diff --git "a/training_graph.json" "b/training_graph.json"
new file mode 100644--- /dev/null
+++ "b/training_graph.json"
@@ -0,0 +1,12878 @@
+[
+    {
+        "current_steps": 0,
+        "loss": 2.7048,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 1,
+        "loss": 2.7526,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 2,
+        "loss": 2.6562,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 3,
+        "loss": 2.7419,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 4,
+        "loss": 2.5812,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 5,
+        "loss": 2.5991,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 6,
+        "loss": 2.5459,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 7,
+        "loss": 2.4803,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 8,
+        "loss": 2.5611,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 9,
+        "loss": 2.4518,
+        "learning_rate": 0.0003,
+        "epoch": 0.0
+    },
+    {
+        "current_steps": 10,
+        "loss": 2.4524,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 11,
+        "loss": 2.5501,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 12,
+        "loss": 2.5409,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 13,
+        "loss": 2.4734,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 14,
+        "loss": 2.493,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 15,
+        "loss": 2.4557,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 16,
+        "loss": 2.4192,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 17,
+        "loss": 2.4948,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 18,
+        "loss": 2.5006,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 19,
+        "loss": 2.4647,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 20,
+        "loss": 2.5465,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 21,
+        "loss": 2.4755,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 22,
+        "loss": 2.5169,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 23,
+        "loss": 2.4674,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 24,
+        "loss": 2.5151,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 25,
+        "loss": 2.4723,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 26,
+        "loss": 2.5178,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 27,
+        "loss": 2.4769,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 28,
+        "loss": 2.5101,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 29,
+        "loss": 2.4615,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 30,
+        "loss": 2.4604,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 31,
+        "loss": 2.4354,
+        "learning_rate": 0.0003,
+        "epoch": 0.01
+    },
+    {
+        "current_steps": 32,
+        "loss": 2.4627,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 33,
+        "loss": 2.4618,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 34,
+        "loss": 2.5047,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 35,
+        "loss": 2.4749,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 36,
+        "loss": 2.4647,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 37,
+        "loss": 2.3877,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 38,
+        "loss": 2.4721,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 39,
+        "loss": 2.4314,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 40,
+        "loss": 2.5077,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 41,
+        "loss": 2.3907,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 42,
+        "loss": 2.4805,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 43,
+        "loss": 2.3909,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 44,
+        "loss": 2.5064,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 45,
+        "loss": 2.3992,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 46,
+        "loss": 2.4499,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 47,
+        "loss": 2.4783,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 48,
+        "loss": 2.5039,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 49,
+        "loss": 2.4714,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 50,
+        "loss": 2.4372,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 51,
+        "loss": 2.4431,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 52,
+        "loss": 2.466,
+        "learning_rate": 0.0003,
+        "epoch": 0.02
+    },
+    {
+        "current_steps": 53,
+        "loss": 2.3709,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 54,
+        "loss": 2.4227,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 55,
+        "loss": 2.3706,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 56,
+        "loss": 2.4684,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 57,
+        "loss": 2.4592,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 58,
+        "loss": 2.4403,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 59,
+        "loss": 2.4586,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 60,
+        "loss": 2.3636,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 61,
+        "loss": 2.4366,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 62,
+        "loss": 2.473,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 63,
+        "loss": 2.4793,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 64,
+        "loss": 2.4306,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 65,
+        "loss": 2.4869,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 66,
+        "loss": 2.3994,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 67,
+        "loss": 2.3857,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 68,
+        "loss": 2.4576,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 69,
+        "loss": 2.5285,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 70,
+        "loss": 2.3401,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 71,
+        "loss": 2.495,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 72,
+        "loss": 2.4957,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 73,
+        "loss": 2.3864,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 74,
+        "loss": 2.3983,
+        "learning_rate": 0.0003,
+        "epoch": 0.03
+    },
+    {
+        "current_steps": 75,
+        "loss": 2.4116,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 76,
+        "loss": 2.4405,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 77,
+        "loss": 2.4557,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 78,
+        "loss": 2.4104,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 79,
+        "loss": 2.459,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 80,
+        "loss": 2.4829,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 81,
+        "loss": 2.3423,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 82,
+        "loss": 2.359,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 83,
+        "loss": 2.4462,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 84,
+        "loss": 2.4407,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 85,
+        "loss": 2.4202,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 86,
+        "loss": 2.3809,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 87,
+        "loss": 2.4178,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 88,
+        "loss": 2.4617,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 89,
+        "loss": 2.3735,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 90,
+        "loss": 2.4576,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 91,
+        "loss": 2.3971,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 92,
+        "loss": 2.4171,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 93,
+        "loss": 2.4328,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 94,
+        "loss": 2.3502,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 95,
+        "loss": 2.4259,
+        "learning_rate": 0.0003,
+        "epoch": 0.04
+    },
+    {
+        "current_steps": 96,
+        "loss": 2.3692,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 97,
+        "loss": 2.4684,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 98,
+        "loss": 2.4487,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 99,
+        "loss": 2.3866,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 100,
+        "loss": 2.4341,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 101,
+        "loss": 2.4028,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 102,
+        "loss": 2.4825,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 103,
+        "loss": 2.4407,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 104,
+        "loss": 2.4803,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 105,
+        "loss": 2.4122,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 106,
+        "loss": 2.4217,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 107,
+        "loss": 2.3827,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 108,
+        "loss": 2.4521,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 109,
+        "loss": 2.4381,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 110,
+        "loss": 2.3614,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 111,
+        "loss": 2.372,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 112,
+        "loss": 2.3868,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 113,
+        "loss": 2.4503,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 114,
+        "loss": 2.439,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 115,
+        "loss": 2.3945,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 116,
+        "loss": 2.4461,
+        "learning_rate": 0.0003,
+        "epoch": 0.05
+    },
+    {
+        "current_steps": 117,
+        "loss": 2.4181,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 118,
+        "loss": 2.4085,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 119,
+        "loss": 2.4674,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 120,
+        "loss": 2.3792,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 121,
+        "loss": 2.4042,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 122,
+        "loss": 2.419,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 123,
+        "loss": 2.3457,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 124,
+        "loss": 2.4798,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 125,
+        "loss": 2.4898,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 126,
+        "loss": 2.4138,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 127,
+        "loss": 2.368,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 128,
+        "loss": 2.404,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 129,
+        "loss": 2.3565,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 130,
+        "loss": 2.3522,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 131,
+        "loss": 2.4006,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 132,
+        "loss": 2.4622,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 133,
+        "loss": 2.3683,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 134,
+        "loss": 2.3623,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 135,
+        "loss": 2.4193,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 136,
+        "loss": 2.3053,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 137,
+        "loss": 2.4266,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 138,
+        "loss": 2.3901,
+        "learning_rate": 0.0003,
+        "epoch": 0.06
+    },
+    {
+        "current_steps": 139,
+        "loss": 2.3806,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 140,
+        "loss": 2.4307,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 141,
+        "loss": 2.4044,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 142,
+        "loss": 2.4146,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 143,
+        "loss": 2.3919,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 144,
+        "loss": 2.4446,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 145,
+        "loss": 2.4773,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 146,
+        "loss": 2.3613,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 147,
+        "loss": 2.4627,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 148,
+        "loss": 2.4126,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 149,
+        "loss": 2.3061,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 150,
+        "loss": 2.4791,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 151,
+        "loss": 2.4377,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 152,
+        "loss": 2.3723,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 153,
+        "loss": 2.426,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 154,
+        "loss": 2.3773,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 155,
+        "loss": 2.4231,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 156,
+        "loss": 2.4029,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 157,
+        "loss": 2.4321,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 158,
+        "loss": 2.4207,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 159,
+        "loss": 2.4618,
+        "learning_rate": 0.0003,
+        "epoch": 0.07
+    },
+    {
+        "current_steps": 160,
+        "loss": 2.3932,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 161,
+        "loss": 2.4255,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 162,
+        "loss": 2.4686,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 163,
+        "loss": 2.4145,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 164,
+        "loss": 2.3652,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 165,
+        "loss": 2.269,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 166,
+        "loss": 2.3516,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 167,
+        "loss": 2.4036,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 168,
+        "loss": 2.397,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 169,
+        "loss": 2.3099,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 170,
+        "loss": 2.4232,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 171,
+        "loss": 2.3914,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 172,
+        "loss": 2.3736,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 173,
+        "loss": 2.4413,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 174,
+        "loss": 2.42,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 175,
+        "loss": 2.3916,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 176,
+        "loss": 2.4026,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 177,
+        "loss": 2.3642,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 178,
+        "loss": 2.4132,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 179,
+        "loss": 2.4125,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 180,
+        "loss": 2.4622,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 181,
+        "loss": 2.4256,
+        "learning_rate": 0.0003,
+        "epoch": 0.08
+    },
+    {
+        "current_steps": 182,
+        "loss": 2.3814,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 183,
+        "loss": 2.3784,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 184,
+        "loss": 2.3985,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 185,
+        "loss": 2.3616,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 186,
+        "loss": 2.3555,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 187,
+        "loss": 2.3334,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 188,
+        "loss": 2.4409,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 189,
+        "loss": 2.392,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 190,
+        "loss": 2.3879,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 191,
+        "loss": 2.4073,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 192,
+        "loss": 2.3886,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 193,
+        "loss": 2.3888,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 194,
+        "loss": 2.3977,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 195,
+        "loss": 2.3958,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 196,
+        "loss": 2.3586,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 197,
+        "loss": 2.3506,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 198,
+        "loss": 2.3957,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 199,
+        "loss": 2.393,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 200,
+        "loss": 2.3216,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 201,
+        "loss": 2.4215,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 202,
+        "loss": 2.3683,
+        "learning_rate": 0.0003,
+        "epoch": 0.09
+    },
+    {
+        "current_steps": 203,
+        "loss": 2.4368,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 204,
+        "loss": 2.4653,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 205,
+        "loss": 2.5357,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 206,
+        "loss": 2.5669,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 207,
+        "loss": 2.5965,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 208,
+        "loss": 2.5148,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 209,
+        "loss": 2.4388,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 210,
+        "loss": 2.4941,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 211,
+        "loss": 2.4111,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 212,
+        "loss": 2.4525,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 213,
+        "loss": 2.3939,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 214,
+        "loss": 2.4177,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 215,
+        "loss": 2.4107,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 216,
+        "loss": 2.3744,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 217,
+        "loss": 2.3617,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 218,
+        "loss": 2.4126,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 219,
+        "loss": 2.3641,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 220,
+        "loss": 2.4387,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 221,
+        "loss": 2.4007,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 222,
+        "loss": 2.4105,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 223,
+        "loss": 2.4269,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 224,
+        "loss": 2.4272,
+        "learning_rate": 0.0003,
+        "epoch": 0.1
+    },
+    {
+        "current_steps": 225,
+        "loss": 2.4023,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 226,
+        "loss": 2.423,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 227,
+        "loss": 2.2845,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 228,
+        "loss": 2.3589,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 229,
+        "loss": 2.4157,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 230,
+        "loss": 2.3965,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 231,
+        "loss": 2.3801,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 232,
+        "loss": 2.3356,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 233,
+        "loss": 2.3803,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 234,
+        "loss": 2.3355,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 235,
+        "loss": 2.3934,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 236,
+        "loss": 2.3537,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 237,
+        "loss": 2.3376,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 238,
+        "loss": 2.431,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 239,
+        "loss": 2.3611,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 240,
+        "loss": 2.3805,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 241,
+        "loss": 2.4057,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 242,
+        "loss": 2.3619,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 243,
+        "loss": 2.401,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 244,
+        "loss": 2.3644,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 245,
+        "loss": 2.3545,
+        "learning_rate": 0.0003,
+        "epoch": 0.11
+    },
+    {
+        "current_steps": 246,
+        "loss": 2.3792,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 247,
+        "loss": 2.3284,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 248,
+        "loss": 2.3724,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 249,
+        "loss": 2.3795,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 250,
+        "loss": 2.4504,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 251,
+        "loss": 2.395,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 252,
+        "loss": 2.4181,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 253,
+        "loss": 2.4252,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 254,
+        "loss": 2.3821,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 255,
+        "loss": 2.4408,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 256,
+        "loss": 2.3566,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 257,
+        "loss": 2.3419,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 258,
+        "loss": 2.3332,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 259,
+        "loss": 2.3933,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 260,
+        "loss": 2.3557,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 261,
+        "loss": 2.372,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 262,
+        "loss": 2.4204,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 263,
+        "loss": 2.3867,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 264,
+        "loss": 2.3961,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 265,
+        "loss": 2.3578,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 266,
+        "loss": 2.4194,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 267,
+        "loss": 2.4032,
+        "learning_rate": 0.0003,
+        "epoch": 0.12
+    },
+    {
+        "current_steps": 268,
+        "loss": 2.4088,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 269,
+        "loss": 2.3151,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 270,
+        "loss": 2.3511,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 271,
+        "loss": 2.4129,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 272,
+        "loss": 2.3532,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 273,
+        "loss": 2.347,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 274,
+        "loss": 2.3973,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 275,
+        "loss": 2.3838,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 276,
+        "loss": 2.4066,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 277,
+        "loss": 2.3166,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 278,
+        "loss": 2.3803,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 279,
+        "loss": 2.4185,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 280,
+        "loss": 2.377,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 281,
+        "loss": 2.3686,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 282,
+        "loss": 2.3738,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 283,
+        "loss": 2.368,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 284,
+        "loss": 2.4009,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 285,
+        "loss": 2.3419,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 286,
+        "loss": 2.3713,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 287,
+        "loss": 2.369,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 288,
+        "loss": 2.3872,
+        "learning_rate": 0.0003,
+        "epoch": 0.13
+    },
+    {
+        "current_steps": 289,
+        "loss": 2.361,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 290,
+        "loss": 2.4094,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 291,
+        "loss": 2.3253,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 292,
+        "loss": 2.3933,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 293,
+        "loss": 2.3984,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 294,
+        "loss": 2.4462,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 295,
+        "loss": 2.3644,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 296,
+        "loss": 2.4388,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 297,
+        "loss": 2.3295,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 298,
+        "loss": 2.4371,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 299,
+        "loss": 2.4308,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 300,
+        "loss": 2.3402,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 301,
+        "loss": 2.3613,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 302,
+        "loss": 2.3347,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 303,
+        "loss": 2.3324,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 304,
+        "loss": 2.3118,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 305,
+        "loss": 2.4177,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 306,
+        "loss": 2.4232,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 307,
+        "loss": 2.4408,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 308,
+        "loss": 2.3749,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 309,
+        "loss": 2.4249,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 310,
+        "loss": 2.3784,
+        "learning_rate": 0.0003,
+        "epoch": 0.14
+    },
+    {
+        "current_steps": 311,
+        "loss": 2.3506,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 312,
+        "loss": 2.3761,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 313,
+        "loss": 2.3684,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 314,
+        "loss": 2.4298,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 315,
+        "loss": 2.3208,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 316,
+        "loss": 2.382,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 317,
+        "loss": 2.3729,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 318,
+        "loss": 2.346,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 319,
+        "loss": 2.3379,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 320,
+        "loss": 2.3388,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 321,
+        "loss": 2.408,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 322,
+        "loss": 2.3684,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 323,
+        "loss": 2.34,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 324,
+        "loss": 2.4094,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 325,
+        "loss": 2.4531,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 326,
+        "loss": 2.3732,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 327,
+        "loss": 2.3291,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 328,
+        "loss": 2.385,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 329,
+        "loss": 2.3701,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 330,
+        "loss": 2.4045,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 331,
+        "loss": 2.3906,
+        "learning_rate": 0.0003,
+        "epoch": 0.15
+    },
+    {
+        "current_steps": 332,
+        "loss": 2.3537,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 333,
+        "loss": 2.3398,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 334,
+        "loss": 2.3941,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 335,
+        "loss": 2.3409,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 336,
+        "loss": 2.317,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 337,
+        "loss": 2.3872,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 338,
+        "loss": 2.3777,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 339,
+        "loss": 2.3644,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 340,
+        "loss": 2.4179,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 341,
+        "loss": 2.3748,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 342,
+        "loss": 2.301,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 343,
+        "loss": 2.3742,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 344,
+        "loss": 2.4047,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 345,
+        "loss": 2.3397,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 346,
+        "loss": 2.3272,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 347,
+        "loss": 2.3768,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 348,
+        "loss": 2.2975,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 349,
+        "loss": 2.3691,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 350,
+        "loss": 2.3526,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 351,
+        "loss": 2.3526,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 352,
+        "loss": 2.3358,
+        "learning_rate": 0.0003,
+        "epoch": 0.16
+    },
+    {
+        "current_steps": 353,
+        "loss": 2.4371,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 354,
+        "loss": 2.366,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 355,
+        "loss": 2.3468,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 356,
+        "loss": 2.3395,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 357,
+        "loss": 2.3546,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 358,
+        "loss": 2.4116,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 359,
+        "loss": 2.4655,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 360,
+        "loss": 2.4125,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 361,
+        "loss": 2.3103,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 362,
+        "loss": 2.359,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 363,
+        "loss": 2.3698,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 364,
+        "loss": 2.4184,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 365,
+        "loss": 2.3104,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 366,
+        "loss": 2.332,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 367,
+        "loss": 2.3033,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 368,
+        "loss": 2.3821,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 369,
+        "loss": 2.3692,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 370,
+        "loss": 2.299,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 371,
+        "loss": 2.4045,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 372,
+        "loss": 2.3852,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 373,
+        "loss": 2.3669,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 374,
+        "loss": 2.3453,
+        "learning_rate": 0.0003,
+        "epoch": 0.17
+    },
+    {
+        "current_steps": 375,
+        "loss": 2.3186,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 376,
+        "loss": 2.4389,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 377,
+        "loss": 2.3957,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 378,
+        "loss": 2.315,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 379,
+        "loss": 2.3833,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 380,
+        "loss": 2.4608,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 381,
+        "loss": 2.3811,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 382,
+        "loss": 2.4058,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 383,
+        "loss": 2.3525,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 384,
+        "loss": 2.3528,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 385,
+        "loss": 2.3957,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 386,
+        "loss": 2.3627,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 387,
+        "loss": 2.3597,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 388,
+        "loss": 2.3877,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 389,
+        "loss": 2.3426,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 390,
+        "loss": 2.3938,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 391,
+        "loss": 2.3962,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 392,
+        "loss": 2.3865,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 393,
+        "loss": 2.3666,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 394,
+        "loss": 2.4118,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 395,
+        "loss": 2.4083,
+        "learning_rate": 0.0003,
+        "epoch": 0.18
+    },
+    {
+        "current_steps": 396,
+        "loss": 2.4255,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 397,
+        "loss": 2.3995,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 398,
+        "loss": 2.3328,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 399,
+        "loss": 2.395,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 400,
+        "loss": 2.3599,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 401,
+        "loss": 2.2955,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 402,
+        "loss": 2.379,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 403,
+        "loss": 2.3777,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 404,
+        "loss": 2.317,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 405,
+        "loss": 2.3257,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 406,
+        "loss": 2.3969,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 407,
+        "loss": 2.4085,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 408,
+        "loss": 2.3554,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 409,
+        "loss": 2.3593,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 410,
+        "loss": 2.3738,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 411,
+        "loss": 2.3938,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 412,
+        "loss": 2.2975,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 413,
+        "loss": 2.3102,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 414,
+        "loss": 2.366,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 415,
+        "loss": 2.3496,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 416,
+        "loss": 2.3664,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 417,
+        "loss": 2.3601,
+        "learning_rate": 0.0003,
+        "epoch": 0.19
+    },
+    {
+        "current_steps": 418,
+        "loss": 2.3441,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 419,
+        "loss": 2.2654,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 420,
+        "loss": 2.3727,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 421,
+        "loss": 2.3426,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 422,
+        "loss": 2.4317,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 423,
+        "loss": 2.4174,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 424,
+        "loss": 2.4071,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 425,
+        "loss": 2.3535,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 426,
+        "loss": 2.3619,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 427,
+        "loss": 2.3205,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 428,
+        "loss": 2.4133,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 429,
+        "loss": 2.3687,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 430,
+        "loss": 2.3407,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 431,
+        "loss": 2.4365,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 432,
+        "loss": 2.325,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 433,
+        "loss": 2.3381,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 434,
+        "loss": 2.4272,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 435,
+        "loss": 2.347,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 436,
+        "loss": 2.3205,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 437,
+        "loss": 2.3637,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 438,
+        "loss": 2.3391,
+        "learning_rate": 0.0003,
+        "epoch": 0.2
+    },
+    {
+        "current_steps": 439,
+        "loss": 2.3013,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 440,
+        "loss": 2.4115,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 441,
+        "loss": 2.3472,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 442,
+        "loss": 2.3448,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 443,
+        "loss": 2.3082,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 444,
+        "loss": 2.3488,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 445,
+        "loss": 2.2777,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 446,
+        "loss": 2.337,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 447,
+        "loss": 2.4118,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 448,
+        "loss": 2.4143,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 449,
+        "loss": 2.4422,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 450,
+        "loss": 2.3645,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 451,
+        "loss": 2.3782,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 452,
+        "loss": 2.3208,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 453,
+        "loss": 2.3697,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 454,
+        "loss": 2.4207,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 455,
+        "loss": 2.3714,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 456,
+        "loss": 2.3618,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 457,
+        "loss": 2.3592,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 458,
+        "loss": 2.3187,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 459,
+        "loss": 2.3644,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 460,
+        "loss": 2.3551,
+        "learning_rate": 0.0003,
+        "epoch": 0.21
+    },
+    {
+        "current_steps": 461,
+        "loss": 2.3141,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 462,
+        "loss": 2.3437,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 463,
+        "loss": 2.2803,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 464,
+        "loss": 2.3936,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 465,
+        "loss": 2.362,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 466,
+        "loss": 2.4105,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 467,
+        "loss": 2.3281,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 468,
+        "loss": 2.3924,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 469,
+        "loss": 2.2735,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 470,
+        "loss": 2.3243,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 471,
+        "loss": 2.3633,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 472,
+        "loss": 2.3505,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 473,
+        "loss": 2.2917,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 474,
+        "loss": 2.3391,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 475,
+        "loss": 2.3373,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 476,
+        "loss": 2.3837,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 477,
+        "loss": 2.3273,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 478,
+        "loss": 2.3439,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 479,
+        "loss": 2.2794,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 480,
+        "loss": 2.3705,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 481,
+        "loss": 2.3506,
+        "learning_rate": 0.0003,
+        "epoch": 0.22
+    },
+    {
+        "current_steps": 482,
+        "loss": 2.3103,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 483,
+        "loss": 2.3669,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 484,
+        "loss": 2.3183,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 485,
+        "loss": 2.294,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 486,
+        "loss": 2.3588,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 487,
+        "loss": 2.3862,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 488,
+        "loss": 2.3954,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 489,
+        "loss": 2.3937,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 490,
+        "loss": 2.2862,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 491,
+        "loss": 2.3366,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 492,
+        "loss": 2.3119,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 493,
+        "loss": 2.349,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 494,
+        "loss": 2.3857,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 495,
+        "loss": 2.2601,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 496,
+        "loss": 2.3769,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 497,
+        "loss": 2.3675,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 498,
+        "loss": 2.3694,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 499,
+        "loss": 2.3868,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 500,
+        "loss": 2.3369,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 501,
+        "loss": 2.3437,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 502,
+        "loss": 2.3429,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 503,
+        "loss": 2.3362,
+        "learning_rate": 0.0003,
+        "epoch": 0.23
+    },
+    {
+        "current_steps": 504,
+        "loss": 2.3682,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 505,
+        "loss": 2.3662,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 506,
+        "loss": 2.439,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 507,
+        "loss": 2.4117,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 508,
+        "loss": 2.3799,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 509,
+        "loss": 2.344,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 510,
+        "loss": 2.3505,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 511,
+        "loss": 2.316,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 512,
+        "loss": 2.3286,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 513,
+        "loss": 2.3375,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 514,
+        "loss": 2.3502,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 515,
+        "loss": 2.3985,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 516,
+        "loss": 2.3231,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 517,
+        "loss": 2.3351,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 518,
+        "loss": 2.3882,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 519,
+        "loss": 2.3973,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 520,
+        "loss": 2.3423,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 521,
+        "loss": 2.3473,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 522,
+        "loss": 2.4099,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 523,
+        "loss": 2.3498,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 524,
+        "loss": 2.3189,
+        "learning_rate": 0.0003,
+        "epoch": 0.24
+    },
+    {
+        "current_steps": 525,
+        "loss": 2.3648,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 526,
+        "loss": 2.3607,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 527,
+        "loss": 2.2927,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 528,
+        "loss": 2.2814,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 529,
+        "loss": 2.3557,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 530,
+        "loss": 2.36,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 531,
+        "loss": 2.3257,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 532,
+        "loss": 2.3032,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 533,
+        "loss": 2.3321,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 534,
+        "loss": 2.2952,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 535,
+        "loss": 2.303,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 536,
+        "loss": 2.3236,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 537,
+        "loss": 2.3636,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 538,
+        "loss": 2.3424,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 539,
+        "loss": 2.357,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 540,
+        "loss": 2.341,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 541,
+        "loss": 2.3866,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 542,
+        "loss": 2.3427,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 543,
+        "loss": 2.3818,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 544,
+        "loss": 2.3124,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 545,
+        "loss": 2.3925,
+        "learning_rate": 0.0003,
+        "epoch": 0.25
+    },
+    {
+        "current_steps": 546,
+        "loss": 2.3879,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 547,
+        "loss": 2.3356,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 548,
+        "loss": 2.3629,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 549,
+        "loss": 2.3505,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 550,
+        "loss": 2.4222,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 551,
+        "loss": 2.319,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 552,
+        "loss": 2.3105,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 553,
+        "loss": 2.4,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 554,
+        "loss": 2.3599,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 555,
+        "loss": 2.3454,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 556,
+        "loss": 2.2975,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 557,
+        "loss": 2.349,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 558,
+        "loss": 2.4284,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 559,
+        "loss": 2.3409,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 560,
+        "loss": 2.4098,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 561,
+        "loss": 2.3786,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 562,
+        "loss": 2.3771,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 563,
+        "loss": 2.3698,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 564,
+        "loss": 2.2913,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 565,
+        "loss": 2.2947,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 566,
+        "loss": 2.3901,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 567,
+        "loss": 2.261,
+        "learning_rate": 0.0003,
+        "epoch": 0.26
+    },
+    {
+        "current_steps": 568,
+        "loss": 2.4077,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 569,
+        "loss": 2.3358,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 570,
+        "loss": 2.3394,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 571,
+        "loss": 2.3108,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 572,
+        "loss": 2.3655,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 573,
+        "loss": 2.2939,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 574,
+        "loss": 2.3086,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 575,
+        "loss": 2.362,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 576,
+        "loss": 2.3223,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 577,
+        "loss": 2.3772,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 578,
+        "loss": 2.3187,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 579,
+        "loss": 2.3837,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 580,
+        "loss": 2.3483,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 581,
+        "loss": 2.3313,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 582,
+        "loss": 2.3596,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 583,
+        "loss": 2.3505,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 584,
+        "loss": 2.3592,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 585,
+        "loss": 2.3126,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 586,
+        "loss": 2.274,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 587,
+        "loss": 2.2796,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 588,
+        "loss": 2.3292,
+        "learning_rate": 0.0003,
+        "epoch": 0.27
+    },
+    {
+        "current_steps": 589,
+        "loss": 2.3339,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 590,
+        "loss": 2.3681,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 591,
+        "loss": 2.3529,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 592,
+        "loss": 2.3514,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 593,
+        "loss": 2.4094,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 594,
+        "loss": 2.3499,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 595,
+        "loss": 2.3039,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 596,
+        "loss": 2.373,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 597,
+        "loss": 2.3078,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 598,
+        "loss": 2.3563,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 599,
+        "loss": 2.3145,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 600,
+        "loss": 2.4017,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 601,
+        "loss": 2.2746,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 602,
+        "loss": 2.3305,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 603,
+        "loss": 2.3059,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 604,
+        "loss": 2.3883,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 605,
+        "loss": 2.3686,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 606,
+        "loss": 2.3169,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 607,
+        "loss": 2.3628,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 608,
+        "loss": 2.4059,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 609,
+        "loss": 2.2302,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 610,
+        "loss": 2.394,
+        "learning_rate": 0.0003,
+        "epoch": 0.28
+    },
+    {
+        "current_steps": 611,
+        "loss": 2.3614,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 612,
+        "loss": 2.3225,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 613,
+        "loss": 2.3567,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 614,
+        "loss": 2.2806,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 615,
+        "loss": 2.3661,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 616,
+        "loss": 2.4167,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 617,
+        "loss": 2.2937,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 618,
+        "loss": 2.3421,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 619,
+        "loss": 2.3608,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 620,
+        "loss": 2.3759,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 621,
+        "loss": 2.423,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 622,
+        "loss": 2.2744,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 623,
+        "loss": 2.3943,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 624,
+        "loss": 2.3879,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 625,
+        "loss": 2.3871,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 626,
+        "loss": 2.315,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 627,
+        "loss": 2.3124,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 628,
+        "loss": 2.361,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 629,
+        "loss": 2.3071,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 630,
+        "loss": 2.2922,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 631,
+        "loss": 2.333,
+        "learning_rate": 0.0003,
+        "epoch": 0.29
+    },
+    {
+        "current_steps": 632,
+        "loss": 2.3228,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 633,
+        "loss": 2.3751,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 634,
+        "loss": 2.3352,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 635,
+        "loss": 2.3217,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 636,
+        "loss": 2.3672,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 637,
+        "loss": 2.3361,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 638,
+        "loss": 2.3653,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 639,
+        "loss": 2.2877,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 640,
+        "loss": 2.2828,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 641,
+        "loss": 2.3638,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 642,
+        "loss": 2.3645,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 643,
+        "loss": 2.3472,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 644,
+        "loss": 2.3381,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 645,
+        "loss": 2.3232,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 646,
+        "loss": 2.3514,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 647,
+        "loss": 2.3111,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 648,
+        "loss": 2.3267,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 649,
+        "loss": 2.3063,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 650,
+        "loss": 2.3357,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 651,
+        "loss": 2.312,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 652,
+        "loss": 2.3356,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 653,
+        "loss": 2.3148,
+        "learning_rate": 0.0003,
+        "epoch": 0.3
+    },
+    {
+        "current_steps": 654,
+        "loss": 2.3536,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 655,
+        "loss": 2.2787,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 656,
+        "loss": 2.368,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 657,
+        "loss": 2.3627,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 658,
+        "loss": 2.3005,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 659,
+        "loss": 2.3234,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 660,
+        "loss": 2.361,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 661,
+        "loss": 2.3232,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 662,
+        "loss": 2.2642,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 663,
+        "loss": 2.3594,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 664,
+        "loss": 2.3559,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 665,
+        "loss": 2.3017,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 666,
+        "loss": 2.3693,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 667,
+        "loss": 2.3406,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 668,
+        "loss": 2.367,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 669,
+        "loss": 2.3389,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 670,
+        "loss": 2.3543,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 671,
+        "loss": 2.3386,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 672,
+        "loss": 2.324,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 673,
+        "loss": 2.3065,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 674,
+        "loss": 2.3407,
+        "learning_rate": 0.0003,
+        "epoch": 0.31
+    },
+    {
+        "current_steps": 675,
+        "loss": 2.352,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 676,
+        "loss": 2.3801,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 677,
+        "loss": 2.2867,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 678,
+        "loss": 2.3256,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 679,
+        "loss": 2.3212,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 680,
+        "loss": 2.3664,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 681,
+        "loss": 2.3004,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 682,
+        "loss": 2.3014,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 683,
+        "loss": 2.2672,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 684,
+        "loss": 2.2934,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 685,
+        "loss": 2.3159,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 686,
+        "loss": 2.3558,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 687,
+        "loss": 2.3288,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 688,
+        "loss": 2.301,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 689,
+        "loss": 2.3713,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 690,
+        "loss": 2.304,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 691,
+        "loss": 2.3439,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 692,
+        "loss": 2.3314,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 693,
+        "loss": 2.3191,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 694,
+        "loss": 2.3424,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 695,
+        "loss": 2.4118,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 696,
+        "loss": 2.3345,
+        "learning_rate": 0.0003,
+        "epoch": 0.32
+    },
+    {
+        "current_steps": 697,
+        "loss": 2.2894,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 698,
+        "loss": 2.3158,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 699,
+        "loss": 2.3512,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 700,
+        "loss": 2.3758,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 701,
+        "loss": 2.304,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 702,
+        "loss": 2.3167,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 703,
+        "loss": 2.3249,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 704,
+        "loss": 2.3503,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 705,
+        "loss": 2.3046,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 706,
+        "loss": 2.3571,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 707,
+        "loss": 2.4044,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 708,
+        "loss": 2.3001,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 709,
+        "loss": 2.3492,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 710,
+        "loss": 2.3802,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 711,
+        "loss": 2.3512,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 712,
+        "loss": 2.3941,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 713,
+        "loss": 2.3724,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 714,
+        "loss": 2.2543,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 715,
+        "loss": 2.381,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 716,
+        "loss": 2.2963,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 717,
+        "loss": 2.3463,
+        "learning_rate": 0.0003,
+        "epoch": 0.33
+    },
+    {
+        "current_steps": 718,
+        "loss": 2.3224,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 719,
+        "loss": 2.3545,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 720,
+        "loss": 2.3841,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 721,
+        "loss": 2.308,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 722,
+        "loss": 2.3354,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 723,
+        "loss": 2.3721,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 724,
+        "loss": 2.2985,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 725,
+        "loss": 2.301,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 726,
+        "loss": 2.3321,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 727,
+        "loss": 2.3009,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 728,
+        "loss": 2.379,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 729,
+        "loss": 2.3314,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 730,
+        "loss": 2.3014,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 731,
+        "loss": 2.3515,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 732,
+        "loss": 2.2881,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 733,
+        "loss": 2.3086,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 734,
+        "loss": 2.3113,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 735,
+        "loss": 2.2921,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 736,
+        "loss": 2.3265,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 737,
+        "loss": 2.34,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 738,
+        "loss": 2.3519,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 739,
+        "loss": 2.382,
+        "learning_rate": 0.0003,
+        "epoch": 0.34
+    },
+    {
+        "current_steps": 740,
+        "loss": 2.3988,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 741,
+        "loss": 2.396,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 742,
+        "loss": 2.2379,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 743,
+        "loss": 2.3222,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 744,
+        "loss": 2.3237,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 745,
+        "loss": 2.3878,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 746,
+        "loss": 2.3586,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 747,
+        "loss": 2.2809,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 748,
+        "loss": 2.3219,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 749,
+        "loss": 2.3342,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 750,
+        "loss": 2.3634,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 751,
+        "loss": 2.3362,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 752,
+        "loss": 2.3108,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 753,
+        "loss": 2.3048,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 754,
+        "loss": 2.2799,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 755,
+        "loss": 2.3153,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 756,
+        "loss": 2.2726,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 757,
+        "loss": 2.3378,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 758,
+        "loss": 2.3124,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 759,
+        "loss": 2.3054,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 760,
+        "loss": 2.2571,
+        "learning_rate": 0.0003,
+        "epoch": 0.35
+    },
+    {
+        "current_steps": 761,
+        "loss": 2.3409,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 762,
+        "loss": 2.3268,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 763,
+        "loss": 2.2925,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 764,
+        "loss": 2.3112,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 765,
+        "loss": 2.3725,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 766,
+        "loss": 2.3185,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 767,
+        "loss": 2.2867,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 768,
+        "loss": 2.3389,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 769,
+        "loss": 2.3534,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 770,
+        "loss": 2.2977,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 771,
+        "loss": 2.3578,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 772,
+        "loss": 2.2945,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 773,
+        "loss": 2.3252,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 774,
+        "loss": 2.2721,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 775,
+        "loss": 2.2875,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 776,
+        "loss": 2.2933,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 777,
+        "loss": 2.3066,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 778,
+        "loss": 2.2837,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 779,
+        "loss": 2.359,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 780,
+        "loss": 2.335,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 781,
+        "loss": 2.2545,
+        "learning_rate": 0.0003,
+        "epoch": 0.36
+    },
+    {
+        "current_steps": 782,
+        "loss": 2.3453,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 783,
+        "loss": 2.3315,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 784,
+        "loss": 2.2457,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 785,
+        "loss": 2.3676,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 786,
+        "loss": 2.2858,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 787,
+        "loss": 2.3606,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 788,
+        "loss": 2.2487,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 789,
+        "loss": 2.3831,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 790,
+        "loss": 2.3606,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 791,
+        "loss": 2.391,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 792,
+        "loss": 2.2591,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 793,
+        "loss": 2.3521,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 794,
+        "loss": 2.3385,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 795,
+        "loss": 2.3598,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 796,
+        "loss": 2.3502,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 797,
+        "loss": 2.2823,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 798,
+        "loss": 2.3386,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 799,
+        "loss": 2.3262,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 800,
+        "loss": 2.2982,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 801,
+        "loss": 2.3396,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 802,
+        "loss": 2.3421,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 803,
+        "loss": 2.2627,
+        "learning_rate": 0.0003,
+        "epoch": 0.37
+    },
+    {
+        "current_steps": 804,
+        "loss": 2.3084,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 805,
+        "loss": 2.3251,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 806,
+        "loss": 2.2435,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 807,
+        "loss": 2.4017,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 808,
+        "loss": 2.3111,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 809,
+        "loss": 2.2748,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 810,
+        "loss": 2.2706,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 811,
+        "loss": 2.323,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 812,
+        "loss": 2.2466,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 813,
+        "loss": 2.3603,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 814,
+        "loss": 2.3975,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 815,
+        "loss": 2.3344,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 816,
+        "loss": 2.3432,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 817,
+        "loss": 2.3264,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 818,
+        "loss": 2.3287,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 819,
+        "loss": 2.3003,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 820,
+        "loss": 2.3043,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 821,
+        "loss": 2.3046,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 822,
+        "loss": 2.352,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 823,
+        "loss": 2.3031,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 824,
+        "loss": 2.3127,
+        "learning_rate": 0.0003,
+        "epoch": 0.38
+    },
+    {
+        "current_steps": 825,
+        "loss": 2.3283,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 826,
+        "loss": 2.4202,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 827,
+        "loss": 2.2913,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 828,
+        "loss": 2.3238,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 829,
+        "loss": 2.3297,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 830,
+        "loss": 2.3828,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 831,
+        "loss": 2.376,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 832,
+        "loss": 2.3024,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 833,
+        "loss": 2.2722,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 834,
+        "loss": 2.3864,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 835,
+        "loss": 2.3559,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 836,
+        "loss": 2.3176,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 837,
+        "loss": 2.3436,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 838,
+        "loss": 2.2673,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 839,
+        "loss": 2.3306,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 840,
+        "loss": 2.3008,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 841,
+        "loss": 2.2803,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 842,
+        "loss": 2.2989,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 843,
+        "loss": 2.362,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 844,
+        "loss": 2.2731,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 845,
+        "loss": 2.3318,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 846,
+        "loss": 2.3162,
+        "learning_rate": 0.0003,
+        "epoch": 0.39
+    },
+    {
+        "current_steps": 847,
+        "loss": 2.3379,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 848,
+        "loss": 2.3276,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 849,
+        "loss": 2.2963,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 850,
+        "loss": 2.2945,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 851,
+        "loss": 2.3496,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 852,
+        "loss": 2.385,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 853,
+        "loss": 2.3451,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 854,
+        "loss": 2.3375,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 855,
+        "loss": 2.3178,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 856,
+        "loss": 2.3014,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 857,
+        "loss": 2.3691,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 858,
+        "loss": 2.2855,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 859,
+        "loss": 2.4001,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 860,
+        "loss": 2.282,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 861,
+        "loss": 2.3329,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 862,
+        "loss": 2.38,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 863,
+        "loss": 2.4475,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 864,
+        "loss": 2.2638,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 865,
+        "loss": 2.327,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 866,
+        "loss": 2.2937,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 867,
+        "loss": 2.2513,
+        "learning_rate": 0.0003,
+        "epoch": 0.4
+    },
+    {
+        "current_steps": 868,
+        "loss": 2.3685,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 869,
+        "loss": 2.3097,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 870,
+        "loss": 2.3942,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 871,
+        "loss": 2.317,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 872,
+        "loss": 2.4447,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 873,
+        "loss": 2.3783,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 874,
+        "loss": 2.306,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 875,
+        "loss": 2.2455,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 876,
+        "loss": 2.3604,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 877,
+        "loss": 2.2804,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 878,
+        "loss": 2.3276,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 879,
+        "loss": 2.289,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 880,
+        "loss": 2.3129,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 881,
+        "loss": 2.3385,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 882,
+        "loss": 2.2647,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 883,
+        "loss": 2.3129,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 884,
+        "loss": 2.2881,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 885,
+        "loss": 2.3215,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 886,
+        "loss": 2.3622,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 887,
+        "loss": 2.2869,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 888,
+        "loss": 2.313,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 889,
+        "loss": 2.3201,
+        "learning_rate": 0.0003,
+        "epoch": 0.41
+    },
+    {
+        "current_steps": 890,
+        "loss": 2.34,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 891,
+        "loss": 2.3816,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 892,
+        "loss": 2.33,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 893,
+        "loss": 2.3417,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 894,
+        "loss": 2.3521,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 895,
+        "loss": 2.2432,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 896,
+        "loss": 2.2692,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 897,
+        "loss": 2.3155,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 898,
+        "loss": 2.3472,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 899,
+        "loss": 2.2958,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 900,
+        "loss": 2.3144,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 901,
+        "loss": 2.2343,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 902,
+        "loss": 2.2732,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 903,
+        "loss": 2.3372,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 904,
+        "loss": 2.2697,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 905,
+        "loss": 2.3615,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 906,
+        "loss": 2.3345,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 907,
+        "loss": 2.2963,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 908,
+        "loss": 2.2818,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 909,
+        "loss": 2.2461,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 910,
+        "loss": 2.3286,
+        "learning_rate": 0.0003,
+        "epoch": 0.42
+    },
+    {
+        "current_steps": 911,
+        "loss": 2.2727,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 912,
+        "loss": 2.3383,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 913,
+        "loss": 2.3175,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 914,
+        "loss": 2.3392,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 915,
+        "loss": 2.2798,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 916,
+        "loss": 2.2547,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 917,
+        "loss": 2.4019,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 918,
+        "loss": 2.3376,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 919,
+        "loss": 2.3279,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 920,
+        "loss": 2.3369,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 921,
+        "loss": 2.3503,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 922,
+        "loss": 2.3332,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 923,
+        "loss": 2.3229,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 924,
+        "loss": 2.3298,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 925,
+        "loss": 2.3318,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 926,
+        "loss": 2.2791,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 927,
+        "loss": 2.2874,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 928,
+        "loss": 2.2854,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 929,
+        "loss": 2.3041,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 930,
+        "loss": 2.2984,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 931,
+        "loss": 2.3598,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 932,
+        "loss": 2.2966,
+        "learning_rate": 0.0003,
+        "epoch": 0.43
+    },
+    {
+        "current_steps": 933,
+        "loss": 2.3386,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 934,
+        "loss": 2.2962,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 935,
+        "loss": 2.2881,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 936,
+        "loss": 2.2896,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 937,
+        "loss": 2.3786,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 938,
+        "loss": 2.2541,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 939,
+        "loss": 2.3389,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 940,
+        "loss": 2.3676,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 941,
+        "loss": 2.3391,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 942,
+        "loss": 2.3226,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 943,
+        "loss": 2.355,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 944,
+        "loss": 2.3099,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 945,
+        "loss": 2.2792,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 946,
+        "loss": 2.3516,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 947,
+        "loss": 2.2459,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 948,
+        "loss": 2.3407,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 949,
+        "loss": 2.2493,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 950,
+        "loss": 2.3185,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 951,
+        "loss": 2.3357,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 952,
+        "loss": 2.3059,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 953,
+        "loss": 2.2865,
+        "learning_rate": 0.0003,
+        "epoch": 0.44
+    },
+    {
+        "current_steps": 954,
+        "loss": 2.2969,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 955,
+        "loss": 2.3371,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 956,
+        "loss": 2.2714,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 957,
+        "loss": 2.2804,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 958,
+        "loss": 2.2947,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 959,
+        "loss": 2.3617,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 960,
+        "loss": 2.2383,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 961,
+        "loss": 2.3135,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 962,
+        "loss": 2.3138,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 963,
+        "loss": 2.2507,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 964,
+        "loss": 2.2694,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 965,
+        "loss": 2.3734,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 966,
+        "loss": 2.2944,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 967,
+        "loss": 2.3371,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 968,
+        "loss": 2.2639,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 969,
+        "loss": 2.3711,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 970,
+        "loss": 2.3617,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 971,
+        "loss": 2.3102,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 972,
+        "loss": 2.3238,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 973,
+        "loss": 2.337,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 974,
+        "loss": 2.3224,
+        "learning_rate": 0.0003,
+        "epoch": 0.45
+    },
+    {
+        "current_steps": 975,
+        "loss": 2.3512,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 976,
+        "loss": 2.3054,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 977,
+        "loss": 2.3649,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 978,
+        "loss": 2.2776,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 979,
+        "loss": 2.331,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 980,
+        "loss": 2.3052,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 981,
+        "loss": 2.3561,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 982,
+        "loss": 2.3507,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 983,
+        "loss": 2.3083,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 984,
+        "loss": 2.3453,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 985,
+        "loss": 2.3088,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 986,
+        "loss": 2.3093,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 987,
+        "loss": 2.413,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 988,
+        "loss": 2.3165,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 989,
+        "loss": 2.3251,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 990,
+        "loss": 2.2735,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 991,
+        "loss": 2.2752,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 992,
+        "loss": 2.2826,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 993,
+        "loss": 2.3069,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 994,
+        "loss": 2.3636,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 995,
+        "loss": 2.2692,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 996,
+        "loss": 2.2732,
+        "learning_rate": 0.0003,
+        "epoch": 0.46
+    },
+    {
+        "current_steps": 997,
+        "loss": 2.3183,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 998,
+        "loss": 2.2852,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 999,
+        "loss": 2.3768,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1000,
+        "loss": 2.333,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1001,
+        "loss": 2.2795,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1002,
+        "loss": 2.2315,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1003,
+        "loss": 2.2695,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1004,
+        "loss": 2.1976,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1005,
+        "loss": 2.2455,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1006,
+        "loss": 2.4439,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1007,
+        "loss": 2.2906,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1008,
+        "loss": 2.3146,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1009,
+        "loss": 2.331,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1010,
+        "loss": 2.2135,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1011,
+        "loss": 2.2961,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1012,
+        "loss": 2.3068,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1013,
+        "loss": 2.3135,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1014,
+        "loss": 2.3341,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1015,
+        "loss": 2.2497,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1016,
+        "loss": 2.2737,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1017,
+        "loss": 2.315,
+        "learning_rate": 0.0003,
+        "epoch": 0.47
+    },
+    {
+        "current_steps": 1018,
+        "loss": 2.3367,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1019,
+        "loss": 2.2259,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1020,
+        "loss": 2.3242,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1021,
+        "loss": 2.3295,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1022,
+        "loss": 2.3249,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1023,
+        "loss": 2.3459,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1024,
+        "loss": 2.2621,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1025,
+        "loss": 2.3176,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1026,
+        "loss": 2.3398,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1027,
+        "loss": 2.2588,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1028,
+        "loss": 2.3341,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1029,
+        "loss": 2.287,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1030,
+        "loss": 2.2347,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1031,
+        "loss": 2.3704,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1032,
+        "loss": 2.3494,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1033,
+        "loss": 2.3014,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1034,
+        "loss": 2.2495,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1035,
+        "loss": 2.3635,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1036,
+        "loss": 2.3178,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1037,
+        "loss": 2.2704,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1038,
+        "loss": 2.2858,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1039,
+        "loss": 2.2871,
+        "learning_rate": 0.0003,
+        "epoch": 0.48
+    },
+    {
+        "current_steps": 1040,
+        "loss": 2.2531,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1041,
+        "loss": 2.282,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1042,
+        "loss": 2.3077,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1043,
+        "loss": 2.4233,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1044,
+        "loss": 2.3446,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1045,
+        "loss": 2.2455,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1046,
+        "loss": 2.2722,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1047,
+        "loss": 2.2806,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1048,
+        "loss": 2.3119,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1049,
+        "loss": 2.3029,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1050,
+        "loss": 2.2761,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1051,
+        "loss": 2.3101,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1052,
+        "loss": 2.2863,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1053,
+        "loss": 2.3617,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1054,
+        "loss": 2.3722,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1055,
+        "loss": 2.3057,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1056,
+        "loss": 2.3453,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1057,
+        "loss": 2.3193,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1058,
+        "loss": 2.2917,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1059,
+        "loss": 2.3035,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1060,
+        "loss": 2.2869,
+        "learning_rate": 0.0003,
+        "epoch": 0.49
+    },
+    {
+        "current_steps": 1061,
+        "loss": 2.3241,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1062,
+        "loss": 2.2593,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1063,
+        "loss": 2.3203,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1064,
+        "loss": 2.3327,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1065,
+        "loss": 2.3668,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1066,
+        "loss": 2.2807,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1067,
+        "loss": 2.3826,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1068,
+        "loss": 2.2379,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1069,
+        "loss": 2.2804,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1070,
+        "loss": 2.2603,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1071,
+        "loss": 2.2974,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1072,
+        "loss": 2.2848,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1073,
+        "loss": 2.3342,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1074,
+        "loss": 2.3506,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1075,
+        "loss": 2.2908,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1076,
+        "loss": 2.3134,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1077,
+        "loss": 2.2358,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1078,
+        "loss": 2.3154,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1079,
+        "loss": 2.3316,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1080,
+        "loss": 2.3518,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1081,
+        "loss": 2.3357,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1082,
+        "loss": 2.367,
+        "learning_rate": 0.0003,
+        "epoch": 0.5
+    },
+    {
+        "current_steps": 1083,
+        "loss": 2.2745,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1084,
+        "loss": 2.2553,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1085,
+        "loss": 2.3437,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1086,
+        "loss": 2.3958,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1087,
+        "loss": 2.289,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1088,
+        "loss": 2.2958,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1089,
+        "loss": 2.3691,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1090,
+        "loss": 2.3051,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1091,
+        "loss": 2.2747,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1092,
+        "loss": 2.2434,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1093,
+        "loss": 2.3141,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1094,
+        "loss": 2.336,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1095,
+        "loss": 2.3772,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1096,
+        "loss": 2.305,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1097,
+        "loss": 2.255,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1098,
+        "loss": 2.3155,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1099,
+        "loss": 2.3678,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1100,
+        "loss": 2.2497,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1101,
+        "loss": 2.243,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1102,
+        "loss": 2.2731,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1103,
+        "loss": 2.2671,
+        "learning_rate": 0.0003,
+        "epoch": 0.51
+    },
+    {
+        "current_steps": 1104,
+        "loss": 2.3067,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1105,
+        "loss": 2.2733,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1106,
+        "loss": 2.3241,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1107,
+        "loss": 2.2607,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1108,
+        "loss": 2.4213,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1109,
+        "loss": 2.3533,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1110,
+        "loss": 2.2682,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1111,
+        "loss": 2.3158,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1112,
+        "loss": 2.3351,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1113,
+        "loss": 2.2803,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1114,
+        "loss": 2.2553,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1115,
+        "loss": 2.2727,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1116,
+        "loss": 2.3342,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1117,
+        "loss": 2.3334,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1118,
+        "loss": 2.2703,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1119,
+        "loss": 2.2828,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1120,
+        "loss": 2.3904,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1121,
+        "loss": 2.2842,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1122,
+        "loss": 2.353,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1123,
+        "loss": 2.3161,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1124,
+        "loss": 2.2799,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1125,
+        "loss": 2.3964,
+        "learning_rate": 0.0003,
+        "epoch": 0.52
+    },
+    {
+        "current_steps": 1126,
+        "loss": 2.3052,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1127,
+        "loss": 2.3244,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1128,
+        "loss": 2.3196,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1129,
+        "loss": 2.4233,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1130,
+        "loss": 2.2924,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1131,
+        "loss": 2.3054,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1132,
+        "loss": 2.2871,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1133,
+        "loss": 2.3079,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1134,
+        "loss": 2.2764,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1135,
+        "loss": 2.3025,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1136,
+        "loss": 2.2708,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1137,
+        "loss": 2.3688,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1138,
+        "loss": 2.2958,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1139,
+        "loss": 2.287,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1140,
+        "loss": 2.3072,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1141,
+        "loss": 2.28,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1142,
+        "loss": 2.3123,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1143,
+        "loss": 2.3349,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1144,
+        "loss": 2.2633,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1145,
+        "loss": 2.3353,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1146,
+        "loss": 2.2613,
+        "learning_rate": 0.0003,
+        "epoch": 0.53
+    },
+    {
+        "current_steps": 1147,
+        "loss": 2.3699,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1148,
+        "loss": 2.2525,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1149,
+        "loss": 2.2971,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1150,
+        "loss": 2.2702,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1151,
+        "loss": 2.3112,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1152,
+        "loss": 2.3511,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1153,
+        "loss": 2.2662,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1154,
+        "loss": 2.3152,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1155,
+        "loss": 2.3127,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1156,
+        "loss": 2.3417,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1157,
+        "loss": 2.2727,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1158,
+        "loss": 2.2688,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1159,
+        "loss": 2.4126,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1160,
+        "loss": 2.2909,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1161,
+        "loss": 2.3145,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1162,
+        "loss": 2.3,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1163,
+        "loss": 2.3413,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1164,
+        "loss": 2.3544,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1165,
+        "loss": 2.3571,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1166,
+        "loss": 2.2484,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1167,
+        "loss": 2.31,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1168,
+        "loss": 2.3539,
+        "learning_rate": 0.0003,
+        "epoch": 0.54
+    },
+    {
+        "current_steps": 1169,
+        "loss": 2.2443,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1170,
+        "loss": 2.266,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1171,
+        "loss": 2.2588,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1172,
+        "loss": 2.2252,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1173,
+        "loss": 2.32,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1174,
+        "loss": 2.343,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1175,
+        "loss": 2.3081,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1176,
+        "loss": 2.2617,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1177,
+        "loss": 2.305,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1178,
+        "loss": 2.2545,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1179,
+        "loss": 2.2916,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1180,
+        "loss": 2.3467,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1181,
+        "loss": 2.3185,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1182,
+        "loss": 2.3566,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1183,
+        "loss": 2.3157,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1184,
+        "loss": 2.3168,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1185,
+        "loss": 2.2499,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1186,
+        "loss": 2.3456,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1187,
+        "loss": 2.259,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1188,
+        "loss": 2.3044,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1189,
+        "loss": 2.2634,
+        "learning_rate": 0.0003,
+        "epoch": 0.55
+    },
+    {
+        "current_steps": 1190,
+        "loss": 2.2705,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1191,
+        "loss": 2.3002,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1192,
+        "loss": 2.2426,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1193,
+        "loss": 2.2649,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1194,
+        "loss": 2.3175,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1195,
+        "loss": 2.2951,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1196,
+        "loss": 2.3168,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1197,
+        "loss": 2.2838,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1198,
+        "loss": 2.3351,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1199,
+        "loss": 2.254,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1200,
+        "loss": 2.2996,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1201,
+        "loss": 2.3422,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1202,
+        "loss": 2.2534,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1203,
+        "loss": 2.2424,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1204,
+        "loss": 2.3185,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1205,
+        "loss": 2.2386,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1206,
+        "loss": 2.2647,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1207,
+        "loss": 2.2859,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1208,
+        "loss": 2.2684,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1209,
+        "loss": 2.3214,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1210,
+        "loss": 2.3579,
+        "learning_rate": 0.0003,
+        "epoch": 0.56
+    },
+    {
+        "current_steps": 1211,
+        "loss": 2.2767,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1212,
+        "loss": 2.3159,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1213,
+        "loss": 2.2672,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1214,
+        "loss": 2.2771,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1215,
+        "loss": 2.2706,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1216,
+        "loss": 2.2624,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1217,
+        "loss": 2.3062,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1218,
+        "loss": 2.2689,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1219,
+        "loss": 2.2931,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1220,
+        "loss": 2.2867,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1221,
+        "loss": 2.3409,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1222,
+        "loss": 2.2549,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1223,
+        "loss": 2.2495,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1224,
+        "loss": 2.3283,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1225,
+        "loss": 2.3108,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1226,
+        "loss": 2.2671,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1227,
+        "loss": 2.3198,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1228,
+        "loss": 2.2889,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1229,
+        "loss": 2.2365,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1230,
+        "loss": 2.3046,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1231,
+        "loss": 2.3307,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1232,
+        "loss": 2.2661,
+        "learning_rate": 0.0003,
+        "epoch": 0.57
+    },
+    {
+        "current_steps": 1233,
+        "loss": 2.3519,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1234,
+        "loss": 2.2856,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1235,
+        "loss": 2.3252,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1236,
+        "loss": 2.334,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1237,
+        "loss": 2.26,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1238,
+        "loss": 2.3704,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1239,
+        "loss": 2.3606,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1240,
+        "loss": 2.2904,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1241,
+        "loss": 2.2781,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1242,
+        "loss": 2.2527,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1243,
+        "loss": 2.3456,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1244,
+        "loss": 2.2454,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1245,
+        "loss": 2.3372,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1246,
+        "loss": 2.2372,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1247,
+        "loss": 2.3062,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1248,
+        "loss": 2.3014,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1249,
+        "loss": 2.2721,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1250,
+        "loss": 2.2957,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1251,
+        "loss": 2.3475,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1252,
+        "loss": 2.3317,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1253,
+        "loss": 2.2832,
+        "learning_rate": 0.0003,
+        "epoch": 0.58
+    },
+    {
+        "current_steps": 1254,
+        "loss": 2.2723,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1255,
+        "loss": 2.3086,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1256,
+        "loss": 2.3698,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1257,
+        "loss": 2.2755,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1258,
+        "loss": 2.3084,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1259,
+        "loss": 2.3499,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1260,
+        "loss": 2.2986,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1261,
+        "loss": 2.2705,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1262,
+        "loss": 2.3125,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1263,
+        "loss": 2.3708,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1264,
+        "loss": 2.2525,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1265,
+        "loss": 2.271,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1266,
+        "loss": 2.286,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1267,
+        "loss": 2.3171,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1268,
+        "loss": 2.2793,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1269,
+        "loss": 2.2703,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1270,
+        "loss": 2.3212,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1271,
+        "loss": 2.2701,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1272,
+        "loss": 2.3195,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1273,
+        "loss": 2.2402,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1274,
+        "loss": 2.2906,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1275,
+        "loss": 2.2993,
+        "learning_rate": 0.0003,
+        "epoch": 0.59
+    },
+    {
+        "current_steps": 1276,
+        "loss": 2.262,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1277,
+        "loss": 2.3408,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1278,
+        "loss": 2.3005,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1279,
+        "loss": 2.2577,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1280,
+        "loss": 2.3719,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1281,
+        "loss": 2.2667,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1282,
+        "loss": 2.3624,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1283,
+        "loss": 2.2892,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1284,
+        "loss": 2.2223,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1285,
+        "loss": 2.322,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1286,
+        "loss": 2.3408,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1287,
+        "loss": 2.255,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1288,
+        "loss": 2.2345,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1289,
+        "loss": 2.303,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1290,
+        "loss": 2.3121,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1291,
+        "loss": 2.2469,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1292,
+        "loss": 2.2731,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1293,
+        "loss": 2.2808,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1294,
+        "loss": 2.2895,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1295,
+        "loss": 2.2843,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1296,
+        "loss": 2.2463,
+        "learning_rate": 0.0003,
+        "epoch": 0.6
+    },
+    {
+        "current_steps": 1297,
+        "loss": 2.2996,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1298,
+        "loss": 2.3215,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1299,
+        "loss": 2.316,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1300,
+        "loss": 2.3253,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1301,
+        "loss": 2.2503,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1302,
+        "loss": 2.258,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1303,
+        "loss": 2.2678,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1304,
+        "loss": 2.3344,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1305,
+        "loss": 2.2506,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1306,
+        "loss": 2.3838,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1307,
+        "loss": 2.3046,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1308,
+        "loss": 2.2953,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1309,
+        "loss": 2.3316,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1310,
+        "loss": 2.2124,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1311,
+        "loss": 2.3177,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1312,
+        "loss": 2.2436,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1313,
+        "loss": 2.3016,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1314,
+        "loss": 2.3062,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1315,
+        "loss": 2.3,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1316,
+        "loss": 2.2892,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1317,
+        "loss": 2.3799,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1318,
+        "loss": 2.2499,
+        "learning_rate": 0.0003,
+        "epoch": 0.61
+    },
+    {
+        "current_steps": 1319,
+        "loss": 2.2797,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1320,
+        "loss": 2.252,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1321,
+        "loss": 2.278,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1322,
+        "loss": 2.2812,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1323,
+        "loss": 2.3014,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1324,
+        "loss": 2.2296,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1325,
+        "loss": 2.2549,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1326,
+        "loss": 2.273,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1327,
+        "loss": 2.1917,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1328,
+        "loss": 2.3567,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1329,
+        "loss": 2.3057,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1330,
+        "loss": 2.2978,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1331,
+        "loss": 2.3194,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1332,
+        "loss": 2.2561,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1333,
+        "loss": 2.2688,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1334,
+        "loss": 2.3211,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1335,
+        "loss": 2.2807,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1336,
+        "loss": 2.322,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1337,
+        "loss": 2.3585,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1338,
+        "loss": 2.2962,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1339,
+        "loss": 2.2398,
+        "learning_rate": 0.0003,
+        "epoch": 0.62
+    },
+    {
+        "current_steps": 1340,
+        "loss": 2.2685,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1341,
+        "loss": 2.2529,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1342,
+        "loss": 2.2663,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1343,
+        "loss": 2.2156,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1344,
+        "loss": 2.3421,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1345,
+        "loss": 2.2682,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1346,
+        "loss": 2.3032,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1347,
+        "loss": 2.2534,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1348,
+        "loss": 2.2951,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1349,
+        "loss": 2.3376,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1350,
+        "loss": 2.355,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1351,
+        "loss": 2.2963,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1352,
+        "loss": 2.2843,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1353,
+        "loss": 2.2404,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1354,
+        "loss": 2.2904,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1355,
+        "loss": 2.3029,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1356,
+        "loss": 2.258,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1357,
+        "loss": 2.2504,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1358,
+        "loss": 2.3709,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1359,
+        "loss": 2.2809,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1360,
+        "loss": 2.203,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1361,
+        "loss": 2.2531,
+        "learning_rate": 0.0003,
+        "epoch": 0.63
+    },
+    {
+        "current_steps": 1362,
+        "loss": 2.2833,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1363,
+        "loss": 2.3014,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1364,
+        "loss": 2.3284,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1365,
+        "loss": 2.3335,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1366,
+        "loss": 2.29,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1367,
+        "loss": 2.3941,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1368,
+        "loss": 2.3037,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1369,
+        "loss": 2.2371,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1370,
+        "loss": 2.2952,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1371,
+        "loss": 2.2814,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1372,
+        "loss": 2.3181,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1373,
+        "loss": 2.2707,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1374,
+        "loss": 2.2117,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1375,
+        "loss": 2.2762,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1376,
+        "loss": 2.2134,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1377,
+        "loss": 2.3676,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1378,
+        "loss": 2.3141,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1379,
+        "loss": 2.2387,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1380,
+        "loss": 2.3337,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1381,
+        "loss": 2.2959,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1382,
+        "loss": 2.2891,
+        "learning_rate": 0.0003,
+        "epoch": 0.64
+    },
+    {
+        "current_steps": 1383,
+        "loss": 2.3036,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1384,
+        "loss": 2.2666,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1385,
+        "loss": 2.2641,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1386,
+        "loss": 2.311,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1387,
+        "loss": 2.2964,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1388,
+        "loss": 2.2966,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1389,
+        "loss": 2.3029,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1390,
+        "loss": 2.2617,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1391,
+        "loss": 2.308,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1392,
+        "loss": 2.2734,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1393,
+        "loss": 2.3282,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1394,
+        "loss": 2.2921,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1395,
+        "loss": 2.2958,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1396,
+        "loss": 2.252,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1397,
+        "loss": 2.3021,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1398,
+        "loss": 2.3275,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1399,
+        "loss": 2.3421,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1400,
+        "loss": 2.2812,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1401,
+        "loss": 2.3555,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1402,
+        "loss": 2.2378,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1403,
+        "loss": 2.2866,
+        "learning_rate": 0.0003,
+        "epoch": 0.65
+    },
+    {
+        "current_steps": 1404,
+        "loss": 2.2775,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1405,
+        "loss": 2.2975,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1406,
+        "loss": 2.2837,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1407,
+        "loss": 2.2878,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1408,
+        "loss": 2.249,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1409,
+        "loss": 2.2947,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1410,
+        "loss": 2.2675,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1411,
+        "loss": 2.3232,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1412,
+        "loss": 2.2873,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1413,
+        "loss": 2.2881,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1414,
+        "loss": 2.333,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1415,
+        "loss": 2.2949,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1416,
+        "loss": 2.2311,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1417,
+        "loss": 2.3117,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1418,
+        "loss": 2.2864,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1419,
+        "loss": 2.1878,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1420,
+        "loss": 2.2926,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1421,
+        "loss": 2.2435,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1422,
+        "loss": 2.2934,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1423,
+        "loss": 2.2529,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1424,
+        "loss": 2.2419,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1425,
+        "loss": 2.3108,
+        "learning_rate": 0.0003,
+        "epoch": 0.66
+    },
+    {
+        "current_steps": 1426,
+        "loss": 2.3477,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1427,
+        "loss": 2.3473,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1428,
+        "loss": 2.2902,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1429,
+        "loss": 2.3345,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1430,
+        "loss": 2.285,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1431,
+        "loss": 2.2672,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1432,
+        "loss": 2.2698,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1433,
+        "loss": 2.3027,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1434,
+        "loss": 2.2593,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1435,
+        "loss": 2.267,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1436,
+        "loss": 2.3087,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1437,
+        "loss": 2.3099,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1438,
+        "loss": 2.2948,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1439,
+        "loss": 2.308,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1440,
+        "loss": 2.3447,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1441,
+        "loss": 2.2508,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1442,
+        "loss": 2.2982,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1443,
+        "loss": 2.2596,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1444,
+        "loss": 2.2381,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1445,
+        "loss": 2.3072,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1446,
+        "loss": 2.3308,
+        "learning_rate": 0.0003,
+        "epoch": 0.67
+    },
+    {
+        "current_steps": 1447,
+        "loss": 2.315,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1448,
+        "loss": 2.3256,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1449,
+        "loss": 2.2744,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1450,
+        "loss": 2.2618,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1451,
+        "loss": 2.2501,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1452,
+        "loss": 2.2896,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1453,
+        "loss": 2.3529,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1454,
+        "loss": 2.2477,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1455,
+        "loss": 2.2898,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1456,
+        "loss": 2.353,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1457,
+        "loss": 2.2197,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1458,
+        "loss": 2.3539,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1459,
+        "loss": 2.3037,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1460,
+        "loss": 2.2255,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1461,
+        "loss": 2.2799,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1462,
+        "loss": 2.2212,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1463,
+        "loss": 2.3202,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1464,
+        "loss": 2.2656,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1465,
+        "loss": 2.2917,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1466,
+        "loss": 2.3194,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1467,
+        "loss": 2.3286,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1468,
+        "loss": 2.2955,
+        "learning_rate": 0.0003,
+        "epoch": 0.68
+    },
+    {
+        "current_steps": 1469,
+        "loss": 2.3072,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1470,
+        "loss": 2.2913,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1471,
+        "loss": 2.2505,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1472,
+        "loss": 2.3948,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1473,
+        "loss": 2.3449,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1474,
+        "loss": 2.2889,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1475,
+        "loss": 2.2687,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1476,
+        "loss": 2.248,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1477,
+        "loss": 2.3147,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1478,
+        "loss": 2.2986,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1479,
+        "loss": 2.3198,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1480,
+        "loss": 2.2834,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1481,
+        "loss": 2.2018,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1482,
+        "loss": 2.2678,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1483,
+        "loss": 2.3154,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1484,
+        "loss": 2.266,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1485,
+        "loss": 2.2232,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1486,
+        "loss": 2.256,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1487,
+        "loss": 2.247,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1488,
+        "loss": 2.2435,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1489,
+        "loss": 2.3066,
+        "learning_rate": 0.0003,
+        "epoch": 0.69
+    },
+    {
+        "current_steps": 1490,
+        "loss": 2.2737,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1491,
+        "loss": 2.2226,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1492,
+        "loss": 2.2997,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1493,
+        "loss": 2.2572,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1494,
+        "loss": 2.2509,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1495,
+        "loss": 2.2945,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1496,
+        "loss": 2.3133,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1497,
+        "loss": 2.2524,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1498,
+        "loss": 2.2715,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1499,
+        "loss": 2.269,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1500,
+        "loss": 2.3078,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1501,
+        "loss": 2.2809,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1502,
+        "loss": 2.3093,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1503,
+        "loss": 2.2572,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1504,
+        "loss": 2.3662,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1505,
+        "loss": 2.3084,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1506,
+        "loss": 2.3186,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1507,
+        "loss": 2.2842,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1508,
+        "loss": 2.2981,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1509,
+        "loss": 2.3399,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1510,
+        "loss": 2.2988,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1511,
+        "loss": 2.2108,
+        "learning_rate": 0.0003,
+        "epoch": 0.7
+    },
+    {
+        "current_steps": 1512,
+        "loss": 2.2835,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1513,
+        "loss": 2.2895,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1514,
+        "loss": 2.3165,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1515,
+        "loss": 2.2951,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1516,
+        "loss": 2.3064,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1517,
+        "loss": 2.2922,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1518,
+        "loss": 2.2255,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1519,
+        "loss": 2.3375,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1520,
+        "loss": 2.2545,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1521,
+        "loss": 2.2946,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1522,
+        "loss": 2.3066,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1523,
+        "loss": 2.3262,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1524,
+        "loss": 2.2295,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1525,
+        "loss": 2.2642,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1526,
+        "loss": 2.2681,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1527,
+        "loss": 2.284,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1528,
+        "loss": 2.2628,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1529,
+        "loss": 2.294,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1530,
+        "loss": 2.2569,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1531,
+        "loss": 2.2488,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1532,
+        "loss": 2.2289,
+        "learning_rate": 0.0003,
+        "epoch": 0.71
+    },
+    {
+        "current_steps": 1533,
+        "loss": 2.2849,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1534,
+        "loss": 2.2451,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1535,
+        "loss": 2.2688,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1536,
+        "loss": 2.2526,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1537,
+        "loss": 2.2421,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1538,
+        "loss": 2.3417,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1539,
+        "loss": 2.3315,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1540,
+        "loss": 2.3343,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1541,
+        "loss": 2.3085,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1542,
+        "loss": 2.273,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1543,
+        "loss": 2.3391,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1544,
+        "loss": 2.2308,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1545,
+        "loss": 2.2504,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1546,
+        "loss": 2.3081,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1547,
+        "loss": 2.308,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1548,
+        "loss": 2.279,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1549,
+        "loss": 2.2288,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1550,
+        "loss": 2.2664,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1551,
+        "loss": 2.3049,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1552,
+        "loss": 2.3405,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1553,
+        "loss": 2.2744,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1554,
+        "loss": 2.2724,
+        "learning_rate": 0.0003,
+        "epoch": 0.72
+    },
+    {
+        "current_steps": 1555,
+        "loss": 2.2636,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1556,
+        "loss": 2.3043,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1557,
+        "loss": 2.3224,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1558,
+        "loss": 2.3286,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1559,
+        "loss": 2.3219,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1560,
+        "loss": 2.3442,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1561,
+        "loss": 2.2852,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1562,
+        "loss": 2.2158,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1563,
+        "loss": 2.3258,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1564,
+        "loss": 2.265,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1565,
+        "loss": 2.3149,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1566,
+        "loss": 2.3051,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1567,
+        "loss": 2.2568,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1568,
+        "loss": 2.2115,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1569,
+        "loss": 2.3099,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1570,
+        "loss": 2.2744,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1571,
+        "loss": 2.2085,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1572,
+        "loss": 2.2694,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1573,
+        "loss": 2.3262,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1574,
+        "loss": 2.2503,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1575,
+        "loss": 2.277,
+        "learning_rate": 0.0003,
+        "epoch": 0.73
+    },
+    {
+        "current_steps": 1576,
+        "loss": 2.2989,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1577,
+        "loss": 2.2102,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1578,
+        "loss": 2.3029,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1579,
+        "loss": 2.2489,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1580,
+        "loss": 2.2304,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1581,
+        "loss": 2.3251,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1582,
+        "loss": 2.2846,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1583,
+        "loss": 2.2841,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1584,
+        "loss": 2.2599,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1585,
+        "loss": 2.2828,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1586,
+        "loss": 2.2714,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1587,
+        "loss": 2.25,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1588,
+        "loss": 2.2862,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1589,
+        "loss": 2.2552,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1590,
+        "loss": 2.3398,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1591,
+        "loss": 2.3006,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1592,
+        "loss": 2.2799,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1593,
+        "loss": 2.3629,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1594,
+        "loss": 2.2622,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1595,
+        "loss": 2.3112,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1596,
+        "loss": 2.2634,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1597,
+        "loss": 2.2392,
+        "learning_rate": 0.0003,
+        "epoch": 0.74
+    },
+    {
+        "current_steps": 1598,
+        "loss": 2.2621,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1599,
+        "loss": 2.3598,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1600,
+        "loss": 2.2821,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1601,
+        "loss": 2.3293,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1602,
+        "loss": 2.2847,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1603,
+        "loss": 2.2151,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1604,
+        "loss": 2.2309,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1605,
+        "loss": 2.2926,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1606,
+        "loss": 2.3044,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1607,
+        "loss": 2.2449,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1608,
+        "loss": 2.2627,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1609,
+        "loss": 2.3134,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1610,
+        "loss": 2.3092,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1611,
+        "loss": 2.3249,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1612,
+        "loss": 2.2572,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1613,
+        "loss": 2.2991,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1614,
+        "loss": 2.2972,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1615,
+        "loss": 2.2368,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1616,
+        "loss": 2.2724,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1617,
+        "loss": 2.3026,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1618,
+        "loss": 2.3366,
+        "learning_rate": 0.0003,
+        "epoch": 0.75
+    },
+    {
+        "current_steps": 1619,
+        "loss": 2.2945,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1620,
+        "loss": 2.2674,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1621,
+        "loss": 2.2749,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1622,
+        "loss": 2.3103,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1623,
+        "loss": 2.2274,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1624,
+        "loss": 2.2727,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1625,
+        "loss": 2.317,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1626,
+        "loss": 2.276,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1627,
+        "loss": 2.2379,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1628,
+        "loss": 2.3721,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1629,
+        "loss": 2.2702,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1630,
+        "loss": 2.2408,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1631,
+        "loss": 2.255,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1632,
+        "loss": 2.2704,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1633,
+        "loss": 2.2085,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1634,
+        "loss": 2.2924,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1635,
+        "loss": 2.2483,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1636,
+        "loss": 2.3454,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1637,
+        "loss": 2.2915,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1638,
+        "loss": 2.2589,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1639,
+        "loss": 2.2455,
+        "learning_rate": 0.0003,
+        "epoch": 0.76
+    },
+    {
+        "current_steps": 1640,
+        "loss": 2.3092,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1641,
+        "loss": 2.2955,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1642,
+        "loss": 2.2909,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1643,
+        "loss": 2.2838,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1644,
+        "loss": 2.3469,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1645,
+        "loss": 2.259,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1646,
+        "loss": 2.31,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1647,
+        "loss": 2.2734,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1648,
+        "loss": 2.2936,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1649,
+        "loss": 2.2975,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1650,
+        "loss": 2.2994,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1651,
+        "loss": 2.2515,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1652,
+        "loss": 2.3507,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1653,
+        "loss": 2.2803,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1654,
+        "loss": 2.2778,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1655,
+        "loss": 2.2288,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1656,
+        "loss": 2.3291,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1657,
+        "loss": 2.2075,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1658,
+        "loss": 2.2262,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1659,
+        "loss": 2.2179,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1660,
+        "loss": 2.2717,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1661,
+        "loss": 2.2316,
+        "learning_rate": 0.0003,
+        "epoch": 0.77
+    },
+    {
+        "current_steps": 1662,
+        "loss": 2.2905,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1663,
+        "loss": 2.2899,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1664,
+        "loss": 2.3547,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1665,
+        "loss": 2.2648,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1666,
+        "loss": 2.3173,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1667,
+        "loss": 2.2774,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1668,
+        "loss": 2.3052,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1669,
+        "loss": 2.3003,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1670,
+        "loss": 2.2806,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1671,
+        "loss": 2.3187,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1672,
+        "loss": 2.3126,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1673,
+        "loss": 2.2966,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1674,
+        "loss": 2.2114,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1675,
+        "loss": 2.2674,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1676,
+        "loss": 2.2356,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1677,
+        "loss": 2.3098,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1678,
+        "loss": 2.373,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1679,
+        "loss": 2.2578,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1680,
+        "loss": 2.2623,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1681,
+        "loss": 2.3145,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1682,
+        "loss": 2.2958,
+        "learning_rate": 0.0003,
+        "epoch": 0.78
+    },
+    {
+        "current_steps": 1683,
+        "loss": 2.3075,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1684,
+        "loss": 2.3007,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1685,
+        "loss": 2.2732,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1686,
+        "loss": 2.3104,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1687,
+        "loss": 2.2932,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1688,
+        "loss": 2.2044,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1689,
+        "loss": 2.2769,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1690,
+        "loss": 2.3224,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1691,
+        "loss": 2.3427,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1692,
+        "loss": 2.2848,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1693,
+        "loss": 2.2653,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1694,
+        "loss": 2.277,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1695,
+        "loss": 2.3268,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1696,
+        "loss": 2.3617,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1697,
+        "loss": 2.2687,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1698,
+        "loss": 2.299,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1699,
+        "loss": 2.2411,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1700,
+        "loss": 2.2984,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1701,
+        "loss": 2.2778,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1702,
+        "loss": 2.2908,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1703,
+        "loss": 2.3329,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1704,
+        "loss": 2.2622,
+        "learning_rate": 0.0003,
+        "epoch": 0.79
+    },
+    {
+        "current_steps": 1705,
+        "loss": 2.29,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1706,
+        "loss": 2.3139,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1707,
+        "loss": 2.2741,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1708,
+        "loss": 2.326,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1709,
+        "loss": 2.2941,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1710,
+        "loss": 2.302,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1711,
+        "loss": 2.2953,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1712,
+        "loss": 2.2595,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1713,
+        "loss": 2.2793,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1714,
+        "loss": 2.2812,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1715,
+        "loss": 2.2589,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1716,
+        "loss": 2.255,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1717,
+        "loss": 2.2349,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1718,
+        "loss": 2.2315,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1719,
+        "loss": 2.253,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1720,
+        "loss": 2.3026,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1721,
+        "loss": 2.2814,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1722,
+        "loss": 2.2806,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1723,
+        "loss": 2.3083,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1724,
+        "loss": 2.2425,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1725,
+        "loss": 2.2486,
+        "learning_rate": 0.0003,
+        "epoch": 0.8
+    },
+    {
+        "current_steps": 1726,
+        "loss": 2.2965,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1727,
+        "loss": 2.3153,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1728,
+        "loss": 2.2916,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1729,
+        "loss": 2.2615,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1730,
+        "loss": 2.2894,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1731,
+        "loss": 2.2811,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1732,
+        "loss": 2.2319,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1733,
+        "loss": 2.3052,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1734,
+        "loss": 2.1984,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1735,
+        "loss": 2.2802,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1736,
+        "loss": 2.2936,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1737,
+        "loss": 2.3029,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1738,
+        "loss": 2.3493,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1739,
+        "loss": 2.2548,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1740,
+        "loss": 2.2946,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1741,
+        "loss": 2.2158,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1742,
+        "loss": 2.3266,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1743,
+        "loss": 2.2569,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1744,
+        "loss": 2.292,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1745,
+        "loss": 2.3103,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1746,
+        "loss": 2.2421,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1747,
+        "loss": 2.2885,
+        "learning_rate": 0.0003,
+        "epoch": 0.81
+    },
+    {
+        "current_steps": 1748,
+        "loss": 2.2224,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1749,
+        "loss": 2.2932,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1750,
+        "loss": 2.2842,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1751,
+        "loss": 2.1897,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1752,
+        "loss": 2.3291,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1753,
+        "loss": 2.276,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1754,
+        "loss": 2.2972,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1755,
+        "loss": 2.3312,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1756,
+        "loss": 2.3358,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1757,
+        "loss": 2.2746,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1758,
+        "loss": 2.2115,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1759,
+        "loss": 2.2822,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1760,
+        "loss": 2.271,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1761,
+        "loss": 2.2039,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1762,
+        "loss": 2.2878,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1763,
+        "loss": 2.2863,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1764,
+        "loss": 2.2521,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1765,
+        "loss": 2.3232,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1766,
+        "loss": 2.2939,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1767,
+        "loss": 2.3044,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1768,
+        "loss": 2.2602,
+        "learning_rate": 0.0003,
+        "epoch": 0.82
+    },
+    {
+        "current_steps": 1769,
+        "loss": 2.2702,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1770,
+        "loss": 2.2252,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1771,
+        "loss": 2.278,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1772,
+        "loss": 2.2695,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1773,
+        "loss": 2.32,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1774,
+        "loss": 2.2783,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1775,
+        "loss": 2.2187,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1776,
+        "loss": 2.2971,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1777,
+        "loss": 2.2539,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1778,
+        "loss": 2.3088,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1779,
+        "loss": 2.2869,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1780,
+        "loss": 2.2857,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1781,
+        "loss": 2.247,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1782,
+        "loss": 2.2904,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1783,
+        "loss": 2.2297,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1784,
+        "loss": 2.2908,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1785,
+        "loss": 2.2173,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1786,
+        "loss": 2.2787,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1787,
+        "loss": 2.2989,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1788,
+        "loss": 2.3113,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1789,
+        "loss": 2.2122,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1790,
+        "loss": 2.275,
+        "learning_rate": 0.0003,
+        "epoch": 0.83
+    },
+    {
+        "current_steps": 1791,
+        "loss": 2.3284,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1792,
+        "loss": 2.2385,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1793,
+        "loss": 2.3255,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1794,
+        "loss": 2.2471,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1795,
+        "loss": 2.2893,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1796,
+        "loss": 2.2616,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1797,
+        "loss": 2.2435,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1798,
+        "loss": 2.2675,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1799,
+        "loss": 2.2111,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1800,
+        "loss": 2.2703,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1801,
+        "loss": 2.185,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1802,
+        "loss": 2.3177,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1803,
+        "loss": 2.2366,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1804,
+        "loss": 2.2979,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1805,
+        "loss": 2.2615,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1806,
+        "loss": 2.3009,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1807,
+        "loss": 2.299,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1808,
+        "loss": 2.2753,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1809,
+        "loss": 2.2419,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1810,
+        "loss": 2.2649,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1811,
+        "loss": 2.2312,
+        "learning_rate": 0.0003,
+        "epoch": 0.84
+    },
+    {
+        "current_steps": 1812,
+        "loss": 2.2431,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1813,
+        "loss": 2.331,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1814,
+        "loss": 2.3089,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1815,
+        "loss": 2.3686,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1816,
+        "loss": 2.2527,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1817,
+        "loss": 2.2251,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1818,
+        "loss": 2.2937,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1819,
+        "loss": 2.2868,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1820,
+        "loss": 2.2756,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1821,
+        "loss": 2.3117,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1822,
+        "loss": 2.3389,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1823,
+        "loss": 2.3745,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1824,
+        "loss": 2.3081,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1825,
+        "loss": 2.2374,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1826,
+        "loss": 2.2238,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1827,
+        "loss": 2.3098,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1828,
+        "loss": 2.2547,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1829,
+        "loss": 2.3461,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1830,
+        "loss": 2.2419,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1831,
+        "loss": 2.3202,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1832,
+        "loss": 2.2273,
+        "learning_rate": 0.0003,
+        "epoch": 0.85
+    },
+    {
+        "current_steps": 1833,
+        "loss": 2.2343,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1834,
+        "loss": 2.2452,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1835,
+        "loss": 2.2537,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1836,
+        "loss": 2.2379,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1837,
+        "loss": 2.3374,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1838,
+        "loss": 2.2899,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1839,
+        "loss": 2.2671,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1840,
+        "loss": 2.2539,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1841,
+        "loss": 2.2806,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1842,
+        "loss": 2.2239,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1843,
+        "loss": 2.2589,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1844,
+        "loss": 2.2875,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1845,
+        "loss": 2.2839,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1846,
+        "loss": 2.2391,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1847,
+        "loss": 2.2421,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1848,
+        "loss": 2.289,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1849,
+        "loss": 2.2038,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1850,
+        "loss": 2.2087,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1851,
+        "loss": 2.2253,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1852,
+        "loss": 2.3044,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1853,
+        "loss": 2.2615,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1854,
+        "loss": 2.2523,
+        "learning_rate": 0.0003,
+        "epoch": 0.86
+    },
+    {
+        "current_steps": 1855,
+        "loss": 2.2918,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1856,
+        "loss": 2.2254,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1857,
+        "loss": 2.3228,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1858,
+        "loss": 2.2744,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1859,
+        "loss": 2.2689,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1860,
+        "loss": 2.3054,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1861,
+        "loss": 2.1658,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1862,
+        "loss": 2.3385,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1863,
+        "loss": 2.2685,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1864,
+        "loss": 2.2614,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1865,
+        "loss": 2.1829,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1866,
+        "loss": 2.2771,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1867,
+        "loss": 2.2491,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1868,
+        "loss": 2.2819,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1869,
+        "loss": 2.382,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1870,
+        "loss": 2.2524,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1871,
+        "loss": 2.274,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1872,
+        "loss": 2.2592,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1873,
+        "loss": 2.2633,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1874,
+        "loss": 2.2936,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1875,
+        "loss": 2.268,
+        "learning_rate": 0.0003,
+        "epoch": 0.87
+    },
+    {
+        "current_steps": 1876,
+        "loss": 2.2902,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1877,
+        "loss": 2.3047,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1878,
+        "loss": 2.2701,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1879,
+        "loss": 2.2581,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1880,
+        "loss": 2.2814,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1881,
+        "loss": 2.212,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1882,
+        "loss": 2.3007,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1883,
+        "loss": 2.2526,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1884,
+        "loss": 2.2995,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1885,
+        "loss": 2.2935,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1886,
+        "loss": 2.2104,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1887,
+        "loss": 2.2389,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1888,
+        "loss": 2.2654,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1889,
+        "loss": 2.1908,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1890,
+        "loss": 2.2805,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1891,
+        "loss": 2.2702,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1892,
+        "loss": 2.2223,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1893,
+        "loss": 2.2669,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1894,
+        "loss": 2.288,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1895,
+        "loss": 2.3161,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1896,
+        "loss": 2.329,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1897,
+        "loss": 2.2451,
+        "learning_rate": 0.0003,
+        "epoch": 0.88
+    },
+    {
+        "current_steps": 1898,
+        "loss": 2.3039,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1899,
+        "loss": 2.2636,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1900,
+        "loss": 2.2525,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1901,
+        "loss": 2.2608,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1902,
+        "loss": 2.289,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1903,
+        "loss": 2.3142,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1904,
+        "loss": 2.2912,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1905,
+        "loss": 2.2452,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1906,
+        "loss": 2.2559,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1907,
+        "loss": 2.2866,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1908,
+        "loss": 2.2884,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1909,
+        "loss": 2.232,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1910,
+        "loss": 2.2431,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1911,
+        "loss": 2.2988,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1912,
+        "loss": 2.1809,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1913,
+        "loss": 2.2853,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1914,
+        "loss": 2.2601,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1915,
+        "loss": 2.2322,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1916,
+        "loss": 2.2423,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1917,
+        "loss": 2.3043,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1918,
+        "loss": 2.2473,
+        "learning_rate": 0.0003,
+        "epoch": 0.89
+    },
+    {
+        "current_steps": 1919,
+        "loss": 2.358,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1920,
+        "loss": 2.297,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1921,
+        "loss": 2.2365,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1922,
+        "loss": 2.2417,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1923,
+        "loss": 2.288,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1924,
+        "loss": 2.2706,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1925,
+        "loss": 2.3152,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1926,
+        "loss": 2.2082,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1927,
+        "loss": 2.2161,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1928,
+        "loss": 2.2458,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1929,
+        "loss": 2.3018,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1930,
+        "loss": 2.2252,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1931,
+        "loss": 2.2446,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1932,
+        "loss": 2.2624,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1933,
+        "loss": 2.172,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1934,
+        "loss": 2.276,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1935,
+        "loss": 2.3029,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1936,
+        "loss": 2.2518,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1937,
+        "loss": 2.331,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1938,
+        "loss": 2.2765,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1939,
+        "loss": 2.2994,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1940,
+        "loss": 2.3118,
+        "learning_rate": 0.0003,
+        "epoch": 0.9
+    },
+    {
+        "current_steps": 1941,
+        "loss": 2.2947,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1942,
+        "loss": 2.2363,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1943,
+        "loss": 2.2751,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1944,
+        "loss": 2.227,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1945,
+        "loss": 2.3051,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1946,
+        "loss": 2.2243,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1947,
+        "loss": 2.2569,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1948,
+        "loss": 2.286,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1949,
+        "loss": 2.2795,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1950,
+        "loss": 2.3207,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1951,
+        "loss": 2.3232,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1952,
+        "loss": 2.2727,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1953,
+        "loss": 2.3233,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1954,
+        "loss": 2.3227,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1955,
+        "loss": 2.2858,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1956,
+        "loss": 2.2599,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1957,
+        "loss": 2.2639,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1958,
+        "loss": 2.3121,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1959,
+        "loss": 2.4,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1960,
+        "loss": 2.31,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1961,
+        "loss": 2.2408,
+        "learning_rate": 0.0003,
+        "epoch": 0.91
+    },
+    {
+        "current_steps": 1962,
+        "loss": 2.283,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1963,
+        "loss": 2.2278,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1964,
+        "loss": 2.2841,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1965,
+        "loss": 2.3802,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1966,
+        "loss": 2.2662,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1967,
+        "loss": 2.2936,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1968,
+        "loss": 2.2353,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1969,
+        "loss": 2.2292,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1970,
+        "loss": 2.1807,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1971,
+        "loss": 2.1833,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1972,
+        "loss": 2.1842,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1973,
+        "loss": 2.2951,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1974,
+        "loss": 2.2776,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1975,
+        "loss": 2.2526,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1976,
+        "loss": 2.2848,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1977,
+        "loss": 2.1991,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1978,
+        "loss": 2.2538,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1979,
+        "loss": 2.2231,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1980,
+        "loss": 2.2685,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1981,
+        "loss": 2.2802,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1982,
+        "loss": 2.3362,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1983,
+        "loss": 2.2315,
+        "learning_rate": 0.0003,
+        "epoch": 0.92
+    },
+    {
+        "current_steps": 1984,
+        "loss": 2.2728,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1985,
+        "loss": 2.2509,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1986,
+        "loss": 2.2579,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1987,
+        "loss": 2.2318,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1988,
+        "loss": 2.2827,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1989,
+        "loss": 2.2477,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1990,
+        "loss": 2.3573,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1991,
+        "loss": 2.2618,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1992,
+        "loss": 2.2423,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1993,
+        "loss": 2.232,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1994,
+        "loss": 2.2739,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1995,
+        "loss": 2.256,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1996,
+        "loss": 2.2568,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1997,
+        "loss": 2.2896,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1998,
+        "loss": 2.2416,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 1999,
+        "loss": 2.2691,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 2000,
+        "loss": 2.2776,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 2001,
+        "loss": 2.2092,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 2002,
+        "loss": 2.279,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 2003,
+        "loss": 2.2027,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 2004,
+        "loss": 2.3049,
+        "learning_rate": 0.0003,
+        "epoch": 0.93
+    },
+    {
+        "current_steps": 2005,
+        "loss": 2.308,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2006,
+        "loss": 2.2406,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2007,
+        "loss": 2.2224,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2008,
+        "loss": 2.2423,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2009,
+        "loss": 2.2882,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2010,
+        "loss": 2.2794,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2011,
+        "loss": 2.2848,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2012,
+        "loss": 2.2446,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2013,
+        "loss": 2.2454,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2014,
+        "loss": 2.2659,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2015,
+        "loss": 2.1905,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2016,
+        "loss": 2.2599,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2017,
+        "loss": 2.2922,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2018,
+        "loss": 2.2009,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2019,
+        "loss": 2.3025,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2020,
+        "loss": 2.307,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2021,
+        "loss": 2.2067,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2022,
+        "loss": 2.2621,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2023,
+        "loss": 2.2708,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2024,
+        "loss": 2.2988,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2025,
+        "loss": 2.2663,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2026,
+        "loss": 2.2765,
+        "learning_rate": 0.0003,
+        "epoch": 0.94
+    },
+    {
+        "current_steps": 2027,
+        "loss": 2.3572,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2028,
+        "loss": 2.2893,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2029,
+        "loss": 2.2576,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2030,
+        "loss": 2.2316,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2031,
+        "loss": 2.293,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2032,
+        "loss": 2.2523,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2033,
+        "loss": 2.2927,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2034,
+        "loss": 2.2581,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2035,
+        "loss": 2.2402,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2036,
+        "loss": 2.2915,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2037,
+        "loss": 2.2889,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2038,
+        "loss": 2.2268,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2039,
+        "loss": 2.324,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2040,
+        "loss": 2.2472,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2041,
+        "loss": 2.2189,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2042,
+        "loss": 2.2334,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2043,
+        "loss": 2.3102,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2044,
+        "loss": 2.1705,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2045,
+        "loss": 2.2662,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2046,
+        "loss": 2.2614,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2047,
+        "loss": 2.2613,
+        "learning_rate": 0.0003,
+        "epoch": 0.95
+    },
+    {
+        "current_steps": 2048,
+        "loss": 2.3485,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2049,
+        "loss": 2.2705,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2050,
+        "loss": 2.2206,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2051,
+        "loss": 2.3229,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2052,
+        "loss": 2.2388,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2053,
+        "loss": 2.22,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2054,
+        "loss": 2.2632,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2055,
+        "loss": 2.3049,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2056,
+        "loss": 2.2174,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2057,
+        "loss": 2.2851,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2058,
+        "loss": 2.2548,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2059,
+        "loss": 2.2495,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2060,
+        "loss": 2.351,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2061,
+        "loss": 2.2795,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2062,
+        "loss": 2.2076,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2063,
+        "loss": 2.2773,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2064,
+        "loss": 2.3454,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2065,
+        "loss": 2.274,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2066,
+        "loss": 2.2393,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2067,
+        "loss": 2.2777,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2068,
+        "loss": 2.3141,
+        "learning_rate": 0.0003,
+        "epoch": 0.96
+    },
+    {
+        "current_steps": 2069,
+        "loss": 2.3013,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2070,
+        "loss": 2.2939,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2071,
+        "loss": 2.2872,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2072,
+        "loss": 2.2111,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2073,
+        "loss": 2.2594,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2074,
+        "loss": 2.2835,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2075,
+        "loss": 2.2438,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2076,
+        "loss": 2.2873,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2077,
+        "loss": 2.2968,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2078,
+        "loss": 2.262,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2079,
+        "loss": 2.3385,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2080,
+        "loss": 2.3206,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2081,
+        "loss": 2.2917,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2082,
+        "loss": 2.2311,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2083,
+        "loss": 2.318,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2084,
+        "loss": 2.22,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2085,
+        "loss": 2.3102,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2086,
+        "loss": 2.2898,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2087,
+        "loss": 2.2548,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2088,
+        "loss": 2.2478,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2089,
+        "loss": 2.2002,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2090,
+        "loss": 2.2368,
+        "learning_rate": 0.0003,
+        "epoch": 0.97
+    },
+    {
+        "current_steps": 2091,
+        "loss": 2.3235,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2092,
+        "loss": 2.2392,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2093,
+        "loss": 2.273,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2094,
+        "loss": 2.2998,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2095,
+        "loss": 2.2588,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2096,
+        "loss": 2.3072,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2097,
+        "loss": 2.2621,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2098,
+        "loss": 2.3242,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2099,
+        "loss": 2.2248,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2100,
+        "loss": 2.3738,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2101,
+        "loss": 2.3279,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2102,
+        "loss": 2.2309,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2103,
+        "loss": 2.2724,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2104,
+        "loss": 2.3312,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2105,
+        "loss": 2.269,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2106,
+        "loss": 2.2353,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2107,
+        "loss": 2.2389,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2108,
+        "loss": 2.3425,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2109,
+        "loss": 2.2289,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2110,
+        "loss": 2.2837,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2111,
+        "loss": 2.2357,
+        "learning_rate": 0.0003,
+        "epoch": 0.98
+    },
+    {
+        "current_steps": 2112,
+        "loss": 2.2346,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2113,
+        "loss": 2.318,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2114,
+        "loss": 2.2192,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2115,
+        "loss": 2.2928,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2116,
+        "loss": 2.2402,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2117,
+        "loss": 2.2331,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2118,
+        "loss": 2.3154,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2119,
+        "loss": 2.2703,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2120,
+        "loss": 2.3003,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2121,
+        "loss": 2.2752,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2122,
+        "loss": 2.2814,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2123,
+        "loss": 2.2562,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2124,
+        "loss": 2.2281,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2125,
+        "loss": 2.2618,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2126,
+        "loss": 2.2616,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2127,
+        "loss": 2.2828,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2128,
+        "loss": 2.2527,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2129,
+        "loss": 2.2948,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2130,
+        "loss": 2.2088,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2131,
+        "loss": 2.2882,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2132,
+        "loss": 2.3212,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2133,
+        "loss": 2.2981,
+        "learning_rate": 0.0003,
+        "epoch": 0.99
+    },
+    {
+        "current_steps": 2134,
+        "loss": 2.2704,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2135,
+        "loss": 2.3096,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2136,
+        "loss": 2.3118,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2137,
+        "loss": 2.1919,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2138,
+        "loss": 2.2986,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2139,
+        "loss": 2.2452,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2140,
+        "loss": 2.2719,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2141,
+        "loss": 2.264,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2142,
+        "loss": 2.2058,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2143,
+        "loss": 2.239,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2144,
+        "loss": 2.3056,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    },
+    {
+        "current_steps": 2144,
+        "loss": 2.3056,
+        "learning_rate": 0.0003,
+        "epoch": 1.0
+    }
+]
\ No newline at end of file