diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,12025 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 74.79454103570761,
+  "global_step": 200001,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.04,
+      "learning_rate": 5e-06,
+      "loss": 10.2769,
+      "step": 100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1e-05,
+      "loss": 9.1431,
+      "step": 200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.5e-05,
+      "loss": 8.0625,
+      "step": 300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2e-05,
+      "loss": 7.1226,
+      "step": 400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2.5e-05,
+      "loss": 6.6198,
+      "step": 500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 3e-05,
+      "loss": 6.393,
+      "step": 600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 6.2631,
+      "step": 700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4e-05,
+      "loss": 6.1591,
+      "step": 800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 6.0731,
+      "step": 900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 5e-05,
+      "loss": 5.9878,
+      "step": 1000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.5e-05,
+      "loss": 5.9197,
+      "step": 1100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6e-05,
+      "loss": 5.8554,
+      "step": 1200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 5.7992,
+      "step": 1300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 7.000000000000001e-05,
+      "loss": 5.7512,
+      "step": 1400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 7.5e-05,
+      "loss": 5.7052,
+      "step": 1500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 8e-05,
+      "loss": 5.6635,
+      "step": 1600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.5e-05,
+      "loss": 5.6218,
+      "step": 1700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 5.5903,
+      "step": 1800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.5e-05,
+      "loss": 5.5598,
+      "step": 1900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0001,
+      "loss": 5.5297,
+      "step": 2000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.000105,
+      "loss": 5.5019,
+      "step": 2100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00011,
+      "loss": 5.4745,
+      "step": 2200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.000115,
+      "loss": 5.4513,
+      "step": 2300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.00012,
+      "loss": 5.4302,
+      "step": 2400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.000125,
+      "loss": 5.408,
+      "step": 2500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 5.3916,
+      "step": 2600
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.000135,
+      "loss": 5.4024,
+      "step": 2700
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 5.3568,
+      "step": 2800
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.000145,
+      "loss": 5.338,
+      "step": 2900
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00015,
+      "loss": 5.3311,
+      "step": 3000
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.000155,
+      "loss": 5.3137,
+      "step": 3100
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.00016,
+      "loss": 5.2984,
+      "step": 3200
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.000165,
+      "loss": 5.2857,
+      "step": 3300
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.00017,
+      "loss": 5.2744,
+      "step": 3400
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.000175,
+      "loss": 5.2631,
+      "step": 3500
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 5.2573,
+      "step": 3600
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.000185,
+      "loss": 5.2485,
+      "step": 3700
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00019,
+      "loss": 5.2358,
+      "step": 3800
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00019500000000000002,
+      "loss": 5.23,
+      "step": 3900
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002,
+      "loss": 5.2214,
+      "step": 4000
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.000205,
+      "loss": 5.2122,
+      "step": 4100
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00021,
+      "loss": 5.2077,
+      "step": 4200
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.000215,
+      "loss": 5.2,
+      "step": 4300
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.00022,
+      "loss": 5.1948,
+      "step": 4400
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 5.1821,
+      "step": 4500
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.00023,
+      "loss": 5.1826,
+      "step": 4600
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.000235,
+      "loss": 5.176,
+      "step": 4700
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00024,
+      "loss": 5.1708,
+      "step": 4800
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.000245,
+      "loss": 5.0859,
+      "step": 4900
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00025,
+      "loss": 4.9407,
+      "step": 5000
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.000255,
+      "loss": 4.8315,
+      "step": 5100
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 4.6961,
+      "step": 5200
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00026500000000000004,
+      "loss": 4.5345,
+      "step": 5300
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.00027,
+      "loss": 4.3013,
+      "step": 5400
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.000275,
+      "loss": 3.8932,
+      "step": 5500
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 3.5304,
+      "step": 5600
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.000285,
+      "loss": 3.1432,
+      "step": 5700
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00029,
+      "loss": 2.8713,
+      "step": 5800
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.000295,
+      "loss": 2.6971,
+      "step": 5900
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0003,
+      "loss": 2.5736,
+      "step": 6000
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.000305,
+      "loss": 2.4816,
+      "step": 6100
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00031,
+      "loss": 2.4073,
+      "step": 6200
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.000315,
+      "loss": 2.3448,
+      "step": 6300
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00032,
+      "loss": 2.2917,
+      "step": 6400
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 2.244,
+      "step": 6500
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00033,
+      "loss": 2.2047,
+      "step": 6600
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.000335,
+      "loss": 2.1707,
+      "step": 6700
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00034,
+      "loss": 2.1401,
+      "step": 6800
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.000345,
+      "loss": 2.1109,
+      "step": 6900
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.00035,
+      "loss": 2.079,
+      "step": 7000
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.000355,
+      "loss": 2.0593,
+      "step": 7100
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 2.0377,
+      "step": 7200
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.000365,
+      "loss": 2.0127,
+      "step": 7300
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00037,
+      "loss": 1.9964,
+      "step": 7400
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.000375,
+      "loss": 1.9722,
+      "step": 7500
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00038,
+      "loss": 1.9608,
+      "step": 7600
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00038500000000000003,
+      "loss": 1.9453,
+      "step": 7700
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 1.9231,
+      "step": 7800
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.000395,
+      "loss": 1.9114,
+      "step": 7900
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0004,
+      "loss": 1.9029,
+      "step": 8000
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00040500000000000003,
+      "loss": 1.8942,
+      "step": 8100
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.00041,
+      "loss": 1.8709,
+      "step": 8200
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.000415,
+      "loss": 1.8589,
+      "step": 8300
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00042,
+      "loss": 1.8475,
+      "step": 8400
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.000425,
+      "loss": 1.8435,
+      "step": 8500
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00043,
+      "loss": 1.833,
+      "step": 8600
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.000435,
+      "loss": 1.8249,
+      "step": 8700
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.00044,
+      "loss": 1.811,
+      "step": 8800
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00044500000000000003,
+      "loss": 1.8013,
+      "step": 8900
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 1.7965,
+      "step": 9000
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.000455,
+      "loss": 1.791,
+      "step": 9100
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.00046,
+      "loss": 1.7794,
+      "step": 9200
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.000465,
+      "loss": 1.778,
+      "step": 9300
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.00047,
+      "loss": 1.7656,
+      "step": 9400
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.000475,
+      "loss": 1.7596,
+      "step": 9500
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00048,
+      "loss": 1.7561,
+      "step": 9600
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.00048499999999999997,
+      "loss": 1.7486,
+      "step": 9700
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00049,
+      "loss": 1.7425,
+      "step": 9800
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.000495,
+      "loss": 1.7347,
+      "step": 9900
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.0005,
+      "loss": 1.7352,
+      "step": 10000
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.0004997368421052632,
+      "loss": 1.7255,
+      "step": 10100
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.0004994736842105263,
+      "loss": 1.72,
+      "step": 10200
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0004992105263157895,
+      "loss": 1.7136,
+      "step": 10300
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.0004989473684210527,
+      "loss": 1.7037,
+      "step": 10400
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0004986842105263158,
+      "loss": 1.7013,
+      "step": 10500
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.000498421052631579,
+      "loss": 1.6961,
+      "step": 10600
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.0004981578947368422,
+      "loss": 1.6953,
+      "step": 10700
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0004978947368421053,
+      "loss": 1.6793,
+      "step": 10800
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0004976315789473685,
+      "loss": 1.6719,
+      "step": 10900
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.0004973684210526315,
+      "loss": 1.672,
+      "step": 11000
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0004971052631578947,
+      "loss": 1.6683,
+      "step": 11100
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.0004968421052631579,
+      "loss": 1.6603,
+      "step": 11200
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.000496578947368421,
+      "loss": 1.655,
+      "step": 11300
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.0004963157894736842,
+      "loss": 1.6493,
+      "step": 11400
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.0004960526315789473,
+      "loss": 1.6436,
+      "step": 11500
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.0004957894736842105,
+      "loss": 1.6445,
+      "step": 11600
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.0004955263157894737,
+      "loss": 1.6418,
+      "step": 11700
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0004952631578947369,
+      "loss": 1.6302,
+      "step": 11800
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.000495,
+      "loss": 1.6286,
+      "step": 11900
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0004947368421052632,
+      "loss": 1.6257,
+      "step": 12000
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.0004944736842105264,
+      "loss": 1.6218,
+      "step": 12100
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.0004942105263157895,
+      "loss": 1.617,
+      "step": 12200
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.0004939473684210527,
+      "loss": 1.6124,
+      "step": 12300
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.0004936842105263158,
+      "loss": 1.61,
+      "step": 12400
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.000493421052631579,
+      "loss": 1.6082,
+      "step": 12500
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.0004931578947368422,
+      "loss": 1.6016,
+      "step": 12600
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.0004928947368421052,
+      "loss": 1.595,
+      "step": 12700
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.0004926315789473684,
+      "loss": 1.5895,
+      "step": 12800
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0004923684210526315,
+      "loss": 1.5889,
+      "step": 12900
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.0004921052631578947,
+      "loss": 1.5904,
+      "step": 13000
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0004918421052631579,
+      "loss": 1.5844,
+      "step": 13100
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.000491578947368421,
+      "loss": 1.5834,
+      "step": 13200
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.0004913157894736842,
+      "loss": 1.5764,
+      "step": 13300
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0004910526315789474,
+      "loss": 1.5851,
+      "step": 13400
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.0004907894736842106,
+      "loss": 1.5645,
+      "step": 13500
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0004905263157894737,
+      "loss": 1.5664,
+      "step": 13600
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.0004902631578947369,
+      "loss": 1.5612,
+      "step": 13700
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00049,
+      "loss": 1.5556,
+      "step": 13800
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.0004897368421052632,
+      "loss": 1.5569,
+      "step": 13900
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.0004894736842105264,
+      "loss": 1.5517,
+      "step": 14000
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.0004892105263157895,
+      "loss": 1.551,
+      "step": 14100
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.0004889473684210527,
+      "loss": 1.5531,
+      "step": 14200
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.0004886842105263158,
+      "loss": 1.5484,
+      "step": 14300
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.000488421052631579,
+      "loss": 1.5433,
+      "step": 14400
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.00048815789473684215,
+      "loss": 1.5421,
+      "step": 14500
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.0004878947368421053,
+      "loss": 1.537,
+      "step": 14600
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.00048763157894736843,
+      "loss": 1.5367,
+      "step": 14700
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.0004873684210526316,
+      "loss": 1.5338,
+      "step": 14800
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.0004871052631578948,
+      "loss": 1.5333,
+      "step": 14900
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.0004868421052631579,
+      "loss": 1.5317,
+      "step": 15000
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00048657894736842106,
+      "loss": 1.5269,
+      "step": 15100
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.0004863157894736842,
+      "loss": 1.5218,
+      "step": 15200
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00048605263157894735,
+      "loss": 1.5188,
+      "step": 15300
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00048578947368421054,
+      "loss": 1.5241,
+      "step": 15400
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.0004855263157894737,
+      "loss": 1.5173,
+      "step": 15500
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00048526315789473683,
+      "loss": 1.517,
+      "step": 15600
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.00048499999999999997,
+      "loss": 1.5195,
+      "step": 15700
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.00048473684210526317,
+      "loss": 1.5087,
+      "step": 15800
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00048447368421052637,
+      "loss": 1.5081,
+      "step": 15900
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.0004842105263157895,
+      "loss": 1.5087,
+      "step": 16000
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00048394736842105265,
+      "loss": 1.51,
+      "step": 16100
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.0004836842105263158,
+      "loss": 1.4973,
+      "step": 16200
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.000483421052631579,
+      "loss": 1.497,
+      "step": 16300
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.00048315789473684213,
+      "loss": 1.4946,
+      "step": 16400
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.0004828947368421053,
+      "loss": 1.4965,
+      "step": 16500
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0004826315789473684,
+      "loss": 1.4927,
+      "step": 16600
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00048236842105263156,
+      "loss": 1.4936,
+      "step": 16700
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00048210526315789476,
+      "loss": 1.4885,
+      "step": 16800
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.0004818421052631579,
+      "loss": 1.4893,
+      "step": 16900
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00048157894736842105,
+      "loss": 1.4883,
+      "step": 17000
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.0004813157894736842,
+      "loss": 1.4872,
+      "step": 17100
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.00048105263157894733,
+      "loss": 1.4842,
+      "step": 17200
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00048078947368421053,
+      "loss": 1.4841,
+      "step": 17300
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.0004805263157894737,
+      "loss": 1.4773,
+      "step": 17400
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00048026315789473687,
+      "loss": 1.4785,
+      "step": 17500
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.00048,
+      "loss": 1.4772,
+      "step": 17600
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.0004797368421052632,
+      "loss": 1.473,
+      "step": 17700
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00047947368421052635,
+      "loss": 1.4755,
+      "step": 17800
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.0004792105263157895,
+      "loss": 1.4791,
+      "step": 17900
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00047894736842105264,
+      "loss": 1.4738,
+      "step": 18000
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.0004786842105263158,
+      "loss": 1.473,
+      "step": 18100
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.000478421052631579,
+      "loss": 1.4702,
+      "step": 18200
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0004781578947368421,
+      "loss": 1.4685,
+      "step": 18300
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.00047789473684210526,
+      "loss": 1.466,
+      "step": 18400
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.0004776315789473684,
+      "loss": 1.4656,
+      "step": 18500
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.00047736842105263155,
+      "loss": 1.4647,
+      "step": 18600
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.00047710526315789475,
+      "loss": 1.4604,
+      "step": 18700
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.0004768421052631579,
+      "loss": 1.4625,
+      "step": 18800
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00047657894736842103,
+      "loss": 1.4563,
+      "step": 18900
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.0004763157894736842,
+      "loss": 1.4567,
+      "step": 19000
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00047605263157894743,
+      "loss": 1.4529,
+      "step": 19100
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.00047578947368421057,
+      "loss": 1.454,
+      "step": 19200
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.0004755263157894737,
+      "loss": 1.4524,
+      "step": 19300
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00047526315789473686,
+      "loss": 1.4492,
+      "step": 19400
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 0.000475,
+      "loss": 1.4467,
+      "step": 19500
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.0004747368421052632,
+      "loss": 1.4468,
+      "step": 19600
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00047447368421052634,
+      "loss": 1.4459,
+      "step": 19700
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.0004742105263157895,
+      "loss": 1.4444,
+      "step": 19800
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.0004739473684210526,
+      "loss": 1.442,
+      "step": 19900
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 0.00047368421052631577,
+      "loss": 1.4422,
+      "step": 20000
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 0.00047342105263157897,
+      "loss": 1.4389,
+      "step": 20100
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.0004731578947368421,
+      "loss": 1.4449,
+      "step": 20200
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00047289473684210525,
+      "loss": 1.4429,
+      "step": 20300
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.0004726315789473684,
+      "loss": 1.4364,
+      "step": 20400
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.0004723684210526316,
+      "loss": 1.4394,
+      "step": 20500
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00047210526315789473,
+      "loss": 1.4358,
+      "step": 20600
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00047184210526315793,
+      "loss": 1.4336,
+      "step": 20700
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.0004715789473684211,
+      "loss": 1.4316,
+      "step": 20800
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0004713157894736842,
+      "loss": 1.4372,
+      "step": 20900
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 0.0004710526315789474,
+      "loss": 1.4308,
+      "step": 21000
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00047078947368421056,
+      "loss": 1.4287,
+      "step": 21100
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.0004705263157894737,
+      "loss": 1.4278,
+      "step": 21200
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00047026315789473684,
+      "loss": 1.4325,
+      "step": 21300
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00047,
+      "loss": 1.4303,
+      "step": 21400
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.0004697368421052632,
+      "loss": 1.4242,
+      "step": 21500
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.0004694736842105263,
+      "loss": 1.424,
+      "step": 21600
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00046921052631578947,
+      "loss": 1.4191,
+      "step": 21700
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.0004689473684210526,
+      "loss": 1.419,
+      "step": 21800
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.0004686842105263158,
+      "loss": 1.4185,
+      "step": 21900
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00046842105263157895,
+      "loss": 1.4213,
+      "step": 22000
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.0004681578947368421,
+      "loss": 1.4203,
+      "step": 22100
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00046789473684210524,
+      "loss": 1.4147,
+      "step": 22200
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.0004676315789473684,
+      "loss": 1.4129,
+      "step": 22300
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00046736842105263163,
+      "loss": 1.4099,
+      "step": 22400
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.0004671052631578948,
+      "loss": 1.4147,
+      "step": 22500
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.0004668421052631579,
+      "loss": 1.4136,
+      "step": 22600
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00046657894736842106,
+      "loss": 1.406,
+      "step": 22700
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.0004663157894736842,
+      "loss": 1.4063,
+      "step": 22800
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0004660526315789474,
+      "loss": 1.4048,
+      "step": 22900
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.00046578947368421054,
+      "loss": 1.405,
+      "step": 23000
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.0004655263157894737,
+      "loss": 1.4049,
+      "step": 23100
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00046526315789473683,
+      "loss": 1.4077,
+      "step": 23200
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.000465,
+      "loss": 1.4017,
+      "step": 23300
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00046473684210526317,
+      "loss": 1.4077,
+      "step": 23400
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.0004644736842105263,
+      "loss": 1.4015,
+      "step": 23500
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00046421052631578946,
+      "loss": 1.4028,
+      "step": 23600
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.0004639473684210526,
+      "loss": 1.3991,
+      "step": 23700
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.0004636842105263158,
+      "loss": 1.4002,
+      "step": 23800
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00046342105263157894,
+      "loss": 1.3981,
+      "step": 23900
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.00046315789473684214,
+      "loss": 1.3977,
+      "step": 24000
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.0004628947368421053,
+      "loss": 1.4059,
+      "step": 24100
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.0004626315789473684,
+      "loss": 1.3919,
+      "step": 24200
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.0004623684210526316,
+      "loss": 1.3951,
+      "step": 24300
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00046210526315789476,
+      "loss": 1.3904,
+      "step": 24400
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.0004618421052631579,
+      "loss": 1.3904,
+      "step": 24500
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.00046157894736842105,
+      "loss": 1.3907,
+      "step": 24600
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00046131578947368425,
+      "loss": 1.3887,
+      "step": 24700
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.0004610526315789474,
+      "loss": 1.389,
+      "step": 24800
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00046078947368421053,
+      "loss": 1.3876,
+      "step": 24900
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.0004605263157894737,
+      "loss": 1.384,
+      "step": 25000
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00041611111111111113,
+      "loss": 1.3801,
+      "step": 25100
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00041555555555555557,
+      "loss": 1.3787,
+      "step": 25200
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.000415,
+      "loss": 1.3744,
+      "step": 25300
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00041444444444444444,
+      "loss": 1.3753,
+      "step": 25400
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.0004138888888888889,
+      "loss": 1.3728,
+      "step": 25500
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.0004133333333333333,
+      "loss": 1.3707,
+      "step": 25600
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.0004127777777777778,
+      "loss": 1.368,
+      "step": 25700
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.00041222222222222224,
+      "loss": 1.368,
+      "step": 25800
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.0004116666666666667,
+      "loss": 1.3685,
+      "step": 25900
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.0004111111111111111,
+      "loss": 1.3691,
+      "step": 26000
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00041055555555555555,
+      "loss": 1.3667,
+      "step": 26100
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00041,
+      "loss": 1.3653,
+      "step": 26200
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00040944444444444443,
+      "loss": 1.365,
+      "step": 26300
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.0004088888888888889,
+      "loss": 1.3606,
+      "step": 26400
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00040833333333333336,
+      "loss": 1.3617,
+      "step": 26500
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.0004077777777777778,
+      "loss": 1.3634,
+      "step": 26600
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 0.00040722222222222223,
+      "loss": 1.3618,
+      "step": 26700
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 0.00040666666666666667,
+      "loss": 1.3629,
+      "step": 26800
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.0004061111111111111,
+      "loss": 1.3543,
+      "step": 26900
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00040555555555555554,
+      "loss": 1.3543,
+      "step": 27000
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00040500000000000003,
+      "loss": 1.3501,
+      "step": 27100
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00040444444444444447,
+      "loss": 1.351,
+      "step": 27200
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 0.0004038888888888889,
+      "loss": 1.3533,
+      "step": 27300
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00040333333333333334,
+      "loss": 1.3538,
+      "step": 27400
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.0004027777777777778,
+      "loss": 1.3501,
+      "step": 27500
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.0004022222222222222,
+      "loss": 1.3455,
+      "step": 27600
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00040166666666666665,
+      "loss": 1.3508,
+      "step": 27700
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.0004011111111111111,
+      "loss": 1.3541,
+      "step": 27800
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.0004005555555555556,
+      "loss": 1.3469,
+      "step": 27900
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.0004,
+      "loss": 1.3451,
+      "step": 28000
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00039944444444444446,
+      "loss": 1.347,
+      "step": 28100
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.0003988888888888889,
+      "loss": 1.346,
+      "step": 28200
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 0.00039833333333333333,
+      "loss": 1.3448,
+      "step": 28300
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00039777777777777777,
+      "loss": 1.3432,
+      "step": 28400
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.0003972222222222222,
+      "loss": 1.3433,
+      "step": 28500
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.0003966666666666667,
+      "loss": 1.346,
+      "step": 28600
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00039611111111111113,
+      "loss": 1.3442,
+      "step": 28700
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 0.00039555555555555557,
+      "loss": 1.3413,
+      "step": 28800
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.000395,
+      "loss": 1.3413,
+      "step": 28900
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00039444444444444444,
+      "loss": 1.3389,
+      "step": 29000
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00039388888888888893,
+      "loss": 1.3406,
+      "step": 29100
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.0003933333333333333,
+      "loss": 1.3392,
+      "step": 29200
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.0003927777777777778,
+      "loss": 1.3355,
+      "step": 29300
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00039222222222222225,
+      "loss": 1.3386,
+      "step": 29400
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.0003916666666666667,
+      "loss": 1.3402,
+      "step": 29500
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.0003911111111111111,
+      "loss": 1.3317,
+      "step": 29600
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.00039055555555555556,
+      "loss": 1.3314,
+      "step": 29700
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 1.3319,
+      "step": 29800
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00038944444444444443,
+      "loss": 1.3309,
+      "step": 29900
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.0003888888888888889,
+      "loss": 1.33,
+      "step": 30000
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.0003883333333333333,
+      "loss": 1.3311,
+      "step": 30100
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.0003877777777777778,
+      "loss": 1.3316,
+      "step": 30200
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00038722222222222223,
+      "loss": 1.3262,
+      "step": 30300
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 0.00038666666666666667,
+      "loss": 1.3283,
+      "step": 30400
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.00038611111111111116,
+      "loss": 1.3257,
+      "step": 30500
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00038555555555555554,
+      "loss": 1.3267,
+      "step": 30600
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00038500000000000003,
+      "loss": 1.3248,
+      "step": 30700
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0003844444444444444,
+      "loss": 1.3224,
+      "step": 30800
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.0003838888888888889,
+      "loss": 1.3207,
+      "step": 30900
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00038333333333333334,
+      "loss": 1.3215,
+      "step": 31000
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.0003827777777777778,
+      "loss": 1.3217,
+      "step": 31100
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.0003822222222222223,
+      "loss": 1.3202,
+      "step": 31200
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00038166666666666666,
+      "loss": 1.3222,
+      "step": 31300
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.00038111111111111115,
+      "loss": 1.3199,
+      "step": 31400
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.00038055555555555553,
+      "loss": 1.3203,
+      "step": 31500
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.00038,
+      "loss": 1.3194,
+      "step": 31600
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.0003794444444444444,
+      "loss": 1.3188,
+      "step": 31700
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0003788888888888889,
+      "loss": 1.3192,
+      "step": 31800
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.0003783333333333334,
+      "loss": 1.3171,
+      "step": 31900
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00037777777777777777,
+      "loss": 1.3215,
+      "step": 32000
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00037722222222222226,
+      "loss": 1.3216,
+      "step": 32100
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.00037666666666666664,
+      "loss": 1.3123,
+      "step": 32200
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00037611111111111113,
+      "loss": 1.3093,
+      "step": 32300
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.0003755555555555555,
+      "loss": 1.3136,
+      "step": 32400
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 0.000375,
+      "loss": 1.3114,
+      "step": 32500
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.0003744444444444445,
+      "loss": 1.3089,
+      "step": 32600
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.0003738888888888889,
+      "loss": 1.3109,
+      "step": 32700
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.0003733333333333334,
+      "loss": 1.3082,
+      "step": 32800
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00037277777777777776,
+      "loss": 1.3104,
+      "step": 32900
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00037222222222222225,
+      "loss": 1.3101,
+      "step": 33000
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00037166666666666663,
+      "loss": 1.3074,
+      "step": 33100
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.0003711111111111111,
+      "loss": 1.3055,
+      "step": 33200
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.0003705555555555556,
+      "loss": 1.305,
+      "step": 33300
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00037,
+      "loss": 1.3077,
+      "step": 33400
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.0003694444444444445,
+      "loss": 1.3065,
+      "step": 33500
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00036888888888888887,
+      "loss": 1.3052,
+      "step": 33600
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00036833333333333336,
+      "loss": 1.3078,
+      "step": 33700
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00036777777777777774,
+      "loss": 1.3038,
+      "step": 33800
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00036722222222222223,
+      "loss": 1.302,
+      "step": 33900
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 0.00036666666666666667,
+      "loss": 1.3034,
+      "step": 34000
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.0003661111111111111,
+      "loss": 1.3002,
+      "step": 34100
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.0003655555555555556,
+      "loss": 1.3046,
+      "step": 34200
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.000365,
+      "loss": 1.3343,
+      "step": 34300
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00036444444444444447,
+      "loss": 1.3163,
+      "step": 34400
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00036388888888888886,
+      "loss": 1.3062,
+      "step": 34500
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 0.00036333333333333335,
+      "loss": 1.3042,
+      "step": 34600
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.0003627777777777778,
+      "loss": 1.3036,
+      "step": 34700
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.0003622222222222222,
+      "loss": 1.3137,
+      "step": 34800
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.0003616666666666667,
+      "loss": 1.2952,
+      "step": 34900
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.0003611111111111111,
+      "loss": 1.2961,
+      "step": 35000
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.0003605555555555556,
+      "loss": 1.2932,
+      "step": 35100
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 1.2936,
+      "step": 35200
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00035944444444444446,
+      "loss": 1.2966,
+      "step": 35300
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.0003588888888888889,
+      "loss": 1.2942,
+      "step": 35400
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00035833333333333333,
+      "loss": 1.2914,
+      "step": 35500
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 0.00035777777777777777,
+      "loss": 1.296,
+      "step": 35600
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.0003572222222222222,
+      "loss": 1.2945,
+      "step": 35700
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.0003566666666666667,
+      "loss": 1.2878,
+      "step": 35800
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.0003561111111111111,
+      "loss": 1.2926,
+      "step": 35900
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00035555555555555557,
+      "loss": 1.2918,
+      "step": 36000
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 0.000355,
+      "loss": 1.2888,
+      "step": 36100
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.00035444444444444445,
+      "loss": 1.2903,
+      "step": 36200
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.0003538888888888889,
+      "loss": 1.3005,
+      "step": 36300
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.0003533333333333333,
+      "loss": 1.2875,
+      "step": 36400
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.0003527777777777778,
+      "loss": 1.2918,
+      "step": 36500
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 0.00035222222222222225,
+      "loss": 1.2885,
+      "step": 36600
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.0003516666666666667,
+      "loss": 1.2856,
+      "step": 36700
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 0.0003511111111111111,
+      "loss": 1.2884,
+      "step": 36800
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 0.00035055555555555556,
+      "loss": 1.2922,
+      "step": 36900
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 0.00035,
+      "loss": 1.2875,
+      "step": 37000
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 0.00034944444444444443,
+      "loss": 1.2874,
+      "step": 37100
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 0.0003488888888888889,
+      "loss": 1.2858,
+      "step": 37200
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 0.00034833333333333336,
+      "loss": 1.2849,
+      "step": 37300
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 0.0003477777777777778,
+      "loss": 1.2803,
+      "step": 37400
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 0.00034722222222222224,
+      "loss": 1.2903,
+      "step": 37500
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 0.00034666666666666667,
+      "loss": 1.2778,
+      "step": 37600
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 0.0003461111111111111,
+      "loss": 1.2795,
+      "step": 37700
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 0.00034555555555555555,
+      "loss": 1.277,
+      "step": 37800
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.000345,
+      "loss": 1.2798,
+      "step": 37900
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 0.0003444444444444445,
+      "loss": 1.2795,
+      "step": 38000
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 0.0003438888888888889,
+      "loss": 1.2772,
+      "step": 38100
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 0.00034333333333333335,
+      "loss": 1.2754,
+      "step": 38200
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 0.0003427777777777778,
+      "loss": 1.2762,
+      "step": 38300
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 0.0003422222222222222,
+      "loss": 1.2732,
+      "step": 38400
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 0.00034166666666666666,
+      "loss": 1.2793,
+      "step": 38500
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 0.0003411111111111111,
+      "loss": 1.2751,
+      "step": 38600
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 0.0003405555555555556,
+      "loss": 1.2753,
+      "step": 38700
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 0.00034,
+      "loss": 1.2758,
+      "step": 38800
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 0.00033944444444444446,
+      "loss": 1.2734,
+      "step": 38900
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 0.0003388888888888889,
+      "loss": 1.2723,
+      "step": 39000
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 0.00033833333333333334,
+      "loss": 1.2746,
+      "step": 39100
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 0.00033777777777777777,
+      "loss": 1.275,
+      "step": 39200
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 0.0003372222222222222,
+      "loss": 1.2711,
+      "step": 39300
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 0.0003366666666666667,
+      "loss": 1.2688,
+      "step": 39400
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 0.00033611111111111114,
+      "loss": 1.2713,
+      "step": 39500
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 0.0003355555555555556,
+      "loss": 1.2725,
+      "step": 39600
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 0.000335,
+      "loss": 1.2679,
+      "step": 39700
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 0.00033444444444444445,
+      "loss": 1.2748,
+      "step": 39800
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 0.0003338888888888889,
+      "loss": 1.2708,
+      "step": 39900
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 0.0003333333333333333,
+      "loss": 1.2706,
+      "step": 40000
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 0.0003327777777777778,
+      "loss": 1.2667,
+      "step": 40100
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 0.0003322222222222222,
+      "loss": 1.2703,
+      "step": 40200
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 0.0003316666666666667,
+      "loss": 1.264,
+      "step": 40300
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 0.0003311111111111111,
+      "loss": 1.262,
+      "step": 40400
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 0.00033055555555555556,
+      "loss": 1.2645,
+      "step": 40500
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 0.00033,
+      "loss": 1.2603,
+      "step": 40600
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 0.00032944444444444444,
+      "loss": 1.2648,
+      "step": 40700
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 0.0003288888888888889,
+      "loss": 1.2644,
+      "step": 40800
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 0.0003283333333333333,
+      "loss": 1.2624,
+      "step": 40900
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 0.0003277777777777778,
+      "loss": 1.2623,
+      "step": 41000
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 0.00032722222222222224,
+      "loss": 1.2616,
+      "step": 41100
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 0.0003266666666666667,
+      "loss": 1.2652,
+      "step": 41200
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 0.0003261111111111111,
+      "loss": 1.2621,
+      "step": 41300
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 0.00032555555555555555,
+      "loss": 1.26,
+      "step": 41400
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 1.259,
+      "step": 41500
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 0.0003244444444444444,
+      "loss": 1.2648,
+      "step": 41600
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 0.0003238888888888889,
+      "loss": 1.2615,
+      "step": 41700
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 0.0003233333333333333,
+      "loss": 1.261,
+      "step": 41800
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 0.0003227777777777778,
+      "loss": 1.2573,
+      "step": 41900
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 0.0003222222222222222,
+      "loss": 1.2604,
+      "step": 42000
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 0.00032166666666666666,
+      "loss": 1.2574,
+      "step": 42100
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 0.00032111111111111115,
+      "loss": 1.255,
+      "step": 42200
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 0.00032055555555555554,
+      "loss": 1.2575,
+      "step": 42300
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.00032,
+      "loss": 1.2556,
+      "step": 42400
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 0.0003194444444444444,
+      "loss": 1.2547,
+      "step": 42500
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 0.0003188888888888889,
+      "loss": 1.2567,
+      "step": 42600
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 0.00031833333333333334,
+      "loss": 1.2552,
+      "step": 42700
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 0.0003177777777777778,
+      "loss": 1.2637,
+      "step": 42800
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 0.00031722222222222227,
+      "loss": 1.2495,
+      "step": 42900
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 0.00031666666666666665,
+      "loss": 1.2485,
+      "step": 43000
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 0.00031611111111111114,
+      "loss": 1.2513,
+      "step": 43100
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 0.0003155555555555555,
+      "loss": 1.2494,
+      "step": 43200
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 0.000315,
+      "loss": 1.253,
+      "step": 43300
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 0.0003144444444444445,
+      "loss": 1.2543,
+      "step": 43400
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 0.0003138888888888889,
+      "loss": 1.2488,
+      "step": 43500
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 0.0003133333333333334,
+      "loss": 1.2479,
+      "step": 43600
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 0.00031277777777777776,
+      "loss": 1.2481,
+      "step": 43700
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 0.00031222222222222225,
+      "loss": 1.2528,
+      "step": 43800
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 0.00031166666666666663,
+      "loss": 1.2445,
+      "step": 43900
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 0.0003111111111111111,
+      "loss": 1.2485,
+      "step": 44000
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 0.0003105555555555555,
+      "loss": 1.2495,
+      "step": 44100
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 0.00031,
+      "loss": 1.2483,
+      "step": 44200
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 0.0003094444444444445,
+      "loss": 1.2411,
+      "step": 44300
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 0.0003088888888888889,
+      "loss": 1.2498,
+      "step": 44400
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 0.00030833333333333337,
+      "loss": 1.2436,
+      "step": 44500
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 0.00030777777777777775,
+      "loss": 1.2457,
+      "step": 44600
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.00030722222222222224,
+      "loss": 1.2442,
+      "step": 44700
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 0.0003066666666666667,
+      "loss": 1.2459,
+      "step": 44800
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 0.0003061111111111111,
+      "loss": 1.2394,
+      "step": 44900
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 0.0003055555555555556,
+      "loss": 1.2418,
+      "step": 45000
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 0.000305,
+      "loss": 1.244,
+      "step": 45100
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 0.0003044444444444445,
+      "loss": 1.2428,
+      "step": 45200
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.00030388888888888886,
+      "loss": 1.2445,
+      "step": 45300
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 0.00030333333333333335,
+      "loss": 1.2447,
+      "step": 45400
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 0.0003027777777777778,
+      "loss": 1.2505,
+      "step": 45500
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 0.0003022222222222222,
+      "loss": 1.2316,
+      "step": 45600
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 0.0003016666666666667,
+      "loss": 1.2389,
+      "step": 45700
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 0.0003011111111111111,
+      "loss": 1.2364,
+      "step": 45800
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 0.0003005555555555556,
+      "loss": 1.2389,
+      "step": 45900
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 0.0003,
+      "loss": 1.2379,
+      "step": 46000
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 0.00029944444444444446,
+      "loss": 1.236,
+      "step": 46100
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 0.0002988888888888889,
+      "loss": 1.2386,
+      "step": 46200
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 0.00029833333333333334,
+      "loss": 1.2364,
+      "step": 46300
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 0.0002977777777777778,
+      "loss": 1.2365,
+      "step": 46400
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 0.0002972222222222222,
+      "loss": 1.2332,
+      "step": 46500
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 0.0002966666666666667,
+      "loss": 1.2385,
+      "step": 46600
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 0.0002961111111111111,
+      "loss": 1.2334,
+      "step": 46700
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 0.0002955555555555556,
+      "loss": 1.2361,
+      "step": 46800
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 0.000295,
+      "loss": 1.2319,
+      "step": 46900
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 0.00029444444444444445,
+      "loss": 1.2394,
+      "step": 47000
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 0.0002938888888888889,
+      "loss": 1.232,
+      "step": 47100
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 0.0002933333333333333,
+      "loss": 1.2349,
+      "step": 47200
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 0.0002927777777777778,
+      "loss": 1.2358,
+      "step": 47300
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 0.0002922222222222222,
+      "loss": 1.2308,
+      "step": 47400
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 0.0002916666666666667,
+      "loss": 1.2316,
+      "step": 47500
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 0.00029111111111111113,
+      "loss": 1.2309,
+      "step": 47600
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 0.00029055555555555556,
+      "loss": 1.2327,
+      "step": 47700
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 0.00029,
+      "loss": 1.2279,
+      "step": 47800
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 0.00028944444444444444,
+      "loss": 1.2307,
+      "step": 47900
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 0.0002888888888888889,
+      "loss": 1.23,
+      "step": 48000
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 0.0002883333333333333,
+      "loss": 1.2274,
+      "step": 48100
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 0.0002877777777777778,
+      "loss": 1.2322,
+      "step": 48200
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 0.00028722222222222224,
+      "loss": 1.2269,
+      "step": 48300
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 0.0002866666666666667,
+      "loss": 1.2264,
+      "step": 48400
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 0.0002861111111111111,
+      "loss": 1.2255,
+      "step": 48500
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 0.00028555555555555555,
+      "loss": 1.2243,
+      "step": 48600
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 0.000285,
+      "loss": 1.223,
+      "step": 48700
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 0.0002844444444444444,
+      "loss": 1.2232,
+      "step": 48800
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 0.0002838888888888889,
+      "loss": 1.2238,
+      "step": 48900
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 0.00028333333333333335,
+      "loss": 1.2239,
+      "step": 49000
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 0.0002827777777777778,
+      "loss": 1.2212,
+      "step": 49100
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 0.00028222222222222223,
+      "loss": 1.2234,
+      "step": 49200
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 0.00028166666666666666,
+      "loss": 1.2267,
+      "step": 49300
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 0.0002811111111111111,
+      "loss": 1.2196,
+      "step": 49400
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 0.00028055555555555554,
+      "loss": 1.2264,
+      "step": 49500
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 1.2256,
+      "step": 49600
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 0.00027944444444444447,
+      "loss": 1.2208,
+      "step": 49700
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 0.0002788888888888889,
+      "loss": 1.2228,
+      "step": 49800
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 0.00027833333333333334,
+      "loss": 1.2245,
+      "step": 49900
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 0.0002777777777777778,
+      "loss": 1.2251,
+      "step": 50000
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 0.0002772222222222222,
+      "loss": 1.2167,
+      "step": 50100
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 0.00027666666666666665,
+      "loss": 1.2187,
+      "step": 50200
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 0.0002761111111111111,
+      "loss": 1.2226,
+      "step": 50300
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 0.0002755555555555556,
+      "loss": 1.2198,
+      "step": 50400
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 0.000275,
+      "loss": 1.2221,
+      "step": 50500
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 0.00027444444444444445,
+      "loss": 1.219,
+      "step": 50600
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 0.0002738888888888889,
+      "loss": 1.2193,
+      "step": 50700
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 0.00027333333333333333,
+      "loss": 1.2178,
+      "step": 50800
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 0.00027277777777777776,
+      "loss": 1.2218,
+      "step": 50900
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 0.0002722222222222222,
+      "loss": 1.2159,
+      "step": 51000
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 0.0002716666666666667,
+      "loss": 1.2122,
+      "step": 51100
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 0.00027111111111111113,
+      "loss": 1.2179,
+      "step": 51200
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 0.00027055555555555557,
+      "loss": 1.2161,
+      "step": 51300
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 0.00027,
+      "loss": 1.2144,
+      "step": 51400
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 0.00026944444444444444,
+      "loss": 1.2154,
+      "step": 51500
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 0.00026888888888888893,
+      "loss": 1.2101,
+      "step": 51600
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 0.0002683333333333333,
+      "loss": 1.2083,
+      "step": 51700
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 0.0002677777777777778,
+      "loss": 1.2219,
+      "step": 51800
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 0.00026722222222222224,
+      "loss": 1.2137,
+      "step": 51900
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 0.0002666666666666667,
+      "loss": 1.2127,
+      "step": 52000
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 0.0002661111111111111,
+      "loss": 1.2112,
+      "step": 52100
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 0.00026555555555555555,
+      "loss": 1.2102,
+      "step": 52200
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 0.00026500000000000004,
+      "loss": 1.2115,
+      "step": 52300
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 0.00026444444444444443,
+      "loss": 1.2088,
+      "step": 52400
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 0.0002638888888888889,
+      "loss": 1.2135,
+      "step": 52500
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 0.0002633333333333333,
+      "loss": 1.2133,
+      "step": 52600
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 0.0002627777777777778,
+      "loss": 1.2094,
+      "step": 52700
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 0.00026222222222222223,
+      "loss": 1.2067,
+      "step": 52800
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 0.00026166666666666667,
+      "loss": 1.2097,
+      "step": 52900
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 0.00026111111111111116,
+      "loss": 1.2099,
+      "step": 53000
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 0.00026055555555555554,
+      "loss": 1.2097,
+      "step": 53100
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 1.2044,
+      "step": 53200
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 0.0002594444444444444,
+      "loss": 1.2092,
+      "step": 53300
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 0.0002588888888888889,
+      "loss": 1.2061,
+      "step": 53400
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 0.00025833333333333334,
+      "loss": 1.2142,
+      "step": 53500
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 0.0002577777777777778,
+      "loss": 1.2038,
+      "step": 53600
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 0.00025722222222222227,
+      "loss": 1.2003,
+      "step": 53700
+    },
+    {
+      "epoch": 20.12,
+      "learning_rate": 0.00025666666666666665,
+      "loss": 1.2033,
+      "step": 53800
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 0.00025611111111111114,
+      "loss": 1.2045,
+      "step": 53900
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 0.00025555555555555553,
+      "loss": 1.1993,
+      "step": 54000
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 0.000255,
+      "loss": 1.204,
+      "step": 54100
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 0.0002544444444444444,
+      "loss": 1.2039,
+      "step": 54200
+    },
+    {
+      "epoch": 20.31,
+      "learning_rate": 0.0002538888888888889,
+      "loss": 1.2004,
+      "step": 54300
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 0.0002533333333333334,
+      "loss": 1.2046,
+      "step": 54400
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 0.00025277777777777777,
+      "loss": 1.2048,
+      "step": 54500
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 0.00025222222222222226,
+      "loss": 1.1981,
+      "step": 54600
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 0.00025166666666666664,
+      "loss": 1.2021,
+      "step": 54700
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 0.00025111111111111113,
+      "loss": 1.201,
+      "step": 54800
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 0.0002505555555555555,
+      "loss": 1.2027,
+      "step": 54900
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 0.00025,
+      "loss": 1.201,
+      "step": 55000
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 0.00024944444444444444,
+      "loss": 1.1973,
+      "step": 55100
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 0.0002488888888888889,
+      "loss": 1.1996,
+      "step": 55200
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 0.0002483333333333333,
+      "loss": 1.1996,
+      "step": 55300
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 0.0002477777777777778,
+      "loss": 1.2022,
+      "step": 55400
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 0.00024722222222222224,
+      "loss": 1.1971,
+      "step": 55500
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 0.0002466666666666667,
+      "loss": 1.1992,
+      "step": 55600
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 0.0002461111111111111,
+      "loss": 1.1997,
+      "step": 55700
+    },
+    {
+      "epoch": 20.87,
+      "learning_rate": 0.00024555555555555556,
+      "loss": 1.2,
+      "step": 55800
+    },
+    {
+      "epoch": 20.9,
+      "learning_rate": 0.000245,
+      "loss": 1.1995,
+      "step": 55900
+    },
+    {
+      "epoch": 20.94,
+      "learning_rate": 0.00024444444444444443,
+      "loss": 1.1969,
+      "step": 56000
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 0.0002438888888888889,
+      "loss": 1.1957,
+      "step": 56100
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 0.00024333333333333336,
+      "loss": 1.2026,
+      "step": 56200
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 0.0002427777777777778,
+      "loss": 1.1926,
+      "step": 56300
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 0.00024222222222222223,
+      "loss": 1.1964,
+      "step": 56400
+    },
+    {
+      "epoch": 21.13,
+      "learning_rate": 0.00024166666666666667,
+      "loss": 1.1918,
+      "step": 56500
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 0.0002411111111111111,
+      "loss": 1.1913,
+      "step": 56600
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 0.00024055555555555554,
+      "loss": 1.1956,
+      "step": 56700
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 0.00024,
+      "loss": 1.1932,
+      "step": 56800
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 0.00023944444444444444,
+      "loss": 1.194,
+      "step": 56900
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 0.0002388888888888889,
+      "loss": 1.1918,
+      "step": 57000
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 0.00023833333333333334,
+      "loss": 1.1879,
+      "step": 57100
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 0.00023777777777777778,
+      "loss": 1.1894,
+      "step": 57200
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 0.00023722222222222222,
+      "loss": 1.1909,
+      "step": 57300
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 0.00023666666666666668,
+      "loss": 1.1917,
+      "step": 57400
+    },
+    {
+      "epoch": 21.5,
+      "learning_rate": 0.00023611111111111112,
+      "loss": 1.1908,
+      "step": 57500
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 0.00023555555555555556,
+      "loss": 1.1908,
+      "step": 57600
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 0.000235,
+      "loss": 1.1924,
+      "step": 57700
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 0.00023444444444444446,
+      "loss": 1.1887,
+      "step": 57800
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 0.0002338888888888889,
+      "loss": 1.1903,
+      "step": 57900
+    },
+    {
+      "epoch": 21.69,
+      "learning_rate": 0.00023333333333333333,
+      "loss": 1.1868,
+      "step": 58000
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 0.0002327777777777778,
+      "loss": 1.1889,
+      "step": 58100
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 0.00023222222222222223,
+      "loss": 1.1854,
+      "step": 58200
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 0.00023166666666666667,
+      "loss": 1.1857,
+      "step": 58300
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 0.0002311111111111111,
+      "loss": 1.1878,
+      "step": 58400
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 0.00023055555555555557,
+      "loss": 1.1873,
+      "step": 58500
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 0.00023,
+      "loss": 1.1867,
+      "step": 58600
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 0.00022944444444444444,
+      "loss": 1.1854,
+      "step": 58700
+    },
+    {
+      "epoch": 21.99,
+      "learning_rate": 0.0002288888888888889,
+      "loss": 1.1867,
+      "step": 58800
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 0.00022833333333333334,
+      "loss": 1.1906,
+      "step": 58900
+    },
+    {
+      "epoch": 22.06,
+      "learning_rate": 0.00022777777777777778,
+      "loss": 1.1831,
+      "step": 59000
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 0.00022722222222222222,
+      "loss": 1.1849,
+      "step": 59100
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 0.00022666666666666666,
+      "loss": 1.1792,
+      "step": 59200
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 0.00022611111111111112,
+      "loss": 1.1823,
+      "step": 59300
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 0.00022555555555555556,
+      "loss": 1.1839,
+      "step": 59400
+    },
+    {
+      "epoch": 22.25,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 1.184,
+      "step": 59500
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 0.00022444444444444446,
+      "loss": 1.1813,
+      "step": 59600
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 0.0002238888888888889,
+      "loss": 1.1858,
+      "step": 59700
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 0.00022333333333333333,
+      "loss": 1.1788,
+      "step": 59800
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 0.00022277777777777777,
+      "loss": 1.1813,
+      "step": 59900
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 0.0002222222222222222,
+      "loss": 1.1816,
+      "step": 60000
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 0.00022166666666666667,
+      "loss": 1.1821,
+      "step": 60100
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 0.00022111111111111113,
+      "loss": 1.1807,
+      "step": 60200
+    },
+    {
+      "epoch": 22.55,
+      "learning_rate": 0.00022055555555555557,
+      "loss": 1.1768,
+      "step": 60300
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 0.00022,
+      "loss": 1.1811,
+      "step": 60400
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 0.00021944444444444444,
+      "loss": 1.1799,
+      "step": 60500
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 0.00021888888888888888,
+      "loss": 1.1784,
+      "step": 60600
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 0.00021833333333333332,
+      "loss": 1.1788,
+      "step": 60700
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 0.00021777777777777776,
+      "loss": 1.1769,
+      "step": 60800
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 0.00021722222222222225,
+      "loss": 1.178,
+      "step": 60900
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 0.00021666666666666668,
+      "loss": 1.1784,
+      "step": 61000
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 0.00021611111111111112,
+      "loss": 1.1775,
+      "step": 61100
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 0.00021555555555555556,
+      "loss": 1.1813,
+      "step": 61200
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 0.000215,
+      "loss": 1.1793,
+      "step": 61300
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 0.00021444444444444443,
+      "loss": 1.1745,
+      "step": 61400
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 0.0002138888888888889,
+      "loss": 1.1788,
+      "step": 61500
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 0.00021333333333333336,
+      "loss": 1.1811,
+      "step": 61600
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 0.0002127777777777778,
+      "loss": 1.1697,
+      "step": 61700
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 0.00021222222222222223,
+      "loss": 1.1751,
+      "step": 61800
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 0.00021166666666666667,
+      "loss": 1.1714,
+      "step": 61900
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 0.0002111111111111111,
+      "loss": 1.1754,
+      "step": 62000
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 0.00021055555555555554,
+      "loss": 1.1711,
+      "step": 62100
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 0.00021,
+      "loss": 1.1712,
+      "step": 62200
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 0.00020944444444444445,
+      "loss": 1.1751,
+      "step": 62300
+    },
+    {
+      "epoch": 23.34,
+      "learning_rate": 0.0002088888888888889,
+      "loss": 1.1733,
+      "step": 62400
+    },
+    {
+      "epoch": 23.37,
+      "learning_rate": 0.00020833333333333335,
+      "loss": 1.1722,
+      "step": 62500
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 0.00020777777777777778,
+      "loss": 1.1707,
+      "step": 62600
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 0.00020722222222222222,
+      "loss": 1.1731,
+      "step": 62700
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 0.00020666666666666666,
+      "loss": 1.1697,
+      "step": 62800
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 0.00020611111111111112,
+      "loss": 1.171,
+      "step": 62900
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 0.00020555555555555556,
+      "loss": 1.1706,
+      "step": 63000
+    },
+    {
+      "epoch": 23.6,
+      "learning_rate": 0.000205,
+      "loss": 1.1687,
+      "step": 63100
+    },
+    {
+      "epoch": 23.63,
+      "learning_rate": 0.00020444444444444446,
+      "loss": 1.1685,
+      "step": 63200
+    },
+    {
+      "epoch": 23.67,
+      "learning_rate": 0.0002038888888888889,
+      "loss": 1.1707,
+      "step": 63300
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 0.00020333333333333333,
+      "loss": 1.1681,
+      "step": 63400
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 0.00020277777777777777,
+      "loss": 1.1708,
+      "step": 63500
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 0.00020222222222222223,
+      "loss": 1.1662,
+      "step": 63600
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 0.00020166666666666667,
+      "loss": 1.1678,
+      "step": 63700
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 0.0002011111111111111,
+      "loss": 1.17,
+      "step": 63800
+    },
+    {
+      "epoch": 23.9,
+      "learning_rate": 0.00020055555555555555,
+      "loss": 1.1663,
+      "step": 63900
+    },
+    {
+      "epoch": 23.93,
+      "learning_rate": 0.0002,
+      "loss": 1.1686,
+      "step": 64000
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 0.00019944444444444445,
+      "loss": 1.168,
+      "step": 64100
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 0.00019888888888888888,
+      "loss": 1.1734,
+      "step": 64200
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 0.00019833333333333335,
+      "loss": 1.1634,
+      "step": 64300
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 0.00019777777777777778,
+      "loss": 1.1634,
+      "step": 64400
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 0.00019722222222222222,
+      "loss": 1.164,
+      "step": 64500
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 0.00019666666666666666,
+      "loss": 1.1616,
+      "step": 64600
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 0.00019611111111111112,
+      "loss": 1.1627,
+      "step": 64700
+    },
+    {
+      "epoch": 24.23,
+      "learning_rate": 0.00019555555555555556,
+      "loss": 1.1626,
+      "step": 64800
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 0.00019500000000000002,
+      "loss": 1.1621,
+      "step": 64900
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 0.00019444444444444446,
+      "loss": 1.1624,
+      "step": 65000
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 0.0001938888888888889,
+      "loss": 1.1614,
+      "step": 65100
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 0.00019333333333333333,
+      "loss": 1.1624,
+      "step": 65200
+    },
+    {
+      "epoch": 24.42,
+      "learning_rate": 0.00019277777777777777,
+      "loss": 1.1644,
+      "step": 65300
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 0.0001922222222222222,
+      "loss": 1.1624,
+      "step": 65400
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 0.00019166666666666667,
+      "loss": 1.1629,
+      "step": 65500
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 0.00019111111111111114,
+      "loss": 1.1592,
+      "step": 65600
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 0.00019055555555555557,
+      "loss": 1.1603,
+      "step": 65700
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 0.00019,
+      "loss": 1.1595,
+      "step": 65800
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 0.00018944444444444445,
+      "loss": 1.1612,
+      "step": 65900
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 0.00018888888888888888,
+      "loss": 1.1603,
+      "step": 66000
+    },
+    {
+      "epoch": 24.72,
+      "learning_rate": 0.00018833333333333332,
+      "loss": 1.1602,
+      "step": 66100
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 0.00018777777777777776,
+      "loss": 1.1628,
+      "step": 66200
+    },
+    {
+      "epoch": 24.79,
+      "learning_rate": 0.00018722222222222225,
+      "loss": 1.1579,
+      "step": 66300
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 0.0001866666666666667,
+      "loss": 1.1594,
+      "step": 66400
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 0.00018611111111111112,
+      "loss": 1.1592,
+      "step": 66500
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 0.00018555555555555556,
+      "loss": 1.1599,
+      "step": 66600
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 0.000185,
+      "loss": 1.1574,
+      "step": 66700
+    },
+    {
+      "epoch": 24.98,
+      "learning_rate": 0.00018444444444444443,
+      "loss": 1.1575,
+      "step": 66800
+    },
+    {
+      "epoch": 25.02,
+      "learning_rate": 0.00018388888888888887,
+      "loss": 1.1605,
+      "step": 66900
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 0.00018333333333333334,
+      "loss": 1.1531,
+      "step": 67000
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 0.0001827777777777778,
+      "loss": 1.1529,
+      "step": 67100
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 0.00018222222222222224,
+      "loss": 1.1537,
+      "step": 67200
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 0.00018166666666666667,
+      "loss": 1.1524,
+      "step": 67300
+    },
+    {
+      "epoch": 25.21,
+      "learning_rate": 0.0001811111111111111,
+      "loss": 1.1539,
+      "step": 67400
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 0.00018055555555555555,
+      "loss": 1.1549,
+      "step": 67500
+    },
+    {
+      "epoch": 25.28,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 1.1574,
+      "step": 67600
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 0.00017944444444444445,
+      "loss": 1.1538,
+      "step": 67700
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 0.00017888888888888889,
+      "loss": 1.1521,
+      "step": 67800
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 0.00017833333333333335,
+      "loss": 1.1518,
+      "step": 67900
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 1.1547,
+      "step": 68000
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 0.00017722222222222222,
+      "loss": 1.1524,
+      "step": 68100
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 0.00017666666666666666,
+      "loss": 1.1514,
+      "step": 68200
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 0.00017611111111111112,
+      "loss": 1.1512,
+      "step": 68300
+    },
+    {
+      "epoch": 25.58,
+      "learning_rate": 0.00017555555555555556,
+      "loss": 1.1508,
+      "step": 68400
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 0.000175,
+      "loss": 1.1526,
+      "step": 68500
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 0.00017444444444444446,
+      "loss": 1.147,
+      "step": 68600
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 0.0001738888888888889,
+      "loss": 1.1497,
+      "step": 68700
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 0.00017333333333333334,
+      "loss": 1.1527,
+      "step": 68800
+    },
+    {
+      "epoch": 25.77,
+      "learning_rate": 0.00017277777777777777,
+      "loss": 1.1492,
+      "step": 68900
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 0.00017222222222222224,
+      "loss": 1.1521,
+      "step": 69000
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 0.00017166666666666667,
+      "loss": 1.1517,
+      "step": 69100
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 0.0001711111111111111,
+      "loss": 1.147,
+      "step": 69200
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 0.00017055555555555555,
+      "loss": 1.1506,
+      "step": 69300
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 0.00017,
+      "loss": 1.1496,
+      "step": 69400
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 0.00016944444444444445,
+      "loss": 1.1478,
+      "step": 69500
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 0.00016888888888888889,
+      "loss": 1.1513,
+      "step": 69600
+    },
+    {
+      "epoch": 26.07,
+      "learning_rate": 0.00016833333333333335,
+      "loss": 1.1459,
+      "step": 69700
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 0.0001677777777777778,
+      "loss": 1.1443,
+      "step": 69800
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 0.00016722222222222222,
+      "loss": 1.1453,
+      "step": 69900
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 0.00016666666666666666,
+      "loss": 1.1438,
+      "step": 70000
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 0.0001661111111111111,
+      "loss": 1.1429,
+      "step": 70100
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 0.00016555555555555556,
+      "loss": 1.1465,
+      "step": 70200
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 0.000165,
+      "loss": 1.1436,
+      "step": 70300
+    },
+    {
+      "epoch": 26.33,
+      "learning_rate": 0.00016444444444444446,
+      "loss": 1.1424,
+      "step": 70400
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 0.0001638888888888889,
+      "loss": 1.1434,
+      "step": 70500
+    },
+    {
+      "epoch": 26.4,
+      "learning_rate": 0.00016333333333333334,
+      "loss": 1.1441,
+      "step": 70600
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 0.00016277777777777777,
+      "loss": 1.1443,
+      "step": 70700
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 0.0001622222222222222,
+      "loss": 1.1446,
+      "step": 70800
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 0.00016166666666666665,
+      "loss": 1.144,
+      "step": 70900
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 0.0001611111111111111,
+      "loss": 1.1429,
+      "step": 71000
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 0.00016055555555555558,
+      "loss": 1.1435,
+      "step": 71100
+    },
+    {
+      "epoch": 26.63,
+      "learning_rate": 0.00016,
+      "loss": 1.1425,
+      "step": 71200
+    },
+    {
+      "epoch": 26.66,
+      "learning_rate": 0.00015944444444444445,
+      "loss": 1.1417,
+      "step": 71300
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 0.0001588888888888889,
+      "loss": 1.1432,
+      "step": 71400
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 0.00015833333333333332,
+      "loss": 1.1405,
+      "step": 71500
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 0.00015777777777777776,
+      "loss": 1.1425,
+      "step": 71600
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 0.00015722222222222225,
+      "loss": 1.1438,
+      "step": 71700
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 0.0001566666666666667,
+      "loss": 1.1402,
+      "step": 71800
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 0.00015611111111111113,
+      "loss": 1.1424,
+      "step": 71900
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 1.1396,
+      "step": 72000
+    },
+    {
+      "epoch": 26.96,
+      "learning_rate": 0.000155,
+      "loss": 1.138,
+      "step": 72100
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 0.00015444444444444444,
+      "loss": 1.1486,
+      "step": 72200
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 0.00015388888888888887,
+      "loss": 1.1363,
+      "step": 72300
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 0.00015333333333333334,
+      "loss": 1.1342,
+      "step": 72400
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 0.0001527777777777778,
+      "loss": 1.1333,
+      "step": 72500
+    },
+    {
+      "epoch": 27.15,
+      "learning_rate": 0.00015222222222222224,
+      "loss": 1.1354,
+      "step": 72600
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 0.00015166666666666668,
+      "loss": 1.1376,
+      "step": 72700
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 0.0001511111111111111,
+      "loss": 1.1336,
+      "step": 72800
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 0.00015055555555555555,
+      "loss": 1.1364,
+      "step": 72900
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 0.00015,
+      "loss": 1.137,
+      "step": 73000
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 0.00014944444444444445,
+      "loss": 1.1346,
+      "step": 73100
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 0.0001488888888888889,
+      "loss": 1.1353,
+      "step": 73200
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 0.00014833333333333335,
+      "loss": 1.1372,
+      "step": 73300
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 0.0001477777777777778,
+      "loss": 1.1362,
+      "step": 73400
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 0.00014722222222222223,
+      "loss": 1.1354,
+      "step": 73500
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 0.00014666666666666666,
+      "loss": 1.1357,
+      "step": 73600
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 0.0001461111111111111,
+      "loss": 1.1349,
+      "step": 73700
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 0.00014555555555555556,
+      "loss": 1.1386,
+      "step": 73800
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 0.000145,
+      "loss": 1.1308,
+      "step": 73900
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 0.00014444444444444444,
+      "loss": 1.1322,
+      "step": 74000
+    },
+    {
+      "epoch": 27.71,
+      "learning_rate": 0.0001438888888888889,
+      "loss": 1.1363,
+      "step": 74100
+    },
+    {
+      "epoch": 27.75,
+      "learning_rate": 0.00014333333333333334,
+      "loss": 1.1335,
+      "step": 74200
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 0.00014277777777777778,
+      "loss": 1.1334,
+      "step": 74300
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 0.0001422222222222222,
+      "loss": 1.1361,
+      "step": 74400
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 0.00014166666666666668,
+      "loss": 1.1352,
+      "step": 74500
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 0.00014111111111111111,
+      "loss": 1.1318,
+      "step": 74600
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 0.00014055555555555555,
+      "loss": 1.1346,
+      "step": 74700
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 1.1355,
+      "step": 74800
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 0.00013944444444444445,
+      "loss": 1.1346,
+      "step": 74900
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 0.0001388888888888889,
+      "loss": 1.1273,
+      "step": 75000
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 0.00013833333333333333,
+      "loss": 1.1258,
+      "step": 75100
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 0.0001377777777777778,
+      "loss": 1.1285,
+      "step": 75200
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 0.00013722222222222223,
+      "loss": 1.127,
+      "step": 75300
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 0.00013666666666666666,
+      "loss": 1.1305,
+      "step": 75400
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 0.0001361111111111111,
+      "loss": 1.1296,
+      "step": 75500
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 0.00013555555555555556,
+      "loss": 1.127,
+      "step": 75600
+    },
+    {
+      "epoch": 28.31,
+      "learning_rate": 0.000135,
+      "loss": 1.1279,
+      "step": 75700
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 0.00013444444444444447,
+      "loss": 1.1275,
+      "step": 75800
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 0.0001338888888888889,
+      "loss": 1.1277,
+      "step": 75900
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 1.1275,
+      "step": 76000
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 0.00013277777777777778,
+      "loss": 1.1282,
+      "step": 76100
+    },
+    {
+      "epoch": 28.5,
+      "learning_rate": 0.00013222222222222221,
+      "loss": 1.126,
+      "step": 76200
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 0.00013166666666666665,
+      "loss": 1.1297,
+      "step": 76300
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 0.00013111111111111111,
+      "loss": 1.1275,
+      "step": 76400
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 0.00013055555555555558,
+      "loss": 1.1299,
+      "step": 76500
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 1.1285,
+      "step": 76600
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 0.00012944444444444445,
+      "loss": 1.1254,
+      "step": 76700
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 0.0001288888888888889,
+      "loss": 1.1227,
+      "step": 76800
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 0.00012833333333333333,
+      "loss": 1.1278,
+      "step": 76900
+    },
+    {
+      "epoch": 28.8,
+      "learning_rate": 0.00012777777777777776,
+      "loss": 1.1211,
+      "step": 77000
+    },
+    {
+      "epoch": 28.83,
+      "learning_rate": 0.0001272222222222222,
+      "loss": 1.1255,
+      "step": 77100
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 0.0001266666666666667,
+      "loss": 1.1229,
+      "step": 77200
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 0.00012611111111111113,
+      "loss": 1.1236,
+      "step": 77300
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 0.00012555555555555557,
+      "loss": 1.1227,
+      "step": 77400
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 0.000125,
+      "loss": 1.1209,
+      "step": 77500
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 0.00012444444444444444,
+      "loss": 1.1309,
+      "step": 77600
+    },
+    {
+      "epoch": 29.06,
+      "learning_rate": 0.0001238888888888889,
+      "loss": 1.1172,
+      "step": 77700
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 0.00012333333333333334,
+      "loss": 1.1183,
+      "step": 77800
+    },
+    {
+      "epoch": 29.13,
+      "learning_rate": 0.00012277777777777778,
+      "loss": 1.1227,
+      "step": 77900
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 0.00012222222222222221,
+      "loss": 1.1187,
+      "step": 78000
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 0.00012166666666666668,
+      "loss": 1.1191,
+      "step": 78100
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 0.00012111111111111112,
+      "loss": 1.1209,
+      "step": 78200
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 0.00012055555555555555,
+      "loss": 1.1231,
+      "step": 78300
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 0.00012,
+      "loss": 1.1187,
+      "step": 78400
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 0.00011944444444444445,
+      "loss": 1.1213,
+      "step": 78500
+    },
+    {
+      "epoch": 29.39,
+      "learning_rate": 0.00011888888888888889,
+      "loss": 1.1208,
+      "step": 78600
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 0.00011833333333333334,
+      "loss": 1.1196,
+      "step": 78700
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 0.00011777777777777778,
+      "loss": 1.1206,
+      "step": 78800
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 0.00011722222222222223,
+      "loss": 1.1197,
+      "step": 78900
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 0.00011666666666666667,
+      "loss": 1.1198,
+      "step": 79000
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 0.00011611111111111112,
+      "loss": 1.1184,
+      "step": 79100
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 0.00011555555555555555,
+      "loss": 1.1181,
+      "step": 79200
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 0.000115,
+      "loss": 1.1159,
+      "step": 79300
+    },
+    {
+      "epoch": 29.69,
+      "learning_rate": 0.00011444444444444445,
+      "loss": 1.1174,
+      "step": 79400
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 0.00011388888888888889,
+      "loss": 1.1199,
+      "step": 79500
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 0.00011333333333333333,
+      "loss": 1.1163,
+      "step": 79600
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 0.00011277777777777778,
+      "loss": 1.123,
+      "step": 79700
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 0.00011222222222222223,
+      "loss": 1.1165,
+      "step": 79800
+    },
+    {
+      "epoch": 29.88,
+      "learning_rate": 0.00011166666666666667,
+      "loss": 1.1155,
+      "step": 79900
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 0.0001111111111111111,
+      "loss": 1.1125,
+      "step": 80000
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 0.00011055555555555557,
+      "loss": 1.1135,
+      "step": 80100
+    },
+    {
+      "epoch": 29.99,
+      "learning_rate": 0.00011,
+      "loss": 1.1162,
+      "step": 80200
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 0.00010944444444444444,
+      "loss": 1.1166,
+      "step": 80300
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 0.00010888888888888888,
+      "loss": 1.1105,
+      "step": 80400
+    },
+    {
+      "epoch": 30.1,
+      "learning_rate": 0.00010833333333333334,
+      "loss": 1.1109,
+      "step": 80500
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 0.00010777777777777778,
+      "loss": 1.1119,
+      "step": 80600
+    },
+    {
+      "epoch": 30.18,
+      "learning_rate": 0.00010722222222222222,
+      "loss": 1.1156,
+      "step": 80700
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 0.00010666666666666668,
+      "loss": 1.1136,
+      "step": 80800
+    },
+    {
+      "epoch": 30.25,
+      "learning_rate": 0.00010611111111111112,
+      "loss": 1.1115,
+      "step": 80900
+    },
+    {
+      "epoch": 30.29,
+      "learning_rate": 0.00010555555555555555,
+      "loss": 1.1109,
+      "step": 81000
+    },
+    {
+      "epoch": 30.33,
+      "learning_rate": 0.000105,
+      "loss": 1.1141,
+      "step": 81100
+    },
+    {
+      "epoch": 30.37,
+      "learning_rate": 0.00010444444444444445,
+      "loss": 1.1129,
+      "step": 81200
+    },
+    {
+      "epoch": 30.4,
+      "learning_rate": 0.00010388888888888889,
+      "loss": 1.1155,
+      "step": 81300
+    },
+    {
+      "epoch": 30.44,
+      "learning_rate": 0.00010333333333333333,
+      "loss": 1.1117,
+      "step": 81400
+    },
+    {
+      "epoch": 30.48,
+      "learning_rate": 0.00010277777777777778,
+      "loss": 1.1097,
+      "step": 81500
+    },
+    {
+      "epoch": 30.52,
+      "learning_rate": 0.00010222222222222223,
+      "loss": 1.1097,
+      "step": 81600
+    },
+    {
+      "epoch": 30.55,
+      "learning_rate": 0.00010166666666666667,
+      "loss": 1.1089,
+      "step": 81700
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 0.00010111111111111112,
+      "loss": 1.1124,
+      "step": 81800
+    },
+    {
+      "epoch": 30.63,
+      "learning_rate": 0.00010055555555555555,
+      "loss": 1.1077,
+      "step": 81900
+    },
+    {
+      "epoch": 30.67,
+      "learning_rate": 0.0001,
+      "loss": 1.1085,
+      "step": 82000
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 9.944444444444444e-05,
+      "loss": 1.1097,
+      "step": 82100
+    },
+    {
+      "epoch": 30.74,
+      "learning_rate": 9.888888888888889e-05,
+      "loss": 1.112,
+      "step": 82200
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 9.833333333333333e-05,
+      "loss": 1.1099,
+      "step": 82300
+    },
+    {
+      "epoch": 30.82,
+      "learning_rate": 9.777777777777778e-05,
+      "loss": 1.1106,
+      "step": 82400
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 9.722222222222223e-05,
+      "loss": 1.1099,
+      "step": 82500
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 9.666666666666667e-05,
+      "loss": 1.1086,
+      "step": 82600
+    },
+    {
+      "epoch": 30.93,
+      "learning_rate": 9.61111111111111e-05,
+      "loss": 1.1101,
+      "step": 82700
+    },
+    {
+      "epoch": 30.96,
+      "learning_rate": 9.555555555555557e-05,
+      "loss": 1.1089,
+      "step": 82800
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 9.5e-05,
+      "loss": 1.1116,
+      "step": 82900
+    },
+    {
+      "epoch": 31.04,
+      "learning_rate": 9.444444444444444e-05,
+      "loss": 1.1062,
+      "step": 83000
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 9.388888888888888e-05,
+      "loss": 1.1082,
+      "step": 83100
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 9.333333333333334e-05,
+      "loss": 1.104,
+      "step": 83200
+    },
+    {
+      "epoch": 31.15,
+      "learning_rate": 9.277777777777778e-05,
+      "loss": 1.1039,
+      "step": 83300
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 9.222222222222222e-05,
+      "loss": 1.1027,
+      "step": 83400
+    },
+    {
+      "epoch": 31.23,
+      "learning_rate": 9.166666666666667e-05,
+      "loss": 1.1049,
+      "step": 83500
+    },
+    {
+      "epoch": 31.26,
+      "learning_rate": 9.111111111111112e-05,
+      "loss": 1.1089,
+      "step": 83600
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 9.055555555555556e-05,
+      "loss": 1.1082,
+      "step": 83700
+    },
+    {
+      "epoch": 31.34,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 1.1019,
+      "step": 83800
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 8.944444444444444e-05,
+      "loss": 1.1034,
+      "step": 83900
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 8.888888888888889e-05,
+      "loss": 1.1034,
+      "step": 84000
+    },
+    {
+      "epoch": 31.45,
+      "learning_rate": 8.833333333333333e-05,
+      "loss": 1.1002,
+      "step": 84100
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 8.777777777777778e-05,
+      "loss": 1.1032,
+      "step": 84200
+    },
+    {
+      "epoch": 31.53,
+      "learning_rate": 8.722222222222223e-05,
+      "loss": 1.1081,
+      "step": 84300
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 8.666666666666667e-05,
+      "loss": 1.0995,
+      "step": 84400
+    },
+    {
+      "epoch": 31.6,
+      "learning_rate": 8.611111111111112e-05,
+      "loss": 1.104,
+      "step": 84500
+    },
+    {
+      "epoch": 31.64,
+      "learning_rate": 8.555555555555556e-05,
+      "loss": 1.1015,
+      "step": 84600
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 8.5e-05,
+      "loss": 1.1022,
+      "step": 84700
+    },
+    {
+      "epoch": 31.71,
+      "learning_rate": 8.444444444444444e-05,
+      "loss": 1.1016,
+      "step": 84800
+    },
+    {
+      "epoch": 31.75,
+      "learning_rate": 8.38888888888889e-05,
+      "loss": 1.1008,
+      "step": 84900
+    },
+    {
+      "epoch": 31.79,
+      "learning_rate": 8.333333333333333e-05,
+      "loss": 1.102,
+      "step": 85000
+    },
+    {
+      "epoch": 31.82,
+      "learning_rate": 8.277777777777778e-05,
+      "loss": 1.0997,
+      "step": 85100
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 8.222222222222223e-05,
+      "loss": 1.1036,
+      "step": 85200
+    },
+    {
+      "epoch": 31.9,
+      "learning_rate": 8.166666666666667e-05,
+      "loss": 1.1021,
+      "step": 85300
+    },
+    {
+      "epoch": 31.94,
+      "learning_rate": 8.11111111111111e-05,
+      "loss": 1.0999,
+      "step": 85400
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 8.055555555555556e-05,
+      "loss": 1.0988,
+      "step": 85500
+    },
+    {
+      "epoch": 32.01,
+      "learning_rate": 8e-05,
+      "loss": 1.1106,
+      "step": 85600
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 7.944444444444444e-05,
+      "loss": 1.1002,
+      "step": 85700
+    },
+    {
+      "epoch": 32.09,
+      "learning_rate": 7.888888888888888e-05,
+      "loss": 1.0966,
+      "step": 85800
+    },
+    {
+      "epoch": 32.12,
+      "learning_rate": 7.833333333333334e-05,
+      "loss": 1.0942,
+      "step": 85900
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 7.777777777777778e-05,
+      "loss": 1.0986,
+      "step": 86000
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 7.722222222222222e-05,
+      "loss": 1.0963,
+      "step": 86100
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 7.666666666666667e-05,
+      "loss": 1.0951,
+      "step": 86200
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 7.611111111111112e-05,
+      "loss": 1.0958,
+      "step": 86300
+    },
+    {
+      "epoch": 32.31,
+      "learning_rate": 7.555555555555556e-05,
+      "loss": 1.095,
+      "step": 86400
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 7.5e-05,
+      "loss": 1.1004,
+      "step": 86500
+    },
+    {
+      "epoch": 32.39,
+      "learning_rate": 7.444444444444444e-05,
+      "loss": 1.0977,
+      "step": 86600
+    },
+    {
+      "epoch": 32.42,
+      "learning_rate": 7.38888888888889e-05,
+      "loss": 1.0984,
+      "step": 86700
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 7.333333333333333e-05,
+      "loss": 1.1005,
+      "step": 86800
+    },
+    {
+      "epoch": 32.5,
+      "learning_rate": 7.277777777777778e-05,
+      "loss": 1.0948,
+      "step": 86900
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 7.222222222222222e-05,
+      "loss": 1.0953,
+      "step": 87000
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 7.166666666666667e-05,
+      "loss": 1.0967,
+      "step": 87100
+    },
+    {
+      "epoch": 32.61,
+      "learning_rate": 7.11111111111111e-05,
+      "loss": 1.0986,
+      "step": 87200
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 7.055555555555556e-05,
+      "loss": 1.0938,
+      "step": 87300
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 7.000000000000001e-05,
+      "loss": 1.0934,
+      "step": 87400
+    },
+    {
+      "epoch": 32.72,
+      "learning_rate": 6.944444444444444e-05,
+      "loss": 1.0925,
+      "step": 87500
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 6.88888888888889e-05,
+      "loss": 1.0951,
+      "step": 87600
+    },
+    {
+      "epoch": 32.8,
+      "learning_rate": 6.833333333333333e-05,
+      "loss": 1.0926,
+      "step": 87700
+    },
+    {
+      "epoch": 32.83,
+      "learning_rate": 6.777777777777778e-05,
+      "loss": 1.0941,
+      "step": 87800
+    },
+    {
+      "epoch": 32.87,
+      "learning_rate": 6.722222222222223e-05,
+      "loss": 1.0952,
+      "step": 87900
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 1.0938,
+      "step": 88000
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 6.611111111111111e-05,
+      "loss": 1.0968,
+      "step": 88100
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 6.555555555555556e-05,
+      "loss": 1.0935,
+      "step": 88200
+    },
+    {
+      "epoch": 33.02,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 1.0984,
+      "step": 88300
+    },
+    {
+      "epoch": 33.06,
+      "learning_rate": 6.444444444444444e-05,
+      "loss": 1.0893,
+      "step": 88400
+    },
+    {
+      "epoch": 33.1,
+      "learning_rate": 6.388888888888888e-05,
+      "loss": 1.0907,
+      "step": 88500
+    },
+    {
+      "epoch": 33.13,
+      "learning_rate": 6.333333333333335e-05,
+      "loss": 1.0879,
+      "step": 88600
+    },
+    {
+      "epoch": 33.17,
+      "learning_rate": 6.277777777777778e-05,
+      "loss": 1.0905,
+      "step": 88700
+    },
+    {
+      "epoch": 33.21,
+      "learning_rate": 6.222222222222222e-05,
+      "loss": 1.0884,
+      "step": 88800
+    },
+    {
+      "epoch": 33.25,
+      "learning_rate": 6.166666666666667e-05,
+      "loss": 1.0897,
+      "step": 88900
+    },
+    {
+      "epoch": 33.28,
+      "learning_rate": 6.111111111111111e-05,
+      "loss": 1.093,
+      "step": 89000
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 6.055555555555556e-05,
+      "loss": 1.0889,
+      "step": 89100
+    },
+    {
+      "epoch": 33.36,
+      "learning_rate": 6e-05,
+      "loss": 1.0903,
+      "step": 89200
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 5.9444444444444445e-05,
+      "loss": 1.0895,
+      "step": 89300
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 5.888888888888889e-05,
+      "loss": 1.0928,
+      "step": 89400
+    },
+    {
+      "epoch": 33.47,
+      "learning_rate": 5.833333333333333e-05,
+      "loss": 1.0909,
+      "step": 89500
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 5.7777777777777776e-05,
+      "loss": 1.0889,
+      "step": 89600
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 5.722222222222223e-05,
+      "loss": 1.0913,
+      "step": 89700
+    },
+    {
+      "epoch": 33.58,
+      "learning_rate": 5.6666666666666664e-05,
+      "loss": 1.09,
+      "step": 89800
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 5.6111111111111114e-05,
+      "loss": 1.0908,
+      "step": 89900
+    },
+    {
+      "epoch": 33.66,
+      "learning_rate": 5.555555555555555e-05,
+      "loss": 1.0879,
+      "step": 90000
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 5.5e-05,
+      "loss": 1.0862,
+      "step": 90100
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 5.444444444444444e-05,
+      "loss": 1.086,
+      "step": 90200
+    },
+    {
+      "epoch": 33.77,
+      "learning_rate": 5.388888888888889e-05,
+      "loss": 1.0897,
+      "step": 90300
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 5.333333333333334e-05,
+      "loss": 1.0888,
+      "step": 90400
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 5.277777777777778e-05,
+      "loss": 1.0892,
+      "step": 90500
+    },
+    {
+      "epoch": 33.88,
+      "learning_rate": 5.222222222222223e-05,
+      "loss": 1.0858,
+      "step": 90600
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 5.1666666666666664e-05,
+      "loss": 1.0874,
+      "step": 90700
+    },
+    {
+      "epoch": 33.96,
+      "learning_rate": 5.1111111111111115e-05,
+      "loss": 1.0864,
+      "step": 90800
+    },
+    {
+      "epoch": 33.99,
+      "learning_rate": 5.055555555555556e-05,
+      "loss": 1.088,
+      "step": 90900
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 5e-05,
+      "loss": 1.0882,
+      "step": 91000
+    },
+    {
+      "epoch": 34.07,
+      "learning_rate": 4.9444444444444446e-05,
+      "loss": 1.0818,
+      "step": 91100
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 4.888888888888889e-05,
+      "loss": 1.0873,
+      "step": 91200
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 4.8333333333333334e-05,
+      "loss": 1.0823,
+      "step": 91300
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 4.7777777777777784e-05,
+      "loss": 1.0852,
+      "step": 91400
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 4.722222222222222e-05,
+      "loss": 1.0845,
+      "step": 91500
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.0864,
+      "step": 91600
+    },
+    {
+      "epoch": 34.29,
+      "learning_rate": 4.611111111111111e-05,
+      "loss": 1.0839,
+      "step": 91700
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 4.555555555555556e-05,
+      "loss": 1.0821,
+      "step": 91800
+    },
+    {
+      "epoch": 34.37,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 1.0843,
+      "step": 91900
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 1.0788,
+      "step": 92000
+    },
+    {
+      "epoch": 34.44,
+      "learning_rate": 4.388888888888889e-05,
+      "loss": 1.0852,
+      "step": 92100
+    },
+    {
+      "epoch": 34.48,
+      "learning_rate": 4.3333333333333334e-05,
+      "loss": 1.0842,
+      "step": 92200
+    },
+    {
+      "epoch": 34.52,
+      "learning_rate": 4.277777777777778e-05,
+      "loss": 1.0815,
+      "step": 92300
+    },
+    {
+      "epoch": 34.55,
+      "learning_rate": 4.222222222222222e-05,
+      "loss": 1.0848,
+      "step": 92400
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 4.1666666666666665e-05,
+      "loss": 1.0835,
+      "step": 92500
+    },
+    {
+      "epoch": 34.63,
+      "learning_rate": 4.1111111111111116e-05,
+      "loss": 1.0838,
+      "step": 92600
+    },
+    {
+      "epoch": 34.67,
+      "learning_rate": 4.055555555555555e-05,
+      "loss": 1.0812,
+      "step": 92700
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 4e-05,
+      "loss": 1.0794,
+      "step": 92800
+    },
+    {
+      "epoch": 34.74,
+      "learning_rate": 3.944444444444444e-05,
+      "loss": 1.0808,
+      "step": 92900
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 3.888888888888889e-05,
+      "loss": 1.084,
+      "step": 93000
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 3.8333333333333334e-05,
+      "loss": 1.0825,
+      "step": 93100
+    },
+    {
+      "epoch": 34.85,
+      "learning_rate": 3.777777777777778e-05,
+      "loss": 1.0788,
+      "step": 93200
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 3.722222222222222e-05,
+      "loss": 1.0812,
+      "step": 93300
+    },
+    {
+      "epoch": 34.93,
+      "learning_rate": 3.6666666666666666e-05,
+      "loss": 1.0804,
+      "step": 93400
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 3.611111111111111e-05,
+      "loss": 1.0809,
+      "step": 93500
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 3.555555555555555e-05,
+      "loss": 1.0839,
+      "step": 93600
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 1.0794,
+      "step": 93700
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 3.444444444444445e-05,
+      "loss": 1.0783,
+      "step": 93800
+    },
+    {
+      "epoch": 35.12,
+      "learning_rate": 3.388888888888889e-05,
+      "loss": 1.0794,
+      "step": 93900
+    },
+    {
+      "epoch": 35.15,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.0768,
+      "step": 94000
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 3.277777777777778e-05,
+      "loss": 1.082,
+      "step": 94100
+    },
+    {
+      "epoch": 35.23,
+      "learning_rate": 3.222222222222222e-05,
+      "loss": 1.0793,
+      "step": 94200
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 3.166666666666667e-05,
+      "loss": 1.0776,
+      "step": 94300
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 3.111111111111111e-05,
+      "loss": 1.0769,
+      "step": 94400
+    },
+    {
+      "epoch": 35.34,
+      "learning_rate": 3.0555555555555554e-05,
+      "loss": 1.0783,
+      "step": 94500
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 3e-05,
+      "loss": 1.0774,
+      "step": 94600
+    },
+    {
+      "epoch": 35.42,
+      "learning_rate": 2.9444444444444445e-05,
+      "loss": 1.0772,
+      "step": 94700
+    },
+    {
+      "epoch": 35.45,
+      "learning_rate": 2.8888888888888888e-05,
+      "loss": 1.0791,
+      "step": 94800
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 2.8333333333333332e-05,
+      "loss": 1.0766,
+      "step": 94900
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 2.7777777777777776e-05,
+      "loss": 1.0755,
+      "step": 95000
+    },
+    {
+      "epoch": 35.56,
+      "learning_rate": 2.722222222222222e-05,
+      "loss": 1.0744,
+      "step": 95100
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 2.666666666666667e-05,
+      "loss": 1.0775,
+      "step": 95200
+    },
+    {
+      "epoch": 35.64,
+      "learning_rate": 2.6111111111111114e-05,
+      "loss": 1.0723,
+      "step": 95300
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 2.5555555555555557e-05,
+      "loss": 1.0763,
+      "step": 95400
+    },
+    {
+      "epoch": 35.71,
+      "learning_rate": 2.5e-05,
+      "loss": 1.0765,
+      "step": 95500
+    },
+    {
+      "epoch": 35.75,
+      "learning_rate": 2.4444444444444445e-05,
+      "loss": 1.0735,
+      "step": 95600
+    },
+    {
+      "epoch": 35.79,
+      "learning_rate": 2.3888888888888892e-05,
+      "loss": 1.0765,
+      "step": 95700
+    },
+    {
+      "epoch": 35.83,
+      "learning_rate": 2.3333333333333336e-05,
+      "loss": 1.0743,
+      "step": 95800
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 2.277777777777778e-05,
+      "loss": 1.0742,
+      "step": 95900
+    },
+    {
+      "epoch": 35.9,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 1.0755,
+      "step": 96000
+    },
+    {
+      "epoch": 35.94,
+      "learning_rate": 2.1666666666666667e-05,
+      "loss": 1.0747,
+      "step": 96100
+    },
+    {
+      "epoch": 35.98,
+      "learning_rate": 2.111111111111111e-05,
+      "loss": 1.0753,
+      "step": 96200
+    },
+    {
+      "epoch": 36.01,
+      "learning_rate": 2.0555555555555558e-05,
+      "loss": 1.0784,
+      "step": 96300
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 2e-05,
+      "loss": 1.0724,
+      "step": 96400
+    },
+    {
+      "epoch": 36.09,
+      "learning_rate": 1.9444444444444445e-05,
+      "loss": 1.0725,
+      "step": 96500
+    },
+    {
+      "epoch": 36.13,
+      "learning_rate": 1.888888888888889e-05,
+      "loss": 1.0723,
+      "step": 96600
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 1.8333333333333333e-05,
+      "loss": 1.073,
+      "step": 96700
+    },
+    {
+      "epoch": 36.2,
+      "learning_rate": 1.7777777777777777e-05,
+      "loss": 1.0736,
+      "step": 96800
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 1.7222222222222224e-05,
+      "loss": 1.0724,
+      "step": 96900
+    },
+    {
+      "epoch": 36.28,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 1.075,
+      "step": 97000
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 1.611111111111111e-05,
+      "loss": 1.0739,
+      "step": 97100
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 1.5555555555555555e-05,
+      "loss": 1.0721,
+      "step": 97200
+    },
+    {
+      "epoch": 36.39,
+      "learning_rate": 1.5e-05,
+      "loss": 1.0707,
+      "step": 97300
+    },
+    {
+      "epoch": 36.42,
+      "learning_rate": 1.4444444444444444e-05,
+      "loss": 1.0714,
+      "step": 97400
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 1.3888888888888888e-05,
+      "loss": 1.0712,
+      "step": 97500
+    },
+    {
+      "epoch": 36.5,
+      "learning_rate": 1.3333333333333335e-05,
+      "loss": 1.0741,
+      "step": 97600
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 1.2777777777777779e-05,
+      "loss": 1.0715,
+      "step": 97700
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 1.2222222222222222e-05,
+      "loss": 1.0717,
+      "step": 97800
+    },
+    {
+      "epoch": 36.61,
+      "learning_rate": 1.1666666666666668e-05,
+      "loss": 1.0703,
+      "step": 97900
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 1.1111111111111112e-05,
+      "loss": 1.0683,
+      "step": 98000
+    },
+    {
+      "epoch": 36.69,
+      "learning_rate": 1.0555555555555555e-05,
+      "loss": 1.069,
+      "step": 98100
+    },
+    {
+      "epoch": 36.72,
+      "learning_rate": 1e-05,
+      "loss": 1.0751,
+      "step": 98200
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 9.444444444444445e-06,
+      "loss": 1.0719,
+      "step": 98300
+    },
+    {
+      "epoch": 36.8,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 1.0701,
+      "step": 98400
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 1.0719,
+      "step": 98500
+    },
+    {
+      "epoch": 36.87,
+      "learning_rate": 7.777777777777777e-06,
+      "loss": 1.0692,
+      "step": 98600
+    },
+    {
+      "epoch": 36.91,
+      "learning_rate": 7.222222222222222e-06,
+      "loss": 1.0716,
+      "step": 98700
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 6.6666666666666675e-06,
+      "loss": 1.0712,
+      "step": 98800
+    },
+    {
+      "epoch": 36.99,
+      "learning_rate": 6.111111111111111e-06,
+      "loss": 1.0695,
+      "step": 98900
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 5.555555555555556e-06,
+      "loss": 1.0749,
+      "step": 99000
+    },
+    {
+      "epoch": 37.06,
+      "learning_rate": 5e-06,
+      "loss": 1.0682,
+      "step": 99100
+    },
+    {
+      "epoch": 37.1,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 1.0721,
+      "step": 99200
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 3.888888888888889e-06,
+      "loss": 1.069,
+      "step": 99300
+    },
+    {
+      "epoch": 37.17,
+      "learning_rate": 3.3333333333333337e-06,
+      "loss": 1.0687,
+      "step": 99400
+    },
+    {
+      "epoch": 37.21,
+      "learning_rate": 2.777777777777778e-06,
+      "loss": 1.0688,
+      "step": 99500
+    },
+    {
+      "epoch": 37.25,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 1.0694,
+      "step": 99600
+    },
+    {
+      "epoch": 37.28,
+      "learning_rate": 1.6666666666666669e-06,
+      "loss": 1.0671,
+      "step": 99700
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 1.111111111111111e-06,
+      "loss": 1.0672,
+      "step": 99800
+    },
+    {
+      "epoch": 37.36,
+      "learning_rate": 5.555555555555555e-07,
+      "loss": 1.0654,
+      "step": 99900
+    },
+    {
+      "epoch": 37.4,
+      "learning_rate": 0.0,
+      "loss": 1.0688,
+      "step": 100000
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 0.00026289473684210524,
+      "loss": 1.1059,
+      "step": 100100
+    },
+    {
+      "epoch": 37.47,
+      "learning_rate": 0.00026263157894736844,
+      "loss": 1.1269,
+      "step": 100200
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 0.0002623684210526316,
+      "loss": 1.1303,
+      "step": 100300
+    },
+    {
+      "epoch": 37.55,
+      "learning_rate": 0.0002621052631578947,
+      "loss": 1.1346,
+      "step": 100400
+    },
+    {
+      "epoch": 37.58,
+      "learning_rate": 0.00026184210526315787,
+      "loss": 1.1416,
+      "step": 100500
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 0.00026157894736842107,
+      "loss": 1.1439,
+      "step": 100600
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 0.0002613157894736842,
+      "loss": 1.1447,
+      "step": 100700
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 0.00026105263157894735,
+      "loss": 1.1446,
+      "step": 100800
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 0.0002607894736842105,
+      "loss": 1.1452,
+      "step": 100900
+    },
+    {
+      "epoch": 37.77,
+      "learning_rate": 0.0002605263157894737,
+      "loss": 1.1472,
+      "step": 101000
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 0.0002602631578947369,
+      "loss": 1.149,
+      "step": 101100
+    },
+    {
+      "epoch": 37.85,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 1.1538,
+      "step": 101200
+    },
+    {
+      "epoch": 37.88,
+      "learning_rate": 0.0002597368421052632,
+      "loss": 1.149,
+      "step": 101300
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 0.0002594736842105263,
+      "loss": 1.1556,
+      "step": 101400
+    },
+    {
+      "epoch": 37.96,
+      "learning_rate": 0.00025921052631578946,
+      "loss": 1.1521,
+      "step": 101500
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 0.00025894736842105266,
+      "loss": 1.1535,
+      "step": 101600
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 0.0002586842105263158,
+      "loss": 1.1576,
+      "step": 101700
+    },
+    {
+      "epoch": 38.07,
+      "learning_rate": 0.00025842105263157894,
+      "loss": 1.1513,
+      "step": 101800
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 0.0002581578947368421,
+      "loss": 1.1513,
+      "step": 101900
+    },
+    {
+      "epoch": 38.15,
+      "learning_rate": 0.0002578947368421053,
+      "loss": 1.1516,
+      "step": 102000
+    },
+    {
+      "epoch": 38.18,
+      "learning_rate": 0.0002576315789473684,
+      "loss": 1.152,
+      "step": 102100
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 0.00025736842105263157,
+      "loss": 1.153,
+      "step": 102200
+    },
+    {
+      "epoch": 38.26,
+      "learning_rate": 0.0002571052631578947,
+      "loss": 1.1558,
+      "step": 102300
+    },
+    {
+      "epoch": 38.29,
+      "learning_rate": 0.00025684210526315786,
+      "loss": 1.1527,
+      "step": 102400
+    },
+    {
+      "epoch": 38.33,
+      "learning_rate": 0.00025657894736842105,
+      "loss": 1.1557,
+      "step": 102500
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 0.00025631578947368425,
+      "loss": 1.1532,
+      "step": 102600
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 0.0002560526315789474,
+      "loss": 1.1515,
+      "step": 102700
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 0.00025578947368421054,
+      "loss": 1.1515,
+      "step": 102800
+    },
+    {
+      "epoch": 38.48,
+      "learning_rate": 0.0002555263157894737,
+      "loss": 1.156,
+      "step": 102900
+    },
+    {
+      "epoch": 38.52,
+      "learning_rate": 0.0002552631578947369,
+      "loss": 1.1539,
+      "step": 103000
+    },
+    {
+      "epoch": 38.56,
+      "learning_rate": 0.000255,
+      "loss": 1.151,
+      "step": 103100
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 0.00025473684210526316,
+      "loss": 1.1533,
+      "step": 103200
+    },
+    {
+      "epoch": 38.63,
+      "learning_rate": 0.0002544736842105263,
+      "loss": 1.1538,
+      "step": 103300
+    },
+    {
+      "epoch": 38.67,
+      "learning_rate": 0.0002542105263157895,
+      "loss": 1.1567,
+      "step": 103400
+    },
+    {
+      "epoch": 38.71,
+      "learning_rate": 0.00025394736842105264,
+      "loss": 1.1575,
+      "step": 103500
+    },
+    {
+      "epoch": 38.74,
+      "learning_rate": 0.0002536842105263158,
+      "loss": 1.1549,
+      "step": 103600
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 0.00025342105263157893,
+      "loss": 1.1582,
+      "step": 103700
+    },
+    {
+      "epoch": 38.82,
+      "learning_rate": 0.0002531578947368421,
+      "loss": 1.1554,
+      "step": 103800
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 0.00025289473684210527,
+      "loss": 1.1547,
+      "step": 103900
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 0.0002526315789473684,
+      "loss": 1.1584,
+      "step": 104000
+    },
+    {
+      "epoch": 38.93,
+      "learning_rate": 0.00025236842105263156,
+      "loss": 1.1595,
+      "step": 104100
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 0.00025210526315789475,
+      "loss": 1.1589,
+      "step": 104200
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 0.0002518421052631579,
+      "loss": 1.1614,
+      "step": 104300
+    },
+    {
+      "epoch": 39.04,
+      "learning_rate": 0.0002515789473684211,
+      "loss": 1.1519,
+      "step": 104400
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 0.00025131578947368424,
+      "loss": 1.1525,
+      "step": 104500
+    },
+    {
+      "epoch": 39.12,
+      "learning_rate": 0.0002510526315789474,
+      "loss": 1.1524,
+      "step": 104600
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 0.0002507894736842105,
+      "loss": 1.1555,
+      "step": 104700
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 0.0002505263157894737,
+      "loss": 1.1548,
+      "step": 104800
+    },
+    {
+      "epoch": 39.23,
+      "learning_rate": 0.00025026315789473686,
+      "loss": 1.1541,
+      "step": 104900
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 0.00025,
+      "loss": 1.1527,
+      "step": 105000
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 0.00024973684210526315,
+      "loss": 1.154,
+      "step": 105100
+    },
+    {
+      "epoch": 39.34,
+      "learning_rate": 0.00024947368421052635,
+      "loss": 1.1538,
+      "step": 105200
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 0.0002492105263157895,
+      "loss": 1.154,
+      "step": 105300
+    },
+    {
+      "epoch": 39.42,
+      "learning_rate": 0.00024894736842105263,
+      "loss": 1.1523,
+      "step": 105400
+    },
+    {
+      "epoch": 39.45,
+      "learning_rate": 0.0002486842105263158,
+      "loss": 1.1553,
+      "step": 105500
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 0.00024842105263157897,
+      "loss": 1.1565,
+      "step": 105600
+    },
+    {
+      "epoch": 39.53,
+      "learning_rate": 0.0002481578947368421,
+      "loss": 1.1535,
+      "step": 105700
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 0.00024789473684210526,
+      "loss": 1.1564,
+      "step": 105800
+    },
+    {
+      "epoch": 39.6,
+      "learning_rate": 0.00024763157894736845,
+      "loss": 1.1558,
+      "step": 105900
+    },
+    {
+      "epoch": 39.64,
+      "learning_rate": 0.0002473684210526316,
+      "loss": 1.1538,
+      "step": 106000
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 0.00024710526315789474,
+      "loss": 1.1571,
+      "step": 106100
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 0.0002468421052631579,
+      "loss": 1.1573,
+      "step": 106200
+    },
+    {
+      "epoch": 39.75,
+      "learning_rate": 0.0002465789473684211,
+      "loss": 1.1535,
+      "step": 106300
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 0.0002463157894736842,
+      "loss": 1.1542,
+      "step": 106400
+    },
+    {
+      "epoch": 39.83,
+      "learning_rate": 0.00024605263157894737,
+      "loss": 1.1519,
+      "step": 106500
+    },
+    {
+      "epoch": 39.87,
+      "learning_rate": 0.0002457894736842105,
+      "loss": 1.1534,
+      "step": 106600
+    },
+    {
+      "epoch": 39.9,
+      "learning_rate": 0.0002455263157894737,
+      "loss": 1.1508,
+      "step": 106700
+    },
+    {
+      "epoch": 39.94,
+      "learning_rate": 0.00024526315789473685,
+      "loss": 1.1543,
+      "step": 106800
+    },
+    {
+      "epoch": 39.98,
+      "learning_rate": 0.000245,
+      "loss": 1.1545,
+      "step": 106900
+    },
+    {
+      "epoch": 40.01,
+      "learning_rate": 0.0002447368421052632,
+      "loss": 1.1544,
+      "step": 107000
+    },
+    {
+      "epoch": 40.05,
+      "learning_rate": 0.00024447368421052633,
+      "loss": 1.1492,
+      "step": 107100
+    },
+    {
+      "epoch": 40.09,
+      "learning_rate": 0.0002442105263157895,
+      "loss": 1.152,
+      "step": 107200
+    },
+    {
+      "epoch": 40.13,
+      "learning_rate": 0.00024394736842105265,
+      "loss": 1.1503,
+      "step": 107300
+    },
+    {
+      "epoch": 40.16,
+      "learning_rate": 0.0002436842105263158,
+      "loss": 1.1506,
+      "step": 107400
+    },
+    {
+      "epoch": 40.2,
+      "learning_rate": 0.00024342105263157896,
+      "loss": 1.1509,
+      "step": 107500
+    },
+    {
+      "epoch": 40.24,
+      "learning_rate": 0.0002431578947368421,
+      "loss": 1.1511,
+      "step": 107600
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 0.00024289473684210527,
+      "loss": 1.1493,
+      "step": 107700
+    },
+    {
+      "epoch": 40.31,
+      "learning_rate": 0.00024263157894736841,
+      "loss": 1.151,
+      "step": 107800
+    },
+    {
+      "epoch": 40.35,
+      "learning_rate": 0.00024236842105263158,
+      "loss": 1.1537,
+      "step": 107900
+    },
+    {
+      "epoch": 40.39,
+      "learning_rate": 0.00024210526315789475,
+      "loss": 1.1519,
+      "step": 108000
+    },
+    {
+      "epoch": 40.43,
+      "learning_rate": 0.0002418421052631579,
+      "loss": 1.1534,
+      "step": 108100
+    },
+    {
+      "epoch": 40.46,
+      "learning_rate": 0.00024157894736842107,
+      "loss": 1.1511,
+      "step": 108200
+    },
+    {
+      "epoch": 40.5,
+      "learning_rate": 0.0002413157894736842,
+      "loss": 1.1546,
+      "step": 108300
+    },
+    {
+      "epoch": 40.54,
+      "learning_rate": 0.00024105263157894738,
+      "loss": 1.1518,
+      "step": 108400
+    },
+    {
+      "epoch": 40.58,
+      "learning_rate": 0.00024078947368421052,
+      "loss": 1.1538,
+      "step": 108500
+    },
+    {
+      "epoch": 40.61,
+      "learning_rate": 0.00024052631578947367,
+      "loss": 1.1513,
+      "step": 108600
+    },
+    {
+      "epoch": 40.65,
+      "learning_rate": 0.00024026315789473686,
+      "loss": 1.1466,
+      "step": 108700
+    },
+    {
+      "epoch": 40.69,
+      "learning_rate": 0.00024,
+      "loss": 1.151,
+      "step": 108800
+    },
+    {
+      "epoch": 40.73,
+      "learning_rate": 0.00023973684210526318,
+      "loss": 1.1535,
+      "step": 108900
+    },
+    {
+      "epoch": 40.76,
+      "learning_rate": 0.00023947368421052632,
+      "loss": 1.1516,
+      "step": 109000
+    },
+    {
+      "epoch": 40.8,
+      "learning_rate": 0.0002392105263157895,
+      "loss": 1.1516,
+      "step": 109100
+    },
+    {
+      "epoch": 40.84,
+      "learning_rate": 0.00023894736842105263,
+      "loss": 1.1523,
+      "step": 109200
+    },
+    {
+      "epoch": 40.87,
+      "learning_rate": 0.00023868421052631577,
+      "loss": 1.1517,
+      "step": 109300
+    },
+    {
+      "epoch": 40.91,
+      "learning_rate": 0.00023842105263157895,
+      "loss": 1.1508,
+      "step": 109400
+    },
+    {
+      "epoch": 40.95,
+      "learning_rate": 0.0002381578947368421,
+      "loss": 1.1529,
+      "step": 109500
+    },
+    {
+      "epoch": 40.99,
+      "learning_rate": 0.00023789473684210529,
+      "loss": 1.1531,
+      "step": 109600
+    },
+    {
+      "epoch": 41.02,
+      "learning_rate": 0.00023763157894736843,
+      "loss": 1.1527,
+      "step": 109700
+    },
+    {
+      "epoch": 41.06,
+      "learning_rate": 0.0002373684210526316,
+      "loss": 1.1456,
+      "step": 109800
+    },
+    {
+      "epoch": 41.1,
+      "learning_rate": 0.00023710526315789474,
+      "loss": 1.1491,
+      "step": 109900
+    },
+    {
+      "epoch": 41.14,
+      "learning_rate": 0.00023684210526315788,
+      "loss": 1.1463,
+      "step": 110000
+    },
+    {
+      "epoch": 41.17,
+      "learning_rate": 0.00023657894736842105,
+      "loss": 1.1486,
+      "step": 110100
+    },
+    {
+      "epoch": 41.21,
+      "learning_rate": 0.0002363157894736842,
+      "loss": 1.1481,
+      "step": 110200
+    },
+    {
+      "epoch": 41.25,
+      "learning_rate": 0.00023605263157894737,
+      "loss": 1.1483,
+      "step": 110300
+    },
+    {
+      "epoch": 41.29,
+      "learning_rate": 0.00023578947368421054,
+      "loss": 1.1478,
+      "step": 110400
+    },
+    {
+      "epoch": 41.32,
+      "learning_rate": 0.0002355263157894737,
+      "loss": 1.1514,
+      "step": 110500
+    },
+    {
+      "epoch": 41.36,
+      "learning_rate": 0.00023526315789473685,
+      "loss": 1.1476,
+      "step": 110600
+    },
+    {
+      "epoch": 41.4,
+      "learning_rate": 0.000235,
+      "loss": 1.1474,
+      "step": 110700
+    },
+    {
+      "epoch": 41.44,
+      "learning_rate": 0.00023473684210526316,
+      "loss": 1.1475,
+      "step": 110800
+    },
+    {
+      "epoch": 41.47,
+      "learning_rate": 0.0002344736842105263,
+      "loss": 1.1456,
+      "step": 110900
+    },
+    {
+      "epoch": 41.51,
+      "learning_rate": 0.00023421052631578948,
+      "loss": 1.1478,
+      "step": 111000
+    },
+    {
+      "epoch": 41.55,
+      "learning_rate": 0.00023394736842105262,
+      "loss": 1.1482,
+      "step": 111100
+    },
+    {
+      "epoch": 41.59,
+      "learning_rate": 0.00023368421052631582,
+      "loss": 1.1426,
+      "step": 111200
+    },
+    {
+      "epoch": 41.62,
+      "learning_rate": 0.00023342105263157896,
+      "loss": 1.1472,
+      "step": 111300
+    },
+    {
+      "epoch": 41.66,
+      "learning_rate": 0.0002331578947368421,
+      "loss": 1.1486,
+      "step": 111400
+    },
+    {
+      "epoch": 41.7,
+      "learning_rate": 0.00023289473684210527,
+      "loss": 1.1452,
+      "step": 111500
+    },
+    {
+      "epoch": 41.74,
+      "learning_rate": 0.00023263157894736841,
+      "loss": 1.1501,
+      "step": 111600
+    },
+    {
+      "epoch": 41.77,
+      "learning_rate": 0.00023236842105263158,
+      "loss": 1.1473,
+      "step": 111700
+    },
+    {
+      "epoch": 41.81,
+      "learning_rate": 0.00023210526315789473,
+      "loss": 1.1454,
+      "step": 111800
+    },
+    {
+      "epoch": 41.85,
+      "learning_rate": 0.0002318421052631579,
+      "loss": 1.1464,
+      "step": 111900
+    },
+    {
+      "epoch": 41.88,
+      "learning_rate": 0.00023157894736842107,
+      "loss": 1.1466,
+      "step": 112000
+    },
+    {
+      "epoch": 41.92,
+      "learning_rate": 0.0002313157894736842,
+      "loss": 1.1459,
+      "step": 112100
+    },
+    {
+      "epoch": 41.96,
+      "learning_rate": 0.00023105263157894738,
+      "loss": 1.1454,
+      "step": 112200
+    },
+    {
+      "epoch": 42.0,
+      "learning_rate": 0.00023078947368421052,
+      "loss": 1.1471,
+      "step": 112300
+    },
+    {
+      "epoch": 42.03,
+      "learning_rate": 0.0002305263157894737,
+      "loss": 1.1486,
+      "step": 112400
+    },
+    {
+      "epoch": 42.07,
+      "learning_rate": 0.00023026315789473684,
+      "loss": 1.143,
+      "step": 112500
+    },
+    {
+      "epoch": 42.11,
+      "learning_rate": 0.00023,
+      "loss": 1.1375,
+      "step": 112600
+    },
+    {
+      "epoch": 42.15,
+      "learning_rate": 0.00022973684210526315,
+      "loss": 1.1434,
+      "step": 112700
+    },
+    {
+      "epoch": 42.18,
+      "learning_rate": 0.00022947368421052632,
+      "loss": 1.1454,
+      "step": 112800
+    },
+    {
+      "epoch": 42.22,
+      "learning_rate": 0.0002292105263157895,
+      "loss": 1.1425,
+      "step": 112900
+    },
+    {
+      "epoch": 42.26,
+      "learning_rate": 0.00022894736842105263,
+      "loss": 1.1455,
+      "step": 113000
+    },
+    {
+      "epoch": 42.3,
+      "learning_rate": 0.0002286842105263158,
+      "loss": 1.1413,
+      "step": 113100
+    },
+    {
+      "epoch": 42.33,
+      "learning_rate": 0.00022842105263157895,
+      "loss": 1.141,
+      "step": 113200
+    },
+    {
+      "epoch": 42.37,
+      "learning_rate": 0.00022815789473684212,
+      "loss": 1.1455,
+      "step": 113300
+    },
+    {
+      "epoch": 42.41,
+      "learning_rate": 0.00022789473684210526,
+      "loss": 1.1436,
+      "step": 113400
+    },
+    {
+      "epoch": 42.45,
+      "learning_rate": 0.00022763157894736843,
+      "loss": 1.1412,
+      "step": 113500
+    },
+    {
+      "epoch": 42.48,
+      "learning_rate": 0.0002273684210526316,
+      "loss": 1.144,
+      "step": 113600
+    },
+    {
+      "epoch": 42.52,
+      "learning_rate": 0.00022710526315789474,
+      "loss": 1.1434,
+      "step": 113700
+    },
+    {
+      "epoch": 42.56,
+      "learning_rate": 0.0002268421052631579,
+      "loss": 1.1438,
+      "step": 113800
+    },
+    {
+      "epoch": 42.6,
+      "learning_rate": 0.00022657894736842105,
+      "loss": 1.1451,
+      "step": 113900
+    },
+    {
+      "epoch": 42.63,
+      "learning_rate": 0.00022631578947368422,
+      "loss": 1.1433,
+      "step": 114000
+    },
+    {
+      "epoch": 42.67,
+      "learning_rate": 0.00022605263157894737,
+      "loss": 1.1418,
+      "step": 114100
+    },
+    {
+      "epoch": 42.71,
+      "learning_rate": 0.00022578947368421054,
+      "loss": 1.1429,
+      "step": 114200
+    },
+    {
+      "epoch": 42.74,
+      "learning_rate": 0.00022552631578947368,
+      "loss": 1.1471,
+      "step": 114300
+    },
+    {
+      "epoch": 42.78,
+      "learning_rate": 0.00022526315789473682,
+      "loss": 1.1397,
+      "step": 114400
+    },
+    {
+      "epoch": 42.82,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 1.1401,
+      "step": 114500
+    },
+    {
+      "epoch": 42.86,
+      "learning_rate": 0.00022473684210526316,
+      "loss": 1.1405,
+      "step": 114600
+    },
+    {
+      "epoch": 42.89,
+      "learning_rate": 0.00022447368421052633,
+      "loss": 1.1435,
+      "step": 114700
+    },
+    {
+      "epoch": 42.93,
+      "learning_rate": 0.00022421052631578948,
+      "loss": 1.1395,
+      "step": 114800
+    },
+    {
+      "epoch": 42.97,
+      "learning_rate": 0.00022394736842105265,
+      "loss": 1.1414,
+      "step": 114900
+    },
+    {
+      "epoch": 43.01,
+      "learning_rate": 0.0002236842105263158,
+      "loss": 1.1457,
+      "step": 115000
+    },
+    {
+      "epoch": 43.04,
+      "learning_rate": 0.00022342105263157893,
+      "loss": 1.1396,
+      "step": 115100
+    },
+    {
+      "epoch": 43.08,
+      "learning_rate": 0.0002231578947368421,
+      "loss": 1.1339,
+      "step": 115200
+    },
+    {
+      "epoch": 43.12,
+      "learning_rate": 0.00022289473684210527,
+      "loss": 1.1366,
+      "step": 115300
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 0.00022263157894736844,
+      "loss": 1.142,
+      "step": 115400
+    },
+    {
+      "epoch": 43.19,
+      "learning_rate": 0.00022236842105263159,
+      "loss": 1.138,
+      "step": 115500
+    },
+    {
+      "epoch": 43.23,
+      "learning_rate": 0.00022210526315789476,
+      "loss": 1.1372,
+      "step": 115600
+    },
+    {
+      "epoch": 43.27,
+      "learning_rate": 0.0002218421052631579,
+      "loss": 1.1377,
+      "step": 115700
+    },
+    {
+      "epoch": 43.31,
+      "learning_rate": 0.00022157894736842104,
+      "loss": 1.1382,
+      "step": 115800
+    },
+    {
+      "epoch": 43.34,
+      "learning_rate": 0.0002213157894736842,
+      "loss": 1.1375,
+      "step": 115900
+    },
+    {
+      "epoch": 43.38,
+      "learning_rate": 0.00022105263157894735,
+      "loss": 1.1372,
+      "step": 116000
+    },
+    {
+      "epoch": 43.42,
+      "learning_rate": 0.00022078947368421055,
+      "loss": 1.1373,
+      "step": 116100
+    },
+    {
+      "epoch": 43.46,
+      "learning_rate": 0.0002205263157894737,
+      "loss": 1.1357,
+      "step": 116200
+    },
+    {
+      "epoch": 43.49,
+      "learning_rate": 0.00022026315789473686,
+      "loss": 1.1357,
+      "step": 116300
+    },
+    {
+      "epoch": 43.53,
+      "learning_rate": 0.00022,
+      "loss": 1.1406,
+      "step": 116400
+    },
+    {
+      "epoch": 43.57,
+      "learning_rate": 0.00021973684210526315,
+      "loss": 1.1385,
+      "step": 116500
+    },
+    {
+      "epoch": 43.6,
+      "learning_rate": 0.00021947368421052632,
+      "loss": 1.1415,
+      "step": 116600
+    },
+    {
+      "epoch": 43.64,
+      "learning_rate": 0.00021921052631578946,
+      "loss": 1.1387,
+      "step": 116700
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 0.00021894736842105263,
+      "loss": 1.1385,
+      "step": 116800
+    },
+    {
+      "epoch": 43.72,
+      "learning_rate": 0.0002186842105263158,
+      "loss": 1.1372,
+      "step": 116900
+    },
+    {
+      "epoch": 43.75,
+      "learning_rate": 0.00021842105263157897,
+      "loss": 1.1343,
+      "step": 117000
+    },
+    {
+      "epoch": 43.79,
+      "learning_rate": 0.00021815789473684212,
+      "loss": 1.1374,
+      "step": 117100
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 0.00021789473684210526,
+      "loss": 1.1398,
+      "step": 117200
+    },
+    {
+      "epoch": 43.87,
+      "learning_rate": 0.00021763157894736843,
+      "loss": 1.1385,
+      "step": 117300
+    },
+    {
+      "epoch": 43.9,
+      "learning_rate": 0.00021736842105263157,
+      "loss": 1.1394,
+      "step": 117400
+    },
+    {
+      "epoch": 43.94,
+      "learning_rate": 0.00021710526315789474,
+      "loss": 1.1401,
+      "step": 117500
+    },
+    {
+      "epoch": 43.98,
+      "learning_rate": 0.00021684210526315789,
+      "loss": 1.1371,
+      "step": 117600
+    },
+    {
+      "epoch": 44.02,
+      "learning_rate": 0.00021657894736842108,
+      "loss": 1.1424,
+      "step": 117700
+    },
+    {
+      "epoch": 44.05,
+      "learning_rate": 0.00021631578947368423,
+      "loss": 1.1344,
+      "step": 117800
+    },
+    {
+      "epoch": 44.09,
+      "learning_rate": 0.00021605263157894737,
+      "loss": 1.1324,
+      "step": 117900
+    },
+    {
+      "epoch": 44.13,
+      "learning_rate": 0.00021578947368421054,
+      "loss": 1.1361,
+      "step": 118000
+    },
+    {
+      "epoch": 44.17,
+      "learning_rate": 0.00021552631578947368,
+      "loss": 1.1341,
+      "step": 118100
+    },
+    {
+      "epoch": 44.2,
+      "learning_rate": 0.00021526315789473685,
+      "loss": 1.1312,
+      "step": 118200
+    },
+    {
+      "epoch": 44.24,
+      "learning_rate": 0.000215,
+      "loss": 1.1307,
+      "step": 118300
+    },
+    {
+      "epoch": 44.28,
+      "learning_rate": 0.00021473684210526316,
+      "loss": 1.1305,
+      "step": 118400
+    },
+    {
+      "epoch": 44.32,
+      "learning_rate": 0.0002144736842105263,
+      "loss": 1.134,
+      "step": 118500
+    },
+    {
+      "epoch": 44.35,
+      "learning_rate": 0.00021421052631578948,
+      "loss": 1.1339,
+      "step": 118600
+    },
+    {
+      "epoch": 44.39,
+      "learning_rate": 0.00021394736842105265,
+      "loss": 1.1359,
+      "step": 118700
+    },
+    {
+      "epoch": 44.43,
+      "learning_rate": 0.0002136842105263158,
+      "loss": 1.1354,
+      "step": 118800
+    },
+    {
+      "epoch": 44.47,
+      "learning_rate": 0.00021342105263157896,
+      "loss": 1.1349,
+      "step": 118900
+    },
+    {
+      "epoch": 44.5,
+      "learning_rate": 0.0002131578947368421,
+      "loss": 1.1335,
+      "step": 119000
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 0.00021289473684210527,
+      "loss": 1.1335,
+      "step": 119100
+    },
+    {
+      "epoch": 44.58,
+      "learning_rate": 0.00021263157894736842,
+      "loss": 1.1416,
+      "step": 119200
+    },
+    {
+      "epoch": 44.61,
+      "learning_rate": 0.00021236842105263156,
+      "loss": 1.1344,
+      "step": 119300
+    },
+    {
+      "epoch": 44.65,
+      "learning_rate": 0.00021210526315789476,
+      "loss": 1.1337,
+      "step": 119400
+    },
+    {
+      "epoch": 44.69,
+      "learning_rate": 0.0002118421052631579,
+      "loss": 1.1299,
+      "step": 119500
+    },
+    {
+      "epoch": 44.73,
+      "learning_rate": 0.00021157894736842107,
+      "loss": 1.1337,
+      "step": 119600
+    },
+    {
+      "epoch": 44.76,
+      "learning_rate": 0.0002113157894736842,
+      "loss": 1.1338,
+      "step": 119700
+    },
+    {
+      "epoch": 44.8,
+      "learning_rate": 0.00021105263157894738,
+      "loss": 1.1312,
+      "step": 119800
+    },
+    {
+      "epoch": 44.84,
+      "learning_rate": 0.00021078947368421053,
+      "loss": 1.1329,
+      "step": 119900
+    },
+    {
+      "epoch": 44.88,
+      "learning_rate": 0.00021052631578947367,
+      "loss": 1.1367,
+      "step": 120000
+    },
+    {
+      "epoch": 44.91,
+      "learning_rate": 0.00021026315789473684,
+      "loss": 1.1346,
+      "step": 120100
+    },
+    {
+      "epoch": 44.95,
+      "learning_rate": 0.00021,
+      "loss": 1.1327,
+      "step": 120200
+    },
+    {
+      "epoch": 44.99,
+      "learning_rate": 0.00020973684210526318,
+      "loss": 1.1302,
+      "step": 120300
+    },
+    {
+      "epoch": 45.03,
+      "learning_rate": 0.00020947368421052632,
+      "loss": 1.1346,
+      "step": 120400
+    },
+    {
+      "epoch": 45.06,
+      "learning_rate": 0.0002092105263157895,
+      "loss": 1.1296,
+      "step": 120500
+    },
+    {
+      "epoch": 45.1,
+      "learning_rate": 0.00020894736842105263,
+      "loss": 1.1282,
+      "step": 120600
+    },
+    {
+      "epoch": 45.14,
+      "learning_rate": 0.00020868421052631578,
+      "loss": 1.1292,
+      "step": 120700
+    },
+    {
+      "epoch": 45.18,
+      "learning_rate": 0.00020842105263157895,
+      "loss": 1.1305,
+      "step": 120800
+    },
+    {
+      "epoch": 45.21,
+      "learning_rate": 0.0002081578947368421,
+      "loss": 1.1277,
+      "step": 120900
+    },
+    {
+      "epoch": 45.25,
+      "learning_rate": 0.0002078947368421053,
+      "loss": 1.1272,
+      "step": 121000
+    },
+    {
+      "epoch": 45.29,
+      "learning_rate": 0.00020763157894736843,
+      "loss": 1.1322,
+      "step": 121100
+    },
+    {
+      "epoch": 45.33,
+      "learning_rate": 0.0002073684210526316,
+      "loss": 1.1284,
+      "step": 121200
+    },
+    {
+      "epoch": 45.36,
+      "learning_rate": 0.00020710526315789474,
+      "loss": 1.1294,
+      "step": 121300
+    },
+    {
+      "epoch": 45.4,
+      "learning_rate": 0.0002068421052631579,
+      "loss": 1.1273,
+      "step": 121400
+    },
+    {
+      "epoch": 45.44,
+      "learning_rate": 0.00020657894736842106,
+      "loss": 1.1271,
+      "step": 121500
+    },
+    {
+      "epoch": 45.47,
+      "learning_rate": 0.0002063157894736842,
+      "loss": 1.1282,
+      "step": 121600
+    },
+    {
+      "epoch": 45.51,
+      "learning_rate": 0.00020605263157894737,
+      "loss": 1.1299,
+      "step": 121700
+    },
+    {
+      "epoch": 45.55,
+      "learning_rate": 0.00020578947368421054,
+      "loss": 1.1283,
+      "step": 121800
+    },
+    {
+      "epoch": 45.59,
+      "learning_rate": 0.0002055263157894737,
+      "loss": 1.1275,
+      "step": 121900
+    },
+    {
+      "epoch": 45.62,
+      "learning_rate": 0.00020526315789473685,
+      "loss": 1.1315,
+      "step": 122000
+    },
+    {
+      "epoch": 45.66,
+      "learning_rate": 0.000205,
+      "loss": 1.1278,
+      "step": 122100
+    },
+    {
+      "epoch": 45.7,
+      "learning_rate": 0.00020473684210526317,
+      "loss": 1.1295,
+      "step": 122200
+    },
+    {
+      "epoch": 45.74,
+      "learning_rate": 0.0002044736842105263,
+      "loss": 1.1314,
+      "step": 122300
+    },
+    {
+      "epoch": 45.77,
+      "learning_rate": 0.00020421052631578948,
+      "loss": 1.128,
+      "step": 122400
+    },
+    {
+      "epoch": 45.81,
+      "learning_rate": 0.00020394736842105262,
+      "loss": 1.1263,
+      "step": 122500
+    },
+    {
+      "epoch": 45.85,
+      "learning_rate": 0.00020368421052631582,
+      "loss": 1.1234,
+      "step": 122600
+    },
+    {
+      "epoch": 45.89,
+      "learning_rate": 0.00020342105263157896,
+      "loss": 1.128,
+      "step": 122700
+    },
+    {
+      "epoch": 45.92,
+      "learning_rate": 0.0002031578947368421,
+      "loss": 1.1291,
+      "step": 122800
+    },
+    {
+      "epoch": 45.96,
+      "learning_rate": 0.00020289473684210527,
+      "loss": 1.1292,
+      "step": 122900
+    },
+    {
+      "epoch": 46.0,
+      "learning_rate": 0.00020263157894736842,
+      "loss": 1.1246,
+      "step": 123000
+    },
+    {
+      "epoch": 46.04,
+      "learning_rate": 0.0002023684210526316,
+      "loss": 1.1327,
+      "step": 123100
+    },
+    {
+      "epoch": 46.07,
+      "learning_rate": 0.00020210526315789473,
+      "loss": 1.1204,
+      "step": 123200
+    },
+    {
+      "epoch": 46.11,
+      "learning_rate": 0.0002018421052631579,
+      "loss": 1.1275,
+      "step": 123300
+    },
+    {
+      "epoch": 46.15,
+      "learning_rate": 0.00020157894736842104,
+      "loss": 1.1235,
+      "step": 123400
+    },
+    {
+      "epoch": 46.19,
+      "learning_rate": 0.00020131578947368421,
+      "loss": 1.1244,
+      "step": 123500
+    },
+    {
+      "epoch": 46.22,
+      "learning_rate": 0.00020105263157894738,
+      "loss": 1.1239,
+      "step": 123600
+    },
+    {
+      "epoch": 46.26,
+      "learning_rate": 0.00020078947368421053,
+      "loss": 1.1252,
+      "step": 123700
+    },
+    {
+      "epoch": 46.3,
+      "learning_rate": 0.0002005263157894737,
+      "loss": 1.1249,
+      "step": 123800
+    },
+    {
+      "epoch": 46.34,
+      "learning_rate": 0.00020026315789473684,
+      "loss": 1.1221,
+      "step": 123900
+    },
+    {
+      "epoch": 46.37,
+      "learning_rate": 0.0002,
+      "loss": 1.1267,
+      "step": 124000
+    },
+    {
+      "epoch": 46.41,
+      "learning_rate": 0.00019973684210526315,
+      "loss": 1.1246,
+      "step": 124100
+    },
+    {
+      "epoch": 46.45,
+      "learning_rate": 0.0001994736842105263,
+      "loss": 1.1262,
+      "step": 124200
+    },
+    {
+      "epoch": 46.48,
+      "learning_rate": 0.0001992105263157895,
+      "loss": 1.1257,
+      "step": 124300
+    },
+    {
+      "epoch": 46.52,
+      "learning_rate": 0.00019894736842105264,
+      "loss": 1.1236,
+      "step": 124400
+    },
+    {
+      "epoch": 46.56,
+      "learning_rate": 0.0001986842105263158,
+      "loss": 1.1255,
+      "step": 124500
+    },
+    {
+      "epoch": 46.6,
+      "learning_rate": 0.00019842105263157895,
+      "loss": 1.1282,
+      "step": 124600
+    },
+    {
+      "epoch": 46.63,
+      "learning_rate": 0.00019815789473684212,
+      "loss": 1.1227,
+      "step": 124700
+    },
+    {
+      "epoch": 46.67,
+      "learning_rate": 0.00019789473684210526,
+      "loss": 1.1252,
+      "step": 124800
+    },
+    {
+      "epoch": 46.71,
+      "learning_rate": 0.0001976315789473684,
+      "loss": 1.1229,
+      "step": 124900
+    },
+    {
+      "epoch": 46.75,
+      "learning_rate": 0.00019736842105263157,
+      "loss": 1.1237,
+      "step": 125000
+    },
+    {
+      "epoch": 46.78,
+      "learning_rate": 0.00019710526315789474,
+      "loss": 1.1217,
+      "step": 125100
+    },
+    {
+      "epoch": 46.82,
+      "learning_rate": 0.00019684210526315791,
+      "loss": 1.1219,
+      "step": 125200
+    },
+    {
+      "epoch": 46.86,
+      "learning_rate": 0.00019657894736842106,
+      "loss": 1.1234,
+      "step": 125300
+    },
+    {
+      "epoch": 46.9,
+      "learning_rate": 0.00019631578947368423,
+      "loss": 1.1233,
+      "step": 125400
+    },
+    {
+      "epoch": 46.93,
+      "learning_rate": 0.00019605263157894737,
+      "loss": 1.1198,
+      "step": 125500
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 0.0001957894736842105,
+      "loss": 1.1238,
+      "step": 125600
+    },
+    {
+      "epoch": 47.01,
+      "learning_rate": 0.00019552631578947368,
+      "loss": 1.1341,
+      "step": 125700
+    },
+    {
+      "epoch": 47.05,
+      "learning_rate": 0.00019526315789473683,
+      "loss": 1.1211,
+      "step": 125800
+    },
+    {
+      "epoch": 47.08,
+      "learning_rate": 0.00019500000000000002,
+      "loss": 1.1213,
+      "step": 125900
+    },
+    {
+      "epoch": 47.12,
+      "learning_rate": 0.00019473684210526317,
+      "loss": 1.1205,
+      "step": 126000
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 0.00019447368421052634,
+      "loss": 1.1181,
+      "step": 126100
+    },
+    {
+      "epoch": 47.2,
+      "learning_rate": 0.00019421052631578948,
+      "loss": 1.118,
+      "step": 126200
+    },
+    {
+      "epoch": 47.23,
+      "learning_rate": 0.00019394736842105262,
+      "loss": 1.1193,
+      "step": 126300
+    },
+    {
+      "epoch": 47.27,
+      "learning_rate": 0.0001936842105263158,
+      "loss": 1.1216,
+      "step": 126400
+    },
+    {
+      "epoch": 47.31,
+      "learning_rate": 0.00019342105263157894,
+      "loss": 1.1221,
+      "step": 126500
+    },
+    {
+      "epoch": 47.34,
+      "learning_rate": 0.0001931578947368421,
+      "loss": 1.1226,
+      "step": 126600
+    },
+    {
+      "epoch": 47.38,
+      "learning_rate": 0.00019289473684210525,
+      "loss": 1.1211,
+      "step": 126700
+    },
+    {
+      "epoch": 47.42,
+      "learning_rate": 0.00019263157894736845,
+      "loss": 1.1194,
+      "step": 126800
+    },
+    {
+      "epoch": 47.46,
+      "learning_rate": 0.0001923684210526316,
+      "loss": 1.1192,
+      "step": 126900
+    },
+    {
+      "epoch": 47.49,
+      "learning_rate": 0.00019210526315789473,
+      "loss": 1.1191,
+      "step": 127000
+    },
+    {
+      "epoch": 47.53,
+      "learning_rate": 0.0001918421052631579,
+      "loss": 1.1181,
+      "step": 127100
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 0.00019157894736842104,
+      "loss": 1.1217,
+      "step": 127200
+    },
+    {
+      "epoch": 47.61,
+      "learning_rate": 0.00019131578947368421,
+      "loss": 1.1175,
+      "step": 127300
+    },
+    {
+      "epoch": 47.64,
+      "learning_rate": 0.00019105263157894736,
+      "loss": 1.1186,
+      "step": 127400
+    },
+    {
+      "epoch": 47.68,
+      "learning_rate": 0.00019078947368421053,
+      "loss": 1.1204,
+      "step": 127500
+    },
+    {
+      "epoch": 47.72,
+      "learning_rate": 0.0001905263157894737,
+      "loss": 1.1218,
+      "step": 127600
+    },
+    {
+      "epoch": 47.76,
+      "learning_rate": 0.00019026315789473684,
+      "loss": 1.1205,
+      "step": 127700
+    },
+    {
+      "epoch": 47.79,
+      "learning_rate": 0.00019,
+      "loss": 1.1204,
+      "step": 127800
+    },
+    {
+      "epoch": 47.83,
+      "learning_rate": 0.00018973684210526315,
+      "loss": 1.1189,
+      "step": 127900
+    },
+    {
+      "epoch": 47.87,
+      "learning_rate": 0.00018947368421052632,
+      "loss": 1.1177,
+      "step": 128000
+    },
+    {
+      "epoch": 47.91,
+      "learning_rate": 0.00018921052631578947,
+      "loss": 1.1215,
+      "step": 128100
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 0.00018894736842105264,
+      "loss": 1.1202,
+      "step": 128200
+    },
+    {
+      "epoch": 47.98,
+      "learning_rate": 0.00018868421052631578,
+      "loss": 1.1197,
+      "step": 128300
+    },
+    {
+      "epoch": 48.02,
+      "learning_rate": 0.00018842105263157895,
+      "loss": 1.1234,
+      "step": 128400
+    },
+    {
+      "epoch": 48.06,
+      "learning_rate": 0.00018815789473684212,
+      "loss": 1.1161,
+      "step": 128500
+    },
+    {
+      "epoch": 48.09,
+      "learning_rate": 0.00018789473684210526,
+      "loss": 1.1169,
+      "step": 128600
+    },
+    {
+      "epoch": 48.13,
+      "learning_rate": 0.00018763157894736843,
+      "loss": 1.1157,
+      "step": 128700
+    },
+    {
+      "epoch": 48.17,
+      "learning_rate": 0.00018736842105263158,
+      "loss": 1.114,
+      "step": 128800
+    },
+    {
+      "epoch": 48.2,
+      "learning_rate": 0.00018710526315789475,
+      "loss": 1.1164,
+      "step": 128900
+    },
+    {
+      "epoch": 48.24,
+      "learning_rate": 0.0001868421052631579,
+      "loss": 1.1126,
+      "step": 129000
+    },
+    {
+      "epoch": 48.28,
+      "learning_rate": 0.00018657894736842106,
+      "loss": 1.1171,
+      "step": 129100
+    },
+    {
+      "epoch": 48.32,
+      "learning_rate": 0.00018631578947368423,
+      "loss": 1.1181,
+      "step": 129200
+    },
+    {
+      "epoch": 48.35,
+      "learning_rate": 0.00018605263157894737,
+      "loss": 1.1174,
+      "step": 129300
+    },
+    {
+      "epoch": 48.39,
+      "learning_rate": 0.00018578947368421054,
+      "loss": 1.1141,
+      "step": 129400
+    },
+    {
+      "epoch": 48.43,
+      "learning_rate": 0.00018552631578947368,
+      "loss": 1.1118,
+      "step": 129500
+    },
+    {
+      "epoch": 48.47,
+      "learning_rate": 0.00018526315789473685,
+      "loss": 1.1152,
+      "step": 129600
+    },
+    {
+      "epoch": 48.5,
+      "learning_rate": 0.000185,
+      "loss": 1.1153,
+      "step": 129700
+    },
+    {
+      "epoch": 48.54,
+      "learning_rate": 0.00018473684210526317,
+      "loss": 1.1143,
+      "step": 129800
+    },
+    {
+      "epoch": 48.58,
+      "learning_rate": 0.0001844736842105263,
+      "loss": 1.1117,
+      "step": 129900
+    },
+    {
+      "epoch": 48.62,
+      "learning_rate": 0.00018421052631578948,
+      "loss": 1.1195,
+      "step": 130000
+    },
+    {
+      "epoch": 48.65,
+      "learning_rate": 0.00018394736842105265,
+      "loss": 1.1153,
+      "step": 130100
+    },
+    {
+      "epoch": 48.69,
+      "learning_rate": 0.0001836842105263158,
+      "loss": 1.1142,
+      "step": 130200
+    },
+    {
+      "epoch": 48.73,
+      "learning_rate": 0.00018342105263157896,
+      "loss": 1.1124,
+      "step": 130300
+    },
+    {
+      "epoch": 48.77,
+      "learning_rate": 0.0001831578947368421,
+      "loss": 1.118,
+      "step": 130400
+    },
+    {
+      "epoch": 48.8,
+      "learning_rate": 0.00018289473684210528,
+      "loss": 1.1149,
+      "step": 130500
+    },
+    {
+      "epoch": 48.84,
+      "learning_rate": 0.00018263157894736842,
+      "loss": 1.1127,
+      "step": 130600
+    },
+    {
+      "epoch": 48.88,
+      "learning_rate": 0.00018236842105263156,
+      "loss": 1.1138,
+      "step": 130700
+    },
+    {
+      "epoch": 48.92,
+      "learning_rate": 0.00018210526315789476,
+      "loss": 1.118,
+      "step": 130800
+    },
+    {
+      "epoch": 48.95,
+      "learning_rate": 0.0001818421052631579,
+      "loss": 1.1125,
+      "step": 130900
+    },
+    {
+      "epoch": 48.99,
+      "learning_rate": 0.00018157894736842107,
+      "loss": 1.1168,
+      "step": 131000
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 0.00018131578947368422,
+      "loss": 1.1163,
+      "step": 131100
+    },
+    {
+      "epoch": 49.07,
+      "learning_rate": 0.00018105263157894739,
+      "loss": 1.1119,
+      "step": 131200
+    },
+    {
+      "epoch": 49.1,
+      "learning_rate": 0.00018078947368421053,
+      "loss": 1.1087,
+      "step": 131300
+    },
+    {
+      "epoch": 49.14,
+      "learning_rate": 0.00018052631578947367,
+      "loss": 1.1065,
+      "step": 131400
+    },
+    {
+      "epoch": 49.18,
+      "learning_rate": 0.00018026315789473684,
+      "loss": 1.108,
+      "step": 131500
+    },
+    {
+      "epoch": 49.21,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 1.1118,
+      "step": 131600
+    },
+    {
+      "epoch": 49.25,
+      "learning_rate": 0.00017973684210526318,
+      "loss": 1.109,
+      "step": 131700
+    },
+    {
+      "epoch": 49.29,
+      "learning_rate": 0.00017947368421052632,
+      "loss": 1.1123,
+      "step": 131800
+    },
+    {
+      "epoch": 49.33,
+      "learning_rate": 0.0001792105263157895,
+      "loss": 1.11,
+      "step": 131900
+    },
+    {
+      "epoch": 49.36,
+      "learning_rate": 0.00017894736842105264,
+      "loss": 1.1092,
+      "step": 132000
+    },
+    {
+      "epoch": 49.4,
+      "learning_rate": 0.00017868421052631578,
+      "loss": 1.1135,
+      "step": 132100
+    },
+    {
+      "epoch": 49.44,
+      "learning_rate": 0.00017842105263157895,
+      "loss": 1.1108,
+      "step": 132200
+    },
+    {
+      "epoch": 49.48,
+      "learning_rate": 0.0001781578947368421,
+      "loss": 1.1122,
+      "step": 132300
+    },
+    {
+      "epoch": 49.51,
+      "learning_rate": 0.00017789473684210526,
+      "loss": 1.1104,
+      "step": 132400
+    },
+    {
+      "epoch": 49.55,
+      "learning_rate": 0.00017763157894736843,
+      "loss": 1.1157,
+      "step": 132500
+    },
+    {
+      "epoch": 49.59,
+      "learning_rate": 0.0001773684210526316,
+      "loss": 1.1118,
+      "step": 132600
+    },
+    {
+      "epoch": 49.63,
+      "learning_rate": 0.00017710526315789475,
+      "loss": 1.1077,
+      "step": 132700
+    },
+    {
+      "epoch": 49.66,
+      "learning_rate": 0.0001768421052631579,
+      "loss": 1.1078,
+      "step": 132800
+    },
+    {
+      "epoch": 49.7,
+      "learning_rate": 0.00017657894736842106,
+      "loss": 1.1132,
+      "step": 132900
+    },
+    {
+      "epoch": 49.74,
+      "learning_rate": 0.0001763157894736842,
+      "loss": 1.1107,
+      "step": 133000
+    },
+    {
+      "epoch": 49.78,
+      "learning_rate": 0.00017605263157894737,
+      "loss": 1.1095,
+      "step": 133100
+    },
+    {
+      "epoch": 49.81,
+      "learning_rate": 0.00017578947368421052,
+      "loss": 1.1103,
+      "step": 133200
+    },
+    {
+      "epoch": 49.85,
+      "learning_rate": 0.0001755263157894737,
+      "loss": 1.1128,
+      "step": 133300
+    },
+    {
+      "epoch": 49.89,
+      "learning_rate": 0.00017526315789473686,
+      "loss": 1.1077,
+      "step": 133400
+    },
+    {
+      "epoch": 49.93,
+      "learning_rate": 0.000175,
+      "loss": 1.1111,
+      "step": 133500
+    },
+    {
+      "epoch": 49.96,
+      "learning_rate": 0.00017473684210526317,
+      "loss": 1.1069,
+      "step": 133600
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 0.0001744736842105263,
+      "loss": 1.1077,
+      "step": 133700
+    },
+    {
+      "epoch": 50.04,
+      "learning_rate": 0.00017421052631578948,
+      "loss": 1.1125,
+      "step": 133800
+    },
+    {
+      "epoch": 50.07,
+      "learning_rate": 0.00017394736842105262,
+      "loss": 1.1069,
+      "step": 133900
+    },
+    {
+      "epoch": 50.11,
+      "learning_rate": 0.0001736842105263158,
+      "loss": 1.1044,
+      "step": 134000
+    },
+    {
+      "epoch": 50.15,
+      "learning_rate": 0.00017342105263157896,
+      "loss": 1.1047,
+      "step": 134100
+    },
+    {
+      "epoch": 50.19,
+      "learning_rate": 0.0001731578947368421,
+      "loss": 1.1071,
+      "step": 134200
+    },
+    {
+      "epoch": 50.22,
+      "learning_rate": 0.00017289473684210528,
+      "loss": 1.1068,
+      "step": 134300
+    },
+    {
+      "epoch": 50.26,
+      "learning_rate": 0.00017263157894736842,
+      "loss": 1.1074,
+      "step": 134400
+    },
+    {
+      "epoch": 50.3,
+      "learning_rate": 0.0001723684210526316,
+      "loss": 1.105,
+      "step": 134500
+    },
+    {
+      "epoch": 50.34,
+      "learning_rate": 0.00017210526315789473,
+      "loss": 1.1074,
+      "step": 134600
+    },
+    {
+      "epoch": 50.37,
+      "learning_rate": 0.0001718421052631579,
+      "loss": 1.1067,
+      "step": 134700
+    },
+    {
+      "epoch": 50.41,
+      "learning_rate": 0.00017157894736842105,
+      "loss": 1.1054,
+      "step": 134800
+    },
+    {
+      "epoch": 50.45,
+      "learning_rate": 0.0001713157894736842,
+      "loss": 1.1087,
+      "step": 134900
+    },
+    {
+      "epoch": 50.49,
+      "learning_rate": 0.00017105263157894739,
+      "loss": 1.1037,
+      "step": 135000
+    },
+    {
+      "epoch": 50.52,
+      "learning_rate": 0.00017078947368421053,
+      "loss": 1.1052,
+      "step": 135100
+    },
+    {
+      "epoch": 50.56,
+      "learning_rate": 0.0001705263157894737,
+      "loss": 1.1086,
+      "step": 135200
+    },
+    {
+      "epoch": 50.6,
+      "learning_rate": 0.00017026315789473684,
+      "loss": 1.1066,
+      "step": 135300
+    },
+    {
+      "epoch": 50.64,
+      "learning_rate": 0.00017,
+      "loss": 1.1086,
+      "step": 135400
+    },
+    {
+      "epoch": 50.67,
+      "learning_rate": 0.00016973684210526316,
+      "loss": 1.1059,
+      "step": 135500
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 0.0001694736842105263,
+      "loss": 1.1066,
+      "step": 135600
+    },
+    {
+      "epoch": 50.75,
+      "learning_rate": 0.00016921052631578947,
+      "loss": 1.1048,
+      "step": 135700
+    },
+    {
+      "epoch": 50.79,
+      "learning_rate": 0.00016894736842105264,
+      "loss": 1.1068,
+      "step": 135800
+    },
+    {
+      "epoch": 50.82,
+      "learning_rate": 0.0001686842105263158,
+      "loss": 1.103,
+      "step": 135900
+    },
+    {
+      "epoch": 50.86,
+      "learning_rate": 0.00016842105263157895,
+      "loss": 1.1026,
+      "step": 136000
+    },
+    {
+      "epoch": 50.9,
+      "learning_rate": 0.00016815789473684212,
+      "loss": 1.1075,
+      "step": 136100
+    },
+    {
+      "epoch": 50.93,
+      "learning_rate": 0.00016789473684210526,
+      "loss": 1.1043,
+      "step": 136200
+    },
+    {
+      "epoch": 50.97,
+      "learning_rate": 0.0001676315789473684,
+      "loss": 1.1071,
+      "step": 136300
+    },
+    {
+      "epoch": 51.01,
+      "learning_rate": 0.00016736842105263158,
+      "loss": 1.1114,
+      "step": 136400
+    },
+    {
+      "epoch": 51.05,
+      "learning_rate": 0.00016710526315789472,
+      "loss": 1.0996,
+      "step": 136500
+    },
+    {
+      "epoch": 51.08,
+      "learning_rate": 0.00016684210526315792,
+      "loss": 1.1027,
+      "step": 136600
+    },
+    {
+      "epoch": 51.12,
+      "learning_rate": 0.00016657894736842106,
+      "loss": 1.1027,
+      "step": 136700
+    },
+    {
+      "epoch": 51.16,
+      "learning_rate": 0.00016631578947368423,
+      "loss": 1.1043,
+      "step": 136800
+    },
+    {
+      "epoch": 51.2,
+      "learning_rate": 0.00016605263157894737,
+      "loss": 1.1021,
+      "step": 136900
+    },
+    {
+      "epoch": 51.23,
+      "learning_rate": 0.00016578947368421052,
+      "loss": 1.0988,
+      "step": 137000
+    },
+    {
+      "epoch": 51.27,
+      "learning_rate": 0.00016552631578947369,
+      "loss": 1.0999,
+      "step": 137100
+    },
+    {
+      "epoch": 51.31,
+      "learning_rate": 0.00016526315789473683,
+      "loss": 1.1028,
+      "step": 137200
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 0.000165,
+      "loss": 1.1025,
+      "step": 137300
+    },
+    {
+      "epoch": 51.38,
+      "learning_rate": 0.00016473684210526317,
+      "loss": 1.1005,
+      "step": 137400
+    },
+    {
+      "epoch": 51.42,
+      "learning_rate": 0.00016447368421052634,
+      "loss": 1.1003,
+      "step": 137500
+    },
+    {
+      "epoch": 51.46,
+      "learning_rate": 0.00016421052631578948,
+      "loss": 1.1005,
+      "step": 137600
+    },
+    {
+      "epoch": 51.5,
+      "learning_rate": 0.00016394736842105263,
+      "loss": 1.1003,
+      "step": 137700
+    },
+    {
+      "epoch": 51.53,
+      "learning_rate": 0.0001636842105263158,
+      "loss": 1.1032,
+      "step": 137800
+    },
+    {
+      "epoch": 51.57,
+      "learning_rate": 0.00016342105263157894,
+      "loss": 1.1027,
+      "step": 137900
+    },
+    {
+      "epoch": 51.61,
+      "learning_rate": 0.0001631578947368421,
+      "loss": 1.102,
+      "step": 138000
+    },
+    {
+      "epoch": 51.65,
+      "learning_rate": 0.00016289473684210525,
+      "loss": 1.0996,
+      "step": 138100
+    },
+    {
+      "epoch": 51.68,
+      "learning_rate": 0.00016263157894736845,
+      "loss": 1.1009,
+      "step": 138200
+    },
+    {
+      "epoch": 51.72,
+      "learning_rate": 0.0001623684210526316,
+      "loss": 1.1014,
+      "step": 138300
+    },
+    {
+      "epoch": 51.76,
+      "learning_rate": 0.00016210526315789473,
+      "loss": 1.1025,
+      "step": 138400
+    },
+    {
+      "epoch": 51.79,
+      "learning_rate": 0.0001618421052631579,
+      "loss": 1.0997,
+      "step": 138500
+    },
+    {
+      "epoch": 51.83,
+      "learning_rate": 0.00016157894736842105,
+      "loss": 1.1029,
+      "step": 138600
+    },
+    {
+      "epoch": 51.87,
+      "learning_rate": 0.00016131578947368422,
+      "loss": 1.1045,
+      "step": 138700
+    },
+    {
+      "epoch": 51.91,
+      "learning_rate": 0.00016105263157894736,
+      "loss": 1.103,
+      "step": 138800
+    },
+    {
+      "epoch": 51.94,
+      "learning_rate": 0.00016078947368421053,
+      "loss": 1.099,
+      "step": 138900
+    },
+    {
+      "epoch": 51.98,
+      "learning_rate": 0.0001605263157894737,
+      "loss": 1.0998,
+      "step": 139000
+    },
+    {
+      "epoch": 52.02,
+      "learning_rate": 0.00016026315789473684,
+      "loss": 1.103,
+      "step": 139100
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 0.00016,
+      "loss": 1.0982,
+      "step": 139200
+    },
+    {
+      "epoch": 52.09,
+      "learning_rate": 0.00015973684210526316,
+      "loss": 1.0975,
+      "step": 139300
+    },
+    {
+      "epoch": 52.13,
+      "learning_rate": 0.00015947368421052633,
+      "loss": 1.0981,
+      "step": 139400
+    },
+    {
+      "epoch": 52.17,
+      "learning_rate": 0.00015921052631578947,
+      "loss": 1.0989,
+      "step": 139500
+    },
+    {
+      "epoch": 52.21,
+      "learning_rate": 0.00015894736842105264,
+      "loss": 1.0985,
+      "step": 139600
+    },
+    {
+      "epoch": 52.24,
+      "learning_rate": 0.00015868421052631578,
+      "loss": 1.0977,
+      "step": 139700
+    },
+    {
+      "epoch": 52.28,
+      "learning_rate": 0.00015842105263157892,
+      "loss": 1.0966,
+      "step": 139800
+    },
+    {
+      "epoch": 52.32,
+      "learning_rate": 0.00015815789473684212,
+      "loss": 1.098,
+      "step": 139900
+    },
+    {
+      "epoch": 52.36,
+      "learning_rate": 0.00015789473684210527,
+      "loss": 1.0939,
+      "step": 140000
+    },
+    {
+      "epoch": 52.39,
+      "learning_rate": 0.00015763157894736844,
+      "loss": 1.0951,
+      "step": 140100
+    },
+    {
+      "epoch": 52.43,
+      "learning_rate": 0.00015736842105263158,
+      "loss": 1.0979,
+      "step": 140200
+    },
+    {
+      "epoch": 52.47,
+      "learning_rate": 0.00015710526315789475,
+      "loss": 1.0985,
+      "step": 140300
+    },
+    {
+      "epoch": 52.51,
+      "learning_rate": 0.0001568421052631579,
+      "loss": 1.0971,
+      "step": 140400
+    },
+    {
+      "epoch": 52.54,
+      "learning_rate": 0.00015657894736842103,
+      "loss": 1.0961,
+      "step": 140500
+    },
+    {
+      "epoch": 52.58,
+      "learning_rate": 0.0001563157894736842,
+      "loss": 1.0978,
+      "step": 140600
+    },
+    {
+      "epoch": 52.62,
+      "learning_rate": 0.00015605263157894737,
+      "loss": 1.0953,
+      "step": 140700
+    },
+    {
+      "epoch": 52.66,
+      "learning_rate": 0.00015578947368421054,
+      "loss": 1.0977,
+      "step": 140800
+    },
+    {
+      "epoch": 52.69,
+      "learning_rate": 0.0001555263157894737,
+      "loss": 1.0979,
+      "step": 140900
+    },
+    {
+      "epoch": 52.73,
+      "learning_rate": 0.00015526315789473686,
+      "loss": 1.0952,
+      "step": 141000
+    },
+    {
+      "epoch": 52.77,
+      "learning_rate": 0.000155,
+      "loss": 1.0987,
+      "step": 141100
+    },
+    {
+      "epoch": 52.8,
+      "learning_rate": 0.00015473684210526314,
+      "loss": 1.0992,
+      "step": 141200
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 0.0001544736842105263,
+      "loss": 1.0926,
+      "step": 141300
+    },
+    {
+      "epoch": 52.88,
+      "learning_rate": 0.00015421052631578946,
+      "loss": 1.0964,
+      "step": 141400
+    },
+    {
+      "epoch": 52.92,
+      "learning_rate": 0.00015394736842105265,
+      "loss": 1.0946,
+      "step": 141500
+    },
+    {
+      "epoch": 52.95,
+      "learning_rate": 0.0001536842105263158,
+      "loss": 1.0996,
+      "step": 141600
+    },
+    {
+      "epoch": 52.99,
+      "learning_rate": 0.00015342105263157897,
+      "loss": 1.0962,
+      "step": 141700
+    },
+    {
+      "epoch": 53.03,
+      "learning_rate": 0.0001531578947368421,
+      "loss": 1.0983,
+      "step": 141800
+    },
+    {
+      "epoch": 53.07,
+      "learning_rate": 0.00015289473684210525,
+      "loss": 1.0905,
+      "step": 141900
+    },
+    {
+      "epoch": 53.1,
+      "learning_rate": 0.00015263157894736842,
+      "loss": 1.0931,
+      "step": 142000
+    },
+    {
+      "epoch": 53.14,
+      "learning_rate": 0.00015236842105263156,
+      "loss": 1.0946,
+      "step": 142100
+    },
+    {
+      "epoch": 53.18,
+      "learning_rate": 0.00015210526315789473,
+      "loss": 1.0962,
+      "step": 142200
+    },
+    {
+      "epoch": 53.22,
+      "learning_rate": 0.0001518421052631579,
+      "loss": 1.0924,
+      "step": 142300
+    },
+    {
+      "epoch": 53.25,
+      "learning_rate": 0.00015157894736842108,
+      "loss": 1.0948,
+      "step": 142400
+    },
+    {
+      "epoch": 53.29,
+      "learning_rate": 0.00015131578947368422,
+      "loss": 1.0933,
+      "step": 142500
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 0.00015105263157894736,
+      "loss": 1.0948,
+      "step": 142600
+    },
+    {
+      "epoch": 53.37,
+      "learning_rate": 0.00015078947368421053,
+      "loss": 1.093,
+      "step": 142700
+    },
+    {
+      "epoch": 53.4,
+      "learning_rate": 0.00015052631578947367,
+      "loss": 1.094,
+      "step": 142800
+    },
+    {
+      "epoch": 53.44,
+      "learning_rate": 0.00015026315789473684,
+      "loss": 1.0982,
+      "step": 142900
+    },
+    {
+      "epoch": 53.48,
+      "learning_rate": 0.00015,
+      "loss": 1.0937,
+      "step": 143000
+    },
+    {
+      "epoch": 53.52,
+      "learning_rate": 0.00014973684210526318,
+      "loss": 1.0956,
+      "step": 143100
+    },
+    {
+      "epoch": 53.55,
+      "learning_rate": 0.00014947368421052633,
+      "loss": 1.0937,
+      "step": 143200
+    },
+    {
+      "epoch": 53.59,
+      "learning_rate": 0.00014921052631578947,
+      "loss": 1.0906,
+      "step": 143300
+    },
+    {
+      "epoch": 53.63,
+      "learning_rate": 0.00014894736842105264,
+      "loss": 1.0904,
+      "step": 143400
+    },
+    {
+      "epoch": 53.66,
+      "learning_rate": 0.00014868421052631578,
+      "loss": 1.0892,
+      "step": 143500
+    },
+    {
+      "epoch": 53.7,
+      "learning_rate": 0.00014842105263157895,
+      "loss": 1.0928,
+      "step": 143600
+    },
+    {
+      "epoch": 53.74,
+      "learning_rate": 0.0001481578947368421,
+      "loss": 1.0916,
+      "step": 143700
+    },
+    {
+      "epoch": 53.78,
+      "learning_rate": 0.00014789473684210527,
+      "loss": 1.094,
+      "step": 143800
+    },
+    {
+      "epoch": 53.81,
+      "learning_rate": 0.0001476315789473684,
+      "loss": 1.0913,
+      "step": 143900
+    },
+    {
+      "epoch": 53.85,
+      "learning_rate": 0.00014736842105263158,
+      "loss": 1.0908,
+      "step": 144000
+    },
+    {
+      "epoch": 53.89,
+      "learning_rate": 0.00014710526315789475,
+      "loss": 1.0909,
+      "step": 144100
+    },
+    {
+      "epoch": 53.93,
+      "learning_rate": 0.0001468421052631579,
+      "loss": 1.0936,
+      "step": 144200
+    },
+    {
+      "epoch": 53.96,
+      "learning_rate": 0.00014657894736842106,
+      "loss": 1.0932,
+      "step": 144300
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 0.0001463157894736842,
+      "loss": 1.1001,
+      "step": 144400
+    },
+    {
+      "epoch": 54.04,
+      "learning_rate": 0.00014605263157894737,
+      "loss": 1.0875,
+      "step": 144500
+    },
+    {
+      "epoch": 54.08,
+      "learning_rate": 0.00014578947368421052,
+      "loss": 1.0868,
+      "step": 144600
+    },
+    {
+      "epoch": 54.11,
+      "learning_rate": 0.0001455263157894737,
+      "loss": 1.0879,
+      "step": 144700
+    },
+    {
+      "epoch": 54.15,
+      "learning_rate": 0.00014526315789473686,
+      "loss": 1.088,
+      "step": 144800
+    },
+    {
+      "epoch": 54.19,
+      "learning_rate": 0.000145,
+      "loss": 1.0886,
+      "step": 144900
+    },
+    {
+      "epoch": 54.23,
+      "learning_rate": 0.00014473684210526317,
+      "loss": 1.0897,
+      "step": 145000
+    },
+    {
+      "epoch": 54.26,
+      "learning_rate": 0.00014447368421052631,
+      "loss": 1.091,
+      "step": 145100
+    },
+    {
+      "epoch": 54.3,
+      "learning_rate": 0.00014421052631578948,
+      "loss": 1.0889,
+      "step": 145200
+    },
+    {
+      "epoch": 54.34,
+      "learning_rate": 0.00014394736842105263,
+      "loss": 1.0895,
+      "step": 145300
+    },
+    {
+      "epoch": 54.38,
+      "learning_rate": 0.0001436842105263158,
+      "loss": 1.0869,
+      "step": 145400
+    },
+    {
+      "epoch": 54.41,
+      "learning_rate": 0.00014342105263157894,
+      "loss": 1.0913,
+      "step": 145500
+    },
+    {
+      "epoch": 54.45,
+      "learning_rate": 0.0001431578947368421,
+      "loss": 1.0896,
+      "step": 145600
+    },
+    {
+      "epoch": 54.49,
+      "learning_rate": 0.00014289473684210528,
+      "loss": 1.0877,
+      "step": 145700
+    },
+    {
+      "epoch": 54.52,
+      "learning_rate": 0.00014263157894736842,
+      "loss": 1.0891,
+      "step": 145800
+    },
+    {
+      "epoch": 54.56,
+      "learning_rate": 0.0001423684210526316,
+      "loss": 1.0922,
+      "step": 145900
+    },
+    {
+      "epoch": 54.6,
+      "learning_rate": 0.00014210526315789474,
+      "loss": 1.0908,
+      "step": 146000
+    },
+    {
+      "epoch": 54.64,
+      "learning_rate": 0.0001418421052631579,
+      "loss": 1.0851,
+      "step": 146100
+    },
+    {
+      "epoch": 54.67,
+      "learning_rate": 0.00014157894736842105,
+      "loss": 1.091,
+      "step": 146200
+    },
+    {
+      "epoch": 54.71,
+      "learning_rate": 0.0001413157894736842,
+      "loss": 1.0902,
+      "step": 146300
+    },
+    {
+      "epoch": 54.75,
+      "learning_rate": 0.0001410526315789474,
+      "loss": 1.0858,
+      "step": 146400
+    },
+    {
+      "epoch": 54.79,
+      "learning_rate": 0.00014078947368421053,
+      "loss": 1.0897,
+      "step": 146500
+    },
+    {
+      "epoch": 54.82,
+      "learning_rate": 0.0001405263157894737,
+      "loss": 1.0896,
+      "step": 146600
+    },
+    {
+      "epoch": 54.86,
+      "learning_rate": 0.00014026315789473684,
+      "loss": 1.0879,
+      "step": 146700
+    },
+    {
+      "epoch": 54.9,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 1.0868,
+      "step": 146800
+    },
+    {
+      "epoch": 54.94,
+      "learning_rate": 0.00013973684210526316,
+      "loss": 1.087,
+      "step": 146900
+    },
+    {
+      "epoch": 54.97,
+      "learning_rate": 0.0001394736842105263,
+      "loss": 1.0872,
+      "step": 147000
+    },
+    {
+      "epoch": 55.01,
+      "learning_rate": 0.00013921052631578947,
+      "loss": 1.0933,
+      "step": 147100
+    },
+    {
+      "epoch": 55.05,
+      "learning_rate": 0.00013894736842105264,
+      "loss": 1.0839,
+      "step": 147200
+    },
+    {
+      "epoch": 55.09,
+      "learning_rate": 0.0001386842105263158,
+      "loss": 1.0824,
+      "step": 147300
+    },
+    {
+      "epoch": 55.12,
+      "learning_rate": 0.00013842105263157895,
+      "loss": 1.0868,
+      "step": 147400
+    },
+    {
+      "epoch": 55.16,
+      "learning_rate": 0.00013815789473684212,
+      "loss": 1.0843,
+      "step": 147500
+    },
+    {
+      "epoch": 55.2,
+      "learning_rate": 0.00013789473684210527,
+      "loss": 1.0854,
+      "step": 147600
+    },
+    {
+      "epoch": 55.24,
+      "learning_rate": 0.0001376315789473684,
+      "loss": 1.0847,
+      "step": 147700
+    },
+    {
+      "epoch": 55.27,
+      "learning_rate": 0.00013736842105263158,
+      "loss": 1.0852,
+      "step": 147800
+    },
+    {
+      "epoch": 55.31,
+      "learning_rate": 0.00013710526315789472,
+      "loss": 1.0863,
+      "step": 147900
+    },
+    {
+      "epoch": 55.35,
+      "learning_rate": 0.00013684210526315792,
+      "loss": 1.0825,
+      "step": 148000
+    },
+    {
+      "epoch": 55.39,
+      "learning_rate": 0.00013657894736842106,
+      "loss": 1.0849,
+      "step": 148100
+    },
+    {
+      "epoch": 55.42,
+      "learning_rate": 0.00013631578947368423,
+      "loss": 1.0845,
+      "step": 148200
+    },
+    {
+      "epoch": 55.46,
+      "learning_rate": 0.00013605263157894738,
+      "loss": 1.0839,
+      "step": 148300
+    },
+    {
+      "epoch": 55.5,
+      "learning_rate": 0.00013578947368421052,
+      "loss": 1.0857,
+      "step": 148400
+    },
+    {
+      "epoch": 55.53,
+      "learning_rate": 0.0001355263157894737,
+      "loss": 1.0835,
+      "step": 148500
+    },
+    {
+      "epoch": 55.57,
+      "learning_rate": 0.00013526315789473683,
+      "loss": 1.0816,
+      "step": 148600
+    },
+    {
+      "epoch": 55.61,
+      "learning_rate": 0.000135,
+      "loss": 1.0826,
+      "step": 148700
+    },
+    {
+      "epoch": 55.65,
+      "learning_rate": 0.00013473684210526314,
+      "loss": 1.0853,
+      "step": 148800
+    },
+    {
+      "epoch": 55.68,
+      "learning_rate": 0.00013447368421052634,
+      "loss": 1.0853,
+      "step": 148900
+    },
+    {
+      "epoch": 55.72,
+      "learning_rate": 0.00013421052631578948,
+      "loss": 1.0859,
+      "step": 149000
+    },
+    {
+      "epoch": 55.76,
+      "learning_rate": 0.00013394736842105263,
+      "loss": 1.0852,
+      "step": 149100
+    },
+    {
+      "epoch": 55.8,
+      "learning_rate": 0.0001336842105263158,
+      "loss": 1.0814,
+      "step": 149200
+    },
+    {
+      "epoch": 55.83,
+      "learning_rate": 0.00013342105263157894,
+      "loss": 1.0793,
+      "step": 149300
+    },
+    {
+      "epoch": 55.87,
+      "learning_rate": 0.0001331578947368421,
+      "loss": 1.0802,
+      "step": 149400
+    },
+    {
+      "epoch": 55.91,
+      "learning_rate": 0.00013289473684210525,
+      "loss": 1.0825,
+      "step": 149500
+    },
+    {
+      "epoch": 55.95,
+      "learning_rate": 0.00013263157894736842,
+      "loss": 1.0839,
+      "step": 149600
+    },
+    {
+      "epoch": 55.98,
+      "learning_rate": 0.0001323684210526316,
+      "loss": 1.085,
+      "step": 149700
+    },
+    {
+      "epoch": 56.02,
+      "learning_rate": 0.00013210526315789474,
+      "loss": 1.086,
+      "step": 149800
+    },
+    {
+      "epoch": 56.06,
+      "learning_rate": 0.0001318421052631579,
+      "loss": 1.0833,
+      "step": 149900
+    },
+    {
+      "epoch": 56.1,
+      "learning_rate": 0.00013157894736842105,
+      "loss": 1.0807,
+      "step": 150000
+    },
+    {
+      "epoch": 56.13,
+      "learning_rate": 0.00013131578947368422,
+      "loss": 1.0792,
+      "step": 150100
+    },
+    {
+      "epoch": 56.17,
+      "learning_rate": 0.00013105263157894736,
+      "loss": 1.0811,
+      "step": 150200
+    },
+    {
+      "epoch": 56.21,
+      "learning_rate": 0.00013078947368421053,
+      "loss": 1.0812,
+      "step": 150300
+    },
+    {
+      "epoch": 56.25,
+      "learning_rate": 0.00013052631578947368,
+      "loss": 1.0785,
+      "step": 150400
+    },
+    {
+      "epoch": 56.28,
+      "learning_rate": 0.00013026315789473685,
+      "loss": 1.0814,
+      "step": 150500
+    },
+    {
+      "epoch": 56.32,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 1.0808,
+      "step": 150600
+    },
+    {
+      "epoch": 56.36,
+      "learning_rate": 0.00012973684210526316,
+      "loss": 1.0788,
+      "step": 150700
+    },
+    {
+      "epoch": 56.39,
+      "learning_rate": 0.00012947368421052633,
+      "loss": 1.0761,
+      "step": 150800
+    },
+    {
+      "epoch": 56.43,
+      "learning_rate": 0.00012921052631578947,
+      "loss": 1.0817,
+      "step": 150900
+    },
+    {
+      "epoch": 56.47,
+      "learning_rate": 0.00012894736842105264,
+      "loss": 1.0802,
+      "step": 151000
+    },
+    {
+      "epoch": 56.51,
+      "learning_rate": 0.00012868421052631578,
+      "loss": 1.0768,
+      "step": 151100
+    },
+    {
+      "epoch": 56.54,
+      "learning_rate": 0.00012842105263157893,
+      "loss": 1.0795,
+      "step": 151200
+    },
+    {
+      "epoch": 56.58,
+      "learning_rate": 0.00012815789473684212,
+      "loss": 1.0857,
+      "step": 151300
+    },
+    {
+      "epoch": 56.62,
+      "learning_rate": 0.00012789473684210527,
+      "loss": 1.0796,
+      "step": 151400
+    },
+    {
+      "epoch": 56.66,
+      "learning_rate": 0.00012763157894736844,
+      "loss": 1.0815,
+      "step": 151500
+    },
+    {
+      "epoch": 56.69,
+      "learning_rate": 0.00012736842105263158,
+      "loss": 1.0795,
+      "step": 151600
+    },
+    {
+      "epoch": 56.73,
+      "learning_rate": 0.00012710526315789475,
+      "loss": 1.0782,
+      "step": 151700
+    },
+    {
+      "epoch": 56.77,
+      "learning_rate": 0.0001268421052631579,
+      "loss": 1.0781,
+      "step": 151800
+    },
+    {
+      "epoch": 56.81,
+      "learning_rate": 0.00012657894736842104,
+      "loss": 1.0825,
+      "step": 151900
+    },
+    {
+      "epoch": 56.84,
+      "learning_rate": 0.0001263157894736842,
+      "loss": 1.0808,
+      "step": 152000
+    },
+    {
+      "epoch": 56.88,
+      "learning_rate": 0.00012605263157894738,
+      "loss": 1.0795,
+      "step": 152100
+    },
+    {
+      "epoch": 56.92,
+      "learning_rate": 0.00012578947368421055,
+      "loss": 1.0802,
+      "step": 152200
+    },
+    {
+      "epoch": 56.96,
+      "learning_rate": 0.0001255263157894737,
+      "loss": 1.08,
+      "step": 152300
+    },
+    {
+      "epoch": 56.99,
+      "learning_rate": 0.00012526315789473686,
+      "loss": 1.0782,
+      "step": 152400
+    },
+    {
+      "epoch": 57.03,
+      "learning_rate": 0.000125,
+      "loss": 1.0813,
+      "step": 152500
+    },
+    {
+      "epoch": 57.07,
+      "learning_rate": 0.00012473684210526317,
+      "loss": 1.08,
+      "step": 152600
+    },
+    {
+      "epoch": 57.11,
+      "learning_rate": 0.00012447368421052632,
+      "loss": 1.0749,
+      "step": 152700
+    },
+    {
+      "epoch": 57.14,
+      "learning_rate": 0.00012421052631578949,
+      "loss": 1.0765,
+      "step": 152800
+    },
+    {
+      "epoch": 57.18,
+      "learning_rate": 0.00012394736842105263,
+      "loss": 1.0759,
+      "step": 152900
+    },
+    {
+      "epoch": 57.22,
+      "learning_rate": 0.0001236842105263158,
+      "loss": 1.0736,
+      "step": 153000
+    },
+    {
+      "epoch": 57.26,
+      "learning_rate": 0.00012342105263157894,
+      "loss": 1.0772,
+      "step": 153100
+    },
+    {
+      "epoch": 57.29,
+      "learning_rate": 0.0001231578947368421,
+      "loss": 1.0763,
+      "step": 153200
+    },
+    {
+      "epoch": 57.33,
+      "learning_rate": 0.00012289473684210525,
+      "loss": 1.0771,
+      "step": 153300
+    },
+    {
+      "epoch": 57.37,
+      "learning_rate": 0.00012263157894736842,
+      "loss": 1.0745,
+      "step": 153400
+    },
+    {
+      "epoch": 57.4,
+      "learning_rate": 0.0001223684210526316,
+      "loss": 1.0791,
+      "step": 153500
+    },
+    {
+      "epoch": 57.44,
+      "learning_rate": 0.00012210526315789474,
+      "loss": 1.0738,
+      "step": 153600
+    },
+    {
+      "epoch": 57.48,
+      "learning_rate": 0.0001218421052631579,
+      "loss": 1.0757,
+      "step": 153700
+    },
+    {
+      "epoch": 57.52,
+      "learning_rate": 0.00012157894736842105,
+      "loss": 1.0767,
+      "step": 153800
+    },
+    {
+      "epoch": 57.55,
+      "learning_rate": 0.00012131578947368421,
+      "loss": 1.0791,
+      "step": 153900
+    },
+    {
+      "epoch": 57.59,
+      "learning_rate": 0.00012105263157894738,
+      "loss": 1.076,
+      "step": 154000
+    },
+    {
+      "epoch": 57.63,
+      "learning_rate": 0.00012078947368421053,
+      "loss": 1.0793,
+      "step": 154100
+    },
+    {
+      "epoch": 57.67,
+      "learning_rate": 0.00012052631578947369,
+      "loss": 1.0743,
+      "step": 154200
+    },
+    {
+      "epoch": 57.7,
+      "learning_rate": 0.00012026315789473683,
+      "loss": 1.0768,
+      "step": 154300
+    },
+    {
+      "epoch": 57.74,
+      "learning_rate": 0.00012,
+      "loss": 1.0749,
+      "step": 154400
+    },
+    {
+      "epoch": 57.78,
+      "learning_rate": 0.00011973684210526316,
+      "loss": 1.076,
+      "step": 154500
+    },
+    {
+      "epoch": 57.82,
+      "learning_rate": 0.00011947368421052632,
+      "loss": 1.0751,
+      "step": 154600
+    },
+    {
+      "epoch": 57.85,
+      "learning_rate": 0.00011921052631578947,
+      "loss": 1.0739,
+      "step": 154700
+    },
+    {
+      "epoch": 57.89,
+      "learning_rate": 0.00011894736842105264,
+      "loss": 1.077,
+      "step": 154800
+    },
+    {
+      "epoch": 57.93,
+      "learning_rate": 0.0001186842105263158,
+      "loss": 1.0748,
+      "step": 154900
+    },
+    {
+      "epoch": 57.97,
+      "learning_rate": 0.00011842105263157894,
+      "loss": 1.0752,
+      "step": 155000
+    },
+    {
+      "epoch": 58.0,
+      "learning_rate": 0.0001181578947368421,
+      "loss": 1.082,
+      "step": 155100
+    },
+    {
+      "epoch": 58.04,
+      "learning_rate": 0.00011789473684210527,
+      "loss": 1.0708,
+      "step": 155200
+    },
+    {
+      "epoch": 58.08,
+      "learning_rate": 0.00011763157894736843,
+      "loss": 1.0671,
+      "step": 155300
+    },
+    {
+      "epoch": 58.12,
+      "learning_rate": 0.00011736842105263158,
+      "loss": 1.0712,
+      "step": 155400
+    },
+    {
+      "epoch": 58.15,
+      "learning_rate": 0.00011710526315789474,
+      "loss": 1.0732,
+      "step": 155500
+    },
+    {
+      "epoch": 58.19,
+      "learning_rate": 0.00011684210526315791,
+      "loss": 1.0737,
+      "step": 155600
+    },
+    {
+      "epoch": 58.23,
+      "learning_rate": 0.00011657894736842105,
+      "loss": 1.0727,
+      "step": 155700
+    },
+    {
+      "epoch": 58.26,
+      "learning_rate": 0.00011631578947368421,
+      "loss": 1.0748,
+      "step": 155800
+    },
+    {
+      "epoch": 58.3,
+      "learning_rate": 0.00011605263157894736,
+      "loss": 1.0692,
+      "step": 155900
+    },
+    {
+      "epoch": 58.34,
+      "learning_rate": 0.00011578947368421053,
+      "loss": 1.0705,
+      "step": 156000
+    },
+    {
+      "epoch": 58.38,
+      "learning_rate": 0.00011552631578947369,
+      "loss": 1.0709,
+      "step": 156100
+    },
+    {
+      "epoch": 58.41,
+      "learning_rate": 0.00011526315789473685,
+      "loss": 1.0718,
+      "step": 156200
+    },
+    {
+      "epoch": 58.45,
+      "learning_rate": 0.000115,
+      "loss": 1.0715,
+      "step": 156300
+    },
+    {
+      "epoch": 58.49,
+      "learning_rate": 0.00011473684210526316,
+      "loss": 1.0723,
+      "step": 156400
+    },
+    {
+      "epoch": 58.53,
+      "learning_rate": 0.00011447368421052632,
+      "loss": 1.0728,
+      "step": 156500
+    },
+    {
+      "epoch": 58.56,
+      "learning_rate": 0.00011421052631578947,
+      "loss": 1.0723,
+      "step": 156600
+    },
+    {
+      "epoch": 58.6,
+      "learning_rate": 0.00011394736842105263,
+      "loss": 1.0693,
+      "step": 156700
+    },
+    {
+      "epoch": 58.64,
+      "learning_rate": 0.0001136842105263158,
+      "loss": 1.073,
+      "step": 156800
+    },
+    {
+      "epoch": 58.68,
+      "learning_rate": 0.00011342105263157896,
+      "loss": 1.0743,
+      "step": 156900
+    },
+    {
+      "epoch": 58.71,
+      "learning_rate": 0.00011315789473684211,
+      "loss": 1.0715,
+      "step": 157000
+    },
+    {
+      "epoch": 58.75,
+      "learning_rate": 0.00011289473684210527,
+      "loss": 1.0734,
+      "step": 157100
+    },
+    {
+      "epoch": 58.79,
+      "learning_rate": 0.00011263157894736841,
+      "loss": 1.0731,
+      "step": 157200
+    },
+    {
+      "epoch": 58.83,
+      "learning_rate": 0.00011236842105263158,
+      "loss": 1.0711,
+      "step": 157300
+    },
+    {
+      "epoch": 58.86,
+      "learning_rate": 0.00011210526315789474,
+      "loss": 1.0694,
+      "step": 157400
+    },
+    {
+      "epoch": 58.9,
+      "learning_rate": 0.0001118421052631579,
+      "loss": 1.0706,
+      "step": 157500
+    },
+    {
+      "epoch": 58.94,
+      "learning_rate": 0.00011157894736842105,
+      "loss": 1.0729,
+      "step": 157600
+    },
+    {
+      "epoch": 58.98,
+      "learning_rate": 0.00011131578947368422,
+      "loss": 1.0704,
+      "step": 157700
+    },
+    {
+      "epoch": 59.01,
+      "learning_rate": 0.00011105263157894738,
+      "loss": 1.0741,
+      "step": 157800
+    },
+    {
+      "epoch": 59.05,
+      "learning_rate": 0.00011078947368421052,
+      "loss": 1.0697,
+      "step": 157900
+    },
+    {
+      "epoch": 59.09,
+      "learning_rate": 0.00011052631578947368,
+      "loss": 1.0676,
+      "step": 158000
+    },
+    {
+      "epoch": 59.12,
+      "learning_rate": 0.00011026315789473685,
+      "loss": 1.0689,
+      "step": 158100
+    },
+    {
+      "epoch": 59.16,
+      "learning_rate": 0.00011,
+      "loss": 1.0674,
+      "step": 158200
+    },
+    {
+      "epoch": 59.2,
+      "learning_rate": 0.00010973684210526316,
+      "loss": 1.0707,
+      "step": 158300
+    },
+    {
+      "epoch": 59.24,
+      "learning_rate": 0.00010947368421052632,
+      "loss": 1.0691,
+      "step": 158400
+    },
+    {
+      "epoch": 59.27,
+      "learning_rate": 0.00010921052631578949,
+      "loss": 1.0657,
+      "step": 158500
+    },
+    {
+      "epoch": 59.31,
+      "learning_rate": 0.00010894736842105263,
+      "loss": 1.0676,
+      "step": 158600
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 0.00010868421052631579,
+      "loss": 1.0678,
+      "step": 158700
+    },
+    {
+      "epoch": 59.39,
+      "learning_rate": 0.00010842105263157894,
+      "loss": 1.0683,
+      "step": 158800
+    },
+    {
+      "epoch": 59.42,
+      "learning_rate": 0.00010815789473684211,
+      "loss": 1.0638,
+      "step": 158900
+    },
+    {
+      "epoch": 59.46,
+      "learning_rate": 0.00010789473684210527,
+      "loss": 1.0722,
+      "step": 159000
+    },
+    {
+      "epoch": 59.5,
+      "learning_rate": 0.00010763157894736843,
+      "loss": 1.0673,
+      "step": 159100
+    },
+    {
+      "epoch": 59.54,
+      "learning_rate": 0.00010736842105263158,
+      "loss": 1.0659,
+      "step": 159200
+    },
+    {
+      "epoch": 59.57,
+      "learning_rate": 0.00010710526315789474,
+      "loss": 1.0665,
+      "step": 159300
+    },
+    {
+      "epoch": 59.61,
+      "learning_rate": 0.0001068421052631579,
+      "loss": 1.0662,
+      "step": 159400
+    },
+    {
+      "epoch": 59.65,
+      "learning_rate": 0.00010657894736842105,
+      "loss": 1.0676,
+      "step": 159500
+    },
+    {
+      "epoch": 59.69,
+      "learning_rate": 0.00010631578947368421,
+      "loss": 1.0677,
+      "step": 159600
+    },
+    {
+      "epoch": 59.72,
+      "learning_rate": 0.00010605263157894738,
+      "loss": 1.0668,
+      "step": 159700
+    },
+    {
+      "epoch": 59.76,
+      "learning_rate": 0.00010578947368421053,
+      "loss": 1.0661,
+      "step": 159800
+    },
+    {
+      "epoch": 59.8,
+      "learning_rate": 0.00010552631578947369,
+      "loss": 1.0705,
+      "step": 159900
+    },
+    {
+      "epoch": 59.84,
+      "learning_rate": 0.00010526315789473683,
+      "loss": 1.0675,
+      "step": 160000
+    },
+    {
+      "epoch": 59.87,
+      "learning_rate": 0.000105,
+      "loss": 1.0695,
+      "step": 160100
+    },
+    {
+      "epoch": 59.91,
+      "learning_rate": 0.00010473684210526316,
+      "loss": 1.0655,
+      "step": 160200
+    },
+    {
+      "epoch": 59.95,
+      "learning_rate": 0.00010447368421052632,
+      "loss": 1.0679,
+      "step": 160300
+    },
+    {
+      "epoch": 59.98,
+      "learning_rate": 0.00010421052631578947,
+      "loss": 1.0667,
+      "step": 160400
+    },
+    {
+      "epoch": 60.02,
+      "learning_rate": 0.00010394736842105264,
+      "loss": 1.0693,
+      "step": 160500
+    },
+    {
+      "epoch": 60.06,
+      "learning_rate": 0.0001036842105263158,
+      "loss": 1.0631,
+      "step": 160600
+    },
+    {
+      "epoch": 60.1,
+      "learning_rate": 0.00010342105263157894,
+      "loss": 1.0619,
+      "step": 160700
+    },
+    {
+      "epoch": 60.13,
+      "learning_rate": 0.0001031578947368421,
+      "loss": 1.0652,
+      "step": 160800
+    },
+    {
+      "epoch": 60.17,
+      "learning_rate": 0.00010289473684210527,
+      "loss": 1.0627,
+      "step": 160900
+    },
+    {
+      "epoch": 60.21,
+      "learning_rate": 0.00010263157894736843,
+      "loss": 1.0658,
+      "step": 161000
+    },
+    {
+      "epoch": 60.25,
+      "learning_rate": 0.00010236842105263158,
+      "loss": 1.0642,
+      "step": 161100
+    },
+    {
+      "epoch": 60.28,
+      "learning_rate": 0.00010210526315789474,
+      "loss": 1.0665,
+      "step": 161200
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 0.00010184210526315791,
+      "loss": 1.0673,
+      "step": 161300
+    },
+    {
+      "epoch": 60.36,
+      "learning_rate": 0.00010157894736842105,
+      "loss": 1.0696,
+      "step": 161400
+    },
+    {
+      "epoch": 60.4,
+      "learning_rate": 0.00010131578947368421,
+      "loss": 1.065,
+      "step": 161500
+    },
+    {
+      "epoch": 60.43,
+      "learning_rate": 0.00010105263157894737,
+      "loss": 1.0633,
+      "step": 161600
+    },
+    {
+      "epoch": 60.47,
+      "learning_rate": 0.00010078947368421052,
+      "loss": 1.0618,
+      "step": 161700
+    },
+    {
+      "epoch": 60.51,
+      "learning_rate": 0.00010052631578947369,
+      "loss": 1.0662,
+      "step": 161800
+    },
+    {
+      "epoch": 60.55,
+      "learning_rate": 0.00010026315789473685,
+      "loss": 1.0637,
+      "step": 161900
+    },
+    {
+      "epoch": 60.58,
+      "learning_rate": 0.0001,
+      "loss": 1.0604,
+      "step": 162000
+    },
+    {
+      "epoch": 60.62,
+      "learning_rate": 9.973684210526315e-05,
+      "loss": 1.0604,
+      "step": 162100
+    },
+    {
+      "epoch": 60.66,
+      "learning_rate": 9.947368421052632e-05,
+      "loss": 1.0634,
+      "step": 162200
+    },
+    {
+      "epoch": 60.7,
+      "learning_rate": 9.921052631578947e-05,
+      "loss": 1.0608,
+      "step": 162300
+    },
+    {
+      "epoch": 60.73,
+      "learning_rate": 9.894736842105263e-05,
+      "loss": 1.0651,
+      "step": 162400
+    },
+    {
+      "epoch": 60.77,
+      "learning_rate": 9.868421052631579e-05,
+      "loss": 1.0639,
+      "step": 162500
+    },
+    {
+      "epoch": 60.81,
+      "learning_rate": 9.842105263157896e-05,
+      "loss": 1.0665,
+      "step": 162600
+    },
+    {
+      "epoch": 60.85,
+      "learning_rate": 9.815789473684211e-05,
+      "loss": 1.0665,
+      "step": 162700
+    },
+    {
+      "epoch": 60.88,
+      "learning_rate": 9.789473684210526e-05,
+      "loss": 1.0644,
+      "step": 162800
+    },
+    {
+      "epoch": 60.92,
+      "learning_rate": 9.763157894736841e-05,
+      "loss": 1.063,
+      "step": 162900
+    },
+    {
+      "epoch": 60.96,
+      "learning_rate": 9.736842105263158e-05,
+      "loss": 1.0627,
+      "step": 163000
+    },
+    {
+      "epoch": 60.99,
+      "learning_rate": 9.710526315789474e-05,
+      "loss": 1.0623,
+      "step": 163100
+    },
+    {
+      "epoch": 61.03,
+      "learning_rate": 9.68421052631579e-05,
+      "loss": 1.0722,
+      "step": 163200
+    },
+    {
+      "epoch": 61.07,
+      "learning_rate": 9.657894736842105e-05,
+      "loss": 1.0582,
+      "step": 163300
+    },
+    {
+      "epoch": 61.11,
+      "learning_rate": 9.631578947368422e-05,
+      "loss": 1.0605,
+      "step": 163400
+    },
+    {
+      "epoch": 61.14,
+      "learning_rate": 9.605263157894737e-05,
+      "loss": 1.0618,
+      "step": 163500
+    },
+    {
+      "epoch": 61.18,
+      "learning_rate": 9.578947368421052e-05,
+      "loss": 1.0613,
+      "step": 163600
+    },
+    {
+      "epoch": 61.22,
+      "learning_rate": 9.552631578947368e-05,
+      "loss": 1.0601,
+      "step": 163700
+    },
+    {
+      "epoch": 61.26,
+      "learning_rate": 9.526315789473685e-05,
+      "loss": 1.0635,
+      "step": 163800
+    },
+    {
+      "epoch": 61.29,
+      "learning_rate": 9.5e-05,
+      "loss": 1.0583,
+      "step": 163900
+    },
+    {
+      "epoch": 61.33,
+      "learning_rate": 9.473684210526316e-05,
+      "loss": 1.0622,
+      "step": 164000
+    },
+    {
+      "epoch": 61.37,
+      "learning_rate": 9.447368421052632e-05,
+      "loss": 1.0647,
+      "step": 164100
+    },
+    {
+      "epoch": 61.41,
+      "learning_rate": 9.421052631578947e-05,
+      "loss": 1.0605,
+      "step": 164200
+    },
+    {
+      "epoch": 61.44,
+      "learning_rate": 9.394736842105263e-05,
+      "loss": 1.0636,
+      "step": 164300
+    },
+    {
+      "epoch": 61.48,
+      "learning_rate": 9.368421052631579e-05,
+      "loss": 1.0578,
+      "step": 164400
+    },
+    {
+      "epoch": 61.52,
+      "learning_rate": 9.342105263157894e-05,
+      "loss": 1.0608,
+      "step": 164500
+    },
+    {
+      "epoch": 61.56,
+      "learning_rate": 9.315789473684211e-05,
+      "loss": 1.0621,
+      "step": 164600
+    },
+    {
+      "epoch": 61.59,
+      "learning_rate": 9.289473684210527e-05,
+      "loss": 1.0562,
+      "step": 164700
+    },
+    {
+      "epoch": 61.63,
+      "learning_rate": 9.263157894736843e-05,
+      "loss": 1.06,
+      "step": 164800
+    },
+    {
+      "epoch": 61.67,
+      "learning_rate": 9.236842105263158e-05,
+      "loss": 1.0589,
+      "step": 164900
+    },
+    {
+      "epoch": 61.71,
+      "learning_rate": 9.210526315789474e-05,
+      "loss": 1.0623,
+      "step": 165000
+    },
+    {
+      "epoch": 61.74,
+      "learning_rate": 9.18421052631579e-05,
+      "loss": 1.0593,
+      "step": 165100
+    },
+    {
+      "epoch": 61.78,
+      "learning_rate": 9.157894736842105e-05,
+      "loss": 1.0608,
+      "step": 165200
+    },
+    {
+      "epoch": 61.82,
+      "learning_rate": 9.131578947368421e-05,
+      "loss": 1.0565,
+      "step": 165300
+    },
+    {
+      "epoch": 61.85,
+      "learning_rate": 9.105263157894738e-05,
+      "loss": 1.0597,
+      "step": 165400
+    },
+    {
+      "epoch": 61.89,
+      "learning_rate": 9.078947368421054e-05,
+      "loss": 1.0618,
+      "step": 165500
+    },
+    {
+      "epoch": 61.93,
+      "learning_rate": 9.052631578947369e-05,
+      "loss": 1.0587,
+      "step": 165600
+    },
+    {
+      "epoch": 61.97,
+      "learning_rate": 9.026315789473684e-05,
+      "loss": 1.0569,
+      "step": 165700
+    },
+    {
+      "epoch": 62.0,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 1.0663,
+      "step": 165800
+    },
+    {
+      "epoch": 62.04,
+      "learning_rate": 8.973684210526316e-05,
+      "loss": 1.0524,
+      "step": 165900
+    },
+    {
+      "epoch": 62.08,
+      "learning_rate": 8.947368421052632e-05,
+      "loss": 1.0573,
+      "step": 166000
+    },
+    {
+      "epoch": 62.12,
+      "learning_rate": 8.921052631578948e-05,
+      "loss": 1.0554,
+      "step": 166100
+    },
+    {
+      "epoch": 62.15,
+      "learning_rate": 8.894736842105263e-05,
+      "loss": 1.059,
+      "step": 166200
+    },
+    {
+      "epoch": 62.19,
+      "learning_rate": 8.86842105263158e-05,
+      "loss": 1.0551,
+      "step": 166300
+    },
+    {
+      "epoch": 62.23,
+      "learning_rate": 8.842105263157894e-05,
+      "loss": 1.0565,
+      "step": 166400
+    },
+    {
+      "epoch": 62.27,
+      "learning_rate": 8.81578947368421e-05,
+      "loss": 1.0585,
+      "step": 166500
+    },
+    {
+      "epoch": 62.3,
+      "learning_rate": 8.789473684210526e-05,
+      "loss": 1.0556,
+      "step": 166600
+    },
+    {
+      "epoch": 62.34,
+      "learning_rate": 8.763157894736843e-05,
+      "loss": 1.0543,
+      "step": 166700
+    },
+    {
+      "epoch": 62.38,
+      "learning_rate": 8.736842105263158e-05,
+      "loss": 1.0566,
+      "step": 166800
+    },
+    {
+      "epoch": 62.42,
+      "learning_rate": 8.710526315789474e-05,
+      "loss": 1.058,
+      "step": 166900
+    },
+    {
+      "epoch": 62.45,
+      "learning_rate": 8.68421052631579e-05,
+      "loss": 1.0562,
+      "step": 167000
+    },
+    {
+      "epoch": 62.49,
+      "learning_rate": 8.657894736842105e-05,
+      "loss": 1.0589,
+      "step": 167100
+    },
+    {
+      "epoch": 62.53,
+      "learning_rate": 8.631578947368421e-05,
+      "loss": 1.0546,
+      "step": 167200
+    },
+    {
+      "epoch": 62.57,
+      "learning_rate": 8.605263157894737e-05,
+      "loss": 1.055,
+      "step": 167300
+    },
+    {
+      "epoch": 62.6,
+      "learning_rate": 8.578947368421052e-05,
+      "loss": 1.0566,
+      "step": 167400
+    },
+    {
+      "epoch": 62.64,
+      "learning_rate": 8.552631578947369e-05,
+      "loss": 1.0563,
+      "step": 167500
+    },
+    {
+      "epoch": 62.68,
+      "learning_rate": 8.526315789473685e-05,
+      "loss": 1.0587,
+      "step": 167600
+    },
+    {
+      "epoch": 62.71,
+      "learning_rate": 8.5e-05,
+      "loss": 1.0578,
+      "step": 167700
+    },
+    {
+      "epoch": 62.75,
+      "learning_rate": 8.473684210526315e-05,
+      "loss": 1.0535,
+      "step": 167800
+    },
+    {
+      "epoch": 62.79,
+      "learning_rate": 8.447368421052632e-05,
+      "loss": 1.0527,
+      "step": 167900
+    },
+    {
+      "epoch": 62.83,
+      "learning_rate": 8.421052631578948e-05,
+      "loss": 1.0546,
+      "step": 168000
+    },
+    {
+      "epoch": 62.86,
+      "learning_rate": 8.394736842105263e-05,
+      "loss": 1.0543,
+      "step": 168100
+    },
+    {
+      "epoch": 62.9,
+      "learning_rate": 8.368421052631579e-05,
+      "loss": 1.0562,
+      "step": 168200
+    },
+    {
+      "epoch": 62.94,
+      "learning_rate": 8.342105263157896e-05,
+      "loss": 1.0578,
+      "step": 168300
+    },
+    {
+      "epoch": 62.98,
+      "learning_rate": 8.315789473684212e-05,
+      "loss": 1.0579,
+      "step": 168400
+    },
+    {
+      "epoch": 63.01,
+      "learning_rate": 8.289473684210526e-05,
+      "loss": 1.0604,
+      "step": 168500
+    },
+    {
+      "epoch": 63.05,
+      "learning_rate": 8.263157894736841e-05,
+      "loss": 1.0547,
+      "step": 168600
+    },
+    {
+      "epoch": 63.09,
+      "learning_rate": 8.236842105263158e-05,
+      "loss": 1.0531,
+      "step": 168700
+    },
+    {
+      "epoch": 63.13,
+      "learning_rate": 8.210526315789474e-05,
+      "loss": 1.0547,
+      "step": 168800
+    },
+    {
+      "epoch": 63.16,
+      "learning_rate": 8.18421052631579e-05,
+      "loss": 1.0509,
+      "step": 168900
+    },
+    {
+      "epoch": 63.2,
+      "learning_rate": 8.157894736842105e-05,
+      "loss": 1.0528,
+      "step": 169000
+    },
+    {
+      "epoch": 63.24,
+      "learning_rate": 8.131578947368422e-05,
+      "loss": 1.0534,
+      "step": 169100
+    },
+    {
+      "epoch": 63.28,
+      "learning_rate": 8.105263157894737e-05,
+      "loss": 1.049,
+      "step": 169200
+    },
+    {
+      "epoch": 63.31,
+      "learning_rate": 8.078947368421052e-05,
+      "loss": 1.0526,
+      "step": 169300
+    },
+    {
+      "epoch": 63.35,
+      "learning_rate": 8.052631578947368e-05,
+      "loss": 1.052,
+      "step": 169400
+    },
+    {
+      "epoch": 63.39,
+      "learning_rate": 8.026315789473685e-05,
+      "loss": 1.0539,
+      "step": 169500
+    },
+    {
+      "epoch": 63.43,
+      "learning_rate": 8e-05,
+      "loss": 1.0525,
+      "step": 169600
+    },
+    {
+      "epoch": 63.46,
+      "learning_rate": 7.973684210526316e-05,
+      "loss": 1.0532,
+      "step": 169700
+    },
+    {
+      "epoch": 63.5,
+      "learning_rate": 7.947368421052632e-05,
+      "loss": 1.0504,
+      "step": 169800
+    },
+    {
+      "epoch": 63.54,
+      "learning_rate": 7.921052631578946e-05,
+      "loss": 1.0497,
+      "step": 169900
+    },
+    {
+      "epoch": 63.58,
+      "learning_rate": 7.894736842105263e-05,
+      "loss": 1.0514,
+      "step": 170000
+    },
+    {
+      "epoch": 63.61,
+      "learning_rate": 7.868421052631579e-05,
+      "loss": 1.0527,
+      "step": 170100
+    },
+    {
+      "epoch": 63.65,
+      "learning_rate": 7.842105263157895e-05,
+      "loss": 1.0498,
+      "step": 170200
+    },
+    {
+      "epoch": 63.69,
+      "learning_rate": 7.81578947368421e-05,
+      "loss": 1.0498,
+      "step": 170300
+    },
+    {
+      "epoch": 63.72,
+      "learning_rate": 7.789473684210527e-05,
+      "loss": 1.0524,
+      "step": 170400
+    },
+    {
+      "epoch": 63.76,
+      "learning_rate": 7.763157894736843e-05,
+      "loss": 1.0573,
+      "step": 170500
+    },
+    {
+      "epoch": 63.8,
+      "learning_rate": 7.736842105263157e-05,
+      "loss": 1.0533,
+      "step": 170600
+    },
+    {
+      "epoch": 63.84,
+      "learning_rate": 7.710526315789473e-05,
+      "loss": 1.0513,
+      "step": 170700
+    },
+    {
+      "epoch": 63.87,
+      "learning_rate": 7.68421052631579e-05,
+      "loss": 1.0501,
+      "step": 170800
+    },
+    {
+      "epoch": 63.91,
+      "learning_rate": 7.657894736842105e-05,
+      "loss": 1.052,
+      "step": 170900
+    },
+    {
+      "epoch": 63.95,
+      "learning_rate": 7.631578947368421e-05,
+      "loss": 1.0528,
+      "step": 171000
+    },
+    {
+      "epoch": 63.99,
+      "learning_rate": 7.605263157894737e-05,
+      "loss": 1.0516,
+      "step": 171100
+    },
+    {
+      "epoch": 64.02,
+      "learning_rate": 7.578947368421054e-05,
+      "loss": 1.0559,
+      "step": 171200
+    },
+    {
+      "epoch": 64.06,
+      "learning_rate": 7.552631578947368e-05,
+      "loss": 1.0487,
+      "step": 171300
+    },
+    {
+      "epoch": 64.1,
+      "learning_rate": 7.526315789473684e-05,
+      "loss": 1.0471,
+      "step": 171400
+    },
+    {
+      "epoch": 64.14,
+      "learning_rate": 7.5e-05,
+      "loss": 1.0488,
+      "step": 171500
+    },
+    {
+      "epoch": 64.17,
+      "learning_rate": 7.473684210526316e-05,
+      "loss": 1.0499,
+      "step": 171600
+    },
+    {
+      "epoch": 64.21,
+      "learning_rate": 7.447368421052632e-05,
+      "loss": 1.0487,
+      "step": 171700
+    },
+    {
+      "epoch": 64.25,
+      "learning_rate": 7.421052631578948e-05,
+      "loss": 1.0494,
+      "step": 171800
+    },
+    {
+      "epoch": 64.29,
+      "learning_rate": 7.394736842105263e-05,
+      "loss": 1.0472,
+      "step": 171900
+    },
+    {
+      "epoch": 64.32,
+      "learning_rate": 7.368421052631579e-05,
+      "loss": 1.05,
+      "step": 172000
+    },
+    {
+      "epoch": 64.36,
+      "learning_rate": 7.342105263157895e-05,
+      "loss": 1.0473,
+      "step": 172100
+    },
+    {
+      "epoch": 64.4,
+      "learning_rate": 7.31578947368421e-05,
+      "loss": 1.0478,
+      "step": 172200
+    },
+    {
+      "epoch": 64.44,
+      "learning_rate": 7.289473684210526e-05,
+      "loss": 1.05,
+      "step": 172300
+    },
+    {
+      "epoch": 64.47,
+      "learning_rate": 7.263157894736843e-05,
+      "loss": 1.0475,
+      "step": 172400
+    },
+    {
+      "epoch": 64.51,
+      "learning_rate": 7.236842105263159e-05,
+      "loss": 1.0461,
+      "step": 172500
+    },
+    {
+      "epoch": 64.55,
+      "learning_rate": 7.210526315789474e-05,
+      "loss": 1.0467,
+      "step": 172600
+    },
+    {
+      "epoch": 64.58,
+      "learning_rate": 7.18421052631579e-05,
+      "loss": 1.0495,
+      "step": 172700
+    },
+    {
+      "epoch": 64.62,
+      "learning_rate": 7.157894736842105e-05,
+      "loss": 1.0519,
+      "step": 172800
+    },
+    {
+      "epoch": 64.66,
+      "learning_rate": 7.131578947368421e-05,
+      "loss": 1.053,
+      "step": 172900
+    },
+    {
+      "epoch": 64.7,
+      "learning_rate": 7.105263157894737e-05,
+      "loss": 1.0444,
+      "step": 173000
+    },
+    {
+      "epoch": 64.73,
+      "learning_rate": 7.078947368421052e-05,
+      "loss": 1.0478,
+      "step": 173100
+    },
+    {
+      "epoch": 64.77,
+      "learning_rate": 7.05263157894737e-05,
+      "loss": 1.0477,
+      "step": 173200
+    },
+    {
+      "epoch": 64.81,
+      "learning_rate": 7.026315789473685e-05,
+      "loss": 1.0462,
+      "step": 173300
+    },
+    {
+      "epoch": 64.85,
+      "learning_rate": 7.000000000000001e-05,
+      "loss": 1.0529,
+      "step": 173400
+    },
+    {
+      "epoch": 64.88,
+      "learning_rate": 6.973684210526315e-05,
+      "loss": 1.0502,
+      "step": 173500
+    },
+    {
+      "epoch": 64.92,
+      "learning_rate": 6.947368421052632e-05,
+      "loss": 1.0494,
+      "step": 173600
+    },
+    {
+      "epoch": 64.96,
+      "learning_rate": 6.921052631578948e-05,
+      "loss": 1.0464,
+      "step": 173700
+    },
+    {
+      "epoch": 65.0,
+      "learning_rate": 6.894736842105263e-05,
+      "loss": 1.0447,
+      "step": 173800
+    },
+    {
+      "epoch": 65.03,
+      "learning_rate": 6.868421052631579e-05,
+      "loss": 1.0476,
+      "step": 173900
+    },
+    {
+      "epoch": 65.07,
+      "learning_rate": 6.842105263157896e-05,
+      "loss": 1.0486,
+      "step": 174000
+    },
+    {
+      "epoch": 65.11,
+      "learning_rate": 6.815789473684212e-05,
+      "loss": 1.044,
+      "step": 174100
+    },
+    {
+      "epoch": 65.15,
+      "learning_rate": 6.789473684210526e-05,
+      "loss": 1.0466,
+      "step": 174200
+    },
+    {
+      "epoch": 65.18,
+      "learning_rate": 6.763157894736842e-05,
+      "loss": 1.0408,
+      "step": 174300
+    },
+    {
+      "epoch": 65.22,
+      "learning_rate": 6.736842105263157e-05,
+      "loss": 1.0478,
+      "step": 174400
+    },
+    {
+      "epoch": 65.26,
+      "learning_rate": 6.710526315789474e-05,
+      "loss": 1.0446,
+      "step": 174500
+    },
+    {
+      "epoch": 65.3,
+      "learning_rate": 6.68421052631579e-05,
+      "loss": 1.0471,
+      "step": 174600
+    },
+    {
+      "epoch": 65.33,
+      "learning_rate": 6.657894736842106e-05,
+      "loss": 1.047,
+      "step": 174700
+    },
+    {
+      "epoch": 65.37,
+      "learning_rate": 6.631578947368421e-05,
+      "loss": 1.0452,
+      "step": 174800
+    },
+    {
+      "epoch": 65.41,
+      "learning_rate": 6.605263157894737e-05,
+      "loss": 1.045,
+      "step": 174900
+    },
+    {
+      "epoch": 65.44,
+      "learning_rate": 6.578947368421052e-05,
+      "loss": 1.0466,
+      "step": 175000
+    },
+    {
+      "epoch": 65.48,
+      "learning_rate": 6.552631578947368e-05,
+      "loss": 1.0462,
+      "step": 175100
+    },
+    {
+      "epoch": 65.52,
+      "learning_rate": 6.526315789473684e-05,
+      "loss": 1.0427,
+      "step": 175200
+    },
+    {
+      "epoch": 65.56,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 1.0449,
+      "step": 175300
+    },
+    {
+      "epoch": 65.59,
+      "learning_rate": 6.473684210526316e-05,
+      "loss": 1.0468,
+      "step": 175400
+    },
+    {
+      "epoch": 65.63,
+      "learning_rate": 6.447368421052632e-05,
+      "loss": 1.0466,
+      "step": 175500
+    },
+    {
+      "epoch": 65.67,
+      "learning_rate": 6.421052631578946e-05,
+      "loss": 1.0443,
+      "step": 175600
+    },
+    {
+      "epoch": 65.71,
+      "learning_rate": 6.394736842105263e-05,
+      "loss": 1.0463,
+      "step": 175700
+    },
+    {
+      "epoch": 65.74,
+      "learning_rate": 6.368421052631579e-05,
+      "loss": 1.0433,
+      "step": 175800
+    },
+    {
+      "epoch": 65.78,
+      "learning_rate": 6.342105263157895e-05,
+      "loss": 1.0441,
+      "step": 175900
+    },
+    {
+      "epoch": 65.82,
+      "learning_rate": 6.31578947368421e-05,
+      "loss": 1.0453,
+      "step": 176000
+    },
+    {
+      "epoch": 65.86,
+      "learning_rate": 6.289473684210527e-05,
+      "loss": 1.0459,
+      "step": 176100
+    },
+    {
+      "epoch": 65.89,
+      "learning_rate": 6.263157894736843e-05,
+      "loss": 1.0447,
+      "step": 176200
+    },
+    {
+      "epoch": 65.93,
+      "learning_rate": 6.236842105263159e-05,
+      "loss": 1.0451,
+      "step": 176300
+    },
+    {
+      "epoch": 65.97,
+      "learning_rate": 6.210526315789474e-05,
+      "loss": 1.0418,
+      "step": 176400
+    },
+    {
+      "epoch": 66.01,
+      "learning_rate": 6.18421052631579e-05,
+      "loss": 1.0505,
+      "step": 176500
+    },
+    {
+      "epoch": 66.04,
+      "learning_rate": 6.157894736842106e-05,
+      "loss": 1.0445,
+      "step": 176600
+    },
+    {
+      "epoch": 66.08,
+      "learning_rate": 6.131578947368421e-05,
+      "loss": 1.0397,
+      "step": 176700
+    },
+    {
+      "epoch": 66.12,
+      "learning_rate": 6.105263157894737e-05,
+      "loss": 1.0408,
+      "step": 176800
+    },
+    {
+      "epoch": 66.16,
+      "learning_rate": 6.0789473684210525e-05,
+      "loss": 1.0449,
+      "step": 176900
+    },
+    {
+      "epoch": 66.19,
+      "learning_rate": 6.052631578947369e-05,
+      "loss": 1.0453,
+      "step": 177000
+    },
+    {
+      "epoch": 66.23,
+      "learning_rate": 6.0263157894736845e-05,
+      "loss": 1.0419,
+      "step": 177100
+    },
+    {
+      "epoch": 66.27,
+      "learning_rate": 6e-05,
+      "loss": 1.0371,
+      "step": 177200
+    },
+    {
+      "epoch": 66.31,
+      "learning_rate": 5.973684210526316e-05,
+      "loss": 1.041,
+      "step": 177300
+    },
+    {
+      "epoch": 66.34,
+      "learning_rate": 5.947368421052632e-05,
+      "loss": 1.0418,
+      "step": 177400
+    },
+    {
+      "epoch": 66.38,
+      "learning_rate": 5.921052631578947e-05,
+      "loss": 1.0413,
+      "step": 177500
+    },
+    {
+      "epoch": 66.42,
+      "learning_rate": 5.8947368421052634e-05,
+      "loss": 1.0426,
+      "step": 177600
+    },
+    {
+      "epoch": 66.45,
+      "learning_rate": 5.868421052631579e-05,
+      "loss": 1.0427,
+      "step": 177700
+    },
+    {
+      "epoch": 66.49,
+      "learning_rate": 5.8421052631578954e-05,
+      "loss": 1.042,
+      "step": 177800
+    },
+    {
+      "epoch": 66.53,
+      "learning_rate": 5.8157894736842104e-05,
+      "loss": 1.0425,
+      "step": 177900
+    },
+    {
+      "epoch": 66.57,
+      "learning_rate": 5.789473684210527e-05,
+      "loss": 1.041,
+      "step": 178000
+    },
+    {
+      "epoch": 66.6,
+      "learning_rate": 5.7631578947368423e-05,
+      "loss": 1.042,
+      "step": 178100
+    },
+    {
+      "epoch": 66.64,
+      "learning_rate": 5.736842105263158e-05,
+      "loss": 1.0379,
+      "step": 178200
+    },
+    {
+      "epoch": 66.68,
+      "learning_rate": 5.7105263157894736e-05,
+      "loss": 1.0401,
+      "step": 178300
+    },
+    {
+      "epoch": 66.72,
+      "learning_rate": 5.68421052631579e-05,
+      "loss": 1.0431,
+      "step": 178400
+    },
+    {
+      "epoch": 66.75,
+      "learning_rate": 5.6578947368421056e-05,
+      "loss": 1.0394,
+      "step": 178500
+    },
+    {
+      "epoch": 66.79,
+      "learning_rate": 5.6315789473684206e-05,
+      "loss": 1.0426,
+      "step": 178600
+    },
+    {
+      "epoch": 66.83,
+      "learning_rate": 5.605263157894737e-05,
+      "loss": 1.0433,
+      "step": 178700
+    },
+    {
+      "epoch": 66.87,
+      "learning_rate": 5.5789473684210526e-05,
+      "loss": 1.0434,
+      "step": 178800
+    },
+    {
+      "epoch": 66.9,
+      "learning_rate": 5.552631578947369e-05,
+      "loss": 1.0388,
+      "step": 178900
+    },
+    {
+      "epoch": 66.94,
+      "learning_rate": 5.526315789473684e-05,
+      "loss": 1.041,
+      "step": 179000
+    },
+    {
+      "epoch": 66.98,
+      "learning_rate": 5.5e-05,
+      "loss": 1.0414,
+      "step": 179100
+    },
+    {
+      "epoch": 67.02,
+      "learning_rate": 5.473684210526316e-05,
+      "loss": 1.0458,
+      "step": 179200
+    },
+    {
+      "epoch": 67.05,
+      "learning_rate": 5.4473684210526315e-05,
+      "loss": 1.0394,
+      "step": 179300
+    },
+    {
+      "epoch": 67.09,
+      "learning_rate": 5.421052631578947e-05,
+      "loss": 1.0395,
+      "step": 179400
+    },
+    {
+      "epoch": 67.13,
+      "learning_rate": 5.3947368421052635e-05,
+      "loss": 1.0368,
+      "step": 179500
+    },
+    {
+      "epoch": 67.17,
+      "learning_rate": 5.368421052631579e-05,
+      "loss": 1.0392,
+      "step": 179600
+    },
+    {
+      "epoch": 67.2,
+      "learning_rate": 5.342105263157895e-05,
+      "loss": 1.0382,
+      "step": 179700
+    },
+    {
+      "epoch": 67.24,
+      "learning_rate": 5.3157894736842104e-05,
+      "loss": 1.0377,
+      "step": 179800
+    },
+    {
+      "epoch": 67.28,
+      "learning_rate": 5.289473684210527e-05,
+      "loss": 1.0397,
+      "step": 179900
+    },
+    {
+      "epoch": 67.31,
+      "learning_rate": 5.263157894736842e-05,
+      "loss": 1.0356,
+      "step": 180000
+    },
+    {
+      "epoch": 67.35,
+      "learning_rate": 5.236842105263158e-05,
+      "loss": 1.0383,
+      "step": 180100
+    },
+    {
+      "epoch": 67.39,
+      "learning_rate": 5.210526315789474e-05,
+      "loss": 1.0362,
+      "step": 180200
+    },
+    {
+      "epoch": 67.43,
+      "learning_rate": 5.18421052631579e-05,
+      "loss": 1.0388,
+      "step": 180300
+    },
+    {
+      "epoch": 67.46,
+      "learning_rate": 5.157894736842105e-05,
+      "loss": 1.039,
+      "step": 180400
+    },
+    {
+      "epoch": 67.5,
+      "learning_rate": 5.131578947368421e-05,
+      "loss": 1.0396,
+      "step": 180500
+    },
+    {
+      "epoch": 67.54,
+      "learning_rate": 5.105263157894737e-05,
+      "loss": 1.0379,
+      "step": 180600
+    },
+    {
+      "epoch": 67.58,
+      "learning_rate": 5.0789473684210526e-05,
+      "loss": 1.0353,
+      "step": 180700
+    },
+    {
+      "epoch": 67.61,
+      "learning_rate": 5.052631578947368e-05,
+      "loss": 1.039,
+      "step": 180800
+    },
+    {
+      "epoch": 67.65,
+      "learning_rate": 5.0263157894736846e-05,
+      "loss": 1.0369,
+      "step": 180900
+    },
+    {
+      "epoch": 67.69,
+      "learning_rate": 5e-05,
+      "loss": 1.0371,
+      "step": 181000
+    },
+    {
+      "epoch": 67.73,
+      "learning_rate": 4.973684210526316e-05,
+      "loss": 1.0387,
+      "step": 181100
+    },
+    {
+      "epoch": 67.76,
+      "learning_rate": 4.9473684210526315e-05,
+      "loss": 1.0402,
+      "step": 181200
+    },
+    {
+      "epoch": 67.8,
+      "learning_rate": 4.921052631578948e-05,
+      "loss": 1.0385,
+      "step": 181300
+    },
+    {
+      "epoch": 67.84,
+      "learning_rate": 4.894736842105263e-05,
+      "loss": 1.0349,
+      "step": 181400
+    },
+    {
+      "epoch": 67.88,
+      "learning_rate": 4.868421052631579e-05,
+      "loss": 1.0362,
+      "step": 181500
+    },
+    {
+      "epoch": 67.91,
+      "learning_rate": 4.842105263157895e-05,
+      "loss": 1.0398,
+      "step": 181600
+    },
+    {
+      "epoch": 67.95,
+      "learning_rate": 4.815789473684211e-05,
+      "loss": 1.0386,
+      "step": 181700
+    },
+    {
+      "epoch": 67.99,
+      "learning_rate": 4.789473684210526e-05,
+      "loss": 1.0365,
+      "step": 181800
+    },
+    {
+      "epoch": 68.03,
+      "learning_rate": 4.7631578947368424e-05,
+      "loss": 1.037,
+      "step": 181900
+    },
+    {
+      "epoch": 68.06,
+      "learning_rate": 4.736842105263158e-05,
+      "loss": 1.0344,
+      "step": 182000
+    },
+    {
+      "epoch": 68.1,
+      "learning_rate": 4.710526315789474e-05,
+      "loss": 1.0362,
+      "step": 182100
+    },
+    {
+      "epoch": 68.14,
+      "learning_rate": 4.6842105263157894e-05,
+      "loss": 1.0321,
+      "step": 182200
+    },
+    {
+      "epoch": 68.17,
+      "learning_rate": 4.657894736842106e-05,
+      "loss": 1.0373,
+      "step": 182300
+    },
+    {
+      "epoch": 68.21,
+      "learning_rate": 4.6315789473684214e-05,
+      "loss": 1.0349,
+      "step": 182400
+    },
+    {
+      "epoch": 68.25,
+      "learning_rate": 4.605263157894737e-05,
+      "loss": 1.0335,
+      "step": 182500
+    },
+    {
+      "epoch": 68.29,
+      "learning_rate": 4.5789473684210527e-05,
+      "loss": 1.0359,
+      "step": 182600
+    },
+    {
+      "epoch": 68.32,
+      "learning_rate": 4.552631578947369e-05,
+      "loss": 1.0341,
+      "step": 182700
+    },
+    {
+      "epoch": 68.36,
+      "learning_rate": 4.5263157894736846e-05,
+      "loss": 1.0377,
+      "step": 182800
+    },
+    {
+      "epoch": 68.4,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 1.038,
+      "step": 182900
+    },
+    {
+      "epoch": 68.44,
+      "learning_rate": 4.473684210526316e-05,
+      "loss": 1.0351,
+      "step": 183000
+    },
+    {
+      "epoch": 68.47,
+      "learning_rate": 4.4473684210526316e-05,
+      "loss": 1.0358,
+      "step": 183100
+    },
+    {
+      "epoch": 68.51,
+      "learning_rate": 4.421052631578947e-05,
+      "loss": 1.0336,
+      "step": 183200
+    },
+    {
+      "epoch": 68.55,
+      "learning_rate": 4.394736842105263e-05,
+      "loss": 1.0358,
+      "step": 183300
+    },
+    {
+      "epoch": 68.59,
+      "learning_rate": 4.368421052631579e-05,
+      "loss": 1.0336,
+      "step": 183400
+    },
+    {
+      "epoch": 68.62,
+      "learning_rate": 4.342105263157895e-05,
+      "loss": 1.0368,
+      "step": 183500
+    },
+    {
+      "epoch": 68.66,
+      "learning_rate": 4.3157894736842105e-05,
+      "loss": 1.0374,
+      "step": 183600
+    },
+    {
+      "epoch": 68.7,
+      "learning_rate": 4.289473684210526e-05,
+      "loss": 1.0363,
+      "step": 183700
+    },
+    {
+      "epoch": 68.74,
+      "learning_rate": 4.2631578947368425e-05,
+      "loss": 1.0351,
+      "step": 183800
+    },
+    {
+      "epoch": 68.77,
+      "learning_rate": 4.2368421052631575e-05,
+      "loss": 1.0325,
+      "step": 183900
+    },
+    {
+      "epoch": 68.81,
+      "learning_rate": 4.210526315789474e-05,
+      "loss": 1.0359,
+      "step": 184000
+    },
+    {
+      "epoch": 68.85,
+      "learning_rate": 4.1842105263157894e-05,
+      "loss": 1.0327,
+      "step": 184100
+    },
+    {
+      "epoch": 68.89,
+      "learning_rate": 4.157894736842106e-05,
+      "loss": 1.0319,
+      "step": 184200
+    },
+    {
+      "epoch": 68.92,
+      "learning_rate": 4.131578947368421e-05,
+      "loss": 1.0332,
+      "step": 184300
+    },
+    {
+      "epoch": 68.96,
+      "learning_rate": 4.105263157894737e-05,
+      "loss": 1.0341,
+      "step": 184400
+    },
+    {
+      "epoch": 69.0,
+      "learning_rate": 4.078947368421053e-05,
+      "loss": 1.0318,
+      "step": 184500
+    },
+    {
+      "epoch": 69.04,
+      "learning_rate": 4.0526315789473684e-05,
+      "loss": 1.0382,
+      "step": 184600
+    },
+    {
+      "epoch": 69.07,
+      "learning_rate": 4.026315789473684e-05,
+      "loss": 1.0329,
+      "step": 184700
+    },
+    {
+      "epoch": 69.11,
+      "learning_rate": 4e-05,
+      "loss": 1.0322,
+      "step": 184800
+    },
+    {
+      "epoch": 69.15,
+      "learning_rate": 3.973684210526316e-05,
+      "loss": 1.0318,
+      "step": 184900
+    },
+    {
+      "epoch": 69.18,
+      "learning_rate": 3.9473684210526316e-05,
+      "loss": 1.0319,
+      "step": 185000
+    },
+    {
+      "epoch": 69.22,
+      "learning_rate": 3.921052631578947e-05,
+      "loss": 1.032,
+      "step": 185100
+    },
+    {
+      "epoch": 69.26,
+      "learning_rate": 3.8947368421052636e-05,
+      "loss": 1.0304,
+      "step": 185200
+    },
+    {
+      "epoch": 69.3,
+      "learning_rate": 3.8684210526315786e-05,
+      "loss": 1.0341,
+      "step": 185300
+    },
+    {
+      "epoch": 69.33,
+      "learning_rate": 3.842105263157895e-05,
+      "loss": 1.034,
+      "step": 185400
+    },
+    {
+      "epoch": 69.37,
+      "learning_rate": 3.8157894736842105e-05,
+      "loss": 1.0343,
+      "step": 185500
+    },
+    {
+      "epoch": 69.41,
+      "learning_rate": 3.789473684210527e-05,
+      "loss": 1.0309,
+      "step": 185600
+    },
+    {
+      "epoch": 69.45,
+      "learning_rate": 3.763157894736842e-05,
+      "loss": 1.0292,
+      "step": 185700
+    },
+    {
+      "epoch": 69.48,
+      "learning_rate": 3.736842105263158e-05,
+      "loss": 1.0294,
+      "step": 185800
+    },
+    {
+      "epoch": 69.52,
+      "learning_rate": 3.710526315789474e-05,
+      "loss": 1.0259,
+      "step": 185900
+    },
+    {
+      "epoch": 69.56,
+      "learning_rate": 3.6842105263157895e-05,
+      "loss": 1.0318,
+      "step": 186000
+    },
+    {
+      "epoch": 69.6,
+      "learning_rate": 3.657894736842105e-05,
+      "loss": 1.0326,
+      "step": 186100
+    },
+    {
+      "epoch": 69.63,
+      "learning_rate": 3.6315789473684214e-05,
+      "loss": 1.031,
+      "step": 186200
+    },
+    {
+      "epoch": 69.67,
+      "learning_rate": 3.605263157894737e-05,
+      "loss": 1.0294,
+      "step": 186300
+    },
+    {
+      "epoch": 69.71,
+      "learning_rate": 3.578947368421053e-05,
+      "loss": 1.0303,
+      "step": 186400
+    },
+    {
+      "epoch": 69.75,
+      "learning_rate": 3.5526315789473684e-05,
+      "loss": 1.0321,
+      "step": 186500
+    },
+    {
+      "epoch": 69.78,
+      "learning_rate": 3.526315789473685e-05,
+      "loss": 1.0296,
+      "step": 186600
+    },
+    {
+      "epoch": 69.82,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 1.0322,
+      "step": 186700
+    },
+    {
+      "epoch": 69.86,
+      "learning_rate": 3.473684210526316e-05,
+      "loss": 1.0286,
+      "step": 186800
+    },
+    {
+      "epoch": 69.9,
+      "learning_rate": 3.447368421052632e-05,
+      "loss": 1.0307,
+      "step": 186900
+    },
+    {
+      "epoch": 69.93,
+      "learning_rate": 3.421052631578948e-05,
+      "loss": 1.0308,
+      "step": 187000
+    },
+    {
+      "epoch": 69.97,
+      "learning_rate": 3.394736842105263e-05,
+      "loss": 1.0319,
+      "step": 187100
+    },
+    {
+      "epoch": 70.01,
+      "learning_rate": 3.3684210526315786e-05,
+      "loss": 1.0335,
+      "step": 187200
+    },
+    {
+      "epoch": 70.04,
+      "learning_rate": 3.342105263157895e-05,
+      "loss": 1.0306,
+      "step": 187300
+    },
+    {
+      "epoch": 70.08,
+      "learning_rate": 3.3157894736842106e-05,
+      "loss": 1.0271,
+      "step": 187400
+    },
+    {
+      "epoch": 70.12,
+      "learning_rate": 3.289473684210526e-05,
+      "loss": 1.0281,
+      "step": 187500
+    },
+    {
+      "epoch": 70.16,
+      "learning_rate": 3.263157894736842e-05,
+      "loss": 1.0296,
+      "step": 187600
+    },
+    {
+      "epoch": 70.19,
+      "learning_rate": 3.236842105263158e-05,
+      "loss": 1.0279,
+      "step": 187700
+    },
+    {
+      "epoch": 70.23,
+      "learning_rate": 3.210526315789473e-05,
+      "loss": 1.029,
+      "step": 187800
+    },
+    {
+      "epoch": 70.27,
+      "learning_rate": 3.1842105263157895e-05,
+      "loss": 1.0288,
+      "step": 187900
+    },
+    {
+      "epoch": 70.31,
+      "learning_rate": 3.157894736842105e-05,
+      "loss": 1.0283,
+      "step": 188000
+    },
+    {
+      "epoch": 70.34,
+      "learning_rate": 3.1315789473684215e-05,
+      "loss": 1.0284,
+      "step": 188100
+    },
+    {
+      "epoch": 70.38,
+      "learning_rate": 3.105263157894737e-05,
+      "loss": 1.0245,
+      "step": 188200
+    },
+    {
+      "epoch": 70.42,
+      "learning_rate": 3.078947368421053e-05,
+      "loss": 1.0298,
+      "step": 188300
+    },
+    {
+      "epoch": 70.46,
+      "learning_rate": 3.0526315789473684e-05,
+      "loss": 1.0267,
+      "step": 188400
+    },
+    {
+      "epoch": 70.49,
+      "learning_rate": 3.0263157894736844e-05,
+      "loss": 1.0256,
+      "step": 188500
+    },
+    {
+      "epoch": 70.53,
+      "learning_rate": 3e-05,
+      "loss": 1.0291,
+      "step": 188600
+    },
+    {
+      "epoch": 70.57,
+      "learning_rate": 2.973684210526316e-05,
+      "loss": 1.0269,
+      "step": 188700
+    },
+    {
+      "epoch": 70.61,
+      "learning_rate": 2.9473684210526317e-05,
+      "loss": 1.0276,
+      "step": 188800
+    },
+    {
+      "epoch": 70.64,
+      "learning_rate": 2.9210526315789477e-05,
+      "loss": 1.03,
+      "step": 188900
+    },
+    {
+      "epoch": 70.68,
+      "learning_rate": 2.8947368421052634e-05,
+      "loss": 1.0286,
+      "step": 189000
+    },
+    {
+      "epoch": 70.72,
+      "learning_rate": 2.868421052631579e-05,
+      "loss": 1.0277,
+      "step": 189100
+    },
+    {
+      "epoch": 70.76,
+      "learning_rate": 2.842105263157895e-05,
+      "loss": 1.0288,
+      "step": 189200
+    },
+    {
+      "epoch": 70.79,
+      "learning_rate": 2.8157894736842103e-05,
+      "loss": 1.028,
+      "step": 189300
+    },
+    {
+      "epoch": 70.83,
+      "learning_rate": 2.7894736842105263e-05,
+      "loss": 1.0264,
+      "step": 189400
+    },
+    {
+      "epoch": 70.87,
+      "learning_rate": 2.763157894736842e-05,
+      "loss": 1.0263,
+      "step": 189500
+    },
+    {
+      "epoch": 70.9,
+      "learning_rate": 2.736842105263158e-05,
+      "loss": 1.0269,
+      "step": 189600
+    },
+    {
+      "epoch": 70.94,
+      "learning_rate": 2.7105263157894736e-05,
+      "loss": 1.0272,
+      "step": 189700
+    },
+    {
+      "epoch": 70.98,
+      "learning_rate": 2.6842105263157896e-05,
+      "loss": 1.0309,
+      "step": 189800
+    },
+    {
+      "epoch": 71.02,
+      "learning_rate": 2.6578947368421052e-05,
+      "loss": 1.0341,
+      "step": 189900
+    },
+    {
+      "epoch": 71.05,
+      "learning_rate": 2.631578947368421e-05,
+      "loss": 1.0242,
+      "step": 190000
+    },
+    {
+      "epoch": 71.09,
+      "learning_rate": 2.605263157894737e-05,
+      "loss": 1.0255,
+      "step": 190100
+    },
+    {
+      "epoch": 71.13,
+      "learning_rate": 2.5789473684210525e-05,
+      "loss": 1.021,
+      "step": 190200
+    },
+    {
+      "epoch": 71.17,
+      "learning_rate": 2.5526315789473685e-05,
+      "loss": 1.0281,
+      "step": 190300
+    },
+    {
+      "epoch": 71.2,
+      "learning_rate": 2.526315789473684e-05,
+      "loss": 1.0254,
+      "step": 190400
+    },
+    {
+      "epoch": 71.24,
+      "learning_rate": 2.5e-05,
+      "loss": 1.0225,
+      "step": 190500
+    },
+    {
+      "epoch": 71.28,
+      "learning_rate": 2.4736842105263158e-05,
+      "loss": 1.0265,
+      "step": 190600
+    },
+    {
+      "epoch": 71.32,
+      "learning_rate": 2.4473684210526314e-05,
+      "loss": 1.0269,
+      "step": 190700
+    },
+    {
+      "epoch": 71.35,
+      "learning_rate": 2.4210526315789474e-05,
+      "loss": 1.0285,
+      "step": 190800
+    },
+    {
+      "epoch": 71.39,
+      "learning_rate": 2.394736842105263e-05,
+      "loss": 1.0241,
+      "step": 190900
+    },
+    {
+      "epoch": 71.43,
+      "learning_rate": 2.368421052631579e-05,
+      "loss": 1.0242,
+      "step": 191000
+    },
+    {
+      "epoch": 71.47,
+      "learning_rate": 2.3421052631578947e-05,
+      "loss": 1.0243,
+      "step": 191100
+    },
+    {
+      "epoch": 71.5,
+      "learning_rate": 2.3157894736842107e-05,
+      "loss": 1.0261,
+      "step": 191200
+    },
+    {
+      "epoch": 71.54,
+      "learning_rate": 2.2894736842105263e-05,
+      "loss": 1.0289,
+      "step": 191300
+    },
+    {
+      "epoch": 71.58,
+      "learning_rate": 2.2631578947368423e-05,
+      "loss": 1.0237,
+      "step": 191400
+    },
+    {
+      "epoch": 71.62,
+      "learning_rate": 2.236842105263158e-05,
+      "loss": 1.0262,
+      "step": 191500
+    },
+    {
+      "epoch": 71.65,
+      "learning_rate": 2.2105263157894736e-05,
+      "loss": 1.0249,
+      "step": 191600
+    },
+    {
+      "epoch": 71.69,
+      "learning_rate": 2.1842105263157896e-05,
+      "loss": 1.0255,
+      "step": 191700
+    },
+    {
+      "epoch": 71.73,
+      "learning_rate": 2.1578947368421053e-05,
+      "loss": 1.0242,
+      "step": 191800
+    },
+    {
+      "epoch": 71.77,
+      "learning_rate": 2.1315789473684212e-05,
+      "loss": 1.0227,
+      "step": 191900
+    },
+    {
+      "epoch": 71.8,
+      "learning_rate": 2.105263157894737e-05,
+      "loss": 1.0239,
+      "step": 192000
+    },
+    {
+      "epoch": 71.84,
+      "learning_rate": 2.078947368421053e-05,
+      "loss": 1.0227,
+      "step": 192100
+    },
+    {
+      "epoch": 71.88,
+      "learning_rate": 2.0526315789473685e-05,
+      "loss": 1.0271,
+      "step": 192200
+    },
+    {
+      "epoch": 71.91,
+      "learning_rate": 2.0263157894736842e-05,
+      "loss": 1.0233,
+      "step": 192300
+    },
+    {
+      "epoch": 71.95,
+      "learning_rate": 2e-05,
+      "loss": 1.0243,
+      "step": 192400
+    },
+    {
+      "epoch": 71.99,
+      "learning_rate": 1.9736842105263158e-05,
+      "loss": 1.0216,
+      "step": 192500
+    },
+    {
+      "epoch": 72.03,
+      "learning_rate": 1.9473684210526318e-05,
+      "loss": 1.0283,
+      "step": 192600
+    },
+    {
+      "epoch": 72.06,
+      "learning_rate": 1.9210526315789474e-05,
+      "loss": 1.0226,
+      "step": 192700
+    },
+    {
+      "epoch": 72.1,
+      "learning_rate": 1.8947368421052634e-05,
+      "loss": 1.0209,
+      "step": 192800
+    },
+    {
+      "epoch": 72.14,
+      "learning_rate": 1.868421052631579e-05,
+      "loss": 1.0217,
+      "step": 192900
+    },
+    {
+      "epoch": 72.18,
+      "learning_rate": 1.8421052631578947e-05,
+      "loss": 1.0206,
+      "step": 193000
+    },
+    {
+      "epoch": 72.21,
+      "learning_rate": 1.8157894736842107e-05,
+      "loss": 1.0235,
+      "step": 193100
+    },
+    {
+      "epoch": 72.25,
+      "learning_rate": 1.7894736842105264e-05,
+      "loss": 1.024,
+      "step": 193200
+    },
+    {
+      "epoch": 72.29,
+      "learning_rate": 1.7631578947368424e-05,
+      "loss": 1.0225,
+      "step": 193300
+    },
+    {
+      "epoch": 72.33,
+      "learning_rate": 1.736842105263158e-05,
+      "loss": 1.0217,
+      "step": 193400
+    },
+    {
+      "epoch": 72.36,
+      "learning_rate": 1.710526315789474e-05,
+      "loss": 1.0238,
+      "step": 193500
+    },
+    {
+      "epoch": 72.4,
+      "learning_rate": 1.6842105263157893e-05,
+      "loss": 1.0222,
+      "step": 193600
+    },
+    {
+      "epoch": 72.44,
+      "learning_rate": 1.6578947368421053e-05,
+      "loss": 1.0198,
+      "step": 193700
+    },
+    {
+      "epoch": 72.48,
+      "learning_rate": 1.631578947368421e-05,
+      "loss": 1.0228,
+      "step": 193800
+    },
+    {
+      "epoch": 72.51,
+      "learning_rate": 1.6052631578947366e-05,
+      "loss": 1.0217,
+      "step": 193900
+    },
+    {
+      "epoch": 72.55,
+      "learning_rate": 1.5789473684210526e-05,
+      "loss": 1.021,
+      "step": 194000
+    },
+    {
+      "epoch": 72.59,
+      "learning_rate": 1.5526315789473686e-05,
+      "loss": 1.0239,
+      "step": 194100
+    },
+    {
+      "epoch": 72.63,
+      "learning_rate": 1.5263157894736842e-05,
+      "loss": 1.0197,
+      "step": 194200
+    },
+    {
+      "epoch": 72.66,
+      "learning_rate": 1.5e-05,
+      "loss": 1.0192,
+      "step": 194300
+    },
+    {
+      "epoch": 72.7,
+      "learning_rate": 1.4736842105263159e-05,
+      "loss": 1.025,
+      "step": 194400
+    },
+    {
+      "epoch": 72.74,
+      "learning_rate": 1.4473684210526317e-05,
+      "loss": 1.023,
+      "step": 194500
+    },
+    {
+      "epoch": 72.77,
+      "learning_rate": 1.4210526315789475e-05,
+      "loss": 1.018,
+      "step": 194600
+    },
+    {
+      "epoch": 72.81,
+      "learning_rate": 1.3947368421052631e-05,
+      "loss": 1.0233,
+      "step": 194700
+    },
+    {
+      "epoch": 72.85,
+      "learning_rate": 1.368421052631579e-05,
+      "loss": 1.0205,
+      "step": 194800
+    },
+    {
+      "epoch": 72.89,
+      "learning_rate": 1.3421052631578948e-05,
+      "loss": 1.0253,
+      "step": 194900
+    },
+    {
+      "epoch": 72.92,
+      "learning_rate": 1.3157894736842104e-05,
+      "loss": 1.0193,
+      "step": 195000
+    },
+    {
+      "epoch": 72.96,
+      "learning_rate": 1.2894736842105262e-05,
+      "loss": 1.0221,
+      "step": 195100
+    },
+    {
+      "epoch": 73.0,
+      "learning_rate": 1.263157894736842e-05,
+      "loss": 1.0221,
+      "step": 195200
+    },
+    {
+      "epoch": 73.04,
+      "learning_rate": 1.2368421052631579e-05,
+      "loss": 1.0233,
+      "step": 195300
+    },
+    {
+      "epoch": 73.07,
+      "learning_rate": 1.2105263157894737e-05,
+      "loss": 1.0181,
+      "step": 195400
+    },
+    {
+      "epoch": 73.11,
+      "learning_rate": 1.1842105263157895e-05,
+      "loss": 1.0205,
+      "step": 195500
+    },
+    {
+      "epoch": 73.15,
+      "learning_rate": 1.1578947368421053e-05,
+      "loss": 1.0207,
+      "step": 195600
+    },
+    {
+      "epoch": 73.19,
+      "learning_rate": 1.1315789473684212e-05,
+      "loss": 1.0226,
+      "step": 195700
+    },
+    {
+      "epoch": 73.22,
+      "learning_rate": 1.1052631578947368e-05,
+      "loss": 1.0195,
+      "step": 195800
+    },
+    {
+      "epoch": 73.26,
+      "learning_rate": 1.0789473684210526e-05,
+      "loss": 1.0189,
+      "step": 195900
+    },
+    {
+      "epoch": 73.3,
+      "learning_rate": 1.0526315789473684e-05,
+      "loss": 1.021,
+      "step": 196000
+    },
+    {
+      "epoch": 73.34,
+      "learning_rate": 1.0263157894736843e-05,
+      "loss": 1.0205,
+      "step": 196100
+    },
+    {
+      "epoch": 73.37,
+      "learning_rate": 1e-05,
+      "loss": 1.0199,
+      "step": 196200
+    },
+    {
+      "epoch": 73.41,
+      "learning_rate": 9.736842105263159e-06,
+      "loss": 1.0202,
+      "step": 196300
+    },
+    {
+      "epoch": 73.45,
+      "learning_rate": 9.473684210526317e-06,
+      "loss": 1.0172,
+      "step": 196400
+    },
+    {
+      "epoch": 73.49,
+      "learning_rate": 9.210526315789474e-06,
+      "loss": 1.0191,
+      "step": 196500
+    },
+    {
+      "epoch": 73.52,
+      "learning_rate": 8.947368421052632e-06,
+      "loss": 1.0216,
+      "step": 196600
+    },
+    {
+      "epoch": 73.56,
+      "learning_rate": 8.68421052631579e-06,
+      "loss": 1.0197,
+      "step": 196700
+    },
+    {
+      "epoch": 73.6,
+      "learning_rate": 8.421052631578947e-06,
+      "loss": 1.0202,
+      "step": 196800
+    },
+    {
+      "epoch": 73.63,
+      "learning_rate": 8.157894736842105e-06,
+      "loss": 1.0219,
+      "step": 196900
+    },
+    {
+      "epoch": 73.67,
+      "learning_rate": 7.894736842105263e-06,
+      "loss": 1.0178,
+      "step": 197000
+    },
+    {
+      "epoch": 73.71,
+      "learning_rate": 7.631578947368421e-06,
+      "loss": 1.023,
+      "step": 197100
+    },
+    {
+      "epoch": 73.75,
+      "learning_rate": 7.368421052631579e-06,
+      "loss": 1.0181,
+      "step": 197200
+    },
+    {
+      "epoch": 73.78,
+      "learning_rate": 7.1052631578947375e-06,
+      "loss": 1.0208,
+      "step": 197300
+    },
+    {
+      "epoch": 73.82,
+      "learning_rate": 6.842105263157895e-06,
+      "loss": 1.0191,
+      "step": 197400
+    },
+    {
+      "epoch": 73.86,
+      "learning_rate": 6.578947368421052e-06,
+      "loss": 1.0193,
+      "step": 197500
+    },
+    {
+      "epoch": 73.9,
+      "learning_rate": 6.31578947368421e-06,
+      "loss": 1.0187,
+      "step": 197600
+    },
+    {
+      "epoch": 73.93,
+      "learning_rate": 6.0526315789473685e-06,
+      "loss": 1.0192,
+      "step": 197700
+    },
+    {
+      "epoch": 73.97,
+      "learning_rate": 5.789473684210527e-06,
+      "loss": 1.02,
+      "step": 197800
+    },
+    {
+      "epoch": 74.01,
+      "learning_rate": 5.526315789473684e-06,
+      "loss": 1.0228,
+      "step": 197900
+    },
+    {
+      "epoch": 74.05,
+      "learning_rate": 5.263157894736842e-06,
+      "loss": 1.0162,
+      "step": 198000
+    },
+    {
+      "epoch": 74.08,
+      "learning_rate": 5e-06,
+      "loss": 1.0187,
+      "step": 198100
+    },
+    {
+      "epoch": 74.12,
+      "learning_rate": 4.736842105263159e-06,
+      "loss": 1.0157,
+      "step": 198200
+    },
+    {
+      "epoch": 74.16,
+      "learning_rate": 4.473684210526316e-06,
+      "loss": 1.0182,
+      "step": 198300
+    },
+    {
+      "epoch": 74.2,
+      "learning_rate": 4.210526315789473e-06,
+      "loss": 1.0197,
+      "step": 198400
+    },
+    {
+      "epoch": 74.23,
+      "learning_rate": 3.9473684210526315e-06,
+      "loss": 1.0176,
+      "step": 198500
+    },
+    {
+      "epoch": 74.27,
+      "learning_rate": 3.6842105263157896e-06,
+      "loss": 1.0183,
+      "step": 198600
+    },
+    {
+      "epoch": 74.31,
+      "learning_rate": 3.4210526315789474e-06,
+      "loss": 1.019,
+      "step": 198700
+    },
+    {
+      "epoch": 74.35,
+      "learning_rate": 3.157894736842105e-06,
+      "loss": 1.0174,
+      "step": 198800
+    },
+    {
+      "epoch": 74.38,
+      "learning_rate": 2.8947368421052634e-06,
+      "loss": 1.0188,
+      "step": 198900
+    },
+    {
+      "epoch": 74.42,
+      "learning_rate": 2.631578947368421e-06,
+      "loss": 1.0175,
+      "step": 199000
+    },
+    {
+      "epoch": 74.46,
+      "learning_rate": 2.3684210526315793e-06,
+      "loss": 1.0187,
+      "step": 199100
+    },
+    {
+      "epoch": 74.5,
+      "learning_rate": 2.1052631578947366e-06,
+      "loss": 1.0183,
+      "step": 199200
+    },
+    {
+      "epoch": 74.53,
+      "learning_rate": 1.8421052631578948e-06,
+      "loss": 1.0195,
+      "step": 199300
+    },
+    {
+      "epoch": 74.57,
+      "learning_rate": 1.5789473684210526e-06,
+      "loss": 1.0219,
+      "step": 199400
+    },
+    {
+      "epoch": 74.61,
+      "learning_rate": 1.3157894736842106e-06,
+      "loss": 1.0212,
+      "step": 199500
+    },
+    {
+      "epoch": 74.64,
+      "learning_rate": 1.0526315789473683e-06,
+      "loss": 1.0155,
+      "step": 199600
+    },
+    {
+      "epoch": 74.68,
+      "learning_rate": 7.894736842105263e-07,
+      "loss": 1.0178,
+      "step": 199700
+    },
+    {
+      "epoch": 74.72,
+      "learning_rate": 5.263157894736842e-07,
+      "loss": 1.0159,
+      "step": 199800
+    },
+    {
+      "epoch": 74.76,
+      "learning_rate": 2.631578947368421e-07,
+      "loss": 1.0171,
+      "step": 199900
+    },
+    {
+      "epoch": 74.79,
+      "learning_rate": 0.0,
+      "loss": 1.0182,
+      "step": 200000
+    },
+    {
+      "epoch": 74.79,
+      "step": 200001,
+      "total_flos": 5.392567541842457e+19,
+      "train_loss": 5.0799270787734195e-06,
+      "train_runtime": 2733.1068,
+      "train_samples_per_second": 149866.079,
+      "train_steps_per_second": 73.177
+    }
+  ],
+  "max_steps": 200000,
+  "num_train_epochs": 75,
+  "total_flos": 5.392567541842457e+19,
+  "trial_name": null,
+  "trial_params": null
+}