+      "loss": 3.9879,
+      "step": 322000
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 9.036950291226233e-06,
+      "loss": 3.9873,
+      "step": 322500
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 9.035409411692195e-06,
+      "loss": 3.986,
+      "step": 323000
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 9.033868532158157e-06,
+      "loss": 3.9881,
+      "step": 323500
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 9.032327652624119e-06,
+      "loss": 3.9869,
+      "step": 324000
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.03078677309008e-06,
+      "loss": 3.9876,
+      "step": 324500
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 9.029245893556042e-06,
+      "loss": 3.9935,
+      "step": 325000
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 9.027705014022004e-06,
+      "loss": 3.9877,
+      "step": 325500
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 9.026164134487966e-06,
+      "loss": 3.9892,
+      "step": 326000
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 9.024623254953928e-06,
+      "loss": 3.9868,
+      "step": 326500
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 9.02308237541989e-06,
+      "loss": 3.9865,
+      "step": 327000
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 9.021541495885853e-06,
+      "loss": 3.9871,
+      "step": 327500
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 9.020000616351813e-06,
+      "loss": 3.9816,
+      "step": 328000
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 9.018459736817777e-06,
+      "loss": 3.9865,
+      "step": 328500
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 9.01691885728374e-06,
+      "loss": 3.9841,
+      "step": 329000
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 9.015377977749701e-06,
+      "loss": 3.9861,
+      "step": 329500
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 9.013837098215662e-06,
+      "loss": 3.9885,
+      "step": 330000
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 9.012296218681624e-06,
+      "loss": 3.9881,
+      "step": 330500
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 9.010755339147586e-06,
+      "loss": 3.9849,
+      "step": 331000
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 9.009214459613548e-06,
+      "loss": 3.9874,
+      "step": 331500
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 9.00767358007951e-06,
+      "loss": 3.9816,
+      "step": 332000
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 9.006132700545473e-06,
+      "loss": 3.9812,
+      "step": 332500
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 9.004591821011433e-06,
+      "loss": 3.9806,
+      "step": 333000
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 9.003050941477395e-06,
+      "loss": 3.9827,
+      "step": 333500
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 9.001510061943358e-06,
+      "loss": 3.983,
+      "step": 334000
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 8.99996918240932e-06,
+      "loss": 3.9768,
+      "step": 334500
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 8.998428302875282e-06,
+      "loss": 3.9837,
+      "step": 335000
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 8.996887423341244e-06,
+      "loss": 3.981,
+      "step": 335500
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 8.995346543807206e-06,
+      "loss": 3.9846,
+      "step": 336000
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 8.993805664273167e-06,
+      "loss": 3.9852,
+      "step": 336500
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 8.992264784739129e-06,
+      "loss": 3.9769,
+      "step": 337000
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 8.990723905205093e-06,
+      "loss": 3.9817,
+      "step": 337500
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 8.989183025671053e-06,
+      "loss": 3.9835,
+      "step": 338000
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 8.987642146137016e-06,
+      "loss": 3.9851,
+      "step": 338500
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 8.986101266602978e-06,
+      "loss": 3.9804,
+      "step": 339000
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 8.98456038706894e-06,
+      "loss": 3.9826,
+      "step": 339500
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 8.983019507534902e-06,
+      "loss": 3.9846,
+      "step": 340000
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 8.981478628000864e-06,
+      "loss": 3.9822,
+      "step": 340500
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 8.979937748466826e-06,
+      "loss": 3.9797,
+      "step": 341000
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 8.978396868932787e-06,
+      "loss": 3.9805,
+      "step": 341500
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 8.976855989398749e-06,
+      "loss": 3.9796,
+      "step": 342000
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 8.975315109864711e-06,
+      "loss": 3.9846,
+      "step": 342500
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 8.973774230330673e-06,
+      "loss": 3.9814,
+      "step": 343000
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 8.972233350796636e-06,
+      "loss": 3.9764,
+      "step": 343500
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 8.970692471262598e-06,
+      "loss": 3.9782,
+      "step": 344000
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 8.96915159172856e-06,
+      "loss": 3.9746,
+      "step": 344500
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 8.96761071219452e-06,
+      "loss": 3.9773,
+      "step": 345000
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 8.966069832660483e-06,
+      "loss": 3.9805,
+      "step": 345500
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 8.964528953126445e-06,
+      "loss": 3.9785,
+      "step": 346000
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 8.962988073592407e-06,
+      "loss": 3.9804,
+      "step": 346500
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 8.96144719405837e-06,
+      "loss": 3.9807,
+      "step": 347000
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 8.959906314524331e-06,
+      "loss": 3.9749,
+      "step": 347500
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 8.958365434990292e-06,
+      "loss": 3.9811,
+      "step": 348000
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 8.956824555456256e-06,
+      "loss": 3.9784,
+      "step": 348500
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 8.955283675922218e-06,
+      "loss": 3.9767,
+      "step": 349000
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 8.95374279638818e-06,
+      "loss": 3.9731,
+      "step": 349500
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 8.95220191685414e-06,
+      "loss": 3.9821,
+      "step": 350000
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 8.950661037320103e-06,
+      "loss": 3.9754,
+      "step": 350500
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 8.949120157786065e-06,
+      "loss": 3.9744,
+      "step": 351000
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 8.947579278252027e-06,
+      "loss": 3.9756,
+      "step": 351500
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 8.94603839871799e-06,
+      "loss": 3.9771,
+      "step": 352000
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 8.944497519183951e-06,
+      "loss": 3.9765,
+      "step": 352500
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 8.942956639649914e-06,
+      "loss": 3.9752,
+      "step": 353000
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 8.941415760115874e-06,
+      "loss": 3.974,
+      "step": 353500
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 8.939874880581836e-06,
+      "loss": 3.9777,
+      "step": 354000
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 8.938334001047799e-06,
+      "loss": 3.9731,
+      "step": 354500
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 8.93679312151376e-06,
+      "loss": 3.9765,
+      "step": 355000
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 8.935252241979723e-06,
+      "loss": 3.9703,
+      "step": 355500
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 8.933711362445685e-06,
+      "loss": 3.9728,
+      "step": 356000
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 8.932170482911646e-06,
+      "loss": 3.9768,
+      "step": 356500
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 8.930629603377608e-06,
+      "loss": 3.9718,
+      "step": 357000
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 8.929088723843572e-06,
+      "loss": 3.977,
+      "step": 357500
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 8.927547844309534e-06,
+      "loss": 3.9746,
+      "step": 358000
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 8.926006964775494e-06,
+      "loss": 3.9735,
+      "step": 358500
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 8.924466085241456e-06,
+      "loss": 3.9726,
+      "step": 359000
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 8.922925205707419e-06,
+      "loss": 3.9728,
+      "step": 359500
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 8.92138432617338e-06,
+      "loss": 3.9771,
+      "step": 360000
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 8.919843446639343e-06,
+      "loss": 3.9706,
+      "step": 360500
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 8.918302567105305e-06,
+      "loss": 3.971,
+      "step": 361000
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 8.916761687571266e-06,
+      "loss": 3.971,
+      "step": 361500
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 8.915220808037228e-06,
+      "loss": 3.9723,
+      "step": 362000
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 8.91367992850319e-06,
+      "loss": 3.9739,
+      "step": 362500
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 8.912139048969152e-06,
+      "loss": 3.9745,
+      "step": 363000
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 8.910598169435114e-06,
+      "loss": 3.9707,
+      "step": 363500
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 8.909057289901077e-06,
+      "loss": 3.9723,
+      "step": 364000
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 8.907516410367039e-06,
+      "loss": 3.9718,
+      "step": 364500
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 8.905975530833e-06,
+      "loss": 3.9705,
+      "step": 365000
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 8.904434651298961e-06,
+      "loss": 3.9673,
+      "step": 365500
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 8.902893771764925e-06,
+      "loss": 3.9709,
+      "step": 366000
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 8.901352892230886e-06,
+      "loss": 3.9723,
+      "step": 366500
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 8.899812012696848e-06,
+      "loss": 3.9693,
+      "step": 367000
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 8.89827113316281e-06,
+      "loss": 3.9696,
+      "step": 367500
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 8.896730253628772e-06,
+      "loss": 3.9702,
+      "step": 368000
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 8.895189374094734e-06,
+      "loss": 3.9671,
+      "step": 368500
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 8.893648494560697e-06,
+      "loss": 3.972,
+      "step": 369000
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 8.892107615026659e-06,
+      "loss": 3.9684,
+      "step": 369500
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 8.89056673549262e-06,
+      "loss": 3.9715,
+      "step": 370000
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 8.889025855958581e-06,
+      "loss": 3.9658,
+      "step": 370500
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 8.887484976424544e-06,
+      "loss": 3.9685,
+      "step": 371000
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 8.885944096890506e-06,
+      "loss": 3.9665,
+      "step": 371500
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 8.884403217356468e-06,
+      "loss": 3.9676,
+      "step": 372000
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 8.88286233782243e-06,
+      "loss": 3.9685,
+      "step": 372500
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 8.881321458288392e-06,
+      "loss": 3.9681,
+      "step": 373000
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 8.879780578754353e-06,
+      "loss": 3.9771,
+      "step": 373500
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 8.878239699220315e-06,
+      "loss": 3.9683,
+      "step": 374000
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 8.876698819686277e-06,
+      "loss": 3.9708,
+      "step": 374500
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 8.87515794015224e-06,
+      "loss": 3.9668,
+      "step": 375000
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 8.873617060618202e-06,
+      "loss": 3.9645,
+      "step": 375500
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 8.872076181084164e-06,
+      "loss": 3.9674,
+      "step": 376000
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 8.870535301550124e-06,
+      "loss": 3.9661,
+      "step": 376500
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 8.868994422016088e-06,
+      "loss": 3.9678,
+      "step": 377000
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 8.86745354248205e-06,
+      "loss": 3.9657,
+      "step": 377500
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 8.865912662948012e-06,
+      "loss": 3.9594,
+      "step": 378000
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 8.864371783413973e-06,
+      "loss": 3.9676,
+      "step": 378500
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 8.862830903879935e-06,
+      "loss": 3.9642,
+      "step": 379000
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 8.861290024345897e-06,
+      "loss": 3.9637,
+      "step": 379500
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 8.85974914481186e-06,
+      "loss": 3.966,
+      "step": 380000
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 8.858208265277822e-06,
+      "loss": 3.9694,
+      "step": 380500
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 8.856667385743784e-06,
+      "loss": 3.9647,
+      "step": 381000
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 8.855126506209746e-06,
+      "loss": 3.9598,
+      "step": 381500
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 8.853585626675707e-06,
+      "loss": 3.9658,
+      "step": 382000
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 8.852044747141669e-06,
+      "loss": 3.9636,
+      "step": 382500
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 8.850503867607631e-06,
+      "loss": 3.9633,
+      "step": 383000
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 8.848962988073593e-06,
+      "loss": 3.96,
+      "step": 383500
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 8.847422108539555e-06,
+      "loss": 3.9656,
+      "step": 384000
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 8.845881229005517e-06,
+      "loss": 3.9638,
+      "step": 384500
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 8.844340349471478e-06,
+      "loss": 3.9647,
+      "step": 385000
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 8.84279946993744e-06,
+      "loss": 3.9634,
+      "step": 385500
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 8.841258590403404e-06,
+      "loss": 3.9622,
+      "step": 386000
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 8.839717710869366e-06,
+      "loss": 3.9682,
+      "step": 386500
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 8.838176831335327e-06,
+      "loss": 3.9624,
+      "step": 387000
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 8.836635951801289e-06,
+      "loss": 3.9634,
+      "step": 387500
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 8.835095072267251e-06,
+      "loss": 3.9596,
+      "step": 388000
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 8.833554192733213e-06,
+      "loss": 3.9633,
+      "step": 388500
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 8.832013313199175e-06,
+      "loss": 3.9654,
+      "step": 389000
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 8.830472433665138e-06,
+      "loss": 3.9634,
+      "step": 389500
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 8.828931554131098e-06,
+      "loss": 3.9607,
+      "step": 390000
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 8.82739067459706e-06,
+      "loss": 3.9624,
+      "step": 390500
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 8.825849795063022e-06,
+      "loss": 3.9664,
+      "step": 391000
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 8.824308915528985e-06,
+      "loss": 3.9613,
+      "step": 391500
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 8.822768035994947e-06,
+      "loss": 3.9596,
+      "step": 392000
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 8.821227156460909e-06,
+      "loss": 3.9561,
+      "step": 392500
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 8.819686276926871e-06,
+      "loss": 3.958,
+      "step": 393000
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 8.818145397392832e-06,
+      "loss": 3.9566,
+      "step": 393500
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 8.816604517858794e-06,
+      "loss": 3.9601,
+      "step": 394000
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 8.815063638324756e-06,
+      "loss": 3.9635,
+      "step": 394500
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 8.813522758790718e-06,
+      "loss": 3.9623,
+      "step": 395000
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 8.81198187925668e-06,
+      "loss": 3.9583,
+      "step": 395500
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 8.810440999722642e-06,
+      "loss": 3.9613,
+      "step": 396000
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 8.808900120188605e-06,
+      "loss": 3.9597,
+      "step": 396500
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 8.807359240654567e-06,
+      "loss": 3.9607,
+      "step": 397000
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 8.805818361120529e-06,
+      "loss": 3.9579,
+      "step": 397500
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 8.804277481586491e-06,
+      "loss": 3.9627,
+      "step": 398000
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 8.802736602052452e-06,
+      "loss": 3.9598,
+      "step": 398500
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 8.801195722518414e-06,
+      "loss": 3.9569,
+      "step": 399000
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 8.799654842984376e-06,
+      "loss": 3.9608,
+      "step": 399500
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 8.798113963450338e-06,
+      "loss": 3.9568,
+      "step": 400000
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 8.7965730839163e-06,
+      "loss": 3.9574,
+      "step": 400500
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 8.795032204382263e-06,
+      "loss": 3.9555,
+      "step": 401000
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 8.793491324848225e-06,
+      "loss": 3.9551,
+      "step": 401500
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 8.791950445314185e-06,
+      "loss": 3.9565,
+      "step": 402000
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 8.790409565780147e-06,
+      "loss": 3.955,
+      "step": 402500
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 8.78886868624611e-06,
+      "loss": 3.957,
+      "step": 403000
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 8.787327806712072e-06,
+      "loss": 3.9543,
+      "step": 403500
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 8.785786927178034e-06,
+      "loss": 3.9592,
+      "step": 404000
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 8.784246047643996e-06,
+      "loss": 3.9573,
+      "step": 404500
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 8.782705168109958e-06,
+      "loss": 3.9549,
+      "step": 405000
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 8.781164288575919e-06,
+      "loss": 3.9592,
+      "step": 405500
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 8.779623409041883e-06,
+      "loss": 3.9577,
+      "step": 406000
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 8.778082529507845e-06,
+      "loss": 3.9564,
+      "step": 406500
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 8.776541649973805e-06,
+      "loss": 3.9579,
+      "step": 407000
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 8.775000770439768e-06,
+      "loss": 3.9576,
+      "step": 407500
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 8.77345989090573e-06,
+      "loss": 3.9524,
+      "step": 408000
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 8.771919011371692e-06,
+      "loss": 3.9548,
+      "step": 408500
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 8.770378131837654e-06,
+      "loss": 3.9589,
+      "step": 409000
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 8.768837252303616e-06,
+      "loss": 3.9554,
+      "step": 409500
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 8.767296372769578e-06,
+      "loss": 3.9558,
+      "step": 410000
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 8.765755493235539e-06,
+      "loss": 3.955,
+      "step": 410500
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 8.764214613701501e-06,
+      "loss": 3.9502,
+      "step": 411000
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 8.762673734167463e-06,
+      "loss": 3.9552,
+      "step": 411500
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 8.761132854633425e-06,
+      "loss": 3.9511,
+      "step": 412000
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 8.759591975099388e-06,
+      "loss": 3.9576,
+      "step": 412500
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 8.75805109556535e-06,
+      "loss": 3.9553,
+      "step": 413000
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 8.75651021603131e-06,
+      "loss": 3.956,
+      "step": 413500
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 8.754969336497272e-06,
+      "loss": 3.9538,
+      "step": 414000
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 8.753428456963236e-06,
+      "loss": 3.9515,
+      "step": 414500
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 8.751887577429199e-06,
+      "loss": 3.9548,
+      "step": 415000
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 8.750346697895159e-06,
+      "loss": 3.9533,
+      "step": 415500
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 8.748805818361121e-06,
+      "loss": 3.9575,
+      "step": 416000
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 8.747264938827083e-06,
+      "loss": 3.9508,
+      "step": 416500
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 8.745724059293046e-06,
+      "loss": 3.9525,
+      "step": 417000
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 8.744183179759008e-06,
+      "loss": 3.9533,
+      "step": 417500
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 8.74264230022497e-06,
+      "loss": 3.9489,
+      "step": 418000
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 8.74110142069093e-06,
+      "loss": 3.9556,
+      "step": 418500
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 8.739560541156893e-06,
+      "loss": 3.9517,
+      "step": 419000
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 8.738019661622855e-06,
+      "loss": 3.9517,
+      "step": 419500
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 8.736478782088817e-06,
+      "loss": 3.9525,
+      "step": 420000
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 8.734937902554779e-06,
+      "loss": 3.9518,
+      "step": 420500
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 8.733397023020741e-06,
+      "loss": 3.952,
+      "step": 421000
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 8.731856143486703e-06,
+      "loss": 3.9539,
+      "step": 421500
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 8.730315263952664e-06,
+      "loss": 3.9547,
+      "step": 422000
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 8.728774384418626e-06,
+      "loss": 3.9506,
+      "step": 422500
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 8.727233504884588e-06,
+      "loss": 3.9527,
+      "step": 423000
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 8.72569262535055e-06,
+      "loss": 3.9541,
+      "step": 423500
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 8.724151745816513e-06,
+      "loss": 3.9535,
+      "step": 424000
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 8.722610866282475e-06,
+      "loss": 3.9459,
+      "step": 424500
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 8.721069986748437e-06,
+      "loss": 3.948,
+      "step": 425000
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 8.7195291072144e-06,
+      "loss": 3.9508,
+      "step": 425500
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 8.717988227680361e-06,
+      "loss": 3.9468,
+      "step": 426000
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 8.716447348146324e-06,
+      "loss": 3.9413,
+      "step": 426500
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 8.714906468612284e-06,
+      "loss": 3.9503,
+      "step": 427000
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 8.713365589078246e-06,
+      "loss": 3.9546,
+      "step": 427500
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 8.711824709544208e-06,
+      "loss": 3.948,
+      "step": 428000
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 8.71028383001017e-06,
+      "loss": 3.9466,
+      "step": 428500
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 8.708742950476133e-06,
+      "loss": 3.952,
+      "step": 429000
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 8.707202070942095e-06,
+      "loss": 3.9521,
+      "step": 429500
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 8.705661191408057e-06,
+      "loss": 3.9488,
+      "step": 430000
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 8.704120311874018e-06,
+      "loss": 3.951,
+      "step": 430500
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 8.70257943233998e-06,
+      "loss": 3.948,
+      "step": 431000
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 8.701038552805942e-06,
+      "loss": 3.9462,
+      "step": 431500
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 8.699497673271904e-06,
+      "loss": 3.9465,
+      "step": 432000
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 8.697956793737866e-06,
+      "loss": 3.9468,
+      "step": 432500
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 8.696415914203829e-06,
+      "loss": 3.9483,
+      "step": 433000
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 8.69487503466979e-06,
+      "loss": 3.9468,
+      "step": 433500
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 8.693334155135751e-06,
+      "loss": 3.9479,
+      "step": 434000
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 8.691793275601715e-06,
+      "loss": 3.9467,
+      "step": 434500
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 8.690252396067677e-06,
+      "loss": 3.9492,
+      "step": 435000
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 8.688711516533638e-06,
+      "loss": 3.946,
+      "step": 435500
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 8.6871706369996e-06,
+      "loss": 3.9488,
+      "step": 436000
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 8.685629757465562e-06,
+      "loss": 3.9498,
+      "step": 436500
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 8.684088877931524e-06,
+      "loss": 3.9463,
+      "step": 437000
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 8.682547998397486e-06,
+      "loss": 3.9503,
+      "step": 437500
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 8.681007118863449e-06,
+      "loss": 3.9485,
+      "step": 438000
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 8.67946623932941e-06,
+      "loss": 3.9488,
+      "step": 438500
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 8.677925359795371e-06,
+      "loss": 3.9456,
+      "step": 439000
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 8.676384480261333e-06,
+      "loss": 3.9491,
+      "step": 439500
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 8.674843600727296e-06,
+      "loss": 3.9463,
+      "step": 440000
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 8.673302721193258e-06,
+      "loss": 3.9483,
+      "step": 440500
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 8.67176184165922e-06,
+      "loss": 3.9461,
+      "step": 441000
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 8.670220962125182e-06,
+      "loss": 3.9433,
+      "step": 441500
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 8.668680082591143e-06,
+      "loss": 3.9456,
+      "step": 442000
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 8.667139203057105e-06,
+      "loss": 3.9461,
+      "step": 442500
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 8.665598323523067e-06,
+      "loss": 3.9451,
+      "step": 443000
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 8.664057443989031e-06,
+      "loss": 3.9437,
+      "step": 443500
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 8.662516564454991e-06,
+      "loss": 3.942,
+      "step": 444000
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 8.660975684920954e-06,
+      "loss": 3.9481,
+      "step": 444500
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 8.659434805386916e-06,
+      "loss": 3.9473,
+      "step": 445000
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 8.657893925852878e-06,
+      "loss": 3.948,
+      "step": 445500
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 8.65635304631884e-06,
+      "loss": 3.9454,
+      "step": 446000
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 8.654812166784802e-06,
+      "loss": 3.9421,
+      "step": 446500
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 8.653271287250763e-06,
+      "loss": 3.9416,
+      "step": 447000
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 8.651730407716725e-06,
+      "loss": 3.9442,
+      "step": 447500
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 8.650189528182687e-06,
+      "loss": 3.944,
+      "step": 448000
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 8.64864864864865e-06,
+      "loss": 3.9441,
+      "step": 448500
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 8.647107769114611e-06,
+      "loss": 3.9433,
+      "step": 449000
+    },
+    {
+      "epoch": 13.8,
+      "step": 449051,
+      "total_flos": 9.008844527527526e+16,
+      "train_loss": 4.181136763339477,
+      "train_runtime": 190799.4817,
+      "train_samples_per_second": 3070.636,
+      "train_steps_per_second": 17.059
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 3254900,
+  "num_train_epochs": 100,
+  "save_steps": 1000000,
+  "total_flos": 9.008844527527526e+16,
+  "trial_name": null,
+  "trial_params": null
+}