diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,16620 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 20.0,
+  "eval_steps": 500,
+  "global_step": 55300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.041591320072332e-10,
+      "loss": 8.7035,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8083182640144665e-09,
+      "loss": 8.5842,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.7124773960216998e-09,
+      "loss": 8.6549,
+      "step": 60
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.616636528028933e-09,
+      "loss": 8.6371,
+      "step": 80
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.5207956600361664e-09,
+      "loss": 8.6676,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.4249547920433996e-09,
+      "loss": 8.6693,
+      "step": 120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.329113924050633e-09,
+      "loss": 8.6823,
+      "step": 140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.233273056057866e-09,
+      "loss": 8.6305,
+      "step": 160
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 8.1374321880651e-09,
+      "loss": 8.7476,
+      "step": 180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.041591320072333e-09,
+      "loss": 8.6453,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.945750452079566e-09,
+      "loss": 8.7242,
+      "step": 220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.0849909584086799e-08,
+      "loss": 8.8198,
+      "step": 240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.1754068716094032e-08,
+      "loss": 8.7031,
+      "step": 260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.2658227848101265e-08,
+      "loss": 8.6468,
+      "step": 280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.3562386980108499e-08,
+      "loss": 8.5893,
+      "step": 300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.4466546112115732e-08,
+      "loss": 8.6628,
+      "step": 320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.5370705244122963e-08,
+      "loss": 8.5913,
+      "step": 340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.62748643761302e-08,
+      "loss": 8.6978,
+      "step": 360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.7179023508137433e-08,
+      "loss": 8.6814,
+      "step": 380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.8083182640144666e-08,
+      "loss": 8.7502,
+      "step": 400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89873417721519e-08,
+      "loss": 8.5715,
+      "step": 420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9891500904159132e-08,
+      "loss": 8.7008,
+      "step": 440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.0795660036166365e-08,
+      "loss": 8.6715,
+      "step": 460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2.1699819168173598e-08,
+      "loss": 8.5782,
+      "step": 480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2.260397830018083e-08,
+      "loss": 8.6972,
+      "step": 500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.3508137432188065e-08,
+      "loss": 8.7056,
+      "step": 520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.4412296564195298e-08,
+      "loss": 8.7433,
+      "step": 540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2.531645569620253e-08,
+      "loss": 8.6857,
+      "step": 560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2.6220614828209764e-08,
+      "loss": 8.6377,
+      "step": 580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.7124773960216997e-08,
+      "loss": 8.6514,
+      "step": 600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2.802893309222423e-08,
+      "loss": 8.668,
+      "step": 620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2.8933092224231463e-08,
+      "loss": 8.7431,
+      "step": 640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2.983725135623869e-08,
+      "loss": 8.7326,
+      "step": 660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.0741410488245926e-08,
+      "loss": 8.7252,
+      "step": 680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 3.1645569620253166e-08,
+      "loss": 8.6977,
+      "step": 700
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.25497287522604e-08,
+      "loss": 8.7162,
+      "step": 720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.345388788426763e-08,
+      "loss": 8.6373,
+      "step": 740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 3.4358047016274865e-08,
+      "loss": 8.5484,
+      "step": 760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 3.52622061482821e-08,
+      "loss": 8.6158,
+      "step": 780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 3.616636528028933e-08,
+      "loss": 8.6236,
+      "step": 800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.7070524412296565e-08,
+      "loss": 8.6796,
+      "step": 820
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 3.79746835443038e-08,
+      "loss": 8.6624,
+      "step": 840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.887884267631103e-08,
+      "loss": 8.6699,
+      "step": 860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9783001808318264e-08,
+      "loss": 8.6308,
+      "step": 880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.06871609403255e-08,
+      "loss": 8.5808,
+      "step": 900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.159132007233273e-08,
+      "loss": 8.6027,
+      "step": 920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.2495479204339963e-08,
+      "loss": 8.7472,
+      "step": 940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.3399638336347197e-08,
+      "loss": 8.6128,
+      "step": 960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.430379746835443e-08,
+      "loss": 8.7902,
+      "step": 980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.520795660036166e-08,
+      "loss": 8.6797,
+      "step": 1000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.6112115732368896e-08,
+      "loss": 8.6378,
+      "step": 1020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.701627486437613e-08,
+      "loss": 8.6656,
+      "step": 1040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.792043399638336e-08,
+      "loss": 8.6539,
+      "step": 1060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.8824593128390595e-08,
+      "loss": 8.7486,
+      "step": 1080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.972875226039783e-08,
+      "loss": 8.6913,
+      "step": 1100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.063291139240506e-08,
+      "loss": 8.8054,
+      "step": 1120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.1537070524412295e-08,
+      "loss": 8.6303,
+      "step": 1140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 5.244122965641953e-08,
+      "loss": 8.6419,
+      "step": 1160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.334538878842676e-08,
+      "loss": 8.6868,
+      "step": 1180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.4249547920433994e-08,
+      "loss": 8.7265,
+      "step": 1200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.515370705244123e-08,
+      "loss": 8.6924,
+      "step": 1220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.605786618444846e-08,
+      "loss": 8.6787,
+      "step": 1240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6962025316455693e-08,
+      "loss": 8.693,
+      "step": 1260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.7866184448462927e-08,
+      "loss": 8.5927,
+      "step": 1280
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.877034358047016e-08,
+      "loss": 8.6922,
+      "step": 1300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.967450271247739e-08,
+      "loss": 8.9393,
+      "step": 1320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 6.057866184448463e-08,
+      "loss": 8.7923,
+      "step": 1340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 6.148282097649185e-08,
+      "loss": 8.7394,
+      "step": 1360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 6.238698010849909e-08,
+      "loss": 8.7055,
+      "step": 1380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 6.329113924050633e-08,
+      "loss": 8.7928,
+      "step": 1400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 6.419529837251356e-08,
+      "loss": 8.6412,
+      "step": 1420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 6.50994575045208e-08,
+      "loss": 8.6909,
+      "step": 1440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 6.600361663652802e-08,
+      "loss": 8.6415,
+      "step": 1460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 6.690777576853526e-08,
+      "loss": 8.7029,
+      "step": 1480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 6.781193490054249e-08,
+      "loss": 8.6085,
+      "step": 1500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.871609403254973e-08,
+      "loss": 8.6977,
+      "step": 1520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.962025316455696e-08,
+      "loss": 8.6144,
+      "step": 1540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.05244122965642e-08,
+      "loss": 8.647,
+      "step": 1560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 7.142857142857142e-08,
+      "loss": 8.701,
+      "step": 1580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.233273056057866e-08,
+      "loss": 8.6896,
+      "step": 1600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.323688969258589e-08,
+      "loss": 8.7102,
+      "step": 1620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.414104882459313e-08,
+      "loss": 8.7441,
+      "step": 1640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.504520795660036e-08,
+      "loss": 8.6805,
+      "step": 1660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.59493670886076e-08,
+      "loss": 8.7272,
+      "step": 1680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.685352622061482e-08,
+      "loss": 8.6238,
+      "step": 1700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 7.775768535262206e-08,
+      "loss": 8.8162,
+      "step": 1720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 7.866184448462929e-08,
+      "loss": 8.6112,
+      "step": 1740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 7.956600361663653e-08,
+      "loss": 8.6329,
+      "step": 1760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.047016274864375e-08,
+      "loss": 8.5605,
+      "step": 1780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.1374321880651e-08,
+      "loss": 8.6107,
+      "step": 1800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.227848101265822e-08,
+      "loss": 8.6376,
+      "step": 1820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.318264014466546e-08,
+      "loss": 8.6818,
+      "step": 1840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.408679927667269e-08,
+      "loss": 8.7487,
+      "step": 1860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 8.499095840867993e-08,
+      "loss": 8.6741,
+      "step": 1880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.589511754068715e-08,
+      "loss": 8.6679,
+      "step": 1900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 8.679927667269439e-08,
+      "loss": 8.6636,
+      "step": 1920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 8.770343580470162e-08,
+      "loss": 8.7159,
+      "step": 1940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 8.860759493670886e-08,
+      "loss": 8.5982,
+      "step": 1960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 8.951175406871609e-08,
+      "loss": 8.7386,
+      "step": 1980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.041591320072333e-08,
+      "loss": 8.6729,
+      "step": 2000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.132007233273057e-08,
+      "loss": 8.6625,
+      "step": 2020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.222423146473779e-08,
+      "loss": 8.6866,
+      "step": 2040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.312839059674503e-08,
+      "loss": 8.6284,
+      "step": 2060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.403254972875226e-08,
+      "loss": 8.6068,
+      "step": 2080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.49367088607595e-08,
+      "loss": 8.7023,
+      "step": 2100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.584086799276672e-08,
+      "loss": 8.6516,
+      "step": 2120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.674502712477396e-08,
+      "loss": 8.746,
+      "step": 2140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.764918625678119e-08,
+      "loss": 8.5947,
+      "step": 2160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.855334538878843e-08,
+      "loss": 8.7309,
+      "step": 2180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.945750452079566e-08,
+      "loss": 8.7407,
+      "step": 2200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.003616636528029e-07,
+      "loss": 8.7647,
+      "step": 2220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.0126582278481012e-07,
+      "loss": 8.6067,
+      "step": 2240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0216998191681736e-07,
+      "loss": 8.6241,
+      "step": 2260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.0307414104882459e-07,
+      "loss": 8.6616,
+      "step": 2280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.0397830018083183e-07,
+      "loss": 8.7636,
+      "step": 2300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.0488245931283906e-07,
+      "loss": 8.6465,
+      "step": 2320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.057866184448463e-07,
+      "loss": 8.6704,
+      "step": 2340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0669077757685352e-07,
+      "loss": 8.5897,
+      "step": 2360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0759493670886076e-07,
+      "loss": 8.664,
+      "step": 2380
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.0849909584086799e-07,
+      "loss": 8.7798,
+      "step": 2400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0940325497287523e-07,
+      "loss": 8.6508,
+      "step": 2420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1030741410488245e-07,
+      "loss": 8.7532,
+      "step": 2440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.112115732368897e-07,
+      "loss": 8.6628,
+      "step": 2460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1211573236889692e-07,
+      "loss": 8.6175,
+      "step": 2480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1301989150090416e-07,
+      "loss": 8.7609,
+      "step": 2500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1392405063291139e-07,
+      "loss": 8.7609,
+      "step": 2520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1482820976491863e-07,
+      "loss": 8.5844,
+      "step": 2540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1573236889692585e-07,
+      "loss": 8.6872,
+      "step": 2560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1663652802893309e-07,
+      "loss": 8.7795,
+      "step": 2580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.1754068716094032e-07,
+      "loss": 8.7157,
+      "step": 2600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1844484629294756e-07,
+      "loss": 8.767,
+      "step": 2620
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1934900542495477e-07,
+      "loss": 8.6297,
+      "step": 2640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.20253164556962e-07,
+      "loss": 8.6604,
+      "step": 2660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.2115732368896925e-07,
+      "loss": 8.6753,
+      "step": 2680
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.220614828209765e-07,
+      "loss": 8.8354,
+      "step": 2700
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.229656419529837e-07,
+      "loss": 8.6935,
+      "step": 2720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2386980108499094e-07,
+      "loss": 8.6787,
+      "step": 2740
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2477396021699818e-07,
+      "loss": 8.6191,
+      "step": 2760
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.256781193490054e-07,
+      "loss": 8.6495,
+      "step": 2780
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.2658227848101266e-07,
+      "loss": 8.7641,
+      "step": 2800
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.2748643761301988e-07,
+      "loss": 8.7237,
+      "step": 2820
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.2839059674502712e-07,
+      "loss": 8.6795,
+      "step": 2840
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.2929475587703433e-07,
+      "loss": 8.6644,
+      "step": 2860
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.301989150090416e-07,
+      "loss": 8.5847,
+      "step": 2880
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.311030741410488e-07,
+      "loss": 8.5999,
+      "step": 2900
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.3200723327305605e-07,
+      "loss": 8.6626,
+      "step": 2920
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.3291139240506326e-07,
+      "loss": 8.7718,
+      "step": 2940
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.3381555153707053e-07,
+      "loss": 8.5763,
+      "step": 2960
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.3471971066907774e-07,
+      "loss": 8.6696,
+      "step": 2980
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.3562386980108498e-07,
+      "loss": 8.6623,
+      "step": 3000
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3652802893309222e-07,
+      "loss": 8.6512,
+      "step": 3020
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3743218806509946e-07,
+      "loss": 8.6001,
+      "step": 3040
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3833634719710667e-07,
+      "loss": 8.6426,
+      "step": 3060
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3924050632911391e-07,
+      "loss": 8.7206,
+      "step": 3080
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.4014466546112115e-07,
+      "loss": 8.6603,
+      "step": 3100
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.410488245931284e-07,
+      "loss": 8.6396,
+      "step": 3120
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.419529837251356e-07,
+      "loss": 8.7162,
+      "step": 3140
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.4285714285714285e-07,
+      "loss": 8.6815,
+      "step": 3160
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.4376130198915009e-07,
+      "loss": 8.6283,
+      "step": 3180
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4466546112115733e-07,
+      "loss": 8.76,
+      "step": 3200
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.4556962025316454e-07,
+      "loss": 8.6929,
+      "step": 3220
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.4647377938517178e-07,
+      "loss": 8.5704,
+      "step": 3240
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.4737793851717902e-07,
+      "loss": 8.7375,
+      "step": 3260
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.4828209764918626e-07,
+      "loss": 8.7367,
+      "step": 3280
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.4918625678119347e-07,
+      "loss": 8.7486,
+      "step": 3300
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.500904159132007e-07,
+      "loss": 8.6897,
+      "step": 3320
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.5099457504520795e-07,
+      "loss": 8.6202,
+      "step": 3340
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.518987341772152e-07,
+      "loss": 8.6978,
+      "step": 3360
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.528028933092224e-07,
+      "loss": 8.6673,
+      "step": 3380
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.5370705244122964e-07,
+      "loss": 8.6666,
+      "step": 3400
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.5461121157323688e-07,
+      "loss": 8.6343,
+      "step": 3420
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.5551537070524412e-07,
+      "loss": 8.6778,
+      "step": 3440
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.5641952983725134e-07,
+      "loss": 8.5937,
+      "step": 3460
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.5732368896925858e-07,
+      "loss": 8.7552,
+      "step": 3480
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.5822784810126582e-07,
+      "loss": 8.7656,
+      "step": 3500
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.5913200723327306e-07,
+      "loss": 8.7108,
+      "step": 3520
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.6003616636528027e-07,
+      "loss": 8.8032,
+      "step": 3540
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.609403254972875e-07,
+      "loss": 8.6476,
+      "step": 3560
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.6184448462929475e-07,
+      "loss": 8.6317,
+      "step": 3580
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.62748643761302e-07,
+      "loss": 8.6736,
+      "step": 3600
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.636528028933092e-07,
+      "loss": 8.6469,
+      "step": 3620
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.6455696202531644e-07,
+      "loss": 8.6429,
+      "step": 3640
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.6546112115732368e-07,
+      "loss": 8.6454,
+      "step": 3660
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.6636528028933092e-07,
+      "loss": 8.6146,
+      "step": 3680
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.6726943942133813e-07,
+      "loss": 8.6884,
+      "step": 3700
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.6817359855334537e-07,
+      "loss": 8.586,
+      "step": 3720
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.6907775768535261e-07,
+      "loss": 8.6912,
+      "step": 3740
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.6998191681735985e-07,
+      "loss": 8.6692,
+      "step": 3760
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.7088607594936707e-07,
+      "loss": 8.7503,
+      "step": 3780
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.717902350813743e-07,
+      "loss": 8.8529,
+      "step": 3800
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.7269439421338155e-07,
+      "loss": 8.6796,
+      "step": 3820
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.7359855334538879e-07,
+      "loss": 8.651,
+      "step": 3840
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.74502712477396e-07,
+      "loss": 8.6298,
+      "step": 3860
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.7540687160940324e-07,
+      "loss": 8.7301,
+      "step": 3880
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.7631103074141048e-07,
+      "loss": 8.6036,
+      "step": 3900
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.7721518987341772e-07,
+      "loss": 8.674,
+      "step": 3920
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.7811934900542493e-07,
+      "loss": 8.7339,
+      "step": 3940
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.7902350813743217e-07,
+      "loss": 8.7509,
+      "step": 3960
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.799276672694394e-07,
+      "loss": 8.7087,
+      "step": 3980
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8083182640144665e-07,
+      "loss": 8.6391,
+      "step": 4000
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8173598553345386e-07,
+      "loss": 8.7756,
+      "step": 4020
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.8264014466546113e-07,
+      "loss": 8.6629,
+      "step": 4040
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.8354430379746834e-07,
+      "loss": 8.8972,
+      "step": 4060
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.8444846292947558e-07,
+      "loss": 8.6525,
+      "step": 4080
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.853526220614828e-07,
+      "loss": 8.6576,
+      "step": 4100
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.8625678119349006e-07,
+      "loss": 8.6853,
+      "step": 4120
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.8716094032549728e-07,
+      "loss": 8.5977,
+      "step": 4140
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.8806509945750452e-07,
+      "loss": 8.6767,
+      "step": 4160
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 1.8896925858951173e-07,
+      "loss": 8.5974,
+      "step": 4180
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.89873417721519e-07,
+      "loss": 8.6597,
+      "step": 4200
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.907775768535262e-07,
+      "loss": 8.6174,
+      "step": 4220
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 1.9168173598553345e-07,
+      "loss": 8.657,
+      "step": 4240
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 1.9258589511754066e-07,
+      "loss": 8.6542,
+      "step": 4260
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.9349005424954793e-07,
+      "loss": 8.704,
+      "step": 4280
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.9439421338155514e-07,
+      "loss": 8.6938,
+      "step": 4300
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 1.9529837251356238e-07,
+      "loss": 8.592,
+      "step": 4320
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 1.962025316455696e-07,
+      "loss": 8.7386,
+      "step": 4340
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.9710669077757686e-07,
+      "loss": 8.5609,
+      "step": 4360
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.9801084990958407e-07,
+      "loss": 8.7259,
+      "step": 4380
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 1.9891500904159131e-07,
+      "loss": 8.642,
+      "step": 4400
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 1.9981916817359853e-07,
+      "loss": 8.711,
+      "step": 4420
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.007233273056058e-07,
+      "loss": 8.5766,
+      "step": 4440
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.01627486437613e-07,
+      "loss": 8.8441,
+      "step": 4460
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.0253164556962025e-07,
+      "loss": 8.6968,
+      "step": 4480
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.0343580470162746e-07,
+      "loss": 8.6558,
+      "step": 4500
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.0433996383363473e-07,
+      "loss": 8.708,
+      "step": 4520
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.0524412296564194e-07,
+      "loss": 8.6468,
+      "step": 4540
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.0614828209764918e-07,
+      "loss": 8.76,
+      "step": 4560
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.070524412296564e-07,
+      "loss": 8.6234,
+      "step": 4580
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.0795660036166366e-07,
+      "loss": 8.6519,
+      "step": 4600
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.0886075949367087e-07,
+      "loss": 8.8524,
+      "step": 4620
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.097649186256781e-07,
+      "loss": 8.6722,
+      "step": 4640
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.1066907775768532e-07,
+      "loss": 8.7367,
+      "step": 4660
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.115732368896926e-07,
+      "loss": 8.6933,
+      "step": 4680
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.124773960216998e-07,
+      "loss": 8.5989,
+      "step": 4700
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.1338155515370704e-07,
+      "loss": 8.7641,
+      "step": 4720
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.1428571428571426e-07,
+      "loss": 8.6898,
+      "step": 4740
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.1518987341772152e-07,
+      "loss": 8.7893,
+      "step": 4760
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.1609403254972874e-07,
+      "loss": 8.6064,
+      "step": 4780
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.1699819168173598e-07,
+      "loss": 8.6028,
+      "step": 4800
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.179023508137432e-07,
+      "loss": 8.7767,
+      "step": 4820
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.1880650994575046e-07,
+      "loss": 8.7362,
+      "step": 4840
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.1971066907775767e-07,
+      "loss": 8.6766,
+      "step": 4860
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.206148282097649e-07,
+      "loss": 8.6547,
+      "step": 4880
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.2151898734177212e-07,
+      "loss": 8.6397,
+      "step": 4900
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.224231464737794e-07,
+      "loss": 8.7055,
+      "step": 4920
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.233273056057866e-07,
+      "loss": 8.5466,
+      "step": 4940
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2423146473779384e-07,
+      "loss": 8.6754,
+      "step": 4960
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.2513562386980105e-07,
+      "loss": 8.6826,
+      "step": 4980
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.2603978300180832e-07,
+      "loss": 8.6895,
+      "step": 5000
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.2694394213381553e-07,
+      "loss": 8.7094,
+      "step": 5020
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.2784810126582277e-07,
+      "loss": 8.7513,
+      "step": 5040
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.2875226039783001e-07,
+      "loss": 8.6838,
+      "step": 5060
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.2965641952983725e-07,
+      "loss": 8.6037,
+      "step": 5080
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.3056057866184447e-07,
+      "loss": 8.7219,
+      "step": 5100
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.314647377938517e-07,
+      "loss": 8.6567,
+      "step": 5120
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.3236889692585895e-07,
+      "loss": 8.6228,
+      "step": 5140
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.3327305605786619e-07,
+      "loss": 8.6443,
+      "step": 5160
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2.341772151898734e-07,
+      "loss": 8.6465,
+      "step": 5180
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2.3508137432188064e-07,
+      "loss": 8.6116,
+      "step": 5200
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2.3598553345388788e-07,
+      "loss": 8.7553,
+      "step": 5220
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.3688969258589512e-07,
+      "loss": 8.7313,
+      "step": 5240
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2.3779385171790233e-07,
+      "loss": 8.6952,
+      "step": 5260
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2.3869801084990954e-07,
+      "loss": 8.6394,
+      "step": 5280
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.3960216998191684e-07,
+      "loss": 8.8396,
+      "step": 5300
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2.40506329113924e-07,
+      "loss": 8.6893,
+      "step": 5320
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2.4141048824593126e-07,
+      "loss": 8.5754,
+      "step": 5340
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2.423146473779385e-07,
+      "loss": 8.7712,
+      "step": 5360
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.4321880650994574e-07,
+      "loss": 8.8228,
+      "step": 5380
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2.44122965641953e-07,
+      "loss": 8.6878,
+      "step": 5400
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2.450271247739602e-07,
+      "loss": 8.6763,
+      "step": 5420
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.459312839059674e-07,
+      "loss": 8.6758,
+      "step": 5440
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2.468354430379747e-07,
+      "loss": 8.7743,
+      "step": 5460
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2.477396021699819e-07,
+      "loss": 8.6171,
+      "step": 5480
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2.4864376130198913e-07,
+      "loss": 8.8626,
+      "step": 5500
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.4954792043399637e-07,
+      "loss": 8.7344,
+      "step": 5520
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2.504520795660036e-07,
+      "loss": 8.7143,
+      "step": 5540
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2.513562386980108e-07,
+      "loss": 8.6419,
+      "step": 5560
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2.522603978300181e-07,
+      "loss": 8.6991,
+      "step": 5580
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.5316455696202533e-07,
+      "loss": 8.8916,
+      "step": 5600
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2.540687160940325e-07,
+      "loss": 8.6356,
+      "step": 5620
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2.5497287522603975e-07,
+      "loss": 8.7768,
+      "step": 5640
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.5587703435804705e-07,
+      "loss": 8.6408,
+      "step": 5660
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2.5678119349005423e-07,
+      "loss": 8.7009,
+      "step": 5680
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2.576853526220615e-07,
+      "loss": 8.6798,
+      "step": 5700
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2.5858951175406866e-07,
+      "loss": 8.7363,
+      "step": 5720
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.5949367088607595e-07,
+      "loss": 8.7773,
+      "step": 5740
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2.603978300180832e-07,
+      "loss": 8.5859,
+      "step": 5760
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2.613019891500904e-07,
+      "loss": 8.7,
+      "step": 5780
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.622061482820976e-07,
+      "loss": 8.6803,
+      "step": 5800
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2.631103074141049e-07,
+      "loss": 8.7215,
+      "step": 5820
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2.640144665461121e-07,
+      "loss": 8.6357,
+      "step": 5840
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.6491862567811934e-07,
+      "loss": 8.6427,
+      "step": 5860
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.658227848101265e-07,
+      "loss": 8.7049,
+      "step": 5880
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2.667269439421338e-07,
+      "loss": 8.6825,
+      "step": 5900
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.6763110307414106e-07,
+      "loss": 8.5983,
+      "step": 5920
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2.6853526220614824e-07,
+      "loss": 8.7502,
+      "step": 5940
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.694394213381555e-07,
+      "loss": 8.6624,
+      "step": 5960
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.703435804701628e-07,
+      "loss": 8.586,
+      "step": 5980
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2.7124773960216996e-07,
+      "loss": 8.6519,
+      "step": 6000
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.721518987341772e-07,
+      "loss": 8.6292,
+      "step": 6020
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.7305605786618444e-07,
+      "loss": 8.6325,
+      "step": 6040
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2.739602169981917e-07,
+      "loss": 8.6599,
+      "step": 6060
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.748643761301989e-07,
+      "loss": 8.6505,
+      "step": 6080
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.757685352622061e-07,
+      "loss": 8.6876,
+      "step": 6100
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.7667269439421335e-07,
+      "loss": 8.6015,
+      "step": 6120
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2.7757685352622064e-07,
+      "loss": 8.5816,
+      "step": 6140
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.7848101265822783e-07,
+      "loss": 8.6318,
+      "step": 6160
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.7938517179023507e-07,
+      "loss": 8.6486,
+      "step": 6180
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2.802893309222423e-07,
+      "loss": 8.6428,
+      "step": 6200
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.8119349005424955e-07,
+      "loss": 8.6547,
+      "step": 6220
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.820976491862568e-07,
+      "loss": 8.6352,
+      "step": 6240
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2.83001808318264e-07,
+      "loss": 8.6556,
+      "step": 6260
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.839059674502712e-07,
+      "loss": 8.5903,
+      "step": 6280
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2.848101265822785e-07,
+      "loss": 8.645,
+      "step": 6300
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.857142857142857e-07,
+      "loss": 8.6496,
+      "step": 6320
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.8661844484629293e-07,
+      "loss": 8.678,
+      "step": 6340
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.8752260397830017e-07,
+      "loss": 8.7506,
+      "step": 6360
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.884267631103074e-07,
+      "loss": 8.7135,
+      "step": 6380
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2.8933092224231465e-07,
+      "loss": 8.6221,
+      "step": 6400
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.9023508137432184e-07,
+      "loss": 8.646,
+      "step": 6420
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2.911392405063291e-07,
+      "loss": 8.5481,
+      "step": 6440
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.9204339963833637e-07,
+      "loss": 8.7291,
+      "step": 6460
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.9294755877034356e-07,
+      "loss": 8.6583,
+      "step": 6480
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2.938517179023508e-07,
+      "loss": 8.5624,
+      "step": 6500
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.9475587703435804e-07,
+      "loss": 8.648,
+      "step": 6520
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.956600361663653e-07,
+      "loss": 8.6145,
+      "step": 6540
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2.965641952983725e-07,
+      "loss": 8.6264,
+      "step": 6560
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.974683544303797e-07,
+      "loss": 8.6417,
+      "step": 6580
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.9837251356238694e-07,
+      "loss": 8.6012,
+      "step": 6600
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.9927667269439424e-07,
+      "loss": 8.6653,
+      "step": 6620
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.001808318264014e-07,
+      "loss": 8.7473,
+      "step": 6640
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.0108499095840866e-07,
+      "loss": 8.6582,
+      "step": 6660
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.019891500904159e-07,
+      "loss": 8.679,
+      "step": 6680
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.0289330922242314e-07,
+      "loss": 8.6476,
+      "step": 6700
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.037974683544304e-07,
+      "loss": 8.6075,
+      "step": 6720
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.0470162748643757e-07,
+      "loss": 8.6942,
+      "step": 6740
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.056057866184448e-07,
+      "loss": 8.6298,
+      "step": 6760
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.065099457504521e-07,
+      "loss": 8.6439,
+      "step": 6780
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.074141048824593e-07,
+      "loss": 8.6465,
+      "step": 6800
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.0831826401446653e-07,
+      "loss": 8.6938,
+      "step": 6820
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.0922242314647377e-07,
+      "loss": 8.6629,
+      "step": 6840
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.10126582278481e-07,
+      "loss": 8.6653,
+      "step": 6860
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.1103074141048825e-07,
+      "loss": 8.6678,
+      "step": 6880
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.1193490054249543e-07,
+      "loss": 8.6495,
+      "step": 6900
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.128390596745027e-07,
+      "loss": 8.6629,
+      "step": 6920
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.1374321880650997e-07,
+      "loss": 8.6308,
+      "step": 6940
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1464737793851715e-07,
+      "loss": 8.723,
+      "step": 6960
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.155515370705244e-07,
+      "loss": 8.7843,
+      "step": 6980
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1645569620253163e-07,
+      "loss": 8.7357,
+      "step": 7000
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.1735985533453887e-07,
+      "loss": 8.6289,
+      "step": 7020
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.182640144665461e-07,
+      "loss": 8.722,
+      "step": 7040
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.1916817359855335e-07,
+      "loss": 8.6224,
+      "step": 7060
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.2007233273056054e-07,
+      "loss": 8.6937,
+      "step": 7080
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 3.2097649186256783e-07,
+      "loss": 8.6861,
+      "step": 7100
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.21880650994575e-07,
+      "loss": 8.771,
+      "step": 7120
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.2278481012658226e-07,
+      "loss": 8.6539,
+      "step": 7140
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 3.236889692585895e-07,
+      "loss": 8.6444,
+      "step": 7160
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.2459312839059674e-07,
+      "loss": 8.7097,
+      "step": 7180
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.25497287522604e-07,
+      "loss": 8.6235,
+      "step": 7200
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 3.264014466546112e-07,
+      "loss": 8.6301,
+      "step": 7220
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.273056057866184e-07,
+      "loss": 8.6158,
+      "step": 7240
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.282097649186257e-07,
+      "loss": 8.7017,
+      "step": 7260
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.291139240506329e-07,
+      "loss": 8.5819,
+      "step": 7280
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 3.300180831826401e-07,
+      "loss": 8.6706,
+      "step": 7300
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.3092224231464736e-07,
+      "loss": 8.7503,
+      "step": 7320
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.318264014466546e-07,
+      "loss": 8.651,
+      "step": 7340
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.3273056057866184e-07,
+      "loss": 8.6765,
+      "step": 7360
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 3.336347197106691e-07,
+      "loss": 8.6551,
+      "step": 7380
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.3453887884267627e-07,
+      "loss": 8.5522,
+      "step": 7400
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.3544303797468356e-07,
+      "loss": 8.6139,
+      "step": 7420
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 3.3634719710669075e-07,
+      "loss": 8.6521,
+      "step": 7440
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 3.37251356238698e-07,
+      "loss": 8.7296,
+      "step": 7460
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.3815551537070523e-07,
+      "loss": 8.8073,
+      "step": 7480
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.3905967450271247e-07,
+      "loss": 8.8213,
+      "step": 7500
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 3.399638336347197e-07,
+      "loss": 8.792,
+      "step": 7520
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.4086799276672695e-07,
+      "loss": 8.8606,
+      "step": 7540
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.4177215189873413e-07,
+      "loss": 8.767,
+      "step": 7560
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.4267631103074143e-07,
+      "loss": 8.6718,
+      "step": 7580
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.435804701627486e-07,
+      "loss": 8.7104,
+      "step": 7600
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.4448462929475585e-07,
+      "loss": 8.8865,
+      "step": 7620
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.453887884267631e-07,
+      "loss": 8.7782,
+      "step": 7640
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.4629294755877033e-07,
+      "loss": 8.7133,
+      "step": 7660
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.4719710669077757e-07,
+      "loss": 8.6618,
+      "step": 7680
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.481012658227848e-07,
+      "loss": 8.6329,
+      "step": 7700
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.49005424954792e-07,
+      "loss": 8.6203,
+      "step": 7720
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.499095840867993e-07,
+      "loss": 8.8011,
+      "step": 7740
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.508137432188065e-07,
+      "loss": 8.6341,
+      "step": 7760
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.517179023508137e-07,
+      "loss": 8.7076,
+      "step": 7780
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.5262206148282096e-07,
+      "loss": 8.8637,
+      "step": 7800
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.535262206148282e-07,
+      "loss": 8.8632,
+      "step": 7820
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.5443037974683544e-07,
+      "loss": 8.7457,
+      "step": 7840
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.553345388788427e-07,
+      "loss": 8.8141,
+      "step": 7860
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.5623869801084986e-07,
+      "loss": 8.6005,
+      "step": 7880
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.5714285714285716e-07,
+      "loss": 8.8337,
+      "step": 7900
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.5804701627486434e-07,
+      "loss": 8.5335,
+      "step": 7920
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.589511754068716e-07,
+      "loss": 8.7152,
+      "step": 7940
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.598553345388788e-07,
+      "loss": 8.8067,
+      "step": 7960
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.6075949367088606e-07,
+      "loss": 8.6357,
+      "step": 7980
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.616636528028933e-07,
+      "loss": 8.6081,
+      "step": 8000
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.6256781193490054e-07,
+      "loss": 8.7371,
+      "step": 8020
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.6347197106690773e-07,
+      "loss": 8.6043,
+      "step": 8040
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.64376130198915e-07,
+      "loss": 8.7781,
+      "step": 8060
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.6528028933092226e-07,
+      "loss": 8.7021,
+      "step": 8080
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.6618444846292945e-07,
+      "loss": 8.6577,
+      "step": 8100
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.670886075949367e-07,
+      "loss": 8.7209,
+      "step": 8120
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.6799276672694393e-07,
+      "loss": 8.7849,
+      "step": 8140
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.6889692585895117e-07,
+      "loss": 8.6802,
+      "step": 8160
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.698010849909584e-07,
+      "loss": 8.6445,
+      "step": 8180
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.707052441229656e-07,
+      "loss": 8.8102,
+      "step": 8200
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.716094032549729e-07,
+      "loss": 8.6827,
+      "step": 8220
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.7251356238698013e-07,
+      "loss": 8.5796,
+      "step": 8240
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.734177215189873e-07,
+      "loss": 8.6077,
+      "step": 8260
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.7432188065099455e-07,
+      "loss": 8.6815,
+      "step": 8280
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.752260397830018e-07,
+      "loss": 8.6782,
+      "step": 8300
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.7613019891500903e-07,
+      "loss": 8.6266,
+      "step": 8320
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.7703435804701627e-07,
+      "loss": 8.7515,
+      "step": 8340
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.7793851717902346e-07,
+      "loss": 8.7789,
+      "step": 8360
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 3.7884267631103075e-07,
+      "loss": 8.6853,
+      "step": 8380
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 3.79746835443038e-07,
+      "loss": 8.6932,
+      "step": 8400
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 3.806509945750452e-07,
+      "loss": 8.6517,
+      "step": 8420
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 3.815551537070524e-07,
+      "loss": 8.7313,
+      "step": 8440
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 3.8245931283905966e-07,
+      "loss": 8.6474,
+      "step": 8460
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 3.833634719710669e-07,
+      "loss": 8.5971,
+      "step": 8480
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 3.8426763110307414e-07,
+      "loss": 8.6707,
+      "step": 8500
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 3.851717902350813e-07,
+      "loss": 8.6945,
+      "step": 8520
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 3.860759493670886e-07,
+      "loss": 8.5875,
+      "step": 8540
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 3.8698010849909586e-07,
+      "loss": 8.6613,
+      "step": 8560
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 3.8788426763110304e-07,
+      "loss": 8.6394,
+      "step": 8580
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 3.887884267631103e-07,
+      "loss": 8.6201,
+      "step": 8600
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 3.896925858951175e-07,
+      "loss": 8.7747,
+      "step": 8620
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 3.9059674502712476e-07,
+      "loss": 8.6703,
+      "step": 8640
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 3.91500904159132e-07,
+      "loss": 8.6593,
+      "step": 8660
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 3.924050632911392e-07,
+      "loss": 8.8128,
+      "step": 8680
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 3.933092224231465e-07,
+      "loss": 8.625,
+      "step": 8700
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 3.942133815551537e-07,
+      "loss": 8.7078,
+      "step": 8720
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 3.951175406871609e-07,
+      "loss": 8.7598,
+      "step": 8740
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 3.9602169981916815e-07,
+      "loss": 8.6145,
+      "step": 8760
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 3.969258589511754e-07,
+      "loss": 8.5881,
+      "step": 8780
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 3.9783001808318263e-07,
+      "loss": 8.6745,
+      "step": 8800
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 3.9873417721518987e-07,
+      "loss": 8.7567,
+      "step": 8820
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 3.9963833634719705e-07,
+      "loss": 8.6587,
+      "step": 8840
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.0054249547920435e-07,
+      "loss": 8.6343,
+      "step": 8860
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.014466546112116e-07,
+      "loss": 8.6116,
+      "step": 8880
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.0235081374321877e-07,
+      "loss": 8.6474,
+      "step": 8900
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.03254972875226e-07,
+      "loss": 8.5796,
+      "step": 8920
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.0415913200723325e-07,
+      "loss": 8.8124,
+      "step": 8940
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.050632911392405e-07,
+      "loss": 8.7305,
+      "step": 8960
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.0596745027124773e-07,
+      "loss": 8.6832,
+      "step": 8980
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.068716094032549e-07,
+      "loss": 8.6668,
+      "step": 9000
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.077757685352622e-07,
+      "loss": 8.611,
+      "step": 9020
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.0867992766726945e-07,
+      "loss": 8.6182,
+      "step": 9040
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.0958408679927664e-07,
+      "loss": 8.6509,
+      "step": 9060
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.104882459312839e-07,
+      "loss": 8.7003,
+      "step": 9080
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.1139240506329117e-07,
+      "loss": 8.7466,
+      "step": 9100
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.1229656419529836e-07,
+      "loss": 8.6764,
+      "step": 9120
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.132007233273056e-07,
+      "loss": 8.8049,
+      "step": 9140
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.141048824593128e-07,
+      "loss": 8.6367,
+      "step": 9160
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.150090415913201e-07,
+      "loss": 8.6451,
+      "step": 9180
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.159132007233273e-07,
+      "loss": 8.7919,
+      "step": 9200
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.168173598553345e-07,
+      "loss": 8.6583,
+      "step": 9220
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.1772151898734174e-07,
+      "loss": 8.6932,
+      "step": 9240
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.1862567811934904e-07,
+      "loss": 8.7003,
+      "step": 9260
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.195298372513562e-07,
+      "loss": 8.7071,
+      "step": 9280
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.2043399638336346e-07,
+      "loss": 8.5826,
+      "step": 9300
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.2133815551537065e-07,
+      "loss": 8.8741,
+      "step": 9320
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.2224231464737794e-07,
+      "loss": 8.7105,
+      "step": 9340
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.231464737793852e-07,
+      "loss": 8.7351,
+      "step": 9360
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.2405063291139237e-07,
+      "loss": 8.6307,
+      "step": 9380
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.249547920433996e-07,
+      "loss": 8.6908,
+      "step": 9400
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.258589511754069e-07,
+      "loss": 8.7118,
+      "step": 9420
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.267631103074141e-07,
+      "loss": 8.7018,
+      "step": 9440
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.2766726943942133e-07,
+      "loss": 8.7443,
+      "step": 9460
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.285714285714285e-07,
+      "loss": 8.6614,
+      "step": 9480
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.294755877034358e-07,
+      "loss": 8.7883,
+      "step": 9500
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.3037974683544305e-07,
+      "loss": 8.7251,
+      "step": 9520
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.3128390596745023e-07,
+      "loss": 8.8043,
+      "step": 9540
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.3218806509945747e-07,
+      "loss": 8.6045,
+      "step": 9560
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.3309222423146477e-07,
+      "loss": 8.6793,
+      "step": 9580
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.3399638336347195e-07,
+      "loss": 8.7049,
+      "step": 9600
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.349005424954792e-07,
+      "loss": 8.6692,
+      "step": 9620
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.358047016274864e-07,
+      "loss": 8.79,
+      "step": 9640
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.3670886075949367e-07,
+      "loss": 8.7726,
+      "step": 9660
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.376130198915009e-07,
+      "loss": 8.6092,
+      "step": 9680
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.385171790235081e-07,
+      "loss": 8.7118,
+      "step": 9700
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.3942133815551534e-07,
+      "loss": 8.6013,
+      "step": 9720
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.4032549728752263e-07,
+      "loss": 8.762,
+      "step": 9740
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.412296564195298e-07,
+      "loss": 8.6593,
+      "step": 9760
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.4213381555153706e-07,
+      "loss": 8.6182,
+      "step": 9780
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.4303797468354424e-07,
+      "loss": 8.6179,
+      "step": 9800
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.4394213381555154e-07,
+      "loss": 8.7162,
+      "step": 9820
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.448462929475588e-07,
+      "loss": 8.6335,
+      "step": 9840
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.4575045207956596e-07,
+      "loss": 8.7033,
+      "step": 9860
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.466546112115732e-07,
+      "loss": 8.6223,
+      "step": 9880
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.475587703435805e-07,
+      "loss": 8.6641,
+      "step": 9900
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.484629294755877e-07,
+      "loss": 8.7288,
+      "step": 9920
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.493670886075949e-07,
+      "loss": 8.6236,
+      "step": 9940
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.502712477396021e-07,
+      "loss": 8.6478,
+      "step": 9960
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.511754068716094e-07,
+      "loss": 8.7282,
+      "step": 9980
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.5207956600361664e-07,
+      "loss": 8.6599,
+      "step": 10000
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.5298372513562383e-07,
+      "loss": 8.7713,
+      "step": 10020
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.5388788426763107e-07,
+      "loss": 8.6667,
+      "step": 10040
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.5479204339963836e-07,
+      "loss": 8.7424,
+      "step": 10060
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.5569620253164555e-07,
+      "loss": 8.6676,
+      "step": 10080
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.566003616636528e-07,
+      "loss": 8.7331,
+      "step": 10100
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.5750452079566003e-07,
+      "loss": 8.6343,
+      "step": 10120
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.5840867992766727e-07,
+      "loss": 8.6826,
+      "step": 10140
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.593128390596745e-07,
+      "loss": 8.6749,
+      "step": 10160
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.602169981916817e-07,
+      "loss": 8.6509,
+      "step": 10180
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.6112115732368893e-07,
+      "loss": 8.6455,
+      "step": 10200
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.620253164556962e-07,
+      "loss": 8.6079,
+      "step": 10220
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.629294755877034e-07,
+      "loss": 8.6061,
+      "step": 10240
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.6383363471971065e-07,
+      "loss": 8.6707,
+      "step": 10260
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.647377938517179e-07,
+      "loss": 8.6731,
+      "step": 10280
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.6564195298372513e-07,
+      "loss": 8.6561,
+      "step": 10300
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.6654611211573237e-07,
+      "loss": 8.8374,
+      "step": 10320
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.6745027124773956e-07,
+      "loss": 8.7873,
+      "step": 10340
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.683544303797468e-07,
+      "loss": 8.5839,
+      "step": 10360
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.692585895117541e-07,
+      "loss": 8.7062,
+      "step": 10380
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.701627486437613e-07,
+      "loss": 8.5917,
+      "step": 10400
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.710669077757685e-07,
+      "loss": 8.6824,
+      "step": 10420
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.7197106690777576e-07,
+      "loss": 8.624,
+      "step": 10440
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.72875226039783e-07,
+      "loss": 8.6687,
+      "step": 10460
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.7377938517179024e-07,
+      "loss": 8.8627,
+      "step": 10480
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.746835443037974e-07,
+      "loss": 8.6775,
+      "step": 10500
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.7558770343580466e-07,
+      "loss": 8.6905,
+      "step": 10520
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.7649186256781196e-07,
+      "loss": 8.6904,
+      "step": 10540
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.773960216998191e-07,
+      "loss": 8.6781,
+      "step": 10560
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.783001808318264e-07,
+      "loss": 8.6216,
+      "step": 10580
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.792043399638337e-07,
+      "loss": 8.6004,
+      "step": 10600
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.801084990958408e-07,
+      "loss": 8.6008,
+      "step": 10620
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.81012658227848e-07,
+      "loss": 8.5841,
+      "step": 10640
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.819168173598553e-07,
+      "loss": 8.6029,
+      "step": 10660
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.828209764918625e-07,
+      "loss": 8.7195,
+      "step": 10680
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.837251356238698e-07,
+      "loss": 8.7097,
+      "step": 10700
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.84629294755877e-07,
+      "loss": 8.7943,
+      "step": 10720
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.855334538878842e-07,
+      "loss": 8.6566,
+      "step": 10740
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.864376130198915e-07,
+      "loss": 8.6458,
+      "step": 10760
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.873417721518987e-07,
+      "loss": 8.8129,
+      "step": 10780
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.88245931283906e-07,
+      "loss": 8.6123,
+      "step": 10800
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.891500904159131e-07,
+      "loss": 8.7739,
+      "step": 10820
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.900542495479204e-07,
+      "loss": 8.5971,
+      "step": 10840
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.909584086799277e-07,
+      "loss": 8.6343,
+      "step": 10860
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.918625678119348e-07,
+      "loss": 8.6995,
+      "step": 10880
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.927667269439422e-07,
+      "loss": 8.7568,
+      "step": 10900
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.936708860759494e-07,
+      "loss": 8.6049,
+      "step": 10920
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.945750452079565e-07,
+      "loss": 8.6712,
+      "step": 10940
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.954792043399638e-07,
+      "loss": 8.6592,
+      "step": 10960
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.96383363471971e-07,
+      "loss": 8.6155,
+      "step": 10980
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.972875226039783e-07,
+      "loss": 8.5876,
+      "step": 11000
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.981916817359855e-07,
+      "loss": 8.7089,
+      "step": 11020
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.990958408679927e-07,
+      "loss": 8.67,
+      "step": 11040
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 5e-07,
+      "loss": 8.6221,
+      "step": 11060
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.999997478613401e-07,
+      "loss": 8.7503,
+      "step": 11080
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.999989914458693e-07,
+      "loss": 8.8729,
+      "step": 11100
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.99997730755113e-07,
+      "loss": 8.7594,
+      "step": 11120
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.999959657916146e-07,
+      "loss": 8.6993,
+      "step": 11140
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4.999936965589338e-07,
+      "loss": 8.8509,
+      "step": 11160
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4.999909230616482e-07,
+      "loss": 8.6098,
+      "step": 11180
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 4.99987645305352e-07,
+      "loss": 8.6471,
+      "step": 11200
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 4.999838632966571e-07,
+      "loss": 8.7941,
+      "step": 11220
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 4.999795770431919e-07,
+      "loss": 8.6125,
+      "step": 11240
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 4.999747865536025e-07,
+      "loss": 8.7692,
+      "step": 11260
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 4.999694918375516e-07,
+      "loss": 8.6221,
+      "step": 11280
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 4.999636929057195e-07,
+      "loss": 8.6307,
+      "step": 11300
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 4.99957389769803e-07,
+      "loss": 8.6476,
+      "step": 11320
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 4.999505824425163e-07,
+      "loss": 8.6193,
+      "step": 11340
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 4.999432709375907e-07,
+      "loss": 8.5791,
+      "step": 11360
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 4.999354552697741e-07,
+      "loss": 8.6525,
+      "step": 11380
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 4.999271354548316e-07,
+      "loss": 8.6326,
+      "step": 11400
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 4.999183115095452e-07,
+      "loss": 8.7265,
+      "step": 11420
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 4.999089834517138e-07,
+      "loss": 8.6149,
+      "step": 11440
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 4.998991513001532e-07,
+      "loss": 8.7124,
+      "step": 11460
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 4.998888150746957e-07,
+      "loss": 8.5983,
+      "step": 11480
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 4.998779747961905e-07,
+      "loss": 8.685,
+      "step": 11500
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 4.99866630486504e-07,
+      "loss": 8.5874,
+      "step": 11520
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 4.998547821685187e-07,
+      "loss": 8.6896,
+      "step": 11540
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 4.99842429866134e-07,
+      "loss": 8.6125,
+      "step": 11560
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 4.998295736042658e-07,
+      "loss": 8.6181,
+      "step": 11580
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 4.998162134088466e-07,
+      "loss": 8.6908,
+      "step": 11600
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 4.998023493068254e-07,
+      "loss": 8.611,
+      "step": 11620
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 4.997879813261676e-07,
+      "loss": 8.6569,
+      "step": 11640
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 4.99773109495855e-07,
+      "loss": 8.7846,
+      "step": 11660
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 4.997577338458857e-07,
+      "loss": 8.6122,
+      "step": 11680
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 4.997418544072741e-07,
+      "loss": 8.8447,
+      "step": 11700
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 4.997254712120507e-07,
+      "loss": 8.7572,
+      "step": 11720
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 4.997085842932621e-07,
+      "loss": 8.6279,
+      "step": 11740
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 4.996911936849713e-07,
+      "loss": 8.6707,
+      "step": 11760
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 4.996732994222569e-07,
+      "loss": 8.7163,
+      "step": 11780
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 4.996549015412135e-07,
+      "loss": 8.7067,
+      "step": 11800
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 4.996360000789519e-07,
+      "loss": 8.5964,
+      "step": 11820
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 4.996165950735983e-07,
+      "loss": 8.8121,
+      "step": 11840
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 4.995966865642945e-07,
+      "loss": 8.7009,
+      "step": 11860
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 4.995762745911985e-07,
+      "loss": 8.6199,
+      "step": 11880
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 4.995553591954832e-07,
+      "loss": 8.7139,
+      "step": 11900
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 4.995339404193373e-07,
+      "loss": 8.6098,
+      "step": 11920
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 4.99512018305965e-07,
+      "loss": 8.6987,
+      "step": 11940
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 4.994895928995854e-07,
+      "loss": 8.6441,
+      "step": 11960
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 4.99466664245433e-07,
+      "loss": 8.6625,
+      "step": 11980
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 4.994432323897575e-07,
+      "loss": 8.6652,
+      "step": 12000
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 4.994192973798235e-07,
+      "loss": 8.7255,
+      "step": 12020
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 4.993948592639104e-07,
+      "loss": 8.6808,
+      "step": 12040
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 4.993699180913127e-07,
+      "loss": 8.6249,
+      "step": 12060
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 4.993444739123394e-07,
+      "loss": 8.8293,
+      "step": 12080
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 4.993185267783141e-07,
+      "loss": 8.6877,
+      "step": 12100
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 4.992920767415752e-07,
+      "loss": 8.6937,
+      "step": 12120
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 4.992651238554753e-07,
+      "loss": 8.8042,
+      "step": 12140
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 4.992376681743811e-07,
+      "loss": 8.631,
+      "step": 12160
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 4.992097097536739e-07,
+      "loss": 8.6861,
+      "step": 12180
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 4.991812486497489e-07,
+      "loss": 8.754,
+      "step": 12200
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 4.991522849200152e-07,
+      "loss": 8.8765,
+      "step": 12220
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 4.991228186228956e-07,
+      "loss": 8.7654,
+      "step": 12240
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 4.990928498178273e-07,
+      "loss": 8.7668,
+      "step": 12260
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 4.990623785652603e-07,
+      "loss": 8.711,
+      "step": 12280
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 4.990314049266585e-07,
+      "loss": 8.6748,
+      "step": 12300
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 4.989999289644991e-07,
+      "loss": 8.7944,
+      "step": 12320
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 4.989679507422728e-07,
+      "loss": 8.699,
+      "step": 12340
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 4.989354703244829e-07,
+      "loss": 8.618,
+      "step": 12360
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 4.98902487776646e-07,
+      "loss": 8.7872,
+      "step": 12380
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 4.988690031652916e-07,
+      "loss": 8.6499,
+      "step": 12400
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 4.988350165579616e-07,
+      "loss": 8.641,
+      "step": 12420
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 4.98800528023211e-07,
+      "loss": 8.6022,
+      "step": 12440
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 4.987655376306068e-07,
+      "loss": 8.7124,
+      "step": 12460
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 4.987300454507285e-07,
+      "loss": 8.6464,
+      "step": 12480
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 4.986940515551675e-07,
+      "loss": 8.7243,
+      "step": 12500
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 4.986575560165277e-07,
+      "loss": 8.8642,
+      "step": 12520
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 4.986205589084244e-07,
+      "loss": 8.6188,
+      "step": 12540
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 4.985830603054849e-07,
+      "loss": 8.6786,
+      "step": 12560
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 4.985450602833479e-07,
+      "loss": 8.6291,
+      "step": 12580
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 4.985065589186638e-07,
+      "loss": 8.6776,
+      "step": 12600
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 4.984675562890938e-07,
+      "loss": 8.626,
+      "step": 12620
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 4.984280524733107e-07,
+      "loss": 8.7193,
+      "step": 12640
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 4.983880475509977e-07,
+      "loss": 8.6363,
+      "step": 12660
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 4.983475416028494e-07,
+      "loss": 8.6651,
+      "step": 12680
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 4.983065347105706e-07,
+      "loss": 8.6346,
+      "step": 12700
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 4.982650269568766e-07,
+      "loss": 8.685,
+      "step": 12720
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 4.982230184254932e-07,
+      "loss": 8.6415,
+      "step": 12740
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 4.981805092011564e-07,
+      "loss": 8.5435,
+      "step": 12760
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 4.981374993696115e-07,
+      "loss": 8.6561,
+      "step": 12780
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 4.980939890176143e-07,
+      "loss": 8.6248,
+      "step": 12800
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 4.980499782329299e-07,
+      "loss": 8.709,
+      "step": 12820
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 4.980054671043329e-07,
+      "loss": 8.619,
+      "step": 12840
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 4.979604557216069e-07,
+      "loss": 8.7461,
+      "step": 12860
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 4.979149441755452e-07,
+      "loss": 8.7711,
+      "step": 12880
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 4.978689325579491e-07,
+      "loss": 8.6337,
+      "step": 12900
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 4.978224209616292e-07,
+      "loss": 8.7815,
+      "step": 12920
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 4.977754094804047e-07,
+      "loss": 8.6441,
+      "step": 12940
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 4.977278982091027e-07,
+      "loss": 8.765,
+      "step": 12960
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 4.976798872435586e-07,
+      "loss": 8.5918,
+      "step": 12980
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 4.976313766806159e-07,
+      "loss": 8.5984,
+      "step": 13000
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 4.975823666181255e-07,
+      "loss": 8.7362,
+      "step": 13020
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 4.975328571549462e-07,
+      "loss": 8.6716,
+      "step": 13040
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 4.97482848390944e-07,
+      "loss": 8.7553,
+      "step": 13060
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 4.974323404269921e-07,
+      "loss": 8.6592,
+      "step": 13080
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 4.973813333649703e-07,
+      "loss": 8.7005,
+      "step": 13100
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 4.973298273077657e-07,
+      "loss": 8.6346,
+      "step": 13120
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 4.972778223592717e-07,
+      "loss": 8.6217,
+      "step": 13140
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 4.972253186243876e-07,
+      "loss": 8.6394,
+      "step": 13160
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 4.971723162090196e-07,
+      "loss": 8.8543,
+      "step": 13180
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 4.971188152200791e-07,
+      "loss": 8.6189,
+      "step": 13200
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 4.970648157654835e-07,
+      "loss": 8.6655,
+      "step": 13220
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 4.970103179541556e-07,
+      "loss": 8.7302,
+      "step": 13240
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 4.969553218960234e-07,
+      "loss": 8.629,
+      "step": 13260
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 4.9689982770202e-07,
+      "loss": 8.6618,
+      "step": 13280
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 4.968438354840833e-07,
+      "loss": 8.5832,
+      "step": 13300
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 4.967873453551557e-07,
+      "loss": 8.6059,
+      "step": 13320
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 4.967303574291839e-07,
+      "loss": 8.7163,
+      "step": 13340
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 4.966728718211188e-07,
+      "loss": 8.6663,
+      "step": 13360
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 4.966148886469152e-07,
+      "loss": 8.6538,
+      "step": 13380
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 4.965564080235315e-07,
+      "loss": 8.6983,
+      "step": 13400
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 4.964974300689295e-07,
+      "loss": 8.6548,
+      "step": 13420
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 4.964379549020741e-07,
+      "loss": 8.6463,
+      "step": 13440
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 4.963779826429333e-07,
+      "loss": 8.7331,
+      "step": 13460
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 4.963175134124775e-07,
+      "loss": 8.7252,
+      "step": 13480
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 4.962565473326802e-07,
+      "loss": 8.7028,
+      "step": 13500
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 4.961950845265162e-07,
+      "loss": 8.6745,
+      "step": 13520
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 4.961331251179628e-07,
+      "loss": 8.5407,
+      "step": 13540
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 4.960706692319991e-07,
+      "loss": 8.7639,
+      "step": 13560
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 4.960077169946052e-07,
+      "loss": 8.6626,
+      "step": 13580
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 4.959442685327627e-07,
+      "loss": 8.7467,
+      "step": 13600
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 4.958803239744542e-07,
+      "loss": 8.6506,
+      "step": 13620
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 4.958158834486628e-07,
+      "loss": 8.5932,
+      "step": 13640
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 4.95750947085372e-07,
+      "loss": 8.751,
+      "step": 13660
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 4.956855150155657e-07,
+      "loss": 8.6431,
+      "step": 13680
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 4.956195873712273e-07,
+      "loss": 8.7175,
+      "step": 13700
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 4.955531642853403e-07,
+      "loss": 8.7028,
+      "step": 13720
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 4.954862458918873e-07,
+      "loss": 8.7185,
+      "step": 13740
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 4.954188323258498e-07,
+      "loss": 8.6743,
+      "step": 13760
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 4.953509237232085e-07,
+      "loss": 8.63,
+      "step": 13780
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 4.952825202209426e-07,
+      "loss": 8.5902,
+      "step": 13800
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 4.952136219570291e-07,
+      "loss": 8.7626,
+      "step": 13820
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.951442290704437e-07,
+      "loss": 8.6377,
+      "step": 13840
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 4.950743417011591e-07,
+      "loss": 8.6489,
+      "step": 13860
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 4.950039599901459e-07,
+      "loss": 8.5806,
+      "step": 13880
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.949330840793717e-07,
+      "loss": 8.6237,
+      "step": 13900
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 4.94861714111801e-07,
+      "loss": 8.5744,
+      "step": 13920
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 4.947898502313948e-07,
+      "loss": 8.6261,
+      "step": 13940
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 4.947174925831103e-07,
+      "loss": 8.6471,
+      "step": 13960
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.946446413129011e-07,
+      "loss": 8.7867,
+      "step": 13980
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 4.945712965677158e-07,
+      "loss": 8.6515,
+      "step": 14000
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 4.944974584954988e-07,
+      "loss": 8.7224,
+      "step": 14020
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.944231272451899e-07,
+      "loss": 8.6068,
+      "step": 14040
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 4.94348302966723e-07,
+      "loss": 8.6677,
+      "step": 14060
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 4.94272985811027e-07,
+      "loss": 8.7058,
+      "step": 14080
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 4.941971759300248e-07,
+      "loss": 8.7201,
+      "step": 14100
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.941208734766332e-07,
+      "loss": 8.6761,
+      "step": 14120
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 4.940440786047627e-07,
+      "loss": 8.7667,
+      "step": 14140
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 4.939667914693168e-07,
+      "loss": 8.7497,
+      "step": 14160
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 4.938890122261922e-07,
+      "loss": 8.6883,
+      "step": 14180
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.93810741032278e-07,
+      "loss": 8.6371,
+      "step": 14200
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 4.937319780454559e-07,
+      "loss": 8.5895,
+      "step": 14220
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 4.936527234245994e-07,
+      "loss": 8.7159,
+      "step": 14240
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.935729773295737e-07,
+      "loss": 8.6975,
+      "step": 14260
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 4.934927399212354e-07,
+      "loss": 8.7762,
+      "step": 14280
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 4.934120113614321e-07,
+      "loss": 8.6464,
+      "step": 14300
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 4.933307918130022e-07,
+      "loss": 8.6431,
+      "step": 14320
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.932490814397744e-07,
+      "loss": 8.6835,
+      "step": 14340
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 4.931668804065674e-07,
+      "loss": 8.7269,
+      "step": 14360
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 4.930841888791897e-07,
+      "loss": 8.6963,
+      "step": 14380
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 4.93001007024439e-07,
+      "loss": 8.7272,
+      "step": 14400
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.929173350101024e-07,
+      "loss": 8.6172,
+      "step": 14420
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 4.928331730049555e-07,
+      "loss": 8.6379,
+      "step": 14440
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 4.927485211787622e-07,
+      "loss": 8.6049,
+      "step": 14460
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.926633797022744e-07,
+      "loss": 8.7153,
+      "step": 14480
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 4.925777487472317e-07,
+      "loss": 8.6468,
+      "step": 14500
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 4.924916284863614e-07,
+      "loss": 8.7272,
+      "step": 14520
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 4.924050190933772e-07,
+      "loss": 8.7109,
+      "step": 14540
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.923179207429798e-07,
+      "loss": 8.7212,
+      "step": 14560
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 4.922303336108562e-07,
+      "loss": 8.6957,
+      "step": 14580
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 4.92142257873679e-07,
+      "loss": 8.7531,
+      "step": 14600
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.920536937091067e-07,
+      "loss": 8.6634,
+      "step": 14620
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 4.919646412957829e-07,
+      "loss": 8.687,
+      "step": 14640
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 4.918751008133362e-07,
+      "loss": 8.7248,
+      "step": 14660
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 4.917850724423792e-07,
+      "loss": 8.6262,
+      "step": 14680
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.916945563645093e-07,
+      "loss": 8.7897,
+      "step": 14700
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 4.91603552762307e-07,
+      "loss": 8.6599,
+      "step": 14720
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 4.915120618193368e-07,
+      "loss": 8.6223,
+      "step": 14740
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 4.914200837201458e-07,
+      "loss": 8.807,
+      "step": 14760
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.913276186502639e-07,
+      "loss": 8.7852,
+      "step": 14780
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 4.912346667962032e-07,
+      "loss": 8.6163,
+      "step": 14800
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 4.911412283454578e-07,
+      "loss": 8.5919,
+      "step": 14820
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.910473034865032e-07,
+      "loss": 8.6501,
+      "step": 14840
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 4.909528924087963e-07,
+      "loss": 8.7825,
+      "step": 14860
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 4.908579953027743e-07,
+      "loss": 8.7476,
+      "step": 14880
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 4.907626123598551e-07,
+      "loss": 8.7136,
+      "step": 14900
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.906667437724366e-07,
+      "loss": 8.6728,
+      "step": 14920
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 4.905703897338963e-07,
+      "loss": 8.7006,
+      "step": 14940
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 4.904735504385906e-07,
+      "loss": 8.68,
+      "step": 14960
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.903762260818551e-07,
+      "loss": 8.5956,
+      "step": 14980
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 4.902784168600036e-07,
+      "loss": 8.5962,
+      "step": 15000
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 4.90180122970328e-07,
+      "loss": 8.7187,
+      "step": 15020
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 4.900813446110978e-07,
+      "loss": 8.6258,
+      "step": 15040
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.899820819815598e-07,
+      "loss": 8.6823,
+      "step": 15060
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 4.898823352819375e-07,
+      "loss": 8.7023,
+      "step": 15080
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 4.897821047134309e-07,
+      "loss": 8.7809,
+      "step": 15100
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 4.896813904782162e-07,
+      "loss": 8.6179,
+      "step": 15120
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.895801927794448e-07,
+      "loss": 8.7411,
+      "step": 15140
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 4.894785118212435e-07,
+      "loss": 8.7115,
+      "step": 15160
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 4.893763478087139e-07,
+      "loss": 8.6297,
+      "step": 15180
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.892737009479322e-07,
+      "loss": 8.5999,
+      "step": 15200
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 4.891705714459482e-07,
+      "loss": 8.7592,
+      "step": 15220
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 4.890669595107853e-07,
+      "loss": 8.5834,
+      "step": 15240
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 4.889628653514402e-07,
+      "loss": 8.6487,
+      "step": 15260
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.888582891778821e-07,
+      "loss": 8.6445,
+      "step": 15280
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 4.887532312010527e-07,
+      "loss": 8.697,
+      "step": 15300
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 4.886476916328654e-07,
+      "loss": 8.6094,
+      "step": 15320
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 4.885416706862048e-07,
+      "loss": 8.7348,
+      "step": 15340
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.88435168574927e-07,
+      "loss": 8.7422,
+      "step": 15360
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 4.883281855138585e-07,
+      "loss": 8.7982,
+      "step": 15380
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 4.882207217187954e-07,
+      "loss": 8.64,
+      "step": 15400
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.881127774065044e-07,
+      "loss": 8.74,
+      "step": 15420
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 4.880043527947205e-07,
+      "loss": 8.6966,
+      "step": 15440
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 4.878954481021483e-07,
+      "loss": 8.7114,
+      "step": 15460
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 4.877860635484606e-07,
+      "loss": 8.6878,
+      "step": 15480
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.876761993542975e-07,
+      "loss": 8.6922,
+      "step": 15500
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 4.875658557412676e-07,
+      "loss": 8.6065,
+      "step": 15520
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 4.874550329319457e-07,
+      "loss": 8.6935,
+      "step": 15540
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.873437311498736e-07,
+      "loss": 8.766,
+      "step": 15560
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 4.872319506195592e-07,
+      "loss": 8.6696,
+      "step": 15580
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 4.871196915664761e-07,
+      "loss": 8.6916,
+      "step": 15600
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 4.870069542170629e-07,
+      "loss": 8.6553,
+      "step": 15620
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.868937387987233e-07,
+      "loss": 8.698,
+      "step": 15640
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 4.867800455398251e-07,
+      "loss": 8.6451,
+      "step": 15660
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 4.866658746697001e-07,
+      "loss": 8.6472,
+      "step": 15680
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 4.865512264186433e-07,
+      "loss": 8.5794,
+      "step": 15700
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.864361010179128e-07,
+      "loss": 8.6878,
+      "step": 15720
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 4.863204986997294e-07,
+      "loss": 8.5777,
+      "step": 15740
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 4.862044196972751e-07,
+      "loss": 8.724,
+      "step": 15760
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.860878642446943e-07,
+      "loss": 8.7462,
+      "step": 15780
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 4.859708325770919e-07,
+      "loss": 8.6529,
+      "step": 15800
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 4.858533249305336e-07,
+      "loss": 8.628,
+      "step": 15820
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 4.857353415420452e-07,
+      "loss": 8.6557,
+      "step": 15840
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.856168826496122e-07,
+      "loss": 8.6378,
+      "step": 15860
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 4.854979484921789e-07,
+      "loss": 8.6128,
+      "step": 15880
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 4.853785393096487e-07,
+      "loss": 8.6342,
+      "step": 15900
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.852586553428828e-07,
+      "loss": 8.623,
+      "step": 15920
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 4.851382968337004e-07,
+      "loss": 8.6746,
+      "step": 15940
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 4.850174640248775e-07,
+      "loss": 8.7209,
+      "step": 15960
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 4.848961571601475e-07,
+      "loss": 8.676,
+      "step": 15980
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.847743764841993e-07,
+      "loss": 8.6759,
+      "step": 16000
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 4.84652122242678e-07,
+      "loss": 8.6827,
+      "step": 16020
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 4.845293946821836e-07,
+      "loss": 8.6991,
+      "step": 16040
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 4.844061940502711e-07,
+      "loss": 8.7061,
+      "step": 16060
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.842825205954495e-07,
+      "loss": 8.6462,
+      "step": 16080
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 4.84158374567182e-07,
+      "loss": 8.647,
+      "step": 16100
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 4.840337562158843e-07,
+      "loss": 8.8409,
+      "step": 16120
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.839086657929256e-07,
+      "loss": 8.6913,
+      "step": 16140
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 4.837831035506267e-07,
+      "loss": 8.7391,
+      "step": 16160
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 4.836570697422605e-07,
+      "loss": 8.7379,
+      "step": 16180
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 4.835305646220509e-07,
+      "loss": 8.644,
+      "step": 16200
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.834035884451725e-07,
+      "loss": 8.7277,
+      "step": 16220
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 4.832761414677502e-07,
+      "loss": 8.6412,
+      "step": 16240
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 4.831482239468585e-07,
+      "loss": 8.7398,
+      "step": 16260
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 4.83019836140521e-07,
+      "loss": 8.754,
+      "step": 16280
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.828909783077099e-07,
+      "loss": 8.6548,
+      "step": 16300
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 4.827616507083456e-07,
+      "loss": 8.6114,
+      "step": 16320
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 4.826318536032958e-07,
+      "loss": 8.6163,
+      "step": 16340
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.825015872543758e-07,
+      "loss": 8.7388,
+      "step": 16360
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 4.823708519243467e-07,
+      "loss": 8.7207,
+      "step": 16380
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 4.822396478769162e-07,
+      "loss": 8.7599,
+      "step": 16400
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 4.821079753767371e-07,
+      "loss": 8.7465,
+      "step": 16420
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.819758346894072e-07,
+      "loss": 8.6565,
+      "step": 16440
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 4.818432260814688e-07,
+      "loss": 8.6793,
+      "step": 16460
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 4.817101498204078e-07,
+      "loss": 8.7599,
+      "step": 16480
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.815766061746537e-07,
+      "loss": 8.7688,
+      "step": 16500
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 4.814425954135785e-07,
+      "loss": 8.6962,
+      "step": 16520
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 4.813081178074968e-07,
+      "loss": 8.7069,
+      "step": 16540
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 4.811731736276643e-07,
+      "loss": 8.6498,
+      "step": 16560
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.810377631462785e-07,
+      "loss": 8.6458,
+      "step": 16580
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 4.809018866364766e-07,
+      "loss": 8.6952,
+      "step": 16600
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 4.80765544372337e-07,
+      "loss": 8.6439,
+      "step": 16620
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 4.806287366288766e-07,
+      "loss": 8.5903,
+      "step": 16640
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 4.804914636820516e-07,
+      "loss": 8.7511,
+      "step": 16660
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 4.803537258087566e-07,
+      "loss": 8.6639,
+      "step": 16680
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 4.802155232868239e-07,
+      "loss": 8.9259,
+      "step": 16700
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 4.800768563950231e-07,
+      "loss": 8.662,
+      "step": 16720
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 4.799377254130606e-07,
+      "loss": 8.7333,
+      "step": 16740
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 4.797981306215784e-07,
+      "loss": 8.6613,
+      "step": 16760
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 4.79658072302155e-07,
+      "loss": 8.7945,
+      "step": 16780
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 4.795175507373028e-07,
+      "loss": 8.6238,
+      "step": 16800
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 4.793765662104696e-07,
+      "loss": 8.6998,
+      "step": 16820
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 4.792351190060363e-07,
+      "loss": 8.7753,
+      "step": 16840
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 4.790932094093175e-07,
+      "loss": 8.5856,
+      "step": 16860
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 4.789508377065603e-07,
+      "loss": 8.6769,
+      "step": 16880
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 4.788080041849441e-07,
+      "loss": 8.6106,
+      "step": 16900
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 4.786647091325796e-07,
+      "loss": 8.6677,
+      "step": 16920
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 4.785209528385087e-07,
+      "loss": 8.5975,
+      "step": 16940
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 4.783767355927033e-07,
+      "loss": 8.7969,
+      "step": 16960
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 4.782320576860656e-07,
+      "loss": 8.6307,
+      "step": 16980
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 4.780869194104268e-07,
+      "loss": 8.5835,
+      "step": 17000
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 4.779413210585464e-07,
+      "loss": 8.6421,
+      "step": 17020
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 4.777952629241122e-07,
+      "loss": 8.7197,
+      "step": 17040
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 4.776487453017397e-07,
+      "loss": 8.672,
+      "step": 17060
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 4.775017684869707e-07,
+      "loss": 8.6878,
+      "step": 17080
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 4.773543327762737e-07,
+      "loss": 8.6636,
+      "step": 17100
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 4.772064384670424e-07,
+      "loss": 8.5946,
+      "step": 17120
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 4.77058085857596e-07,
+      "loss": 8.7402,
+      "step": 17140
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 4.769092752471778e-07,
+      "loss": 8.5723,
+      "step": 17160
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 4.7676000693595506e-07,
+      "loss": 8.8408,
+      "step": 17180
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 4.766102812250183e-07,
+      "loss": 8.723,
+      "step": 17200
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 4.764600984163808e-07,
+      "loss": 8.6513,
+      "step": 17220
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 4.7630945881297746e-07,
+      "loss": 8.6348,
+      "step": 17240
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 4.761583627186649e-07,
+      "loss": 8.572,
+      "step": 17260
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 4.7600681043822044e-07,
+      "loss": 8.7423,
+      "step": 17280
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 4.7585480227734163e-07,
+      "loss": 8.6616,
+      "step": 17300
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 4.7570233854264564e-07,
+      "loss": 8.7175,
+      "step": 17320
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 4.7554941954166826e-07,
+      "loss": 8.7237,
+      "step": 17340
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 4.7539604558286395e-07,
+      "loss": 8.8209,
+      "step": 17360
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 4.752422169756047e-07,
+      "loss": 8.6999,
+      "step": 17380
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 4.7508793403017976e-07,
+      "loss": 8.6965,
+      "step": 17400
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 4.749331970577946e-07,
+      "loss": 8.6204,
+      "step": 17420
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 4.747780063705705e-07,
+      "loss": 8.6613,
+      "step": 17440
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 4.7462236228154405e-07,
+      "loss": 8.7295,
+      "step": 17460
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 4.744662651046666e-07,
+      "loss": 8.6806,
+      "step": 17480
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 4.7430971515480304e-07,
+      "loss": 8.6633,
+      "step": 17500
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 4.741527127477317e-07,
+      "loss": 8.676,
+      "step": 17520
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 4.7399525820014376e-07,
+      "loss": 8.6381,
+      "step": 17540
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 4.738373518296421e-07,
+      "loss": 8.7071,
+      "step": 17560
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 4.7367899395474106e-07,
+      "loss": 8.6584,
+      "step": 17580
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 4.7352018489486606e-07,
+      "loss": 8.7591,
+      "step": 17600
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 4.7336092497035207e-07,
+      "loss": 8.7067,
+      "step": 17620
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 4.732012145024439e-07,
+      "loss": 8.6802,
+      "step": 17640
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 4.7304105381329484e-07,
+      "loss": 8.9072,
+      "step": 17660
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 4.7288044322596663e-07,
+      "loss": 8.6963,
+      "step": 17680
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 4.727193830644285e-07,
+      "loss": 8.6309,
+      "step": 17700
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 4.725578736535562e-07,
+      "loss": 8.6925,
+      "step": 17720
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 4.723959153191319e-07,
+      "loss": 8.653,
+      "step": 17740
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 4.722335083878433e-07,
+      "loss": 8.612,
+      "step": 17760
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 4.7207065318728296e-07,
+      "loss": 8.7143,
+      "step": 17780
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 4.7190735004594753e-07,
+      "loss": 8.6625,
+      "step": 17800
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 4.7174359929323735e-07,
+      "loss": 8.6155,
+      "step": 17820
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 4.715794012594555e-07,
+      "loss": 8.7971,
+      "step": 17840
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 4.7141475627580754e-07,
+      "loss": 8.6821,
+      "step": 17860
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 4.712496646744002e-07,
+      "loss": 8.6292,
+      "step": 17880
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 4.7108412678824134e-07,
+      "loss": 8.6244,
+      "step": 17900
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 4.70918142951239e-07,
+      "loss": 8.5691,
+      "step": 17920
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 4.7075171349820077e-07,
+      "loss": 8.6863,
+      "step": 17940
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 4.705848387648329e-07,
+      "loss": 8.6821,
+      "step": 17960
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 4.7041751908774007e-07,
+      "loss": 8.6589,
+      "step": 17980
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 4.702497548044243e-07,
+      "loss": 8.7016,
+      "step": 18000
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 4.700815462532845e-07,
+      "loss": 8.725,
+      "step": 18020
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 4.699128937736157e-07,
+      "loss": 8.6708,
+      "step": 18040
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 4.697437977056084e-07,
+      "loss": 8.8955,
+      "step": 18060
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 4.695742583903478e-07,
+      "loss": 8.7241,
+      "step": 18080
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 4.694042761698134e-07,
+      "loss": 8.6222,
+      "step": 18100
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 4.692338513868776e-07,
+      "loss": 8.7879,
+      "step": 18120
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 4.6906298438530604e-07,
+      "loss": 8.6359,
+      "step": 18140
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 4.6889167550975613e-07,
+      "loss": 8.6056,
+      "step": 18160
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 4.6871992510577644e-07,
+      "loss": 8.6364,
+      "step": 18180
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 4.6854773351980647e-07,
+      "loss": 8.6699,
+      "step": 18200
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 4.683751010991754e-07,
+      "loss": 8.6423,
+      "step": 18220
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 4.682020281921017e-07,
+      "loss": 8.7572,
+      "step": 18240
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 4.6802851514769227e-07,
+      "loss": 8.7399,
+      "step": 18260
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 4.67854562315942e-07,
+      "loss": 8.6589,
+      "step": 18280
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 4.6768017004773263e-07,
+      "loss": 8.7953,
+      "step": 18300
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 4.6750533869483257e-07,
+      "loss": 8.5799,
+      "step": 18320
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 4.6733006860989567e-07,
+      "loss": 8.6455,
+      "step": 18340
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 4.6715436014646077e-07,
+      "loss": 8.6309,
+      "step": 18360
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 4.669782136589512e-07,
+      "loss": 8.6688,
+      "step": 18380
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 4.6680162950267356e-07,
+      "loss": 8.6939,
+      "step": 18400
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 4.666246080338175e-07,
+      "loss": 8.6871,
+      "step": 18420
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 4.6644714960945453e-07,
+      "loss": 8.7004,
+      "step": 18440
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 4.662692545875378e-07,
+      "loss": 8.6696,
+      "step": 18460
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 4.660909233269009e-07,
+      "loss": 8.6115,
+      "step": 18480
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 4.6591215618725766e-07,
+      "loss": 8.6684,
+      "step": 18500
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 4.657329535292007e-07,
+      "loss": 8.7201,
+      "step": 18520
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 4.6555331571420155e-07,
+      "loss": 8.6737,
+      "step": 18540
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 4.653732431046092e-07,
+      "loss": 8.709,
+      "step": 18560
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 4.6519273606364984e-07,
+      "loss": 8.5615,
+      "step": 18580
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 4.6501179495542585e-07,
+      "loss": 8.6053,
+      "step": 18600
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 4.6483042014491527e-07,
+      "loss": 8.6445,
+      "step": 18620
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 4.646486119979709e-07,
+      "loss": 8.7607,
+      "step": 18640
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 4.6446637088131956e-07,
+      "loss": 8.7539,
+      "step": 18660
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 4.642836971625616e-07,
+      "loss": 8.6076,
+      "step": 18680
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 4.6410059121016984e-07,
+      "loss": 8.7657,
+      "step": 18700
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 4.639170533934891e-07,
+      "loss": 8.5961,
+      "step": 18720
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 4.6373308408273495e-07,
+      "loss": 8.7081,
+      "step": 18740
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 4.635486836489938e-07,
+      "loss": 8.8006,
+      "step": 18760
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 4.633638524642215e-07,
+      "loss": 8.6963,
+      "step": 18780
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 4.631785909012426e-07,
+      "loss": 8.6344,
+      "step": 18800
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 4.6299289933375007e-07,
+      "loss": 8.7319,
+      "step": 18820
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 4.6280677813630397e-07,
+      "loss": 8.7458,
+      "step": 18840
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 4.626202276843311e-07,
+      "loss": 8.6465,
+      "step": 18860
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 4.624332483541242e-07,
+      "loss": 8.5764,
+      "step": 18880
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 4.62245840522841e-07,
+      "loss": 8.7452,
+      "step": 18900
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 4.6205800456850343e-07,
+      "loss": 8.6529,
+      "step": 18920
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 4.618697408699973e-07,
+      "loss": 8.6897,
+      "step": 18940
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 4.6168104980707103e-07,
+      "loss": 8.7269,
+      "step": 18960
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 4.6149193176033505e-07,
+      "loss": 8.6609,
+      "step": 18980
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 4.6130238711126123e-07,
+      "loss": 8.6924,
+      "step": 19000
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 4.6111241624218166e-07,
+      "loss": 8.7371,
+      "step": 19020
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 4.609220195362886e-07,
+      "loss": 8.6188,
+      "step": 19040
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 4.607311973776328e-07,
+      "loss": 8.7018,
+      "step": 19060
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 4.6053995015112343e-07,
+      "loss": 8.5875,
+      "step": 19080
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 4.6034827824252715e-07,
+      "loss": 8.7276,
+      "step": 19100
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 4.601561820384671e-07,
+      "loss": 8.5814,
+      "step": 19120
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 4.5996366192642226e-07,
+      "loss": 8.6103,
+      "step": 19140
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 4.597707182947268e-07,
+      "loss": 8.6523,
+      "step": 19160
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 4.595773515325691e-07,
+      "loss": 8.6586,
+      "step": 19180
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 4.593835620299911e-07,
+      "loss": 8.6554,
+      "step": 19200
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 4.5918935017788724e-07,
+      "loss": 8.6933,
+      "step": 19220
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 4.589947163680041e-07,
+      "loss": 8.6822,
+      "step": 19240
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 4.5879966099293955e-07,
+      "loss": 8.7641,
+      "step": 19260
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 4.5860418444614133e-07,
+      "loss": 8.6162,
+      "step": 19280
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 4.5840828712190717e-07,
+      "loss": 8.6388,
+      "step": 19300
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 4.5821196941538334e-07,
+      "loss": 8.8059,
+      "step": 19320
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 4.580152317225641e-07,
+      "loss": 8.6918,
+      "step": 19340
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 4.5781807444029075e-07,
+      "loss": 8.6036,
+      "step": 19360
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 4.5762049796625124e-07,
+      "loss": 8.697,
+      "step": 19380
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 4.5742250269897884e-07,
+      "loss": 8.65,
+      "step": 19400
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 4.572240890378517e-07,
+      "loss": 8.6848,
+      "step": 19420
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 4.570252573830918e-07,
+      "loss": 8.7286,
+      "step": 19440
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 4.568260081357643e-07,
+      "loss": 8.5952,
+      "step": 19460
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 4.5662634169777674e-07,
+      "loss": 8.6095,
+      "step": 19480
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 4.5642625847187813e-07,
+      "loss": 8.7558,
+      "step": 19500
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 4.5622575886165826e-07,
+      "loss": 8.6469,
+      "step": 19520
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 4.5602484327154666e-07,
+      "loss": 8.6547,
+      "step": 19540
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 4.55823512106812e-07,
+      "loss": 8.6466,
+      "step": 19560
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 4.5562176577356146e-07,
+      "loss": 8.7095,
+      "step": 19580
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 4.554196046787392e-07,
+      "loss": 8.7402,
+      "step": 19600
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 4.552170292301264e-07,
+      "loss": 8.6959,
+      "step": 19620
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 4.550140398363398e-07,
+      "loss": 8.6991,
+      "step": 19640
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 4.5481063690683116e-07,
+      "loss": 8.6932,
+      "step": 19660
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 4.546068208518865e-07,
+      "loss": 8.702,
+      "step": 19680
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 4.5440259208262497e-07,
+      "loss": 8.7574,
+      "step": 19700
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 4.5419795101099847e-07,
+      "loss": 8.7992,
+      "step": 19720
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 4.539928980497902e-07,
+      "loss": 8.7671,
+      "step": 19740
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 4.537874336126146e-07,
+      "loss": 8.7184,
+      "step": 19760
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 4.535815581139157e-07,
+      "loss": 8.6308,
+      "step": 19780
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 4.5337527196896715e-07,
+      "loss": 8.7107,
+      "step": 19800
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 4.5316857559387036e-07,
+      "loss": 8.6339,
+      "step": 19820
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 4.529614694055546e-07,
+      "loss": 8.676,
+      "step": 19840
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 4.527539538217757e-07,
+      "loss": 8.686,
+      "step": 19860
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 4.5254602926111533e-07,
+      "loss": 8.7237,
+      "step": 19880
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 4.5233769614298e-07,
+      "loss": 8.6244,
+      "step": 19900
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 4.521289548876003e-07,
+      "loss": 8.6432,
+      "step": 19920
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 4.519198059160303e-07,
+      "loss": 8.6141,
+      "step": 19940
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 4.517102496501462e-07,
+      "loss": 8.6889,
+      "step": 19960
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 4.5150028651264596e-07,
+      "loss": 8.5727,
+      "step": 19980
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 4.512899169270481e-07,
+      "loss": 8.6158,
+      "step": 20000
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 4.510791413176912e-07,
+      "loss": 8.7124,
+      "step": 20020
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 4.508679601097326e-07,
+      "loss": 8.6239,
+      "step": 20040
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 4.5065637372914784e-07,
+      "loss": 8.7528,
+      "step": 20060
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 4.504443826027298e-07,
+      "loss": 8.6065,
+      "step": 20080
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 4.5023198715808783e-07,
+      "loss": 8.7053,
+      "step": 20100
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 4.5001918782364665e-07,
+      "loss": 8.6622,
+      "step": 20120
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 4.498059850286459e-07,
+      "loss": 8.6676,
+      "step": 20140
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 4.4959237920313877e-07,
+      "loss": 8.6896,
+      "step": 20160
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 4.493783707779916e-07,
+      "loss": 8.7231,
+      "step": 20180
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 4.491639601848828e-07,
+      "loss": 8.6187,
+      "step": 20200
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 4.489491478563019e-07,
+      "loss": 8.7279,
+      "step": 20220
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 4.4873393422554894e-07,
+      "loss": 8.6275,
+      "step": 20240
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 4.4851831972673324e-07,
+      "loss": 8.6786,
+      "step": 20260
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 4.483023047947729e-07,
+      "loss": 8.679,
+      "step": 20280
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 4.4808588986539355e-07,
+      "loss": 8.6271,
+      "step": 20300
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 4.478690753751278e-07,
+      "loss": 8.6284,
+      "step": 20320
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 4.476518617613142e-07,
+      "loss": 8.7289,
+      "step": 20340
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 4.4743424946209627e-07,
+      "loss": 8.7282,
+      "step": 20360
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 4.4721623891642185e-07,
+      "loss": 8.709,
+      "step": 20380
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 4.46997830564042e-07,
+      "loss": 8.8007,
+      "step": 20400
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 4.4677902484551023e-07,
+      "loss": 8.7284,
+      "step": 20420
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 4.4655982220218167e-07,
+      "loss": 8.6282,
+      "step": 20440
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 4.463402230762119e-07,
+      "loss": 8.7566,
+      "step": 20460
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 4.461202279105565e-07,
+      "loss": 8.7809,
+      "step": 20480
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 4.458998371489695e-07,
+      "loss": 8.6693,
+      "step": 20500
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 4.4567905123600345e-07,
+      "loss": 8.6117,
+      "step": 20520
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 4.4545787061700746e-07,
+      "loss": 8.5389,
+      "step": 20540
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 4.4523629573812705e-07,
+      "loss": 8.6785,
+      "step": 20560
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 4.45014327046303e-07,
+      "loss": 8.6557,
+      "step": 20580
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 4.447919649892704e-07,
+      "loss": 8.7228,
+      "step": 20600
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 4.445692100155579e-07,
+      "loss": 8.661,
+      "step": 20620
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 4.443460625744865e-07,
+      "loss": 8.5727,
+      "step": 20640
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 4.44122523116169e-07,
+      "loss": 8.6239,
+      "step": 20660
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 4.438985920915089e-07,
+      "loss": 8.62,
+      "step": 20680
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 4.436742699521997e-07,
+      "loss": 8.7129,
+      "step": 20700
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 4.434495571507234e-07,
+      "loss": 8.6496,
+      "step": 20720
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 4.432244541403506e-07,
+      "loss": 8.7242,
+      "step": 20740
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 4.4299896137513837e-07,
+      "loss": 8.7956,
+      "step": 20760
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 4.4277307930993045e-07,
+      "loss": 8.7197,
+      "step": 20780
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 4.4254680840035554e-07,
+      "loss": 8.8465,
+      "step": 20800
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 4.423201491028269e-07,
+      "loss": 8.6558,
+      "step": 20820
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 4.42093101874541e-07,
+      "loss": 8.7996,
+      "step": 20840
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 4.4186566717347693e-07,
+      "loss": 8.8602,
+      "step": 20860
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 4.4163784545839543e-07,
+      "loss": 8.5926,
+      "step": 20880
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 4.414096371888377e-07,
+      "loss": 8.6053,
+      "step": 20900
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 4.411810428251248e-07,
+      "loss": 8.6233,
+      "step": 20920
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 4.409520628283565e-07,
+      "loss": 8.7897,
+      "step": 20940
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 4.407226976604105e-07,
+      "loss": 8.6926,
+      "step": 20960
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 4.404929477839414e-07,
+      "loss": 8.6766,
+      "step": 20980
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 4.402628136623798e-07,
+      "loss": 8.6537,
+      "step": 21000
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 4.400322957599314e-07,
+      "loss": 8.697,
+      "step": 21020
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 4.3980139454157607e-07,
+      "loss": 8.6,
+      "step": 21040
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 4.3957011047306656e-07,
+      "loss": 8.7509,
+      "step": 21060
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 4.393384440209284e-07,
+      "loss": 8.8166,
+      "step": 21080
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 4.39106395652458e-07,
+      "loss": 8.7277,
+      "step": 21100
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 4.3887396583572225e-07,
+      "loss": 8.6775,
+      "step": 21120
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 4.386411550395575e-07,
+      "loss": 8.637,
+      "step": 21140
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 4.3840796373356864e-07,
+      "loss": 8.6908,
+      "step": 21160
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 4.381743923881279e-07,
+      "loss": 8.6484,
+      "step": 21180
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 4.3794044147437437e-07,
+      "loss": 8.6396,
+      "step": 21200
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 4.377061114642125e-07,
+      "loss": 8.7759,
+      "step": 21220
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 4.3747140283031153e-07,
+      "loss": 8.6919,
+      "step": 21240
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 4.3723631604610447e-07,
+      "loss": 8.6463,
+      "step": 21260
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 4.3700085158578694e-07,
+      "loss": 8.8272,
+      "step": 21280
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 4.367650099243166e-07,
+      "loss": 8.6321,
+      "step": 21300
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 4.365287915374118e-07,
+      "loss": 8.6541,
+      "step": 21320
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 4.362921969015509e-07,
+      "loss": 8.6122,
+      "step": 21340
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 4.360552264939712e-07,
+      "loss": 8.6558,
+      "step": 21360
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 4.358178807926677e-07,
+      "loss": 8.7851,
+      "step": 21380
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 4.355801602763927e-07,
+      "loss": 8.7331,
+      "step": 21400
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 4.353420654246546e-07,
+      "loss": 8.6654,
+      "step": 21420
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 4.3510359671771647e-07,
+      "loss": 8.6424,
+      "step": 21440
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 4.3486475463659593e-07,
+      "loss": 8.6382,
+      "step": 21460
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 4.3462553966306357e-07,
+      "loss": 8.7952,
+      "step": 21480
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 4.3438595227964205e-07,
+      "loss": 8.6571,
+      "step": 21500
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 4.341459929696054e-07,
+      "loss": 8.615,
+      "step": 21520
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 4.3390566221697764e-07,
+      "loss": 8.6564,
+      "step": 21540
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 4.3366496050653235e-07,
+      "loss": 8.7334,
+      "step": 21560
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 4.3342388832379094e-07,
+      "loss": 8.5915,
+      "step": 21580
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 4.3318244615502254e-07,
+      "loss": 8.7452,
+      "step": 21600
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 4.329406344872423e-07,
+      "loss": 8.62,
+      "step": 21620
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 4.326984538082108e-07,
+      "loss": 8.6302,
+      "step": 21640
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 4.3245590460643293e-07,
+      "loss": 8.6475,
+      "step": 21660
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 4.32212987371157e-07,
+      "loss": 8.771,
+      "step": 21680
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 4.3196970259237355e-07,
+      "loss": 8.6497,
+      "step": 21700
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 4.3172605076081456e-07,
+      "loss": 8.8053,
+      "step": 21720
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 4.3148203236795234e-07,
+      "loss": 8.6689,
+      "step": 21740
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 4.312376479059988e-07,
+      "loss": 8.6764,
+      "step": 21760
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 4.3099289786790405e-07,
+      "loss": 8.6942,
+      "step": 21780
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 4.307477827473556e-07,
+      "loss": 8.7068,
+      "step": 21800
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 4.305023030387775e-07,
+      "loss": 8.6783,
+      "step": 21820
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 4.302564592373292e-07,
+      "loss": 8.7257,
+      "step": 21840
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 4.300102518389044e-07,
+      "loss": 8.7192,
+      "step": 21860
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 4.2976368134013033e-07,
+      "loss": 8.7052,
+      "step": 21880
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 4.295167482383667e-07,
+      "loss": 8.687,
+      "step": 21900
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 4.292694530317046e-07,
+      "loss": 8.6849,
+      "step": 21920
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 4.2902179621896534e-07,
+      "loss": 8.6997,
+      "step": 21940
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 4.2877377829969983e-07,
+      "loss": 8.6191,
+      "step": 21960
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 4.2852539977418745e-07,
+      "loss": 8.6416,
+      "step": 21980
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 4.2827666114343463e-07,
+      "loss": 8.6274,
+      "step": 22000
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 4.2802756290917446e-07,
+      "loss": 8.6461,
+      "step": 22020
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 4.2777810557386534e-07,
+      "loss": 8.6694,
+      "step": 22040
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 4.2752828964068996e-07,
+      "loss": 8.6447,
+      "step": 22060
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 4.2727811561355423e-07,
+      "loss": 8.7354,
+      "step": 22080
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 4.2702758399708674e-07,
+      "loss": 8.7295,
+      "step": 22100
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 4.2677669529663686e-07,
+      "loss": 8.6513,
+      "step": 22120
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 4.2652545001827474e-07,
+      "loss": 8.5954,
+      "step": 22140
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 4.262738486687895e-07,
+      "loss": 8.6644,
+      "step": 22160
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 4.2602189175568847e-07,
+      "loss": 8.6703,
+      "step": 22180
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 4.2576957978719636e-07,
+      "loss": 8.7068,
+      "step": 22200
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 4.255169132722539e-07,
+      "loss": 8.7908,
+      "step": 22220
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 4.252638927205172e-07,
+      "loss": 8.7085,
+      "step": 22240
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 4.250105186423563e-07,
+      "loss": 8.6468,
+      "step": 22260
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 4.2475679154885443e-07,
+      "loss": 8.6714,
+      "step": 22280
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 4.2450271195180675e-07,
+      "loss": 8.7756,
+      "step": 22300
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 4.242482803637197e-07,
+      "loss": 8.8086,
+      "step": 22320
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 4.2399349729780954e-07,
+      "loss": 8.7139,
+      "step": 22340
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 4.237383632680015e-07,
+      "loss": 8.7571,
+      "step": 22360
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 4.2348287878892896e-07,
+      "loss": 8.6348,
+      "step": 22380
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 4.232270443759319e-07,
+      "loss": 8.7034,
+      "step": 22400
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 4.2297086054505626e-07,
+      "loss": 8.6211,
+      "step": 22420
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 4.2271432781305293e-07,
+      "loss": 8.578,
+      "step": 22440
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 4.224574466973765e-07,
+      "loss": 8.6008,
+      "step": 22460
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 4.222002177161841e-07,
+      "loss": 8.6203,
+      "step": 22480
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 4.219426413883348e-07,
+      "loss": 8.7373,
+      "step": 22500
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 4.216847182333881e-07,
+      "loss": 8.712,
+      "step": 22520
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 4.2142644877160325e-07,
+      "loss": 8.6409,
+      "step": 22540
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 4.2116783352393803e-07,
+      "loss": 8.7292,
+      "step": 22560
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 4.2090887301204763e-07,
+      "loss": 8.602,
+      "step": 22580
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 4.2064956775828366e-07,
+      "loss": 8.8186,
+      "step": 22600
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 4.203899182856932e-07,
+      "loss": 8.6242,
+      "step": 22620
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 4.201299251180176e-07,
+      "loss": 8.7123,
+      "step": 22640
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 4.198695887796914e-07,
+      "loss": 8.8066,
+      "step": 22660
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 4.1960890979584155e-07,
+      "loss": 8.6635,
+      "step": 22680
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 4.1934788869228603e-07,
+      "loss": 8.6928,
+      "step": 22700
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 4.1908652599553293e-07,
+      "loss": 8.6065,
+      "step": 22720
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 4.188248222327794e-07,
+      "loss": 8.651,
+      "step": 22740
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 4.1856277793191044e-07,
+      "loss": 8.5583,
+      "step": 22760
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 4.1830039362149807e-07,
+      "loss": 8.6621,
+      "step": 22780
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 4.1803766983080006e-07,
+      "loss": 8.7382,
+      "step": 22800
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 4.177746070897592e-07,
+      "loss": 8.6508,
+      "step": 22820
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 4.1751120592900156e-07,
+      "loss": 8.6554,
+      "step": 22840
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 4.1724746687983623e-07,
+      "loss": 8.6529,
+      "step": 22860
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 4.169833904742537e-07,
+      "loss": 8.6613,
+      "step": 22880
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 4.1671897724492475e-07,
+      "loss": 8.7068,
+      "step": 22900
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 4.164542277252e-07,
+      "loss": 8.7057,
+      "step": 22920
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 4.1618914244910797e-07,
+      "loss": 8.6056,
+      "step": 22940
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 4.159237219513547e-07,
+      "loss": 8.6663,
+      "step": 22960
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 4.1565796676732237e-07,
+      "loss": 8.838,
+      "step": 22980
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 4.153918774330682e-07,
+      "loss": 8.6266,
+      "step": 23000
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 4.151254544853234e-07,
+      "loss": 8.6716,
+      "step": 23020
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 4.1485869846149233e-07,
+      "loss": 8.8722,
+      "step": 23040
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 4.1459160989965087e-07,
+      "loss": 8.7373,
+      "step": 23060
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.1432418933854586e-07,
+      "loss": 8.5846,
+      "step": 23080
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 4.140564373175939e-07,
+      "loss": 8.6194,
+      "step": 23100
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 4.1378835437687996e-07,
+      "loss": 8.6622,
+      "step": 23120
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 4.135199410571567e-07,
+      "loss": 8.6823,
+      "step": 23140
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.132511978998432e-07,
+      "loss": 8.6646,
+      "step": 23160
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 4.1298212544702356e-07,
+      "loss": 8.7136,
+      "step": 23180
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 4.1271272424144645e-07,
+      "loss": 8.6156,
+      "step": 23200
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 4.124429948265235e-07,
+      "loss": 8.6579,
+      "step": 23220
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.1217293774632844e-07,
+      "loss": 8.5948,
+      "step": 23240
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 4.11902553545596e-07,
+      "loss": 8.5581,
+      "step": 23260
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 4.116318427697205e-07,
+      "loss": 8.7029,
+      "step": 23280
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.113608059647552e-07,
+      "loss": 8.6475,
+      "step": 23300
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 4.1108944367741105e-07,
+      "loss": 8.6973,
+      "step": 23320
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 4.108177564550554e-07,
+      "loss": 8.6234,
+      "step": 23340
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 4.1054574484571105e-07,
+      "loss": 8.7258,
+      "step": 23360
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.10273409398055e-07,
+      "loss": 8.648,
+      "step": 23380
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 4.100007506614178e-07,
+      "loss": 8.5639,
+      "step": 23400
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 4.097277691857819e-07,
+      "loss": 8.7568,
+      "step": 23420
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.094544655217807e-07,
+      "loss": 8.818,
+      "step": 23440
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 4.091808402206976e-07,
+      "loss": 8.6586,
+      "step": 23460
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 4.0890689383446476e-07,
+      "loss": 8.7856,
+      "step": 23480
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 4.086326269156618e-07,
+      "loss": 8.5939,
+      "step": 23500
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.083580400175153e-07,
+      "loss": 8.647,
+      "step": 23520
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 4.0808313369389693e-07,
+      "loss": 8.6309,
+      "step": 23540
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 4.078079084993227e-07,
+      "loss": 8.827,
+      "step": 23560
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 4.0753236498895215e-07,
+      "loss": 8.7282,
+      "step": 23580
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.0725650371858646e-07,
+      "loss": 8.6681,
+      "step": 23600
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 4.069803252446679e-07,
+      "loss": 8.6488,
+      "step": 23620
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 4.0670383012427877e-07,
+      "loss": 8.6818,
+      "step": 23640
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.0642701891513996e-07,
+      "loss": 8.8065,
+      "step": 23660
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 4.0614989217560983e-07,
+      "loss": 8.8094,
+      "step": 23680
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 4.058724504646834e-07,
+      "loss": 8.7127,
+      "step": 23700
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 4.0559469434199077e-07,
+      "loss": 8.6676,
+      "step": 23720
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.0531662436779654e-07,
+      "loss": 8.7212,
+      "step": 23740
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 4.050382411029981e-07,
+      "loss": 8.721,
+      "step": 23760
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 4.0475954510912513e-07,
+      "loss": 8.6397,
+      "step": 23780
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.044805369483377e-07,
+      "loss": 8.6683,
+      "step": 23800
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 4.0420121718342583e-07,
+      "loss": 8.6248,
+      "step": 23820
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 4.0392158637780794e-07,
+      "loss": 8.5869,
+      "step": 23840
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 4.0364164509553e-07,
+      "loss": 8.6588,
+      "step": 23860
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0336139390126424e-07,
+      "loss": 8.692,
+      "step": 23880
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 4.0308083336030784e-07,
+      "loss": 8.6193,
+      "step": 23900
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 4.027999640385821e-07,
+      "loss": 8.6798,
+      "step": 23920
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 4.0251878650263107e-07,
+      "loss": 8.7117,
+      "step": 23940
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.022373013196206e-07,
+      "loss": 8.6933,
+      "step": 23960
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 4.019555090573372e-07,
+      "loss": 8.5723,
+      "step": 23980
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 4.0167341028418655e-07,
+      "loss": 8.6345,
+      "step": 24000
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 4.0139100556919266e-07,
+      "loss": 8.6287,
+      "step": 24020
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 4.0110829548199667e-07,
+      "loss": 8.621,
+      "step": 24040
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 4.008252805928559e-07,
+      "loss": 8.7364,
+      "step": 24060
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 4.005419614726421e-07,
+      "loss": 8.7642,
+      "step": 24080
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 4.00258338692841e-07,
+      "loss": 8.6653,
+      "step": 24100
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 3.999744128255508e-07,
+      "loss": 8.6865,
+      "step": 24120
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 3.996901844434809e-07,
+      "loss": 8.6638,
+      "step": 24140
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 3.994056541199511e-07,
+      "loss": 8.6719,
+      "step": 24160
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.9912082242889e-07,
+      "loss": 8.6338,
+      "step": 24180
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 3.988356899448344e-07,
+      "loss": 8.7657,
+      "step": 24200
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 3.9855025724292763e-07,
+      "loss": 8.6599,
+      "step": 24220
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.982645248989186e-07,
+      "loss": 8.5804,
+      "step": 24240
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 3.9797849348916074e-07,
+      "loss": 8.7593,
+      "step": 24260
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 3.9769216359061063e-07,
+      "loss": 8.7266,
+      "step": 24280
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 3.974055357808269e-07,
+      "loss": 8.683,
+      "step": 24300
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.971186106379693e-07,
+      "loss": 8.7335,
+      "step": 24320
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 3.968313887407971e-07,
+      "loss": 8.7124,
+      "step": 24340
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 3.9654387066866833e-07,
+      "loss": 8.6624,
+      "step": 24360
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.962560570015383e-07,
+      "loss": 8.6476,
+      "step": 24380
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 3.9596794831995863e-07,
+      "loss": 8.6399,
+      "step": 24400
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 3.9567954520507594e-07,
+      "loss": 8.6308,
+      "step": 24420
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 3.953908482386311e-07,
+      "loss": 8.7086,
+      "step": 24440
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.9510185800295715e-07,
+      "loss": 8.7306,
+      "step": 24460
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 3.94812575080979e-07,
+      "loss": 8.688,
+      "step": 24480
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 3.9452300005621206e-07,
+      "loss": 8.7237,
+      "step": 24500
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 3.9423313351276075e-07,
+      "loss": 8.7392,
+      "step": 24520
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.9394297603531756e-07,
+      "loss": 8.6499,
+      "step": 24540
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 3.9365252820916186e-07,
+      "loss": 8.6272,
+      "step": 24560
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 3.933617906201585e-07,
+      "loss": 8.7747,
+      "step": 24580
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.930707638547571e-07,
+      "loss": 8.744,
+      "step": 24600
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 3.927794484999905e-07,
+      "loss": 8.6325,
+      "step": 24620
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 3.924878451434735e-07,
+      "loss": 8.8209,
+      "step": 24640
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 3.9219595437340205e-07,
+      "loss": 8.6263,
+      "step": 24660
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.9190377677855155e-07,
+      "loss": 8.5573,
+      "step": 24680
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 3.916113129482762e-07,
+      "loss": 8.6047,
+      "step": 24700
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 3.913185634725077e-07,
+      "loss": 8.7075,
+      "step": 24720
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.9102552894175347e-07,
+      "loss": 8.6059,
+      "step": 24740
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 3.907322099470963e-07,
+      "loss": 8.7004,
+      "step": 24760
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 3.904386070801927e-07,
+      "loss": 8.645,
+      "step": 24780
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 3.9014472093327164e-07,
+      "loss": 8.7493,
+      "step": 24800
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.8985055209913367e-07,
+      "loss": 8.7113,
+      "step": 24820
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 3.8955610117114946e-07,
+      "loss": 8.7381,
+      "step": 24840
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 3.8926136874325867e-07,
+      "loss": 8.6343,
+      "step": 24860
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 3.889663554099688e-07,
+      "loss": 8.6153,
+      "step": 24880
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 3.88671061766354e-07,
+      "loss": 8.7392,
+      "step": 24900
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 3.883754884080539e-07,
+      "loss": 8.7271,
+      "step": 24920
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 3.880796359312722e-07,
+      "loss": 8.7976,
+      "step": 24940
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 3.8778350493277566e-07,
+      "loss": 8.6362,
+      "step": 24960
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 3.8748709600989296e-07,
+      "loss": 8.7061,
+      "step": 24980
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 3.871904097605131e-07,
+      "loss": 8.6856,
+      "step": 25000
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 3.8689344678308476e-07,
+      "loss": 8.6853,
+      "step": 25020
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 3.8659620767661483e-07,
+      "loss": 8.6812,
+      "step": 25040
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 3.862986930406669e-07,
+      "loss": 8.5669,
+      "step": 25060
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 3.8600090347536064e-07,
+      "loss": 8.747,
+      "step": 25080
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 3.8570283958137e-07,
+      "loss": 8.7162,
+      "step": 25100
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 3.8540450195992255e-07,
+      "loss": 8.6484,
+      "step": 25120
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 3.8510589121279787e-07,
+      "loss": 8.6183,
+      "step": 25140
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 3.8480700794232634e-07,
+      "loss": 8.5655,
+      "step": 25160
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 3.845078527513883e-07,
+      "loss": 8.8672,
+      "step": 25180
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 3.842084262434125e-07,
+      "loss": 8.6426,
+      "step": 25200
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 3.839087290223748e-07,
+      "loss": 8.6784,
+      "step": 25220
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 3.8360876169279734e-07,
+      "loss": 8.686,
+      "step": 25240
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 3.8330852485974697e-07,
+      "loss": 8.6127,
+      "step": 25260
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 3.8300801912883414e-07,
+      "loss": 8.6752,
+      "step": 25280
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 3.8270724510621177e-07,
+      "loss": 8.7257,
+      "step": 25300
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 3.82406203398574e-07,
+      "loss": 8.706,
+      "step": 25320
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 3.8210489461315485e-07,
+      "loss": 8.6395,
+      "step": 25340
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 3.81803319357727e-07,
+      "loss": 8.6969,
+      "step": 25360
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 3.8150147824060075e-07,
+      "loss": 8.8156,
+      "step": 25380
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 3.8119937187062254e-07,
+      "loss": 8.634,
+      "step": 25400
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 3.8089700085717405e-07,
+      "loss": 8.7115,
+      "step": 25420
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 3.8059436581017044e-07,
+      "loss": 8.7182,
+      "step": 25440
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 3.802914673400599e-07,
+      "loss": 8.6773,
+      "step": 25460
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 3.7998830605782175e-07,
+      "loss": 8.5986,
+      "step": 25480
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 3.796848825749652e-07,
+      "loss": 8.6632,
+      "step": 25500
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 3.7938119750352885e-07,
+      "loss": 8.7261,
+      "step": 25520
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 3.790772514560785e-07,
+      "loss": 8.9573,
+      "step": 25540
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 3.787730450457065e-07,
+      "loss": 8.8924,
+      "step": 25560
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 3.7846857888603056e-07,
+      "loss": 8.6284,
+      "step": 25580
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 3.781638535911922e-07,
+      "loss": 8.6416,
+      "step": 25600
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 3.7785886977585555e-07,
+      "loss": 8.6983,
+      "step": 25620
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 3.775536280552063e-07,
+      "loss": 8.6153,
+      "step": 25640
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 3.7724812904495035e-07,
+      "loss": 8.6753,
+      "step": 25660
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 3.769423733613126e-07,
+      "loss": 8.7574,
+      "step": 25680
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 3.7663636162103577e-07,
+      "loss": 8.6958,
+      "step": 25700
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 3.76330094441379e-07,
+      "loss": 8.6696,
+      "step": 25720
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 3.760235724401164e-07,
+      "loss": 8.6922,
+      "step": 25740
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 3.757167962355365e-07,
+      "loss": 8.7126,
+      "step": 25760
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 3.754097664464405e-07,
+      "loss": 8.9095,
+      "step": 25780
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 3.7510248369214093e-07,
+      "loss": 8.6746,
+      "step": 25800
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 3.7479494859246073e-07,
+      "loss": 8.5895,
+      "step": 25820
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 3.744871617677319e-07,
+      "loss": 8.6395,
+      "step": 25840
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 3.7417912383879394e-07,
+      "loss": 8.7634,
+      "step": 25860
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 3.73870835426993e-07,
+      "loss": 8.7353,
+      "step": 25880
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 3.735622971541807e-07,
+      "loss": 8.6576,
+      "step": 25900
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 3.732535096427123e-07,
+      "loss": 8.7598,
+      "step": 25920
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 3.7294447351544594e-07,
+      "loss": 8.669,
+      "step": 25940
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 3.7263518939574136e-07,
+      "loss": 8.595,
+      "step": 25960
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 3.723256579074583e-07,
+      "loss": 8.6576,
+      "step": 25980
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 3.720158796749556e-07,
+      "loss": 8.6325,
+      "step": 26000
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 3.7170585532308995e-07,
+      "loss": 8.6388,
+      "step": 26020
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 3.713955854772143e-07,
+      "loss": 8.6669,
+      "step": 26040
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 3.710850707631767e-07,
+      "loss": 8.6721,
+      "step": 26060
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 3.707743118073195e-07,
+      "loss": 8.7287,
+      "step": 26080
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 3.704633092364773e-07,
+      "loss": 8.6791,
+      "step": 26100
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 3.7015206367797627e-07,
+      "loss": 8.6707,
+      "step": 26120
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 3.698405757596327e-07,
+      "loss": 8.6335,
+      "step": 26140
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 3.695288461097519e-07,
+      "loss": 8.6813,
+      "step": 26160
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 3.692168753571265e-07,
+      "loss": 8.6498,
+      "step": 26180
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 3.6890466413103574e-07,
+      "loss": 8.6455,
+      "step": 26200
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 3.6859221306124353e-07,
+      "loss": 8.626,
+      "step": 26220
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 3.682795227779981e-07,
+      "loss": 8.6599,
+      "step": 26240
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 3.6796659391202976e-07,
+      "loss": 8.6244,
+      "step": 26260
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 3.6765342709455035e-07,
+      "loss": 8.6675,
+      "step": 26280
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 3.673400229572514e-07,
+      "loss": 8.6763,
+      "step": 26300
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 3.670263821323034e-07,
+      "loss": 8.7124,
+      "step": 26320
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 3.667125052523542e-07,
+      "loss": 8.6563,
+      "step": 26340
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 3.6639839295052776e-07,
+      "loss": 8.6153,
+      "step": 26360
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 3.660840458604228e-07,
+      "loss": 8.663,
+      "step": 26380
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 3.657694646161119e-07,
+      "loss": 8.7112,
+      "step": 26400
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 3.654546498521397e-07,
+      "loss": 8.6633,
+      "step": 26420
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 3.6513960220352204e-07,
+      "loss": 8.6195,
+      "step": 26440
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 3.6482432230574445e-07,
+      "loss": 8.7926,
+      "step": 26460
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 3.645088107947609e-07,
+      "loss": 8.6659,
+      "step": 26480
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 3.641930683069927e-07,
+      "loss": 8.6512,
+      "step": 26500
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 3.638770954793268e-07,
+      "loss": 8.6814,
+      "step": 26520
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 3.6356089294911494e-07,
+      "loss": 8.7023,
+      "step": 26540
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 3.632444613541723e-07,
+      "loss": 8.7315,
+      "step": 26560
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 3.629278013327759e-07,
+      "loss": 8.5947,
+      "step": 26580
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 3.6261091352366363e-07,
+      "loss": 8.8138,
+      "step": 26600
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 3.622937985660328e-07,
+      "loss": 8.7118,
+      "step": 26620
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 3.6197645709953895e-07,
+      "loss": 8.681,
+      "step": 26640
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 3.6165888976429447e-07,
+      "loss": 8.7761,
+      "step": 26660
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 3.613410972008674e-07,
+      "loss": 8.6516,
+      "step": 26680
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 3.610230800502802e-07,
+      "loss": 8.6794,
+      "step": 26700
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 3.60704838954008e-07,
+      "loss": 8.7636,
+      "step": 26720
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 3.6038637455397796e-07,
+      "loss": 8.6292,
+      "step": 26740
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 3.6006768749256755e-07,
+      "loss": 8.7857,
+      "step": 26760
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 3.597487784126035e-07,
+      "loss": 8.6254,
+      "step": 26780
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 3.594296479573602e-07,
+      "loss": 8.6402,
+      "step": 26800
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 3.591102967705586e-07,
+      "loss": 8.6597,
+      "step": 26820
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 3.5879072549636494e-07,
+      "loss": 8.7317,
+      "step": 26840
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 3.584709347793895e-07,
+      "loss": 8.6803,
+      "step": 26860
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 3.581509252646851e-07,
+      "loss": 8.6608,
+      "step": 26880
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 3.5783069759774587e-07,
+      "loss": 8.6517,
+      "step": 26900
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 3.5751025242450596e-07,
+      "loss": 8.6721,
+      "step": 26920
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 3.5718959039133836e-07,
+      "loss": 8.7057,
+      "step": 26940
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 3.568687121450533e-07,
+      "loss": 8.5605,
+      "step": 26960
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 3.565476183328975e-07,
+      "loss": 8.666,
+      "step": 26980
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 3.5622630960255215e-07,
+      "loss": 8.6459,
+      "step": 27000
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 3.5590478660213206e-07,
+      "loss": 8.6509,
+      "step": 27020
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 3.5558304998018426e-07,
+      "loss": 8.6461,
+      "step": 27040
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 3.5526110038568664e-07,
+      "loss": 8.7623,
+      "step": 27060
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 3.5493893846804673e-07,
+      "loss": 8.6492,
+      "step": 27080
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 3.546165648771004e-07,
+      "loss": 8.6686,
+      "step": 27100
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 3.5429398026311037e-07,
+      "loss": 8.7239,
+      "step": 27120
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 3.5397118527676505e-07,
+      "loss": 8.6809,
+      "step": 27140
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 3.5364818056917724e-07,
+      "loss": 8.6912,
+      "step": 27160
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 3.5332496679188264e-07,
+      "loss": 8.6091,
+      "step": 27180
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 3.530015445968388e-07,
+      "loss": 8.7204,
+      "step": 27200
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 3.5267791463642367e-07,
+      "loss": 8.6489,
+      "step": 27220
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 3.5235407756343416e-07,
+      "loss": 8.5754,
+      "step": 27240
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 3.520300340310852e-07,
+      "loss": 8.6814,
+      "step": 27260
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.517057846930078e-07,
+      "loss": 8.6847,
+      "step": 27280
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 3.5138133020324844e-07,
+      "loss": 8.6579,
+      "step": 27300
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 3.510566712162673e-07,
+      "loss": 8.716,
+      "step": 27320
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 3.5073180838693694e-07,
+      "loss": 8.6446,
+      "step": 27340
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.5040674237054125e-07,
+      "loss": 8.7245,
+      "step": 27360
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 3.500814738227739e-07,
+      "loss": 8.7127,
+      "step": 27380
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 3.49756003399737e-07,
+      "loss": 8.6278,
+      "step": 27400
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 3.494303317579401e-07,
+      "loss": 8.781,
+      "step": 27420
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 3.491044595542985e-07,
+      "loss": 8.6123,
+      "step": 27440
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 3.4877838744613194e-07,
+      "loss": 8.6652,
+      "step": 27460
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 3.4845211609116354e-07,
+      "loss": 8.6108,
+      "step": 27480
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 3.481256461475182e-07,
+      "loss": 8.7342,
+      "step": 27500
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 3.4779897827372164e-07,
+      "loss": 8.6786,
+      "step": 27520
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 3.474721131286985e-07,
+      "loss": 8.7928,
+      "step": 27540
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 3.4714505137177163e-07,
+      "loss": 8.6573,
+      "step": 27560
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 3.468177936626603e-07,
+      "loss": 8.6552,
+      "step": 27580
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 3.4649034066147894e-07,
+      "loss": 8.6789,
+      "step": 27600
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 3.461626930287361e-07,
+      "loss": 8.6438,
+      "step": 27620
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 3.4583485142533303e-07,
+      "loss": 8.6694,
+      "step": 27640
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 3.455068165125619e-07,
+      "loss": 8.746,
+      "step": 27660
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 3.4517858895210493e-07,
+      "loss": 8.705,
+      "step": 27680
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 3.448501694060332e-07,
+      "loss": 8.6114,
+      "step": 27700
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 3.4452155853680454e-07,
+      "loss": 8.7115,
+      "step": 27720
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 3.441927570072632e-07,
+      "loss": 8.923,
+      "step": 27740
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 3.438637654806378e-07,
+      "loss": 8.8009,
+      "step": 27760
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 3.4353458462053995e-07,
+      "loss": 8.6399,
+      "step": 27780
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 3.432052150909637e-07,
+      "loss": 8.941,
+      "step": 27800
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 3.428756575562832e-07,
+      "loss": 8.7679,
+      "step": 27820
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 3.4254591268125214e-07,
+      "loss": 8.7031,
+      "step": 27840
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 3.4221598113100194e-07,
+      "loss": 8.7045,
+      "step": 27860
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 3.418858635710406e-07,
+      "loss": 8.7079,
+      "step": 27880
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 3.415555606672512e-07,
+      "loss": 8.6618,
+      "step": 27900
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 3.412250730858909e-07,
+      "loss": 8.7842,
+      "step": 27920
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 3.408944014935892e-07,
+      "loss": 8.7194,
+      "step": 27940
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 3.4056354655734686e-07,
+      "loss": 8.7642,
+      "step": 27960
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 3.402325089445346e-07,
+      "loss": 8.6615,
+      "step": 27980
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 3.399012893228912e-07,
+      "loss": 8.7468,
+      "step": 28000
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 3.3956988836052293e-07,
+      "loss": 8.671,
+      "step": 28020
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 3.392383067259018e-07,
+      "loss": 8.6229,
+      "step": 28040
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 3.389065450878641e-07,
+      "loss": 8.6014,
+      "step": 28060
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 3.3857460411560943e-07,
+      "loss": 8.7231,
+      "step": 28080
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 3.38242484478699e-07,
+      "loss": 8.5853,
+      "step": 28100
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 3.379101868470543e-07,
+      "loss": 8.6662,
+      "step": 28120
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 3.375777118909561e-07,
+      "loss": 8.6895,
+      "step": 28140
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 3.372450602810426e-07,
+      "loss": 8.6535,
+      "step": 28160
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 3.3691223268830846e-07,
+      "loss": 8.6333,
+      "step": 28180
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 3.3657922978410335e-07,
+      "loss": 8.7247,
+      "step": 28200
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 3.3624605224013054e-07,
+      "loss": 8.5743,
+      "step": 28220
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 3.3591270072844547e-07,
+      "loss": 8.6759,
+      "step": 28240
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 3.355791759214546e-07,
+      "loss": 8.6982,
+      "step": 28260
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 3.3524547849191396e-07,
+      "loss": 8.7385,
+      "step": 28280
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 3.3491160911292774e-07,
+      "loss": 8.6972,
+      "step": 28300
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 3.3457756845794687e-07,
+      "loss": 8.6619,
+      "step": 28320
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 3.342433572007679e-07,
+      "loss": 8.659,
+      "step": 28340
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 3.3390897601553146e-07,
+      "loss": 8.6246,
+      "step": 28360
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 3.3357442557672096e-07,
+      "loss": 8.6084,
+      "step": 28380
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 3.3323970655916115e-07,
+      "loss": 8.5871,
+      "step": 28400
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 3.3290481963801696e-07,
+      "loss": 8.6602,
+      "step": 28420
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 3.325697654887918e-07,
+      "loss": 8.6066,
+      "step": 28440
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 3.3223454478732647e-07,
+      "loss": 8.6817,
+      "step": 28460
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 3.3189915820979785e-07,
+      "loss": 8.7466,
+      "step": 28480
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 3.3156360643271736e-07,
+      "loss": 8.8099,
+      "step": 28500
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 3.312278901329295e-07,
+      "loss": 8.6193,
+      "step": 28520
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 3.3089200998761077e-07,
+      "loss": 8.7632,
+      "step": 28540
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 3.305559666742682e-07,
+      "loss": 8.7607,
+      "step": 28560
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 3.3021976087073767e-07,
+      "loss": 8.5753,
+      "step": 28580
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 3.298833932551832e-07,
+      "loss": 8.7299,
+      "step": 28600
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 3.295468645060951e-07,
+      "loss": 8.7158,
+      "step": 28620
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 3.2921017530228845e-07,
+      "loss": 8.6602,
+      "step": 28640
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 3.288733263229022e-07,
+      "loss": 8.6871,
+      "step": 28660
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 3.2853631824739756e-07,
+      "loss": 8.751,
+      "step": 28680
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 3.281991517555568e-07,
+      "loss": 8.7443,
+      "step": 28700
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 3.278618275274814e-07,
+      "loss": 8.6701,
+      "step": 28720
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 3.2752434624359127e-07,
+      "loss": 8.8152,
+      "step": 28740
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 3.2718670858462296e-07,
+      "loss": 8.6241,
+      "step": 28760
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 3.2684891523162854e-07,
+      "loss": 8.6202,
+      "step": 28780
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 3.2651096686597423e-07,
+      "loss": 8.7507,
+      "step": 28800
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 3.261728641693387e-07,
+      "loss": 8.6473,
+      "step": 28820
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 3.2583460782371215e-07,
+      "loss": 8.582,
+      "step": 28840
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 3.254961985113944e-07,
+      "loss": 8.77,
+      "step": 28860
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 3.2515763691499425e-07,
+      "loss": 8.7138,
+      "step": 28880
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 3.2481892371742725e-07,
+      "loss": 8.6842,
+      "step": 28900
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 3.2448005960191507e-07,
+      "loss": 8.67,
+      "step": 28920
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 3.241410452519835e-07,
+      "loss": 8.7029,
+      "step": 28940
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 3.2380188135146173e-07,
+      "loss": 8.658,
+      "step": 28960
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 3.234625685844802e-07,
+      "loss": 8.6853,
+      "step": 28980
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 3.2312310763547005e-07,
+      "loss": 8.5897,
+      "step": 29000
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 3.227834991891609e-07,
+      "loss": 8.7498,
+      "step": 29020
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 3.224437439305803e-07,
+      "loss": 8.6559,
+      "step": 29040
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 3.2210384254505164e-07,
+      "loss": 8.6665,
+      "step": 29060
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 3.2176379571819314e-07,
+      "loss": 8.6583,
+      "step": 29080
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 3.214236041359164e-07,
+      "loss": 8.6153,
+      "step": 29100
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 3.2108326848442503e-07,
+      "loss": 8.6621,
+      "step": 29120
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 3.2074278945021326e-07,
+      "loss": 8.6406,
+      "step": 29140
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 3.2040216772006457e-07,
+      "loss": 8.7462,
+      "step": 29160
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 3.200614039810501e-07,
+      "loss": 8.6706,
+      "step": 29180
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 3.197204989205276e-07,
+      "loss": 8.6145,
+      "step": 29200
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 3.1937945322613976e-07,
+      "loss": 8.6873,
+      "step": 29220
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 3.190382675858131e-07,
+      "loss": 8.6631,
+      "step": 29240
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 3.186969426877563e-07,
+      "loss": 8.8381,
+      "step": 29260
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 3.18355479220459e-07,
+      "loss": 8.6286,
+      "step": 29280
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 3.1801387787269043e-07,
+      "loss": 8.6504,
+      "step": 29300
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 3.1767213933349756e-07,
+      "loss": 8.6191,
+      "step": 29320
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 3.173302642922046e-07,
+      "loss": 8.7694,
+      "step": 29340
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 3.1698825343841086e-07,
+      "loss": 8.7551,
+      "step": 29360
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 3.166461074619895e-07,
+      "loss": 8.6852,
+      "step": 29380
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 3.1630382705308637e-07,
+      "loss": 8.582,
+      "step": 29400
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 3.1596141290211854e-07,
+      "loss": 8.7058,
+      "step": 29420
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 3.156188656997727e-07,
+      "loss": 8.7603,
+      "step": 29440
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 3.1527618613700396e-07,
+      "loss": 8.591,
+      "step": 29460
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 3.1493337490503457e-07,
+      "loss": 8.7316,
+      "step": 29480
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 3.145904326953521e-07,
+      "loss": 8.6253,
+      "step": 29500
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 3.142473601997086e-07,
+      "loss": 8.5578,
+      "step": 29520
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 3.1390415811011864e-07,
+      "loss": 8.6595,
+      "step": 29540
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 3.1356082711885846e-07,
+      "loss": 8.6187,
+      "step": 29560
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 3.1321736791846416e-07,
+      "loss": 8.7339,
+      "step": 29580
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 3.1287378120173045e-07,
+      "loss": 8.7281,
+      "step": 29600
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 3.125300676617093e-07,
+      "loss": 8.6875,
+      "step": 29620
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 3.121862279917084e-07,
+      "loss": 8.5509,
+      "step": 29640
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 3.118422628852901e-07,
+      "loss": 8.6831,
+      "step": 29660
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 3.1149817303626947e-07,
+      "loss": 8.6986,
+      "step": 29680
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 3.111539591387135e-07,
+      "loss": 8.6567,
+      "step": 29700
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 3.1080962188693907e-07,
+      "loss": 8.7286,
+      "step": 29720
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 3.1046516197551204e-07,
+      "loss": 8.6032,
+      "step": 29740
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 3.101205800992458e-07,
+      "loss": 8.6943,
+      "step": 29760
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 3.097758769531996e-07,
+      "loss": 8.7576,
+      "step": 29780
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 3.0943105323267746e-07,
+      "loss": 8.6276,
+      "step": 29800
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 3.0908610963322626e-07,
+      "loss": 8.6576,
+      "step": 29820
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 3.0874104685063515e-07,
+      "loss": 8.6753,
+      "step": 29840
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 3.0839586558093333e-07,
+      "loss": 8.6131,
+      "step": 29860
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 3.080505665203893e-07,
+      "loss": 8.667,
+      "step": 29880
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 3.077051503655089e-07,
+      "loss": 8.6642,
+      "step": 29900
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 3.073596178130342e-07,
+      "loss": 8.6796,
+      "step": 29920
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 3.070139695599423e-07,
+      "loss": 8.7781,
+      "step": 29940
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 3.066682063034433e-07,
+      "loss": 8.6049,
+      "step": 29960
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 3.063223287409797e-07,
+      "loss": 8.5715,
+      "step": 29980
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 3.059763375702241e-07,
+      "loss": 8.7563,
+      "step": 30000
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 3.056302334890786e-07,
+      "loss": 8.6739,
+      "step": 30020
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 3.05284017195673e-07,
+      "loss": 8.6724,
+      "step": 30040
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 3.049376893883633e-07,
+      "loss": 8.6921,
+      "step": 30060
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 3.0459125076573063e-07,
+      "loss": 8.9008,
+      "step": 30080
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 3.0424470202657946e-07,
+      "loss": 8.7367,
+      "step": 30100
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 3.038980438699366e-07,
+      "loss": 8.5934,
+      "step": 30120
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 3.035512769950493e-07,
+      "loss": 8.5653,
+      "step": 30140
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 3.0320440210138433e-07,
+      "loss": 8.6243,
+      "step": 30160
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 3.028574198886262e-07,
+      "loss": 8.6315,
+      "step": 30180
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 3.0251033105667594e-07,
+      "loss": 8.651,
+      "step": 30200
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 3.021631363056497e-07,
+      "loss": 8.5832,
+      "step": 30220
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 3.018158363358773e-07,
+      "loss": 8.6438,
+      "step": 30240
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 3.0146843184790056e-07,
+      "loss": 8.7393,
+      "step": 30260
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 3.0112092354247235e-07,
+      "loss": 8.7362,
+      "step": 30280
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 3.00773312120555e-07,
+      "loss": 8.7914,
+      "step": 30300
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 3.004255982833186e-07,
+      "loss": 8.6087,
+      "step": 30320
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 3.0007778273214015e-07,
+      "loss": 8.6884,
+      "step": 30340
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.997298661686014e-07,
+      "loss": 8.6022,
+      "step": 30360
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.9938184929448816e-07,
+      "loss": 8.6424,
+      "step": 30380
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.990337328117886e-07,
+      "loss": 8.6903,
+      "step": 30400
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.986855174226915e-07,
+      "loss": 8.6494,
+      "step": 30420
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.983372038295855e-07,
+      "loss": 8.8256,
+      "step": 30440
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.979887927350573e-07,
+      "loss": 8.7223,
+      "step": 30460
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.9764028484188985e-07,
+      "loss": 8.7295,
+      "step": 30480
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.972916808530619e-07,
+      "loss": 8.7298,
+      "step": 30500
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.969429814717456e-07,
+      "loss": 8.8727,
+      "step": 30520
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.9659418740130587e-07,
+      "loss": 8.6341,
+      "step": 30540
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.9624529934529845e-07,
+      "loss": 8.7136,
+      "step": 30560
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.9589631800746864e-07,
+      "loss": 8.712,
+      "step": 30580
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.955472440917498e-07,
+      "loss": 8.6917,
+      "step": 30600
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.9519807830226234e-07,
+      "loss": 8.655,
+      "step": 30620
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.948488213433118e-07,
+      "loss": 8.6994,
+      "step": 30640
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.944994739193876e-07,
+      "loss": 8.6317,
+      "step": 30660
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.9415003673516165e-07,
+      "loss": 8.5908,
+      "step": 30680
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.9380051049548695e-07,
+      "loss": 8.6299,
+      "step": 30700
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.9345089590539605e-07,
+      "loss": 8.715,
+      "step": 30720
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.9310119367009987e-07,
+      "loss": 8.7997,
+      "step": 30740
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.927514044949861e-07,
+      "loss": 8.6665,
+      "step": 30760
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.9240152908561765e-07,
+      "loss": 8.646,
+      "step": 30780
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.9205156814773143e-07,
+      "loss": 8.6162,
+      "step": 30800
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.917015223872369e-07,
+      "loss": 8.6797,
+      "step": 30820
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.913513925102146e-07,
+      "loss": 8.7125,
+      "step": 30840
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.9100117922291476e-07,
+      "loss": 8.7651,
+      "step": 30860
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.9065088323175594e-07,
+      "loss": 8.7213,
+      "step": 30880
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.903005052433234e-07,
+      "loss": 8.7114,
+      "step": 30900
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.8995004596436774e-07,
+      "loss": 8.7222,
+      "step": 30920
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.8959950610180373e-07,
+      "loss": 8.7304,
+      "step": 30940
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.892488863627085e-07,
+      "loss": 8.6837,
+      "step": 30960
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.888981874543205e-07,
+      "loss": 8.6943,
+      "step": 30980
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.8854741008403753e-07,
+      "loss": 8.5617,
+      "step": 31000
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.881965549594161e-07,
+      "loss": 8.8403,
+      "step": 31020
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.878456227881692e-07,
+      "loss": 8.7576,
+      "step": 31040
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.8749461427816546e-07,
+      "loss": 8.6772,
+      "step": 31060
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.871435301374273e-07,
+      "loss": 8.6206,
+      "step": 31080
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.8679237107413e-07,
+      "loss": 8.6233,
+      "step": 31100
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.864411377965995e-07,
+      "loss": 8.6301,
+      "step": 31120
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.860898310133119e-07,
+      "loss": 8.7266,
+      "step": 31140
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.8573845143289123e-07,
+      "loss": 8.7034,
+      "step": 31160
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.853869997641086e-07,
+      "loss": 8.6918,
+      "step": 31180
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.850354767158804e-07,
+      "loss": 8.6865,
+      "step": 31200
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.846838829972671e-07,
+      "loss": 8.6556,
+      "step": 31220
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.843322193174715e-07,
+      "loss": 8.8643,
+      "step": 31240
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.8398048638583774e-07,
+      "loss": 8.623,
+      "step": 31260
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.8362868491184965e-07,
+      "loss": 8.6568,
+      "step": 31280
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.8327681560512925e-07,
+      "loss": 8.6121,
+      "step": 31300
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.829248791754353e-07,
+      "loss": 8.66,
+      "step": 31320
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.8257287633266205e-07,
+      "loss": 8.6959,
+      "step": 31340
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.8222080778683766e-07,
+      "loss": 8.9021,
+      "step": 31360
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.81868674248123e-07,
+      "loss": 8.6877,
+      "step": 31380
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.8151647642680976e-07,
+      "loss": 8.7424,
+      "step": 31400
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.811642150333196e-07,
+      "loss": 8.8318,
+      "step": 31420
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.8081189077820206e-07,
+      "loss": 8.6563,
+      "step": 31440
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.804595043721337e-07,
+      "loss": 8.5683,
+      "step": 31460
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.801070565259165e-07,
+      "loss": 8.6405,
+      "step": 31480
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.797545479504762e-07,
+      "loss": 8.602,
+      "step": 31500
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.7940197935686123e-07,
+      "loss": 8.6209,
+      "step": 31520
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.790493514562408e-07,
+      "loss": 8.7519,
+      "step": 31540
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.78696664959904e-07,
+      "loss": 8.639,
+      "step": 31560
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.783439205792581e-07,
+      "loss": 8.7752,
+      "step": 31580
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.7799111902582693e-07,
+      "loss": 8.5395,
+      "step": 31600
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.7763826101124996e-07,
+      "loss": 8.6828,
+      "step": 31620
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.7728534724728023e-07,
+      "loss": 8.6239,
+      "step": 31640
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.7693237844578336e-07,
+      "loss": 8.6411,
+      "step": 31660
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.7657935531873606e-07,
+      "loss": 8.61,
+      "step": 31680
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.762262785782245e-07,
+      "loss": 8.6629,
+      "step": 31700
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.758731489364431e-07,
+      "loss": 8.7071,
+      "step": 31720
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.7551996710569294e-07,
+      "loss": 8.7268,
+      "step": 31740
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.751667337983803e-07,
+      "loss": 8.6602,
+      "step": 31760
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.7481344972701545e-07,
+      "loss": 8.8078,
+      "step": 31780
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.7446011560421087e-07,
+      "loss": 8.7245,
+      "step": 31800
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.7410673214268017e-07,
+      "loss": 8.7388,
+      "step": 31820
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.737533000552363e-07,
+      "loss": 8.8027,
+      "step": 31840
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.733998200547906e-07,
+      "loss": 8.6749,
+      "step": 31860
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.730462928543507e-07,
+      "loss": 8.6042,
+      "step": 31880
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.726927191670197e-07,
+      "loss": 8.6362,
+      "step": 31900
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.7233909970599426e-07,
+      "loss": 8.7159,
+      "step": 31920
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.7198543518456356e-07,
+      "loss": 8.8243,
+      "step": 31940
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.716317263161076e-07,
+      "loss": 8.7407,
+      "step": 31960
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.712779738140957e-07,
+      "loss": 8.7114,
+      "step": 31980
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.7092417839208537e-07,
+      "loss": 8.6751,
+      "step": 32000
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.7057034076372073e-07,
+      "loss": 8.6862,
+      "step": 32020
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.7021646164273084e-07,
+      "loss": 8.6095,
+      "step": 32040
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.698625417429286e-07,
+      "loss": 8.6848,
+      "step": 32060
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.695085817782091e-07,
+      "loss": 8.7513,
+      "step": 32080
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.691545824625483e-07,
+      "loss": 8.6242,
+      "step": 32100
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.6880054451000144e-07,
+      "loss": 8.7933,
+      "step": 32120
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.6844646863470185e-07,
+      "loss": 8.6506,
+      "step": 32140
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.6809235555085923e-07,
+      "loss": 8.6979,
+      "step": 32160
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.677382059727583e-07,
+      "loss": 8.6129,
+      "step": 32180
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.673840206147576e-07,
+      "loss": 8.6981,
+      "step": 32200
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.670298001912875e-07,
+      "loss": 8.6077,
+      "step": 32220
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.666755454168495e-07,
+      "loss": 8.6092,
+      "step": 32240
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.663212570060141e-07,
+      "loss": 8.651,
+      "step": 32260
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.659669356734198e-07,
+      "loss": 8.7201,
+      "step": 32280
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.6561258213377133e-07,
+      "loss": 8.6061,
+      "step": 32300
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.6525819710183867e-07,
+      "loss": 8.7201,
+      "step": 32320
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.6490378129245496e-07,
+      "loss": 8.6382,
+      "step": 32340
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.645493354205158e-07,
+      "loss": 8.7026,
+      "step": 32360
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.6419486020097713e-07,
+      "loss": 8.6788,
+      "step": 32380
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.638403563488542e-07,
+      "loss": 8.6216,
+      "step": 32400
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.6348582457922006e-07,
+      "loss": 8.7143,
+      "step": 32420
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.6313126560720413e-07,
+      "loss": 8.6853,
+      "step": 32440
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.627766801479904e-07,
+      "loss": 8.676,
+      "step": 32460
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.6242206891681663e-07,
+      "loss": 8.7085,
+      "step": 32480
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.620674326289725e-07,
+      "loss": 8.7902,
+      "step": 32500
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.6171277199979785e-07,
+      "loss": 8.6366,
+      "step": 32520
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.613580877446822e-07,
+      "loss": 8.5595,
+      "step": 32540
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.6100338057906243e-07,
+      "loss": 8.6858,
+      "step": 32560
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.606486512184215e-07,
+      "loss": 8.6326,
+      "step": 32580
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.602939003782875e-07,
+      "loss": 8.6178,
+      "step": 32600
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.5993912877423147e-07,
+      "loss": 8.6427,
+      "step": 32620
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.5958433712186656e-07,
+      "loss": 8.7161,
+      "step": 32640
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.5922952613684627e-07,
+      "loss": 8.6176,
+      "step": 32660
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.5887469653486327e-07,
+      "loss": 8.5744,
+      "step": 32680
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.585198490316475e-07,
+      "loss": 8.6941,
+      "step": 32700
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.5816498434296513e-07,
+      "loss": 8.7242,
+      "step": 32720
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.5781010318461714e-07,
+      "loss": 8.758,
+      "step": 32740
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.5745520627243756e-07,
+      "loss": 8.7095,
+      "step": 32760
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.571002943222922e-07,
+      "loss": 8.6995,
+      "step": 32780
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.567453680500774e-07,
+      "loss": 8.6771,
+      "step": 32800
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.5639042817171804e-07,
+      "loss": 8.6903,
+      "step": 32820
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.560354754031667e-07,
+      "loss": 8.7104,
+      "step": 32840
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.5568051046040197e-07,
+      "loss": 8.7613,
+      "step": 32860
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.553255340594268e-07,
+      "loss": 8.6661,
+      "step": 32880
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.549705469162675e-07,
+      "loss": 8.581,
+      "step": 32900
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.54615549746972e-07,
+      "loss": 8.6058,
+      "step": 32920
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.5426054326760816e-07,
+      "loss": 8.6671,
+      "step": 32940
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.53905528194263e-07,
+      "loss": 8.7079,
+      "step": 32960
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.5355050524304067e-07,
+      "loss": 8.7816,
+      "step": 32980
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.5319547513006124e-07,
+      "loss": 8.7454,
+      "step": 33000
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.528404385714594e-07,
+      "loss": 8.692,
+      "step": 33020
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.524853962833824e-07,
+      "loss": 8.6828,
+      "step": 33040
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.521303489819896e-07,
+      "loss": 8.6124,
+      "step": 33060
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.5177529738345005e-07,
+      "loss": 8.6774,
+      "step": 33080
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.514202422039417e-07,
+      "loss": 8.6826,
+      "step": 33100
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.510651841596496e-07,
+      "loss": 8.6866,
+      "step": 33120
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.5071012396676473e-07,
+      "loss": 8.5997,
+      "step": 33140
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.5035506234148213e-07,
+      "loss": 8.6878,
+      "step": 33160
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.5e-07,
+      "loss": 8.6028,
+      "step": 33180
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.4964493765851795e-07,
+      "loss": 8.5968,
+      "step": 33200
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.492898760332353e-07,
+      "loss": 8.7196,
+      "step": 33220
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.4893481584035043e-07,
+      "loss": 8.5782,
+      "step": 33240
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 2.485797577960583e-07,
+      "loss": 8.6388,
+      "step": 33260
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.4822470261655e-07,
+      "loss": 8.6747,
+      "step": 33280
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 2.478696510180105e-07,
+      "loss": 8.678,
+      "step": 33300
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 2.475146037166176e-07,
+      "loss": 8.7884,
+      "step": 33320
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 2.471595614285407e-07,
+      "loss": 8.6724,
+      "step": 33340
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.4680452486993874e-07,
+      "loss": 8.5924,
+      "step": 33360
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 2.4644949475695936e-07,
+      "loss": 8.6632,
+      "step": 33380
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 2.460944718057371e-07,
+      "loss": 8.7362,
+      "step": 33400
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.4573945673239187e-07,
+      "loss": 8.7602,
+      "step": 33420
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 2.453844502530281e-07,
+      "loss": 8.6372,
+      "step": 33440
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 2.4502945308373243e-07,
+      "loss": 8.6209,
+      "step": 33460
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 2.446744659405732e-07,
+      "loss": 8.6308,
+      "step": 33480
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.44319489539598e-07,
+      "loss": 8.6478,
+      "step": 33500
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 2.439645245968333e-07,
+      "loss": 8.6312,
+      "step": 33520
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 2.43609571828282e-07,
+      "loss": 8.6993,
+      "step": 33540
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.432546319499226e-07,
+      "loss": 8.7386,
+      "step": 33560
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 2.4289970567770775e-07,
+      "loss": 8.6758,
+      "step": 33580
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 2.4254479372756236e-07,
+      "loss": 8.7626,
+      "step": 33600
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 2.4218989681538284e-07,
+      "loss": 8.7846,
+      "step": 33620
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.4183501565703485e-07,
+      "loss": 8.8279,
+      "step": 33640
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 2.4148015096835255e-07,
+      "loss": 8.759,
+      "step": 33660
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 2.4112530346513676e-07,
+      "loss": 8.6495,
+      "step": 33680
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 2.407704738631537e-07,
+      "loss": 8.6556,
+      "step": 33700
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.404156628781335e-07,
+      "loss": 8.6345,
+      "step": 33720
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 2.400608712257686e-07,
+      "loss": 8.6331,
+      "step": 33740
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 2.3970609962171255e-07,
+      "loss": 8.617,
+      "step": 33760
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.3935134878157853e-07,
+      "loss": 8.7142,
+      "step": 33780
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 2.3899661942093755e-07,
+      "loss": 8.6639,
+      "step": 33800
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 2.386419122553178e-07,
+      "loss": 8.6693,
+      "step": 33820
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 2.382872280002022e-07,
+      "loss": 8.7309,
+      "step": 33840
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.3793256737102757e-07,
+      "loss": 8.7177,
+      "step": 33860
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 2.3757793108318337e-07,
+      "loss": 8.6647,
+      "step": 33880
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 2.3722331985200956e-07,
+      "loss": 8.725,
+      "step": 33900
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.368687343927959e-07,
+      "loss": 8.5963,
+      "step": 33920
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 2.3651417542077994e-07,
+      "loss": 8.694,
+      "step": 33940
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 2.361596436511458e-07,
+      "loss": 8.7634,
+      "step": 33960
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 2.3580513979902295e-07,
+      "loss": 8.9021,
+      "step": 33980
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.354506645794842e-07,
+      "loss": 8.6636,
+      "step": 34000
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 2.3509621870754504e-07,
+      "loss": 8.6979,
+      "step": 34020
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 2.347418028981614e-07,
+      "loss": 8.6839,
+      "step": 34040
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 2.3438741786622862e-07,
+      "loss": 8.6451,
+      "step": 34060
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.3403306432658023e-07,
+      "loss": 8.6216,
+      "step": 34080
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 2.3367874299398583e-07,
+      "loss": 8.7277,
+      "step": 34100
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 2.3332445458315048e-07,
+      "loss": 8.6828,
+      "step": 34120
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.3297019980871242e-07,
+      "loss": 8.7317,
+      "step": 34140
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 2.3261597938524244e-07,
+      "loss": 8.7596,
+      "step": 34160
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 2.3226179402724173e-07,
+      "loss": 8.6388,
+      "step": 34180
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 2.3190764444914078e-07,
+      "loss": 8.6952,
+      "step": 34200
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.3155353136529818e-07,
+      "loss": 8.5665,
+      "step": 34220
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 2.311994554899985e-07,
+      "loss": 8.669,
+      "step": 34240
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 2.3084541753745173e-07,
+      "loss": 8.6576,
+      "step": 34260
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 2.3049141822179097e-07,
+      "loss": 8.6854,
+      "step": 34280
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.301374582570714e-07,
+      "loss": 8.6454,
+      "step": 34300
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 2.2978353835726919e-07,
+      "loss": 8.7634,
+      "step": 34320
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 2.2942965923627925e-07,
+      "loss": 8.6704,
+      "step": 34340
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.290758216079146e-07,
+      "loss": 8.7067,
+      "step": 34360
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 2.2872202618590437e-07,
+      "loss": 8.6651,
+      "step": 34380
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 2.2836827368389245e-07,
+      "loss": 8.7543,
+      "step": 34400
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 2.2801456481543645e-07,
+      "loss": 8.6127,
+      "step": 34420
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.276609002940057e-07,
+      "loss": 8.6231,
+      "step": 34440
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 2.2730728083298032e-07,
+      "loss": 8.7681,
+      "step": 34460
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 2.2695370714564925e-07,
+      "loss": 8.636,
+      "step": 34480
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.2660017994520938e-07,
+      "loss": 8.6136,
+      "step": 34500
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 2.2624669994476368e-07,
+      "loss": 8.7001,
+      "step": 34520
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 2.2589326785731986e-07,
+      "loss": 8.6366,
+      "step": 34540
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 2.2553988439578914e-07,
+      "loss": 8.6689,
+      "step": 34560
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.2518655027298464e-07,
+      "loss": 8.5895,
+      "step": 34580
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 2.2483326620161975e-07,
+      "loss": 8.5908,
+      "step": 34600
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 2.2448003289430712e-07,
+      "loss": 8.6957,
+      "step": 34620
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 2.2412685106355693e-07,
+      "loss": 8.6366,
+      "step": 34640
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.237737214217755e-07,
+      "loss": 8.7087,
+      "step": 34660
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 2.2342064468126395e-07,
+      "loss": 8.6485,
+      "step": 34680
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 2.2306762155421662e-07,
+      "loss": 8.6545,
+      "step": 34700
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.2271465275271983e-07,
+      "loss": 8.7019,
+      "step": 34720
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 2.2236173898875002e-07,
+      "loss": 8.69,
+      "step": 34740
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 2.2200888097417302e-07,
+      "loss": 8.638,
+      "step": 34760
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 2.216560794207419e-07,
+      "loss": 8.7119,
+      "step": 34780
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.21303335040096e-07,
+      "loss": 8.7205,
+      "step": 34800
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 2.2095064854375928e-07,
+      "loss": 8.7169,
+      "step": 34820
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 2.2059802064313882e-07,
+      "loss": 8.5366,
+      "step": 34840
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.2024545204952382e-07,
+      "loss": 8.7096,
+      "step": 34860
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 2.1989294347408347e-07,
+      "loss": 8.5825,
+      "step": 34880
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 2.195404956278663e-07,
+      "loss": 8.6663,
+      "step": 34900
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 2.1918810922179803e-07,
+      "loss": 8.6597,
+      "step": 34920
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.1883578496668043e-07,
+      "loss": 8.586,
+      "step": 34940
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 2.1848352357319022e-07,
+      "loss": 8.7007,
+      "step": 34960
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 2.1813132575187697e-07,
+      "loss": 8.7152,
+      "step": 34980
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 2.1777919221316232e-07,
+      "loss": 8.7078,
+      "step": 35000
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.1742712366733803e-07,
+      "loss": 8.7375,
+      "step": 35020
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 2.1707512082456473e-07,
+      "loss": 8.7171,
+      "step": 35040
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 2.167231843948708e-07,
+      "loss": 8.6852,
+      "step": 35060
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.1637131508815027e-07,
+      "loss": 8.6224,
+      "step": 35080
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 2.1601951361416223e-07,
+      "loss": 8.7685,
+      "step": 35100
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 2.1566778068252858e-07,
+      "loss": 8.7018,
+      "step": 35120
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 2.1531611700273295e-07,
+      "loss": 8.7495,
+      "step": 35140
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.1496452328411964e-07,
+      "loss": 8.7644,
+      "step": 35160
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 2.146130002358914e-07,
+      "loss": 8.674,
+      "step": 35180
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 2.142615485671088e-07,
+      "loss": 8.7433,
+      "step": 35200
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 2.139101689866881e-07,
+      "loss": 8.615,
+      "step": 35220
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.135588622034005e-07,
+      "loss": 8.642,
+      "step": 35240
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 2.1320762892587008e-07,
+      "loss": 8.8543,
+      "step": 35260
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 2.128564698625726e-07,
+      "loss": 8.562,
+      "step": 35280
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.1250538572183457e-07,
+      "loss": 8.7055,
+      "step": 35300
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 2.1215437721183074e-07,
+      "loss": 8.6682,
+      "step": 35320
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 2.1180344504058392e-07,
+      "loss": 8.6209,
+      "step": 35340
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 2.1145258991596245e-07,
+      "loss": 8.6196,
+      "step": 35360
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.1110181254567955e-07,
+      "loss": 8.6175,
+      "step": 35380
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 2.1075111363729154e-07,
+      "loss": 8.813,
+      "step": 35400
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 2.1040049389819624e-07,
+      "loss": 8.5952,
+      "step": 35420
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.1004995403563224e-07,
+      "loss": 8.7259,
+      "step": 35440
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 2.096994947566766e-07,
+      "loss": 8.7216,
+      "step": 35460
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 2.0934911676824403e-07,
+      "loss": 8.7122,
+      "step": 35480
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 2.089988207770852e-07,
+      "loss": 8.614,
+      "step": 35500
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.086486074897854e-07,
+      "loss": 8.6394,
+      "step": 35520
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 2.0829847761276316e-07,
+      "loss": 8.6364,
+      "step": 35540
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 2.0794843185226865e-07,
+      "loss": 8.7904,
+      "step": 35560
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 2.0759847091438238e-07,
+      "loss": 8.6759,
+      "step": 35580
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.0724859550501393e-07,
+      "loss": 8.5717,
+      "step": 35600
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 2.0689880632990008e-07,
+      "loss": 8.7717,
+      "step": 35620
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 2.0654910409460396e-07,
+      "loss": 8.6799,
+      "step": 35640
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.0619948950451316e-07,
+      "loss": 8.6777,
+      "step": 35660
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 2.0584996326483838e-07,
+      "loss": 8.6712,
+      "step": 35680
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 2.0550052608061248e-07,
+      "loss": 8.6452,
+      "step": 35700
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 2.0515117865668815e-07,
+      "loss": 8.6933,
+      "step": 35720
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.0480192169773763e-07,
+      "loss": 8.7019,
+      "step": 35740
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 2.0445275590825024e-07,
+      "loss": 8.7944,
+      "step": 35760
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 2.0410368199253142e-07,
+      "loss": 8.7143,
+      "step": 35780
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.0375470065470158e-07,
+      "loss": 8.6063,
+      "step": 35800
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 2.0340581259869405e-07,
+      "loss": 8.6562,
+      "step": 35820
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 2.0305701852825438e-07,
+      "loss": 8.6519,
+      "step": 35840
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 2.027083191469381e-07,
+      "loss": 8.703,
+      "step": 35860
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.0235971515811013e-07,
+      "loss": 8.6644,
+      "step": 35880
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 2.0201120726494278e-07,
+      "loss": 8.751,
+      "step": 35900
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 2.016627961704144e-07,
+      "loss": 8.6513,
+      "step": 35920
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 2.013144825773085e-07,
+      "loss": 8.5942,
+      "step": 35940
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.0096626718821143e-07,
+      "loss": 8.739,
+      "step": 35960
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 2.0061815070551184e-07,
+      "loss": 8.6871,
+      "step": 35980
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 2.002701338313987e-07,
+      "loss": 8.6548,
+      "step": 36000
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.9992221726785988e-07,
+      "loss": 8.7896,
+      "step": 36020
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.995744017166814e-07,
+      "loss": 8.5878,
+      "step": 36040
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.9922668787944497e-07,
+      "loss": 8.6677,
+      "step": 36060
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.9887907645752765e-07,
+      "loss": 8.5799,
+      "step": 36080
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.9853156815209955e-07,
+      "loss": 8.6493,
+      "step": 36100
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.9818416366412275e-07,
+      "loss": 8.7074,
+      "step": 36120
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.9783686369435031e-07,
+      "loss": 8.6966,
+      "step": 36140
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.9748966894332404e-07,
+      "loss": 8.623,
+      "step": 36160
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.9714258011137384e-07,
+      "loss": 8.6557,
+      "step": 36180
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.9679559789861575e-07,
+      "loss": 8.6353,
+      "step": 36200
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.9644872300495068e-07,
+      "loss": 8.5737,
+      "step": 36220
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.9610195613006343e-07,
+      "loss": 8.7244,
+      "step": 36240
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.9575529797342047e-07,
+      "loss": 8.6988,
+      "step": 36260
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.954087492342694e-07,
+      "loss": 8.6604,
+      "step": 36280
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.950623106116367e-07,
+      "loss": 8.614,
+      "step": 36300
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.9471598280432705e-07,
+      "loss": 8.7385,
+      "step": 36320
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.9436976651092142e-07,
+      "loss": 8.6884,
+      "step": 36340
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.9402366242977592e-07,
+      "loss": 8.6888,
+      "step": 36360
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.936776712590203e-07,
+      "loss": 8.7554,
+      "step": 36380
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.933317936965566e-07,
+      "loss": 8.7202,
+      "step": 36400
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.9298603044005774e-07,
+      "loss": 8.7588,
+      "step": 36420
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.9264038218696576e-07,
+      "loss": 8.6336,
+      "step": 36440
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.9229484963449112e-07,
+      "loss": 8.6737,
+      "step": 36460
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.919494334796107e-07,
+      "loss": 8.7338,
+      "step": 36480
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.9160413441906665e-07,
+      "loss": 8.6593,
+      "step": 36500
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.9125895314936488e-07,
+      "loss": 8.6246,
+      "step": 36520
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.909138903667738e-07,
+      "loss": 8.6588,
+      "step": 36540
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.905689467673226e-07,
+      "loss": 8.655,
+      "step": 36560
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.9022412304680042e-07,
+      "loss": 8.6502,
+      "step": 36580
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.8987941990075415e-07,
+      "loss": 8.6792,
+      "step": 36600
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.8953483802448796e-07,
+      "loss": 8.6347,
+      "step": 36620
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.8919037811306104e-07,
+      "loss": 8.64,
+      "step": 36640
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.8884604086128654e-07,
+      "loss": 8.681,
+      "step": 36660
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.885018269637305e-07,
+      "loss": 8.6094,
+      "step": 36680
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.8815773711470987e-07,
+      "loss": 8.6675,
+      "step": 36700
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.8781377200829156e-07,
+      "loss": 8.6894,
+      "step": 36720
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.8746993233829079e-07,
+      "loss": 8.6208,
+      "step": 36740
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.8712621879826955e-07,
+      "loss": 8.6134,
+      "step": 36760
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.867826320815359e-07,
+      "loss": 8.7218,
+      "step": 36780
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.8643917288114146e-07,
+      "loss": 8.8159,
+      "step": 36800
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.8609584188988133e-07,
+      "loss": 8.7762,
+      "step": 36820
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.8575263980029147e-07,
+      "loss": 8.7979,
+      "step": 36840
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.8540956730464785e-07,
+      "loss": 8.6362,
+      "step": 36860
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.8506662509496546e-07,
+      "loss": 8.7675,
+      "step": 36880
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.8472381386299596e-07,
+      "loss": 8.6792,
+      "step": 36900
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.8438113430022733e-07,
+      "loss": 8.6449,
+      "step": 36920
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.840385870978815e-07,
+      "loss": 8.6976,
+      "step": 36940
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.8369617294691358e-07,
+      "loss": 8.7641,
+      "step": 36960
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.8335389253801055e-07,
+      "loss": 8.7507,
+      "step": 36980
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.8301174656158912e-07,
+      "loss": 8.6479,
+      "step": 37000
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.826697357077954e-07,
+      "loss": 8.6015,
+      "step": 37020
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.823278606665024e-07,
+      "loss": 8.74,
+      "step": 37040
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.8198612212730963e-07,
+      "loss": 8.6362,
+      "step": 37060
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.81644520779541e-07,
+      "loss": 8.5745,
+      "step": 37080
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.8130305731224365e-07,
+      "loss": 8.7794,
+      "step": 37100
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.8096173241418695e-07,
+      "loss": 8.5795,
+      "step": 37120
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.8062054677386021e-07,
+      "loss": 8.6263,
+      "step": 37140
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.8027950107947246e-07,
+      "loss": 8.7687,
+      "step": 37160
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.7993859601894992e-07,
+      "loss": 8.6807,
+      "step": 37180
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.795978322799354e-07,
+      "loss": 8.7346,
+      "step": 37200
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.7925721054978674e-07,
+      "loss": 8.6137,
+      "step": 37220
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.7891673151557492e-07,
+      "loss": 8.7427,
+      "step": 37240
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.7857639586408364e-07,
+      "loss": 8.6359,
+      "step": 37260
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.7823620428180692e-07,
+      "loss": 8.7831,
+      "step": 37280
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.778961574549484e-07,
+      "loss": 8.6557,
+      "step": 37300
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.775562560694197e-07,
+      "loss": 8.645,
+      "step": 37320
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.7721650081083905e-07,
+      "loss": 8.6701,
+      "step": 37340
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.7687689236452995e-07,
+      "loss": 8.7074,
+      "step": 37360
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.765374314155198e-07,
+      "loss": 8.7768,
+      "step": 37380
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.7619811864853827e-07,
+      "loss": 8.5979,
+      "step": 37400
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.758589547480165e-07,
+      "loss": 8.6545,
+      "step": 37420
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.75519940398085e-07,
+      "loss": 8.6859,
+      "step": 37440
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.7518107628257273e-07,
+      "loss": 8.6216,
+      "step": 37460
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.748423630850058e-07,
+      "loss": 8.8257,
+      "step": 37480
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.7450380148860556e-07,
+      "loss": 8.6443,
+      "step": 37500
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.741653921762879e-07,
+      "loss": 8.5728,
+      "step": 37520
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.7382713583066125e-07,
+      "loss": 8.6605,
+      "step": 37540
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.734890331340258e-07,
+      "loss": 8.6944,
+      "step": 37560
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.731510847683714e-07,
+      "loss": 8.6816,
+      "step": 37580
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.728132914153771e-07,
+      "loss": 8.7527,
+      "step": 37600
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.7247565375640881e-07,
+      "loss": 8.9439,
+      "step": 37620
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.7213817247251862e-07,
+      "loss": 8.6915,
+      "step": 37640
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.7180084824444325e-07,
+      "loss": 8.6954,
+      "step": 37660
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.7146368175260234e-07,
+      "loss": 8.6402,
+      "step": 37680
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.711266736770978e-07,
+      "loss": 8.7081,
+      "step": 37700
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.7078982469771163e-07,
+      "loss": 8.5867,
+      "step": 37720
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.704531354939049e-07,
+      "loss": 8.6839,
+      "step": 37740
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.7011660674481676e-07,
+      "loss": 8.7227,
+      "step": 37760
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.6978023912926225e-07,
+      "loss": 8.5406,
+      "step": 37780
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.6944403332573185e-07,
+      "loss": 8.6903,
+      "step": 37800
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.6910799001238923e-07,
+      "loss": 8.6448,
+      "step": 37820
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.6877210986707046e-07,
+      "loss": 8.6563,
+      "step": 37840
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.6843639356728267e-07,
+      "loss": 8.7411,
+      "step": 37860
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.6810084179020208e-07,
+      "loss": 8.644,
+      "step": 37880
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.6776545521267354e-07,
+      "loss": 8.6785,
+      "step": 37900
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.674302345112083e-07,
+      "loss": 8.6754,
+      "step": 37920
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.6709518036198307e-07,
+      "loss": 8.6037,
+      "step": 37940
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.6676029344083885e-07,
+      "loss": 8.6496,
+      "step": 37960
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.66425574423279e-07,
+      "loss": 8.5986,
+      "step": 37980
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.6609102398446852e-07,
+      "loss": 8.6326,
+      "step": 38000
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.6575664279923212e-07,
+      "loss": 8.8621,
+      "step": 38020
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.654224315420531e-07,
+      "loss": 8.5913,
+      "step": 38040
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.650883908870723e-07,
+      "loss": 8.6037,
+      "step": 38060
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.6475452150808597e-07,
+      "loss": 8.631,
+      "step": 38080
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.6442082407854538e-07,
+      "loss": 8.6065,
+      "step": 38100
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.6408729927155453e-07,
+      "loss": 8.7009,
+      "step": 38120
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.6375394775986952e-07,
+      "loss": 8.6739,
+      "step": 38140
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.6342077021589669e-07,
+      "loss": 8.5858,
+      "step": 38160
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.6308776731169154e-07,
+      "loss": 8.9962,
+      "step": 38180
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.6275493971895743e-07,
+      "loss": 8.6618,
+      "step": 38200
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.624222881090439e-07,
+      "loss": 8.7069,
+      "step": 38220
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.620898131529457e-07,
+      "loss": 8.7122,
+      "step": 38240
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.61757515521301e-07,
+      "loss": 8.6128,
+      "step": 38260
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.6142539588439052e-07,
+      "loss": 8.9103,
+      "step": 38280
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.6109345491213585e-07,
+      "loss": 8.6976,
+      "step": 38300
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.607616932740982e-07,
+      "loss": 8.6669,
+      "step": 38320
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.6043011163947707e-07,
+      "loss": 8.6298,
+      "step": 38340
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.600987106771089e-07,
+      "loss": 8.6765,
+      "step": 38360
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.5976749105546545e-07,
+      "loss": 8.7179,
+      "step": 38380
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.5943645344265312e-07,
+      "loss": 8.7319,
+      "step": 38400
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5910559850641076e-07,
+      "loss": 8.6038,
+      "step": 38420
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5877492691410913e-07,
+      "loss": 8.6624,
+      "step": 38440
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.5844443933274886e-07,
+      "loss": 8.6136,
+      "step": 38460
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5811413642895943e-07,
+      "loss": 8.8174,
+      "step": 38480
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5778401886899806e-07,
+      "loss": 8.7944,
+      "step": 38500
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5745408731874776e-07,
+      "loss": 8.5984,
+      "step": 38520
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.5712434244371675e-07,
+      "loss": 8.8729,
+      "step": 38540
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5679478490903635e-07,
+      "loss": 8.6856,
+      "step": 38560
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5646541537946003e-07,
+      "loss": 8.6329,
+      "step": 38580
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.5613623451936232e-07,
+      "loss": 8.6972,
+      "step": 38600
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.5580724299273677e-07,
+      "loss": 8.7132,
+      "step": 38620
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.5547844146319544e-07,
+      "loss": 8.6481,
+      "step": 38640
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.551498305939669e-07,
+      "loss": 8.6931,
+      "step": 38660
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.5482141104789504e-07,
+      "loss": 8.7225,
+      "step": 38680
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.5449318348743817e-07,
+      "loss": 8.7001,
+      "step": 38700
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.5416514857466695e-07,
+      "loss": 8.6532,
+      "step": 38720
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.5383730697126386e-07,
+      "loss": 8.6646,
+      "step": 38740
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.5350965933852104e-07,
+      "loss": 8.5851,
+      "step": 38760
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.5318220633733975e-07,
+      "loss": 8.7631,
+      "step": 38780
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.528549486282284e-07,
+      "loss": 8.7466,
+      "step": 38800
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.5252788687130143e-07,
+      "loss": 8.7216,
+      "step": 38820
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.5220102172627837e-07,
+      "loss": 8.5501,
+      "step": 38840
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.518743538524817e-07,
+      "loss": 8.6958,
+      "step": 38860
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.515478839088365e-07,
+      "loss": 8.6093,
+      "step": 38880
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.5122161255386812e-07,
+      "loss": 8.6491,
+      "step": 38900
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.5089554044570149e-07,
+      "loss": 8.6376,
+      "step": 38920
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.5056966824205988e-07,
+      "loss": 8.729,
+      "step": 38940
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.5024399660026294e-07,
+      "loss": 8.7798,
+      "step": 38960
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.4991852617722617e-07,
+      "loss": 8.6727,
+      "step": 38980
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.495932576294588e-07,
+      "loss": 8.6122,
+      "step": 39000
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.4926819161306306e-07,
+      "loss": 8.7075,
+      "step": 39020
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.4894332878373276e-07,
+      "loss": 8.7238,
+      "step": 39040
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.4861866979675152e-07,
+      "loss": 8.7556,
+      "step": 39060
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.4829421530699222e-07,
+      "loss": 8.7966,
+      "step": 39080
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4796996596891487e-07,
+      "loss": 8.7137,
+      "step": 39100
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.4764592243656582e-07,
+      "loss": 8.6169,
+      "step": 39120
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4732208536357636e-07,
+      "loss": 8.6715,
+      "step": 39140
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4699845540316123e-07,
+      "loss": 8.6655,
+      "step": 39160
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.466750332081174e-07,
+      "loss": 8.6515,
+      "step": 39180
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4635181943082284e-07,
+      "loss": 8.6899,
+      "step": 39200
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4602881472323498e-07,
+      "loss": 8.731,
+      "step": 39220
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.4570601973688966e-07,
+      "loss": 8.6511,
+      "step": 39240
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.4538343512289957e-07,
+      "loss": 8.6555,
+      "step": 39260
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.450610615319533e-07,
+      "loss": 8.6992,
+      "step": 39280
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.447388996143134e-07,
+      "loss": 8.6656,
+      "step": 39300
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.4441695001981585e-07,
+      "loss": 8.6907,
+      "step": 39320
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.4409521339786808e-07,
+      "loss": 8.8347,
+      "step": 39340
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.437736903974479e-07,
+      "loss": 8.6904,
+      "step": 39360
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4345238166710254e-07,
+      "loss": 8.6437,
+      "step": 39380
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.4313128785494667e-07,
+      "loss": 8.6775,
+      "step": 39400
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4281040960866175e-07,
+      "loss": 8.5845,
+      "step": 39420
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4248974757549415e-07,
+      "loss": 8.6379,
+      "step": 39440
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.421693024022542e-07,
+      "loss": 8.6482,
+      "step": 39460
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.4184907473531496e-07,
+      "loss": 8.7291,
+      "step": 39480
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.4152906522061047e-07,
+      "loss": 8.7168,
+      "step": 39500
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.412092745036351e-07,
+      "loss": 8.6693,
+      "step": 39520
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.4088970322944145e-07,
+      "loss": 8.6391,
+      "step": 39540
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.405703520426399e-07,
+      "loss": 8.705,
+      "step": 39560
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.402512215873965e-07,
+      "loss": 8.84,
+      "step": 39580
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.3993231250743243e-07,
+      "loss": 8.7896,
+      "step": 39600
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.3961362544602212e-07,
+      "loss": 8.599,
+      "step": 39620
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.3929516104599202e-07,
+      "loss": 8.7522,
+      "step": 39640
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.3897691994971985e-07,
+      "loss": 8.6363,
+      "step": 39660
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.386589027991325e-07,
+      "loss": 8.6951,
+      "step": 39680
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.3834111023570556e-07,
+      "loss": 8.8011,
+      "step": 39700
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.3802354290046103e-07,
+      "loss": 8.5725,
+      "step": 39720
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.377062014339672e-07,
+      "loss": 8.7008,
+      "step": 39740
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.3738908647633634e-07,
+      "loss": 8.6939,
+      "step": 39760
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.3707219866722408e-07,
+      "loss": 8.6877,
+      "step": 39780
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.367555386458276e-07,
+      "loss": 8.634,
+      "step": 39800
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.3643910705088501e-07,
+      "loss": 8.719,
+      "step": 39820
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.3612290452067322e-07,
+      "loss": 8.6471,
+      "step": 39840
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3580693169300727e-07,
+      "loss": 8.6649,
+      "step": 39860
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3549118920523905e-07,
+      "loss": 8.7158,
+      "step": 39880
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.3517567769425548e-07,
+      "loss": 8.6799,
+      "step": 39900
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3486039779647793e-07,
+      "loss": 8.6747,
+      "step": 39920
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3454535014786023e-07,
+      "loss": 8.6028,
+      "step": 39940
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.3423053538388808e-07,
+      "loss": 8.6774,
+      "step": 39960
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.3391595413957717e-07,
+      "loss": 8.6907,
+      "step": 39980
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3360160704947221e-07,
+      "loss": 8.6227,
+      "step": 40000
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3328749474764577e-07,
+      "loss": 8.6221,
+      "step": 40020
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.329736178676965e-07,
+      "loss": 8.6608,
+      "step": 40040
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.3265997704274856e-07,
+      "loss": 8.7676,
+      "step": 40060
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.323465729054497e-07,
+      "loss": 8.6608,
+      "step": 40080
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.3203340608797016e-07,
+      "loss": 8.7215,
+      "step": 40100
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.317204772220019e-07,
+      "loss": 8.6288,
+      "step": 40120
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3140778693875637e-07,
+      "loss": 8.5937,
+      "step": 40140
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3109533586896432e-07,
+      "loss": 8.6781,
+      "step": 40160
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.3078312464287354e-07,
+      "loss": 8.6522,
+      "step": 40180
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.304711538902481e-07,
+      "loss": 8.5842,
+      "step": 40200
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.301594242403673e-07,
+      "loss": 8.695,
+      "step": 40220
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.2984793632202373e-07,
+      "loss": 8.6941,
+      "step": 40240
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.2953669076352274e-07,
+      "loss": 8.6796,
+      "step": 40260
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.2922568819268054e-07,
+      "loss": 8.7716,
+      "step": 40280
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.2891492923682323e-07,
+      "loss": 8.6689,
+      "step": 40300
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.2860441452278574e-07,
+      "loss": 8.4937,
+      "step": 40320
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.2829414467691e-07,
+      "loss": 8.695,
+      "step": 40340
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.2798412032504437e-07,
+      "loss": 8.6764,
+      "step": 40360
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.276743420925418e-07,
+      "loss": 8.5718,
+      "step": 40380
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.273648106042587e-07,
+      "loss": 8.7333,
+      "step": 40400
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.270555264845541e-07,
+      "loss": 8.6268,
+      "step": 40420
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.2674649035728768e-07,
+      "loss": 8.8738,
+      "step": 40440
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.264377028458193e-07,
+      "loss": 8.6499,
+      "step": 40460
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.2612916457300687e-07,
+      "loss": 8.667,
+      "step": 40480
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.2582087616120607e-07,
+      "loss": 8.816,
+      "step": 40500
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.2551283823226812e-07,
+      "loss": 8.6488,
+      "step": 40520
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.2520505140753917e-07,
+      "loss": 8.6484,
+      "step": 40540
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.2489751630785905e-07,
+      "loss": 8.6768,
+      "step": 40560
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.2459023355355946e-07,
+      "loss": 8.7502,
+      "step": 40580
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.2428320376446348e-07,
+      "loss": 8.6466,
+      "step": 40600
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.2397642755988368e-07,
+      "loss": 8.6937,
+      "step": 40620
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.2366990555862106e-07,
+      "loss": 8.7133,
+      "step": 40640
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.233636383789642e-07,
+      "loss": 8.6059,
+      "step": 40660
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.2305762663868728e-07,
+      "loss": 8.5764,
+      "step": 40680
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.2275187095504962e-07,
+      "loss": 8.8081,
+      "step": 40700
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.2244637194479376e-07,
+      "loss": 8.6386,
+      "step": 40720
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.2214113022414446e-07,
+      "loss": 8.762,
+      "step": 40740
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.2183614640880783e-07,
+      "loss": 8.6669,
+      "step": 40760
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.2153142111396937e-07,
+      "loss": 8.6712,
+      "step": 40780
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.2122695495429347e-07,
+      "loss": 8.719,
+      "step": 40800
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.2092274854392156e-07,
+      "loss": 8.7279,
+      "step": 40820
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.2061880249647113e-07,
+      "loss": 8.6048,
+      "step": 40840
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.2031511742503478e-07,
+      "loss": 8.7292,
+      "step": 40860
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.2001169394217825e-07,
+      "loss": 8.6321,
+      "step": 40880
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.1970853265994007e-07,
+      "loss": 8.6249,
+      "step": 40900
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.1940563418982959e-07,
+      "loss": 8.5311,
+      "step": 40920
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.1910299914282601e-07,
+      "loss": 8.8256,
+      "step": 40940
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.1880062812937753e-07,
+      "loss": 8.9261,
+      "step": 40960
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.1849852175939928e-07,
+      "loss": 8.6666,
+      "step": 40980
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.1819668064227303e-07,
+      "loss": 8.6694,
+      "step": 41000
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.1789510538684522e-07,
+      "loss": 8.7188,
+      "step": 41020
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.1759379660142597e-07,
+      "loss": 8.6856,
+      "step": 41040
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.1729275489378826e-07,
+      "loss": 8.5903,
+      "step": 41060
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.1699198087116588e-07,
+      "loss": 8.6546,
+      "step": 41080
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.166914751402531e-07,
+      "loss": 8.6771,
+      "step": 41100
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.1639123830720265e-07,
+      "loss": 8.7304,
+      "step": 41120
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.160912709776252e-07,
+      "loss": 8.707,
+      "step": 41140
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.1579157375658755e-07,
+      "loss": 8.6712,
+      "step": 41160
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.1549214724861168e-07,
+      "loss": 8.7817,
+      "step": 41180
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.151929920576737e-07,
+      "loss": 8.7075,
+      "step": 41200
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.1489410878720216e-07,
+      "loss": 8.7929,
+      "step": 41220
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.1459549804007748e-07,
+      "loss": 8.6006,
+      "step": 41240
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.1429716041863008e-07,
+      "loss": 8.7001,
+      "step": 41260
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.1399909652463943e-07,
+      "loss": 8.6809,
+      "step": 41280
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.1370130695933316e-07,
+      "loss": 8.6832,
+      "step": 41300
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.134037923233852e-07,
+      "loss": 8.5958,
+      "step": 41320
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.1310655321691525e-07,
+      "loss": 8.754,
+      "step": 41340
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.1280959023948692e-07,
+      "loss": 8.7372,
+      "step": 41360
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.1251290399010712e-07,
+      "loss": 8.6279,
+      "step": 41380
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.122164950672243e-07,
+      "loss": 8.6381,
+      "step": 41400
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.1192036406872781e-07,
+      "loss": 8.6657,
+      "step": 41420
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.1162451159194614e-07,
+      "loss": 8.6235,
+      "step": 41440
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.1132893823364594e-07,
+      "loss": 8.684,
+      "step": 41460
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.1103364459003126e-07,
+      "loss": 8.651,
+      "step": 41480
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.1073863125674135e-07,
+      "loss": 8.7139,
+      "step": 41500
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.1044389882885058e-07,
+      "loss": 8.6635,
+      "step": 41520
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.1014944790086631e-07,
+      "loss": 8.6469,
+      "step": 41540
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.0985527906672834e-07,
+      "loss": 8.6706,
+      "step": 41560
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.0956139291980726e-07,
+      "loss": 8.6693,
+      "step": 41580
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.0926779005290365e-07,
+      "loss": 8.6948,
+      "step": 41600
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.0897447105824645e-07,
+      "loss": 8.6314,
+      "step": 41620
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.0868143652749228e-07,
+      "loss": 8.6982,
+      "step": 41640
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.0838868705172377e-07,
+      "loss": 8.6151,
+      "step": 41660
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.0809622322144843e-07,
+      "loss": 8.6785,
+      "step": 41680
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.07804045626598e-07,
+      "loss": 8.8313,
+      "step": 41700
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.0751215485652643e-07,
+      "loss": 8.7213,
+      "step": 41720
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.0722055150000947e-07,
+      "loss": 8.6961,
+      "step": 41740
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.0692923614524279e-07,
+      "loss": 8.73,
+      "step": 41760
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.0663820937984147e-07,
+      "loss": 8.7056,
+      "step": 41780
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.063474717908382e-07,
+      "loss": 8.6131,
+      "step": 41800
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.0605702396468238e-07,
+      "loss": 8.6164,
+      "step": 41820
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.0576686648723923e-07,
+      "loss": 8.6729,
+      "step": 41840
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.0547699994378786e-07,
+      "loss": 8.7336,
+      "step": 41860
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.0518742491902097e-07,
+      "loss": 8.7729,
+      "step": 41880
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.0489814199704292e-07,
+      "loss": 8.7094,
+      "step": 41900
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.0460915176136892e-07,
+      "loss": 8.7918,
+      "step": 41920
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.0432045479492399e-07,
+      "loss": 8.6686,
+      "step": 41940
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.0403205168004132e-07,
+      "loss": 8.5945,
+      "step": 41960
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.0374394299846168e-07,
+      "loss": 8.6095,
+      "step": 41980
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.0345612933133166e-07,
+      "loss": 8.687,
+      "step": 42000
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.0316861125920281e-07,
+      "loss": 8.6749,
+      "step": 42020
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.0288138936203067e-07,
+      "loss": 8.6988,
+      "step": 42040
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.02594464219173e-07,
+      "loss": 8.645,
+      "step": 42060
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.0230783640938936e-07,
+      "loss": 8.6744,
+      "step": 42080
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.0202150651083929e-07,
+      "loss": 8.6319,
+      "step": 42100
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.0173547510108136e-07,
+      "loss": 8.6984,
+      "step": 42120
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.0144974275707241e-07,
+      "loss": 8.6916,
+      "step": 42140
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.0116431005516557e-07,
+      "loss": 8.6657,
+      "step": 42160
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.0087917757111e-07,
+      "loss": 8.6635,
+      "step": 42180
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.0059434588004897e-07,
+      "loss": 8.759,
+      "step": 42200
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.0030981555651908e-07,
+      "loss": 8.6307,
+      "step": 42220
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.0002558717444922e-07,
+      "loss": 8.6877,
+      "step": 42240
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 9.974166130715894e-08,
+      "loss": 8.6019,
+      "step": 42260
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 9.945803852735793e-08,
+      "loss": 8.6728,
+      "step": 42280
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 9.917471940714412e-08,
+      "loss": 8.7323,
+      "step": 42300
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 9.889170451800332e-08,
+      "loss": 8.6797,
+      "step": 42320
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 9.860899443080741e-08,
+      "loss": 8.6671,
+      "step": 42340
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 9.832658971581346e-08,
+      "loss": 8.6717,
+      "step": 42360
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 9.804449094266279e-08,
+      "loss": 8.6571,
+      "step": 42380
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 9.776269868037928e-08,
+      "loss": 8.7102,
+      "step": 42400
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 9.748121349736891e-08,
+      "loss": 8.6032,
+      "step": 42420
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 9.720003596141796e-08,
+      "loss": 8.6771,
+      "step": 42440
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 9.691916663969214e-08,
+      "loss": 8.6988,
+      "step": 42460
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 9.663860609873575e-08,
+      "loss": 8.6573,
+      "step": 42480
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 9.635835490446992e-08,
+      "loss": 8.7768,
+      "step": 42500
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 9.607841362219207e-08,
+      "loss": 8.6382,
+      "step": 42520
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 9.579878281657428e-08,
+      "loss": 8.6746,
+      "step": 42540
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 9.551946305166233e-08,
+      "loss": 8.7247,
+      "step": 42560
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 9.524045489087493e-08,
+      "loss": 8.6537,
+      "step": 42580
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 9.496175889700184e-08,
+      "loss": 8.69,
+      "step": 42600
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 9.46833756322035e-08,
+      "loss": 8.6793,
+      "step": 42620
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 9.440530565800927e-08,
+      "loss": 8.6738,
+      "step": 42640
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 9.412754953531663e-08,
+      "loss": 8.6723,
+      "step": 42660
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 9.385010782439018e-08,
+      "loss": 8.6072,
+      "step": 42680
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 9.357298108486003e-08,
+      "loss": 8.6054,
+      "step": 42700
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 9.329616987572122e-08,
+      "loss": 8.6386,
+      "step": 42720
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 9.301967475533215e-08,
+      "loss": 8.7196,
+      "step": 42740
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 9.274349628141359e-08,
+      "loss": 8.7016,
+      "step": 42760
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 9.246763501104793e-08,
+      "loss": 8.657,
+      "step": 42780
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 9.219209150067725e-08,
+      "loss": 8.8457,
+      "step": 42800
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 9.191686630610313e-08,
+      "loss": 8.591,
+      "step": 42820
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 9.164195998248469e-08,
+      "loss": 8.5781,
+      "step": 42840
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 9.13673730843382e-08,
+      "loss": 8.5978,
+      "step": 42860
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 9.109310616553534e-08,
+      "loss": 8.64,
+      "step": 42880
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 9.08191597793024e-08,
+      "loss": 8.6924,
+      "step": 42900
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 9.05455344782193e-08,
+      "loss": 8.5971,
+      "step": 42920
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 9.027223081421806e-08,
+      "loss": 8.6123,
+      "step": 42940
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 8.999924933858219e-08,
+      "loss": 8.5796,
+      "step": 42960
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 8.972659060194504e-08,
+      "loss": 8.6404,
+      "step": 42980
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 8.945425515428904e-08,
+      "loss": 8.7294,
+      "step": 43000
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 8.918224354494466e-08,
+      "loss": 8.7276,
+      "step": 43020
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 8.891055632258892e-08,
+      "loss": 8.6615,
+      "step": 43040
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 8.863919403524478e-08,
+      "loss": 8.6805,
+      "step": 43060
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 8.836815723027957e-08,
+      "loss": 8.6901,
+      "step": 43080
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 8.809744645440403e-08,
+      "loss": 8.6524,
+      "step": 43100
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 8.78270622536716e-08,
+      "loss": 8.9045,
+      "step": 43120
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 8.75570051734765e-08,
+      "loss": 8.7131,
+      "step": 43140
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 8.728727575855363e-08,
+      "loss": 8.7249,
+      "step": 43160
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 8.701787455297646e-08,
+      "loss": 8.6845,
+      "step": 43180
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 8.67488021001569e-08,
+      "loss": 8.6726,
+      "step": 43200
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 8.648005894284324e-08,
+      "loss": 8.7113,
+      "step": 43220
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 8.621164562312003e-08,
+      "loss": 8.6752,
+      "step": 43240
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 8.594356268240616e-08,
+      "loss": 8.6732,
+      "step": 43260
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 8.567581066145413e-08,
+      "loss": 8.686,
+      "step": 43280
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 8.54083901003492e-08,
+      "loss": 8.7233,
+      "step": 43300
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 8.514130153850768e-08,
+      "loss": 8.7432,
+      "step": 43320
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 8.487454551467657e-08,
+      "loss": 8.7446,
+      "step": 43340
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 8.460812256693178e-08,
+      "loss": 8.9778,
+      "step": 43360
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 8.434203323267764e-08,
+      "loss": 8.7359,
+      "step": 43380
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 8.407627804864523e-08,
+      "loss": 8.7208,
+      "step": 43400
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 8.381085755089201e-08,
+      "loss": 8.5855,
+      "step": 43420
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 8.354577227479995e-08,
+      "loss": 8.6439,
+      "step": 43440
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 8.328102275507518e-08,
+      "loss": 8.7329,
+      "step": 43460
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 8.301660952574633e-08,
+      "loss": 8.6073,
+      "step": 43480
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 8.27525331201637e-08,
+      "loss": 8.7215,
+      "step": 43500
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 8.24887940709984e-08,
+      "loss": 8.6948,
+      "step": 43520
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 8.222539291024077e-08,
+      "loss": 8.7347,
+      "step": 43540
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 8.19623301691999e-08,
+      "loss": 8.6756,
+      "step": 43560
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 8.169960637850192e-08,
+      "loss": 8.6516,
+      "step": 43580
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 8.143722206808959e-08,
+      "loss": 8.6615,
+      "step": 43600
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 8.117517776722066e-08,
+      "loss": 8.6372,
+      "step": 43620
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 8.091347400446702e-08,
+      "loss": 8.6507,
+      "step": 43640
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 8.065211130771393e-08,
+      "loss": 8.7741,
+      "step": 43660
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 8.039109020415838e-08,
+      "loss": 8.5572,
+      "step": 43680
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 8.013041122030856e-08,
+      "loss": 8.6496,
+      "step": 43700
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 7.987007488198244e-08,
+      "loss": 8.6822,
+      "step": 43720
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 7.961008171430677e-08,
+      "loss": 8.723,
+      "step": 43740
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 7.935043224171631e-08,
+      "loss": 8.6827,
+      "step": 43760
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 7.909112698795231e-08,
+      "loss": 8.6871,
+      "step": 43780
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 7.883216647606192e-08,
+      "loss": 8.6921,
+      "step": 43800
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 7.857355122839673e-08,
+      "loss": 8.7701,
+      "step": 43820
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 7.831528176661189e-08,
+      "loss": 8.5903,
+      "step": 43840
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 7.805735861166527e-08,
+      "loss": 8.6362,
+      "step": 43860
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 7.77997822838159e-08,
+      "loss": 8.6959,
+      "step": 43880
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 7.754255330262353e-08,
+      "loss": 8.6671,
+      "step": 43900
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 7.728567218694706e-08,
+      "loss": 8.7909,
+      "step": 43920
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 7.702913945494368e-08,
+      "loss": 8.6083,
+      "step": 43940
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 7.677295562406812e-08,
+      "loss": 8.639,
+      "step": 43960
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 7.651712121107101e-08,
+      "loss": 8.8751,
+      "step": 43980
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 7.626163673199848e-08,
+      "loss": 8.6571,
+      "step": 44000
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 7.600650270219044e-08,
+      "loss": 8.8169,
+      "step": 44020
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 7.57517196362803e-08,
+      "loss": 8.6811,
+      "step": 44040
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 7.549728804819325e-08,
+      "loss": 8.6199,
+      "step": 44060
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 7.524320845114557e-08,
+      "loss": 8.6188,
+      "step": 44080
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 7.498948135764368e-08,
+      "loss": 8.6039,
+      "step": 44100
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 7.473610727948271e-08,
+      "loss": 8.6675,
+      "step": 44120
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 7.448308672774605e-08,
+      "loss": 8.6623,
+      "step": 44140
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 7.423042021280369e-08,
+      "loss": 8.8081,
+      "step": 44160
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 7.397810824431155e-08,
+      "loss": 8.6835,
+      "step": 44180
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 7.372615133121057e-08,
+      "loss": 8.6718,
+      "step": 44200
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 7.347454998172522e-08,
+      "loss": 8.6558,
+      "step": 44220
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 7.322330470336313e-08,
+      "loss": 8.6661,
+      "step": 44240
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 7.297241600291334e-08,
+      "loss": 8.6431,
+      "step": 44260
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 7.272188438644574e-08,
+      "loss": 8.6932,
+      "step": 44280
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 7.24717103593101e-08,
+      "loss": 8.6602,
+      "step": 44300
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 7.222189442613464e-08,
+      "loss": 8.6308,
+      "step": 44320
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 7.197243709082554e-08,
+      "loss": 8.6254,
+      "step": 44340
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 7.172333885656542e-08,
+      "loss": 8.6904,
+      "step": 44360
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 7.147460022581255e-08,
+      "loss": 8.6725,
+      "step": 44380
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 7.122622170030016e-08,
+      "loss": 8.763,
+      "step": 44400
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 7.097820378103464e-08,
+      "loss": 8.6677,
+      "step": 44420
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 7.073054696829545e-08,
+      "loss": 8.6222,
+      "step": 44440
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 7.04832517616333e-08,
+      "loss": 8.7767,
+      "step": 44460
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 7.023631865986965e-08,
+      "loss": 8.7476,
+      "step": 44480
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 6.998974816109565e-08,
+      "loss": 8.7494,
+      "step": 44500
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.97435407626708e-08,
+      "loss": 8.7064,
+      "step": 44520
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.949769696122249e-08,
+      "loss": 8.7511,
+      "step": 44540
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.925221725264436e-08,
+      "loss": 8.6708,
+      "step": 44560
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 6.900710213209596e-08,
+      "loss": 8.6829,
+      "step": 44580
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 6.876235209400123e-08,
+      "loss": 8.5974,
+      "step": 44600
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.851796763204765e-08,
+      "loss": 8.612,
+      "step": 44620
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 6.827394923918553e-08,
+      "loss": 8.5953,
+      "step": 44640
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 6.803029740762648e-08,
+      "loss": 8.6415,
+      "step": 44660
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 6.778701262884304e-08,
+      "loss": 8.6311,
+      "step": 44680
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.75440953935671e-08,
+      "loss": 8.708,
+      "step": 44700
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 6.730154619178918e-08,
+      "loss": 8.6652,
+      "step": 44720
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 6.705936551275773e-08,
+      "loss": 8.6269,
+      "step": 44740
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 6.681755384497748e-08,
+      "loss": 8.7657,
+      "step": 44760
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 6.657611167620908e-08,
+      "loss": 8.709,
+      "step": 44780
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 6.633503949346775e-08,
+      "loss": 8.6648,
+      "step": 44800
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 6.609433778302234e-08,
+      "loss": 8.6801,
+      "step": 44820
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 6.585400703039465e-08,
+      "loss": 8.7038,
+      "step": 44840
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 6.561404772035792e-08,
+      "loss": 8.7411,
+      "step": 44860
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 6.537446033693645e-08,
+      "loss": 8.6095,
+      "step": 44880
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 6.513524536340412e-08,
+      "loss": 8.7588,
+      "step": 44900
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 6.489640328228354e-08,
+      "loss": 8.6095,
+      "step": 44920
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 6.465793457534552e-08,
+      "loss": 8.6027,
+      "step": 44940
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 6.441983972360729e-08,
+      "loss": 8.5969,
+      "step": 44960
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 6.418211920733235e-08,
+      "loss": 8.6493,
+      "step": 44980
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 6.39447735060288e-08,
+      "loss": 8.6815,
+      "step": 45000
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 6.370780309844906e-08,
+      "loss": 8.6133,
+      "step": 45020
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 6.347120846258818e-08,
+      "loss": 8.743,
+      "step": 45040
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 6.323499007568336e-08,
+      "loss": 8.6719,
+      "step": 45060
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 6.299914841421309e-08,
+      "loss": 8.7164,
+      "step": 45080
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 6.276368395389556e-08,
+      "loss": 8.5924,
+      "step": 45100
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 6.25285971696885e-08,
+      "loss": 8.7207,
+      "step": 45120
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 6.229388853578748e-08,
+      "loss": 8.691,
+      "step": 45140
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 6.20595585256256e-08,
+      "loss": 8.6939,
+      "step": 45160
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 6.1825607611872e-08,
+      "loss": 8.7609,
+      "step": 45180
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 6.159203626643137e-08,
+      "loss": 8.6598,
+      "step": 45200
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 6.135884496044244e-08,
+      "loss": 8.6806,
+      "step": 45220
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 6.112603416427776e-08,
+      "loss": 8.6477,
+      "step": 45240
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 6.089360434754203e-08,
+      "loss": 8.7188,
+      "step": 45260
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 6.066155597907157e-08,
+      "loss": 8.6378,
+      "step": 45280
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 6.04298895269334e-08,
+      "loss": 8.6899,
+      "step": 45300
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 6.019860545842392e-08,
+      "loss": 8.6343,
+      "step": 45320
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.996770424006856e-08,
+      "loss": 8.741,
+      "step": 45340
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.973718633762015e-08,
+      "loss": 8.6815,
+      "step": 45360
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 5.950705221605859e-08,
+      "loss": 8.6599,
+      "step": 45380
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.927730233958947e-08,
+      "loss": 8.692,
+      "step": 45400
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.9047937171643494e-08,
+      "loss": 8.7429,
+      "step": 45420
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.881895717487523e-08,
+      "loss": 8.7445,
+      "step": 45440
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.8590362811162254e-08,
+      "loss": 8.6593,
+      "step": 45460
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.836215454160453e-08,
+      "loss": 8.8061,
+      "step": 45480
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.813433282652297e-08,
+      "loss": 8.6217,
+      "step": 45500
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 5.7906898125458984e-08,
+      "loss": 8.7685,
+      "step": 45520
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.767985089717312e-08,
+      "loss": 8.7338,
+      "step": 45540
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.7453191599644405e-08,
+      "loss": 8.6323,
+      "step": 45560
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.722692069006957e-08,
+      "loss": 8.5866,
+      "step": 45580
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.700103862486158e-08,
+      "loss": 8.7144,
+      "step": 45600
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.6775545859649446e-08,
+      "loss": 8.6515,
+      "step": 45620
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.655044284927657e-08,
+      "loss": 8.7815,
+      "step": 45640
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 5.632573004780031e-08,
+      "loss": 8.6392,
+      "step": 45660
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.610140790849108e-08,
+      "loss": 8.6613,
+      "step": 45680
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.587747688383099e-08,
+      "loss": 8.7389,
+      "step": 45700
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.5653937425513526e-08,
+      "loss": 8.6548,
+      "step": 45720
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.5430789984442064e-08,
+      "loss": 8.661,
+      "step": 45740
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 5.520803501072954e-08,
+      "loss": 8.7147,
+      "step": 45760
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.4985672953696995e-08,
+      "loss": 8.7411,
+      "step": 45780
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.4763704261872906e-08,
+      "loss": 8.6135,
+      "step": 45800
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.454212938299255e-08,
+      "loss": 8.7409,
+      "step": 45820
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.432094876399654e-08,
+      "loss": 8.6923,
+      "step": 45840
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.4100162851030447e-08,
+      "loss": 8.5929,
+      "step": 45860
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.387977208944355e-08,
+      "loss": 8.644,
+      "step": 45880
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 5.3659776923788036e-08,
+      "loss": 8.648,
+      "step": 45900
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.344017779781834e-08,
+      "loss": 8.594,
+      "step": 45920
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.32209751544897e-08,
+      "loss": 8.7064,
+      "step": 45940
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.3002169435958e-08,
+      "loss": 8.7087,
+      "step": 45960
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.278376108357818e-08,
+      "loss": 8.8258,
+      "step": 45980
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.2565750537903716e-08,
+      "loss": 8.758,
+      "step": 46000
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.2348138238685835e-08,
+      "loss": 8.6735,
+      "step": 46020
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 5.213092462487215e-08,
+      "loss": 8.6791,
+      "step": 46040
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.1914110134606445e-08,
+      "loss": 8.6668,
+      "step": 46060
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.1697695205227126e-08,
+      "loss": 8.6343,
+      "step": 46080
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.1481680273266713e-08,
+      "loss": 8.6874,
+      "step": 46100
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.1266065774451086e-08,
+      "loss": 8.6916,
+      "step": 46120
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.105085214369806e-08,
+      "loss": 8.7685,
+      "step": 46140
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 5.0836039815117224e-08,
+      "loss": 8.7846,
+      "step": 46160
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.062162922200844e-08,
+      "loss": 8.7066,
+      "step": 46180
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.040762079686123e-08,
+      "loss": 8.6493,
+      "step": 46200
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.019401497135414e-08,
+      "loss": 8.6473,
+      "step": 46220
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 4.9980812176353274e-08,
+      "loss": 8.6748,
+      "step": 46240
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 4.9768012841912147e-08,
+      "loss": 8.7513,
+      "step": 46260
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 4.955561739727013e-08,
+      "loss": 8.7193,
+      "step": 46280
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 4.9343626270852174e-08,
+      "loss": 8.5842,
+      "step": 46300
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 4.9132039890267456e-08,
+      "loss": 8.6781,
+      "step": 46320
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 4.89208586823088e-08,
+      "loss": 8.7505,
+      "step": 46340
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 4.8710083072951904e-08,
+      "loss": 8.839,
+      "step": 46360
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 4.849971348735405e-08,
+      "loss": 8.6246,
+      "step": 46380
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 4.8289750349853834e-08,
+      "loss": 8.6899,
+      "step": 46400
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 4.8080194083969764e-08,
+      "loss": 8.7211,
+      "step": 46420
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 4.7871045112399674e-08,
+      "loss": 8.7762,
+      "step": 46440
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 4.7662303857020056e-08,
+      "loss": 8.7086,
+      "step": 46460
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 4.745397073888463e-08,
+      "loss": 8.6422,
+      "step": 46480
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 4.724604617822428e-08,
+      "loss": 8.6367,
+      "step": 46500
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 4.703853059444543e-08,
+      "loss": 8.6699,
+      "step": 46520
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 4.683142440612967e-08,
+      "loss": 8.7553,
+      "step": 46540
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.66247280310329e-08,
+      "loss": 8.6603,
+      "step": 46560
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.641844188608421e-08,
+      "loss": 8.7499,
+      "step": 46580
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.621256638738541e-08,
+      "loss": 8.6107,
+      "step": 46600
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.600710195020982e-08,
+      "loss": 8.5863,
+      "step": 46620
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.5802048989001556e-08,
+      "loss": 8.8368,
+      "step": 46640
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.559740791737504e-08,
+      "loss": 8.7823,
+      "step": 46660
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 4.539317914811353e-08,
+      "loss": 8.5468,
+      "step": 46680
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.518936309316887e-08,
+      "loss": 8.5938,
+      "step": 46700
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.498596016366027e-08,
+      "loss": 8.6411,
+      "step": 46720
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.4782970769873614e-08,
+      "loss": 8.6245,
+      "step": 46740
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.458039532126082e-08,
+      "loss": 8.5955,
+      "step": 46760
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 4.4378234226438546e-08,
+      "loss": 8.665,
+      "step": 46780
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.4176487893187956e-08,
+      "loss": 8.7599,
+      "step": 46800
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.3975156728453336e-08,
+      "loss": 8.6156,
+      "step": 46820
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.3774241138341775e-08,
+      "loss": 8.6159,
+      "step": 46840
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.35737415281219e-08,
+      "loss": 8.6734,
+      "step": 46860
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.3373658302223253e-08,
+      "loss": 8.652,
+      "step": 46880
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.317399186423573e-08,
+      "loss": 8.6976,
+      "step": 46900
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 4.29747426169082e-08,
+      "loss": 8.6995,
+      "step": 46920
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.2775910962148334e-08,
+      "loss": 8.6838,
+      "step": 46940
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.257749730102112e-08,
+      "loss": 8.7379,
+      "step": 46960
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.2379502033748764e-08,
+      "loss": 8.6209,
+      "step": 46980
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.218192555970923e-08,
+      "loss": 8.7024,
+      "step": 47000
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.198476827743597e-08,
+      "loss": 8.6773,
+      "step": 47020
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 4.1788030584616634e-08,
+      "loss": 8.6747,
+      "step": 47040
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.159171287809279e-08,
+      "loss": 8.8099,
+      "step": 47060
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.1395815553858624e-08,
+      "loss": 8.7174,
+      "step": 47080
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.120033900706041e-08,
+      "loss": 8.6592,
+      "step": 47100
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.1005283631995816e-08,
+      "loss": 8.6115,
+      "step": 47120
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.081064982211274e-08,
+      "loss": 8.8176,
+      "step": 47140
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.061643797000894e-08,
+      "loss": 8.7223,
+      "step": 47160
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 4.042264846743085e-08,
+      "loss": 8.6302,
+      "step": 47180
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.022928170527315e-08,
+      "loss": 8.6813,
+      "step": 47200
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.003633807357767e-08,
+      "loss": 8.6455,
+      "step": 47220
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 3.984381796153288e-08,
+      "loss": 8.5833,
+      "step": 47240
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 3.9651721757472835e-08,
+      "loss": 8.6188,
+      "step": 47260
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 3.946004984887652e-08,
+      "loss": 8.6727,
+      "step": 47280
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 3.9268802622367234e-08,
+      "loss": 8.6855,
+      "step": 47300
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 3.9077980463711384e-08,
+      "loss": 8.81,
+      "step": 47320
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 3.888758375781828e-08,
+      "loss": 8.761,
+      "step": 47340
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 3.869761288873882e-08,
+      "loss": 8.6576,
+      "step": 47360
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 3.850806823966491e-08,
+      "loss": 8.6554,
+      "step": 47380
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 3.831895019292897e-08,
+      "loss": 8.6861,
+      "step": 47400
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 3.813025913000265e-08,
+      "loss": 8.7043,
+      "step": 47420
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 3.7941995431496535e-08,
+      "loss": 8.7619,
+      "step": 47440
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 3.775415947715899e-08,
+      "loss": 8.726,
+      "step": 47460
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 3.7566751645875776e-08,
+      "loss": 8.6323,
+      "step": 47480
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 3.7379772315668885e-08,
+      "loss": 8.7214,
+      "step": 47500
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 3.7193221863696026e-08,
+      "loss": 8.7395,
+      "step": 47520
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 3.700710066624993e-08,
+      "loss": 8.6951,
+      "step": 47540
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 3.6821409098757336e-08,
+      "loss": 8.6347,
+      "step": 47560
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 3.6636147535778483e-08,
+      "loss": 8.6968,
+      "step": 47580
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 3.6451316351006153e-08,
+      "loss": 8.6192,
+      "step": 47600
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 3.6266915917264994e-08,
+      "loss": 8.6344,
+      "step": 47620
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 3.6082946606510956e-08,
+      "loss": 8.8091,
+      "step": 47640
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 3.589940878983008e-08,
+      "loss": 8.802,
+      "step": 47660
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 3.571630283743837e-08,
+      "loss": 8.6752,
+      "step": 47680
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 3.5533629118680436e-08,
+      "loss": 8.7103,
+      "step": 47700
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 3.535138800202911e-08,
+      "loss": 8.7011,
+      "step": 47720
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 3.5169579855084755e-08,
+      "loss": 8.7138,
+      "step": 47740
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 3.498820504457414e-08,
+      "loss": 8.6862,
+      "step": 47760
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 3.4807263936350166e-08,
+      "loss": 8.6514,
+      "step": 47780
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 3.4626756895390824e-08,
+      "loss": 8.6566,
+      "step": 47800
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 3.444668428579844e-08,
+      "loss": 8.6684,
+      "step": 47820
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 3.426704647079928e-08,
+      "loss": 8.6983,
+      "step": 47840
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 3.4087843812742354e-08,
+      "loss": 8.6276,
+      "step": 47860
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 3.390907667309906e-08,
+      "loss": 8.6916,
+      "step": 47880
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 3.373074541246224e-08,
+      "loss": 8.6749,
+      "step": 47900
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 3.355285039054545e-08,
+      "loss": 8.6663,
+      "step": 47920
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 3.3375391966182526e-08,
+      "loss": 8.626,
+      "step": 47940
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 3.3198370497326405e-08,
+      "loss": 8.7064,
+      "step": 47960
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 3.3021786341048824e-08,
+      "loss": 8.6254,
+      "step": 47980
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 3.284563985353925e-08,
+      "loss": 8.715,
+      "step": 48000
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 3.2669931390104374e-08,
+      "loss": 8.7273,
+      "step": 48020
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 3.249466130516745e-08,
+      "loss": 8.6447,
+      "step": 48040
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 3.231982995226731e-08,
+      "loss": 8.7037,
+      "step": 48060
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 3.2145437684058e-08,
+      "loss": 8.6636,
+      "step": 48080
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 3.197148485230769e-08,
+      "loss": 8.6266,
+      "step": 48100
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 3.179797180789831e-08,
+      "loss": 8.7095,
+      "step": 48120
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 3.162489890082459e-08,
+      "loss": 8.6365,
+      "step": 48140
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 3.14522664801935e-08,
+      "loss": 8.6417,
+      "step": 48160
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 3.1280074894223545e-08,
+      "loss": 8.6506,
+      "step": 48180
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 3.1108324490243864e-08,
+      "loss": 8.771,
+      "step": 48200
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 3.093701561469394e-08,
+      "loss": 8.7983,
+      "step": 48220
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 3.07661486131224e-08,
+      "loss": 8.6677,
+      "step": 48240
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 3.059572383018666e-08,
+      "loss": 8.6159,
+      "step": 48260
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 3.0425741609652166e-08,
+      "loss": 8.5923,
+      "step": 48280
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 3.0256202294391576e-08,
+      "loss": 8.6792,
+      "step": 48300
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 3.00871062263843e-08,
+      "loss": 8.6226,
+      "step": 48320
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 2.991845374671553e-08,
+      "loss": 8.6312,
+      "step": 48340
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 2.9750245195575703e-08,
+      "loss": 8.6999,
+      "step": 48360
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 2.9582480912259984e-08,
+      "loss": 8.6946,
+      "step": 48380
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 2.94151612351671e-08,
+      "loss": 8.6892,
+      "step": 48400
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 2.924828650179928e-08,
+      "loss": 8.6228,
+      "step": 48420
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 2.908185704876101e-08,
+      "loss": 8.6398,
+      "step": 48440
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 2.8915873211758645e-08,
+      "loss": 8.6749,
+      "step": 48460
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 2.875033532559984e-08,
+      "loss": 8.7345,
+      "step": 48480
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 2.8585243724192466e-08,
+      "loss": 8.6511,
+      "step": 48500
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 2.8420598740544476e-08,
+      "loss": 8.616,
+      "step": 48520
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 2.825640070676269e-08,
+      "loss": 8.8774,
+      "step": 48540
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 2.8092649954052473e-08,
+      "loss": 8.6434,
+      "step": 48560
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 2.792934681271708e-08,
+      "loss": 8.5202,
+      "step": 48580
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 2.7766491612156663e-08,
+      "loss": 8.6703,
+      "step": 48600
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 2.7604084680868112e-08,
+      "loss": 8.6421,
+      "step": 48620
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 2.74421263464438e-08,
+      "loss": 8.717,
+      "step": 48640
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 2.7280616935571516e-08,
+      "loss": 8.6583,
+      "step": 48660
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 2.7119556774033327e-08,
+      "loss": 8.7085,
+      "step": 48680
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 2.6958946186705162e-08,
+      "loss": 8.6937,
+      "step": 48700
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 2.679878549755618e-08,
+      "loss": 8.6624,
+      "step": 48720
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 2.6639075029647935e-08,
+      "loss": 8.6899,
+      "step": 48740
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 2.6479815105133974e-08,
+      "loss": 8.7277,
+      "step": 48760
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 2.632100604525886e-08,
+      "loss": 8.6808,
+      "step": 48780
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 2.616264817035793e-08,
+      "loss": 8.722,
+      "step": 48800
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 2.6004741799856207e-08,
+      "loss": 8.6183,
+      "step": 48820
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 2.5847287252268228e-08,
+      "loss": 8.6058,
+      "step": 48840
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 2.569028484519692e-08,
+      "loss": 8.6472,
+      "step": 48860
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 2.5533734895333363e-08,
+      "loss": 8.8245,
+      "step": 48880
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 2.5377637718455884e-08,
+      "loss": 8.6572,
+      "step": 48900
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 2.5221993629429505e-08,
+      "loss": 8.5695,
+      "step": 48920
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 2.5066802942205452e-08,
+      "loss": 8.7051,
+      "step": 48940
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 2.4912065969820206e-08,
+      "loss": 8.688,
+      "step": 48960
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 2.475778302439524e-08,
+      "loss": 8.732,
+      "step": 48980
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 2.4603954417136e-08,
+      "loss": 8.6216,
+      "step": 49000
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 2.445058045833173e-08,
+      "loss": 8.6969,
+      "step": 49020
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 2.4297661457354346e-08,
+      "loss": 8.8897,
+      "step": 49040
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 2.41451977226583e-08,
+      "loss": 8.552,
+      "step": 49060
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 2.3993189561779537e-08,
+      "loss": 8.6289,
+      "step": 49080
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 2.3841637281335064e-08,
+      "loss": 8.793,
+      "step": 49100
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 2.3690541187022545e-08,
+      "loss": 8.6217,
+      "step": 49120
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 2.3539901583619183e-08,
+      "loss": 8.6762,
+      "step": 49140
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 2.338971877498161e-08,
+      "loss": 8.619,
+      "step": 49160
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 2.323999306404492e-08,
+      "loss": 8.7034,
+      "step": 49180
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 2.30907247528222e-08,
+      "loss": 8.613,
+      "step": 49200
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 2.2941914142404013e-08,
+      "loss": 8.6475,
+      "step": 49220
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 2.2793561532957555e-08,
+      "loss": 8.575,
+      "step": 49240
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 2.2645667223726322e-08,
+      "loss": 8.7523,
+      "step": 49260
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 2.2498231513029236e-08,
+      "loss": 8.7378,
+      "step": 49280
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 2.2351254698260296e-08,
+      "loss": 8.5931,
+      "step": 49300
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 2.220473707588777e-08,
+      "loss": 8.663,
+      "step": 49320
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 2.205867894145366e-08,
+      "loss": 8.6425,
+      "step": 49340
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 2.191308058957328e-08,
+      "loss": 8.55,
+      "step": 49360
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 2.1767942313934334e-08,
+      "loss": 8.7109,
+      "step": 49380
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 2.1623264407296642e-08,
+      "loss": 8.6994,
+      "step": 49400
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 2.147904716149135e-08,
+      "loss": 8.6966,
+      "step": 49420
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 2.1335290867420337e-08,
+      "loss": 8.6716,
+      "step": 49440
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 2.1191995815055876e-08,
+      "loss": 8.6833,
+      "step": 49460
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 2.1049162293439587e-08,
+      "loss": 8.7627,
+      "step": 49480
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 2.0906790590682455e-08,
+      "loss": 8.6788,
+      "step": 49500
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 2.0764880993963675e-08,
+      "loss": 8.6624,
+      "step": 49520
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 2.062343378953038e-08,
+      "loss": 8.778,
+      "step": 49540
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 2.0482449262697126e-08,
+      "loss": 8.7197,
+      "step": 49560
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 2.0341927697845012e-08,
+      "loss": 8.7374,
+      "step": 49580
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 2.0201869378421497e-08,
+      "loss": 8.6552,
+      "step": 49600
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 2.006227458693946e-08,
+      "loss": 8.7903,
+      "step": 49620
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 1.9923143604976823e-08,
+      "loss": 8.6071,
+      "step": 49640
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 1.978447671317604e-08,
+      "loss": 8.6615,
+      "step": 49660
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.9646274191243318e-08,
+      "loss": 8.5126,
+      "step": 49680
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 1.9508536317948356e-08,
+      "loss": 8.6629,
+      "step": 49700
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 1.937126337112338e-08,
+      "loss": 8.6126,
+      "step": 49720
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 1.923445562766296e-08,
+      "loss": 8.9988,
+      "step": 49740
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.909811336352332e-08,
+      "loss": 8.7915,
+      "step": 49760
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 1.8962236853721586e-08,
+      "loss": 8.6122,
+      "step": 49780
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 1.882682637233568e-08,
+      "loss": 8.7854,
+      "step": 49800
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 1.86918821925032e-08,
+      "loss": 8.6883,
+      "step": 49820
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.855740458642141e-08,
+      "loss": 8.6263,
+      "step": 49840
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 1.84233938253463e-08,
+      "loss": 8.6004,
+      "step": 49860
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 1.828985017959217e-08,
+      "loss": 8.5865,
+      "step": 49880
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.815677391853124e-08,
+      "loss": 8.5881,
+      "step": 49900
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 1.8024165310592754e-08,
+      "loss": 8.6077,
+      "step": 49920
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 1.789202462326289e-08,
+      "loss": 8.6032,
+      "step": 49940
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 1.7760352123083798e-08,
+      "loss": 8.6446,
+      "step": 49960
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.7629148075653243e-08,
+      "loss": 8.6063,
+      "step": 49980
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 1.749841274562422e-08,
+      "loss": 8.6,
+      "step": 50000
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 1.7368146396704113e-08,
+      "loss": 8.5862,
+      "step": 50020
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.7238349291654435e-08,
+      "loss": 8.6713,
+      "step": 50040
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 1.7109021692290114e-08,
+      "loss": 8.5957,
+      "step": 50060
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 1.6980163859479007e-08,
+      "loss": 8.6717,
+      "step": 50080
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 1.6851776053141503e-08,
+      "loss": 8.6833,
+      "step": 50100
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.6723858532249778e-08,
+      "loss": 8.6057,
+      "step": 50120
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 1.6596411554827522e-08,
+      "loss": 8.7165,
+      "step": 50140
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 1.6469435377949175e-08,
+      "loss": 8.6832,
+      "step": 50160
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 1.634293025773953e-08,
+      "loss": 8.6771,
+      "step": 50180
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.6216896449373295e-08,
+      "loss": 8.7166,
+      "step": 50200
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 1.6091334207074398e-08,
+      "loss": 8.8346,
+      "step": 50220
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 1.596624378411565e-08,
+      "loss": 8.7444,
+      "step": 50240
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.5841625432818057e-08,
+      "loss": 8.7195,
+      "step": 50260
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 1.5717479404550455e-08,
+      "loss": 8.7715,
+      "step": 50280
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 1.5593805949728977e-08,
+      "loss": 8.7242,
+      "step": 50300
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 1.5470605317816436e-08,
+      "loss": 8.7336,
+      "step": 50320
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.5347877757322076e-08,
+      "loss": 8.7456,
+      "step": 50340
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 1.5225623515800673e-08,
+      "loss": 8.9196,
+      "step": 50360
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 1.5103842839852527e-08,
+      "loss": 8.6397,
+      "step": 50380
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 1.4982535975122474e-08,
+      "loss": 8.7685,
+      "step": 50400
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.48617031662997e-08,
+      "loss": 8.7352,
+      "step": 50420
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 1.4741344657117238e-08,
+      "loss": 8.6774,
+      "step": 50440
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 1.4621460690351334e-08,
+      "loss": 8.6874,
+      "step": 50460
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.4502051507821106e-08,
+      "loss": 8.6548,
+      "step": 50480
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 1.438311735038783e-08,
+      "loss": 8.6336,
+      "step": 50500
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 1.4264658457954743e-08,
+      "loss": 8.5954,
+      "step": 50520
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 1.4146675069466401e-08,
+      "loss": 8.6314,
+      "step": 50540
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.4029167422908105e-08,
+      "loss": 8.6163,
+      "step": 50560
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 1.3912135755305753e-08,
+      "loss": 8.756,
+      "step": 50580
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 1.3795580302724874e-08,
+      "loss": 8.7309,
+      "step": 50600
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.3679501300270652e-08,
+      "loss": 8.6124,
+      "step": 50620
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 1.3563898982087069e-08,
+      "loss": 8.7536,
+      "step": 50640
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 1.3448773581356653e-08,
+      "loss": 8.6723,
+      "step": 50660
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 1.3334125330299928e-08,
+      "loss": 8.6517,
+      "step": 50680
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.3219954460174876e-08,
+      "loss": 8.7087,
+      "step": 50700
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 1.3106261201276724e-08,
+      "loss": 8.6847,
+      "step": 50720
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 1.2993045782937084e-08,
+      "loss": 8.6975,
+      "step": 50740
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 1.2880308433523945e-08,
+      "loss": 8.6035,
+      "step": 50760
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.2768049380440765e-08,
+      "loss": 8.7271,
+      "step": 50780
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 1.2656268850126411e-08,
+      "loss": 8.7072,
+      "step": 50800
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 1.254496706805433e-08,
+      "loss": 8.806,
+      "step": 50820
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.243414425873246e-08,
+      "loss": 8.6377,
+      "step": 50840
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 1.2323800645702431e-08,
+      "loss": 8.7319,
+      "step": 50860
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 1.221393645153948e-08,
+      "loss": 8.6606,
+      "step": 50880
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 1.2104551897851644e-08,
+      "loss": 8.6349,
+      "step": 50900
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.1995647205279457e-08,
+      "loss": 8.7337,
+      "step": 50920
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 1.1887222593495699e-08,
+      "loss": 8.7268,
+      "step": 50940
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 1.1779278281204536e-08,
+      "loss": 8.7171,
+      "step": 50960
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.1671814486141546e-08,
+      "loss": 8.6429,
+      "step": 50980
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 1.156483142507289e-08,
+      "loss": 8.6306,
+      "step": 51000
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 1.1458329313795146e-08,
+      "loss": 8.7194,
+      "step": 51020
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 1.135230836713466e-08,
+      "loss": 8.6686,
+      "step": 51040
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.1246768798947287e-08,
+      "loss": 8.6497,
+      "step": 51060
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 1.1141710822117872e-08,
+      "loss": 8.6738,
+      "step": 51080
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 1.1037134648559793e-08,
+      "loss": 8.6014,
+      "step": 51100
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 1.0933040489214674e-08,
+      "loss": 8.6016,
+      "step": 51120
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.0829428554051834e-08,
+      "loss": 8.7236,
+      "step": 51140
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 1.0726299052067761e-08,
+      "loss": 8.6591,
+      "step": 51160
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 1.0623652191286026e-08,
+      "loss": 8.7178,
+      "step": 51180
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.0521488178756532e-08,
+      "loss": 8.8135,
+      "step": 51200
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 1.0419807220555271e-08,
+      "loss": 8.6145,
+      "step": 51220
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 1.0318609521783817e-08,
+      "loss": 8.6767,
+      "step": 51240
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 1.0217895286568995e-08,
+      "loss": 8.6303,
+      "step": 51260
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.0117664718062469e-08,
+      "loss": 8.5655,
+      "step": 51280
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 1.001791801844018e-08,
+      "loss": 8.6803,
+      "step": 51300
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 9.918655388902158e-09,
+      "loss": 8.6766,
+      "step": 51320
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 9.81987702967202e-09,
+      "loss": 8.6313,
+      "step": 51340
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 9.721583139996382e-09,
+      "loss": 8.7152,
+      "step": 51360
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 9.623773918144895e-09,
+      "loss": 8.6508,
+      "step": 51380
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 9.526449561409356e-09,
+      "loss": 8.742,
+      "step": 51400
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 9.429610266103699e-09,
+      "loss": 8.6364,
+      "step": 51420
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 9.333256227563341e-09,
+      "loss": 8.8212,
+      "step": 51440
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 9.237387640144867e-09,
+      "loss": 8.6756,
+      "step": 51460
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 9.14200469722573e-09,
+      "loss": 8.5969,
+      "step": 51480
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 9.047107591203723e-09,
+      "loss": 8.6711,
+      "step": 51500
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 8.952696513496755e-09,
+      "loss": 8.6413,
+      "step": 51520
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 8.858771654542185e-09,
+      "loss": 8.6178,
+      "step": 51540
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 8.76533320379677e-09,
+      "loss": 8.7109,
+      "step": 51560
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 8.672381349736108e-09,
+      "loss": 8.858,
+      "step": 51580
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 8.57991627985416e-09,
+      "loss": 8.745,
+      "step": 51600
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 8.48793818066315e-09,
+      "loss": 8.6337,
+      "step": 51620
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 8.396447237692921e-09,
+      "loss": 8.7487,
+      "step": 51640
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 8.305443635490712e-09,
+      "loss": 8.5949,
+      "step": 51660
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 8.214927557620766e-09,
+      "loss": 8.6928,
+      "step": 51680
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 8.124899186663815e-09,
+      "loss": 8.7684,
+      "step": 51700
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 8.035358704217039e-09,
+      "loss": 8.6277,
+      "step": 51720
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 7.94630629089324e-09,
+      "loss": 8.6413,
+      "step": 51740
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 7.85774212632101e-09,
+      "loss": 8.6884,
+      "step": 51760
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 7.769666389143864e-09,
+      "loss": 8.7835,
+      "step": 51780
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 7.682079257020163e-09,
+      "loss": 8.6372,
+      "step": 51800
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 7.594980906622805e-09,
+      "loss": 8.6274,
+      "step": 51820
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 7.508371513638618e-09,
+      "loss": 8.6954,
+      "step": 51840
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 7.422251252768269e-09,
+      "loss": 8.6183,
+      "step": 51860
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 7.336620297725666e-09,
+      "loss": 8.6687,
+      "step": 51880
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 7.251478821237833e-09,
+      "loss": 8.5802,
+      "step": 51900
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 7.1668269950444784e-09,
+      "loss": 8.6163,
+      "step": 51920
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 7.0826649898974856e-09,
+      "loss": 8.6173,
+      "step": 51940
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 6.998992975560919e-09,
+      "loss": 8.6898,
+      "step": 51960
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 6.915811120810355e-09,
+      "loss": 8.7128,
+      "step": 51980
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 6.833119593432607e-09,
+      "loss": 8.6544,
+      "step": 52000
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 6.750918560225583e-09,
+      "loss": 8.6416,
+      "step": 52020
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 6.66920818699776e-09,
+      "loss": 8.7659,
+      "step": 52040
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 6.587988638567881e-09,
+      "loss": 8.6386,
+      "step": 52060
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 6.50726007876462e-09,
+      "loss": 8.7111,
+      "step": 52080
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 6.427022670426329e-09,
+      "loss": 8.6797,
+      "step": 52100
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 6.347276575400628e-09,
+      "loss": 8.664,
+      "step": 52120
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 6.268021954544095e-09,
+      "loss": 8.6207,
+      "step": 52140
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 6.189258967721989e-09,
+      "loss": 8.6631,
+      "step": 52160
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 6.110987773807835e-09,
+      "loss": 8.7244,
+      "step": 52180
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 6.033208530683204e-09,
+      "loss": 8.6522,
+      "step": 52200
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 5.955921395237318e-09,
+      "loss": 8.698,
+      "step": 52220
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 5.879126523366751e-09,
+      "loss": 8.6421,
+      "step": 52240
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 5.802824069975176e-09,
+      "loss": 8.6436,
+      "step": 52260
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 5.727014188972979e-09,
+      "loss": 8.6355,
+      "step": 52280
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 5.651697033277003e-09,
+      "loss": 8.6781,
+      "step": 52300
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 5.576872754810113e-09,
+      "loss": 8.7829,
+      "step": 52320
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 5.5025415045011066e-09,
+      "loss": 8.9203,
+      "step": 52340
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 5.428703432284243e-09,
+      "loss": 8.6602,
+      "step": 52360
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 5.355358687098938e-09,
+      "loss": 8.7053,
+      "step": 52380
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 5.282507416889625e-09,
+      "loss": 8.7101,
+      "step": 52400
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 5.210149768605177e-09,
+      "loss": 8.7651,
+      "step": 52420
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 5.138285888199007e-09,
+      "loss": 8.7112,
+      "step": 52440
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 5.066915920628301e-09,
+      "loss": 8.7172,
+      "step": 52460
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 4.996040009854152e-09,
+      "loss": 8.7541,
+      "step": 52480
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 4.925658298840979e-09,
+      "loss": 8.7614,
+      "step": 52500
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 4.855770929556385e-09,
+      "loss": 8.807,
+      "step": 52520
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 4.786378042970884e-09,
+      "loss": 8.6,
+      "step": 52540
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 4.7174797790574264e-09,
+      "loss": 8.7356,
+      "step": 52560
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 4.649076276791425e-09,
+      "loss": 8.6023,
+      "step": 52580
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 4.5811676741501496e-09,
+      "loss": 8.6362,
+      "step": 52600
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 4.513754108112722e-09,
+      "loss": 8.748,
+      "step": 52620
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 4.446835714659647e-09,
+      "loss": 8.679,
+      "step": 52640
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 4.380412628772645e-09,
+      "loss": 8.6705,
+      "step": 52660
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 4.314484984434319e-09,
+      "loss": 8.6436,
+      "step": 52680
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 4.249052914627988e-09,
+      "loss": 8.6,
+      "step": 52700
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 4.184116551337241e-09,
+      "loss": 8.9462,
+      "step": 52720
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 4.119676025545777e-09,
+      "loss": 8.7455,
+      "step": 52740
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 4.055731467237283e-09,
+      "loss": 8.6577,
+      "step": 52760
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 3.992283005394837e-09,
+      "loss": 8.6922,
+      "step": 52780
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 3.929330768000949e-09,
+      "loss": 8.6823,
+      "step": 52800
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 3.866874882037157e-09,
+      "loss": 8.7113,
+      "step": 52820
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 3.8049154734838275e-09,
+      "loss": 8.6948,
+      "step": 52840
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 3.743452667319846e-09,
+      "loss": 8.6673,
+      "step": 52860
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 3.6824865875224043e-09,
+      "loss": 8.7135,
+      "step": 52880
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 3.6220173570667424e-09,
+      "loss": 8.6514,
+      "step": 52900
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 3.562045097925903e-09,
+      "loss": 8.7106,
+      "step": 52920
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 3.502569931070509e-09,
+      "loss": 8.6637,
+      "step": 52940
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 3.4435919764684572e-09,
+      "loss": 8.6529,
+      "step": 52960
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 3.385111353084724e-09,
+      "loss": 8.6605,
+      "step": 52980
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 3.3271281788811444e-09,
+      "loss": 8.7399,
+      "step": 53000
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 3.26964257081605e-09,
+      "loss": 8.6828,
+      "step": 53020
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 3.2126546448442704e-09,
+      "loss": 8.7259,
+      "step": 53040
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 3.1561645159166596e-09,
+      "loss": 8.7069,
+      "step": 53060
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 3.1001722979799306e-09,
+      "loss": 8.6755,
+      "step": 53080
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 3.0446781039765725e-09,
+      "loss": 8.637,
+      "step": 53100
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 2.989682045844405e-09,
+      "loss": 8.6292,
+      "step": 53120
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 2.935184234516497e-09,
+      "loss": 8.7152,
+      "step": 53140
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 2.8811847799208868e-09,
+      "loss": 8.6,
+      "step": 53160
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 2.827683790980362e-09,
+      "loss": 8.6343,
+      "step": 53180
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 2.774681375612292e-09,
+      "loss": 8.6721,
+      "step": 53200
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 2.722177640728324e-09,
+      "loss": 8.7229,
+      "step": 53220
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 2.6701726922342126e-09,
+      "loss": 8.5951,
+      "step": 53240
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 2.6186666350296594e-09,
+      "loss": 8.6535,
+      "step": 53260
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 2.5676595730079742e-09,
+      "loss": 8.6983,
+      "step": 53280
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 2.517151609055995e-09,
+      "loss": 8.5833,
+      "step": 53300
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 2.4671428450537824e-09,
+      "loss": 8.6913,
+      "step": 53320
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 2.417633381874534e-09,
+      "loss": 8.759,
+      "step": 53340
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 2.3686233193841722e-09,
+      "loss": 8.6451,
+      "step": 53360
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 2.3201127564414223e-09,
+      "loss": 8.6675,
+      "step": 53380
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 2.272101790897346e-09,
+      "loss": 8.7818,
+      "step": 53400
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 2.224590519595282e-09,
+      "loss": 8.5935,
+      "step": 53420
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 2.177579038370736e-09,
+      "loss": 8.6203,
+      "step": 53440
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 2.13106744205091e-09,
+      "loss": 8.6421,
+      "step": 53460
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 2.085055824454868e-09,
+      "loss": 8.6868,
+      "step": 53480
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 2.039544278393007e-09,
+      "loss": 8.7173,
+      "step": 53500
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.994532895667117e-09,
+      "loss": 8.7867,
+      "step": 53520
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.95002176707007e-09,
+      "loss": 8.6087,
+      "step": 53540
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.9060109823856583e-09,
+      "loss": 8.7634,
+      "step": 53560
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.8625006303884527e-09,
+      "loss": 8.5889,
+      "step": 53580
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.8194907988436093e-09,
+      "loss": 8.5872,
+      "step": 53600
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.7769815745066474e-09,
+      "loss": 8.629,
+      "step": 53620
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.7349730431233111e-09,
+      "loss": 8.6705,
+      "step": 53640
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.6934652894294022e-09,
+      "loss": 8.7446,
+      "step": 53660
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.6524583971505857e-09,
+      "loss": 8.5836,
+      "step": 53680
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.6119524490022795e-09,
+      "loss": 8.6287,
+      "step": 53700
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.5719475266893489e-09,
+      "loss": 8.5938,
+      "step": 53720
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.5324437109061616e-09,
+      "loss": 8.6851,
+      "step": 53740
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.493441081336172e-09,
+      "loss": 8.7353,
+      "step": 53760
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.454939716651976e-09,
+      "loss": 8.7456,
+      "step": 53780
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.4169396945150346e-09,
+      "loss": 8.5975,
+      "step": 53800
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.3794410915755339e-09,
+      "loss": 8.6356,
+      "step": 53820
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.3424439834722746e-09,
+      "loss": 8.6342,
+      "step": 53840
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.3059484448324221e-09,
+      "loss": 8.6636,
+      "step": 53860
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.269954549271507e-09,
+      "loss": 8.7235,
+      "step": 53880
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2344623693931467e-09,
+      "loss": 8.7035,
+      "step": 53900
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.199471976788935e-09,
+      "loss": 8.6849,
+      "step": 53920
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.1649834420383032e-09,
+      "loss": 8.6341,
+      "step": 53940
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.1309968347084364e-09,
+      "loss": 8.7291,
+      "step": 53960
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.0975122233539968e-09,
+      "loss": 8.5852,
+      "step": 53980
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.0645296755171229e-09,
+      "loss": 8.6188,
+      "step": 54000
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.0320492577272077e-09,
+      "loss": 8.7887,
+      "step": 54020
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.0000710355008157e-09,
+      "loss": 8.9253,
+      "step": 54040
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 9.685950733414882e-10,
+      "loss": 8.5889,
+      "step": 54060
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 9.376214347397437e-10,
+      "loss": 8.7428,
+      "step": 54080
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 9.071501821727167e-10,
+      "loss": 8.7254,
+      "step": 54100
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 8.771813771042968e-10,
+      "loss": 8.6894,
+      "step": 54120
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 8.47715079984851e-10,
+      "loss": 8.7289,
+      "step": 54140
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 8.187513502510846e-10,
+      "loss": 8.6529,
+      "step": 54160
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 7.902902463260419e-10,
+      "loss": 8.6643,
+      "step": 54180
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 7.62331825618856e-10,
+      "loss": 8.6873,
+      "step": 54200
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 7.348761445247209e-10,
+      "loss": 8.6534,
+      "step": 54220
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 7.079232584247252e-10,
+      "loss": 8.5672,
+      "step": 54240
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.814732216858243e-10,
+      "loss": 8.6725,
+      "step": 54260
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 6.555260876606183e-10,
+      "loss": 8.751,
+      "step": 54280
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 6.300819086873243e-10,
+      "loss": 8.6916,
+      "step": 54300
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 6.051407360895822e-10,
+      "loss": 8.6952,
+      "step": 54320
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 5.807026201765098e-10,
+      "loss": 8.6289,
+      "step": 54340
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.567676102424534e-10,
+      "loss": 8.7357,
+      "step": 54360
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.333357545669325e-10,
+      "loss": 8.5972,
+      "step": 54380
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.10407100414556e-10,
+      "loss": 8.6977,
+      "step": 54400
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 4.87981694034939e-10,
+      "loss": 8.6525,
+      "step": 54420
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 4.660595806625645e-10,
+      "loss": 8.7391,
+      "step": 54440
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.446408045167549e-10,
+      "loss": 8.7213,
+      "step": 54460
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 4.2372540880147854e-10,
+      "loss": 8.6799,
+      "step": 54480
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.0331343570540466e-10,
+      "loss": 8.6531,
+      "step": 54500
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 3.834049264017092e-10,
+      "loss": 8.7059,
+      "step": 54520
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 3.6399992104804713e-10,
+      "loss": 8.6514,
+      "step": 54540
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 3.450984587863859e-10,
+      "loss": 8.5528,
+      "step": 54560
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 3.267005777430887e-10,
+      "loss": 8.944,
+      "step": 54580
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 3.088063150286924e-10,
+      "loss": 8.6987,
+      "step": 54600
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 2.91415706737852e-10,
+      "loss": 8.6739,
+      "step": 54620
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 2.74528787949313e-10,
+      "loss": 8.702,
+      "step": 54640
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 2.5814559272588353e-10,
+      "loss": 8.7295,
+      "step": 54660
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 2.4226615411424013e-10,
+      "loss": 8.6574,
+      "step": 54680
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 2.268905041449276e-10,
+      "loss": 8.6305,
+      "step": 54700
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 2.1201867383233153e-10,
+      "loss": 8.6852,
+      "step": 54720
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.9765069317453918e-10,
+      "loss": 8.6515,
+      "step": 54740
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 1.8378659115333984e-10,
+      "loss": 8.7308,
+      "step": 54760
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 1.7042639573419672e-10,
+      "loss": 8.6392,
+      "step": 54780
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 1.5757013386599738e-10,
+      "loss": 8.6741,
+      "step": 54800
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.452178314813035e-10,
+      "loss": 8.7132,
+      "step": 54820
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 1.3336951349599e-10,
+      "loss": 8.6422,
+      "step": 54840
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 1.2202520380946713e-10,
+      "loss": 8.6508,
+      "step": 54860
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 1.1118492530443058e-10,
+      "loss": 8.5966,
+      "step": 54880
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 1.0084869984686162e-10,
+      "loss": 8.7772,
+      "step": 54900
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 9.101654828613803e-11,
+      "loss": 8.7014,
+      "step": 54920
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 8.168849045472881e-11,
+      "loss": 8.7369,
+      "step": 54940
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 7.286454516833296e-11,
+      "loss": 8.7147,
+      "step": 54960
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 6.454473022587947e-11,
+      "loss": 8.593,
+      "step": 54980
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 5.672906240927755e-11,
+      "loss": 8.6138,
+      "step": 55000
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 4.9417557483610875e-11,
+      "loss": 8.6584,
+      "step": 55020
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 4.261023019697108e-11,
+      "loss": 8.6232,
+      "step": 55040
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 3.630709428051326e-11,
+      "loss": 8.7377,
+      "step": 55060
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 3.050816244831722e-11,
+      "loss": 8.6403,
+      "step": 55080
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 2.5213446397470693e-11,
+      "loss": 8.7605,
+      "step": 55100
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 2.0422956808013868e-11,
+      "loss": 8.571,
+      "step": 55120
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.6136703342856107e-11,
+      "loss": 8.6668,
+      "step": 55140
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.235469464785921e-11,
+      "loss": 8.7839,
+      "step": 55160
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 9.07693835175416e-12,
+      "loss": 8.7593,
+      "step": 55180
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 6.3034410661133574e-12,
+      "loss": 8.732,
+      "step": 55200
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 4.034208385378379e-12,
+      "loss": 8.725,
+      "step": 55220
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 2.2692448868877334e-12,
+      "loss": 8.8461,
+      "step": 55240
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.0085541307103262e-12,
+      "loss": 8.6615,
+      "step": 55260
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.521386598119957e-13,
+      "loss": 8.6046,
+      "step": 55280
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.0,
+      "loss": 8.6345,
+      "step": 55300
+    },
+    {
+      "epoch": 20.0,
+      "step": 55300,
+      "total_flos": 1.644510409710981e+18,
+      "train_loss": 8.68246454138868,
+      "train_runtime": 29891.7849,
+      "train_samples_per_second": 3.7,
+      "train_steps_per_second": 1.85
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 55300,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 20,
+  "save_steps": 100,
+  "total_flos": 1.644510409710981e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}