diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,22295 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 100.0,
+  "global_step": 3700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0,
+      "loss": 15.6195,
+      "step": 1
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0,
+      "loss": 15.5073,
+      "step": 2
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1e-06,
+      "loss": 14.4581,
+      "step": 3
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2e-06,
+      "loss": 14.5043,
+      "step": 4
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 3e-06,
+      "loss": 13.8272,
+      "step": 5
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4e-06,
+      "loss": 15.1966,
+      "step": 6
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5e-06,
+      "loss": 16.2804,
+      "step": 7
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6e-06,
+      "loss": 15.0119,
+      "step": 8
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7e-06,
+      "loss": 14.653,
+      "step": 9
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8e-06,
+      "loss": 14.6559,
+      "step": 10
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9e-06,
+      "loss": 13.6463,
+      "step": 11
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1e-05,
+      "loss": 14.1526,
+      "step": 12
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.1e-05,
+      "loss": 15.4964,
+      "step": 13
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.2e-05,
+      "loss": 15.1013,
+      "step": 14
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3e-05,
+      "loss": 14.5165,
+      "step": 15
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.4e-05,
+      "loss": 13.8915,
+      "step": 16
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5e-05,
+      "loss": 13.5514,
+      "step": 17
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6e-05,
+      "loss": 13.2658,
+      "step": 18
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 13.5287,
+      "step": 19
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.8e-05,
+      "loss": 15.0066,
+      "step": 20
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.9e-05,
+      "loss": 13.8309,
+      "step": 21
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2e-05,
+      "loss": 13.4415,
+      "step": 22
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.1000000000000002e-05,
+      "loss": 12.4292,
+      "step": 23
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.2e-05,
+      "loss": 12.418,
+      "step": 24
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2e-05,
+      "loss": 12.0329,
+      "step": 25
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.3e-05,
+      "loss": 13.3658,
+      "step": 26
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.4e-05,
+      "loss": 12.2996,
+      "step": 27
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5e-05,
+      "loss": 11.3261,
+      "step": 28
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.6e-05,
+      "loss": 10.5315,
+      "step": 29
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7e-05,
+      "loss": 9.8041,
+      "step": 30
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.8e-05,
+      "loss": 9.523,
+      "step": 31
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.9000000000000004e-05,
+      "loss": 10.7327,
+      "step": 32
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3e-05,
+      "loss": 9.1251,
+      "step": 33
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1e-05,
+      "loss": 8.3973,
+      "step": 34
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.2e-05,
+      "loss": 7.8455,
+      "step": 35
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3e-05,
+      "loss": 7.5146,
+      "step": 36
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 8.0064,
+      "step": 37
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 7.3989,
+      "step": 38
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.6e-05,
+      "loss": 7.2097,
+      "step": 39
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.7e-05,
+      "loss": 6.5015,
+      "step": 40
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.8e-05,
+      "loss": 6.3267,
+      "step": 41
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.9e-05,
+      "loss": 5.8663,
+      "step": 42
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4e-05,
+      "loss": 5.5293,
+      "step": 43
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.1e-05,
+      "loss": 6.6383,
+      "step": 44
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.2000000000000004e-05,
+      "loss": 5.8792,
+      "step": 45
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.2999999999999995e-05,
+      "loss": 5.5792,
+      "step": 46
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.4e-05,
+      "loss": 5.4197,
+      "step": 47
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 5.2121,
+      "step": 48
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.6e-05,
+      "loss": 4.9884,
+      "step": 49
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.7000000000000004e-05,
+      "loss": 5.4546,
+      "step": 50
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.8e-05,
+      "loss": 5.0865,
+      "step": 51
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.9000000000000005e-05,
+      "loss": 4.9148,
+      "step": 52
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5e-05,
+      "loss": 4.8803,
+      "step": 53
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.1e-05,
+      "loss": 4.6615,
+      "step": 54
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2e-05,
+      "loss": 4.7224,
+      "step": 55
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.3e-05,
+      "loss": 4.9522,
+      "step": 56
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 5.4e-05,
+      "loss": 4.6776,
+      "step": 57
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 5.5e-05,
+      "loss": 4.5132,
+      "step": 58
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 5.6e-05,
+      "loss": 4.4838,
+      "step": 59
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.7e-05,
+      "loss": 4.4392,
+      "step": 60
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.800000000000001e-05,
+      "loss": 4.3198,
+      "step": 61
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 5.9e-05,
+      "loss": 4.3567,
+      "step": 62
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 6e-05,
+      "loss": 4.539,
+      "step": 63
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 6.1e-05,
+      "loss": 4.3051,
+      "step": 64
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 6.2e-05,
+      "loss": 4.2226,
+      "step": 65
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 6.3e-05,
+      "loss": 4.2058,
+      "step": 66
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 6.4e-05,
+      "loss": 4.0903,
+      "step": 67
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 4.0985,
+      "step": 68
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 6.6e-05,
+      "loss": 4.2325,
+      "step": 69
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 6.7e-05,
+      "loss": 4.1052,
+      "step": 70
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 6.800000000000001e-05,
+      "loss": 3.9697,
+      "step": 71
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.900000000000001e-05,
+      "loss": 3.9088,
+      "step": 72
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 7.000000000000001e-05,
+      "loss": 3.9698,
+      "step": 73
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 7.099999999999999e-05,
+      "loss": 4.0541,
+      "step": 74
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 7.2e-05,
+      "loss": 3.9794,
+      "step": 75
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 7.3e-05,
+      "loss": 3.8948,
+      "step": 76
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 7.4e-05,
+      "loss": 3.8883,
+      "step": 77
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 7.5e-05,
+      "loss": 3.8033,
+      "step": 78
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 7.6e-05,
+      "loss": 3.8101,
+      "step": 79
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 7.7e-05,
+      "loss": 3.7674,
+      "step": 80
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 7.8e-05,
+      "loss": 3.8785,
+      "step": 81
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 7.9e-05,
+      "loss": 3.7836,
+      "step": 82
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 8e-05,
+      "loss": 3.7129,
+      "step": 83
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 8.1e-05,
+      "loss": 3.6691,
+      "step": 84
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 8.2e-05,
+      "loss": 3.6402,
+      "step": 85
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 8.300000000000001e-05,
+      "loss": 3.6414,
+      "step": 86
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 8.400000000000001e-05,
+      "loss": 3.6958,
+      "step": 87
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 8.5e-05,
+      "loss": 3.6505,
+      "step": 88
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 8.599999999999999e-05,
+      "loss": 3.5752,
+      "step": 89
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 8.7e-05,
+      "loss": 3.5321,
+      "step": 90
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 8.8e-05,
+      "loss": 3.5134,
+      "step": 91
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 8.9e-05,
+      "loss": 3.5408,
+      "step": 92
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 3.6869,
+      "step": 93
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 9.1e-05,
+      "loss": 3.5319,
+      "step": 94
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 9.2e-05,
+      "loss": 3.5053,
+      "step": 95
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.3e-05,
+      "loss": 3.4387,
+      "step": 96
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 9.400000000000001e-05,
+      "loss": 3.443,
+      "step": 97
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 9.5e-05,
+      "loss": 3.3938,
+      "step": 98
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 9.6e-05,
+      "loss": 3.4723,
+      "step": 99
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 9.7e-05,
+      "loss": 3.4324,
+      "step": 100
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 9.800000000000001e-05,
+      "loss": 3.3759,
+      "step": 101
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 9.900000000000001e-05,
+      "loss": 3.3655,
+      "step": 102
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0001,
+      "loss": 3.3568,
+      "step": 103
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.000101,
+      "loss": 3.3286,
+      "step": 104
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.000102,
+      "loss": 3.3444,
+      "step": 105
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.000103,
+      "loss": 3.3888,
+      "step": 106
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.000104,
+      "loss": 3.2907,
+      "step": 107
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.000105,
+      "loss": 3.275,
+      "step": 108
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.000106,
+      "loss": 3.2697,
+      "step": 109
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.000107,
+      "loss": 3.2901,
+      "step": 110
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.000108,
+      "loss": 3.4175,
+      "step": 111
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.000109,
+      "loss": 3.2505,
+      "step": 112
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00011,
+      "loss": 3.2408,
+      "step": 113
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.000111,
+      "loss": 3.2383,
+      "step": 114
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.000112,
+      "loss": 3.2229,
+      "step": 115
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00011300000000000001,
+      "loss": 3.216,
+      "step": 116
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.000114,
+      "loss": 3.2522,
+      "step": 117
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.000115,
+      "loss": 3.2635,
+      "step": 118
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00011600000000000001,
+      "loss": 3.2141,
+      "step": 119
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00011700000000000001,
+      "loss": 3.1859,
+      "step": 120
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.000118,
+      "loss": 3.2102,
+      "step": 121
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.00011899999999999999,
+      "loss": 3.1732,
+      "step": 122
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00012,
+      "loss": 3.2272,
+      "step": 123
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.000121,
+      "loss": 3.2217,
+      "step": 124
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.000122,
+      "loss": 3.1856,
+      "step": 125
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.000123,
+      "loss": 3.1681,
+      "step": 126
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.000124,
+      "loss": 3.1774,
+      "step": 127
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.000125,
+      "loss": 3.1537,
+      "step": 128
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.000126,
+      "loss": 3.1598,
+      "step": 129
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.000127,
+      "loss": 3.3046,
+      "step": 130
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.000128,
+      "loss": 3.1543,
+      "step": 131
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.00012900000000000002,
+      "loss": 3.141,
+      "step": 132
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 3.141,
+      "step": 133
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.000131,
+      "loss": 3.1555,
+      "step": 134
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.000132,
+      "loss": 3.1532,
+      "step": 135
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.000133,
+      "loss": 3.2472,
+      "step": 136
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.000134,
+      "loss": 3.1378,
+      "step": 137
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.000135,
+      "loss": 3.1197,
+      "step": 138
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00013600000000000003,
+      "loss": 3.1166,
+      "step": 139
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.00013700000000000002,
+      "loss": 3.1301,
+      "step": 140
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00013800000000000002,
+      "loss": 3.136,
+      "step": 141
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.00013900000000000002,
+      "loss": 3.1341,
+      "step": 142
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 3.1548,
+      "step": 143
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 3.0931,
+      "step": 144
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.00014199999999999998,
+      "loss": 3.1108,
+      "step": 145
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.00014299999999999998,
+      "loss": 3.1194,
+      "step": 146
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.000144,
+      "loss": 3.1408,
+      "step": 147
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.000145,
+      "loss": 3.2052,
+      "step": 148
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.000146,
+      "loss": 3.0969,
+      "step": 149
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.000147,
+      "loss": 3.0824,
+      "step": 150
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.000148,
+      "loss": 3.0856,
+      "step": 151
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.000149,
+      "loss": 3.082,
+      "step": 152
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.00015,
+      "loss": 3.0866,
+      "step": 153
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.000151,
+      "loss": 3.1255,
+      "step": 154
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.000152,
+      "loss": 3.1303,
+      "step": 155
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.000153,
+      "loss": 3.0832,
+      "step": 156
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.000154,
+      "loss": 3.0556,
+      "step": 157
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.000155,
+      "loss": 3.09,
+      "step": 158
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.000156,
+      "loss": 3.0485,
+      "step": 159
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.000157,
+      "loss": 3.073,
+      "step": 160
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.000158,
+      "loss": 3.0777,
+      "step": 161
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00015900000000000002,
+      "loss": 3.0702,
+      "step": 162
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.00016,
+      "loss": 3.063,
+      "step": 163
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.000161,
+      "loss": 3.0981,
+      "step": 164
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.000162,
+      "loss": 3.0606,
+      "step": 165
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.000163,
+      "loss": 3.0546,
+      "step": 166
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.000164,
+      "loss": 3.1347,
+      "step": 167
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.000165,
+      "loss": 3.0804,
+      "step": 168
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.00016600000000000002,
+      "loss": 3.0687,
+      "step": 169
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00016700000000000002,
+      "loss": 3.0529,
+      "step": 170
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.00016800000000000002,
+      "loss": 3.0602,
+      "step": 171
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.00016900000000000002,
+      "loss": 3.0525,
+      "step": 172
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00017,
+      "loss": 3.0561,
+      "step": 173
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.000171,
+      "loss": 3.043,
+      "step": 174
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00017199999999999998,
+      "loss": 3.0517,
+      "step": 175
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.000173,
+      "loss": 3.0373,
+      "step": 176
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.000174,
+      "loss": 3.0465,
+      "step": 177
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.000175,
+      "loss": 3.0227,
+      "step": 178
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.000176,
+      "loss": 3.0355,
+      "step": 179
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.000177,
+      "loss": 3.0533,
+      "step": 180
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.000178,
+      "loss": 3.0461,
+      "step": 181
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.000179,
+      "loss": 3.0513,
+      "step": 182
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 3.0374,
+      "step": 183
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.000181,
+      "loss": 3.0652,
+      "step": 184
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.000182,
+      "loss": 3.0148,
+      "step": 185
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.000183,
+      "loss": 3.0489,
+      "step": 186
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.000184,
+      "loss": 3.0469,
+      "step": 187
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.000185,
+      "loss": 3.017,
+      "step": 188
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.000186,
+      "loss": 3.065,
+      "step": 189
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.000187,
+      "loss": 3.0279,
+      "step": 190
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.00018800000000000002,
+      "loss": 3.0205,
+      "step": 191
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.000189,
+      "loss": 3.0336,
+      "step": 192
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.00019,
+      "loss": 3.0377,
+      "step": 193
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.000191,
+      "loss": 3.015,
+      "step": 194
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.000192,
+      "loss": 3.0273,
+      "step": 195
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.000193,
+      "loss": 2.9981,
+      "step": 196
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.000194,
+      "loss": 3.0142,
+      "step": 197
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.00019500000000000002,
+      "loss": 3.0201,
+      "step": 198
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00019600000000000002,
+      "loss": 3.0186,
+      "step": 199
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00019700000000000002,
+      "loss": 2.9975,
+      "step": 200
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00019800000000000002,
+      "loss": 3.0249,
+      "step": 201
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.000199,
+      "loss": 3.0019,
+      "step": 202
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.0002,
+      "loss": 2.9881,
+      "step": 203
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.000201,
+      "loss": 3.0163,
+      "step": 204
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.000202,
+      "loss": 2.9916,
+      "step": 205
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00020300000000000003,
+      "loss": 2.9989,
+      "step": 206
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.000204,
+      "loss": 2.9941,
+      "step": 207
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.000205,
+      "loss": 2.9812,
+      "step": 208
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.000206,
+      "loss": 2.9564,
+      "step": 209
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.000207,
+      "loss": 2.9762,
+      "step": 210
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.000208,
+      "loss": 2.9835,
+      "step": 211
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00020899999999999998,
+      "loss": 3.0011,
+      "step": 212
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00021,
+      "loss": 2.9919,
+      "step": 213
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.000211,
+      "loss": 3.0058,
+      "step": 214
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.000212,
+      "loss": 3.012,
+      "step": 215
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.000213,
+      "loss": 2.9486,
+      "step": 216
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.000214,
+      "loss": 2.9711,
+      "step": 217
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.000215,
+      "loss": 2.995,
+      "step": 218
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.000216,
+      "loss": 2.9519,
+      "step": 219
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.00021700000000000002,
+      "loss": 2.943,
+      "step": 220
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.000218,
+      "loss": 2.9152,
+      "step": 221
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.000219,
+      "loss": 3.0891,
+      "step": 222
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.00022,
+      "loss": 2.9716,
+      "step": 223
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.000221,
+      "loss": 2.9667,
+      "step": 224
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.000222,
+      "loss": 2.9128,
+      "step": 225
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.000223,
+      "loss": 2.9892,
+      "step": 226
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.000224,
+      "loss": 2.8884,
+      "step": 227
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 2.857,
+      "step": 228
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00022600000000000002,
+      "loss": 2.9918,
+      "step": 229
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00022700000000000002,
+      "loss": 2.8965,
+      "step": 230
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.000228,
+      "loss": 2.9051,
+      "step": 231
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.000229,
+      "loss": 2.909,
+      "step": 232
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00023,
+      "loss": 2.9174,
+      "step": 233
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.000231,
+      "loss": 2.8701,
+      "step": 234
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.00023200000000000003,
+      "loss": 2.886,
+      "step": 235
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00023300000000000003,
+      "loss": 2.8591,
+      "step": 236
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00023400000000000002,
+      "loss": 2.8473,
+      "step": 237
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.000235,
+      "loss": 2.8596,
+      "step": 238
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.000236,
+      "loss": 2.7874,
+      "step": 239
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.000237,
+      "loss": 2.7928,
+      "step": 240
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.00023799999999999998,
+      "loss": 2.7892,
+      "step": 241
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00023899999999999998,
+      "loss": 2.7667,
+      "step": 242
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00024,
+      "loss": 2.7637,
+      "step": 243
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.000241,
+      "loss": 2.727,
+      "step": 244
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.000242,
+      "loss": 2.719,
+      "step": 245
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.000243,
+      "loss": 2.6695,
+      "step": 246
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.000244,
+      "loss": 2.6643,
+      "step": 247
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.000245,
+      "loss": 2.6983,
+      "step": 248
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.000246,
+      "loss": 2.6443,
+      "step": 249
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.000247,
+      "loss": 2.5917,
+      "step": 250
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.000248,
+      "loss": 2.5677,
+      "step": 251
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.000249,
+      "loss": 2.5834,
+      "step": 252
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.00025,
+      "loss": 2.5845,
+      "step": 253
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00025100000000000003,
+      "loss": 2.5713,
+      "step": 254
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.000252,
+      "loss": 2.595,
+      "step": 255
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 0.000253,
+      "loss": 2.4907,
+      "step": 256
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.000254,
+      "loss": 2.5373,
+      "step": 257
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.000255,
+      "loss": 2.5167,
+      "step": 258
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.000256,
+      "loss": 2.4347,
+      "step": 259
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.000257,
+      "loss": 2.3908,
+      "step": 260
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.00025800000000000004,
+      "loss": 2.3528,
+      "step": 261
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.000259,
+      "loss": 2.3008,
+      "step": 262
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 2.2657,
+      "step": 263
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.000261,
+      "loss": 2.2046,
+      "step": 264
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.000262,
+      "loss": 2.2167,
+      "step": 265
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.000263,
+      "loss": 2.1376,
+      "step": 266
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.000264,
+      "loss": 2.1094,
+      "step": 267
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00026500000000000004,
+      "loss": 2.118,
+      "step": 268
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.000266,
+      "loss": 2.0502,
+      "step": 269
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00026700000000000004,
+      "loss": 1.9736,
+      "step": 270
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.000268,
+      "loss": 1.9367,
+      "step": 271
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00026900000000000003,
+      "loss": 1.9111,
+      "step": 272
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00027,
+      "loss": 1.8512,
+      "step": 273
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00027100000000000003,
+      "loss": 1.8298,
+      "step": 274
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.00027200000000000005,
+      "loss": 1.8157,
+      "step": 275
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.000273,
+      "loss": 1.7201,
+      "step": 276
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.00027400000000000005,
+      "loss": 1.6966,
+      "step": 277
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.000275,
+      "loss": 1.7841,
+      "step": 278
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00027600000000000004,
+      "loss": 1.6715,
+      "step": 279
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.000277,
+      "loss": 1.6931,
+      "step": 280
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00027800000000000004,
+      "loss": 1.5213,
+      "step": 281
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 0.000279,
+      "loss": 1.5803,
+      "step": 282
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 1.5692,
+      "step": 283
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00028100000000000005,
+      "loss": 1.5354,
+      "step": 284
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 1.5139,
+      "step": 285
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.000283,
+      "loss": 1.3925,
+      "step": 286
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00028399999999999996,
+      "loss": 1.4602,
+      "step": 287
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.000285,
+      "loss": 1.4092,
+      "step": 288
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00028599999999999996,
+      "loss": 1.4256,
+      "step": 289
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.000287,
+      "loss": 1.361,
+      "step": 290
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.000288,
+      "loss": 1.3903,
+      "step": 291
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.000289,
+      "loss": 1.3391,
+      "step": 292
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.00029,
+      "loss": 1.282,
+      "step": 293
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 1.3109,
+      "step": 294
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.000292,
+      "loss": 1.2758,
+      "step": 295
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00029299999999999997,
+      "loss": 1.292,
+      "step": 296
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.000294,
+      "loss": 1.2738,
+      "step": 297
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.000295,
+      "loss": 1.2446,
+      "step": 298
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.000296,
+      "loss": 1.2255,
+      "step": 299
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.000297,
+      "loss": 1.2218,
+      "step": 300
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.000298,
+      "loss": 1.1992,
+      "step": 301
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.000299,
+      "loss": 1.1966,
+      "step": 302
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.0003,
+      "loss": 1.1669,
+      "step": 303
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.000301,
+      "loss": 1.1022,
+      "step": 304
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.000302,
+      "loss": 1.0532,
+      "step": 305
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.000303,
+      "loss": 1.1824,
+      "step": 306
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.000304,
+      "loss": 1.0715,
+      "step": 307
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.000305,
+      "loss": 1.0672,
+      "step": 308
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.000306,
+      "loss": 1.1031,
+      "step": 309
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.000307,
+      "loss": 1.0676,
+      "step": 310
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.000308,
+      "loss": 1.0736,
+      "step": 311
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00030900000000000003,
+      "loss": 1.0582,
+      "step": 312
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00031,
+      "loss": 1.0729,
+      "step": 313
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.000311,
+      "loss": 1.022,
+      "step": 314
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.000312,
+      "loss": 1.0355,
+      "step": 315
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 0.000313,
+      "loss": 1.0295,
+      "step": 316
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.000314,
+      "loss": 1.0201,
+      "step": 317
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.000315,
+      "loss": 0.9915,
+      "step": 318
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.000316,
+      "loss": 1.0132,
+      "step": 319
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.000317,
+      "loss": 1.0596,
+      "step": 320
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.00031800000000000003,
+      "loss": 1.0613,
+      "step": 321
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.000319,
+      "loss": 1.0086,
+      "step": 322
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00032,
+      "loss": 0.9607,
+      "step": 323
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.000321,
+      "loss": 0.9375,
+      "step": 324
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.000322,
+      "loss": 0.9815,
+      "step": 325
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.000323,
+      "loss": 0.9932,
+      "step": 326
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.000324,
+      "loss": 1.0713,
+      "step": 327
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 1.0306,
+      "step": 328
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.000326,
+      "loss": 0.9587,
+      "step": 329
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.00032700000000000003,
+      "loss": 0.9478,
+      "step": 330
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.000328,
+      "loss": 0.9995,
+      "step": 331
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00032900000000000003,
+      "loss": 1.0039,
+      "step": 332
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 0.00033,
+      "loss": 1.0722,
+      "step": 333
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.000331,
+      "loss": 0.9033,
+      "step": 334
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00033200000000000005,
+      "loss": 0.8853,
+      "step": 335
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.000333,
+      "loss": 0.89,
+      "step": 336
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00033400000000000004,
+      "loss": 0.8817,
+      "step": 337
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.000335,
+      "loss": 0.8728,
+      "step": 338
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.00033600000000000004,
+      "loss": 0.8701,
+      "step": 339
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 0.000337,
+      "loss": 0.887,
+      "step": 340
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00033800000000000003,
+      "loss": 0.8735,
+      "step": 341
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00033900000000000005,
+      "loss": 0.8405,
+      "step": 342
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00034,
+      "loss": 0.9496,
+      "step": 343
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00034100000000000005,
+      "loss": 0.834,
+      "step": 344
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.000342,
+      "loss": 0.7885,
+      "step": 345
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00034300000000000004,
+      "loss": 0.7772,
+      "step": 346
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00034399999999999996,
+      "loss": 0.82,
+      "step": 347
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.000345,
+      "loss": 0.8291,
+      "step": 348
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.000346,
+      "loss": 0.7856,
+      "step": 349
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.000347,
+      "loss": 0.8347,
+      "step": 350
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.000348,
+      "loss": 0.8251,
+      "step": 351
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00034899999999999997,
+      "loss": 0.8366,
+      "step": 352
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00035,
+      "loss": 0.917,
+      "step": 353
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00035099999999999997,
+      "loss": 0.7599,
+      "step": 354
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.000352,
+      "loss": 0.8401,
+      "step": 355
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00035299999999999996,
+      "loss": 0.8041,
+      "step": 356
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 0.000354,
+      "loss": 0.7712,
+      "step": 357
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.000355,
+      "loss": 0.8697,
+      "step": 358
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.000356,
+      "loss": 0.828,
+      "step": 359
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.000357,
+      "loss": 0.7671,
+      "step": 360
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.000358,
+      "loss": 0.8003,
+      "step": 361
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.000359,
+      "loss": 0.7936,
+      "step": 362
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 0.8536,
+      "step": 363
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.000361,
+      "loss": 0.813,
+      "step": 364
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.000362,
+      "loss": 0.8407,
+      "step": 365
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.000363,
+      "loss": 0.7772,
+      "step": 366
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.000364,
+      "loss": 0.7694,
+      "step": 367
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.000365,
+      "loss": 0.7412,
+      "step": 368
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.000366,
+      "loss": 0.7526,
+      "step": 369
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.000367,
+      "loss": 0.8785,
+      "step": 370
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.000368,
+      "loss": 0.7873,
+      "step": 371
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.000369,
+      "loss": 0.7219,
+      "step": 372
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00037,
+      "loss": 0.7949,
+      "step": 373
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.000371,
+      "loss": 0.6976,
+      "step": 374
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.000372,
+      "loss": 0.7232,
+      "step": 375
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.000373,
+      "loss": 0.7772,
+      "step": 376
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.000374,
+      "loss": 0.6925,
+      "step": 377
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.000375,
+      "loss": 0.7404,
+      "step": 378
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.00037600000000000003,
+      "loss": 0.6883,
+      "step": 379
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.000377,
+      "loss": 0.6623,
+      "step": 380
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.000378,
+      "loss": 0.6327,
+      "step": 381
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.000379,
+      "loss": 0.7286,
+      "step": 382
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 0.00038,
+      "loss": 0.7415,
+      "step": 383
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.000381,
+      "loss": 0.6802,
+      "step": 384
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.000382,
+      "loss": 0.7032,
+      "step": 385
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00038300000000000004,
+      "loss": 0.6416,
+      "step": 386
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.000384,
+      "loss": 0.6953,
+      "step": 387
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00038500000000000003,
+      "loss": 0.7168,
+      "step": 388
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.000386,
+      "loss": 0.7073,
+      "step": 389
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.00038700000000000003,
+      "loss": 0.7204,
+      "step": 390
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.000388,
+      "loss": 0.6645,
+      "step": 391
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.000389,
+      "loss": 0.7079,
+      "step": 392
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 0.6952,
+      "step": 393
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.000391,
+      "loss": 0.7412,
+      "step": 394
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00039200000000000004,
+      "loss": 0.7123,
+      "step": 395
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.000393,
+      "loss": 0.707,
+      "step": 396
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00039400000000000004,
+      "loss": 0.6581,
+      "step": 397
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.000395,
+      "loss": 0.6476,
+      "step": 398
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00039600000000000003,
+      "loss": 0.6017,
+      "step": 399
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 0.00039700000000000005,
+      "loss": 0.6368,
+      "step": 400
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.000398,
+      "loss": 0.653,
+      "step": 401
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.00039900000000000005,
+      "loss": 0.6691,
+      "step": 402
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.0004,
+      "loss": 0.6485,
+      "step": 403
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00040100000000000004,
+      "loss": 0.6911,
+      "step": 404
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.000402,
+      "loss": 0.6976,
+      "step": 405
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00040300000000000004,
+      "loss": 0.6353,
+      "step": 406
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 0.000404,
+      "loss": 0.7716,
+      "step": 407
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00040500000000000003,
+      "loss": 0.6539,
+      "step": 408
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00040600000000000006,
+      "loss": 0.5896,
+      "step": 409
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00040699999999999997,
+      "loss": 0.5657,
+      "step": 410
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.000408,
+      "loss": 0.6307,
+      "step": 411
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00040899999999999997,
+      "loss": 0.5966,
+      "step": 412
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00041,
+      "loss": 0.5523,
+      "step": 413
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00041099999999999996,
+      "loss": 0.6163,
+      "step": 414
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.000412,
+      "loss": 0.5707,
+      "step": 415
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.000413,
+      "loss": 0.5876,
+      "step": 416
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 0.000414,
+      "loss": 0.6326,
+      "step": 417
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.000415,
+      "loss": 0.619,
+      "step": 418
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.000416,
+      "loss": 0.5776,
+      "step": 419
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.000417,
+      "loss": 0.541,
+      "step": 420
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00041799999999999997,
+      "loss": 0.6606,
+      "step": 421
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.000419,
+      "loss": 0.591,
+      "step": 422
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00042,
+      "loss": 0.6034,
+      "step": 423
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 0.000421,
+      "loss": 0.5981,
+      "step": 424
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.000422,
+      "loss": 0.5892,
+      "step": 425
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.000423,
+      "loss": 0.6287,
+      "step": 426
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.000424,
+      "loss": 0.6319,
+      "step": 427
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.000425,
+      "loss": 0.6303,
+      "step": 428
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.000426,
+      "loss": 0.6006,
+      "step": 429
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.000427,
+      "loss": 0.602,
+      "step": 430
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.000428,
+      "loss": 0.61,
+      "step": 431
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.000429,
+      "loss": 0.668,
+      "step": 432
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00043,
+      "loss": 0.5766,
+      "step": 433
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.000431,
+      "loss": 0.6228,
+      "step": 434
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.000432,
+      "loss": 0.5953,
+      "step": 435
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.000433,
+      "loss": 0.5916,
+      "step": 436
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00043400000000000003,
+      "loss": 0.5253,
+      "step": 437
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.000435,
+      "loss": 0.6327,
+      "step": 438
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.000436,
+      "loss": 0.6082,
+      "step": 439
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.000437,
+      "loss": 0.6057,
+      "step": 440
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 0.000438,
+      "loss": 0.5861,
+      "step": 441
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.000439,
+      "loss": 0.6716,
+      "step": 442
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00044,
+      "loss": 0.6413,
+      "step": 443
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.000441,
+      "loss": 0.5029,
+      "step": 444
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.000442,
+      "loss": 0.6002,
+      "step": 445
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00044300000000000003,
+      "loss": 0.5832,
+      "step": 446
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.000444,
+      "loss": 0.6211,
+      "step": 447
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.00044500000000000003,
+      "loss": 0.5511,
+      "step": 448
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.000446,
+      "loss": 0.5532,
+      "step": 449
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.000447,
+      "loss": 0.6131,
+      "step": 450
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.000448,
+      "loss": 0.5755,
+      "step": 451
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.000449,
+      "loss": 0.5313,
+      "step": 452
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 0.6135,
+      "step": 453
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.000451,
+      "loss": 0.6025,
+      "step": 454
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00045200000000000004,
+      "loss": 0.6089,
+      "step": 455
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.000453,
+      "loss": 0.6059,
+      "step": 456
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00045400000000000003,
+      "loss": 0.5757,
+      "step": 457
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.000455,
+      "loss": 0.5453,
+      "step": 458
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.000456,
+      "loss": 0.5203,
+      "step": 459
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.00045700000000000005,
+      "loss": 0.5415,
+      "step": 460
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.000458,
+      "loss": 0.6215,
+      "step": 461
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00045900000000000004,
+      "loss": 0.5171,
+      "step": 462
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00046,
+      "loss": 0.6054,
+      "step": 463
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00046100000000000004,
+      "loss": 0.5604,
+      "step": 464
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.000462,
+      "loss": 0.5532,
+      "step": 465
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00046300000000000003,
+      "loss": 0.6361,
+      "step": 466
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 0.00046400000000000006,
+      "loss": 0.5939,
+      "step": 467
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.000465,
+      "loss": 0.5972,
+      "step": 468
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00046600000000000005,
+      "loss": 0.6193,
+      "step": 469
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.000467,
+      "loss": 0.5929,
+      "step": 470
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00046800000000000005,
+      "loss": 0.5685,
+      "step": 471
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.00046899999999999996,
+      "loss": 0.6227,
+      "step": 472
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00047,
+      "loss": 0.6359,
+      "step": 473
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.000471,
+      "loss": 0.5884,
+      "step": 474
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.000472,
+      "loss": 0.5663,
+      "step": 475
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.000473,
+      "loss": 0.5656,
+      "step": 476
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.000474,
+      "loss": 0.5575,
+      "step": 477
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.000475,
+      "loss": 0.5337,
+      "step": 478
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00047599999999999997,
+      "loss": 0.6036,
+      "step": 479
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.000477,
+      "loss": 0.5857,
+      "step": 480
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00047799999999999996,
+      "loss": 0.5805,
+      "step": 481
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.000479,
+      "loss": 0.5903,
+      "step": 482
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00048,
+      "loss": 0.579,
+      "step": 483
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 0.000481,
+      "loss": 0.5388,
+      "step": 484
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.000482,
+      "loss": 0.5776,
+      "step": 485
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.000483,
+      "loss": 0.5283,
+      "step": 486
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.000484,
+      "loss": 0.5263,
+      "step": 487
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00048499999999999997,
+      "loss": 0.5491,
+      "step": 488
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.000486,
+      "loss": 0.5532,
+      "step": 489
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.000487,
+      "loss": 0.5478,
+      "step": 490
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.000488,
+      "loss": 0.5684,
+      "step": 491
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.000489,
+      "loss": 0.5944,
+      "step": 492
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00049,
+      "loss": 0.5145,
+      "step": 493
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.000491,
+      "loss": 0.5446,
+      "step": 494
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.000492,
+      "loss": 0.5772,
+      "step": 495
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.0004930000000000001,
+      "loss": 0.4986,
+      "step": 496
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.000494,
+      "loss": 0.5385,
+      "step": 497
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.000495,
+      "loss": 0.5794,
+      "step": 498
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.000496,
+      "loss": 0.5338,
+      "step": 499
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.000497,
+      "loss": 0.5462,
+      "step": 500
+    },
+    {
+      "epoch": 13.51,
+      "eval_cer": 0.11878738811676337,
+      "eval_loss": 0.4422529637813568,
+      "eval_runtime": 11.2014,
+      "eval_samples_per_second": 147.035,
+      "eval_steps_per_second": 2.321,
+      "eval_wer": 0.48074762537023796,
+      "step": 500
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 0.000498,
+      "loss": 0.5576,
+      "step": 501
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 0.000499,
+      "loss": 0.5489,
+      "step": 502
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.0005,
+      "loss": 0.5573,
+      "step": 503
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00049984375,
+      "loss": 0.5769,
+      "step": 504
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.0004996875,
+      "loss": 0.5356,
+      "step": 505
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 0.00049953125,
+      "loss": 0.558,
+      "step": 506
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 0.000499375,
+      "loss": 0.586,
+      "step": 507
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 0.00049921875,
+      "loss": 0.5821,
+      "step": 508
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 0.0004990625,
+      "loss": 0.4896,
+      "step": 509
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 0.00049890625,
+      "loss": 0.5897,
+      "step": 510
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 0.0004987500000000001,
+      "loss": 0.5668,
+      "step": 511
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 0.00049859375,
+      "loss": 0.5911,
+      "step": 512
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 0.0004984375,
+      "loss": 0.5914,
+      "step": 513
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 0.00049828125,
+      "loss": 0.5261,
+      "step": 514
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 0.000498125,
+      "loss": 0.5169,
+      "step": 515
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 0.00049796875,
+      "loss": 0.5666,
+      "step": 516
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 0.0004978125,
+      "loss": 0.5451,
+      "step": 517
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 0.00049765625,
+      "loss": 0.4495,
+      "step": 518
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 0.0004975,
+      "loss": 0.5501,
+      "step": 519
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 0.00049734375,
+      "loss": 0.4783,
+      "step": 520
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 0.0004971875,
+      "loss": 0.5412,
+      "step": 521
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 0.00049703125,
+      "loss": 0.4519,
+      "step": 522
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 0.000496875,
+      "loss": 0.5525,
+      "step": 523
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 0.00049671875,
+      "loss": 0.5191,
+      "step": 524
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 0.0004965625,
+      "loss": 0.4767,
+      "step": 525
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 0.00049640625,
+      "loss": 0.471,
+      "step": 526
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 0.0004962500000000001,
+      "loss": 0.4541,
+      "step": 527
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 0.00049609375,
+      "loss": 0.5077,
+      "step": 528
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 0.0004959375,
+      "loss": 0.5107,
+      "step": 529
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 0.00049578125,
+      "loss": 0.4686,
+      "step": 530
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 0.000495625,
+      "loss": 0.568,
+      "step": 531
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 0.00049546875,
+      "loss": 0.5221,
+      "step": 532
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 0.0004953125,
+      "loss": 0.4672,
+      "step": 533
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 0.00049515625,
+      "loss": 0.5154,
+      "step": 534
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 0.000495,
+      "loss": 0.4685,
+      "step": 535
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 0.0004948437500000001,
+      "loss": 0.4375,
+      "step": 536
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 0.0004946875,
+      "loss": 0.4489,
+      "step": 537
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 0.00049453125,
+      "loss": 0.5482,
+      "step": 538
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 0.000494375,
+      "loss": 0.4504,
+      "step": 539
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 0.00049421875,
+      "loss": 0.444,
+      "step": 540
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 0.0004940625,
+      "loss": 0.4491,
+      "step": 541
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 0.00049390625,
+      "loss": 0.4409,
+      "step": 542
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 0.00049375,
+      "loss": 0.4856,
+      "step": 543
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 0.00049359375,
+      "loss": 0.5121,
+      "step": 544
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 0.0004934375,
+      "loss": 0.4936,
+      "step": 545
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 0.00049328125,
+      "loss": 0.4599,
+      "step": 546
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 0.000493125,
+      "loss": 0.4495,
+      "step": 547
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 0.00049296875,
+      "loss": 0.5432,
+      "step": 548
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 0.0004928125,
+      "loss": 0.4812,
+      "step": 549
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 0.00049265625,
+      "loss": 0.4801,
+      "step": 550
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 0.0004925,
+      "loss": 0.4536,
+      "step": 551
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 0.00049234375,
+      "loss": 0.4686,
+      "step": 552
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 0.0004921875,
+      "loss": 0.4971,
+      "step": 553
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 0.00049203125,
+      "loss": 0.4328,
+      "step": 554
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 0.000491875,
+      "loss": 0.461,
+      "step": 555
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 0.00049171875,
+      "loss": 0.481,
+      "step": 556
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 0.0004915625,
+      "loss": 0.427,
+      "step": 557
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 0.00049140625,
+      "loss": 0.3956,
+      "step": 558
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 0.00049125,
+      "loss": 0.4571,
+      "step": 559
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 0.00049109375,
+      "loss": 0.4666,
+      "step": 560
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 0.0004909375000000001,
+      "loss": 0.4915,
+      "step": 561
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 0.00049078125,
+      "loss": 0.4677,
+      "step": 562
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 0.000490625,
+      "loss": 0.4502,
+      "step": 563
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 0.00049046875,
+      "loss": 0.4586,
+      "step": 564
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 0.0004903125,
+      "loss": 0.4459,
+      "step": 565
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 0.00049015625,
+      "loss": 0.4871,
+      "step": 566
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 0.00049,
+      "loss": 0.4243,
+      "step": 567
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 0.00048984375,
+      "loss": 0.3963,
+      "step": 568
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 0.0004896875,
+      "loss": 0.4528,
+      "step": 569
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 0.00048953125,
+      "loss": 0.4822,
+      "step": 570
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 0.000489375,
+      "loss": 0.4259,
+      "step": 571
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 0.00048921875,
+      "loss": 0.484,
+      "step": 572
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 0.0004890625,
+      "loss": 0.4906,
+      "step": 573
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 0.00048890625,
+      "loss": 0.5433,
+      "step": 574
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 0.00048875,
+      "loss": 0.4747,
+      "step": 575
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 0.00048859375,
+      "loss": 0.4609,
+      "step": 576
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 0.0004884375,
+      "loss": 0.5152,
+      "step": 577
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 0.00048828125,
+      "loss": 0.4443,
+      "step": 578
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 0.000488125,
+      "loss": 0.4826,
+      "step": 579
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 0.00048796875,
+      "loss": 0.4851,
+      "step": 580
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 0.0004878125,
+      "loss": 0.4914,
+      "step": 581
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 0.00048765625,
+      "loss": 0.4764,
+      "step": 582
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 0.0004875,
+      "loss": 0.466,
+      "step": 583
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 0.00048734375000000004,
+      "loss": 0.4444,
+      "step": 584
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 0.0004871875,
+      "loss": 0.4954,
+      "step": 585
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 0.00048703125000000005,
+      "loss": 0.4896,
+      "step": 586
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.000486875,
+      "loss": 0.4577,
+      "step": 587
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 0.00048671875,
+      "loss": 0.4527,
+      "step": 588
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 0.0004865625,
+      "loss": 0.432,
+      "step": 589
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 0.00048640624999999997,
+      "loss": 0.4598,
+      "step": 590
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 0.00048625000000000003,
+      "loss": 0.5165,
+      "step": 591
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 0.00048609375,
+      "loss": 0.4573,
+      "step": 592
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 0.00048593750000000004,
+      "loss": 0.467,
+      "step": 593
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 0.00048578125,
+      "loss": 0.4421,
+      "step": 594
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 0.000485625,
+      "loss": 0.4424,
+      "step": 595
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 0.00048546875,
+      "loss": 0.4931,
+      "step": 596
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 0.0004853125,
+      "loss": 0.4909,
+      "step": 597
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 0.00048515625,
+      "loss": 0.4491,
+      "step": 598
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 0.00048499999999999997,
+      "loss": 0.4332,
+      "step": 599
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 0.00048484375000000003,
+      "loss": 0.4759,
+      "step": 600
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 0.0004846875,
+      "loss": 0.3903,
+      "step": 601
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 0.00048453125000000004,
+      "loss": 0.4066,
+      "step": 602
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 0.000484375,
+      "loss": 0.4021,
+      "step": 603
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 0.00048421875,
+      "loss": 0.3495,
+      "step": 604
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 0.0004840625,
+      "loss": 0.4126,
+      "step": 605
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 0.00048390625,
+      "loss": 0.4195,
+      "step": 606
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 0.00048375,
+      "loss": 0.405,
+      "step": 607
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 0.00048359375,
+      "loss": 0.4001,
+      "step": 608
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 0.00048343750000000003,
+      "loss": 0.445,
+      "step": 609
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 0.00048328125,
+      "loss": 0.4208,
+      "step": 610
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 0.00048312500000000005,
+      "loss": 0.4482,
+      "step": 611
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 0.00048296875,
+      "loss": 0.4277,
+      "step": 612
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 0.0004828125,
+      "loss": 0.4451,
+      "step": 613
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 0.00048265625,
+      "loss": 0.3962,
+      "step": 614
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 0.0004825,
+      "loss": 0.3939,
+      "step": 615
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 0.00048234375,
+      "loss": 0.4085,
+      "step": 616
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 0.0004821875,
+      "loss": 0.4514,
+      "step": 617
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 0.00048203125000000004,
+      "loss": 0.4463,
+      "step": 618
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 0.000481875,
+      "loss": 0.4361,
+      "step": 619
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 0.00048171875,
+      "loss": 0.475,
+      "step": 620
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 0.0004815625,
+      "loss": 0.4494,
+      "step": 621
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 0.00048140625,
+      "loss": 0.4475,
+      "step": 622
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 0.00048125,
+      "loss": 0.3955,
+      "step": 623
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 0.00048109375,
+      "loss": 0.4256,
+      "step": 624
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 0.00048093750000000003,
+      "loss": 0.4003,
+      "step": 625
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 0.00048078125,
+      "loss": 0.3701,
+      "step": 626
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 0.00048062500000000004,
+      "loss": 0.3917,
+      "step": 627
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 0.00048046875,
+      "loss": 0.4247,
+      "step": 628
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 0.0004803125,
+      "loss": 0.247,
+      "step": 629
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 0.00048015625,
+      "loss": 0.4448,
+      "step": 630
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 0.00048,
+      "loss": 0.4094,
+      "step": 631
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 0.00047984375,
+      "loss": 0.4156,
+      "step": 632
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 0.0004796875,
+      "loss": 0.3851,
+      "step": 633
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 0.00047953125000000003,
+      "loss": 0.421,
+      "step": 634
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 0.000479375,
+      "loss": 0.378,
+      "step": 635
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 0.00047921875000000004,
+      "loss": 0.4467,
+      "step": 636
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 0.0004790625,
+      "loss": 0.4361,
+      "step": 637
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 0.00047890625,
+      "loss": 0.365,
+      "step": 638
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 0.00047875,
+      "loss": 0.3968,
+      "step": 639
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 0.00047859375,
+      "loss": 0.4067,
+      "step": 640
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 0.0004784375,
+      "loss": 0.4095,
+      "step": 641
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 0.00047828125000000003,
+      "loss": 0.3985,
+      "step": 642
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 0.00047812500000000003,
+      "loss": 0.4526,
+      "step": 643
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 0.00047796875,
+      "loss": 0.4126,
+      "step": 644
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 0.0004778125,
+      "loss": 0.4185,
+      "step": 645
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 0.00047765625,
+      "loss": 0.4179,
+      "step": 646
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 0.0004775,
+      "loss": 0.398,
+      "step": 647
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 0.00047734375,
+      "loss": 0.3764,
+      "step": 648
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 0.0004771875,
+      "loss": 0.4822,
+      "step": 649
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 0.00047703125,
+      "loss": 0.4272,
+      "step": 650
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 0.000476875,
+      "loss": 0.4143,
+      "step": 651
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 0.00047671875000000004,
+      "loss": 0.427,
+      "step": 652
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 0.0004765625,
+      "loss": 0.3829,
+      "step": 653
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 0.00047640625,
+      "loss": 0.3733,
+      "step": 654
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 0.00047625,
+      "loss": 0.4096,
+      "step": 655
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 0.00047609375,
+      "loss": 0.3885,
+      "step": 656
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 0.0004759375,
+      "loss": 0.4104,
+      "step": 657
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 0.00047578125,
+      "loss": 0.3337,
+      "step": 658
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 0.00047562500000000003,
+      "loss": 0.3659,
+      "step": 659
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 0.00047546875,
+      "loss": 0.443,
+      "step": 660
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 0.00047531250000000004,
+      "loss": 0.4227,
+      "step": 661
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 0.00047515625,
+      "loss": 0.3942,
+      "step": 662
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 0.000475,
+      "loss": 0.4098,
+      "step": 663
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 0.00047484375,
+      "loss": 0.3735,
+      "step": 664
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 0.0004746875,
+      "loss": 0.3683,
+      "step": 665
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 0.00047453125,
+      "loss": 0.4971,
+      "step": 666
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 0.000474375,
+      "loss": 0.3767,
+      "step": 667
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 0.00047421875000000003,
+      "loss": 0.3758,
+      "step": 668
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 0.0004740625,
+      "loss": 0.4115,
+      "step": 669
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 0.00047390625,
+      "loss": 0.3362,
+      "step": 670
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 0.00047375,
+      "loss": 0.3954,
+      "step": 671
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 0.00047359375,
+      "loss": 0.3436,
+      "step": 672
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 0.0004734375,
+      "loss": 0.4042,
+      "step": 673
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 0.00047328125,
+      "loss": 0.3958,
+      "step": 674
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 0.000473125,
+      "loss": 0.4201,
+      "step": 675
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 0.00047296875000000003,
+      "loss": 0.3464,
+      "step": 676
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 0.00047281250000000003,
+      "loss": 0.4417,
+      "step": 677
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 0.00047265625,
+      "loss": 0.404,
+      "step": 678
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 0.0004725,
+      "loss": 0.3864,
+      "step": 679
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 0.00047234375,
+      "loss": 0.4228,
+      "step": 680
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 0.0004721875,
+      "loss": 0.3534,
+      "step": 681
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 0.00047203125,
+      "loss": 0.4131,
+      "step": 682
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 0.000471875,
+      "loss": 0.3887,
+      "step": 683
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 0.00047171875,
+      "loss": 0.3768,
+      "step": 684
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 0.00047156250000000003,
+      "loss": 0.4303,
+      "step": 685
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 0.00047140625000000004,
+      "loss": 0.3616,
+      "step": 686
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 0.00047125,
+      "loss": 0.3823,
+      "step": 687
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 0.00047109375,
+      "loss": 0.4175,
+      "step": 688
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 0.0004709375,
+      "loss": 0.375,
+      "step": 689
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 0.00047078125,
+      "loss": 0.374,
+      "step": 690
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 0.000470625,
+      "loss": 0.3753,
+      "step": 691
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 0.00047046875,
+      "loss": 0.3884,
+      "step": 692
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 0.00047031250000000003,
+      "loss": 0.3507,
+      "step": 693
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 0.00047015625000000003,
+      "loss": 0.3769,
+      "step": 694
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 0.00047,
+      "loss": 0.4085,
+      "step": 695
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 0.00046984375,
+      "loss": 0.3951,
+      "step": 696
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 0.0004696875,
+      "loss": 0.3985,
+      "step": 697
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 0.00046953125,
+      "loss": 0.4197,
+      "step": 698
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 0.000469375,
+      "loss": 0.4367,
+      "step": 699
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 0.00046921875,
+      "loss": 0.3656,
+      "step": 700
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 0.0004690625,
+      "loss": 0.4606,
+      "step": 701
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 0.00046890625000000003,
+      "loss": 0.45,
+      "step": 702
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 0.00046875,
+      "loss": 0.3504,
+      "step": 703
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 0.00046859375,
+      "loss": 0.3664,
+      "step": 704
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 0.0004684375,
+      "loss": 0.3484,
+      "step": 705
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 0.00046828125,
+      "loss": 0.3604,
+      "step": 706
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 0.000468125,
+      "loss": 0.3417,
+      "step": 707
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 0.00046796875,
+      "loss": 0.4207,
+      "step": 708
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 0.0004678125,
+      "loss": 0.4157,
+      "step": 709
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 0.00046765625000000003,
+      "loss": 0.3858,
+      "step": 710
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 0.00046750000000000003,
+      "loss": 0.3776,
+      "step": 711
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 0.00046734375,
+      "loss": 0.3682,
+      "step": 712
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 0.0004671875,
+      "loss": 0.393,
+      "step": 713
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 0.00046703125,
+      "loss": 0.4005,
+      "step": 714
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 0.000466875,
+      "loss": 0.372,
+      "step": 715
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 0.00046671875,
+      "loss": 0.395,
+      "step": 716
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 0.0004665625,
+      "loss": 0.3819,
+      "step": 717
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 0.00046640625,
+      "loss": 0.3658,
+      "step": 718
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 0.00046625000000000003,
+      "loss": 0.4281,
+      "step": 719
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 0.00046609375,
+      "loss": 0.3786,
+      "step": 720
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 0.0004659375,
+      "loss": 0.3631,
+      "step": 721
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 0.00046578125,
+      "loss": 0.3944,
+      "step": 722
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 0.000465625,
+      "loss": 0.4075,
+      "step": 723
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 0.00046546875,
+      "loss": 0.3782,
+      "step": 724
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 0.0004653125,
+      "loss": 0.3684,
+      "step": 725
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 0.00046515625,
+      "loss": 0.3702,
+      "step": 726
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 0.000465,
+      "loss": 0.3692,
+      "step": 727
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 0.00046484375000000003,
+      "loss": 0.3698,
+      "step": 728
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 0.0004646875,
+      "loss": 0.4136,
+      "step": 729
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 0.00046453125,
+      "loss": 0.3609,
+      "step": 730
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 0.000464375,
+      "loss": 0.33,
+      "step": 731
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 0.00046421875,
+      "loss": 0.3468,
+      "step": 732
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 0.0004640625,
+      "loss": 0.3618,
+      "step": 733
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 0.00046390625,
+      "loss": 0.3889,
+      "step": 734
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 0.00046375,
+      "loss": 0.4332,
+      "step": 735
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 0.00046359375000000003,
+      "loss": 0.3822,
+      "step": 736
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 0.00046343750000000004,
+      "loss": 0.3894,
+      "step": 737
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 0.00046328125,
+      "loss": 0.3933,
+      "step": 738
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 0.000463125,
+      "loss": 0.3892,
+      "step": 739
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.00046296875,
+      "loss": 0.3467,
+      "step": 740
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 0.0004628125,
+      "loss": 0.3728,
+      "step": 741
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 0.00046265625,
+      "loss": 0.3418,
+      "step": 742
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 0.0004625,
+      "loss": 0.3682,
+      "step": 743
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 0.00046234375,
+      "loss": 0.4034,
+      "step": 744
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 0.0004621875,
+      "loss": 0.3592,
+      "step": 745
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 0.00046203125000000004,
+      "loss": 0.3874,
+      "step": 746
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 0.000461875,
+      "loss": 0.3832,
+      "step": 747
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 0.00046171875,
+      "loss": 0.3565,
+      "step": 748
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 0.0004615625,
+      "loss": 0.3593,
+      "step": 749
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 0.00046140625,
+      "loss": 0.3761,
+      "step": 750
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 0.00046125,
+      "loss": 0.3531,
+      "step": 751
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 0.00046109375,
+      "loss": 0.3296,
+      "step": 752
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 0.00046093750000000003,
+      "loss": 0.3895,
+      "step": 753
+    },
+    {
+      "epoch": 20.38,
+      "learning_rate": 0.00046078125,
+      "loss": 0.354,
+      "step": 754
+    },
+    {
+      "epoch": 20.41,
+      "learning_rate": 0.00046062500000000004,
+      "loss": 0.3536,
+      "step": 755
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 0.00046046875,
+      "loss": 0.3447,
+      "step": 756
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 0.0004603125,
+      "loss": 0.363,
+      "step": 757
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 0.00046015625,
+      "loss": 0.3859,
+      "step": 758
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 0.00046,
+      "loss": 0.3768,
+      "step": 759
+    },
+    {
+      "epoch": 20.54,
+      "learning_rate": 0.00045984375,
+      "loss": 0.3673,
+      "step": 760
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 0.0004596875,
+      "loss": 0.3454,
+      "step": 761
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 0.00045953125000000003,
+      "loss": 0.3464,
+      "step": 762
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 0.000459375,
+      "loss": 0.3531,
+      "step": 763
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 0.00045921875,
+      "loss": 0.3838,
+      "step": 764
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 0.0004590625,
+      "loss": 0.3324,
+      "step": 765
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 0.00045890625,
+      "loss": 0.3783,
+      "step": 766
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 0.00045875,
+      "loss": 0.3618,
+      "step": 767
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 0.00045859375,
+      "loss": 0.3398,
+      "step": 768
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 0.0004584375,
+      "loss": 0.3729,
+      "step": 769
+    },
+    {
+      "epoch": 20.81,
+      "learning_rate": 0.00045828125,
+      "loss": 0.4049,
+      "step": 770
+    },
+    {
+      "epoch": 20.84,
+      "learning_rate": 0.00045812500000000004,
+      "loss": 0.3408,
+      "step": 771
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 0.00045796875,
+      "loss": 0.4052,
+      "step": 772
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 0.0004578125,
+      "loss": 0.3795,
+      "step": 773
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 0.00045765625,
+      "loss": 0.3156,
+      "step": 774
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 0.0004575,
+      "loss": 0.338,
+      "step": 775
+    },
+    {
+      "epoch": 20.97,
+      "learning_rate": 0.00045734375,
+      "loss": 0.394,
+      "step": 776
+    },
+    {
+      "epoch": 21.0,
+      "learning_rate": 0.0004571875,
+      "loss": 0.2016,
+      "step": 777
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 0.00045703125,
+      "loss": 0.3412,
+      "step": 778
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 0.000456875,
+      "loss": 0.3324,
+      "step": 779
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 0.00045671875000000004,
+      "loss": 0.3393,
+      "step": 780
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 0.0004565625,
+      "loss": 0.3472,
+      "step": 781
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 0.00045640625,
+      "loss": 0.368,
+      "step": 782
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 0.00045625,
+      "loss": 0.3041,
+      "step": 783
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 0.00045609375,
+      "loss": 0.3685,
+      "step": 784
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 0.0004559375,
+      "loss": 0.3335,
+      "step": 785
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 0.00045578125,
+      "loss": 0.3399,
+      "step": 786
+    },
+    {
+      "epoch": 21.27,
+      "learning_rate": 0.00045562500000000003,
+      "loss": 0.3829,
+      "step": 787
+    },
+    {
+      "epoch": 21.3,
+      "learning_rate": 0.00045546875,
+      "loss": 0.4077,
+      "step": 788
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 0.00045531250000000004,
+      "loss": 0.3433,
+      "step": 789
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 0.00045515625,
+      "loss": 0.3608,
+      "step": 790
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 0.000455,
+      "loss": 0.3616,
+      "step": 791
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 0.00045484375,
+      "loss": 0.3468,
+      "step": 792
+    },
+    {
+      "epoch": 21.43,
+      "learning_rate": 0.0004546875,
+      "loss": 0.3191,
+      "step": 793
+    },
+    {
+      "epoch": 21.46,
+      "learning_rate": 0.00045453125,
+      "loss": 0.3292,
+      "step": 794
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 0.00045437499999999997,
+      "loss": 0.3889,
+      "step": 795
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 0.00045421875000000003,
+      "loss": 0.3291,
+      "step": 796
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 0.0004540625,
+      "loss": 0.3697,
+      "step": 797
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 0.00045390625000000005,
+      "loss": 0.3502,
+      "step": 798
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 0.00045375,
+      "loss": 0.3455,
+      "step": 799
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 0.00045359375,
+      "loss": 0.3473,
+      "step": 800
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 0.0004534375,
+      "loss": 0.3743,
+      "step": 801
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 0.00045328125,
+      "loss": 0.3757,
+      "step": 802
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 0.000453125,
+      "loss": 0.3797,
+      "step": 803
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 0.00045296875,
+      "loss": 0.3105,
+      "step": 804
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 0.00045281250000000004,
+      "loss": 0.3291,
+      "step": 805
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 0.00045265625,
+      "loss": 0.3841,
+      "step": 806
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 0.00045250000000000005,
+      "loss": 0.3561,
+      "step": 807
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 0.00045234375,
+      "loss": 0.3425,
+      "step": 808
+    },
+    {
+      "epoch": 21.86,
+      "learning_rate": 0.0004521875,
+      "loss": 0.3822,
+      "step": 809
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 0.00045203125,
+      "loss": 0.3855,
+      "step": 810
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 0.000451875,
+      "loss": 0.3657,
+      "step": 811
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 0.00045171875,
+      "loss": 0.3706,
+      "step": 812
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 0.0004515625,
+      "loss": 0.3612,
+      "step": 813
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 0.00045140625000000004,
+      "loss": 0.4476,
+      "step": 814
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 0.00045125,
+      "loss": 0.3423,
+      "step": 815
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 0.00045109375000000005,
+      "loss": 0.3275,
+      "step": 816
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 0.0004509375,
+      "loss": 0.3283,
+      "step": 817
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 0.00045078125,
+      "loss": 0.4102,
+      "step": 818
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 0.000450625,
+      "loss": 0.4033,
+      "step": 819
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 0.00045046874999999997,
+      "loss": 0.3673,
+      "step": 820
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 0.00045031250000000003,
+      "loss": 0.3428,
+      "step": 821
+    },
+    {
+      "epoch": 22.22,
+      "learning_rate": 0.00045015625,
+      "loss": 0.3397,
+      "step": 822
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 0.3195,
+      "step": 823
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 0.00044984375,
+      "loss": 0.3438,
+      "step": 824
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 0.0004496875,
+      "loss": 0.3535,
+      "step": 825
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 0.00044953125,
+      "loss": 0.2877,
+      "step": 826
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 0.000449375,
+      "loss": 0.3432,
+      "step": 827
+    },
+    {
+      "epoch": 22.38,
+      "learning_rate": 0.00044921875,
+      "loss": 0.3614,
+      "step": 828
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 0.00044906249999999997,
+      "loss": 0.3254,
+      "step": 829
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 0.00044890625000000003,
+      "loss": 0.3125,
+      "step": 830
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 0.00044875,
+      "loss": 0.3879,
+      "step": 831
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 0.00044859375000000004,
+      "loss": 0.353,
+      "step": 832
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 0.0004484375,
+      "loss": 0.3943,
+      "step": 833
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 0.00044828125,
+      "loss": 0.3662,
+      "step": 834
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 0.000448125,
+      "loss": 0.3374,
+      "step": 835
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 0.00044796875,
+      "loss": 0.3133,
+      "step": 836
+    },
+    {
+      "epoch": 22.62,
+      "learning_rate": 0.0004478125,
+      "loss": 0.3543,
+      "step": 837
+    },
+    {
+      "epoch": 22.65,
+      "learning_rate": 0.00044765625,
+      "loss": 0.3487,
+      "step": 838
+    },
+    {
+      "epoch": 22.68,
+      "learning_rate": 0.00044750000000000004,
+      "loss": 0.3661,
+      "step": 839
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 0.00044734375,
+      "loss": 0.4186,
+      "step": 840
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 0.00044718750000000005,
+      "loss": 0.3228,
+      "step": 841
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 0.00044703125,
+      "loss": 0.3484,
+      "step": 842
+    },
+    {
+      "epoch": 22.78,
+      "learning_rate": 0.000446875,
+      "loss": 0.3657,
+      "step": 843
+    },
+    {
+      "epoch": 22.81,
+      "learning_rate": 0.00044671875,
+      "loss": 0.3418,
+      "step": 844
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 0.00044656249999999997,
+      "loss": 0.3558,
+      "step": 845
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 0.00044640625,
+      "loss": 0.3415,
+      "step": 846
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 0.00044625,
+      "loss": 0.3314,
+      "step": 847
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 0.00044609375000000004,
+      "loss": 0.4194,
+      "step": 848
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 0.0004459375,
+      "loss": 0.4192,
+      "step": 849
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 0.00044578125000000005,
+      "loss": 0.3462,
+      "step": 850
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 0.000445625,
+      "loss": 0.382,
+      "step": 851
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 0.00044546875,
+      "loss": 0.3774,
+      "step": 852
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 0.0004453125,
+      "loss": 0.3321,
+      "step": 853
+    },
+    {
+      "epoch": 23.08,
+      "learning_rate": 0.00044515624999999997,
+      "loss": 0.3248,
+      "step": 854
+    },
+    {
+      "epoch": 23.11,
+      "learning_rate": 0.00044500000000000003,
+      "loss": 0.3162,
+      "step": 855
+    },
+    {
+      "epoch": 23.14,
+      "learning_rate": 0.00044484375,
+      "loss": 0.3593,
+      "step": 856
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 0.00044468750000000004,
+      "loss": 0.3259,
+      "step": 857
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 0.00044453125,
+      "loss": 0.3489,
+      "step": 858
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 0.00044437500000000005,
+      "loss": 0.3422,
+      "step": 859
+    },
+    {
+      "epoch": 23.24,
+      "learning_rate": 0.00044421875,
+      "loss": 0.3302,
+      "step": 860
+    },
+    {
+      "epoch": 23.27,
+      "learning_rate": 0.0004440625,
+      "loss": 0.327,
+      "step": 861
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 0.00044390625,
+      "loss": 0.3204,
+      "step": 862
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 0.00044374999999999997,
+      "loss": 0.3958,
+      "step": 863
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 0.00044359375000000003,
+      "loss": 0.3156,
+      "step": 864
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 0.0004434375,
+      "loss": 0.3161,
+      "step": 865
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 0.00044328125000000004,
+      "loss": 0.3309,
+      "step": 866
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 0.000443125,
+      "loss": 0.2737,
+      "step": 867
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 0.00044296875000000006,
+      "loss": 0.3644,
+      "step": 868
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 0.0004428125,
+      "loss": 0.3038,
+      "step": 869
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 0.00044265624999999996,
+      "loss": 0.2938,
+      "step": 870
+    },
+    {
+      "epoch": 23.54,
+      "learning_rate": 0.0004425,
+      "loss": 0.3202,
+      "step": 871
+    },
+    {
+      "epoch": 23.57,
+      "learning_rate": 0.00044234375,
+      "loss": 0.3057,
+      "step": 872
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 0.00044218750000000003,
+      "loss": 0.3078,
+      "step": 873
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 0.00044203125,
+      "loss": 0.3168,
+      "step": 874
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 0.00044187500000000005,
+      "loss": 0.3279,
+      "step": 875
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 0.00044171875,
+      "loss": 0.3272,
+      "step": 876
+    },
+    {
+      "epoch": 23.7,
+      "learning_rate": 0.00044156250000000006,
+      "loss": 0.409,
+      "step": 877
+    },
+    {
+      "epoch": 23.73,
+      "learning_rate": 0.00044140625,
+      "loss": 0.3368,
+      "step": 878
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 0.00044124999999999996,
+      "loss": 0.3371,
+      "step": 879
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 0.00044109375,
+      "loss": 0.3592,
+      "step": 880
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 0.0004409375,
+      "loss": 0.3379,
+      "step": 881
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 0.00044078125000000004,
+      "loss": 0.2931,
+      "step": 882
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 0.000440625,
+      "loss": 0.3106,
+      "step": 883
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 0.00044046875000000005,
+      "loss": 0.3284,
+      "step": 884
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 0.0004403125,
+      "loss": 0.3434,
+      "step": 885
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 0.00044015625,
+      "loss": 0.3239,
+      "step": 886
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 0.00044,
+      "loss": 0.3601,
+      "step": 887
+    },
+    {
+      "epoch": 24.0,
+      "learning_rate": 0.00043984374999999997,
+      "loss": 0.2733,
+      "step": 888
+    },
+    {
+      "epoch": 24.03,
+      "learning_rate": 0.00043968750000000003,
+      "loss": 0.3334,
+      "step": 889
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 0.00043953125,
+      "loss": 0.3358,
+      "step": 890
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 0.00043937500000000004,
+      "loss": 0.3319,
+      "step": 891
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 0.00043921875,
+      "loss": 0.3855,
+      "step": 892
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 0.00043906250000000005,
+      "loss": 0.307,
+      "step": 893
+    },
+    {
+      "epoch": 24.16,
+      "learning_rate": 0.00043890625,
+      "loss": 0.3047,
+      "step": 894
+    },
+    {
+      "epoch": 24.19,
+      "learning_rate": 0.00043874999999999996,
+      "loss": 0.3395,
+      "step": 895
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 0.00043859375,
+      "loss": 0.3152,
+      "step": 896
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 0.00043843749999999997,
+      "loss": 0.3084,
+      "step": 897
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 0.00043828125000000003,
+      "loss": 0.3001,
+      "step": 898
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 0.000438125,
+      "loss": 0.3585,
+      "step": 899
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 0.00043796875000000004,
+      "loss": 0.2977,
+      "step": 900
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 0.0004378125,
+      "loss": 0.2801,
+      "step": 901
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 0.00043765625000000006,
+      "loss": 0.3053,
+      "step": 902
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 0.0004375,
+      "loss": 0.2924,
+      "step": 903
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 0.00043734374999999996,
+      "loss": 0.3382,
+      "step": 904
+    },
+    {
+      "epoch": 24.46,
+      "learning_rate": 0.0004371875,
+      "loss": 0.3306,
+      "step": 905
+    },
+    {
+      "epoch": 24.49,
+      "learning_rate": 0.00043703125,
+      "loss": 0.3278,
+      "step": 906
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 0.00043687500000000003,
+      "loss": 0.3058,
+      "step": 907
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 0.00043671875,
+      "loss": 0.3073,
+      "step": 908
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 0.00043656250000000005,
+      "loss": 0.3372,
+      "step": 909
+    },
+    {
+      "epoch": 24.59,
+      "learning_rate": 0.00043640625,
+      "loss": 0.3285,
+      "step": 910
+    },
+    {
+      "epoch": 24.62,
+      "learning_rate": 0.00043625000000000006,
+      "loss": 0.3177,
+      "step": 911
+    },
+    {
+      "epoch": 24.65,
+      "learning_rate": 0.00043609375,
+      "loss": 0.3099,
+      "step": 912
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 0.00043593749999999996,
+      "loss": 0.3075,
+      "step": 913
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 0.00043578125,
+      "loss": 0.3439,
+      "step": 914
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 0.000435625,
+      "loss": 0.3115,
+      "step": 915
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 0.00043546875000000004,
+      "loss": 0.2684,
+      "step": 916
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 0.0004353125,
+      "loss": 0.3387,
+      "step": 917
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 0.00043515625000000005,
+      "loss": 0.3547,
+      "step": 918
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 0.000435,
+      "loss": 0.294,
+      "step": 919
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 0.00043484375,
+      "loss": 0.3064,
+      "step": 920
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 0.0004346875,
+      "loss": 0.3313,
+      "step": 921
+    },
+    {
+      "epoch": 24.92,
+      "learning_rate": 0.00043453124999999997,
+      "loss": 0.2962,
+      "step": 922
+    },
+    {
+      "epoch": 24.95,
+      "learning_rate": 0.00043437500000000003,
+      "loss": 0.3099,
+      "step": 923
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 0.00043421875,
+      "loss": 0.3181,
+      "step": 924
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 0.00043406250000000004,
+      "loss": 0.2542,
+      "step": 925
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 0.00043390625,
+      "loss": 0.2641,
+      "step": 926
+    },
+    {
+      "epoch": 25.05,
+      "learning_rate": 0.00043375000000000005,
+      "loss": 0.3167,
+      "step": 927
+    },
+    {
+      "epoch": 25.08,
+      "learning_rate": 0.00043359375,
+      "loss": 0.3068,
+      "step": 928
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 0.00043343749999999996,
+      "loss": 0.3631,
+      "step": 929
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 0.00043328125,
+      "loss": 0.3684,
+      "step": 930
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 0.00043312499999999997,
+      "loss": 0.3105,
+      "step": 931
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 0.00043296875000000003,
+      "loss": 0.3315,
+      "step": 932
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 0.0004328125,
+      "loss": 0.3096,
+      "step": 933
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 0.00043265625000000004,
+      "loss": 0.3244,
+      "step": 934
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 0.0004325,
+      "loss": 0.3359,
+      "step": 935
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 0.00043234375000000006,
+      "loss": 0.2669,
+      "step": 936
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 0.0004321875,
+      "loss": 0.3118,
+      "step": 937
+    },
+    {
+      "epoch": 25.35,
+      "learning_rate": 0.00043203124999999996,
+      "loss": 0.2854,
+      "step": 938
+    },
+    {
+      "epoch": 25.38,
+      "learning_rate": 0.000431875,
+      "loss": 0.3421,
+      "step": 939
+    },
+    {
+      "epoch": 25.41,
+      "learning_rate": 0.00043171875,
+      "loss": 0.3357,
+      "step": 940
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 0.00043156250000000003,
+      "loss": 0.3658,
+      "step": 941
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 0.00043140625,
+      "loss": 0.2996,
+      "step": 942
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 0.00043125000000000005,
+      "loss": 0.2979,
+      "step": 943
+    },
+    {
+      "epoch": 25.51,
+      "learning_rate": 0.00043109375,
+      "loss": 0.3141,
+      "step": 944
+    },
+    {
+      "epoch": 25.54,
+      "learning_rate": 0.0004309375,
+      "loss": 0.2801,
+      "step": 945
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 0.00043078125,
+      "loss": 0.3117,
+      "step": 946
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 0.00043062499999999996,
+      "loss": 0.3145,
+      "step": 947
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 0.00043046875,
+      "loss": 0.3356,
+      "step": 948
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 0.0004303125,
+      "loss": 0.333,
+      "step": 949
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 0.00043015625000000004,
+      "loss": 0.3004,
+      "step": 950
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 0.00043,
+      "loss": 0.3118,
+      "step": 951
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 0.00042984375000000005,
+      "loss": 0.2761,
+      "step": 952
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 0.0004296875,
+      "loss": 0.2824,
+      "step": 953
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 0.00042953125,
+      "loss": 0.3009,
+      "step": 954
+    },
+    {
+      "epoch": 25.81,
+      "learning_rate": 0.000429375,
+      "loss": 0.3518,
+      "step": 955
+    },
+    {
+      "epoch": 25.84,
+      "learning_rate": 0.00042921874999999997,
+      "loss": 0.3299,
+      "step": 956
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 0.00042906250000000003,
+      "loss": 0.3333,
+      "step": 957
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 0.00042890625,
+      "loss": 0.3276,
+      "step": 958
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 0.00042875000000000004,
+      "loss": 0.3049,
+      "step": 959
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 0.00042859375,
+      "loss": 0.247,
+      "step": 960
+    },
+    {
+      "epoch": 25.97,
+      "learning_rate": 0.00042843750000000005,
+      "loss": 0.3278,
+      "step": 961
+    },
+    {
+      "epoch": 26.0,
+      "learning_rate": 0.00042828125,
+      "loss": 0.2277,
+      "step": 962
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 0.000428125,
+      "loss": 0.361,
+      "step": 963
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 0.00042796875,
+      "loss": 0.3074,
+      "step": 964
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 0.00042781249999999997,
+      "loss": 0.3401,
+      "step": 965
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 0.00042765625000000003,
+      "loss": 0.3153,
+      "step": 966
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 0.0004275,
+      "loss": 0.3044,
+      "step": 967
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 0.00042734375000000004,
+      "loss": 0.3048,
+      "step": 968
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 0.0004271875,
+      "loss": 0.3674,
+      "step": 969
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 0.00042703125,
+      "loss": 0.3624,
+      "step": 970
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 0.000426875,
+      "loss": 0.3079,
+      "step": 971
+    },
+    {
+      "epoch": 26.27,
+      "learning_rate": 0.00042671875,
+      "loss": 0.2771,
+      "step": 972
+    },
+    {
+      "epoch": 26.3,
+      "learning_rate": 0.0004265625,
+      "loss": 0.3573,
+      "step": 973
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 0.00042640625,
+      "loss": 0.3229,
+      "step": 974
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 0.00042625000000000003,
+      "loss": 0.3294,
+      "step": 975
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 0.00042609375,
+      "loss": 0.3785,
+      "step": 976
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 0.00042593750000000005,
+      "loss": 0.3129,
+      "step": 977
+    },
+    {
+      "epoch": 26.43,
+      "learning_rate": 0.00042578125,
+      "loss": 0.3029,
+      "step": 978
+    },
+    {
+      "epoch": 26.46,
+      "learning_rate": 0.000425625,
+      "loss": 0.3248,
+      "step": 979
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 0.00042546875,
+      "loss": 0.3343,
+      "step": 980
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 0.0004253125,
+      "loss": 0.3006,
+      "step": 981
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 0.00042515625,
+      "loss": 0.3334,
+      "step": 982
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 0.000425,
+      "loss": 0.3328,
+      "step": 983
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 0.00042484375000000004,
+      "loss": 0.3219,
+      "step": 984
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 0.0004246875,
+      "loss": 0.3432,
+      "step": 985
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 0.00042453125000000005,
+      "loss": 0.2762,
+      "step": 986
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 0.000424375,
+      "loss": 0.3086,
+      "step": 987
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 0.00042421875,
+      "loss": 0.3312,
+      "step": 988
+    },
+    {
+      "epoch": 26.73,
+      "learning_rate": 0.0004240625,
+      "loss": 0.3004,
+      "step": 989
+    },
+    {
+      "epoch": 26.76,
+      "learning_rate": 0.00042390624999999997,
+      "loss": 0.2712,
+      "step": 990
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 0.00042375000000000003,
+      "loss": 0.3154,
+      "step": 991
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 0.00042359375,
+      "loss": 0.3398,
+      "step": 992
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 0.00042343750000000004,
+      "loss": 0.3102,
+      "step": 993
+    },
+    {
+      "epoch": 26.86,
+      "learning_rate": 0.00042328125,
+      "loss": 0.3164,
+      "step": 994
+    },
+    {
+      "epoch": 26.89,
+      "learning_rate": 0.000423125,
+      "loss": 0.2897,
+      "step": 995
+    },
+    {
+      "epoch": 26.92,
+      "learning_rate": 0.00042296875,
+      "loss": 0.3487,
+      "step": 996
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 0.0004228125,
+      "loss": 0.3611,
+      "step": 997
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 0.00042265625,
+      "loss": 0.2976,
+      "step": 998
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 0.00042249999999999997,
+      "loss": 0.3477,
+      "step": 999
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 0.00042234375000000003,
+      "loss": 0.342,
+      "step": 1000
+    },
+    {
+      "epoch": 27.03,
+      "eval_cer": 0.09674807718256645,
+      "eval_loss": 0.3781281113624573,
+      "eval_runtime": 11.6311,
+      "eval_samples_per_second": 141.603,
+      "eval_steps_per_second": 2.235,
+      "eval_wer": 0.39536308855070984,
+      "step": 1000
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 0.0004221875,
+      "loss": 0.3297,
+      "step": 1001
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 0.00042203125000000004,
+      "loss": 0.2976,
+      "step": 1002
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 0.000421875,
+      "loss": 0.291,
+      "step": 1003
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 0.00042171875,
+      "loss": 0.3292,
+      "step": 1004
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 0.0004215625,
+      "loss": 0.3013,
+      "step": 1005
+    },
+    {
+      "epoch": 27.19,
+      "learning_rate": 0.00042140625,
+      "loss": 0.2919,
+      "step": 1006
+    },
+    {
+      "epoch": 27.22,
+      "learning_rate": 0.00042125,
+      "loss": 0.3129,
+      "step": 1007
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 0.00042109374999999997,
+      "loss": 0.2611,
+      "step": 1008
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 0.00042093750000000003,
+      "loss": 0.2592,
+      "step": 1009
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 0.00042078125,
+      "loss": 0.3075,
+      "step": 1010
+    },
+    {
+      "epoch": 27.32,
+      "learning_rate": 0.00042062500000000005,
+      "loss": 0.3171,
+      "step": 1011
+    },
+    {
+      "epoch": 27.35,
+      "learning_rate": 0.00042046875,
+      "loss": 0.3161,
+      "step": 1012
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 0.0004203125,
+      "loss": 0.3107,
+      "step": 1013
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 0.00042015625,
+      "loss": 0.2784,
+      "step": 1014
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 0.00042,
+      "loss": 0.2644,
+      "step": 1015
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 0.00041984375,
+      "loss": 0.3201,
+      "step": 1016
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 0.0004196875,
+      "loss": 0.3115,
+      "step": 1017
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 0.00041953125000000004,
+      "loss": 0.2358,
+      "step": 1018
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 0.000419375,
+      "loss": 0.3181,
+      "step": 1019
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 0.00041921875,
+      "loss": 0.3112,
+      "step": 1020
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 0.0004190625,
+      "loss": 0.2925,
+      "step": 1021
+    },
+    {
+      "epoch": 27.62,
+      "learning_rate": 0.00041890625,
+      "loss": 0.2663,
+      "step": 1022
+    },
+    {
+      "epoch": 27.65,
+      "learning_rate": 0.00041875,
+      "loss": 0.3093,
+      "step": 1023
+    },
+    {
+      "epoch": 27.68,
+      "learning_rate": 0.00041859375,
+      "loss": 0.2696,
+      "step": 1024
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 0.0004184375,
+      "loss": 0.3062,
+      "step": 1025
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 0.00041828125,
+      "loss": 0.3226,
+      "step": 1026
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 0.00041812500000000004,
+      "loss": 0.3055,
+      "step": 1027
+    },
+    {
+      "epoch": 27.78,
+      "learning_rate": 0.00041796875,
+      "loss": 0.2804,
+      "step": 1028
+    },
+    {
+      "epoch": 27.81,
+      "learning_rate": 0.0004178125,
+      "loss": 0.264,
+      "step": 1029
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 0.00041765625,
+      "loss": 0.2733,
+      "step": 1030
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 0.0004175,
+      "loss": 0.3638,
+      "step": 1031
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 0.00041734375,
+      "loss": 0.2552,
+      "step": 1032
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 0.0004171875,
+      "loss": 0.2686,
+      "step": 1033
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 0.00041703125000000003,
+      "loss": 0.3347,
+      "step": 1034
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 0.000416875,
+      "loss": 0.3714,
+      "step": 1035
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 0.00041671875000000004,
+      "loss": 0.2994,
+      "step": 1036
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 0.0004165625,
+      "loss": 0.3675,
+      "step": 1037
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 0.00041640625,
+      "loss": 0.2816,
+      "step": 1038
+    },
+    {
+      "epoch": 28.08,
+      "learning_rate": 0.00041625,
+      "loss": 0.2587,
+      "step": 1039
+    },
+    {
+      "epoch": 28.11,
+      "learning_rate": 0.00041609375,
+      "loss": 0.3026,
+      "step": 1040
+    },
+    {
+      "epoch": 28.14,
+      "learning_rate": 0.0004159375,
+      "loss": 0.3318,
+      "step": 1041
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 0.00041578125,
+      "loss": 0.297,
+      "step": 1042
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 0.00041562500000000003,
+      "loss": 0.3752,
+      "step": 1043
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 0.00041546875,
+      "loss": 0.2493,
+      "step": 1044
+    },
+    {
+      "epoch": 28.24,
+      "learning_rate": 0.0004153125,
+      "loss": 0.309,
+      "step": 1045
+    },
+    {
+      "epoch": 28.27,
+      "learning_rate": 0.00041515625,
+      "loss": 0.287,
+      "step": 1046
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 0.000415,
+      "loss": 0.3041,
+      "step": 1047
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 0.00041484375,
+      "loss": 0.3021,
+      "step": 1048
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 0.0004146875,
+      "loss": 0.2998,
+      "step": 1049
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 0.00041453125,
+      "loss": 0.2714,
+      "step": 1050
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 0.000414375,
+      "loss": 0.2931,
+      "step": 1051
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 0.00041421875000000004,
+      "loss": 0.257,
+      "step": 1052
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 0.0004140625,
+      "loss": 0.2719,
+      "step": 1053
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 0.00041390625,
+      "loss": 0.2423,
+      "step": 1054
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 0.00041375,
+      "loss": 0.289,
+      "step": 1055
+    },
+    {
+      "epoch": 28.54,
+      "learning_rate": 0.00041359375,
+      "loss": 0.3765,
+      "step": 1056
+    },
+    {
+      "epoch": 28.57,
+      "learning_rate": 0.0004134375,
+      "loss": 0.293,
+      "step": 1057
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 0.00041328125,
+      "loss": 0.3189,
+      "step": 1058
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 0.000413125,
+      "loss": 0.3036,
+      "step": 1059
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 0.00041296875,
+      "loss": 0.3105,
+      "step": 1060
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 0.00041281250000000004,
+      "loss": 0.2786,
+      "step": 1061
+    },
+    {
+      "epoch": 28.7,
+      "learning_rate": 0.00041265625,
+      "loss": 0.3123,
+      "step": 1062
+    },
+    {
+      "epoch": 28.73,
+      "learning_rate": 0.0004125,
+      "loss": 0.2676,
+      "step": 1063
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 0.00041234375,
+      "loss": 0.2988,
+      "step": 1064
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 0.0004121875,
+      "loss": 0.3537,
+      "step": 1065
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 0.00041203125,
+      "loss": 0.2642,
+      "step": 1066
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 0.000411875,
+      "loss": 0.2879,
+      "step": 1067
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 0.00041171875000000003,
+      "loss": 0.2931,
+      "step": 1068
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 0.0004115625,
+      "loss": 0.3033,
+      "step": 1069
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 0.00041140625,
+      "loss": 0.2638,
+      "step": 1070
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 0.00041125,
+      "loss": 0.2754,
+      "step": 1071
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 0.00041109375,
+      "loss": 0.3027,
+      "step": 1072
+    },
+    {
+      "epoch": 29.0,
+      "learning_rate": 0.0004109375,
+      "loss": 0.337,
+      "step": 1073
+    },
+    {
+      "epoch": 29.03,
+      "learning_rate": 0.00041078125,
+      "loss": 0.283,
+      "step": 1074
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 0.000410625,
+      "loss": 0.3149,
+      "step": 1075
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 0.00041046875,
+      "loss": 0.3193,
+      "step": 1076
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 0.00041031250000000003,
+      "loss": 0.2432,
+      "step": 1077
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 0.00041015625,
+      "loss": 0.3273,
+      "step": 1078
+    },
+    {
+      "epoch": 29.16,
+      "learning_rate": 0.00041,
+      "loss": 0.2777,
+      "step": 1079
+    },
+    {
+      "epoch": 29.19,
+      "learning_rate": 0.00040984375,
+      "loss": 0.2773,
+      "step": 1080
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 0.0004096875,
+      "loss": 0.3112,
+      "step": 1081
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 0.00040953125,
+      "loss": 0.2719,
+      "step": 1082
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 0.000409375,
+      "loss": 0.3099,
+      "step": 1083
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 0.00040921875,
+      "loss": 0.2553,
+      "step": 1084
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 0.00040906250000000003,
+      "loss": 0.2717,
+      "step": 1085
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 0.00040890625000000004,
+      "loss": 0.2813,
+      "step": 1086
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 0.00040875,
+      "loss": 0.3127,
+      "step": 1087
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 0.00040859375,
+      "loss": 0.264,
+      "step": 1088
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 0.0004084375,
+      "loss": 0.2949,
+      "step": 1089
+    },
+    {
+      "epoch": 29.46,
+      "learning_rate": 0.00040828125,
+      "loss": 0.2569,
+      "step": 1090
+    },
+    {
+      "epoch": 29.49,
+      "learning_rate": 0.000408125,
+      "loss": 0.2892,
+      "step": 1091
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 0.00040796875,
+      "loss": 0.255,
+      "step": 1092
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 0.0004078125,
+      "loss": 0.3312,
+      "step": 1093
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 0.00040765625000000003,
+      "loss": 0.2974,
+      "step": 1094
+    },
+    {
+      "epoch": 29.59,
+      "learning_rate": 0.0004075,
+      "loss": 0.3469,
+      "step": 1095
+    },
+    {
+      "epoch": 29.62,
+      "learning_rate": 0.00040734375,
+      "loss": 0.2993,
+      "step": 1096
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 0.0004071875,
+      "loss": 0.2904,
+      "step": 1097
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 0.00040703125,
+      "loss": 0.2386,
+      "step": 1098
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 0.000406875,
+      "loss": 0.29,
+      "step": 1099
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 0.00040671875,
+      "loss": 0.2851,
+      "step": 1100
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 0.0004065625,
+      "loss": 0.3097,
+      "step": 1101
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 0.00040640625000000003,
+      "loss": 0.2724,
+      "step": 1102
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 0.00040625000000000004,
+      "loss": 0.3279,
+      "step": 1103
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 0.00040609375,
+      "loss": 0.2954,
+      "step": 1104
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 0.0004059375,
+      "loss": 0.2984,
+      "step": 1105
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 0.00040578125,
+      "loss": 0.2515,
+      "step": 1106
+    },
+    {
+      "epoch": 29.92,
+      "learning_rate": 0.000405625,
+      "loss": 0.258,
+      "step": 1107
+    },
+    {
+      "epoch": 29.95,
+      "learning_rate": 0.00040546875,
+      "loss": 0.3383,
+      "step": 1108
+    },
+    {
+      "epoch": 29.97,
+      "learning_rate": 0.0004053125,
+      "loss": 0.2863,
+      "step": 1109
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 0.00040515625,
+      "loss": 0.2741,
+      "step": 1110
+    },
+    {
+      "epoch": 30.03,
+      "learning_rate": 0.00040500000000000003,
+      "loss": 0.3321,
+      "step": 1111
+    },
+    {
+      "epoch": 30.05,
+      "learning_rate": 0.00040484375,
+      "loss": 0.3029,
+      "step": 1112
+    },
+    {
+      "epoch": 30.08,
+      "learning_rate": 0.0004046875,
+      "loss": 0.2891,
+      "step": 1113
+    },
+    {
+      "epoch": 30.11,
+      "learning_rate": 0.00040453125,
+      "loss": 0.243,
+      "step": 1114
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 0.000404375,
+      "loss": 0.2723,
+      "step": 1115
+    },
+    {
+      "epoch": 30.16,
+      "learning_rate": 0.00040421875,
+      "loss": 0.2733,
+      "step": 1116
+    },
+    {
+      "epoch": 30.19,
+      "learning_rate": 0.0004040625,
+      "loss": 0.3069,
+      "step": 1117
+    },
+    {
+      "epoch": 30.22,
+      "learning_rate": 0.00040390625,
+      "loss": 0.3379,
+      "step": 1118
+    },
+    {
+      "epoch": 30.24,
+      "learning_rate": 0.00040375000000000003,
+      "loss": 0.2506,
+      "step": 1119
+    },
+    {
+      "epoch": 30.27,
+      "learning_rate": 0.00040359375,
+      "loss": 0.2593,
+      "step": 1120
+    },
+    {
+      "epoch": 30.3,
+      "learning_rate": 0.0004034375,
+      "loss": 0.2974,
+      "step": 1121
+    },
+    {
+      "epoch": 30.32,
+      "learning_rate": 0.00040328125,
+      "loss": 0.3071,
+      "step": 1122
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 0.000403125,
+      "loss": 0.3374,
+      "step": 1123
+    },
+    {
+      "epoch": 30.38,
+      "learning_rate": 0.00040296875,
+      "loss": 0.2875,
+      "step": 1124
+    },
+    {
+      "epoch": 30.41,
+      "learning_rate": 0.0004028125,
+      "loss": 0.2945,
+      "step": 1125
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 0.00040265625,
+      "loss": 0.2931,
+      "step": 1126
+    },
+    {
+      "epoch": 30.46,
+      "learning_rate": 0.0004025,
+      "loss": 0.2942,
+      "step": 1127
+    },
+    {
+      "epoch": 30.49,
+      "learning_rate": 0.00040234375000000003,
+      "loss": 0.3032,
+      "step": 1128
+    },
+    {
+      "epoch": 30.51,
+      "learning_rate": 0.0004021875,
+      "loss": 0.26,
+      "step": 1129
+    },
+    {
+      "epoch": 30.54,
+      "learning_rate": 0.00040203125,
+      "loss": 0.3371,
+      "step": 1130
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 0.000401875,
+      "loss": 0.2952,
+      "step": 1131
+    },
+    {
+      "epoch": 30.59,
+      "learning_rate": 0.00040171875,
+      "loss": 0.2898,
+      "step": 1132
+    },
+    {
+      "epoch": 30.62,
+      "learning_rate": 0.0004015625,
+      "loss": 0.2989,
+      "step": 1133
+    },
+    {
+      "epoch": 30.65,
+      "learning_rate": 0.00040140625,
+      "loss": 0.1991,
+      "step": 1134
+    },
+    {
+      "epoch": 30.68,
+      "learning_rate": 0.00040125,
+      "loss": 0.2305,
+      "step": 1135
+    },
+    {
+      "epoch": 30.7,
+      "learning_rate": 0.00040109375000000003,
+      "loss": 0.3114,
+      "step": 1136
+    },
+    {
+      "epoch": 30.73,
+      "learning_rate": 0.00040093750000000004,
+      "loss": 0.3078,
+      "step": 1137
+    },
+    {
+      "epoch": 30.76,
+      "learning_rate": 0.00040078125,
+      "loss": 0.291,
+      "step": 1138
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 0.000400625,
+      "loss": 0.2613,
+      "step": 1139
+    },
+    {
+      "epoch": 30.81,
+      "learning_rate": 0.00040046875,
+      "loss": 0.2524,
+      "step": 1140
+    },
+    {
+      "epoch": 30.84,
+      "learning_rate": 0.0004003125,
+      "loss": 0.2638,
+      "step": 1141
+    },
+    {
+      "epoch": 30.86,
+      "learning_rate": 0.00040015625,
+      "loss": 0.308,
+      "step": 1142
+    },
+    {
+      "epoch": 30.89,
+      "learning_rate": 0.0004,
+      "loss": 0.2491,
+      "step": 1143
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 0.00039984375,
+      "loss": 0.2975,
+      "step": 1144
+    },
+    {
+      "epoch": 30.95,
+      "learning_rate": 0.0003996875,
+      "loss": 0.2255,
+      "step": 1145
+    },
+    {
+      "epoch": 30.97,
+      "learning_rate": 0.00039953125000000004,
+      "loss": 0.2716,
+      "step": 1146
+    },
+    {
+      "epoch": 31.0,
+      "learning_rate": 0.000399375,
+      "loss": 0.3114,
+      "step": 1147
+    },
+    {
+      "epoch": 31.03,
+      "learning_rate": 0.00039921875,
+      "loss": 0.298,
+      "step": 1148
+    },
+    {
+      "epoch": 31.05,
+      "learning_rate": 0.0003990625,
+      "loss": 0.2996,
+      "step": 1149
+    },
+    {
+      "epoch": 31.08,
+      "learning_rate": 0.00039890625,
+      "loss": 0.2899,
+      "step": 1150
+    },
+    {
+      "epoch": 31.11,
+      "learning_rate": 0.00039875,
+      "loss": 0.2795,
+      "step": 1151
+    },
+    {
+      "epoch": 31.14,
+      "learning_rate": 0.00039859375,
+      "loss": 0.2591,
+      "step": 1152
+    },
+    {
+      "epoch": 31.16,
+      "learning_rate": 0.00039843750000000003,
+      "loss": 0.2635,
+      "step": 1153
+    },
+    {
+      "epoch": 31.19,
+      "learning_rate": 0.00039828125,
+      "loss": 0.3001,
+      "step": 1154
+    },
+    {
+      "epoch": 31.22,
+      "learning_rate": 0.00039812500000000004,
+      "loss": 0.2909,
+      "step": 1155
+    },
+    {
+      "epoch": 31.24,
+      "learning_rate": 0.00039796875,
+      "loss": 0.2818,
+      "step": 1156
+    },
+    {
+      "epoch": 31.27,
+      "learning_rate": 0.0003978125,
+      "loss": 0.3039,
+      "step": 1157
+    },
+    {
+      "epoch": 31.3,
+      "learning_rate": 0.00039765625,
+      "loss": 0.2561,
+      "step": 1158
+    },
+    {
+      "epoch": 31.32,
+      "learning_rate": 0.0003975,
+      "loss": 0.3022,
+      "step": 1159
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 0.00039734375,
+      "loss": 0.2827,
+      "step": 1160
+    },
+    {
+      "epoch": 31.38,
+      "learning_rate": 0.0003971875,
+      "loss": 0.302,
+      "step": 1161
+    },
+    {
+      "epoch": 31.41,
+      "learning_rate": 0.00039703125000000003,
+      "loss": 0.2048,
+      "step": 1162
+    },
+    {
+      "epoch": 31.43,
+      "learning_rate": 0.000396875,
+      "loss": 0.2487,
+      "step": 1163
+    },
+    {
+      "epoch": 31.46,
+      "learning_rate": 0.00039671875,
+      "loss": 0.2531,
+      "step": 1164
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 0.0003965625,
+      "loss": 0.2661,
+      "step": 1165
+    },
+    {
+      "epoch": 31.51,
+      "learning_rate": 0.00039640625,
+      "loss": 0.2403,
+      "step": 1166
+    },
+    {
+      "epoch": 31.54,
+      "learning_rate": 0.00039625,
+      "loss": 0.2842,
+      "step": 1167
+    },
+    {
+      "epoch": 31.57,
+      "learning_rate": 0.00039609375,
+      "loss": 0.3,
+      "step": 1168
+    },
+    {
+      "epoch": 31.59,
+      "learning_rate": 0.0003959375,
+      "loss": 0.2928,
+      "step": 1169
+    },
+    {
+      "epoch": 31.62,
+      "learning_rate": 0.00039578125,
+      "loss": 0.275,
+      "step": 1170
+    },
+    {
+      "epoch": 31.65,
+      "learning_rate": 0.00039562500000000003,
+      "loss": 0.3024,
+      "step": 1171
+    },
+    {
+      "epoch": 31.68,
+      "learning_rate": 0.00039546875,
+      "loss": 0.3165,
+      "step": 1172
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 0.0003953125,
+      "loss": 0.3611,
+      "step": 1173
+    },
+    {
+      "epoch": 31.73,
+      "learning_rate": 0.00039515625,
+      "loss": 0.2733,
+      "step": 1174
+    },
+    {
+      "epoch": 31.76,
+      "learning_rate": 0.000395,
+      "loss": 0.2722,
+      "step": 1175
+    },
+    {
+      "epoch": 31.78,
+      "learning_rate": 0.00039484375,
+      "loss": 0.3164,
+      "step": 1176
+    },
+    {
+      "epoch": 31.81,
+      "learning_rate": 0.0003946875,
+      "loss": 0.2914,
+      "step": 1177
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 0.00039453125,
+      "loss": 0.2922,
+      "step": 1178
+    },
+    {
+      "epoch": 31.86,
+      "learning_rate": 0.000394375,
+      "loss": 0.2473,
+      "step": 1179
+    },
+    {
+      "epoch": 31.89,
+      "learning_rate": 0.00039421875000000004,
+      "loss": 0.2423,
+      "step": 1180
+    },
+    {
+      "epoch": 31.92,
+      "learning_rate": 0.0003940625,
+      "loss": 0.2976,
+      "step": 1181
+    },
+    {
+      "epoch": 31.95,
+      "learning_rate": 0.00039390625,
+      "loss": 0.3021,
+      "step": 1182
+    },
+    {
+      "epoch": 31.97,
+      "learning_rate": 0.00039375,
+      "loss": 0.2851,
+      "step": 1183
+    },
+    {
+      "epoch": 32.0,
+      "learning_rate": 0.00039359375,
+      "loss": 0.2612,
+      "step": 1184
+    },
+    {
+      "epoch": 32.03,
+      "learning_rate": 0.0003934375,
+      "loss": 0.307,
+      "step": 1185
+    },
+    {
+      "epoch": 32.05,
+      "learning_rate": 0.00039328125,
+      "loss": 0.1999,
+      "step": 1186
+    },
+    {
+      "epoch": 32.08,
+      "learning_rate": 0.00039312500000000003,
+      "loss": 0.2849,
+      "step": 1187
+    },
+    {
+      "epoch": 32.11,
+      "learning_rate": 0.00039296875,
+      "loss": 0.2473,
+      "step": 1188
+    },
+    {
+      "epoch": 32.14,
+      "learning_rate": 0.00039281250000000004,
+      "loss": 0.238,
+      "step": 1189
+    },
+    {
+      "epoch": 32.16,
+      "learning_rate": 0.00039265625,
+      "loss": 0.3027,
+      "step": 1190
+    },
+    {
+      "epoch": 32.19,
+      "learning_rate": 0.0003925,
+      "loss": 0.2685,
+      "step": 1191
+    },
+    {
+      "epoch": 32.22,
+      "learning_rate": 0.00039234375,
+      "loss": 0.3147,
+      "step": 1192
+    },
+    {
+      "epoch": 32.24,
+      "learning_rate": 0.0003921875,
+      "loss": 0.2317,
+      "step": 1193
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 0.00039203125,
+      "loss": 0.2424,
+      "step": 1194
+    },
+    {
+      "epoch": 32.3,
+      "learning_rate": 0.00039187499999999997,
+      "loss": 0.2308,
+      "step": 1195
+    },
+    {
+      "epoch": 32.32,
+      "learning_rate": 0.00039171875000000003,
+      "loss": 0.2866,
+      "step": 1196
+    },
+    {
+      "epoch": 32.35,
+      "learning_rate": 0.0003915625,
+      "loss": 0.2461,
+      "step": 1197
+    },
+    {
+      "epoch": 32.38,
+      "learning_rate": 0.00039140625000000004,
+      "loss": 0.267,
+      "step": 1198
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 0.00039125,
+      "loss": 0.2571,
+      "step": 1199
+    },
+    {
+      "epoch": 32.43,
+      "learning_rate": 0.00039109375,
+      "loss": 0.2977,
+      "step": 1200
+    },
+    {
+      "epoch": 32.46,
+      "learning_rate": 0.0003909375,
+      "loss": 0.2717,
+      "step": 1201
+    },
+    {
+      "epoch": 32.49,
+      "learning_rate": 0.00039078125,
+      "loss": 0.3235,
+      "step": 1202
+    },
+    {
+      "epoch": 32.51,
+      "learning_rate": 0.000390625,
+      "loss": 0.2746,
+      "step": 1203
+    },
+    {
+      "epoch": 32.54,
+      "learning_rate": 0.00039046875,
+      "loss": 0.2854,
+      "step": 1204
+    },
+    {
+      "epoch": 32.57,
+      "learning_rate": 0.00039031250000000003,
+      "loss": 0.2692,
+      "step": 1205
+    },
+    {
+      "epoch": 32.59,
+      "learning_rate": 0.00039015625,
+      "loss": 0.2708,
+      "step": 1206
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 0.2718,
+      "step": 1207
+    },
+    {
+      "epoch": 32.65,
+      "learning_rate": 0.00038984375,
+      "loss": 0.2253,
+      "step": 1208
+    },
+    {
+      "epoch": 32.68,
+      "learning_rate": 0.0003896875,
+      "loss": 0.2415,
+      "step": 1209
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 0.00038953125,
+      "loss": 0.3131,
+      "step": 1210
+    },
+    {
+      "epoch": 32.73,
+      "learning_rate": 0.000389375,
+      "loss": 0.2411,
+      "step": 1211
+    },
+    {
+      "epoch": 32.76,
+      "learning_rate": 0.00038921875,
+      "loss": 0.235,
+      "step": 1212
+    },
+    {
+      "epoch": 32.78,
+      "learning_rate": 0.0003890625,
+      "loss": 0.26,
+      "step": 1213
+    },
+    {
+      "epoch": 32.81,
+      "learning_rate": 0.00038890625000000004,
+      "loss": 0.2629,
+      "step": 1214
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 0.00038875,
+      "loss": 0.2511,
+      "step": 1215
+    },
+    {
+      "epoch": 32.86,
+      "learning_rate": 0.00038859375000000005,
+      "loss": 0.2611,
+      "step": 1216
+    },
+    {
+      "epoch": 32.89,
+      "learning_rate": 0.0003884375,
+      "loss": 0.2504,
+      "step": 1217
+    },
+    {
+      "epoch": 32.92,
+      "learning_rate": 0.00038828125,
+      "loss": 0.288,
+      "step": 1218
+    },
+    {
+      "epoch": 32.95,
+      "learning_rate": 0.000388125,
+      "loss": 0.3153,
+      "step": 1219
+    },
+    {
+      "epoch": 32.97,
+      "learning_rate": 0.00038796874999999997,
+      "loss": 0.2737,
+      "step": 1220
+    },
+    {
+      "epoch": 33.0,
+      "learning_rate": 0.00038781250000000003,
+      "loss": 0.3181,
+      "step": 1221
+    },
+    {
+      "epoch": 33.03,
+      "learning_rate": 0.00038765625,
+      "loss": 0.3191,
+      "step": 1222
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 0.00038750000000000004,
+      "loss": 0.2739,
+      "step": 1223
+    },
+    {
+      "epoch": 33.08,
+      "learning_rate": 0.00038734375,
+      "loss": 0.2422,
+      "step": 1224
+    },
+    {
+      "epoch": 33.11,
+      "learning_rate": 0.0003871875,
+      "loss": 0.2182,
+      "step": 1225
+    },
+    {
+      "epoch": 33.14,
+      "learning_rate": 0.00038703125,
+      "loss": 0.2975,
+      "step": 1226
+    },
+    {
+      "epoch": 33.16,
+      "learning_rate": 0.000386875,
+      "loss": 0.2621,
+      "step": 1227
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 0.00038671875,
+      "loss": 0.2087,
+      "step": 1228
+    },
+    {
+      "epoch": 33.22,
+      "learning_rate": 0.00038656249999999997,
+      "loss": 0.2646,
+      "step": 1229
+    },
+    {
+      "epoch": 33.24,
+      "learning_rate": 0.00038640625000000003,
+      "loss": 0.2579,
+      "step": 1230
+    },
+    {
+      "epoch": 33.27,
+      "learning_rate": 0.00038625,
+      "loss": 0.2294,
+      "step": 1231
+    },
+    {
+      "epoch": 33.3,
+      "learning_rate": 0.00038609375000000004,
+      "loss": 0.2426,
+      "step": 1232
+    },
+    {
+      "epoch": 33.32,
+      "learning_rate": 0.0003859375,
+      "loss": 0.2669,
+      "step": 1233
+    },
+    {
+      "epoch": 33.35,
+      "learning_rate": 0.00038578125,
+      "loss": 0.2358,
+      "step": 1234
+    },
+    {
+      "epoch": 33.38,
+      "learning_rate": 0.000385625,
+      "loss": 0.2392,
+      "step": 1235
+    },
+    {
+      "epoch": 33.41,
+      "learning_rate": 0.00038546875,
+      "loss": 0.2496,
+      "step": 1236
+    },
+    {
+      "epoch": 33.43,
+      "learning_rate": 0.0003853125,
+      "loss": 0.2579,
+      "step": 1237
+    },
+    {
+      "epoch": 33.46,
+      "learning_rate": 0.00038515625,
+      "loss": 0.2318,
+      "step": 1238
+    },
+    {
+      "epoch": 33.49,
+      "learning_rate": 0.00038500000000000003,
+      "loss": 0.2592,
+      "step": 1239
+    },
+    {
+      "epoch": 33.51,
+      "learning_rate": 0.00038484375,
+      "loss": 0.2182,
+      "step": 1240
+    },
+    {
+      "epoch": 33.54,
+      "learning_rate": 0.00038468750000000005,
+      "loss": 0.2774,
+      "step": 1241
+    },
+    {
+      "epoch": 33.57,
+      "learning_rate": 0.00038453125,
+      "loss": 0.2663,
+      "step": 1242
+    },
+    {
+      "epoch": 33.59,
+      "learning_rate": 0.000384375,
+      "loss": 0.3021,
+      "step": 1243
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 0.00038421875,
+      "loss": 0.2631,
+      "step": 1244
+    },
+    {
+      "epoch": 33.65,
+      "learning_rate": 0.00038406249999999996,
+      "loss": 0.3048,
+      "step": 1245
+    },
+    {
+      "epoch": 33.68,
+      "learning_rate": 0.00038390625,
+      "loss": 0.2571,
+      "step": 1246
+    },
+    {
+      "epoch": 33.7,
+      "learning_rate": 0.00038375,
+      "loss": 0.2723,
+      "step": 1247
+    },
+    {
+      "epoch": 33.73,
+      "learning_rate": 0.00038359375000000004,
+      "loss": 0.2555,
+      "step": 1248
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 0.0003834375,
+      "loss": 0.253,
+      "step": 1249
+    },
+    {
+      "epoch": 33.78,
+      "learning_rate": 0.00038328125000000005,
+      "loss": 0.2917,
+      "step": 1250
+    },
+    {
+      "epoch": 33.81,
+      "learning_rate": 0.000383125,
+      "loss": 0.2835,
+      "step": 1251
+    },
+    {
+      "epoch": 33.84,
+      "learning_rate": 0.00038296875,
+      "loss": 0.2662,
+      "step": 1252
+    },
+    {
+      "epoch": 33.86,
+      "learning_rate": 0.0003828125,
+      "loss": 0.2809,
+      "step": 1253
+    },
+    {
+      "epoch": 33.89,
+      "learning_rate": 0.00038265624999999997,
+      "loss": 0.2913,
+      "step": 1254
+    },
+    {
+      "epoch": 33.92,
+      "learning_rate": 0.00038250000000000003,
+      "loss": 0.2689,
+      "step": 1255
+    },
+    {
+      "epoch": 33.95,
+      "learning_rate": 0.00038234375,
+      "loss": 0.278,
+      "step": 1256
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 0.00038218750000000004,
+      "loss": 0.2577,
+      "step": 1257
+    },
+    {
+      "epoch": 34.0,
+      "learning_rate": 0.00038203125,
+      "loss": 0.1845,
+      "step": 1258
+    },
+    {
+      "epoch": 34.03,
+      "learning_rate": 0.00038187500000000005,
+      "loss": 0.2865,
+      "step": 1259
+    },
+    {
+      "epoch": 34.05,
+      "learning_rate": 0.00038171875,
+      "loss": 0.2595,
+      "step": 1260
+    },
+    {
+      "epoch": 34.08,
+      "learning_rate": 0.0003815625,
+      "loss": 0.309,
+      "step": 1261
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 0.00038140625,
+      "loss": 0.2659,
+      "step": 1262
+    },
+    {
+      "epoch": 34.14,
+      "learning_rate": 0.00038124999999999997,
+      "loss": 0.2765,
+      "step": 1263
+    },
+    {
+      "epoch": 34.16,
+      "learning_rate": 0.00038109375000000003,
+      "loss": 0.2494,
+      "step": 1264
+    },
+    {
+      "epoch": 34.19,
+      "learning_rate": 0.0003809375,
+      "loss": 0.2731,
+      "step": 1265
+    },
+    {
+      "epoch": 34.22,
+      "learning_rate": 0.00038078125000000004,
+      "loss": 0.2693,
+      "step": 1266
+    },
+    {
+      "epoch": 34.24,
+      "learning_rate": 0.000380625,
+      "loss": 0.2754,
+      "step": 1267
+    },
+    {
+      "epoch": 34.27,
+      "learning_rate": 0.00038046875000000006,
+      "loss": 0.292,
+      "step": 1268
+    },
+    {
+      "epoch": 34.3,
+      "learning_rate": 0.0003803125,
+      "loss": 0.3076,
+      "step": 1269
+    },
+    {
+      "epoch": 34.32,
+      "learning_rate": 0.00038015624999999996,
+      "loss": 0.2697,
+      "step": 1270
+    },
+    {
+      "epoch": 34.35,
+      "learning_rate": 0.00038,
+      "loss": 0.2679,
+      "step": 1271
+    },
+    {
+      "epoch": 34.38,
+      "learning_rate": 0.00037984375,
+      "loss": 0.2615,
+      "step": 1272
+    },
+    {
+      "epoch": 34.41,
+      "learning_rate": 0.00037968750000000003,
+      "loss": 0.2502,
+      "step": 1273
+    },
+    {
+      "epoch": 34.43,
+      "learning_rate": 0.00037953125,
+      "loss": 0.3082,
+      "step": 1274
+    },
+    {
+      "epoch": 34.46,
+      "learning_rate": 0.00037937500000000005,
+      "loss": 0.353,
+      "step": 1275
+    },
+    {
+      "epoch": 34.49,
+      "learning_rate": 0.00037921875,
+      "loss": 0.2162,
+      "step": 1276
+    },
+    {
+      "epoch": 34.51,
+      "learning_rate": 0.00037906250000000006,
+      "loss": 0.2569,
+      "step": 1277
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 0.00037890625,
+      "loss": 0.32,
+      "step": 1278
+    },
+    {
+      "epoch": 34.57,
+      "learning_rate": 0.00037874999999999996,
+      "loss": 0.2494,
+      "step": 1279
+    },
+    {
+      "epoch": 34.59,
+      "learning_rate": 0.00037859375,
+      "loss": 0.2221,
+      "step": 1280
+    },
+    {
+      "epoch": 34.62,
+      "learning_rate": 0.0003784375,
+      "loss": 0.2745,
+      "step": 1281
+    },
+    {
+      "epoch": 34.65,
+      "learning_rate": 0.00037828125000000004,
+      "loss": 0.2787,
+      "step": 1282
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 0.000378125,
+      "loss": 0.2545,
+      "step": 1283
+    },
+    {
+      "epoch": 34.7,
+      "learning_rate": 0.00037796875000000005,
+      "loss": 0.2681,
+      "step": 1284
+    },
+    {
+      "epoch": 34.73,
+      "learning_rate": 0.0003778125,
+      "loss": 0.2491,
+      "step": 1285
+    },
+    {
+      "epoch": 34.76,
+      "learning_rate": 0.00037765625,
+      "loss": 0.2625,
+      "step": 1286
+    },
+    {
+      "epoch": 34.78,
+      "learning_rate": 0.0003775,
+      "loss": 0.2593,
+      "step": 1287
+    },
+    {
+      "epoch": 34.81,
+      "learning_rate": 0.00037734374999999997,
+      "loss": 0.2229,
+      "step": 1288
+    },
+    {
+      "epoch": 34.84,
+      "learning_rate": 0.0003771875,
+      "loss": 0.3269,
+      "step": 1289
+    },
+    {
+      "epoch": 34.86,
+      "learning_rate": 0.00037703125,
+      "loss": 0.2459,
+      "step": 1290
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 0.00037687500000000004,
+      "loss": 0.2884,
+      "step": 1291
+    },
+    {
+      "epoch": 34.92,
+      "learning_rate": 0.00037671875,
+      "loss": 0.2305,
+      "step": 1292
+    },
+    {
+      "epoch": 34.95,
+      "learning_rate": 0.00037656250000000005,
+      "loss": 0.2474,
+      "step": 1293
+    },
+    {
+      "epoch": 34.97,
+      "learning_rate": 0.00037640625,
+      "loss": 0.2604,
+      "step": 1294
+    },
+    {
+      "epoch": 35.0,
+      "learning_rate": 0.00037624999999999996,
+      "loss": 0.1428,
+      "step": 1295
+    },
+    {
+      "epoch": 35.03,
+      "learning_rate": 0.00037609375,
+      "loss": 0.2871,
+      "step": 1296
+    },
+    {
+      "epoch": 35.05,
+      "learning_rate": 0.00037593749999999997,
+      "loss": 0.2621,
+      "step": 1297
+    },
+    {
+      "epoch": 35.08,
+      "learning_rate": 0.00037578125000000003,
+      "loss": 0.2869,
+      "step": 1298
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 0.000375625,
+      "loss": 0.2569,
+      "step": 1299
+    },
+    {
+      "epoch": 35.14,
+      "learning_rate": 0.00037546875000000004,
+      "loss": 0.3093,
+      "step": 1300
+    },
+    {
+      "epoch": 35.16,
+      "learning_rate": 0.0003753125,
+      "loss": 0.2788,
+      "step": 1301
+    },
+    {
+      "epoch": 35.19,
+      "learning_rate": 0.00037515625000000006,
+      "loss": 0.2752,
+      "step": 1302
+    },
+    {
+      "epoch": 35.22,
+      "learning_rate": 0.000375,
+      "loss": 0.2806,
+      "step": 1303
+    },
+    {
+      "epoch": 35.24,
+      "learning_rate": 0.00037484374999999996,
+      "loss": 0.228,
+      "step": 1304
+    },
+    {
+      "epoch": 35.27,
+      "learning_rate": 0.0003746875,
+      "loss": 0.2561,
+      "step": 1305
+    },
+    {
+      "epoch": 35.3,
+      "learning_rate": 0.00037453124999999997,
+      "loss": 0.2807,
+      "step": 1306
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 0.00037437500000000003,
+      "loss": 0.2864,
+      "step": 1307
+    },
+    {
+      "epoch": 35.35,
+      "learning_rate": 0.00037421875,
+      "loss": 0.2399,
+      "step": 1308
+    },
+    {
+      "epoch": 35.38,
+      "learning_rate": 0.00037406250000000005,
+      "loss": 0.2617,
+      "step": 1309
+    },
+    {
+      "epoch": 35.41,
+      "learning_rate": 0.00037390625,
+      "loss": 0.2464,
+      "step": 1310
+    },
+    {
+      "epoch": 35.43,
+      "learning_rate": 0.00037375000000000006,
+      "loss": 0.2432,
+      "step": 1311
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 0.00037359375,
+      "loss": 0.2762,
+      "step": 1312
+    },
+    {
+      "epoch": 35.49,
+      "learning_rate": 0.00037343749999999996,
+      "loss": 0.2765,
+      "step": 1313
+    },
+    {
+      "epoch": 35.51,
+      "learning_rate": 0.00037328125,
+      "loss": 0.278,
+      "step": 1314
+    },
+    {
+      "epoch": 35.54,
+      "learning_rate": 0.000373125,
+      "loss": 0.2584,
+      "step": 1315
+    },
+    {
+      "epoch": 35.57,
+      "learning_rate": 0.00037296875000000004,
+      "loss": 0.2352,
+      "step": 1316
+    },
+    {
+      "epoch": 35.59,
+      "learning_rate": 0.0003728125,
+      "loss": 0.2384,
+      "step": 1317
+    },
+    {
+      "epoch": 35.62,
+      "learning_rate": 0.00037265625000000005,
+      "loss": 0.254,
+      "step": 1318
+    },
+    {
+      "epoch": 35.65,
+      "learning_rate": 0.0003725,
+      "loss": 0.2348,
+      "step": 1319
+    },
+    {
+      "epoch": 35.68,
+      "learning_rate": 0.00037234375,
+      "loss": 0.2252,
+      "step": 1320
+    },
+    {
+      "epoch": 35.7,
+      "learning_rate": 0.0003721875,
+      "loss": 0.3198,
+      "step": 1321
+    },
+    {
+      "epoch": 35.73,
+      "learning_rate": 0.00037203124999999997,
+      "loss": 0.256,
+      "step": 1322
+    },
+    {
+      "epoch": 35.76,
+      "learning_rate": 0.000371875,
+      "loss": 0.2975,
+      "step": 1323
+    },
+    {
+      "epoch": 35.78,
+      "learning_rate": 0.00037171875,
+      "loss": 0.3407,
+      "step": 1324
+    },
+    {
+      "epoch": 35.81,
+      "learning_rate": 0.00037156250000000004,
+      "loss": 0.2467,
+      "step": 1325
+    },
+    {
+      "epoch": 35.84,
+      "learning_rate": 0.00037140625,
+      "loss": 0.1863,
+      "step": 1326
+    },
+    {
+      "epoch": 35.86,
+      "learning_rate": 0.00037125000000000005,
+      "loss": 0.2882,
+      "step": 1327
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 0.00037109375,
+      "loss": 0.2907,
+      "step": 1328
+    },
+    {
+      "epoch": 35.92,
+      "learning_rate": 0.00037093749999999996,
+      "loss": 0.2739,
+      "step": 1329
+    },
+    {
+      "epoch": 35.95,
+      "learning_rate": 0.00037078125,
+      "loss": 0.2688,
+      "step": 1330
+    },
+    {
+      "epoch": 35.97,
+      "learning_rate": 0.00037062499999999997,
+      "loss": 0.3172,
+      "step": 1331
+    },
+    {
+      "epoch": 36.0,
+      "learning_rate": 0.00037046875000000003,
+      "loss": 0.1966,
+      "step": 1332
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 0.0003703125,
+      "loss": 0.2845,
+      "step": 1333
+    },
+    {
+      "epoch": 36.05,
+      "learning_rate": 0.00037015625000000004,
+      "loss": 0.2556,
+      "step": 1334
+    },
+    {
+      "epoch": 36.08,
+      "learning_rate": 0.00037,
+      "loss": 0.3027,
+      "step": 1335
+    },
+    {
+      "epoch": 36.11,
+      "learning_rate": 0.00036984375000000006,
+      "loss": 0.2566,
+      "step": 1336
+    },
+    {
+      "epoch": 36.14,
+      "learning_rate": 0.0003696875,
+      "loss": 0.2356,
+      "step": 1337
+    },
+    {
+      "epoch": 36.16,
+      "learning_rate": 0.00036953124999999996,
+      "loss": 0.2336,
+      "step": 1338
+    },
+    {
+      "epoch": 36.19,
+      "learning_rate": 0.000369375,
+      "loss": 0.2507,
+      "step": 1339
+    },
+    {
+      "epoch": 36.22,
+      "learning_rate": 0.00036921874999999997,
+      "loss": 0.2114,
+      "step": 1340
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 0.00036906250000000003,
+      "loss": 0.2439,
+      "step": 1341
+    },
+    {
+      "epoch": 36.27,
+      "learning_rate": 0.00036890625,
+      "loss": 0.2934,
+      "step": 1342
+    },
+    {
+      "epoch": 36.3,
+      "learning_rate": 0.00036875000000000005,
+      "loss": 0.2676,
+      "step": 1343
+    },
+    {
+      "epoch": 36.32,
+      "learning_rate": 0.00036859375,
+      "loss": 0.2487,
+      "step": 1344
+    },
+    {
+      "epoch": 36.35,
+      "learning_rate": 0.0003684375,
+      "loss": 0.2622,
+      "step": 1345
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 0.00036828125,
+      "loss": 0.2717,
+      "step": 1346
+    },
+    {
+      "epoch": 36.41,
+      "learning_rate": 0.00036812499999999996,
+      "loss": 0.254,
+      "step": 1347
+    },
+    {
+      "epoch": 36.43,
+      "learning_rate": 0.00036796875,
+      "loss": 0.2576,
+      "step": 1348
+    },
+    {
+      "epoch": 36.46,
+      "learning_rate": 0.0003678125,
+      "loss": 0.2703,
+      "step": 1349
+    },
+    {
+      "epoch": 36.49,
+      "learning_rate": 0.00036765625000000004,
+      "loss": 0.2986,
+      "step": 1350
+    },
+    {
+      "epoch": 36.51,
+      "learning_rate": 0.0003675,
+      "loss": 0.2225,
+      "step": 1351
+    },
+    {
+      "epoch": 36.54,
+      "learning_rate": 0.00036734375000000005,
+      "loss": 0.2832,
+      "step": 1352
+    },
+    {
+      "epoch": 36.57,
+      "learning_rate": 0.0003671875,
+      "loss": 0.2398,
+      "step": 1353
+    },
+    {
+      "epoch": 36.59,
+      "learning_rate": 0.00036703125,
+      "loss": 0.3037,
+      "step": 1354
+    },
+    {
+      "epoch": 36.62,
+      "learning_rate": 0.000366875,
+      "loss": 0.1957,
+      "step": 1355
+    },
+    {
+      "epoch": 36.65,
+      "learning_rate": 0.00036671874999999997,
+      "loss": 0.2435,
+      "step": 1356
+    },
+    {
+      "epoch": 36.68,
+      "learning_rate": 0.0003665625,
+      "loss": 0.2561,
+      "step": 1357
+    },
+    {
+      "epoch": 36.7,
+      "learning_rate": 0.00036640625,
+      "loss": 0.281,
+      "step": 1358
+    },
+    {
+      "epoch": 36.73,
+      "learning_rate": 0.00036625000000000004,
+      "loss": 0.2924,
+      "step": 1359
+    },
+    {
+      "epoch": 36.76,
+      "learning_rate": 0.00036609375,
+      "loss": 0.2324,
+      "step": 1360
+    },
+    {
+      "epoch": 36.78,
+      "learning_rate": 0.00036593750000000005,
+      "loss": 0.2639,
+      "step": 1361
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 0.00036578125,
+      "loss": 0.2773,
+      "step": 1362
+    },
+    {
+      "epoch": 36.84,
+      "learning_rate": 0.000365625,
+      "loss": 0.2643,
+      "step": 1363
+    },
+    {
+      "epoch": 36.86,
+      "learning_rate": 0.00036546875,
+      "loss": 0.2645,
+      "step": 1364
+    },
+    {
+      "epoch": 36.89,
+      "learning_rate": 0.00036531249999999997,
+      "loss": 0.262,
+      "step": 1365
+    },
+    {
+      "epoch": 36.92,
+      "learning_rate": 0.00036515625000000003,
+      "loss": 0.2548,
+      "step": 1366
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 0.000365,
+      "loss": 0.2431,
+      "step": 1367
+    },
+    {
+      "epoch": 36.97,
+      "learning_rate": 0.00036484375000000004,
+      "loss": 0.2436,
+      "step": 1368
+    },
+    {
+      "epoch": 37.0,
+      "learning_rate": 0.0003646875,
+      "loss": 0.2491,
+      "step": 1369
+    },
+    {
+      "epoch": 37.03,
+      "learning_rate": 0.00036453125,
+      "loss": 0.2936,
+      "step": 1370
+    },
+    {
+      "epoch": 37.05,
+      "learning_rate": 0.000364375,
+      "loss": 0.2369,
+      "step": 1371
+    },
+    {
+      "epoch": 37.08,
+      "learning_rate": 0.00036421875,
+      "loss": 0.2508,
+      "step": 1372
+    },
+    {
+      "epoch": 37.11,
+      "learning_rate": 0.0003640625,
+      "loss": 0.2642,
+      "step": 1373
+    },
+    {
+      "epoch": 37.14,
+      "learning_rate": 0.00036390624999999997,
+      "loss": 0.2744,
+      "step": 1374
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 0.00036375000000000003,
+      "loss": 0.2805,
+      "step": 1375
+    },
+    {
+      "epoch": 37.19,
+      "learning_rate": 0.00036359375,
+      "loss": 0.291,
+      "step": 1376
+    },
+    {
+      "epoch": 37.22,
+      "learning_rate": 0.00036343750000000005,
+      "loss": 0.2859,
+      "step": 1377
+    },
+    {
+      "epoch": 37.24,
+      "learning_rate": 0.00036328125,
+      "loss": 0.2647,
+      "step": 1378
+    },
+    {
+      "epoch": 37.27,
+      "learning_rate": 0.000363125,
+      "loss": 0.2041,
+      "step": 1379
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 0.00036296875,
+      "loss": 0.2831,
+      "step": 1380
+    },
+    {
+      "epoch": 37.32,
+      "learning_rate": 0.0003628125,
+      "loss": 0.2359,
+      "step": 1381
+    },
+    {
+      "epoch": 37.35,
+      "learning_rate": 0.00036265625,
+      "loss": 0.2659,
+      "step": 1382
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 0.0003625,
+      "loss": 0.2308,
+      "step": 1383
+    },
+    {
+      "epoch": 37.41,
+      "learning_rate": 0.00036234375000000004,
+      "loss": 0.2871,
+      "step": 1384
+    },
+    {
+      "epoch": 37.43,
+      "learning_rate": 0.0003621875,
+      "loss": 0.2799,
+      "step": 1385
+    },
+    {
+      "epoch": 37.46,
+      "learning_rate": 0.00036203125000000005,
+      "loss": 0.245,
+      "step": 1386
+    },
+    {
+      "epoch": 37.49,
+      "learning_rate": 0.000361875,
+      "loss": 0.2227,
+      "step": 1387
+    },
+    {
+      "epoch": 37.51,
+      "learning_rate": 0.00036171875,
+      "loss": 0.2848,
+      "step": 1388
+    },
+    {
+      "epoch": 37.54,
+      "learning_rate": 0.0003615625,
+      "loss": 0.295,
+      "step": 1389
+    },
+    {
+      "epoch": 37.57,
+      "learning_rate": 0.00036140624999999997,
+      "loss": 0.3183,
+      "step": 1390
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 0.00036125,
+      "loss": 0.2482,
+      "step": 1391
+    },
+    {
+      "epoch": 37.62,
+      "learning_rate": 0.00036109375,
+      "loss": 0.2796,
+      "step": 1392
+    },
+    {
+      "epoch": 37.65,
+      "learning_rate": 0.00036093750000000004,
+      "loss": 0.2315,
+      "step": 1393
+    },
+    {
+      "epoch": 37.68,
+      "learning_rate": 0.00036078125,
+      "loss": 0.2512,
+      "step": 1394
+    },
+    {
+      "epoch": 37.7,
+      "learning_rate": 0.000360625,
+      "loss": 0.3152,
+      "step": 1395
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 0.00036046875,
+      "loss": 0.2948,
+      "step": 1396
+    },
+    {
+      "epoch": 37.76,
+      "learning_rate": 0.0003603125,
+      "loss": 0.2724,
+      "step": 1397
+    },
+    {
+      "epoch": 37.78,
+      "learning_rate": 0.00036015625,
+      "loss": 0.2319,
+      "step": 1398
+    },
+    {
+      "epoch": 37.81,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 0.2932,
+      "step": 1399
+    },
+    {
+      "epoch": 37.84,
+      "learning_rate": 0.00035984375000000003,
+      "loss": 0.2684,
+      "step": 1400
+    },
+    {
+      "epoch": 37.86,
+      "learning_rate": 0.0003596875,
+      "loss": 0.2277,
+      "step": 1401
+    },
+    {
+      "epoch": 37.89,
+      "learning_rate": 0.00035953125000000004,
+      "loss": 0.2326,
+      "step": 1402
+    },
+    {
+      "epoch": 37.92,
+      "learning_rate": 0.000359375,
+      "loss": 0.2512,
+      "step": 1403
+    },
+    {
+      "epoch": 37.95,
+      "learning_rate": 0.00035921875,
+      "loss": 0.3007,
+      "step": 1404
+    },
+    {
+      "epoch": 37.97,
+      "learning_rate": 0.0003590625,
+      "loss": 0.2752,
+      "step": 1405
+    },
+    {
+      "epoch": 38.0,
+      "learning_rate": 0.00035890625,
+      "loss": 0.1775,
+      "step": 1406
+    },
+    {
+      "epoch": 38.03,
+      "learning_rate": 0.00035875,
+      "loss": 0.2509,
+      "step": 1407
+    },
+    {
+      "epoch": 38.05,
+      "learning_rate": 0.00035859374999999997,
+      "loss": 0.2456,
+      "step": 1408
+    },
+    {
+      "epoch": 38.08,
+      "learning_rate": 0.00035843750000000003,
+      "loss": 0.2756,
+      "step": 1409
+    },
+    {
+      "epoch": 38.11,
+      "learning_rate": 0.00035828125,
+      "loss": 0.2311,
+      "step": 1410
+    },
+    {
+      "epoch": 38.14,
+      "learning_rate": 0.00035812500000000004,
+      "loss": 0.2813,
+      "step": 1411
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 0.00035796875,
+      "loss": 0.2283,
+      "step": 1412
+    },
+    {
+      "epoch": 38.19,
+      "learning_rate": 0.0003578125,
+      "loss": 0.3162,
+      "step": 1413
+    },
+    {
+      "epoch": 38.22,
+      "learning_rate": 0.00035765625,
+      "loss": 0.2754,
+      "step": 1414
+    },
+    {
+      "epoch": 38.24,
+      "learning_rate": 0.0003575,
+      "loss": 0.2696,
+      "step": 1415
+    },
+    {
+      "epoch": 38.27,
+      "learning_rate": 0.00035734375,
+      "loss": 0.2654,
+      "step": 1416
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 0.0003571875,
+      "loss": 0.257,
+      "step": 1417
+    },
+    {
+      "epoch": 38.32,
+      "learning_rate": 0.00035703125000000004,
+      "loss": 0.2543,
+      "step": 1418
+    },
+    {
+      "epoch": 38.35,
+      "learning_rate": 0.000356875,
+      "loss": 0.2457,
+      "step": 1419
+    },
+    {
+      "epoch": 38.38,
+      "learning_rate": 0.00035671875,
+      "loss": 0.2793,
+      "step": 1420
+    },
+    {
+      "epoch": 38.41,
+      "learning_rate": 0.0003565625,
+      "loss": 0.2695,
+      "step": 1421
+    },
+    {
+      "epoch": 38.43,
+      "learning_rate": 0.00035640625,
+      "loss": 0.2563,
+      "step": 1422
+    },
+    {
+      "epoch": 38.46,
+      "learning_rate": 0.00035625,
+      "loss": 0.2666,
+      "step": 1423
+    },
+    {
+      "epoch": 38.49,
+      "learning_rate": 0.00035609375,
+      "loss": 0.2587,
+      "step": 1424
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 0.0003559375,
+      "loss": 0.2538,
+      "step": 1425
+    },
+    {
+      "epoch": 38.54,
+      "learning_rate": 0.00035578125,
+      "loss": 0.3061,
+      "step": 1426
+    },
+    {
+      "epoch": 38.57,
+      "learning_rate": 0.00035562500000000004,
+      "loss": 0.2799,
+      "step": 1427
+    },
+    {
+      "epoch": 38.59,
+      "learning_rate": 0.00035546875,
+      "loss": 0.2416,
+      "step": 1428
+    },
+    {
+      "epoch": 38.62,
+      "learning_rate": 0.0003553125,
+      "loss": 0.285,
+      "step": 1429
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 0.00035515625,
+      "loss": 0.3062,
+      "step": 1430
+    },
+    {
+      "epoch": 38.68,
+      "learning_rate": 0.000355,
+      "loss": 0.2468,
+      "step": 1431
+    },
+    {
+      "epoch": 38.7,
+      "learning_rate": 0.00035484375,
+      "loss": 0.2476,
+      "step": 1432
+    },
+    {
+      "epoch": 38.73,
+      "learning_rate": 0.0003546875,
+      "loss": 0.2426,
+      "step": 1433
+    },
+    {
+      "epoch": 38.76,
+      "learning_rate": 0.00035453125000000003,
+      "loss": 0.256,
+      "step": 1434
+    },
+    {
+      "epoch": 38.78,
+      "learning_rate": 0.000354375,
+      "loss": 0.3021,
+      "step": 1435
+    },
+    {
+      "epoch": 38.81,
+      "learning_rate": 0.00035421875000000004,
+      "loss": 0.2649,
+      "step": 1436
+    },
+    {
+      "epoch": 38.84,
+      "learning_rate": 0.0003540625,
+      "loss": 0.2316,
+      "step": 1437
+    },
+    {
+      "epoch": 38.86,
+      "learning_rate": 0.00035390625,
+      "loss": 0.2756,
+      "step": 1438
+    },
+    {
+      "epoch": 38.89,
+      "learning_rate": 0.00035375,
+      "loss": 0.272,
+      "step": 1439
+    },
+    {
+      "epoch": 38.92,
+      "learning_rate": 0.00035359375,
+      "loss": 0.2772,
+      "step": 1440
+    },
+    {
+      "epoch": 38.95,
+      "learning_rate": 0.0003534375,
+      "loss": 0.2652,
+      "step": 1441
+    },
+    {
+      "epoch": 38.97,
+      "learning_rate": 0.00035328125,
+      "loss": 0.2778,
+      "step": 1442
+    },
+    {
+      "epoch": 39.0,
+      "learning_rate": 0.00035312500000000003,
+      "loss": 0.1365,
+      "step": 1443
+    },
+    {
+      "epoch": 39.03,
+      "learning_rate": 0.00035296875,
+      "loss": 0.3031,
+      "step": 1444
+    },
+    {
+      "epoch": 39.05,
+      "learning_rate": 0.0003528125,
+      "loss": 0.2771,
+      "step": 1445
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 0.00035265625,
+      "loss": 0.2373,
+      "step": 1446
+    },
+    {
+      "epoch": 39.11,
+      "learning_rate": 0.0003525,
+      "loss": 0.2392,
+      "step": 1447
+    },
+    {
+      "epoch": 39.14,
+      "learning_rate": 0.00035234375,
+      "loss": 0.2292,
+      "step": 1448
+    },
+    {
+      "epoch": 39.16,
+      "learning_rate": 0.0003521875,
+      "loss": 0.2281,
+      "step": 1449
+    },
+    {
+      "epoch": 39.19,
+      "learning_rate": 0.00035203125,
+      "loss": 0.2499,
+      "step": 1450
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 0.000351875,
+      "loss": 0.2233,
+      "step": 1451
+    },
+    {
+      "epoch": 39.24,
+      "learning_rate": 0.00035171875000000003,
+      "loss": 0.2383,
+      "step": 1452
+    },
+    {
+      "epoch": 39.27,
+      "learning_rate": 0.0003515625,
+      "loss": 0.2647,
+      "step": 1453
+    },
+    {
+      "epoch": 39.3,
+      "learning_rate": 0.00035140625,
+      "loss": 0.2568,
+      "step": 1454
+    },
+    {
+      "epoch": 39.32,
+      "learning_rate": 0.00035125,
+      "loss": 0.2279,
+      "step": 1455
+    },
+    {
+      "epoch": 39.35,
+      "learning_rate": 0.00035109375,
+      "loss": 0.2379,
+      "step": 1456
+    },
+    {
+      "epoch": 39.38,
+      "learning_rate": 0.0003509375,
+      "loss": 0.2487,
+      "step": 1457
+    },
+    {
+      "epoch": 39.41,
+      "learning_rate": 0.00035078125,
+      "loss": 0.2143,
+      "step": 1458
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 0.000350625,
+      "loss": 0.2233,
+      "step": 1459
+    },
+    {
+      "epoch": 39.46,
+      "learning_rate": 0.00035046875,
+      "loss": 0.2578,
+      "step": 1460
+    },
+    {
+      "epoch": 39.49,
+      "learning_rate": 0.00035031250000000004,
+      "loss": 0.2269,
+      "step": 1461
+    },
+    {
+      "epoch": 39.51,
+      "learning_rate": 0.00035015625,
+      "loss": 0.2306,
+      "step": 1462
+    },
+    {
+      "epoch": 39.54,
+      "learning_rate": 0.00035,
+      "loss": 0.3121,
+      "step": 1463
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 0.00034984375,
+      "loss": 0.2481,
+      "step": 1464
+    },
+    {
+      "epoch": 39.59,
+      "learning_rate": 0.0003496875,
+      "loss": 0.2568,
+      "step": 1465
+    },
+    {
+      "epoch": 39.62,
+      "learning_rate": 0.00034953125,
+      "loss": 0.2518,
+      "step": 1466
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 0.000349375,
+      "loss": 0.2218,
+      "step": 1467
+    },
+    {
+      "epoch": 39.68,
+      "learning_rate": 0.00034921875000000003,
+      "loss": 0.1909,
+      "step": 1468
+    },
+    {
+      "epoch": 39.7,
+      "learning_rate": 0.0003490625,
+      "loss": 0.2841,
+      "step": 1469
+    },
+    {
+      "epoch": 39.73,
+      "learning_rate": 0.00034890625,
+      "loss": 0.2433,
+      "step": 1470
+    },
+    {
+      "epoch": 39.76,
+      "learning_rate": 0.00034875,
+      "loss": 0.2101,
+      "step": 1471
+    },
+    {
+      "epoch": 39.78,
+      "learning_rate": 0.00034859375,
+      "loss": 0.261,
+      "step": 1472
+    },
+    {
+      "epoch": 39.81,
+      "learning_rate": 0.0003484375,
+      "loss": 0.2531,
+      "step": 1473
+    },
+    {
+      "epoch": 39.84,
+      "learning_rate": 0.00034828125,
+      "loss": 0.2533,
+      "step": 1474
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 0.000348125,
+      "loss": 0.2106,
+      "step": 1475
+    },
+    {
+      "epoch": 39.89,
+      "learning_rate": 0.00034796875,
+      "loss": 0.215,
+      "step": 1476
+    },
+    {
+      "epoch": 39.92,
+      "learning_rate": 0.00034781250000000003,
+      "loss": 0.2192,
+      "step": 1477
+    },
+    {
+      "epoch": 39.95,
+      "learning_rate": 0.00034765625,
+      "loss": 0.2729,
+      "step": 1478
+    },
+    {
+      "epoch": 39.97,
+      "learning_rate": 0.0003475,
+      "loss": 0.2219,
+      "step": 1479
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 0.00034734375,
+      "loss": 0.1871,
+      "step": 1480
+    },
+    {
+      "epoch": 40.03,
+      "learning_rate": 0.0003471875,
+      "loss": 0.2624,
+      "step": 1481
+    },
+    {
+      "epoch": 40.05,
+      "learning_rate": 0.00034703125,
+      "loss": 0.2038,
+      "step": 1482
+    },
+    {
+      "epoch": 40.08,
+      "learning_rate": 0.000346875,
+      "loss": 0.213,
+      "step": 1483
+    },
+    {
+      "epoch": 40.11,
+      "learning_rate": 0.00034671875,
+      "loss": 0.2515,
+      "step": 1484
+    },
+    {
+      "epoch": 40.14,
+      "learning_rate": 0.00034656250000000003,
+      "loss": 0.1913,
+      "step": 1485
+    },
+    {
+      "epoch": 40.16,
+      "learning_rate": 0.00034640625000000003,
+      "loss": 0.2318,
+      "step": 1486
+    },
+    {
+      "epoch": 40.19,
+      "learning_rate": 0.00034625,
+      "loss": 0.1975,
+      "step": 1487
+    },
+    {
+      "epoch": 40.22,
+      "learning_rate": 0.00034609375,
+      "loss": 0.1874,
+      "step": 1488
+    },
+    {
+      "epoch": 40.24,
+      "learning_rate": 0.0003459375,
+      "loss": 0.2282,
+      "step": 1489
+    },
+    {
+      "epoch": 40.27,
+      "learning_rate": 0.00034578125,
+      "loss": 0.2318,
+      "step": 1490
+    },
+    {
+      "epoch": 40.3,
+      "learning_rate": 0.000345625,
+      "loss": 0.2211,
+      "step": 1491
+    },
+    {
+      "epoch": 40.32,
+      "learning_rate": 0.00034546875,
+      "loss": 0.2525,
+      "step": 1492
+    },
+    {
+      "epoch": 40.35,
+      "learning_rate": 0.0003453125,
+      "loss": 0.2073,
+      "step": 1493
+    },
+    {
+      "epoch": 40.38,
+      "learning_rate": 0.00034515625000000003,
+      "loss": 0.2245,
+      "step": 1494
+    },
+    {
+      "epoch": 40.41,
+      "learning_rate": 0.000345,
+      "loss": 0.2251,
+      "step": 1495
+    },
+    {
+      "epoch": 40.43,
+      "learning_rate": 0.00034484375,
+      "loss": 0.2145,
+      "step": 1496
+    },
+    {
+      "epoch": 40.46,
+      "learning_rate": 0.0003446875,
+      "loss": 0.2136,
+      "step": 1497
+    },
+    {
+      "epoch": 40.49,
+      "learning_rate": 0.00034453125,
+      "loss": 0.1771,
+      "step": 1498
+    },
+    {
+      "epoch": 40.51,
+      "learning_rate": 0.000344375,
+      "loss": 0.2187,
+      "step": 1499
+    },
+    {
+      "epoch": 40.54,
+      "learning_rate": 0.00034421875,
+      "loss": 0.2272,
+      "step": 1500
+    },
+    {
+      "epoch": 40.54,
+      "eval_cer": 0.08926670564776083,
+      "eval_loss": 0.3816392421722412,
+      "eval_runtime": 11.6877,
+      "eval_samples_per_second": 140.918,
+      "eval_steps_per_second": 2.225,
+      "eval_wer": 0.3595138392401185,
+      "step": 1500
+    },
+    {
+      "epoch": 40.57,
+      "learning_rate": 0.0003440625,
+      "loss": 0.2268,
+      "step": 1501
+    },
+    {
+      "epoch": 40.59,
+      "learning_rate": 0.00034390625000000003,
+      "loss": 0.2479,
+      "step": 1502
+    },
+    {
+      "epoch": 40.62,
+      "learning_rate": 0.00034375,
+      "loss": 0.2183,
+      "step": 1503
+    },
+    {
+      "epoch": 40.65,
+      "learning_rate": 0.00034359375,
+      "loss": 0.2006,
+      "step": 1504
+    },
+    {
+      "epoch": 40.68,
+      "learning_rate": 0.0003434375,
+      "loss": 0.2382,
+      "step": 1505
+    },
+    {
+      "epoch": 40.7,
+      "learning_rate": 0.00034328125,
+      "loss": 0.2559,
+      "step": 1506
+    },
+    {
+      "epoch": 40.73,
+      "learning_rate": 0.000343125,
+      "loss": 0.2184,
+      "step": 1507
+    },
+    {
+      "epoch": 40.76,
+      "learning_rate": 0.00034296875,
+      "loss": 0.2354,
+      "step": 1508
+    },
+    {
+      "epoch": 40.78,
+      "learning_rate": 0.0003428125,
+      "loss": 0.26,
+      "step": 1509
+    },
+    {
+      "epoch": 40.81,
+      "learning_rate": 0.00034265625,
+      "loss": 0.2571,
+      "step": 1510
+    },
+    {
+      "epoch": 40.84,
+      "learning_rate": 0.00034250000000000003,
+      "loss": 0.2411,
+      "step": 1511
+    },
+    {
+      "epoch": 40.86,
+      "learning_rate": 0.00034234375,
+      "loss": 0.2315,
+      "step": 1512
+    },
+    {
+      "epoch": 40.89,
+      "learning_rate": 0.0003421875,
+      "loss": 0.239,
+      "step": 1513
+    },
+    {
+      "epoch": 40.92,
+      "learning_rate": 0.00034203125,
+      "loss": 0.2418,
+      "step": 1514
+    },
+    {
+      "epoch": 40.95,
+      "learning_rate": 0.000341875,
+      "loss": 0.2118,
+      "step": 1515
+    },
+    {
+      "epoch": 40.97,
+      "learning_rate": 0.00034171875,
+      "loss": 0.1818,
+      "step": 1516
+    },
+    {
+      "epoch": 41.0,
+      "learning_rate": 0.0003415625,
+      "loss": 0.1319,
+      "step": 1517
+    },
+    {
+      "epoch": 41.03,
+      "learning_rate": 0.00034140625,
+      "loss": 0.2657,
+      "step": 1518
+    },
+    {
+      "epoch": 41.05,
+      "learning_rate": 0.00034125000000000003,
+      "loss": 0.2172,
+      "step": 1519
+    },
+    {
+      "epoch": 41.08,
+      "learning_rate": 0.00034109375,
+      "loss": 0.2475,
+      "step": 1520
+    },
+    {
+      "epoch": 41.11,
+      "learning_rate": 0.0003409375,
+      "loss": 0.2251,
+      "step": 1521
+    },
+    {
+      "epoch": 41.14,
+      "learning_rate": 0.00034078125,
+      "loss": 0.2644,
+      "step": 1522
+    },
+    {
+      "epoch": 41.16,
+      "learning_rate": 0.000340625,
+      "loss": 0.2274,
+      "step": 1523
+    },
+    {
+      "epoch": 41.19,
+      "learning_rate": 0.00034046875,
+      "loss": 0.1755,
+      "step": 1524
+    },
+    {
+      "epoch": 41.22,
+      "learning_rate": 0.0003403125,
+      "loss": 0.2139,
+      "step": 1525
+    },
+    {
+      "epoch": 41.24,
+      "learning_rate": 0.00034015625,
+      "loss": 0.2251,
+      "step": 1526
+    },
+    {
+      "epoch": 41.27,
+      "learning_rate": 0.00034,
+      "loss": 0.2556,
+      "step": 1527
+    },
+    {
+      "epoch": 41.3,
+      "learning_rate": 0.00033984375000000003,
+      "loss": 0.2221,
+      "step": 1528
+    },
+    {
+      "epoch": 41.32,
+      "learning_rate": 0.0003396875,
+      "loss": 0.244,
+      "step": 1529
+    },
+    {
+      "epoch": 41.35,
+      "learning_rate": 0.00033953125,
+      "loss": 0.2161,
+      "step": 1530
+    },
+    {
+      "epoch": 41.38,
+      "learning_rate": 0.000339375,
+      "loss": 0.2487,
+      "step": 1531
+    },
+    {
+      "epoch": 41.41,
+      "learning_rate": 0.00033921875,
+      "loss": 0.2506,
+      "step": 1532
+    },
+    {
+      "epoch": 41.43,
+      "learning_rate": 0.0003390625,
+      "loss": 0.2177,
+      "step": 1533
+    },
+    {
+      "epoch": 41.46,
+      "learning_rate": 0.00033890625,
+      "loss": 0.2677,
+      "step": 1534
+    },
+    {
+      "epoch": 41.49,
+      "learning_rate": 0.00033875,
+      "loss": 0.2512,
+      "step": 1535
+    },
+    {
+      "epoch": 41.51,
+      "learning_rate": 0.00033859375000000003,
+      "loss": 0.2436,
+      "step": 1536
+    },
+    {
+      "epoch": 41.54,
+      "learning_rate": 0.00033843750000000003,
+      "loss": 0.26,
+      "step": 1537
+    },
+    {
+      "epoch": 41.57,
+      "learning_rate": 0.00033828125,
+      "loss": 0.2362,
+      "step": 1538
+    },
+    {
+      "epoch": 41.59,
+      "learning_rate": 0.000338125,
+      "loss": 0.2299,
+      "step": 1539
+    },
+    {
+      "epoch": 41.62,
+      "learning_rate": 0.00033796875,
+      "loss": 0.2438,
+      "step": 1540
+    },
+    {
+      "epoch": 41.65,
+      "learning_rate": 0.0003378125,
+      "loss": 0.2274,
+      "step": 1541
+    },
+    {
+      "epoch": 41.68,
+      "learning_rate": 0.00033765625,
+      "loss": 0.2159,
+      "step": 1542
+    },
+    {
+      "epoch": 41.7,
+      "learning_rate": 0.0003375,
+      "loss": 0.2581,
+      "step": 1543
+    },
+    {
+      "epoch": 41.73,
+      "learning_rate": 0.00033734375,
+      "loss": 0.237,
+      "step": 1544
+    },
+    {
+      "epoch": 41.76,
+      "learning_rate": 0.0003371875,
+      "loss": 0.2297,
+      "step": 1545
+    },
+    {
+      "epoch": 41.78,
+      "learning_rate": 0.00033703125000000004,
+      "loss": 0.2494,
+      "step": 1546
+    },
+    {
+      "epoch": 41.81,
+      "learning_rate": 0.000336875,
+      "loss": 0.2067,
+      "step": 1547
+    },
+    {
+      "epoch": 41.84,
+      "learning_rate": 0.00033671875,
+      "loss": 0.2804,
+      "step": 1548
+    },
+    {
+      "epoch": 41.86,
+      "learning_rate": 0.0003365625,
+      "loss": 0.222,
+      "step": 1549
+    },
+    {
+      "epoch": 41.89,
+      "learning_rate": 0.00033640625,
+      "loss": 0.2387,
+      "step": 1550
+    },
+    {
+      "epoch": 41.92,
+      "learning_rate": 0.00033625,
+      "loss": 0.1936,
+      "step": 1551
+    },
+    {
+      "epoch": 41.95,
+      "learning_rate": 0.00033609375,
+      "loss": 0.2032,
+      "step": 1552
+    },
+    {
+      "epoch": 41.97,
+      "learning_rate": 0.00033593750000000003,
+      "loss": 0.2457,
+      "step": 1553
+    },
+    {
+      "epoch": 42.0,
+      "learning_rate": 0.00033578125,
+      "loss": 0.3371,
+      "step": 1554
+    },
+    {
+      "epoch": 42.03,
+      "learning_rate": 0.00033562500000000004,
+      "loss": 0.2235,
+      "step": 1555
+    },
+    {
+      "epoch": 42.05,
+      "learning_rate": 0.00033546875,
+      "loss": 0.218,
+      "step": 1556
+    },
+    {
+      "epoch": 42.08,
+      "learning_rate": 0.0003353125,
+      "loss": 0.2231,
+      "step": 1557
+    },
+    {
+      "epoch": 42.11,
+      "learning_rate": 0.00033515625,
+      "loss": 0.2524,
+      "step": 1558
+    },
+    {
+      "epoch": 42.14,
+      "learning_rate": 0.000335,
+      "loss": 0.1948,
+      "step": 1559
+    },
+    {
+      "epoch": 42.16,
+      "learning_rate": 0.00033484375,
+      "loss": 0.2572,
+      "step": 1560
+    },
+    {
+      "epoch": 42.19,
+      "learning_rate": 0.0003346875,
+      "loss": 0.2231,
+      "step": 1561
+    },
+    {
+      "epoch": 42.22,
+      "learning_rate": 0.00033453125000000003,
+      "loss": 0.2234,
+      "step": 1562
+    },
+    {
+      "epoch": 42.24,
+      "learning_rate": 0.000334375,
+      "loss": 0.216,
+      "step": 1563
+    },
+    {
+      "epoch": 42.27,
+      "learning_rate": 0.00033421875,
+      "loss": 0.203,
+      "step": 1564
+    },
+    {
+      "epoch": 42.3,
+      "learning_rate": 0.0003340625,
+      "loss": 0.209,
+      "step": 1565
+    },
+    {
+      "epoch": 42.32,
+      "learning_rate": 0.00033390625,
+      "loss": 0.1901,
+      "step": 1566
+    },
+    {
+      "epoch": 42.35,
+      "learning_rate": 0.00033375,
+      "loss": 0.2105,
+      "step": 1567
+    },
+    {
+      "epoch": 42.38,
+      "learning_rate": 0.00033359375,
+      "loss": 0.2414,
+      "step": 1568
+    },
+    {
+      "epoch": 42.41,
+      "learning_rate": 0.0003334375,
+      "loss": 0.2358,
+      "step": 1569
+    },
+    {
+      "epoch": 42.43,
+      "learning_rate": 0.00033328125,
+      "loss": 0.1979,
+      "step": 1570
+    },
+    {
+      "epoch": 42.46,
+      "learning_rate": 0.00033312500000000003,
+      "loss": 0.2017,
+      "step": 1571
+    },
+    {
+      "epoch": 42.49,
+      "learning_rate": 0.00033296875,
+      "loss": 0.2146,
+      "step": 1572
+    },
+    {
+      "epoch": 42.51,
+      "learning_rate": 0.0003328125,
+      "loss": 0.2258,
+      "step": 1573
+    },
+    {
+      "epoch": 42.54,
+      "learning_rate": 0.00033265625,
+      "loss": 0.2477,
+      "step": 1574
+    },
+    {
+      "epoch": 42.57,
+      "learning_rate": 0.0003325,
+      "loss": 0.2024,
+      "step": 1575
+    },
+    {
+      "epoch": 42.59,
+      "learning_rate": 0.00033234375,
+      "loss": 0.1935,
+      "step": 1576
+    },
+    {
+      "epoch": 42.62,
+      "learning_rate": 0.0003321875,
+      "loss": 0.2366,
+      "step": 1577
+    },
+    {
+      "epoch": 42.65,
+      "learning_rate": 0.00033203125,
+      "loss": 0.2312,
+      "step": 1578
+    },
+    {
+      "epoch": 42.68,
+      "learning_rate": 0.000331875,
+      "loss": 0.1853,
+      "step": 1579
+    },
+    {
+      "epoch": 42.7,
+      "learning_rate": 0.00033171875000000004,
+      "loss": 0.2663,
+      "step": 1580
+    },
+    {
+      "epoch": 42.73,
+      "learning_rate": 0.0003315625,
+      "loss": 0.21,
+      "step": 1581
+    },
+    {
+      "epoch": 42.76,
+      "learning_rate": 0.00033140625,
+      "loss": 0.2329,
+      "step": 1582
+    },
+    {
+      "epoch": 42.78,
+      "learning_rate": 0.00033125,
+      "loss": 0.2356,
+      "step": 1583
+    },
+    {
+      "epoch": 42.81,
+      "learning_rate": 0.00033109375,
+      "loss": 0.2176,
+      "step": 1584
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 0.0003309375,
+      "loss": 0.2552,
+      "step": 1585
+    },
+    {
+      "epoch": 42.86,
+      "learning_rate": 0.00033078125,
+      "loss": 0.2416,
+      "step": 1586
+    },
+    {
+      "epoch": 42.89,
+      "learning_rate": 0.000330625,
+      "loss": 0.206,
+      "step": 1587
+    },
+    {
+      "epoch": 42.92,
+      "learning_rate": 0.00033046875,
+      "loss": 0.225,
+      "step": 1588
+    },
+    {
+      "epoch": 42.95,
+      "learning_rate": 0.00033031250000000004,
+      "loss": 0.1926,
+      "step": 1589
+    },
+    {
+      "epoch": 42.97,
+      "learning_rate": 0.00033015625,
+      "loss": 0.2022,
+      "step": 1590
+    },
+    {
+      "epoch": 43.0,
+      "learning_rate": 0.00033,
+      "loss": 0.1418,
+      "step": 1591
+    },
+    {
+      "epoch": 43.03,
+      "learning_rate": 0.00032984375,
+      "loss": 0.2431,
+      "step": 1592
+    },
+    {
+      "epoch": 43.05,
+      "learning_rate": 0.0003296875,
+      "loss": 0.1884,
+      "step": 1593
+    },
+    {
+      "epoch": 43.08,
+      "learning_rate": 0.00032953125,
+      "loss": 0.2043,
+      "step": 1594
+    },
+    {
+      "epoch": 43.11,
+      "learning_rate": 0.00032937499999999997,
+      "loss": 0.2462,
+      "step": 1595
+    },
+    {
+      "epoch": 43.14,
+      "learning_rate": 0.00032921875000000003,
+      "loss": 0.2167,
+      "step": 1596
+    },
+    {
+      "epoch": 43.16,
+      "learning_rate": 0.0003290625,
+      "loss": 0.242,
+      "step": 1597
+    },
+    {
+      "epoch": 43.19,
+      "learning_rate": 0.00032890625000000004,
+      "loss": 0.2624,
+      "step": 1598
+    },
+    {
+      "epoch": 43.22,
+      "learning_rate": 0.00032875,
+      "loss": 0.2234,
+      "step": 1599
+    },
+    {
+      "epoch": 43.24,
+      "learning_rate": 0.00032859375,
+      "loss": 0.2234,
+      "step": 1600
+    },
+    {
+      "epoch": 43.27,
+      "learning_rate": 0.0003284375,
+      "loss": 0.2374,
+      "step": 1601
+    },
+    {
+      "epoch": 43.3,
+      "learning_rate": 0.00032828125,
+      "loss": 0.256,
+      "step": 1602
+    },
+    {
+      "epoch": 43.32,
+      "learning_rate": 0.000328125,
+      "loss": 0.2231,
+      "step": 1603
+    },
+    {
+      "epoch": 43.35,
+      "learning_rate": 0.00032796874999999997,
+      "loss": 0.2611,
+      "step": 1604
+    },
+    {
+      "epoch": 43.38,
+      "learning_rate": 0.00032781250000000003,
+      "loss": 0.2133,
+      "step": 1605
+    },
+    {
+      "epoch": 43.41,
+      "learning_rate": 0.00032765625,
+      "loss": 0.2448,
+      "step": 1606
+    },
+    {
+      "epoch": 43.43,
+      "learning_rate": 0.00032750000000000005,
+      "loss": 0.2378,
+      "step": 1607
+    },
+    {
+      "epoch": 43.46,
+      "learning_rate": 0.00032734375,
+      "loss": 0.2149,
+      "step": 1608
+    },
+    {
+      "epoch": 43.49,
+      "learning_rate": 0.0003271875,
+      "loss": 0.2157,
+      "step": 1609
+    },
+    {
+      "epoch": 43.51,
+      "learning_rate": 0.00032703125,
+      "loss": 0.188,
+      "step": 1610
+    },
+    {
+      "epoch": 43.54,
+      "learning_rate": 0.000326875,
+      "loss": 0.2564,
+      "step": 1611
+    },
+    {
+      "epoch": 43.57,
+      "learning_rate": 0.00032671875,
+      "loss": 0.2702,
+      "step": 1612
+    },
+    {
+      "epoch": 43.59,
+      "learning_rate": 0.0003265625,
+      "loss": 0.2148,
+      "step": 1613
+    },
+    {
+      "epoch": 43.62,
+      "learning_rate": 0.00032640625000000004,
+      "loss": 0.2029,
+      "step": 1614
+    },
+    {
+      "epoch": 43.65,
+      "learning_rate": 0.00032625,
+      "loss": 0.19,
+      "step": 1615
+    },
+    {
+      "epoch": 43.68,
+      "learning_rate": 0.00032609375000000005,
+      "loss": 0.2393,
+      "step": 1616
+    },
+    {
+      "epoch": 43.7,
+      "learning_rate": 0.0003259375,
+      "loss": 0.2412,
+      "step": 1617
+    },
+    {
+      "epoch": 43.73,
+      "learning_rate": 0.00032578125,
+      "loss": 0.2231,
+      "step": 1618
+    },
+    {
+      "epoch": 43.76,
+      "learning_rate": 0.000325625,
+      "loss": 0.1955,
+      "step": 1619
+    },
+    {
+      "epoch": 43.78,
+      "learning_rate": 0.00032546874999999997,
+      "loss": 0.2642,
+      "step": 1620
+    },
+    {
+      "epoch": 43.81,
+      "learning_rate": 0.0003253125,
+      "loss": 0.179,
+      "step": 1621
+    },
+    {
+      "epoch": 43.84,
+      "learning_rate": 0.00032515625,
+      "loss": 0.2057,
+      "step": 1622
+    },
+    {
+      "epoch": 43.86,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 0.218,
+      "step": 1623
+    },
+    {
+      "epoch": 43.89,
+      "learning_rate": 0.00032484375,
+      "loss": 0.1894,
+      "step": 1624
+    },
+    {
+      "epoch": 43.92,
+      "learning_rate": 0.0003246875,
+      "loss": 0.177,
+      "step": 1625
+    },
+    {
+      "epoch": 43.95,
+      "learning_rate": 0.00032453125,
+      "loss": 0.2282,
+      "step": 1626
+    },
+    {
+      "epoch": 43.97,
+      "learning_rate": 0.000324375,
+      "loss": 0.1838,
+      "step": 1627
+    },
+    {
+      "epoch": 44.0,
+      "learning_rate": 0.00032421875,
+      "loss": 0.1557,
+      "step": 1628
+    },
+    {
+      "epoch": 44.03,
+      "learning_rate": 0.00032406249999999997,
+      "loss": 0.2211,
+      "step": 1629
+    },
+    {
+      "epoch": 44.05,
+      "learning_rate": 0.00032390625000000003,
+      "loss": 0.2168,
+      "step": 1630
+    },
+    {
+      "epoch": 44.08,
+      "learning_rate": 0.00032375,
+      "loss": 0.2292,
+      "step": 1631
+    },
+    {
+      "epoch": 44.11,
+      "learning_rate": 0.00032359375000000004,
+      "loss": 0.2195,
+      "step": 1632
+    },
+    {
+      "epoch": 44.14,
+      "learning_rate": 0.0003234375,
+      "loss": 0.2535,
+      "step": 1633
+    },
+    {
+      "epoch": 44.16,
+      "learning_rate": 0.00032328125,
+      "loss": 0.2519,
+      "step": 1634
+    },
+    {
+      "epoch": 44.19,
+      "learning_rate": 0.000323125,
+      "loss": 0.2358,
+      "step": 1635
+    },
+    {
+      "epoch": 44.22,
+      "learning_rate": 0.00032296875,
+      "loss": 0.2217,
+      "step": 1636
+    },
+    {
+      "epoch": 44.24,
+      "learning_rate": 0.0003228125,
+      "loss": 0.2156,
+      "step": 1637
+    },
+    {
+      "epoch": 44.27,
+      "learning_rate": 0.00032265624999999997,
+      "loss": 0.2208,
+      "step": 1638
+    },
+    {
+      "epoch": 44.3,
+      "learning_rate": 0.00032250000000000003,
+      "loss": 0.1936,
+      "step": 1639
+    },
+    {
+      "epoch": 44.32,
+      "learning_rate": 0.00032234375,
+      "loss": 0.1963,
+      "step": 1640
+    },
+    {
+      "epoch": 44.35,
+      "learning_rate": 0.00032218750000000005,
+      "loss": 0.2381,
+      "step": 1641
+    },
+    {
+      "epoch": 44.38,
+      "learning_rate": 0.00032203125,
+      "loss": 0.2458,
+      "step": 1642
+    },
+    {
+      "epoch": 44.41,
+      "learning_rate": 0.000321875,
+      "loss": 0.2422,
+      "step": 1643
+    },
+    {
+      "epoch": 44.43,
+      "learning_rate": 0.00032171875,
+      "loss": 0.1991,
+      "step": 1644
+    },
+    {
+      "epoch": 44.46,
+      "learning_rate": 0.00032156249999999996,
+      "loss": 0.2379,
+      "step": 1645
+    },
+    {
+      "epoch": 44.49,
+      "learning_rate": 0.00032140625,
+      "loss": 0.2336,
+      "step": 1646
+    },
+    {
+      "epoch": 44.51,
+      "learning_rate": 0.00032125,
+      "loss": 0.2491,
+      "step": 1647
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 0.00032109375000000004,
+      "loss": 0.226,
+      "step": 1648
+    },
+    {
+      "epoch": 44.57,
+      "learning_rate": 0.0003209375,
+      "loss": 0.2105,
+      "step": 1649
+    },
+    {
+      "epoch": 44.59,
+      "learning_rate": 0.00032078125000000005,
+      "loss": 0.214,
+      "step": 1650
+    },
+    {
+      "epoch": 44.62,
+      "learning_rate": 0.000320625,
+      "loss": 0.1865,
+      "step": 1651
+    },
+    {
+      "epoch": 44.65,
+      "learning_rate": 0.00032046875,
+      "loss": 0.2409,
+      "step": 1652
+    },
+    {
+      "epoch": 44.68,
+      "learning_rate": 0.0003203125,
+      "loss": 0.2215,
+      "step": 1653
+    },
+    {
+      "epoch": 44.7,
+      "learning_rate": 0.00032015624999999997,
+      "loss": 0.2394,
+      "step": 1654
+    },
+    {
+      "epoch": 44.73,
+      "learning_rate": 0.00032,
+      "loss": 0.2589,
+      "step": 1655
+    },
+    {
+      "epoch": 44.76,
+      "learning_rate": 0.00031984375,
+      "loss": 0.2064,
+      "step": 1656
+    },
+    {
+      "epoch": 44.78,
+      "learning_rate": 0.00031968750000000004,
+      "loss": 0.2602,
+      "step": 1657
+    },
+    {
+      "epoch": 44.81,
+      "learning_rate": 0.00031953125,
+      "loss": 0.2294,
+      "step": 1658
+    },
+    {
+      "epoch": 44.84,
+      "learning_rate": 0.00031937500000000005,
+      "loss": 0.1784,
+      "step": 1659
+    },
+    {
+      "epoch": 44.86,
+      "learning_rate": 0.00031921875,
+      "loss": 0.2219,
+      "step": 1660
+    },
+    {
+      "epoch": 44.89,
+      "learning_rate": 0.0003190625,
+      "loss": 0.234,
+      "step": 1661
+    },
+    {
+      "epoch": 44.92,
+      "learning_rate": 0.00031890625,
+      "loss": 0.2576,
+      "step": 1662
+    },
+    {
+      "epoch": 44.95,
+      "learning_rate": 0.00031874999999999997,
+      "loss": 0.2865,
+      "step": 1663
+    },
+    {
+      "epoch": 44.97,
+      "learning_rate": 0.00031859375000000003,
+      "loss": 0.2315,
+      "step": 1664
+    },
+    {
+      "epoch": 45.0,
+      "learning_rate": 0.0003184375,
+      "loss": 0.1582,
+      "step": 1665
+    },
+    {
+      "epoch": 45.03,
+      "learning_rate": 0.00031828125000000004,
+      "loss": 0.2281,
+      "step": 1666
+    },
+    {
+      "epoch": 45.05,
+      "learning_rate": 0.000318125,
+      "loss": 0.2091,
+      "step": 1667
+    },
+    {
+      "epoch": 45.08,
+      "learning_rate": 0.00031796875000000005,
+      "loss": 0.189,
+      "step": 1668
+    },
+    {
+      "epoch": 45.11,
+      "learning_rate": 0.0003178125,
+      "loss": 0.2141,
+      "step": 1669
+    },
+    {
+      "epoch": 45.14,
+      "learning_rate": 0.00031765624999999996,
+      "loss": 0.204,
+      "step": 1670
+    },
+    {
+      "epoch": 45.16,
+      "learning_rate": 0.0003175,
+      "loss": 0.1425,
+      "step": 1671
+    },
+    {
+      "epoch": 45.19,
+      "learning_rate": 0.00031734374999999997,
+      "loss": 0.1802,
+      "step": 1672
+    },
+    {
+      "epoch": 45.22,
+      "learning_rate": 0.00031718750000000003,
+      "loss": 0.2464,
+      "step": 1673
+    },
+    {
+      "epoch": 45.24,
+      "learning_rate": 0.00031703125,
+      "loss": 0.2442,
+      "step": 1674
+    },
+    {
+      "epoch": 45.27,
+      "learning_rate": 0.00031687500000000005,
+      "loss": 0.2576,
+      "step": 1675
+    },
+    {
+      "epoch": 45.3,
+      "learning_rate": 0.00031671875,
+      "loss": 0.2396,
+      "step": 1676
+    },
+    {
+      "epoch": 45.32,
+      "learning_rate": 0.00031656250000000006,
+      "loss": 0.2161,
+      "step": 1677
+    },
+    {
+      "epoch": 45.35,
+      "learning_rate": 0.00031640625,
+      "loss": 0.2177,
+      "step": 1678
+    },
+    {
+      "epoch": 45.38,
+      "learning_rate": 0.00031624999999999996,
+      "loss": 0.2172,
+      "step": 1679
+    },
+    {
+      "epoch": 45.41,
+      "learning_rate": 0.00031609375,
+      "loss": 0.1942,
+      "step": 1680
+    },
+    {
+      "epoch": 45.43,
+      "learning_rate": 0.0003159375,
+      "loss": 0.2096,
+      "step": 1681
+    },
+    {
+      "epoch": 45.46,
+      "learning_rate": 0.00031578125000000004,
+      "loss": 0.2238,
+      "step": 1682
+    },
+    {
+      "epoch": 45.49,
+      "learning_rate": 0.000315625,
+      "loss": 0.1876,
+      "step": 1683
+    },
+    {
+      "epoch": 45.51,
+      "learning_rate": 0.00031546875000000005,
+      "loss": 0.2014,
+      "step": 1684
+    },
+    {
+      "epoch": 45.54,
+      "learning_rate": 0.0003153125,
+      "loss": 0.2527,
+      "step": 1685
+    },
+    {
+      "epoch": 45.57,
+      "learning_rate": 0.00031515625,
+      "loss": 0.2339,
+      "step": 1686
+    },
+    {
+      "epoch": 45.59,
+      "learning_rate": 0.000315,
+      "loss": 0.1752,
+      "step": 1687
+    },
+    {
+      "epoch": 45.62,
+      "learning_rate": 0.00031484374999999997,
+      "loss": 0.2059,
+      "step": 1688
+    },
+    {
+      "epoch": 45.65,
+      "learning_rate": 0.0003146875,
+      "loss": 0.2261,
+      "step": 1689
+    },
+    {
+      "epoch": 45.68,
+      "learning_rate": 0.00031453125,
+      "loss": 0.2197,
+      "step": 1690
+    },
+    {
+      "epoch": 45.7,
+      "learning_rate": 0.00031437500000000004,
+      "loss": 0.2133,
+      "step": 1691
+    },
+    {
+      "epoch": 45.73,
+      "learning_rate": 0.00031421875,
+      "loss": 0.205,
+      "step": 1692
+    },
+    {
+      "epoch": 45.76,
+      "learning_rate": 0.00031406250000000005,
+      "loss": 0.1839,
+      "step": 1693
+    },
+    {
+      "epoch": 45.78,
+      "learning_rate": 0.00031390625,
+      "loss": 0.22,
+      "step": 1694
+    },
+    {
+      "epoch": 45.81,
+      "learning_rate": 0.00031374999999999996,
+      "loss": 0.2433,
+      "step": 1695
+    },
+    {
+      "epoch": 45.84,
+      "learning_rate": 0.00031359375,
+      "loss": 0.2342,
+      "step": 1696
+    },
+    {
+      "epoch": 45.86,
+      "learning_rate": 0.00031343749999999997,
+      "loss": 0.2026,
+      "step": 1697
+    },
+    {
+      "epoch": 45.89,
+      "learning_rate": 0.00031328125000000003,
+      "loss": 0.2192,
+      "step": 1698
+    },
+    {
+      "epoch": 45.92,
+      "learning_rate": 0.000313125,
+      "loss": 0.2001,
+      "step": 1699
+    },
+    {
+      "epoch": 45.95,
+      "learning_rate": 0.00031296875000000004,
+      "loss": 0.182,
+      "step": 1700
+    },
+    {
+      "epoch": 45.97,
+      "learning_rate": 0.0003128125,
+      "loss": 0.1877,
+      "step": 1701
+    },
+    {
+      "epoch": 46.0,
+      "learning_rate": 0.00031265625000000005,
+      "loss": 0.176,
+      "step": 1702
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 0.0003125,
+      "loss": 0.2381,
+      "step": 1703
+    },
+    {
+      "epoch": 46.05,
+      "learning_rate": 0.00031234374999999996,
+      "loss": 0.1941,
+      "step": 1704
+    },
+    {
+      "epoch": 46.08,
+      "learning_rate": 0.0003121875,
+      "loss": 0.2221,
+      "step": 1705
+    },
+    {
+      "epoch": 46.11,
+      "learning_rate": 0.00031203124999999997,
+      "loss": 0.2197,
+      "step": 1706
+    },
+    {
+      "epoch": 46.14,
+      "learning_rate": 0.00031187500000000003,
+      "loss": 0.2075,
+      "step": 1707
+    },
+    {
+      "epoch": 46.16,
+      "learning_rate": 0.00031171875,
+      "loss": 0.1721,
+      "step": 1708
+    },
+    {
+      "epoch": 46.19,
+      "learning_rate": 0.00031156250000000004,
+      "loss": 0.27,
+      "step": 1709
+    },
+    {
+      "epoch": 46.22,
+      "learning_rate": 0.00031140625,
+      "loss": 0.2283,
+      "step": 1710
+    },
+    {
+      "epoch": 46.24,
+      "learning_rate": 0.00031125000000000006,
+      "loss": 0.2003,
+      "step": 1711
+    },
+    {
+      "epoch": 46.27,
+      "learning_rate": 0.00031109375,
+      "loss": 0.2267,
+      "step": 1712
+    },
+    {
+      "epoch": 46.3,
+      "learning_rate": 0.00031093749999999996,
+      "loss": 0.2369,
+      "step": 1713
+    },
+    {
+      "epoch": 46.32,
+      "learning_rate": 0.00031078125,
+      "loss": 0.2032,
+      "step": 1714
+    },
+    {
+      "epoch": 46.35,
+      "learning_rate": 0.000310625,
+      "loss": 0.1933,
+      "step": 1715
+    },
+    {
+      "epoch": 46.38,
+      "learning_rate": 0.00031046875000000004,
+      "loss": 0.2209,
+      "step": 1716
+    },
+    {
+      "epoch": 46.41,
+      "learning_rate": 0.0003103125,
+      "loss": 0.2124,
+      "step": 1717
+    },
+    {
+      "epoch": 46.43,
+      "learning_rate": 0.00031015625000000005,
+      "loss": 0.2262,
+      "step": 1718
+    },
+    {
+      "epoch": 46.46,
+      "learning_rate": 0.00031,
+      "loss": 0.214,
+      "step": 1719
+    },
+    {
+      "epoch": 46.49,
+      "learning_rate": 0.00030984375,
+      "loss": 0.2071,
+      "step": 1720
+    },
+    {
+      "epoch": 46.51,
+      "learning_rate": 0.0003096875,
+      "loss": 0.18,
+      "step": 1721
+    },
+    {
+      "epoch": 46.54,
+      "learning_rate": 0.00030953124999999996,
+      "loss": 0.2356,
+      "step": 1722
+    },
+    {
+      "epoch": 46.57,
+      "learning_rate": 0.000309375,
+      "loss": 0.2402,
+      "step": 1723
+    },
+    {
+      "epoch": 46.59,
+      "learning_rate": 0.00030921875,
+      "loss": 0.2258,
+      "step": 1724
+    },
+    {
+      "epoch": 46.62,
+      "learning_rate": 0.00030906250000000004,
+      "loss": 0.2786,
+      "step": 1725
+    },
+    {
+      "epoch": 46.65,
+      "learning_rate": 0.00030890625,
+      "loss": 0.1946,
+      "step": 1726
+    },
+    {
+      "epoch": 46.68,
+      "learning_rate": 0.00030875000000000005,
+      "loss": 0.1948,
+      "step": 1727
+    },
+    {
+      "epoch": 46.7,
+      "learning_rate": 0.00030859375,
+      "loss": 0.227,
+      "step": 1728
+    },
+    {
+      "epoch": 46.73,
+      "learning_rate": 0.00030843749999999996,
+      "loss": 0.2131,
+      "step": 1729
+    },
+    {
+      "epoch": 46.76,
+      "learning_rate": 0.00030828125,
+      "loss": 0.1949,
+      "step": 1730
+    },
+    {
+      "epoch": 46.78,
+      "learning_rate": 0.00030812499999999997,
+      "loss": 0.2207,
+      "step": 1731
+    },
+    {
+      "epoch": 46.81,
+      "learning_rate": 0.00030796875000000003,
+      "loss": 0.2171,
+      "step": 1732
+    },
+    {
+      "epoch": 46.84,
+      "learning_rate": 0.0003078125,
+      "loss": 0.2042,
+      "step": 1733
+    },
+    {
+      "epoch": 46.86,
+      "learning_rate": 0.00030765625000000004,
+      "loss": 0.2161,
+      "step": 1734
+    },
+    {
+      "epoch": 46.89,
+      "learning_rate": 0.0003075,
+      "loss": 0.2159,
+      "step": 1735
+    },
+    {
+      "epoch": 46.92,
+      "learning_rate": 0.00030734375000000005,
+      "loss": 0.2289,
+      "step": 1736
+    },
+    {
+      "epoch": 46.95,
+      "learning_rate": 0.0003071875,
+      "loss": 0.1966,
+      "step": 1737
+    },
+    {
+      "epoch": 46.97,
+      "learning_rate": 0.00030703124999999996,
+      "loss": 0.1722,
+      "step": 1738
+    },
+    {
+      "epoch": 47.0,
+      "learning_rate": 0.000306875,
+      "loss": 0.2589,
+      "step": 1739
+    },
+    {
+      "epoch": 47.03,
+      "learning_rate": 0.00030671874999999997,
+      "loss": 0.2107,
+      "step": 1740
+    },
+    {
+      "epoch": 47.05,
+      "learning_rate": 0.00030656250000000003,
+      "loss": 0.2353,
+      "step": 1741
+    },
+    {
+      "epoch": 47.08,
+      "learning_rate": 0.00030640625,
+      "loss": 0.1972,
+      "step": 1742
+    },
+    {
+      "epoch": 47.11,
+      "learning_rate": 0.00030625000000000004,
+      "loss": 0.2501,
+      "step": 1743
+    },
+    {
+      "epoch": 47.14,
+      "learning_rate": 0.00030609375,
+      "loss": 0.2483,
+      "step": 1744
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 0.0003059375,
+      "loss": 0.185,
+      "step": 1745
+    },
+    {
+      "epoch": 47.19,
+      "learning_rate": 0.00030578125,
+      "loss": 0.2362,
+      "step": 1746
+    },
+    {
+      "epoch": 47.22,
+      "learning_rate": 0.00030562499999999996,
+      "loss": 0.1945,
+      "step": 1747
+    },
+    {
+      "epoch": 47.24,
+      "learning_rate": 0.00030546875,
+      "loss": 0.1911,
+      "step": 1748
+    },
+    {
+      "epoch": 47.27,
+      "learning_rate": 0.0003053125,
+      "loss": 0.2144,
+      "step": 1749
+    },
+    {
+      "epoch": 47.3,
+      "learning_rate": 0.00030515625000000003,
+      "loss": 0.1632,
+      "step": 1750
+    },
+    {
+      "epoch": 47.32,
+      "learning_rate": 0.000305,
+      "loss": 0.1924,
+      "step": 1751
+    },
+    {
+      "epoch": 47.35,
+      "learning_rate": 0.00030484375000000005,
+      "loss": 0.1897,
+      "step": 1752
+    },
+    {
+      "epoch": 47.38,
+      "learning_rate": 0.0003046875,
+      "loss": 0.2078,
+      "step": 1753
+    },
+    {
+      "epoch": 47.41,
+      "learning_rate": 0.00030453125,
+      "loss": 0.18,
+      "step": 1754
+    },
+    {
+      "epoch": 47.43,
+      "learning_rate": 0.000304375,
+      "loss": 0.1977,
+      "step": 1755
+    },
+    {
+      "epoch": 47.46,
+      "learning_rate": 0.00030421874999999996,
+      "loss": 0.1859,
+      "step": 1756
+    },
+    {
+      "epoch": 47.49,
+      "learning_rate": 0.0003040625,
+      "loss": 0.2018,
+      "step": 1757
+    },
+    {
+      "epoch": 47.51,
+      "learning_rate": 0.00030390625,
+      "loss": 0.2056,
+      "step": 1758
+    },
+    {
+      "epoch": 47.54,
+      "learning_rate": 0.00030375000000000004,
+      "loss": 0.2478,
+      "step": 1759
+    },
+    {
+      "epoch": 47.57,
+      "learning_rate": 0.00030359375,
+      "loss": 0.1877,
+      "step": 1760
+    },
+    {
+      "epoch": 47.59,
+      "learning_rate": 0.00030343750000000005,
+      "loss": 0.224,
+      "step": 1761
+    },
+    {
+      "epoch": 47.62,
+      "learning_rate": 0.00030328125,
+      "loss": 0.1906,
+      "step": 1762
+    },
+    {
+      "epoch": 47.65,
+      "learning_rate": 0.000303125,
+      "loss": 0.1807,
+      "step": 1763
+    },
+    {
+      "epoch": 47.68,
+      "learning_rate": 0.00030296875,
+      "loss": 0.1464,
+      "step": 1764
+    },
+    {
+      "epoch": 47.7,
+      "learning_rate": 0.00030281249999999997,
+      "loss": 0.227,
+      "step": 1765
+    },
+    {
+      "epoch": 47.73,
+      "learning_rate": 0.00030265625000000003,
+      "loss": 0.222,
+      "step": 1766
+    },
+    {
+      "epoch": 47.76,
+      "learning_rate": 0.0003025,
+      "loss": 0.183,
+      "step": 1767
+    },
+    {
+      "epoch": 47.78,
+      "learning_rate": 0.00030234375000000004,
+      "loss": 0.1827,
+      "step": 1768
+    },
+    {
+      "epoch": 47.81,
+      "learning_rate": 0.0003021875,
+      "loss": 0.2414,
+      "step": 1769
+    },
+    {
+      "epoch": 47.84,
+      "learning_rate": 0.00030203125,
+      "loss": 0.2161,
+      "step": 1770
+    },
+    {
+      "epoch": 47.86,
+      "learning_rate": 0.000301875,
+      "loss": 0.214,
+      "step": 1771
+    },
+    {
+      "epoch": 47.89,
+      "learning_rate": 0.00030171875,
+      "loss": 0.1935,
+      "step": 1772
+    },
+    {
+      "epoch": 47.92,
+      "learning_rate": 0.0003015625,
+      "loss": 0.1803,
+      "step": 1773
+    },
+    {
+      "epoch": 47.95,
+      "learning_rate": 0.00030140624999999997,
+      "loss": 0.1986,
+      "step": 1774
+    },
+    {
+      "epoch": 47.97,
+      "learning_rate": 0.00030125000000000003,
+      "loss": 0.1882,
+      "step": 1775
+    },
+    {
+      "epoch": 48.0,
+      "learning_rate": 0.00030109375,
+      "loss": 0.2363,
+      "step": 1776
+    },
+    {
+      "epoch": 48.03,
+      "learning_rate": 0.00030093750000000004,
+      "loss": 0.2459,
+      "step": 1777
+    },
+    {
+      "epoch": 48.05,
+      "learning_rate": 0.00030078125,
+      "loss": 0.2009,
+      "step": 1778
+    },
+    {
+      "epoch": 48.08,
+      "learning_rate": 0.000300625,
+      "loss": 0.1981,
+      "step": 1779
+    },
+    {
+      "epoch": 48.11,
+      "learning_rate": 0.00030046875,
+      "loss": 0.1735,
+      "step": 1780
+    },
+    {
+      "epoch": 48.14,
+      "learning_rate": 0.0003003125,
+      "loss": 0.2067,
+      "step": 1781
+    },
+    {
+      "epoch": 48.16,
+      "learning_rate": 0.00030015625,
+      "loss": 0.1975,
+      "step": 1782
+    },
+    {
+      "epoch": 48.19,
+      "learning_rate": 0.0003,
+      "loss": 0.2019,
+      "step": 1783
+    },
+    {
+      "epoch": 48.22,
+      "learning_rate": 0.00029984375000000003,
+      "loss": 0.221,
+      "step": 1784
+    },
+    {
+      "epoch": 48.24,
+      "learning_rate": 0.0002996875,
+      "loss": 0.2019,
+      "step": 1785
+    },
+    {
+      "epoch": 48.27,
+      "learning_rate": 0.00029953125000000005,
+      "loss": 0.201,
+      "step": 1786
+    },
+    {
+      "epoch": 48.3,
+      "learning_rate": 0.000299375,
+      "loss": 0.213,
+      "step": 1787
+    },
+    {
+      "epoch": 48.32,
+      "learning_rate": 0.00029921875,
+      "loss": 0.1627,
+      "step": 1788
+    },
+    {
+      "epoch": 48.35,
+      "learning_rate": 0.0002990625,
+      "loss": 0.1872,
+      "step": 1789
+    },
+    {
+      "epoch": 48.38,
+      "learning_rate": 0.00029890624999999996,
+      "loss": 0.2078,
+      "step": 1790
+    },
+    {
+      "epoch": 48.41,
+      "learning_rate": 0.00029875,
+      "loss": 0.2557,
+      "step": 1791
+    },
+    {
+      "epoch": 48.43,
+      "learning_rate": 0.00029859375,
+      "loss": 0.2098,
+      "step": 1792
+    },
+    {
+      "epoch": 48.46,
+      "learning_rate": 0.00029843750000000004,
+      "loss": 0.2339,
+      "step": 1793
+    },
+    {
+      "epoch": 48.49,
+      "learning_rate": 0.00029828125,
+      "loss": 0.1959,
+      "step": 1794
+    },
+    {
+      "epoch": 48.51,
+      "learning_rate": 0.000298125,
+      "loss": 0.2412,
+      "step": 1795
+    },
+    {
+      "epoch": 48.54,
+      "learning_rate": 0.00029796875,
+      "loss": 0.212,
+      "step": 1796
+    },
+    {
+      "epoch": 48.57,
+      "learning_rate": 0.0002978125,
+      "loss": 0.2165,
+      "step": 1797
+    },
+    {
+      "epoch": 48.59,
+      "learning_rate": 0.00029765625,
+      "loss": 0.2367,
+      "step": 1798
+    },
+    {
+      "epoch": 48.62,
+      "learning_rate": 0.00029749999999999997,
+      "loss": 0.223,
+      "step": 1799
+    },
+    {
+      "epoch": 48.65,
+      "learning_rate": 0.00029734375000000003,
+      "loss": 0.1818,
+      "step": 1800
+    },
+    {
+      "epoch": 48.68,
+      "learning_rate": 0.0002971875,
+      "loss": 0.1858,
+      "step": 1801
+    },
+    {
+      "epoch": 48.7,
+      "learning_rate": 0.00029703125000000004,
+      "loss": 0.2155,
+      "step": 1802
+    },
+    {
+      "epoch": 48.73,
+      "learning_rate": 0.000296875,
+      "loss": 0.2136,
+      "step": 1803
+    },
+    {
+      "epoch": 48.76,
+      "learning_rate": 0.00029671875,
+      "loss": 0.2027,
+      "step": 1804
+    },
+    {
+      "epoch": 48.78,
+      "learning_rate": 0.0002965625,
+      "loss": 0.2021,
+      "step": 1805
+    },
+    {
+      "epoch": 48.81,
+      "learning_rate": 0.00029640625,
+      "loss": 0.2347,
+      "step": 1806
+    },
+    {
+      "epoch": 48.84,
+      "learning_rate": 0.00029625,
+      "loss": 0.2174,
+      "step": 1807
+    },
+    {
+      "epoch": 48.86,
+      "learning_rate": 0.00029609374999999997,
+      "loss": 0.228,
+      "step": 1808
+    },
+    {
+      "epoch": 48.89,
+      "learning_rate": 0.00029593750000000003,
+      "loss": 0.2109,
+      "step": 1809
+    },
+    {
+      "epoch": 48.92,
+      "learning_rate": 0.00029578125,
+      "loss": 0.1917,
+      "step": 1810
+    },
+    {
+      "epoch": 48.95,
+      "learning_rate": 0.00029562500000000004,
+      "loss": 0.2414,
+      "step": 1811
+    },
+    {
+      "epoch": 48.97,
+      "learning_rate": 0.00029546875,
+      "loss": 0.2037,
+      "step": 1812
+    },
+    {
+      "epoch": 49.0,
+      "learning_rate": 0.0002953125,
+      "loss": 0.1,
+      "step": 1813
+    },
+    {
+      "epoch": 49.03,
+      "learning_rate": 0.00029515625,
+      "loss": 0.2182,
+      "step": 1814
+    },
+    {
+      "epoch": 49.05,
+      "learning_rate": 0.000295,
+      "loss": 0.2309,
+      "step": 1815
+    },
+    {
+      "epoch": 49.08,
+      "learning_rate": 0.00029484375,
+      "loss": 0.2168,
+      "step": 1816
+    },
+    {
+      "epoch": 49.11,
+      "learning_rate": 0.0002946875,
+      "loss": 0.2159,
+      "step": 1817
+    },
+    {
+      "epoch": 49.14,
+      "learning_rate": 0.00029453125000000003,
+      "loss": 0.2469,
+      "step": 1818
+    },
+    {
+      "epoch": 49.16,
+      "learning_rate": 0.000294375,
+      "loss": 0.2069,
+      "step": 1819
+    },
+    {
+      "epoch": 49.19,
+      "learning_rate": 0.00029421875,
+      "loss": 0.2265,
+      "step": 1820
+    },
+    {
+      "epoch": 49.22,
+      "learning_rate": 0.0002940625,
+      "loss": 0.2442,
+      "step": 1821
+    },
+    {
+      "epoch": 49.24,
+      "learning_rate": 0.00029390625,
+      "loss": 0.1888,
+      "step": 1822
+    },
+    {
+      "epoch": 49.27,
+      "learning_rate": 0.00029375,
+      "loss": 0.1651,
+      "step": 1823
+    },
+    {
+      "epoch": 49.3,
+      "learning_rate": 0.00029359375,
+      "loss": 0.1502,
+      "step": 1824
+    },
+    {
+      "epoch": 49.32,
+      "learning_rate": 0.0002934375,
+      "loss": 0.1995,
+      "step": 1825
+    },
+    {
+      "epoch": 49.35,
+      "learning_rate": 0.00029328125,
+      "loss": 0.254,
+      "step": 1826
+    },
+    {
+      "epoch": 49.38,
+      "learning_rate": 0.00029312500000000004,
+      "loss": 0.2538,
+      "step": 1827
+    },
+    {
+      "epoch": 49.41,
+      "learning_rate": 0.00029296875,
+      "loss": 0.1797,
+      "step": 1828
+    },
+    {
+      "epoch": 49.43,
+      "learning_rate": 0.0002928125,
+      "loss": 0.2179,
+      "step": 1829
+    },
+    {
+      "epoch": 49.46,
+      "learning_rate": 0.00029265625,
+      "loss": 0.2242,
+      "step": 1830
+    },
+    {
+      "epoch": 49.49,
+      "learning_rate": 0.0002925,
+      "loss": 0.1851,
+      "step": 1831
+    },
+    {
+      "epoch": 49.51,
+      "learning_rate": 0.00029234375,
+      "loss": 0.2325,
+      "step": 1832
+    },
+    {
+      "epoch": 49.54,
+      "learning_rate": 0.0002921875,
+      "loss": 0.2359,
+      "step": 1833
+    },
+    {
+      "epoch": 49.57,
+      "learning_rate": 0.00029203125000000003,
+      "loss": 0.2289,
+      "step": 1834
+    },
+    {
+      "epoch": 49.59,
+      "learning_rate": 0.000291875,
+      "loss": 0.22,
+      "step": 1835
+    },
+    {
+      "epoch": 49.62,
+      "learning_rate": 0.00029171875000000004,
+      "loss": 0.2589,
+      "step": 1836
+    },
+    {
+      "epoch": 49.65,
+      "learning_rate": 0.0002915625,
+      "loss": 0.1997,
+      "step": 1837
+    },
+    {
+      "epoch": 49.68,
+      "learning_rate": 0.00029140625,
+      "loss": 0.1926,
+      "step": 1838
+    },
+    {
+      "epoch": 49.7,
+      "learning_rate": 0.00029125,
+      "loss": 0.2629,
+      "step": 1839
+    },
+    {
+      "epoch": 49.73,
+      "learning_rate": 0.00029109375,
+      "loss": 0.2163,
+      "step": 1840
+    },
+    {
+      "epoch": 49.76,
+      "learning_rate": 0.0002909375,
+      "loss": 0.219,
+      "step": 1841
+    },
+    {
+      "epoch": 49.78,
+      "learning_rate": 0.00029078125,
+      "loss": 0.2218,
+      "step": 1842
+    },
+    {
+      "epoch": 49.81,
+      "learning_rate": 0.00029062500000000003,
+      "loss": 0.2206,
+      "step": 1843
+    },
+    {
+      "epoch": 49.84,
+      "learning_rate": 0.00029046875,
+      "loss": 0.1939,
+      "step": 1844
+    },
+    {
+      "epoch": 49.86,
+      "learning_rate": 0.0002903125,
+      "loss": 0.194,
+      "step": 1845
+    },
+    {
+      "epoch": 49.89,
+      "learning_rate": 0.00029015625,
+      "loss": 0.22,
+      "step": 1846
+    },
+    {
+      "epoch": 49.92,
+      "learning_rate": 0.00029,
+      "loss": 0.2454,
+      "step": 1847
+    },
+    {
+      "epoch": 49.95,
+      "learning_rate": 0.00028984375,
+      "loss": 0.2543,
+      "step": 1848
+    },
+    {
+      "epoch": 49.97,
+      "learning_rate": 0.0002896875,
+      "loss": 0.1951,
+      "step": 1849
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 0.00028953125,
+      "loss": 0.1142,
+      "step": 1850
+    },
+    {
+      "epoch": 50.03,
+      "learning_rate": 0.000289375,
+      "loss": 0.215,
+      "step": 1851
+    },
+    {
+      "epoch": 50.05,
+      "learning_rate": 0.00028921875000000003,
+      "loss": 0.2273,
+      "step": 1852
+    },
+    {
+      "epoch": 50.08,
+      "learning_rate": 0.0002890625,
+      "loss": 0.2073,
+      "step": 1853
+    },
+    {
+      "epoch": 50.11,
+      "learning_rate": 0.00028890625,
+      "loss": 0.2393,
+      "step": 1854
+    },
+    {
+      "epoch": 50.14,
+      "learning_rate": 0.00028875,
+      "loss": 0.1703,
+      "step": 1855
+    },
+    {
+      "epoch": 50.16,
+      "learning_rate": 0.00028859375,
+      "loss": 0.1696,
+      "step": 1856
+    },
+    {
+      "epoch": 50.19,
+      "learning_rate": 0.0002884375,
+      "loss": 0.2318,
+      "step": 1857
+    },
+    {
+      "epoch": 50.22,
+      "learning_rate": 0.00028828125,
+      "loss": 0.1982,
+      "step": 1858
+    },
+    {
+      "epoch": 50.24,
+      "learning_rate": 0.000288125,
+      "loss": 0.1984,
+      "step": 1859
+    },
+    {
+      "epoch": 50.27,
+      "learning_rate": 0.00028796875,
+      "loss": 0.2193,
+      "step": 1860
+    },
+    {
+      "epoch": 50.3,
+      "learning_rate": 0.00028781250000000004,
+      "loss": 0.1838,
+      "step": 1861
+    },
+    {
+      "epoch": 50.32,
+      "learning_rate": 0.00028765625,
+      "loss": 0.1662,
+      "step": 1862
+    },
+    {
+      "epoch": 50.35,
+      "learning_rate": 0.0002875,
+      "loss": 0.2345,
+      "step": 1863
+    },
+    {
+      "epoch": 50.38,
+      "learning_rate": 0.00028734375,
+      "loss": 0.2145,
+      "step": 1864
+    },
+    {
+      "epoch": 50.41,
+      "learning_rate": 0.0002871875,
+      "loss": 0.1951,
+      "step": 1865
+    },
+    {
+      "epoch": 50.43,
+      "learning_rate": 0.00028703125,
+      "loss": 0.1947,
+      "step": 1866
+    },
+    {
+      "epoch": 50.46,
+      "learning_rate": 0.000286875,
+      "loss": 0.1664,
+      "step": 1867
+    },
+    {
+      "epoch": 50.49,
+      "learning_rate": 0.00028671875,
+      "loss": 0.1781,
+      "step": 1868
+    },
+    {
+      "epoch": 50.51,
+      "learning_rate": 0.0002865625,
+      "loss": 0.1752,
+      "step": 1869
+    },
+    {
+      "epoch": 50.54,
+      "learning_rate": 0.00028640625,
+      "loss": 0.1935,
+      "step": 1870
+    },
+    {
+      "epoch": 50.57,
+      "learning_rate": 0.00028625,
+      "loss": 0.1896,
+      "step": 1871
+    },
+    {
+      "epoch": 50.59,
+      "learning_rate": 0.00028609375,
+      "loss": 0.175,
+      "step": 1872
+    },
+    {
+      "epoch": 50.62,
+      "learning_rate": 0.0002859375,
+      "loss": 0.2108,
+      "step": 1873
+    },
+    {
+      "epoch": 50.65,
+      "learning_rate": 0.00028578125,
+      "loss": 0.2031,
+      "step": 1874
+    },
+    {
+      "epoch": 50.68,
+      "learning_rate": 0.000285625,
+      "loss": 0.1554,
+      "step": 1875
+    },
+    {
+      "epoch": 50.7,
+      "learning_rate": 0.00028546875,
+      "loss": 0.2189,
+      "step": 1876
+    },
+    {
+      "epoch": 50.73,
+      "learning_rate": 0.00028531250000000003,
+      "loss": 0.2048,
+      "step": 1877
+    },
+    {
+      "epoch": 50.76,
+      "learning_rate": 0.00028515625,
+      "loss": 0.2031,
+      "step": 1878
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 0.000285,
+      "loss": 0.1972,
+      "step": 1879
+    },
+    {
+      "epoch": 50.81,
+      "learning_rate": 0.00028484375,
+      "loss": 0.1903,
+      "step": 1880
+    },
+    {
+      "epoch": 50.84,
+      "learning_rate": 0.0002846875,
+      "loss": 0.2236,
+      "step": 1881
+    },
+    {
+      "epoch": 50.86,
+      "learning_rate": 0.00028453125,
+      "loss": 0.2074,
+      "step": 1882
+    },
+    {
+      "epoch": 50.89,
+      "learning_rate": 0.000284375,
+      "loss": 0.202,
+      "step": 1883
+    },
+    {
+      "epoch": 50.92,
+      "learning_rate": 0.00028421875,
+      "loss": 0.2241,
+      "step": 1884
+    },
+    {
+      "epoch": 50.95,
+      "learning_rate": 0.0002840625,
+      "loss": 0.192,
+      "step": 1885
+    },
+    {
+      "epoch": 50.97,
+      "learning_rate": 0.00028390625000000003,
+      "loss": 0.2307,
+      "step": 1886
+    },
+    {
+      "epoch": 51.0,
+      "learning_rate": 0.00028375,
+      "loss": 0.1615,
+      "step": 1887
+    },
+    {
+      "epoch": 51.03,
+      "learning_rate": 0.00028359375,
+      "loss": 0.2087,
+      "step": 1888
+    },
+    {
+      "epoch": 51.05,
+      "learning_rate": 0.0002834375,
+      "loss": 0.2024,
+      "step": 1889
+    },
+    {
+      "epoch": 51.08,
+      "learning_rate": 0.00028328125,
+      "loss": 0.2027,
+      "step": 1890
+    },
+    {
+      "epoch": 51.11,
+      "learning_rate": 0.000283125,
+      "loss": 0.2176,
+      "step": 1891
+    },
+    {
+      "epoch": 51.14,
+      "learning_rate": 0.00028296875,
+      "loss": 0.2128,
+      "step": 1892
+    },
+    {
+      "epoch": 51.16,
+      "learning_rate": 0.0002828125,
+      "loss": 0.1515,
+      "step": 1893
+    },
+    {
+      "epoch": 51.19,
+      "learning_rate": 0.00028265625000000003,
+      "loss": 0.1878,
+      "step": 1894
+    },
+    {
+      "epoch": 51.22,
+      "learning_rate": 0.0002825,
+      "loss": 0.2056,
+      "step": 1895
+    },
+    {
+      "epoch": 51.24,
+      "learning_rate": 0.00028234375,
+      "loss": 0.1681,
+      "step": 1896
+    },
+    {
+      "epoch": 51.27,
+      "learning_rate": 0.0002821875,
+      "loss": 0.1696,
+      "step": 1897
+    },
+    {
+      "epoch": 51.3,
+      "learning_rate": 0.00028203125,
+      "loss": 0.2137,
+      "step": 1898
+    },
+    {
+      "epoch": 51.32,
+      "learning_rate": 0.000281875,
+      "loss": 0.1484,
+      "step": 1899
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 0.00028171875,
+      "loss": 0.2389,
+      "step": 1900
+    },
+    {
+      "epoch": 51.38,
+      "learning_rate": 0.0002815625,
+      "loss": 0.2378,
+      "step": 1901
+    },
+    {
+      "epoch": 51.41,
+      "learning_rate": 0.00028140625,
+      "loss": 0.2052,
+      "step": 1902
+    },
+    {
+      "epoch": 51.43,
+      "learning_rate": 0.00028125000000000003,
+      "loss": 0.1732,
+      "step": 1903
+    },
+    {
+      "epoch": 51.46,
+      "learning_rate": 0.00028109375,
+      "loss": 0.157,
+      "step": 1904
+    },
+    {
+      "epoch": 51.49,
+      "learning_rate": 0.0002809375,
+      "loss": 0.1458,
+      "step": 1905
+    },
+    {
+      "epoch": 51.51,
+      "learning_rate": 0.00028078125,
+      "loss": 0.1828,
+      "step": 1906
+    },
+    {
+      "epoch": 51.54,
+      "learning_rate": 0.000280625,
+      "loss": 0.2103,
+      "step": 1907
+    },
+    {
+      "epoch": 51.57,
+      "learning_rate": 0.00028046875,
+      "loss": 0.2093,
+      "step": 1908
+    },
+    {
+      "epoch": 51.59,
+      "learning_rate": 0.0002803125,
+      "loss": 0.2086,
+      "step": 1909
+    },
+    {
+      "epoch": 51.62,
+      "learning_rate": 0.00028015625,
+      "loss": 0.1951,
+      "step": 1910
+    },
+    {
+      "epoch": 51.65,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 0.18,
+      "step": 1911
+    },
+    {
+      "epoch": 51.68,
+      "learning_rate": 0.00027984375,
+      "loss": 0.2043,
+      "step": 1912
+    },
+    {
+      "epoch": 51.7,
+      "learning_rate": 0.0002796875,
+      "loss": 0.1871,
+      "step": 1913
+    },
+    {
+      "epoch": 51.73,
+      "learning_rate": 0.00027953125,
+      "loss": 0.2197,
+      "step": 1914
+    },
+    {
+      "epoch": 51.76,
+      "learning_rate": 0.000279375,
+      "loss": 0.2047,
+      "step": 1915
+    },
+    {
+      "epoch": 51.78,
+      "learning_rate": 0.00027921875,
+      "loss": 0.2108,
+      "step": 1916
+    },
+    {
+      "epoch": 51.81,
+      "learning_rate": 0.0002790625,
+      "loss": 0.2286,
+      "step": 1917
+    },
+    {
+      "epoch": 51.84,
+      "learning_rate": 0.00027890625,
+      "loss": 0.1818,
+      "step": 1918
+    },
+    {
+      "epoch": 51.86,
+      "learning_rate": 0.00027875,
+      "loss": 0.2094,
+      "step": 1919
+    },
+    {
+      "epoch": 51.89,
+      "learning_rate": 0.00027859375,
+      "loss": 0.1706,
+      "step": 1920
+    },
+    {
+      "epoch": 51.92,
+      "learning_rate": 0.0002784375,
+      "loss": 0.1623,
+      "step": 1921
+    },
+    {
+      "epoch": 51.95,
+      "learning_rate": 0.00027828125,
+      "loss": 0.1688,
+      "step": 1922
+    },
+    {
+      "epoch": 51.97,
+      "learning_rate": 0.000278125,
+      "loss": 0.1795,
+      "step": 1923
+    },
+    {
+      "epoch": 52.0,
+      "learning_rate": 0.00027796875,
+      "loss": 0.1726,
+      "step": 1924
+    },
+    {
+      "epoch": 52.03,
+      "learning_rate": 0.0002778125,
+      "loss": 0.215,
+      "step": 1925
+    },
+    {
+      "epoch": 52.05,
+      "learning_rate": 0.00027765625,
+      "loss": 0.2058,
+      "step": 1926
+    },
+    {
+      "epoch": 52.08,
+      "learning_rate": 0.0002775,
+      "loss": 0.1817,
+      "step": 1927
+    },
+    {
+      "epoch": 52.11,
+      "learning_rate": 0.00027734375000000003,
+      "loss": 0.1827,
+      "step": 1928
+    },
+    {
+      "epoch": 52.14,
+      "learning_rate": 0.0002771875,
+      "loss": 0.1993,
+      "step": 1929
+    },
+    {
+      "epoch": 52.16,
+      "learning_rate": 0.00027703125,
+      "loss": 0.2007,
+      "step": 1930
+    },
+    {
+      "epoch": 52.19,
+      "learning_rate": 0.000276875,
+      "loss": 0.1918,
+      "step": 1931
+    },
+    {
+      "epoch": 52.22,
+      "learning_rate": 0.00027671875,
+      "loss": 0.2049,
+      "step": 1932
+    },
+    {
+      "epoch": 52.24,
+      "learning_rate": 0.0002765625,
+      "loss": 0.1876,
+      "step": 1933
+    },
+    {
+      "epoch": 52.27,
+      "learning_rate": 0.00027640625,
+      "loss": 0.1891,
+      "step": 1934
+    },
+    {
+      "epoch": 52.3,
+      "learning_rate": 0.00027625,
+      "loss": 0.2231,
+      "step": 1935
+    },
+    {
+      "epoch": 52.32,
+      "learning_rate": 0.00027609375,
+      "loss": 0.1727,
+      "step": 1936
+    },
+    {
+      "epoch": 52.35,
+      "learning_rate": 0.00027593750000000003,
+      "loss": 0.1915,
+      "step": 1937
+    },
+    {
+      "epoch": 52.38,
+      "learning_rate": 0.00027578125,
+      "loss": 0.1904,
+      "step": 1938
+    },
+    {
+      "epoch": 52.41,
+      "learning_rate": 0.000275625,
+      "loss": 0.2137,
+      "step": 1939
+    },
+    {
+      "epoch": 52.43,
+      "learning_rate": 0.00027546875,
+      "loss": 0.1968,
+      "step": 1940
+    },
+    {
+      "epoch": 52.46,
+      "learning_rate": 0.0002753125,
+      "loss": 0.1574,
+      "step": 1941
+    },
+    {
+      "epoch": 52.49,
+      "learning_rate": 0.00027515625,
+      "loss": 0.2204,
+      "step": 1942
+    },
+    {
+      "epoch": 52.51,
+      "learning_rate": 0.000275,
+      "loss": 0.2068,
+      "step": 1943
+    },
+    {
+      "epoch": 52.54,
+      "learning_rate": 0.00027484375,
+      "loss": 0.2211,
+      "step": 1944
+    },
+    {
+      "epoch": 52.57,
+      "learning_rate": 0.0002746875,
+      "loss": 0.2033,
+      "step": 1945
+    },
+    {
+      "epoch": 52.59,
+      "learning_rate": 0.00027453125000000004,
+      "loss": 0.1727,
+      "step": 1946
+    },
+    {
+      "epoch": 52.62,
+      "learning_rate": 0.000274375,
+      "loss": 0.2215,
+      "step": 1947
+    },
+    {
+      "epoch": 52.65,
+      "learning_rate": 0.00027421875,
+      "loss": 0.2297,
+      "step": 1948
+    },
+    {
+      "epoch": 52.68,
+      "learning_rate": 0.0002740625,
+      "loss": 0.1399,
+      "step": 1949
+    },
+    {
+      "epoch": 52.7,
+      "learning_rate": 0.00027390625,
+      "loss": 0.2085,
+      "step": 1950
+    },
+    {
+      "epoch": 52.73,
+      "learning_rate": 0.00027375,
+      "loss": 0.1999,
+      "step": 1951
+    },
+    {
+      "epoch": 52.76,
+      "learning_rate": 0.00027359375,
+      "loss": 0.1749,
+      "step": 1952
+    },
+    {
+      "epoch": 52.78,
+      "learning_rate": 0.0002734375,
+      "loss": 0.1714,
+      "step": 1953
+    },
+    {
+      "epoch": 52.81,
+      "learning_rate": 0.00027328125,
+      "loss": 0.2082,
+      "step": 1954
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 0.00027312500000000004,
+      "loss": 0.1535,
+      "step": 1955
+    },
+    {
+      "epoch": 52.86,
+      "learning_rate": 0.00027296875,
+      "loss": 0.1537,
+      "step": 1956
+    },
+    {
+      "epoch": 52.89,
+      "learning_rate": 0.0002728125,
+      "loss": 0.2035,
+      "step": 1957
+    },
+    {
+      "epoch": 52.92,
+      "learning_rate": 0.00027265625,
+      "loss": 0.1871,
+      "step": 1958
+    },
+    {
+      "epoch": 52.95,
+      "learning_rate": 0.0002725,
+      "loss": 0.2086,
+      "step": 1959
+    },
+    {
+      "epoch": 52.97,
+      "learning_rate": 0.00027234375,
+      "loss": 0.1365,
+      "step": 1960
+    },
+    {
+      "epoch": 53.0,
+      "learning_rate": 0.0002721875,
+      "loss": 0.1595,
+      "step": 1961
+    },
+    {
+      "epoch": 53.03,
+      "learning_rate": 0.00027203125000000003,
+      "loss": 0.2085,
+      "step": 1962
+    },
+    {
+      "epoch": 53.05,
+      "learning_rate": 0.000271875,
+      "loss": 0.1801,
+      "step": 1963
+    },
+    {
+      "epoch": 53.08,
+      "learning_rate": 0.00027171875,
+      "loss": 0.1888,
+      "step": 1964
+    },
+    {
+      "epoch": 53.11,
+      "learning_rate": 0.0002715625,
+      "loss": 0.2026,
+      "step": 1965
+    },
+    {
+      "epoch": 53.14,
+      "learning_rate": 0.00027140625,
+      "loss": 0.2227,
+      "step": 1966
+    },
+    {
+      "epoch": 53.16,
+      "learning_rate": 0.00027125,
+      "loss": 0.1523,
+      "step": 1967
+    },
+    {
+      "epoch": 53.19,
+      "learning_rate": 0.00027109375,
+      "loss": 0.241,
+      "step": 1968
+    },
+    {
+      "epoch": 53.22,
+      "learning_rate": 0.0002709375,
+      "loss": 0.1697,
+      "step": 1969
+    },
+    {
+      "epoch": 53.24,
+      "learning_rate": 0.00027078124999999997,
+      "loss": 0.1666,
+      "step": 1970
+    },
+    {
+      "epoch": 53.27,
+      "learning_rate": 0.00027062500000000003,
+      "loss": 0.1874,
+      "step": 1971
+    },
+    {
+      "epoch": 53.3,
+      "learning_rate": 0.00027046875,
+      "loss": 0.2019,
+      "step": 1972
+    },
+    {
+      "epoch": 53.32,
+      "learning_rate": 0.0002703125,
+      "loss": 0.173,
+      "step": 1973
+    },
+    {
+      "epoch": 53.35,
+      "learning_rate": 0.00027015625,
+      "loss": 0.1736,
+      "step": 1974
+    },
+    {
+      "epoch": 53.38,
+      "learning_rate": 0.00027,
+      "loss": 0.1776,
+      "step": 1975
+    },
+    {
+      "epoch": 53.41,
+      "learning_rate": 0.00026984375,
+      "loss": 0.1841,
+      "step": 1976
+    },
+    {
+      "epoch": 53.43,
+      "learning_rate": 0.0002696875,
+      "loss": 0.1541,
+      "step": 1977
+    },
+    {
+      "epoch": 53.46,
+      "learning_rate": 0.00026953125,
+      "loss": 0.1841,
+      "step": 1978
+    },
+    {
+      "epoch": 53.49,
+      "learning_rate": 0.000269375,
+      "loss": 0.1787,
+      "step": 1979
+    },
+    {
+      "epoch": 53.51,
+      "learning_rate": 0.00026921875000000004,
+      "loss": 0.1719,
+      "step": 1980
+    },
+    {
+      "epoch": 53.54,
+      "learning_rate": 0.0002690625,
+      "loss": 0.1819,
+      "step": 1981
+    },
+    {
+      "epoch": 53.57,
+      "learning_rate": 0.00026890625,
+      "loss": 0.1652,
+      "step": 1982
+    },
+    {
+      "epoch": 53.59,
+      "learning_rate": 0.00026875,
+      "loss": 0.2322,
+      "step": 1983
+    },
+    {
+      "epoch": 53.62,
+      "learning_rate": 0.00026859375,
+      "loss": 0.1996,
+      "step": 1984
+    },
+    {
+      "epoch": 53.65,
+      "learning_rate": 0.0002684375,
+      "loss": 0.1656,
+      "step": 1985
+    },
+    {
+      "epoch": 53.68,
+      "learning_rate": 0.00026828125,
+      "loss": 0.1536,
+      "step": 1986
+    },
+    {
+      "epoch": 53.7,
+      "learning_rate": 0.000268125,
+      "loss": 0.2002,
+      "step": 1987
+    },
+    {
+      "epoch": 53.73,
+      "learning_rate": 0.00026796875,
+      "loss": 0.2064,
+      "step": 1988
+    },
+    {
+      "epoch": 53.76,
+      "learning_rate": 0.00026781250000000004,
+      "loss": 0.1724,
+      "step": 1989
+    },
+    {
+      "epoch": 53.78,
+      "learning_rate": 0.00026765625,
+      "loss": 0.2075,
+      "step": 1990
+    },
+    {
+      "epoch": 53.81,
+      "learning_rate": 0.0002675,
+      "loss": 0.1723,
+      "step": 1991
+    },
+    {
+      "epoch": 53.84,
+      "learning_rate": 0.00026734375,
+      "loss": 0.1993,
+      "step": 1992
+    },
+    {
+      "epoch": 53.86,
+      "learning_rate": 0.0002671875,
+      "loss": 0.1933,
+      "step": 1993
+    },
+    {
+      "epoch": 53.89,
+      "learning_rate": 0.00026703125,
+      "loss": 0.191,
+      "step": 1994
+    },
+    {
+      "epoch": 53.92,
+      "learning_rate": 0.00026687499999999997,
+      "loss": 0.1886,
+      "step": 1995
+    },
+    {
+      "epoch": 53.95,
+      "learning_rate": 0.00026671875000000003,
+      "loss": 0.1488,
+      "step": 1996
+    },
+    {
+      "epoch": 53.97,
+      "learning_rate": 0.0002665625,
+      "loss": 0.1522,
+      "step": 1997
+    },
+    {
+      "epoch": 54.0,
+      "learning_rate": 0.00026640625000000004,
+      "loss": 0.1316,
+      "step": 1998
+    },
+    {
+      "epoch": 54.03,
+      "learning_rate": 0.00026625,
+      "loss": 0.1915,
+      "step": 1999
+    },
+    {
+      "epoch": 54.05,
+      "learning_rate": 0.00026609375,
+      "loss": 0.1805,
+      "step": 2000
+    },
+    {
+      "epoch": 54.05,
+      "eval_cer": 0.08536859623082804,
+      "eval_loss": 0.3942892849445343,
+      "eval_runtime": 11.6653,
+      "eval_samples_per_second": 141.187,
+      "eval_steps_per_second": 2.229,
+      "eval_wer": 0.3486875702175467,
+      "step": 2000
+    },
+    {
+      "epoch": 54.08,
+      "learning_rate": 0.0002659375,
+      "loss": 0.1914,
+      "step": 2001
+    },
+    {
+      "epoch": 54.11,
+      "learning_rate": 0.00026578125,
+      "loss": 0.2135,
+      "step": 2002
+    },
+    {
+      "epoch": 54.14,
+      "learning_rate": 0.000265625,
+      "loss": 0.1761,
+      "step": 2003
+    },
+    {
+      "epoch": 54.16,
+      "learning_rate": 0.00026546874999999997,
+      "loss": 0.1678,
+      "step": 2004
+    },
+    {
+      "epoch": 54.19,
+      "learning_rate": 0.00026531250000000003,
+      "loss": 0.2055,
+      "step": 2005
+    },
+    {
+      "epoch": 54.22,
+      "learning_rate": 0.00026515625,
+      "loss": 0.1933,
+      "step": 2006
+    },
+    {
+      "epoch": 54.24,
+      "learning_rate": 0.00026500000000000004,
+      "loss": 0.216,
+      "step": 2007
+    },
+    {
+      "epoch": 54.27,
+      "learning_rate": 0.00026484375,
+      "loss": 0.1997,
+      "step": 2008
+    },
+    {
+      "epoch": 54.3,
+      "learning_rate": 0.0002646875,
+      "loss": 0.1682,
+      "step": 2009
+    },
+    {
+      "epoch": 54.32,
+      "learning_rate": 0.00026453125,
+      "loss": 0.1545,
+      "step": 2010
+    },
+    {
+      "epoch": 54.35,
+      "learning_rate": 0.000264375,
+      "loss": 0.2493,
+      "step": 2011
+    },
+    {
+      "epoch": 54.38,
+      "learning_rate": 0.00026421875,
+      "loss": 0.2055,
+      "step": 2012
+    },
+    {
+      "epoch": 54.41,
+      "learning_rate": 0.0002640625,
+      "loss": 0.1623,
+      "step": 2013
+    },
+    {
+      "epoch": 54.43,
+      "learning_rate": 0.00026390625000000003,
+      "loss": 0.1665,
+      "step": 2014
+    },
+    {
+      "epoch": 54.46,
+      "learning_rate": 0.00026375,
+      "loss": 0.1835,
+      "step": 2015
+    },
+    {
+      "epoch": 54.49,
+      "learning_rate": 0.00026359375000000005,
+      "loss": 0.193,
+      "step": 2016
+    },
+    {
+      "epoch": 54.51,
+      "learning_rate": 0.0002634375,
+      "loss": 0.1723,
+      "step": 2017
+    },
+    {
+      "epoch": 54.54,
+      "learning_rate": 0.00026328125,
+      "loss": 0.2119,
+      "step": 2018
+    },
+    {
+      "epoch": 54.57,
+      "learning_rate": 0.000263125,
+      "loss": 0.1862,
+      "step": 2019
+    },
+    {
+      "epoch": 54.59,
+      "learning_rate": 0.00026296874999999996,
+      "loss": 0.164,
+      "step": 2020
+    },
+    {
+      "epoch": 54.62,
+      "learning_rate": 0.0002628125,
+      "loss": 0.2151,
+      "step": 2021
+    },
+    {
+      "epoch": 54.65,
+      "learning_rate": 0.00026265625,
+      "loss": 0.192,
+      "step": 2022
+    },
+    {
+      "epoch": 54.68,
+      "learning_rate": 0.00026250000000000004,
+      "loss": 0.1742,
+      "step": 2023
+    },
+    {
+      "epoch": 54.7,
+      "learning_rate": 0.00026234375,
+      "loss": 0.1875,
+      "step": 2024
+    },
+    {
+      "epoch": 54.73,
+      "learning_rate": 0.0002621875,
+      "loss": 0.1999,
+      "step": 2025
+    },
+    {
+      "epoch": 54.76,
+      "learning_rate": 0.00026203125,
+      "loss": 0.1707,
+      "step": 2026
+    },
+    {
+      "epoch": 54.78,
+      "learning_rate": 0.000261875,
+      "loss": 0.1872,
+      "step": 2027
+    },
+    {
+      "epoch": 54.81,
+      "learning_rate": 0.00026171875,
+      "loss": 0.1682,
+      "step": 2028
+    },
+    {
+      "epoch": 54.84,
+      "learning_rate": 0.00026156249999999997,
+      "loss": 0.1449,
+      "step": 2029
+    },
+    {
+      "epoch": 54.86,
+      "learning_rate": 0.00026140625000000003,
+      "loss": 0.1707,
+      "step": 2030
+    },
+    {
+      "epoch": 54.89,
+      "learning_rate": 0.00026125,
+      "loss": 0.1866,
+      "step": 2031
+    },
+    {
+      "epoch": 54.92,
+      "learning_rate": 0.00026109375000000004,
+      "loss": 0.1942,
+      "step": 2032
+    },
+    {
+      "epoch": 54.95,
+      "learning_rate": 0.0002609375,
+      "loss": 0.1633,
+      "step": 2033
+    },
+    {
+      "epoch": 54.97,
+      "learning_rate": 0.00026078125,
+      "loss": 0.1807,
+      "step": 2034
+    },
+    {
+      "epoch": 55.0,
+      "learning_rate": 0.000260625,
+      "loss": 0.0834,
+      "step": 2035
+    },
+    {
+      "epoch": 55.03,
+      "learning_rate": 0.00026046875,
+      "loss": 0.1575,
+      "step": 2036
+    },
+    {
+      "epoch": 55.05,
+      "learning_rate": 0.0002603125,
+      "loss": 0.1744,
+      "step": 2037
+    },
+    {
+      "epoch": 55.08,
+      "learning_rate": 0.00026015624999999997,
+      "loss": 0.1807,
+      "step": 2038
+    },
+    {
+      "epoch": 55.11,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 0.2208,
+      "step": 2039
+    },
+    {
+      "epoch": 55.14,
+      "learning_rate": 0.00025984375,
+      "loss": 0.1612,
+      "step": 2040
+    },
+    {
+      "epoch": 55.16,
+      "learning_rate": 0.00025968750000000004,
+      "loss": 0.1386,
+      "step": 2041
+    },
+    {
+      "epoch": 55.19,
+      "learning_rate": 0.00025953125,
+      "loss": 0.2042,
+      "step": 2042
+    },
+    {
+      "epoch": 55.22,
+      "learning_rate": 0.000259375,
+      "loss": 0.1752,
+      "step": 2043
+    },
+    {
+      "epoch": 55.24,
+      "learning_rate": 0.00025921875,
+      "loss": 0.1927,
+      "step": 2044
+    },
+    {
+      "epoch": 55.27,
+      "learning_rate": 0.00025906249999999996,
+      "loss": 0.1859,
+      "step": 2045
+    },
+    {
+      "epoch": 55.3,
+      "learning_rate": 0.00025890625,
+      "loss": 0.1779,
+      "step": 2046
+    },
+    {
+      "epoch": 55.32,
+      "learning_rate": 0.00025875,
+      "loss": 0.1666,
+      "step": 2047
+    },
+    {
+      "epoch": 55.35,
+      "learning_rate": 0.00025859375000000003,
+      "loss": 0.164,
+      "step": 2048
+    },
+    {
+      "epoch": 55.38,
+      "learning_rate": 0.0002584375,
+      "loss": 0.1765,
+      "step": 2049
+    },
+    {
+      "epoch": 55.41,
+      "learning_rate": 0.00025828125000000005,
+      "loss": 0.1896,
+      "step": 2050
+    },
+    {
+      "epoch": 55.43,
+      "learning_rate": 0.000258125,
+      "loss": 0.1616,
+      "step": 2051
+    },
+    {
+      "epoch": 55.46,
+      "learning_rate": 0.00025796875,
+      "loss": 0.234,
+      "step": 2052
+    },
+    {
+      "epoch": 55.49,
+      "learning_rate": 0.0002578125,
+      "loss": 0.1871,
+      "step": 2053
+    },
+    {
+      "epoch": 55.51,
+      "learning_rate": 0.00025765624999999996,
+      "loss": 0.1723,
+      "step": 2054
+    },
+    {
+      "epoch": 55.54,
+      "learning_rate": 0.0002575,
+      "loss": 0.1958,
+      "step": 2055
+    },
+    {
+      "epoch": 55.57,
+      "learning_rate": 0.00025734375,
+      "loss": 0.2252,
+      "step": 2056
+    },
+    {
+      "epoch": 55.59,
+      "learning_rate": 0.00025718750000000004,
+      "loss": 0.1898,
+      "step": 2057
+    },
+    {
+      "epoch": 55.62,
+      "learning_rate": 0.00025703125,
+      "loss": 0.1735,
+      "step": 2058
+    },
+    {
+      "epoch": 55.65,
+      "learning_rate": 0.00025687500000000005,
+      "loss": 0.1987,
+      "step": 2059
+    },
+    {
+      "epoch": 55.68,
+      "learning_rate": 0.00025671875,
+      "loss": 0.1512,
+      "step": 2060
+    },
+    {
+      "epoch": 55.7,
+      "learning_rate": 0.0002565625,
+      "loss": 0.2083,
+      "step": 2061
+    },
+    {
+      "epoch": 55.73,
+      "learning_rate": 0.00025640625,
+      "loss": 0.2016,
+      "step": 2062
+    },
+    {
+      "epoch": 55.76,
+      "learning_rate": 0.00025624999999999997,
+      "loss": 0.2306,
+      "step": 2063
+    },
+    {
+      "epoch": 55.78,
+      "learning_rate": 0.00025609375000000003,
+      "loss": 0.1883,
+      "step": 2064
+    },
+    {
+      "epoch": 55.81,
+      "learning_rate": 0.0002559375,
+      "loss": 0.1733,
+      "step": 2065
+    },
+    {
+      "epoch": 55.84,
+      "learning_rate": 0.00025578125000000004,
+      "loss": 0.1727,
+      "step": 2066
+    },
+    {
+      "epoch": 55.86,
+      "learning_rate": 0.000255625,
+      "loss": 0.1833,
+      "step": 2067
+    },
+    {
+      "epoch": 55.89,
+      "learning_rate": 0.00025546875000000005,
+      "loss": 0.215,
+      "step": 2068
+    },
+    {
+      "epoch": 55.92,
+      "learning_rate": 0.0002553125,
+      "loss": 0.1626,
+      "step": 2069
+    },
+    {
+      "epoch": 55.95,
+      "learning_rate": 0.00025515624999999996,
+      "loss": 0.2208,
+      "step": 2070
+    },
+    {
+      "epoch": 55.97,
+      "learning_rate": 0.000255,
+      "loss": 0.1963,
+      "step": 2071
+    },
+    {
+      "epoch": 56.0,
+      "learning_rate": 0.00025484374999999997,
+      "loss": 0.1679,
+      "step": 2072
+    },
+    {
+      "epoch": 56.03,
+      "learning_rate": 0.00025468750000000003,
+      "loss": 0.19,
+      "step": 2073
+    },
+    {
+      "epoch": 56.05,
+      "learning_rate": 0.00025453125,
+      "loss": 0.1839,
+      "step": 2074
+    },
+    {
+      "epoch": 56.08,
+      "learning_rate": 0.00025437500000000004,
+      "loss": 0.2009,
+      "step": 2075
+    },
+    {
+      "epoch": 56.11,
+      "learning_rate": 0.00025421875,
+      "loss": 0.1859,
+      "step": 2076
+    },
+    {
+      "epoch": 56.14,
+      "learning_rate": 0.00025406250000000006,
+      "loss": 0.2139,
+      "step": 2077
+    },
+    {
+      "epoch": 56.16,
+      "learning_rate": 0.00025390625,
+      "loss": 0.1816,
+      "step": 2078
+    },
+    {
+      "epoch": 56.19,
+      "learning_rate": 0.00025374999999999996,
+      "loss": 0.1767,
+      "step": 2079
+    },
+    {
+      "epoch": 56.22,
+      "learning_rate": 0.00025359375,
+      "loss": 0.1976,
+      "step": 2080
+    },
+    {
+      "epoch": 56.24,
+      "learning_rate": 0.0002534375,
+      "loss": 0.1755,
+      "step": 2081
+    },
+    {
+      "epoch": 56.27,
+      "learning_rate": 0.00025328125000000003,
+      "loss": 0.1635,
+      "step": 2082
+    },
+    {
+      "epoch": 56.3,
+      "learning_rate": 0.000253125,
+      "loss": 0.1627,
+      "step": 2083
+    },
+    {
+      "epoch": 56.32,
+      "learning_rate": 0.00025296875000000005,
+      "loss": 0.1852,
+      "step": 2084
+    },
+    {
+      "epoch": 56.35,
+      "learning_rate": 0.0002528125,
+      "loss": 0.1933,
+      "step": 2085
+    },
+    {
+      "epoch": 56.38,
+      "learning_rate": 0.00025265625,
+      "loss": 0.1757,
+      "step": 2086
+    },
+    {
+      "epoch": 56.41,
+      "learning_rate": 0.0002525,
+      "loss": 0.1919,
+      "step": 2087
+    },
+    {
+      "epoch": 56.43,
+      "learning_rate": 0.00025234374999999996,
+      "loss": 0.1824,
+      "step": 2088
+    },
+    {
+      "epoch": 56.46,
+      "learning_rate": 0.0002521875,
+      "loss": 0.1934,
+      "step": 2089
+    },
+    {
+      "epoch": 56.49,
+      "learning_rate": 0.00025203125,
+      "loss": 0.1619,
+      "step": 2090
+    },
+    {
+      "epoch": 56.51,
+      "learning_rate": 0.00025187500000000004,
+      "loss": 0.1593,
+      "step": 2091
+    },
+    {
+      "epoch": 56.54,
+      "learning_rate": 0.00025171875,
+      "loss": 0.2113,
+      "step": 2092
+    },
+    {
+      "epoch": 56.57,
+      "learning_rate": 0.00025156250000000005,
+      "loss": 0.1756,
+      "step": 2093
+    },
+    {
+      "epoch": 56.59,
+      "learning_rate": 0.00025140625,
+      "loss": 0.1492,
+      "step": 2094
+    },
+    {
+      "epoch": 56.62,
+      "learning_rate": 0.00025124999999999995,
+      "loss": 0.1804,
+      "step": 2095
+    },
+    {
+      "epoch": 56.65,
+      "learning_rate": 0.00025109375,
+      "loss": 0.1853,
+      "step": 2096
+    },
+    {
+      "epoch": 56.68,
+      "learning_rate": 0.00025093749999999997,
+      "loss": 0.1659,
+      "step": 2097
+    },
+    {
+      "epoch": 56.7,
+      "learning_rate": 0.00025078125000000003,
+      "loss": 0.1873,
+      "step": 2098
+    },
+    {
+      "epoch": 56.73,
+      "learning_rate": 0.000250625,
+      "loss": 0.1611,
+      "step": 2099
+    },
+    {
+      "epoch": 56.76,
+      "learning_rate": 0.00025046875000000004,
+      "loss": 0.1875,
+      "step": 2100
+    },
+    {
+      "epoch": 56.78,
+      "learning_rate": 0.0002503125,
+      "loss": 0.1801,
+      "step": 2101
+    },
+    {
+      "epoch": 56.81,
+      "learning_rate": 0.00025015625000000005,
+      "loss": 0.1729,
+      "step": 2102
+    },
+    {
+      "epoch": 56.84,
+      "learning_rate": 0.00025,
+      "loss": 0.1653,
+      "step": 2103
+    },
+    {
+      "epoch": 56.86,
+      "learning_rate": 0.00024984375,
+      "loss": 0.2101,
+      "step": 2104
+    },
+    {
+      "epoch": 56.89,
+      "learning_rate": 0.0002496875,
+      "loss": 0.2044,
+      "step": 2105
+    },
+    {
+      "epoch": 56.92,
+      "learning_rate": 0.00024953125,
+      "loss": 0.1525,
+      "step": 2106
+    },
+    {
+      "epoch": 56.95,
+      "learning_rate": 0.00024937500000000003,
+      "loss": 0.1628,
+      "step": 2107
+    },
+    {
+      "epoch": 56.97,
+      "learning_rate": 0.00024921875,
+      "loss": 0.2203,
+      "step": 2108
+    },
+    {
+      "epoch": 57.0,
+      "learning_rate": 0.0002490625,
+      "loss": 0.1252,
+      "step": 2109
+    },
+    {
+      "epoch": 57.03,
+      "learning_rate": 0.00024890625,
+      "loss": 0.2005,
+      "step": 2110
+    },
+    {
+      "epoch": 57.05,
+      "learning_rate": 0.00024875,
+      "loss": 0.1609,
+      "step": 2111
+    },
+    {
+      "epoch": 57.08,
+      "learning_rate": 0.00024859375,
+      "loss": 0.1588,
+      "step": 2112
+    },
+    {
+      "epoch": 57.11,
+      "learning_rate": 0.0002484375,
+      "loss": 0.202,
+      "step": 2113
+    },
+    {
+      "epoch": 57.14,
+      "learning_rate": 0.00024828125,
+      "loss": 0.1961,
+      "step": 2114
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 0.00024812500000000003,
+      "loss": 0.1893,
+      "step": 2115
+    },
+    {
+      "epoch": 57.19,
+      "learning_rate": 0.00024796875,
+      "loss": 0.1738,
+      "step": 2116
+    },
+    {
+      "epoch": 57.22,
+      "learning_rate": 0.0002478125,
+      "loss": 0.2127,
+      "step": 2117
+    },
+    {
+      "epoch": 57.24,
+      "learning_rate": 0.00024765625,
+      "loss": 0.16,
+      "step": 2118
+    },
+    {
+      "epoch": 57.27,
+      "learning_rate": 0.0002475,
+      "loss": 0.1785,
+      "step": 2119
+    },
+    {
+      "epoch": 57.3,
+      "learning_rate": 0.00024734375,
+      "loss": 0.138,
+      "step": 2120
+    },
+    {
+      "epoch": 57.32,
+      "learning_rate": 0.0002471875,
+      "loss": 0.1505,
+      "step": 2121
+    },
+    {
+      "epoch": 57.35,
+      "learning_rate": 0.00024703125,
+      "loss": 0.1389,
+      "step": 2122
+    },
+    {
+      "epoch": 57.38,
+      "learning_rate": 0.000246875,
+      "loss": 0.1985,
+      "step": 2123
+    },
+    {
+      "epoch": 57.41,
+      "learning_rate": 0.00024671875,
+      "loss": 0.1656,
+      "step": 2124
+    },
+    {
+      "epoch": 57.43,
+      "learning_rate": 0.0002465625,
+      "loss": 0.183,
+      "step": 2125
+    },
+    {
+      "epoch": 57.46,
+      "learning_rate": 0.00024640625,
+      "loss": 0.2115,
+      "step": 2126
+    },
+    {
+      "epoch": 57.49,
+      "learning_rate": 0.00024625,
+      "loss": 0.1648,
+      "step": 2127
+    },
+    {
+      "epoch": 57.51,
+      "learning_rate": 0.00024609375,
+      "loss": 0.14,
+      "step": 2128
+    },
+    {
+      "epoch": 57.54,
+      "learning_rate": 0.0002459375,
+      "loss": 0.2063,
+      "step": 2129
+    },
+    {
+      "epoch": 57.57,
+      "learning_rate": 0.00024578125,
+      "loss": 0.1754,
+      "step": 2130
+    },
+    {
+      "epoch": 57.59,
+      "learning_rate": 0.000245625,
+      "loss": 0.1575,
+      "step": 2131
+    },
+    {
+      "epoch": 57.62,
+      "learning_rate": 0.00024546875000000003,
+      "loss": 0.1704,
+      "step": 2132
+    },
+    {
+      "epoch": 57.65,
+      "learning_rate": 0.0002453125,
+      "loss": 0.1935,
+      "step": 2133
+    },
+    {
+      "epoch": 57.68,
+      "learning_rate": 0.00024515625,
+      "loss": 0.1431,
+      "step": 2134
+    },
+    {
+      "epoch": 57.7,
+      "learning_rate": 0.000245,
+      "loss": 0.2051,
+      "step": 2135
+    },
+    {
+      "epoch": 57.73,
+      "learning_rate": 0.00024484375,
+      "loss": 0.1766,
+      "step": 2136
+    },
+    {
+      "epoch": 57.76,
+      "learning_rate": 0.0002446875,
+      "loss": 0.1685,
+      "step": 2137
+    },
+    {
+      "epoch": 57.78,
+      "learning_rate": 0.00024453125,
+      "loss": 0.1818,
+      "step": 2138
+    },
+    {
+      "epoch": 57.81,
+      "learning_rate": 0.000244375,
+      "loss": 0.1619,
+      "step": 2139
+    },
+    {
+      "epoch": 57.84,
+      "learning_rate": 0.00024421875,
+      "loss": 0.1392,
+      "step": 2140
+    },
+    {
+      "epoch": 57.86,
+      "learning_rate": 0.0002440625,
+      "loss": 0.1825,
+      "step": 2141
+    },
+    {
+      "epoch": 57.89,
+      "learning_rate": 0.00024390625,
+      "loss": 0.176,
+      "step": 2142
+    },
+    {
+      "epoch": 57.92,
+      "learning_rate": 0.00024375,
+      "loss": 0.1836,
+      "step": 2143
+    },
+    {
+      "epoch": 57.95,
+      "learning_rate": 0.00024359375,
+      "loss": 0.162,
+      "step": 2144
+    },
+    {
+      "epoch": 57.97,
+      "learning_rate": 0.0002434375,
+      "loss": 0.161,
+      "step": 2145
+    },
+    {
+      "epoch": 58.0,
+      "learning_rate": 0.00024328125,
+      "loss": 0.1654,
+      "step": 2146
+    },
+    {
+      "epoch": 58.03,
+      "learning_rate": 0.00024312500000000001,
+      "loss": 0.1773,
+      "step": 2147
+    },
+    {
+      "epoch": 58.05,
+      "learning_rate": 0.00024296875000000002,
+      "loss": 0.1672,
+      "step": 2148
+    },
+    {
+      "epoch": 58.08,
+      "learning_rate": 0.0002428125,
+      "loss": 0.1478,
+      "step": 2149
+    },
+    {
+      "epoch": 58.11,
+      "learning_rate": 0.00024265625,
+      "loss": 0.1924,
+      "step": 2150
+    },
+    {
+      "epoch": 58.14,
+      "learning_rate": 0.00024249999999999999,
+      "loss": 0.1918,
+      "step": 2151
+    },
+    {
+      "epoch": 58.16,
+      "learning_rate": 0.00024234375,
+      "loss": 0.1223,
+      "step": 2152
+    },
+    {
+      "epoch": 58.19,
+      "learning_rate": 0.0002421875,
+      "loss": 0.1754,
+      "step": 2153
+    },
+    {
+      "epoch": 58.22,
+      "learning_rate": 0.00024203125,
+      "loss": 0.1402,
+      "step": 2154
+    },
+    {
+      "epoch": 58.24,
+      "learning_rate": 0.000241875,
+      "loss": 0.1678,
+      "step": 2155
+    },
+    {
+      "epoch": 58.27,
+      "learning_rate": 0.00024171875000000002,
+      "loss": 0.172,
+      "step": 2156
+    },
+    {
+      "epoch": 58.3,
+      "learning_rate": 0.00024156250000000002,
+      "loss": 0.1997,
+      "step": 2157
+    },
+    {
+      "epoch": 58.32,
+      "learning_rate": 0.00024140625,
+      "loss": 0.1295,
+      "step": 2158
+    },
+    {
+      "epoch": 58.35,
+      "learning_rate": 0.00024125,
+      "loss": 0.1792,
+      "step": 2159
+    },
+    {
+      "epoch": 58.38,
+      "learning_rate": 0.00024109375,
+      "loss": 0.1947,
+      "step": 2160
+    },
+    {
+      "epoch": 58.41,
+      "learning_rate": 0.0002409375,
+      "loss": 0.1351,
+      "step": 2161
+    },
+    {
+      "epoch": 58.43,
+      "learning_rate": 0.00024078125,
+      "loss": 0.1712,
+      "step": 2162
+    },
+    {
+      "epoch": 58.46,
+      "learning_rate": 0.000240625,
+      "loss": 0.1717,
+      "step": 2163
+    },
+    {
+      "epoch": 58.49,
+      "learning_rate": 0.00024046875000000001,
+      "loss": 0.1626,
+      "step": 2164
+    },
+    {
+      "epoch": 58.51,
+      "learning_rate": 0.00024031250000000002,
+      "loss": 0.1833,
+      "step": 2165
+    },
+    {
+      "epoch": 58.54,
+      "learning_rate": 0.00024015625,
+      "loss": 0.1787,
+      "step": 2166
+    },
+    {
+      "epoch": 58.57,
+      "learning_rate": 0.00024,
+      "loss": 0.1688,
+      "step": 2167
+    },
+    {
+      "epoch": 58.59,
+      "learning_rate": 0.00023984375,
+      "loss": 0.1945,
+      "step": 2168
+    },
+    {
+      "epoch": 58.62,
+      "learning_rate": 0.0002396875,
+      "loss": 0.1934,
+      "step": 2169
+    },
+    {
+      "epoch": 58.65,
+      "learning_rate": 0.00023953125,
+      "loss": 0.1761,
+      "step": 2170
+    },
+    {
+      "epoch": 58.68,
+      "learning_rate": 0.000239375,
+      "loss": 0.1656,
+      "step": 2171
+    },
+    {
+      "epoch": 58.7,
+      "learning_rate": 0.00023921875,
+      "loss": 0.2099,
+      "step": 2172
+    },
+    {
+      "epoch": 58.73,
+      "learning_rate": 0.00023906250000000002,
+      "loss": 0.1375,
+      "step": 2173
+    },
+    {
+      "epoch": 58.76,
+      "learning_rate": 0.00023890625,
+      "loss": 0.1677,
+      "step": 2174
+    },
+    {
+      "epoch": 58.78,
+      "learning_rate": 0.00023875,
+      "loss": 0.1552,
+      "step": 2175
+    },
+    {
+      "epoch": 58.81,
+      "learning_rate": 0.00023859375,
+      "loss": 0.1893,
+      "step": 2176
+    },
+    {
+      "epoch": 58.84,
+      "learning_rate": 0.0002384375,
+      "loss": 0.1833,
+      "step": 2177
+    },
+    {
+      "epoch": 58.86,
+      "learning_rate": 0.00023828125,
+      "loss": 0.1335,
+      "step": 2178
+    },
+    {
+      "epoch": 58.89,
+      "learning_rate": 0.000238125,
+      "loss": 0.1513,
+      "step": 2179
+    },
+    {
+      "epoch": 58.92,
+      "learning_rate": 0.00023796875,
+      "loss": 0.1807,
+      "step": 2180
+    },
+    {
+      "epoch": 58.95,
+      "learning_rate": 0.00023781250000000001,
+      "loss": 0.184,
+      "step": 2181
+    },
+    {
+      "epoch": 58.97,
+      "learning_rate": 0.00023765625000000002,
+      "loss": 0.1703,
+      "step": 2182
+    },
+    {
+      "epoch": 59.0,
+      "learning_rate": 0.0002375,
+      "loss": 0.1508,
+      "step": 2183
+    },
+    {
+      "epoch": 59.03,
+      "learning_rate": 0.00023734375,
+      "loss": 0.1752,
+      "step": 2184
+    },
+    {
+      "epoch": 59.05,
+      "learning_rate": 0.0002371875,
+      "loss": 0.186,
+      "step": 2185
+    },
+    {
+      "epoch": 59.08,
+      "learning_rate": 0.00023703125,
+      "loss": 0.1583,
+      "step": 2186
+    },
+    {
+      "epoch": 59.11,
+      "learning_rate": 0.000236875,
+      "loss": 0.1482,
+      "step": 2187
+    },
+    {
+      "epoch": 59.14,
+      "learning_rate": 0.00023671875,
+      "loss": 0.1955,
+      "step": 2188
+    },
+    {
+      "epoch": 59.16,
+      "learning_rate": 0.0002365625,
+      "loss": 0.1543,
+      "step": 2189
+    },
+    {
+      "epoch": 59.19,
+      "learning_rate": 0.00023640625000000002,
+      "loss": 0.155,
+      "step": 2190
+    },
+    {
+      "epoch": 59.22,
+      "learning_rate": 0.00023625,
+      "loss": 0.1618,
+      "step": 2191
+    },
+    {
+      "epoch": 59.24,
+      "learning_rate": 0.00023609375,
+      "loss": 0.1484,
+      "step": 2192
+    },
+    {
+      "epoch": 59.27,
+      "learning_rate": 0.0002359375,
+      "loss": 0.1572,
+      "step": 2193
+    },
+    {
+      "epoch": 59.3,
+      "learning_rate": 0.00023578125000000002,
+      "loss": 0.1808,
+      "step": 2194
+    },
+    {
+      "epoch": 59.32,
+      "learning_rate": 0.000235625,
+      "loss": 0.1364,
+      "step": 2195
+    },
+    {
+      "epoch": 59.35,
+      "learning_rate": 0.00023546875,
+      "loss": 0.1854,
+      "step": 2196
+    },
+    {
+      "epoch": 59.38,
+      "learning_rate": 0.0002353125,
+      "loss": 0.1785,
+      "step": 2197
+    },
+    {
+      "epoch": 59.41,
+      "learning_rate": 0.00023515625000000001,
+      "loss": 0.1805,
+      "step": 2198
+    },
+    {
+      "epoch": 59.43,
+      "learning_rate": 0.000235,
+      "loss": 0.1663,
+      "step": 2199
+    },
+    {
+      "epoch": 59.46,
+      "learning_rate": 0.00023484375,
+      "loss": 0.1813,
+      "step": 2200
+    },
+    {
+      "epoch": 59.49,
+      "learning_rate": 0.0002346875,
+      "loss": 0.1637,
+      "step": 2201
+    },
+    {
+      "epoch": 59.51,
+      "learning_rate": 0.00023453125,
+      "loss": 0.144,
+      "step": 2202
+    },
+    {
+      "epoch": 59.54,
+      "learning_rate": 0.000234375,
+      "loss": 0.1828,
+      "step": 2203
+    },
+    {
+      "epoch": 59.57,
+      "learning_rate": 0.00023421875,
+      "loss": 0.1628,
+      "step": 2204
+    },
+    {
+      "epoch": 59.59,
+      "learning_rate": 0.0002340625,
+      "loss": 0.1397,
+      "step": 2205
+    },
+    {
+      "epoch": 59.62,
+      "learning_rate": 0.00023390625,
+      "loss": 0.1486,
+      "step": 2206
+    },
+    {
+      "epoch": 59.65,
+      "learning_rate": 0.00023375000000000002,
+      "loss": 0.1566,
+      "step": 2207
+    },
+    {
+      "epoch": 59.68,
+      "learning_rate": 0.00023359375,
+      "loss": 0.1106,
+      "step": 2208
+    },
+    {
+      "epoch": 59.7,
+      "learning_rate": 0.0002334375,
+      "loss": 0.1967,
+      "step": 2209
+    },
+    {
+      "epoch": 59.73,
+      "learning_rate": 0.00023328125,
+      "loss": 0.1935,
+      "step": 2210
+    },
+    {
+      "epoch": 59.76,
+      "learning_rate": 0.00023312500000000002,
+      "loss": 0.1953,
+      "step": 2211
+    },
+    {
+      "epoch": 59.78,
+      "learning_rate": 0.00023296875,
+      "loss": 0.17,
+      "step": 2212
+    },
+    {
+      "epoch": 59.81,
+      "learning_rate": 0.0002328125,
+      "loss": 0.1523,
+      "step": 2213
+    },
+    {
+      "epoch": 59.84,
+      "learning_rate": 0.00023265625,
+      "loss": 0.1685,
+      "step": 2214
+    },
+    {
+      "epoch": 59.86,
+      "learning_rate": 0.0002325,
+      "loss": 0.1675,
+      "step": 2215
+    },
+    {
+      "epoch": 59.89,
+      "learning_rate": 0.00023234375,
+      "loss": 0.1546,
+      "step": 2216
+    },
+    {
+      "epoch": 59.92,
+      "learning_rate": 0.0002321875,
+      "loss": 0.2071,
+      "step": 2217
+    },
+    {
+      "epoch": 59.95,
+      "learning_rate": 0.00023203125,
+      "loss": 0.1562,
+      "step": 2218
+    },
+    {
+      "epoch": 59.97,
+      "learning_rate": 0.000231875,
+      "loss": 0.1609,
+      "step": 2219
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 0.00023171875000000002,
+      "loss": 0.2023,
+      "step": 2220
+    },
+    {
+      "epoch": 60.03,
+      "learning_rate": 0.0002315625,
+      "loss": 0.1868,
+      "step": 2221
+    },
+    {
+      "epoch": 60.05,
+      "learning_rate": 0.00023140625,
+      "loss": 0.1756,
+      "step": 2222
+    },
+    {
+      "epoch": 60.08,
+      "learning_rate": 0.00023125,
+      "loss": 0.1471,
+      "step": 2223
+    },
+    {
+      "epoch": 60.11,
+      "learning_rate": 0.00023109375,
+      "loss": 0.1725,
+      "step": 2224
+    },
+    {
+      "epoch": 60.14,
+      "learning_rate": 0.0002309375,
+      "loss": 0.1493,
+      "step": 2225
+    },
+    {
+      "epoch": 60.16,
+      "learning_rate": 0.00023078125,
+      "loss": 0.1271,
+      "step": 2226
+    },
+    {
+      "epoch": 60.19,
+      "learning_rate": 0.000230625,
+      "loss": 0.1837,
+      "step": 2227
+    },
+    {
+      "epoch": 60.22,
+      "learning_rate": 0.00023046875000000001,
+      "loss": 0.1551,
+      "step": 2228
+    },
+    {
+      "epoch": 60.24,
+      "learning_rate": 0.00023031250000000002,
+      "loss": 0.1725,
+      "step": 2229
+    },
+    {
+      "epoch": 60.27,
+      "learning_rate": 0.00023015625,
+      "loss": 0.2208,
+      "step": 2230
+    },
+    {
+      "epoch": 60.3,
+      "learning_rate": 0.00023,
+      "loss": 0.1534,
+      "step": 2231
+    },
+    {
+      "epoch": 60.32,
+      "learning_rate": 0.00022984375,
+      "loss": 0.1867,
+      "step": 2232
+    },
+    {
+      "epoch": 60.35,
+      "learning_rate": 0.0002296875,
+      "loss": 0.17,
+      "step": 2233
+    },
+    {
+      "epoch": 60.38,
+      "learning_rate": 0.00022953125,
+      "loss": 0.1865,
+      "step": 2234
+    },
+    {
+      "epoch": 60.41,
+      "learning_rate": 0.000229375,
+      "loss": 0.1573,
+      "step": 2235
+    },
+    {
+      "epoch": 60.43,
+      "learning_rate": 0.00022921875,
+      "loss": 0.1462,
+      "step": 2236
+    },
+    {
+      "epoch": 60.46,
+      "learning_rate": 0.00022906250000000002,
+      "loss": 0.1428,
+      "step": 2237
+    },
+    {
+      "epoch": 60.49,
+      "learning_rate": 0.00022890625,
+      "loss": 0.1533,
+      "step": 2238
+    },
+    {
+      "epoch": 60.51,
+      "learning_rate": 0.00022875,
+      "loss": 0.1328,
+      "step": 2239
+    },
+    {
+      "epoch": 60.54,
+      "learning_rate": 0.00022859375,
+      "loss": 0.1933,
+      "step": 2240
+    },
+    {
+      "epoch": 60.57,
+      "learning_rate": 0.0002284375,
+      "loss": 0.1801,
+      "step": 2241
+    },
+    {
+      "epoch": 60.59,
+      "learning_rate": 0.00022828125,
+      "loss": 0.1605,
+      "step": 2242
+    },
+    {
+      "epoch": 60.62,
+      "learning_rate": 0.000228125,
+      "loss": 0.1707,
+      "step": 2243
+    },
+    {
+      "epoch": 60.65,
+      "learning_rate": 0.00022796875,
+      "loss": 0.1437,
+      "step": 2244
+    },
+    {
+      "epoch": 60.68,
+      "learning_rate": 0.00022781250000000001,
+      "loss": 0.1678,
+      "step": 2245
+    },
+    {
+      "epoch": 60.7,
+      "learning_rate": 0.00022765625000000002,
+      "loss": 0.18,
+      "step": 2246
+    },
+    {
+      "epoch": 60.73,
+      "learning_rate": 0.0002275,
+      "loss": 0.1608,
+      "step": 2247
+    },
+    {
+      "epoch": 60.76,
+      "learning_rate": 0.00022734375,
+      "loss": 0.1129,
+      "step": 2248
+    },
+    {
+      "epoch": 60.78,
+      "learning_rate": 0.00022718749999999999,
+      "loss": 0.1968,
+      "step": 2249
+    },
+    {
+      "epoch": 60.81,
+      "learning_rate": 0.00022703125,
+      "loss": 0.155,
+      "step": 2250
+    },
+    {
+      "epoch": 60.84,
+      "learning_rate": 0.000226875,
+      "loss": 0.1864,
+      "step": 2251
+    },
+    {
+      "epoch": 60.86,
+      "learning_rate": 0.00022671875,
+      "loss": 0.1874,
+      "step": 2252
+    },
+    {
+      "epoch": 60.89,
+      "learning_rate": 0.0002265625,
+      "loss": 0.1195,
+      "step": 2253
+    },
+    {
+      "epoch": 60.92,
+      "learning_rate": 0.00022640625000000002,
+      "loss": 0.1462,
+      "step": 2254
+    },
+    {
+      "epoch": 60.95,
+      "learning_rate": 0.00022625000000000002,
+      "loss": 0.1603,
+      "step": 2255
+    },
+    {
+      "epoch": 60.97,
+      "learning_rate": 0.00022609375,
+      "loss": 0.1408,
+      "step": 2256
+    },
+    {
+      "epoch": 61.0,
+      "learning_rate": 0.0002259375,
+      "loss": 0.1843,
+      "step": 2257
+    },
+    {
+      "epoch": 61.03,
+      "learning_rate": 0.00022578125,
+      "loss": 0.1791,
+      "step": 2258
+    },
+    {
+      "epoch": 61.05,
+      "learning_rate": 0.000225625,
+      "loss": 0.173,
+      "step": 2259
+    },
+    {
+      "epoch": 61.08,
+      "learning_rate": 0.00022546875,
+      "loss": 0.1463,
+      "step": 2260
+    },
+    {
+      "epoch": 61.11,
+      "learning_rate": 0.0002253125,
+      "loss": 0.1584,
+      "step": 2261
+    },
+    {
+      "epoch": 61.14,
+      "learning_rate": 0.00022515625000000001,
+      "loss": 0.1378,
+      "step": 2262
+    },
+    {
+      "epoch": 61.16,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 0.1717,
+      "step": 2263
+    },
+    {
+      "epoch": 61.19,
+      "learning_rate": 0.00022484375,
+      "loss": 0.2092,
+      "step": 2264
+    },
+    {
+      "epoch": 61.22,
+      "learning_rate": 0.0002246875,
+      "loss": 0.1955,
+      "step": 2265
+    },
+    {
+      "epoch": 61.24,
+      "learning_rate": 0.00022453124999999999,
+      "loss": 0.1914,
+      "step": 2266
+    },
+    {
+      "epoch": 61.27,
+      "learning_rate": 0.000224375,
+      "loss": 0.1617,
+      "step": 2267
+    },
+    {
+      "epoch": 61.3,
+      "learning_rate": 0.00022421875,
+      "loss": 0.154,
+      "step": 2268
+    },
+    {
+      "epoch": 61.32,
+      "learning_rate": 0.0002240625,
+      "loss": 0.1252,
+      "step": 2269
+    },
+    {
+      "epoch": 61.35,
+      "learning_rate": 0.00022390625,
+      "loss": 0.1409,
+      "step": 2270
+    },
+    {
+      "epoch": 61.38,
+      "learning_rate": 0.00022375000000000002,
+      "loss": 0.1438,
+      "step": 2271
+    },
+    {
+      "epoch": 61.41,
+      "learning_rate": 0.00022359375000000002,
+      "loss": 0.1404,
+      "step": 2272
+    },
+    {
+      "epoch": 61.43,
+      "learning_rate": 0.0002234375,
+      "loss": 0.1585,
+      "step": 2273
+    },
+    {
+      "epoch": 61.46,
+      "learning_rate": 0.00022328124999999998,
+      "loss": 0.1741,
+      "step": 2274
+    },
+    {
+      "epoch": 61.49,
+      "learning_rate": 0.000223125,
+      "loss": 0.129,
+      "step": 2275
+    },
+    {
+      "epoch": 61.51,
+      "learning_rate": 0.00022296875,
+      "loss": 0.1803,
+      "step": 2276
+    },
+    {
+      "epoch": 61.54,
+      "learning_rate": 0.0002228125,
+      "loss": 0.1728,
+      "step": 2277
+    },
+    {
+      "epoch": 61.57,
+      "learning_rate": 0.00022265625,
+      "loss": 0.1703,
+      "step": 2278
+    },
+    {
+      "epoch": 61.59,
+      "learning_rate": 0.00022250000000000001,
+      "loss": 0.1477,
+      "step": 2279
+    },
+    {
+      "epoch": 61.62,
+      "learning_rate": 0.00022234375000000002,
+      "loss": 0.1724,
+      "step": 2280
+    },
+    {
+      "epoch": 61.65,
+      "learning_rate": 0.00022218750000000003,
+      "loss": 0.1423,
+      "step": 2281
+    },
+    {
+      "epoch": 61.68,
+      "learning_rate": 0.00022203125,
+      "loss": 0.1334,
+      "step": 2282
+    },
+    {
+      "epoch": 61.7,
+      "learning_rate": 0.00022187499999999999,
+      "loss": 0.2058,
+      "step": 2283
+    },
+    {
+      "epoch": 61.73,
+      "learning_rate": 0.00022171875,
+      "loss": 0.1626,
+      "step": 2284
+    },
+    {
+      "epoch": 61.76,
+      "learning_rate": 0.0002215625,
+      "loss": 0.1741,
+      "step": 2285
+    },
+    {
+      "epoch": 61.78,
+      "learning_rate": 0.00022140625,
+      "loss": 0.1195,
+      "step": 2286
+    },
+    {
+      "epoch": 61.81,
+      "learning_rate": 0.00022125,
+      "loss": 0.1378,
+      "step": 2287
+    },
+    {
+      "epoch": 61.84,
+      "learning_rate": 0.00022109375000000002,
+      "loss": 0.1746,
+      "step": 2288
+    },
+    {
+      "epoch": 61.86,
+      "learning_rate": 0.00022093750000000002,
+      "loss": 0.1378,
+      "step": 2289
+    },
+    {
+      "epoch": 61.89,
+      "learning_rate": 0.00022078125000000003,
+      "loss": 0.1602,
+      "step": 2290
+    },
+    {
+      "epoch": 61.92,
+      "learning_rate": 0.00022062499999999998,
+      "loss": 0.1521,
+      "step": 2291
+    },
+    {
+      "epoch": 61.95,
+      "learning_rate": 0.00022046875,
+      "loss": 0.1428,
+      "step": 2292
+    },
+    {
+      "epoch": 61.97,
+      "learning_rate": 0.0002203125,
+      "loss": 0.1939,
+      "step": 2293
+    },
+    {
+      "epoch": 62.0,
+      "learning_rate": 0.00022015625,
+      "loss": 0.1129,
+      "step": 2294
+    },
+    {
+      "epoch": 62.03,
+      "learning_rate": 0.00022,
+      "loss": 0.207,
+      "step": 2295
+    },
+    {
+      "epoch": 62.05,
+      "learning_rate": 0.00021984375000000001,
+      "loss": 0.1142,
+      "step": 2296
+    },
+    {
+      "epoch": 62.08,
+      "learning_rate": 0.00021968750000000002,
+      "loss": 0.1479,
+      "step": 2297
+    },
+    {
+      "epoch": 62.11,
+      "learning_rate": 0.00021953125000000003,
+      "loss": 0.1405,
+      "step": 2298
+    },
+    {
+      "epoch": 62.14,
+      "learning_rate": 0.00021937499999999998,
+      "loss": 0.1428,
+      "step": 2299
+    },
+    {
+      "epoch": 62.16,
+      "learning_rate": 0.00021921874999999999,
+      "loss": 0.1161,
+      "step": 2300
+    },
+    {
+      "epoch": 62.19,
+      "learning_rate": 0.0002190625,
+      "loss": 0.1788,
+      "step": 2301
+    },
+    {
+      "epoch": 62.22,
+      "learning_rate": 0.00021890625,
+      "loss": 0.1363,
+      "step": 2302
+    },
+    {
+      "epoch": 62.24,
+      "learning_rate": 0.00021875,
+      "loss": 0.1683,
+      "step": 2303
+    },
+    {
+      "epoch": 62.27,
+      "learning_rate": 0.00021859375,
+      "loss": 0.1577,
+      "step": 2304
+    },
+    {
+      "epoch": 62.3,
+      "learning_rate": 0.00021843750000000002,
+      "loss": 0.1034,
+      "step": 2305
+    },
+    {
+      "epoch": 62.32,
+      "learning_rate": 0.00021828125000000002,
+      "loss": 0.1566,
+      "step": 2306
+    },
+    {
+      "epoch": 62.35,
+      "learning_rate": 0.00021812500000000003,
+      "loss": 0.1535,
+      "step": 2307
+    },
+    {
+      "epoch": 62.38,
+      "learning_rate": 0.00021796874999999998,
+      "loss": 0.1436,
+      "step": 2308
+    },
+    {
+      "epoch": 62.41,
+      "learning_rate": 0.0002178125,
+      "loss": 0.1383,
+      "step": 2309
+    },
+    {
+      "epoch": 62.43,
+      "learning_rate": 0.00021765625,
+      "loss": 0.1988,
+      "step": 2310
+    },
+    {
+      "epoch": 62.46,
+      "learning_rate": 0.0002175,
+      "loss": 0.1635,
+      "step": 2311
+    },
+    {
+      "epoch": 62.49,
+      "learning_rate": 0.00021734375,
+      "loss": 0.1315,
+      "step": 2312
+    },
+    {
+      "epoch": 62.51,
+      "learning_rate": 0.00021718750000000001,
+      "loss": 0.1757,
+      "step": 2313
+    },
+    {
+      "epoch": 62.54,
+      "learning_rate": 0.00021703125000000002,
+      "loss": 0.1621,
+      "step": 2314
+    },
+    {
+      "epoch": 62.57,
+      "learning_rate": 0.00021687500000000003,
+      "loss": 0.1667,
+      "step": 2315
+    },
+    {
+      "epoch": 62.59,
+      "learning_rate": 0.00021671874999999998,
+      "loss": 0.149,
+      "step": 2316
+    },
+    {
+      "epoch": 62.62,
+      "learning_rate": 0.00021656249999999999,
+      "loss": 0.1509,
+      "step": 2317
+    },
+    {
+      "epoch": 62.65,
+      "learning_rate": 0.00021640625,
+      "loss": 0.1475,
+      "step": 2318
+    },
+    {
+      "epoch": 62.68,
+      "learning_rate": 0.00021625,
+      "loss": 0.1376,
+      "step": 2319
+    },
+    {
+      "epoch": 62.7,
+      "learning_rate": 0.00021609375,
+      "loss": 0.1845,
+      "step": 2320
+    },
+    {
+      "epoch": 62.73,
+      "learning_rate": 0.0002159375,
+      "loss": 0.204,
+      "step": 2321
+    },
+    {
+      "epoch": 62.76,
+      "learning_rate": 0.00021578125000000002,
+      "loss": 0.1274,
+      "step": 2322
+    },
+    {
+      "epoch": 62.78,
+      "learning_rate": 0.00021562500000000002,
+      "loss": 0.1378,
+      "step": 2323
+    },
+    {
+      "epoch": 62.81,
+      "learning_rate": 0.00021546875,
+      "loss": 0.1464,
+      "step": 2324
+    },
+    {
+      "epoch": 62.84,
+      "learning_rate": 0.00021531249999999998,
+      "loss": 0.1764,
+      "step": 2325
+    },
+    {
+      "epoch": 62.86,
+      "learning_rate": 0.00021515625,
+      "loss": 0.1549,
+      "step": 2326
+    },
+    {
+      "epoch": 62.89,
+      "learning_rate": 0.000215,
+      "loss": 0.1864,
+      "step": 2327
+    },
+    {
+      "epoch": 62.92,
+      "learning_rate": 0.00021484375,
+      "loss": 0.1483,
+      "step": 2328
+    },
+    {
+      "epoch": 62.95,
+      "learning_rate": 0.0002146875,
+      "loss": 0.1208,
+      "step": 2329
+    },
+    {
+      "epoch": 62.97,
+      "learning_rate": 0.00021453125000000001,
+      "loss": 0.147,
+      "step": 2330
+    },
+    {
+      "epoch": 63.0,
+      "learning_rate": 0.00021437500000000002,
+      "loss": 0.0936,
+      "step": 2331
+    },
+    {
+      "epoch": 63.03,
+      "learning_rate": 0.00021421875000000003,
+      "loss": 0.1675,
+      "step": 2332
+    },
+    {
+      "epoch": 63.05,
+      "learning_rate": 0.0002140625,
+      "loss": 0.1467,
+      "step": 2333
+    },
+    {
+      "epoch": 63.08,
+      "learning_rate": 0.00021390624999999999,
+      "loss": 0.1417,
+      "step": 2334
+    },
+    {
+      "epoch": 63.11,
+      "learning_rate": 0.00021375,
+      "loss": 0.157,
+      "step": 2335
+    },
+    {
+      "epoch": 63.14,
+      "learning_rate": 0.00021359375,
+      "loss": 0.14,
+      "step": 2336
+    },
+    {
+      "epoch": 63.16,
+      "learning_rate": 0.0002134375,
+      "loss": 0.137,
+      "step": 2337
+    },
+    {
+      "epoch": 63.19,
+      "learning_rate": 0.00021328125,
+      "loss": 0.1841,
+      "step": 2338
+    },
+    {
+      "epoch": 63.22,
+      "learning_rate": 0.00021312500000000002,
+      "loss": 0.1915,
+      "step": 2339
+    },
+    {
+      "epoch": 63.24,
+      "learning_rate": 0.00021296875000000002,
+      "loss": 0.1757,
+      "step": 2340
+    },
+    {
+      "epoch": 63.27,
+      "learning_rate": 0.0002128125,
+      "loss": 0.154,
+      "step": 2341
+    },
+    {
+      "epoch": 63.3,
+      "learning_rate": 0.00021265625,
+      "loss": 0.1648,
+      "step": 2342
+    },
+    {
+      "epoch": 63.32,
+      "learning_rate": 0.0002125,
+      "loss": 0.1482,
+      "step": 2343
+    },
+    {
+      "epoch": 63.35,
+      "learning_rate": 0.00021234375,
+      "loss": 0.2012,
+      "step": 2344
+    },
+    {
+      "epoch": 63.38,
+      "learning_rate": 0.0002121875,
+      "loss": 0.1665,
+      "step": 2345
+    },
+    {
+      "epoch": 63.41,
+      "learning_rate": 0.00021203125,
+      "loss": 0.1321,
+      "step": 2346
+    },
+    {
+      "epoch": 63.43,
+      "learning_rate": 0.00021187500000000001,
+      "loss": 0.1188,
+      "step": 2347
+    },
+    {
+      "epoch": 63.46,
+      "learning_rate": 0.00021171875000000002,
+      "loss": 0.174,
+      "step": 2348
+    },
+    {
+      "epoch": 63.49,
+      "learning_rate": 0.0002115625,
+      "loss": 0.1552,
+      "step": 2349
+    },
+    {
+      "epoch": 63.51,
+      "learning_rate": 0.00021140625,
+      "loss": 0.1338,
+      "step": 2350
+    },
+    {
+      "epoch": 63.54,
+      "learning_rate": 0.00021124999999999998,
+      "loss": 0.1734,
+      "step": 2351
+    },
+    {
+      "epoch": 63.57,
+      "learning_rate": 0.00021109375,
+      "loss": 0.1158,
+      "step": 2352
+    },
+    {
+      "epoch": 63.59,
+      "learning_rate": 0.0002109375,
+      "loss": 0.1748,
+      "step": 2353
+    },
+    {
+      "epoch": 63.62,
+      "learning_rate": 0.00021078125,
+      "loss": 0.1347,
+      "step": 2354
+    },
+    {
+      "epoch": 63.65,
+      "learning_rate": 0.000210625,
+      "loss": 0.1394,
+      "step": 2355
+    },
+    {
+      "epoch": 63.68,
+      "learning_rate": 0.00021046875000000002,
+      "loss": 0.1339,
+      "step": 2356
+    },
+    {
+      "epoch": 63.7,
+      "learning_rate": 0.00021031250000000002,
+      "loss": 0.1977,
+      "step": 2357
+    },
+    {
+      "epoch": 63.73,
+      "learning_rate": 0.00021015625,
+      "loss": 0.1649,
+      "step": 2358
+    },
+    {
+      "epoch": 63.76,
+      "learning_rate": 0.00021,
+      "loss": 0.1366,
+      "step": 2359
+    },
+    {
+      "epoch": 63.78,
+      "learning_rate": 0.00020984375,
+      "loss": 0.2111,
+      "step": 2360
+    },
+    {
+      "epoch": 63.81,
+      "learning_rate": 0.0002096875,
+      "loss": 0.2033,
+      "step": 2361
+    },
+    {
+      "epoch": 63.84,
+      "learning_rate": 0.00020953125,
+      "loss": 0.137,
+      "step": 2362
+    },
+    {
+      "epoch": 63.86,
+      "learning_rate": 0.000209375,
+      "loss": 0.1442,
+      "step": 2363
+    },
+    {
+      "epoch": 63.89,
+      "learning_rate": 0.00020921875,
+      "loss": 0.1618,
+      "step": 2364
+    },
+    {
+      "epoch": 63.92,
+      "learning_rate": 0.00020906250000000002,
+      "loss": 0.1821,
+      "step": 2365
+    },
+    {
+      "epoch": 63.95,
+      "learning_rate": 0.00020890625,
+      "loss": 0.1551,
+      "step": 2366
+    },
+    {
+      "epoch": 63.97,
+      "learning_rate": 0.00020875,
+      "loss": 0.1236,
+      "step": 2367
+    },
+    {
+      "epoch": 64.0,
+      "learning_rate": 0.00020859375,
+      "loss": 0.1082,
+      "step": 2368
+    },
+    {
+      "epoch": 64.03,
+      "learning_rate": 0.0002084375,
+      "loss": 0.1514,
+      "step": 2369
+    },
+    {
+      "epoch": 64.05,
+      "learning_rate": 0.00020828125,
+      "loss": 0.1213,
+      "step": 2370
+    },
+    {
+      "epoch": 64.08,
+      "learning_rate": 0.000208125,
+      "loss": 0.1605,
+      "step": 2371
+    },
+    {
+      "epoch": 64.11,
+      "learning_rate": 0.00020796875,
+      "loss": 0.1476,
+      "step": 2372
+    },
+    {
+      "epoch": 64.14,
+      "learning_rate": 0.00020781250000000002,
+      "loss": 0.1323,
+      "step": 2373
+    },
+    {
+      "epoch": 64.16,
+      "learning_rate": 0.00020765625,
+      "loss": 0.1109,
+      "step": 2374
+    },
+    {
+      "epoch": 64.19,
+      "learning_rate": 0.0002075,
+      "loss": 0.1488,
+      "step": 2375
+    },
+    {
+      "epoch": 64.22,
+      "learning_rate": 0.00020734375,
+      "loss": 0.1449,
+      "step": 2376
+    },
+    {
+      "epoch": 64.24,
+      "learning_rate": 0.0002071875,
+      "loss": 0.1768,
+      "step": 2377
+    },
+    {
+      "epoch": 64.27,
+      "learning_rate": 0.00020703125,
+      "loss": 0.1614,
+      "step": 2378
+    },
+    {
+      "epoch": 64.3,
+      "learning_rate": 0.000206875,
+      "loss": 0.1121,
+      "step": 2379
+    },
+    {
+      "epoch": 64.32,
+      "learning_rate": 0.00020671875,
+      "loss": 0.1206,
+      "step": 2380
+    },
+    {
+      "epoch": 64.35,
+      "learning_rate": 0.0002065625,
+      "loss": 0.1262,
+      "step": 2381
+    },
+    {
+      "epoch": 64.38,
+      "learning_rate": 0.00020640625000000002,
+      "loss": 0.1645,
+      "step": 2382
+    },
+    {
+      "epoch": 64.41,
+      "learning_rate": 0.00020625,
+      "loss": 0.1508,
+      "step": 2383
+    },
+    {
+      "epoch": 64.43,
+      "learning_rate": 0.00020609375,
+      "loss": 0.1705,
+      "step": 2384
+    },
+    {
+      "epoch": 64.46,
+      "learning_rate": 0.0002059375,
+      "loss": 0.1481,
+      "step": 2385
+    },
+    {
+      "epoch": 64.49,
+      "learning_rate": 0.00020578125,
+      "loss": 0.1288,
+      "step": 2386
+    },
+    {
+      "epoch": 64.51,
+      "learning_rate": 0.000205625,
+      "loss": 0.1538,
+      "step": 2387
+    },
+    {
+      "epoch": 64.54,
+      "learning_rate": 0.00020546875,
+      "loss": 0.1714,
+      "step": 2388
+    },
+    {
+      "epoch": 64.57,
+      "learning_rate": 0.0002053125,
+      "loss": 0.1557,
+      "step": 2389
+    },
+    {
+      "epoch": 64.59,
+      "learning_rate": 0.00020515625000000002,
+      "loss": 0.1365,
+      "step": 2390
+    },
+    {
+      "epoch": 64.62,
+      "learning_rate": 0.000205,
+      "loss": 0.1293,
+      "step": 2391
+    },
+    {
+      "epoch": 64.65,
+      "learning_rate": 0.00020484375,
+      "loss": 0.1235,
+      "step": 2392
+    },
+    {
+      "epoch": 64.68,
+      "learning_rate": 0.0002046875,
+      "loss": 0.088,
+      "step": 2393
+    },
+    {
+      "epoch": 64.7,
+      "learning_rate": 0.00020453125000000001,
+      "loss": 0.1811,
+      "step": 2394
+    },
+    {
+      "epoch": 64.73,
+      "learning_rate": 0.000204375,
+      "loss": 0.1743,
+      "step": 2395
+    },
+    {
+      "epoch": 64.76,
+      "learning_rate": 0.00020421875,
+      "loss": 0.1596,
+      "step": 2396
+    },
+    {
+      "epoch": 64.78,
+      "learning_rate": 0.0002040625,
+      "loss": 0.1486,
+      "step": 2397
+    },
+    {
+      "epoch": 64.81,
+      "learning_rate": 0.00020390625,
+      "loss": 0.1869,
+      "step": 2398
+    },
+    {
+      "epoch": 64.84,
+      "learning_rate": 0.00020375,
+      "loss": 0.1249,
+      "step": 2399
+    },
+    {
+      "epoch": 64.86,
+      "learning_rate": 0.00020359375,
+      "loss": 0.1812,
+      "step": 2400
+    },
+    {
+      "epoch": 64.89,
+      "learning_rate": 0.0002034375,
+      "loss": 0.1675,
+      "step": 2401
+    },
+    {
+      "epoch": 64.92,
+      "learning_rate": 0.00020328125,
+      "loss": 0.1449,
+      "step": 2402
+    },
+    {
+      "epoch": 64.95,
+      "learning_rate": 0.00020312500000000002,
+      "loss": 0.1622,
+      "step": 2403
+    },
+    {
+      "epoch": 64.97,
+      "learning_rate": 0.00020296875,
+      "loss": 0.1658,
+      "step": 2404
+    },
+    {
+      "epoch": 65.0,
+      "learning_rate": 0.0002028125,
+      "loss": 0.0548,
+      "step": 2405
+    },
+    {
+      "epoch": 65.03,
+      "learning_rate": 0.00020265625,
+      "loss": 0.1962,
+      "step": 2406
+    },
+    {
+      "epoch": 65.05,
+      "learning_rate": 0.00020250000000000002,
+      "loss": 0.1313,
+      "step": 2407
+    },
+    {
+      "epoch": 65.08,
+      "learning_rate": 0.00020234375,
+      "loss": 0.1547,
+      "step": 2408
+    },
+    {
+      "epoch": 65.11,
+      "learning_rate": 0.0002021875,
+      "loss": 0.1403,
+      "step": 2409
+    },
+    {
+      "epoch": 65.14,
+      "learning_rate": 0.00020203125,
+      "loss": 0.1204,
+      "step": 2410
+    },
+    {
+      "epoch": 65.16,
+      "learning_rate": 0.00020187500000000001,
+      "loss": 0.1881,
+      "step": 2411
+    },
+    {
+      "epoch": 65.19,
+      "learning_rate": 0.00020171875,
+      "loss": 0.1497,
+      "step": 2412
+    },
+    {
+      "epoch": 65.22,
+      "learning_rate": 0.0002015625,
+      "loss": 0.1427,
+      "step": 2413
+    },
+    {
+      "epoch": 65.24,
+      "learning_rate": 0.00020140625,
+      "loss": 0.1269,
+      "step": 2414
+    },
+    {
+      "epoch": 65.27,
+      "learning_rate": 0.00020125,
+      "loss": 0.1575,
+      "step": 2415
+    },
+    {
+      "epoch": 65.3,
+      "learning_rate": 0.00020125,
+      "loss": 0.163,
+      "step": 2416
+    },
+    {
+      "epoch": 65.32,
+      "learning_rate": 0.00020109375,
+      "loss": 0.1285,
+      "step": 2417
+    },
+    {
+      "epoch": 65.35,
+      "learning_rate": 0.0002009375,
+      "loss": 0.1321,
+      "step": 2418
+    },
+    {
+      "epoch": 65.38,
+      "learning_rate": 0.00020078125,
+      "loss": 0.1683,
+      "step": 2419
+    },
+    {
+      "epoch": 65.41,
+      "learning_rate": 0.000200625,
+      "loss": 0.1377,
+      "step": 2420
+    },
+    {
+      "epoch": 65.43,
+      "learning_rate": 0.00020046875000000002,
+      "loss": 0.1531,
+      "step": 2421
+    },
+    {
+      "epoch": 65.46,
+      "learning_rate": 0.0002003125,
+      "loss": 0.1499,
+      "step": 2422
+    },
+    {
+      "epoch": 65.49,
+      "learning_rate": 0.00020015625,
+      "loss": 0.1428,
+      "step": 2423
+    },
+    {
+      "epoch": 65.51,
+      "learning_rate": 0.0002,
+      "loss": 0.1433,
+      "step": 2424
+    },
+    {
+      "epoch": 65.54,
+      "learning_rate": 0.00019984375,
+      "loss": 0.156,
+      "step": 2425
+    },
+    {
+      "epoch": 65.57,
+      "learning_rate": 0.0001996875,
+      "loss": 0.1439,
+      "step": 2426
+    },
+    {
+      "epoch": 65.59,
+      "learning_rate": 0.00019953125,
+      "loss": 0.1562,
+      "step": 2427
+    },
+    {
+      "epoch": 65.62,
+      "learning_rate": 0.000199375,
+      "loss": 0.1432,
+      "step": 2428
+    },
+    {
+      "epoch": 65.65,
+      "learning_rate": 0.00019921875000000001,
+      "loss": 0.1448,
+      "step": 2429
+    },
+    {
+      "epoch": 65.68,
+      "learning_rate": 0.00019906250000000002,
+      "loss": 0.128,
+      "step": 2430
+    },
+    {
+      "epoch": 65.7,
+      "learning_rate": 0.00019890625,
+      "loss": 0.1648,
+      "step": 2431
+    },
+    {
+      "epoch": 65.73,
+      "learning_rate": 0.00019875,
+      "loss": 0.1333,
+      "step": 2432
+    },
+    {
+      "epoch": 65.76,
+      "learning_rate": 0.00019859375,
+      "loss": 0.1528,
+      "step": 2433
+    },
+    {
+      "epoch": 65.78,
+      "learning_rate": 0.0001984375,
+      "loss": 0.1682,
+      "step": 2434
+    },
+    {
+      "epoch": 65.81,
+      "learning_rate": 0.00019828125,
+      "loss": 0.1423,
+      "step": 2435
+    },
+    {
+      "epoch": 65.84,
+      "learning_rate": 0.000198125,
+      "loss": 0.1286,
+      "step": 2436
+    },
+    {
+      "epoch": 65.86,
+      "learning_rate": 0.00019796875,
+      "loss": 0.1174,
+      "step": 2437
+    },
+    {
+      "epoch": 65.89,
+      "learning_rate": 0.00019781250000000002,
+      "loss": 0.1597,
+      "step": 2438
+    },
+    {
+      "epoch": 65.92,
+      "learning_rate": 0.00019765625,
+      "loss": 0.1618,
+      "step": 2439
+    },
+    {
+      "epoch": 65.95,
+      "learning_rate": 0.0001975,
+      "loss": 0.1638,
+      "step": 2440
+    },
+    {
+      "epoch": 65.97,
+      "learning_rate": 0.00019734375,
+      "loss": 0.1444,
+      "step": 2441
+    },
+    {
+      "epoch": 66.0,
+      "learning_rate": 0.0001971875,
+      "loss": 0.127,
+      "step": 2442
+    },
+    {
+      "epoch": 66.03,
+      "learning_rate": 0.00019703125,
+      "loss": 0.1898,
+      "step": 2443
+    },
+    {
+      "epoch": 66.05,
+      "learning_rate": 0.000196875,
+      "loss": 0.1586,
+      "step": 2444
+    },
+    {
+      "epoch": 66.08,
+      "learning_rate": 0.00019671875,
+      "loss": 0.1614,
+      "step": 2445
+    },
+    {
+      "epoch": 66.11,
+      "learning_rate": 0.00019656250000000001,
+      "loss": 0.1576,
+      "step": 2446
+    },
+    {
+      "epoch": 66.14,
+      "learning_rate": 0.00019640625000000002,
+      "loss": 0.1702,
+      "step": 2447
+    },
+    {
+      "epoch": 66.16,
+      "learning_rate": 0.00019625,
+      "loss": 0.1722,
+      "step": 2448
+    },
+    {
+      "epoch": 66.19,
+      "learning_rate": 0.00019609375,
+      "loss": 0.1404,
+      "step": 2449
+    },
+    {
+      "epoch": 66.22,
+      "learning_rate": 0.00019593749999999999,
+      "loss": 0.1339,
+      "step": 2450
+    },
+    {
+      "epoch": 66.24,
+      "learning_rate": 0.00019578125,
+      "loss": 0.1364,
+      "step": 2451
+    },
+    {
+      "epoch": 66.27,
+      "learning_rate": 0.000195625,
+      "loss": 0.1402,
+      "step": 2452
+    },
+    {
+      "epoch": 66.3,
+      "learning_rate": 0.00019546875,
+      "loss": 0.1296,
+      "step": 2453
+    },
+    {
+      "epoch": 66.32,
+      "learning_rate": 0.0001953125,
+      "loss": 0.1587,
+      "step": 2454
+    },
+    {
+      "epoch": 66.35,
+      "learning_rate": 0.00019515625000000002,
+      "loss": 0.1574,
+      "step": 2455
+    },
+    {
+      "epoch": 66.38,
+      "learning_rate": 0.00019500000000000002,
+      "loss": 0.1581,
+      "step": 2456
+    },
+    {
+      "epoch": 66.41,
+      "learning_rate": 0.00019484375,
+      "loss": 0.1468,
+      "step": 2457
+    },
+    {
+      "epoch": 66.43,
+      "learning_rate": 0.0001946875,
+      "loss": 0.1408,
+      "step": 2458
+    },
+    {
+      "epoch": 66.46,
+      "learning_rate": 0.00019453125,
+      "loss": 0.1475,
+      "step": 2459
+    },
+    {
+      "epoch": 66.49,
+      "learning_rate": 0.000194375,
+      "loss": 0.1241,
+      "step": 2460
+    },
+    {
+      "epoch": 66.51,
+      "learning_rate": 0.00019421875,
+      "loss": 0.1159,
+      "step": 2461
+    },
+    {
+      "epoch": 66.54,
+      "learning_rate": 0.0001940625,
+      "loss": 0.1229,
+      "step": 2462
+    },
+    {
+      "epoch": 66.57,
+      "learning_rate": 0.00019390625000000001,
+      "loss": 0.1943,
+      "step": 2463
+    },
+    {
+      "epoch": 66.59,
+      "learning_rate": 0.00019375000000000002,
+      "loss": 0.1754,
+      "step": 2464
+    },
+    {
+      "epoch": 66.62,
+      "learning_rate": 0.00019359375,
+      "loss": 0.1498,
+      "step": 2465
+    },
+    {
+      "epoch": 66.65,
+      "learning_rate": 0.0001934375,
+      "loss": 0.1583,
+      "step": 2466
+    },
+    {
+      "epoch": 66.68,
+      "learning_rate": 0.00019328124999999999,
+      "loss": 0.1479,
+      "step": 2467
+    },
+    {
+      "epoch": 66.7,
+      "learning_rate": 0.000193125,
+      "loss": 0.1444,
+      "step": 2468
+    },
+    {
+      "epoch": 66.73,
+      "learning_rate": 0.00019296875,
+      "loss": 0.1648,
+      "step": 2469
+    },
+    {
+      "epoch": 66.76,
+      "learning_rate": 0.0001928125,
+      "loss": 0.1673,
+      "step": 2470
+    },
+    {
+      "epoch": 66.78,
+      "learning_rate": 0.00019265625,
+      "loss": 0.1874,
+      "step": 2471
+    },
+    {
+      "epoch": 66.81,
+      "learning_rate": 0.00019250000000000002,
+      "loss": 0.154,
+      "step": 2472
+    },
+    {
+      "epoch": 66.84,
+      "learning_rate": 0.00019234375000000002,
+      "loss": 0.1186,
+      "step": 2473
+    },
+    {
+      "epoch": 66.86,
+      "learning_rate": 0.0001921875,
+      "loss": 0.1399,
+      "step": 2474
+    },
+    {
+      "epoch": 66.89,
+      "learning_rate": 0.00019203124999999998,
+      "loss": 0.1488,
+      "step": 2475
+    },
+    {
+      "epoch": 66.92,
+      "learning_rate": 0.000191875,
+      "loss": 0.1339,
+      "step": 2476
+    },
+    {
+      "epoch": 66.95,
+      "learning_rate": 0.00019171875,
+      "loss": 0.1266,
+      "step": 2477
+    },
+    {
+      "epoch": 66.97,
+      "learning_rate": 0.0001915625,
+      "loss": 0.1211,
+      "step": 2478
+    },
+    {
+      "epoch": 67.0,
+      "learning_rate": 0.00019140625,
+      "loss": 0.1296,
+      "step": 2479
+    },
+    {
+      "epoch": 67.03,
+      "learning_rate": 0.00019125000000000001,
+      "loss": 0.1746,
+      "step": 2480
+    },
+    {
+      "epoch": 67.05,
+      "learning_rate": 0.00019109375000000002,
+      "loss": 0.1296,
+      "step": 2481
+    },
+    {
+      "epoch": 67.08,
+      "learning_rate": 0.00019093750000000003,
+      "loss": 0.1472,
+      "step": 2482
+    },
+    {
+      "epoch": 67.11,
+      "learning_rate": 0.00019078125,
+      "loss": 0.1619,
+      "step": 2483
+    },
+    {
+      "epoch": 67.14,
+      "learning_rate": 0.00019062499999999998,
+      "loss": 0.1358,
+      "step": 2484
+    },
+    {
+      "epoch": 67.16,
+      "learning_rate": 0.00019046875,
+      "loss": 0.1305,
+      "step": 2485
+    },
+    {
+      "epoch": 67.19,
+      "learning_rate": 0.0001903125,
+      "loss": 0.1462,
+      "step": 2486
+    },
+    {
+      "epoch": 67.22,
+      "learning_rate": 0.00019015625,
+      "loss": 0.1502,
+      "step": 2487
+    },
+    {
+      "epoch": 67.24,
+      "learning_rate": 0.00019,
+      "loss": 0.1417,
+      "step": 2488
+    },
+    {
+      "epoch": 67.27,
+      "learning_rate": 0.00018984375000000002,
+      "loss": 0.128,
+      "step": 2489
+    },
+    {
+      "epoch": 67.3,
+      "learning_rate": 0.00018968750000000002,
+      "loss": 0.1441,
+      "step": 2490
+    },
+    {
+      "epoch": 67.32,
+      "learning_rate": 0.00018953125000000003,
+      "loss": 0.172,
+      "step": 2491
+    },
+    {
+      "epoch": 67.35,
+      "learning_rate": 0.00018937499999999998,
+      "loss": 0.1643,
+      "step": 2492
+    },
+    {
+      "epoch": 67.38,
+      "learning_rate": 0.00018921875,
+      "loss": 0.1272,
+      "step": 2493
+    },
+    {
+      "epoch": 67.41,
+      "learning_rate": 0.0001890625,
+      "loss": 0.1527,
+      "step": 2494
+    },
+    {
+      "epoch": 67.43,
+      "learning_rate": 0.00018890625,
+      "loss": 0.1707,
+      "step": 2495
+    },
+    {
+      "epoch": 67.46,
+      "learning_rate": 0.00018875,
+      "loss": 0.1494,
+      "step": 2496
+    },
+    {
+      "epoch": 67.49,
+      "learning_rate": 0.00018859375,
+      "loss": 0.1494,
+      "step": 2497
+    },
+    {
+      "epoch": 67.51,
+      "learning_rate": 0.00018843750000000002,
+      "loss": 0.1172,
+      "step": 2498
+    },
+    {
+      "epoch": 67.54,
+      "learning_rate": 0.00018828125000000003,
+      "loss": 0.1586,
+      "step": 2499
+    },
+    {
+      "epoch": 67.57,
+      "learning_rate": 0.00018812499999999998,
+      "loss": 0.1318,
+      "step": 2500
+    },
+    {
+      "epoch": 67.57,
+      "eval_cer": 0.08009115578944212,
+      "eval_loss": 0.3818117380142212,
+      "eval_runtime": 11.6346,
+      "eval_samples_per_second": 141.56,
+      "eval_steps_per_second": 2.235,
+      "eval_wer": 0.3262179552650393,
+      "step": 2500
+    },
+    {
+      "epoch": 67.59,
+      "learning_rate": 0.00018796874999999998,
+      "loss": 0.155,
+      "step": 2501
+    },
+    {
+      "epoch": 67.62,
+      "learning_rate": 0.0001878125,
+      "loss": 0.1461,
+      "step": 2502
+    },
+    {
+      "epoch": 67.65,
+      "learning_rate": 0.00018765625,
+      "loss": 0.1303,
+      "step": 2503
+    },
+    {
+      "epoch": 67.68,
+      "learning_rate": 0.0001875,
+      "loss": 0.1167,
+      "step": 2504
+    },
+    {
+      "epoch": 67.7,
+      "learning_rate": 0.00018734375,
+      "loss": 0.1227,
+      "step": 2505
+    },
+    {
+      "epoch": 67.73,
+      "learning_rate": 0.00018718750000000002,
+      "loss": 0.1925,
+      "step": 2506
+    },
+    {
+      "epoch": 67.76,
+      "learning_rate": 0.00018703125000000002,
+      "loss": 0.1507,
+      "step": 2507
+    },
+    {
+      "epoch": 67.78,
+      "learning_rate": 0.00018687500000000003,
+      "loss": 0.1389,
+      "step": 2508
+    },
+    {
+      "epoch": 67.81,
+      "learning_rate": 0.00018671874999999998,
+      "loss": 0.1358,
+      "step": 2509
+    },
+    {
+      "epoch": 67.84,
+      "learning_rate": 0.0001865625,
+      "loss": 0.1157,
+      "step": 2510
+    },
+    {
+      "epoch": 67.86,
+      "learning_rate": 0.00018640625,
+      "loss": 0.1308,
+      "step": 2511
+    },
+    {
+      "epoch": 67.89,
+      "learning_rate": 0.00018625,
+      "loss": 0.137,
+      "step": 2512
+    },
+    {
+      "epoch": 67.92,
+      "learning_rate": 0.00018609375,
+      "loss": 0.1123,
+      "step": 2513
+    },
+    {
+      "epoch": 67.95,
+      "learning_rate": 0.0001859375,
+      "loss": 0.1253,
+      "step": 2514
+    },
+    {
+      "epoch": 67.97,
+      "learning_rate": 0.00018578125000000002,
+      "loss": 0.1305,
+      "step": 2515
+    },
+    {
+      "epoch": 68.0,
+      "learning_rate": 0.00018562500000000003,
+      "loss": 0.1963,
+      "step": 2516
+    },
+    {
+      "epoch": 68.03,
+      "learning_rate": 0.00018546874999999998,
+      "loss": 0.1412,
+      "step": 2517
+    },
+    {
+      "epoch": 68.05,
+      "learning_rate": 0.00018531249999999998,
+      "loss": 0.1438,
+      "step": 2518
+    },
+    {
+      "epoch": 68.08,
+      "learning_rate": 0.00018515625,
+      "loss": 0.1376,
+      "step": 2519
+    },
+    {
+      "epoch": 68.11,
+      "learning_rate": 0.000185,
+      "loss": 0.1631,
+      "step": 2520
+    },
+    {
+      "epoch": 68.14,
+      "learning_rate": 0.00018484375,
+      "loss": 0.1077,
+      "step": 2521
+    },
+    {
+      "epoch": 68.16,
+      "learning_rate": 0.0001846875,
+      "loss": 0.1655,
+      "step": 2522
+    },
+    {
+      "epoch": 68.19,
+      "learning_rate": 0.00018453125000000002,
+      "loss": 0.1354,
+      "step": 2523
+    },
+    {
+      "epoch": 68.22,
+      "learning_rate": 0.00018437500000000002,
+      "loss": 0.1854,
+      "step": 2524
+    },
+    {
+      "epoch": 68.24,
+      "learning_rate": 0.00018421875,
+      "loss": 0.1465,
+      "step": 2525
+    },
+    {
+      "epoch": 68.27,
+      "learning_rate": 0.00018406249999999998,
+      "loss": 0.1601,
+      "step": 2526
+    },
+    {
+      "epoch": 68.3,
+      "learning_rate": 0.00018390625,
+      "loss": 0.1467,
+      "step": 2527
+    },
+    {
+      "epoch": 68.32,
+      "learning_rate": 0.00018375,
+      "loss": 0.1599,
+      "step": 2528
+    },
+    {
+      "epoch": 68.35,
+      "learning_rate": 0.00018359375,
+      "loss": 0.1545,
+      "step": 2529
+    },
+    {
+      "epoch": 68.38,
+      "learning_rate": 0.0001834375,
+      "loss": 0.1382,
+      "step": 2530
+    },
+    {
+      "epoch": 68.41,
+      "learning_rate": 0.00018328125,
+      "loss": 0.1405,
+      "step": 2531
+    },
+    {
+      "epoch": 68.43,
+      "learning_rate": 0.00018312500000000002,
+      "loss": 0.1481,
+      "step": 2532
+    },
+    {
+      "epoch": 68.46,
+      "learning_rate": 0.00018296875000000003,
+      "loss": 0.1478,
+      "step": 2533
+    },
+    {
+      "epoch": 68.49,
+      "learning_rate": 0.0001828125,
+      "loss": 0.1458,
+      "step": 2534
+    },
+    {
+      "epoch": 68.51,
+      "learning_rate": 0.00018265624999999998,
+      "loss": 0.1321,
+      "step": 2535
+    },
+    {
+      "epoch": 68.54,
+      "learning_rate": 0.0001825,
+      "loss": 0.1121,
+      "step": 2536
+    },
+    {
+      "epoch": 68.57,
+      "learning_rate": 0.00018234375,
+      "loss": 0.1235,
+      "step": 2537
+    },
+    {
+      "epoch": 68.59,
+      "learning_rate": 0.0001821875,
+      "loss": 0.1541,
+      "step": 2538
+    },
+    {
+      "epoch": 68.62,
+      "learning_rate": 0.00018203125,
+      "loss": 0.1401,
+      "step": 2539
+    },
+    {
+      "epoch": 68.65,
+      "learning_rate": 0.00018187500000000002,
+      "loss": 0.1285,
+      "step": 2540
+    },
+    {
+      "epoch": 68.68,
+      "learning_rate": 0.00018171875000000002,
+      "loss": 0.1444,
+      "step": 2541
+    },
+    {
+      "epoch": 68.7,
+      "learning_rate": 0.0001815625,
+      "loss": 0.1727,
+      "step": 2542
+    },
+    {
+      "epoch": 68.73,
+      "learning_rate": 0.00018140625,
+      "loss": 0.1631,
+      "step": 2543
+    },
+    {
+      "epoch": 68.76,
+      "learning_rate": 0.00018125,
+      "loss": 0.1166,
+      "step": 2544
+    },
+    {
+      "epoch": 68.78,
+      "learning_rate": 0.00018109375,
+      "loss": 0.1382,
+      "step": 2545
+    },
+    {
+      "epoch": 68.81,
+      "learning_rate": 0.0001809375,
+      "loss": 0.137,
+      "step": 2546
+    },
+    {
+      "epoch": 68.84,
+      "learning_rate": 0.00018078125,
+      "loss": 0.1103,
+      "step": 2547
+    },
+    {
+      "epoch": 68.86,
+      "learning_rate": 0.000180625,
+      "loss": 0.1914,
+      "step": 2548
+    },
+    {
+      "epoch": 68.89,
+      "learning_rate": 0.00018046875000000002,
+      "loss": 0.1546,
+      "step": 2549
+    },
+    {
+      "epoch": 68.92,
+      "learning_rate": 0.0001803125,
+      "loss": 0.13,
+      "step": 2550
+    },
+    {
+      "epoch": 68.95,
+      "learning_rate": 0.00018015625,
+      "loss": 0.1775,
+      "step": 2551
+    },
+    {
+      "epoch": 68.97,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.1258,
+      "step": 2552
+    },
+    {
+      "epoch": 69.0,
+      "learning_rate": 0.00017984375,
+      "loss": 0.1129,
+      "step": 2553
+    },
+    {
+      "epoch": 69.03,
+      "learning_rate": 0.0001796875,
+      "loss": 0.1572,
+      "step": 2554
+    },
+    {
+      "epoch": 69.05,
+      "learning_rate": 0.00017953125,
+      "loss": 0.131,
+      "step": 2555
+    },
+    {
+      "epoch": 69.08,
+      "learning_rate": 0.000179375,
+      "loss": 0.1443,
+      "step": 2556
+    },
+    {
+      "epoch": 69.11,
+      "learning_rate": 0.00017921875000000002,
+      "loss": 0.163,
+      "step": 2557
+    },
+    {
+      "epoch": 69.14,
+      "learning_rate": 0.00017906250000000002,
+      "loss": 0.1397,
+      "step": 2558
+    },
+    {
+      "epoch": 69.16,
+      "learning_rate": 0.00017890625,
+      "loss": 0.1023,
+      "step": 2559
+    },
+    {
+      "epoch": 69.19,
+      "learning_rate": 0.00017875,
+      "loss": 0.1305,
+      "step": 2560
+    },
+    {
+      "epoch": 69.22,
+      "learning_rate": 0.00017859375,
+      "loss": 0.1524,
+      "step": 2561
+    },
+    {
+      "epoch": 69.24,
+      "learning_rate": 0.0001784375,
+      "loss": 0.1269,
+      "step": 2562
+    },
+    {
+      "epoch": 69.27,
+      "learning_rate": 0.00017828125,
+      "loss": 0.177,
+      "step": 2563
+    },
+    {
+      "epoch": 69.3,
+      "learning_rate": 0.000178125,
+      "loss": 0.1427,
+      "step": 2564
+    },
+    {
+      "epoch": 69.32,
+      "learning_rate": 0.00017796875,
+      "loss": 0.1498,
+      "step": 2565
+    },
+    {
+      "epoch": 69.35,
+      "learning_rate": 0.00017781250000000002,
+      "loss": 0.1193,
+      "step": 2566
+    },
+    {
+      "epoch": 69.38,
+      "learning_rate": 0.00017765625,
+      "loss": 0.1467,
+      "step": 2567
+    },
+    {
+      "epoch": 69.41,
+      "learning_rate": 0.0001775,
+      "loss": 0.142,
+      "step": 2568
+    },
+    {
+      "epoch": 69.43,
+      "learning_rate": 0.00017734375,
+      "loss": 0.12,
+      "step": 2569
+    },
+    {
+      "epoch": 69.46,
+      "learning_rate": 0.0001771875,
+      "loss": 0.1323,
+      "step": 2570
+    },
+    {
+      "epoch": 69.49,
+      "learning_rate": 0.00017703125,
+      "loss": 0.1289,
+      "step": 2571
+    },
+    {
+      "epoch": 69.51,
+      "learning_rate": 0.000176875,
+      "loss": 0.1312,
+      "step": 2572
+    },
+    {
+      "epoch": 69.54,
+      "learning_rate": 0.00017671875,
+      "loss": 0.1307,
+      "step": 2573
+    },
+    {
+      "epoch": 69.57,
+      "learning_rate": 0.00017656250000000002,
+      "loss": 0.1282,
+      "step": 2574
+    },
+    {
+      "epoch": 69.59,
+      "learning_rate": 0.00017640625,
+      "loss": 0.1512,
+      "step": 2575
+    },
+    {
+      "epoch": 69.62,
+      "learning_rate": 0.00017625,
+      "loss": 0.1178,
+      "step": 2576
+    },
+    {
+      "epoch": 69.65,
+      "learning_rate": 0.00017609375,
+      "loss": 0.1454,
+      "step": 2577
+    },
+    {
+      "epoch": 69.68,
+      "learning_rate": 0.0001759375,
+      "loss": 0.1224,
+      "step": 2578
+    },
+    {
+      "epoch": 69.7,
+      "learning_rate": 0.00017578125,
+      "loss": 0.1821,
+      "step": 2579
+    },
+    {
+      "epoch": 69.73,
+      "learning_rate": 0.000175625,
+      "loss": 0.1333,
+      "step": 2580
+    },
+    {
+      "epoch": 69.76,
+      "learning_rate": 0.00017546875,
+      "loss": 0.1403,
+      "step": 2581
+    },
+    {
+      "epoch": 69.78,
+      "learning_rate": 0.0001753125,
+      "loss": 0.1598,
+      "step": 2582
+    },
+    {
+      "epoch": 69.81,
+      "learning_rate": 0.00017515625000000002,
+      "loss": 0.11,
+      "step": 2583
+    },
+    {
+      "epoch": 69.84,
+      "learning_rate": 0.000175,
+      "loss": 0.1107,
+      "step": 2584
+    },
+    {
+      "epoch": 69.86,
+      "learning_rate": 0.00017484375,
+      "loss": 0.1604,
+      "step": 2585
+    },
+    {
+      "epoch": 69.89,
+      "learning_rate": 0.0001746875,
+      "loss": 0.1204,
+      "step": 2586
+    },
+    {
+      "epoch": 69.92,
+      "learning_rate": 0.00017453125,
+      "loss": 0.1135,
+      "step": 2587
+    },
+    {
+      "epoch": 69.95,
+      "learning_rate": 0.000174375,
+      "loss": 0.1185,
+      "step": 2588
+    },
+    {
+      "epoch": 69.97,
+      "learning_rate": 0.00017421875,
+      "loss": 0.1368,
+      "step": 2589
+    },
+    {
+      "epoch": 70.0,
+      "learning_rate": 0.0001740625,
+      "loss": 0.0557,
+      "step": 2590
+    },
+    {
+      "epoch": 70.03,
+      "learning_rate": 0.00017390625000000002,
+      "loss": 0.1763,
+      "step": 2591
+    },
+    {
+      "epoch": 70.05,
+      "learning_rate": 0.00017375,
+      "loss": 0.1455,
+      "step": 2592
+    },
+    {
+      "epoch": 70.08,
+      "learning_rate": 0.00017359375,
+      "loss": 0.1223,
+      "step": 2593
+    },
+    {
+      "epoch": 70.11,
+      "learning_rate": 0.0001734375,
+      "loss": 0.1447,
+      "step": 2594
+    },
+    {
+      "epoch": 70.14,
+      "learning_rate": 0.00017328125000000001,
+      "loss": 0.1317,
+      "step": 2595
+    },
+    {
+      "epoch": 70.16,
+      "learning_rate": 0.000173125,
+      "loss": 0.1304,
+      "step": 2596
+    },
+    {
+      "epoch": 70.19,
+      "learning_rate": 0.00017296875,
+      "loss": 0.1425,
+      "step": 2597
+    },
+    {
+      "epoch": 70.22,
+      "learning_rate": 0.0001728125,
+      "loss": 0.1196,
+      "step": 2598
+    },
+    {
+      "epoch": 70.24,
+      "learning_rate": 0.00017265625,
+      "loss": 0.155,
+      "step": 2599
+    },
+    {
+      "epoch": 70.27,
+      "learning_rate": 0.0001725,
+      "loss": 0.1406,
+      "step": 2600
+    },
+    {
+      "epoch": 70.3,
+      "learning_rate": 0.00017234375,
+      "loss": 0.1213,
+      "step": 2601
+    },
+    {
+      "epoch": 70.32,
+      "learning_rate": 0.0001721875,
+      "loss": 0.1114,
+      "step": 2602
+    },
+    {
+      "epoch": 70.35,
+      "learning_rate": 0.00017203125,
+      "loss": 0.1355,
+      "step": 2603
+    },
+    {
+      "epoch": 70.38,
+      "learning_rate": 0.000171875,
+      "loss": 0.1096,
+      "step": 2604
+    },
+    {
+      "epoch": 70.41,
+      "learning_rate": 0.00017171875,
+      "loss": 0.1147,
+      "step": 2605
+    },
+    {
+      "epoch": 70.43,
+      "learning_rate": 0.0001715625,
+      "loss": 0.1511,
+      "step": 2606
+    },
+    {
+      "epoch": 70.46,
+      "learning_rate": 0.00017140625,
+      "loss": 0.1069,
+      "step": 2607
+    },
+    {
+      "epoch": 70.49,
+      "learning_rate": 0.00017125000000000002,
+      "loss": 0.0884,
+      "step": 2608
+    },
+    {
+      "epoch": 70.51,
+      "learning_rate": 0.00017109375,
+      "loss": 0.0956,
+      "step": 2609
+    },
+    {
+      "epoch": 70.54,
+      "learning_rate": 0.0001709375,
+      "loss": 0.1214,
+      "step": 2610
+    },
+    {
+      "epoch": 70.57,
+      "learning_rate": 0.00017078125,
+      "loss": 0.1339,
+      "step": 2611
+    },
+    {
+      "epoch": 70.59,
+      "learning_rate": 0.00017062500000000001,
+      "loss": 0.1606,
+      "step": 2612
+    },
+    {
+      "epoch": 70.62,
+      "learning_rate": 0.00017046875,
+      "loss": 0.1384,
+      "step": 2613
+    },
+    {
+      "epoch": 70.65,
+      "learning_rate": 0.0001703125,
+      "loss": 0.1082,
+      "step": 2614
+    },
+    {
+      "epoch": 70.68,
+      "learning_rate": 0.00017015625,
+      "loss": 0.1147,
+      "step": 2615
+    },
+    {
+      "epoch": 70.7,
+      "learning_rate": 0.00017,
+      "loss": 0.1796,
+      "step": 2616
+    },
+    {
+      "epoch": 70.73,
+      "learning_rate": 0.00016984375,
+      "loss": 0.1164,
+      "step": 2617
+    },
+    {
+      "epoch": 70.76,
+      "learning_rate": 0.0001696875,
+      "loss": 0.1364,
+      "step": 2618
+    },
+    {
+      "epoch": 70.78,
+      "learning_rate": 0.00016953125,
+      "loss": 0.1096,
+      "step": 2619
+    },
+    {
+      "epoch": 70.81,
+      "learning_rate": 0.000169375,
+      "loss": 0.1036,
+      "step": 2620
+    },
+    {
+      "epoch": 70.84,
+      "learning_rate": 0.00016921875000000002,
+      "loss": 0.1247,
+      "step": 2621
+    },
+    {
+      "epoch": 70.86,
+      "learning_rate": 0.0001690625,
+      "loss": 0.1371,
+      "step": 2622
+    },
+    {
+      "epoch": 70.89,
+      "learning_rate": 0.00016890625,
+      "loss": 0.1191,
+      "step": 2623
+    },
+    {
+      "epoch": 70.92,
+      "learning_rate": 0.00016875,
+      "loss": 0.1143,
+      "step": 2624
+    },
+    {
+      "epoch": 70.95,
+      "learning_rate": 0.00016859375,
+      "loss": 0.1427,
+      "step": 2625
+    },
+    {
+      "epoch": 70.97,
+      "learning_rate": 0.0001684375,
+      "loss": 0.1288,
+      "step": 2626
+    },
+    {
+      "epoch": 71.0,
+      "learning_rate": 0.00016828125,
+      "loss": 0.0732,
+      "step": 2627
+    },
+    {
+      "epoch": 71.03,
+      "learning_rate": 0.000168125,
+      "loss": 0.1452,
+      "step": 2628
+    },
+    {
+      "epoch": 71.05,
+      "learning_rate": 0.00016796875000000001,
+      "loss": 0.1382,
+      "step": 2629
+    },
+    {
+      "epoch": 71.08,
+      "learning_rate": 0.00016781250000000002,
+      "loss": 0.1344,
+      "step": 2630
+    },
+    {
+      "epoch": 71.11,
+      "learning_rate": 0.00016765625,
+      "loss": 0.138,
+      "step": 2631
+    },
+    {
+      "epoch": 71.14,
+      "learning_rate": 0.0001675,
+      "loss": 0.1162,
+      "step": 2632
+    },
+    {
+      "epoch": 71.16,
+      "learning_rate": 0.00016734375,
+      "loss": 0.107,
+      "step": 2633
+    },
+    {
+      "epoch": 71.19,
+      "learning_rate": 0.0001671875,
+      "loss": 0.1347,
+      "step": 2634
+    },
+    {
+      "epoch": 71.22,
+      "learning_rate": 0.00016703125,
+      "loss": 0.0989,
+      "step": 2635
+    },
+    {
+      "epoch": 71.24,
+      "learning_rate": 0.000166875,
+      "loss": 0.128,
+      "step": 2636
+    },
+    {
+      "epoch": 71.27,
+      "learning_rate": 0.00016671875,
+      "loss": 0.1567,
+      "step": 2637
+    },
+    {
+      "epoch": 71.3,
+      "learning_rate": 0.00016656250000000002,
+      "loss": 0.1275,
+      "step": 2638
+    },
+    {
+      "epoch": 71.32,
+      "learning_rate": 0.00016640625,
+      "loss": 0.1022,
+      "step": 2639
+    },
+    {
+      "epoch": 71.35,
+      "learning_rate": 0.00016625,
+      "loss": 0.1142,
+      "step": 2640
+    },
+    {
+      "epoch": 71.38,
+      "learning_rate": 0.00016609375,
+      "loss": 0.1532,
+      "step": 2641
+    },
+    {
+      "epoch": 71.41,
+      "learning_rate": 0.0001659375,
+      "loss": 0.1534,
+      "step": 2642
+    },
+    {
+      "epoch": 71.43,
+      "learning_rate": 0.00016578125,
+      "loss": 0.1094,
+      "step": 2643
+    },
+    {
+      "epoch": 71.46,
+      "learning_rate": 0.000165625,
+      "loss": 0.1442,
+      "step": 2644
+    },
+    {
+      "epoch": 71.49,
+      "learning_rate": 0.00016546875,
+      "loss": 0.0795,
+      "step": 2645
+    },
+    {
+      "epoch": 71.51,
+      "learning_rate": 0.0001653125,
+      "loss": 0.1041,
+      "step": 2646
+    },
+    {
+      "epoch": 71.54,
+      "learning_rate": 0.00016515625000000002,
+      "loss": 0.1424,
+      "step": 2647
+    },
+    {
+      "epoch": 71.57,
+      "learning_rate": 0.000165,
+      "loss": 0.1438,
+      "step": 2648
+    },
+    {
+      "epoch": 71.59,
+      "learning_rate": 0.00016484375,
+      "loss": 0.1116,
+      "step": 2649
+    },
+    {
+      "epoch": 71.62,
+      "learning_rate": 0.00016468749999999998,
+      "loss": 0.1284,
+      "step": 2650
+    },
+    {
+      "epoch": 71.65,
+      "learning_rate": 0.00016453125,
+      "loss": 0.1358,
+      "step": 2651
+    },
+    {
+      "epoch": 71.68,
+      "learning_rate": 0.000164375,
+      "loss": 0.0985,
+      "step": 2652
+    },
+    {
+      "epoch": 71.7,
+      "learning_rate": 0.00016421875,
+      "loss": 0.1412,
+      "step": 2653
+    },
+    {
+      "epoch": 71.73,
+      "learning_rate": 0.0001640625,
+      "loss": 0.1142,
+      "step": 2654
+    },
+    {
+      "epoch": 71.76,
+      "learning_rate": 0.00016390625000000002,
+      "loss": 0.135,
+      "step": 2655
+    },
+    {
+      "epoch": 71.78,
+      "learning_rate": 0.00016375000000000002,
+      "loss": 0.1183,
+      "step": 2656
+    },
+    {
+      "epoch": 71.81,
+      "learning_rate": 0.00016359375,
+      "loss": 0.1262,
+      "step": 2657
+    },
+    {
+      "epoch": 71.84,
+      "learning_rate": 0.0001634375,
+      "loss": 0.1196,
+      "step": 2658
+    },
+    {
+      "epoch": 71.86,
+      "learning_rate": 0.00016328125,
+      "loss": 0.1321,
+      "step": 2659
+    },
+    {
+      "epoch": 71.89,
+      "learning_rate": 0.000163125,
+      "loss": 0.1424,
+      "step": 2660
+    },
+    {
+      "epoch": 71.92,
+      "learning_rate": 0.00016296875,
+      "loss": 0.1175,
+      "step": 2661
+    },
+    {
+      "epoch": 71.95,
+      "learning_rate": 0.0001628125,
+      "loss": 0.1188,
+      "step": 2662
+    },
+    {
+      "epoch": 71.97,
+      "learning_rate": 0.00016265625,
+      "loss": 0.1207,
+      "step": 2663
+    },
+    {
+      "epoch": 72.0,
+      "learning_rate": 0.00016250000000000002,
+      "loss": 0.1063,
+      "step": 2664
+    },
+    {
+      "epoch": 72.03,
+      "learning_rate": 0.00016234375,
+      "loss": 0.1488,
+      "step": 2665
+    },
+    {
+      "epoch": 72.05,
+      "learning_rate": 0.0001621875,
+      "loss": 0.1219,
+      "step": 2666
+    },
+    {
+      "epoch": 72.08,
+      "learning_rate": 0.00016203124999999998,
+      "loss": 0.1367,
+      "step": 2667
+    },
+    {
+      "epoch": 72.11,
+      "learning_rate": 0.000161875,
+      "loss": 0.1287,
+      "step": 2668
+    },
+    {
+      "epoch": 72.14,
+      "learning_rate": 0.00016171875,
+      "loss": 0.1033,
+      "step": 2669
+    },
+    {
+      "epoch": 72.16,
+      "learning_rate": 0.0001615625,
+      "loss": 0.1304,
+      "step": 2670
+    },
+    {
+      "epoch": 72.19,
+      "learning_rate": 0.00016140625,
+      "loss": 0.1225,
+      "step": 2671
+    },
+    {
+      "epoch": 72.22,
+      "learning_rate": 0.00016125000000000002,
+      "loss": 0.1429,
+      "step": 2672
+    },
+    {
+      "epoch": 72.24,
+      "learning_rate": 0.00016109375000000002,
+      "loss": 0.153,
+      "step": 2673
+    },
+    {
+      "epoch": 72.27,
+      "learning_rate": 0.0001609375,
+      "loss": 0.1368,
+      "step": 2674
+    },
+    {
+      "epoch": 72.3,
+      "learning_rate": 0.00016078124999999998,
+      "loss": 0.1313,
+      "step": 2675
+    },
+    {
+      "epoch": 72.32,
+      "learning_rate": 0.000160625,
+      "loss": 0.0957,
+      "step": 2676
+    },
+    {
+      "epoch": 72.35,
+      "learning_rate": 0.00016046875,
+      "loss": 0.1225,
+      "step": 2677
+    },
+    {
+      "epoch": 72.38,
+      "learning_rate": 0.0001603125,
+      "loss": 0.1601,
+      "step": 2678
+    },
+    {
+      "epoch": 72.41,
+      "learning_rate": 0.00016015625,
+      "loss": 0.1187,
+      "step": 2679
+    },
+    {
+      "epoch": 72.43,
+      "learning_rate": 0.00016,
+      "loss": 0.1479,
+      "step": 2680
+    },
+    {
+      "epoch": 72.46,
+      "learning_rate": 0.00015984375000000002,
+      "loss": 0.1206,
+      "step": 2681
+    },
+    {
+      "epoch": 72.49,
+      "learning_rate": 0.00015968750000000003,
+      "loss": 0.129,
+      "step": 2682
+    },
+    {
+      "epoch": 72.51,
+      "learning_rate": 0.00015953125,
+      "loss": 0.1418,
+      "step": 2683
+    },
+    {
+      "epoch": 72.54,
+      "learning_rate": 0.00015937499999999998,
+      "loss": 0.1424,
+      "step": 2684
+    },
+    {
+      "epoch": 72.57,
+      "learning_rate": 0.00015921875,
+      "loss": 0.1174,
+      "step": 2685
+    },
+    {
+      "epoch": 72.59,
+      "learning_rate": 0.0001590625,
+      "loss": 0.1434,
+      "step": 2686
+    },
+    {
+      "epoch": 72.62,
+      "learning_rate": 0.00015890625,
+      "loss": 0.1258,
+      "step": 2687
+    },
+    {
+      "epoch": 72.65,
+      "learning_rate": 0.00015875,
+      "loss": 0.1158,
+      "step": 2688
+    },
+    {
+      "epoch": 72.68,
+      "learning_rate": 0.00015859375000000002,
+      "loss": 0.0993,
+      "step": 2689
+    },
+    {
+      "epoch": 72.7,
+      "learning_rate": 0.00015843750000000002,
+      "loss": 0.1555,
+      "step": 2690
+    },
+    {
+      "epoch": 72.73,
+      "learning_rate": 0.00015828125000000003,
+      "loss": 0.1351,
+      "step": 2691
+    },
+    {
+      "epoch": 72.76,
+      "learning_rate": 0.00015812499999999998,
+      "loss": 0.13,
+      "step": 2692
+    },
+    {
+      "epoch": 72.78,
+      "learning_rate": 0.00015796875,
+      "loss": 0.1492,
+      "step": 2693
+    },
+    {
+      "epoch": 72.81,
+      "learning_rate": 0.0001578125,
+      "loss": 0.141,
+      "step": 2694
+    },
+    {
+      "epoch": 72.84,
+      "learning_rate": 0.00015765625,
+      "loss": 0.0839,
+      "step": 2695
+    },
+    {
+      "epoch": 72.86,
+      "learning_rate": 0.0001575,
+      "loss": 0.1342,
+      "step": 2696
+    },
+    {
+      "epoch": 72.89,
+      "learning_rate": 0.00015734375,
+      "loss": 0.1065,
+      "step": 2697
+    },
+    {
+      "epoch": 72.92,
+      "learning_rate": 0.00015718750000000002,
+      "loss": 0.1371,
+      "step": 2698
+    },
+    {
+      "epoch": 72.95,
+      "learning_rate": 0.00015703125000000003,
+      "loss": 0.1324,
+      "step": 2699
+    },
+    {
+      "epoch": 72.97,
+      "learning_rate": 0.00015687499999999998,
+      "loss": 0.0829,
+      "step": 2700
+    },
+    {
+      "epoch": 73.0,
+      "learning_rate": 0.00015671874999999998,
+      "loss": 0.066,
+      "step": 2701
+    },
+    {
+      "epoch": 73.03,
+      "learning_rate": 0.0001565625,
+      "loss": 0.1457,
+      "step": 2702
+    },
+    {
+      "epoch": 73.05,
+      "learning_rate": 0.00015640625,
+      "loss": 0.1457,
+      "step": 2703
+    },
+    {
+      "epoch": 73.08,
+      "learning_rate": 0.00015625,
+      "loss": 0.1254,
+      "step": 2704
+    },
+    {
+      "epoch": 73.11,
+      "learning_rate": 0.00015609375,
+      "loss": 0.1232,
+      "step": 2705
+    },
+    {
+      "epoch": 73.14,
+      "learning_rate": 0.00015593750000000002,
+      "loss": 0.127,
+      "step": 2706
+    },
+    {
+      "epoch": 73.16,
+      "learning_rate": 0.00015578125000000002,
+      "loss": 0.1143,
+      "step": 2707
+    },
+    {
+      "epoch": 73.19,
+      "learning_rate": 0.00015562500000000003,
+      "loss": 0.1051,
+      "step": 2708
+    },
+    {
+      "epoch": 73.22,
+      "learning_rate": 0.00015546874999999998,
+      "loss": 0.1664,
+      "step": 2709
+    },
+    {
+      "epoch": 73.24,
+      "learning_rate": 0.0001553125,
+      "loss": 0.1397,
+      "step": 2710
+    },
+    {
+      "epoch": 73.27,
+      "learning_rate": 0.00015515625,
+      "loss": 0.12,
+      "step": 2711
+    },
+    {
+      "epoch": 73.3,
+      "learning_rate": 0.000155,
+      "loss": 0.1428,
+      "step": 2712
+    },
+    {
+      "epoch": 73.32,
+      "learning_rate": 0.00015484375,
+      "loss": 0.0995,
+      "step": 2713
+    },
+    {
+      "epoch": 73.35,
+      "learning_rate": 0.0001546875,
+      "loss": 0.1059,
+      "step": 2714
+    },
+    {
+      "epoch": 73.38,
+      "learning_rate": 0.00015453125000000002,
+      "loss": 0.1237,
+      "step": 2715
+    },
+    {
+      "epoch": 73.41,
+      "learning_rate": 0.00015437500000000003,
+      "loss": 0.1277,
+      "step": 2716
+    },
+    {
+      "epoch": 73.43,
+      "learning_rate": 0.00015421874999999998,
+      "loss": 0.1221,
+      "step": 2717
+    },
+    {
+      "epoch": 73.46,
+      "learning_rate": 0.00015406249999999998,
+      "loss": 0.1443,
+      "step": 2718
+    },
+    {
+      "epoch": 73.49,
+      "learning_rate": 0.00015390625,
+      "loss": 0.0994,
+      "step": 2719
+    },
+    {
+      "epoch": 73.51,
+      "learning_rate": 0.00015375,
+      "loss": 0.0863,
+      "step": 2720
+    },
+    {
+      "epoch": 73.54,
+      "learning_rate": 0.00015359375,
+      "loss": 0.1399,
+      "step": 2721
+    },
+    {
+      "epoch": 73.57,
+      "learning_rate": 0.0001534375,
+      "loss": 0.1062,
+      "step": 2722
+    },
+    {
+      "epoch": 73.59,
+      "learning_rate": 0.00015328125000000002,
+      "loss": 0.1396,
+      "step": 2723
+    },
+    {
+      "epoch": 73.62,
+      "learning_rate": 0.00015312500000000002,
+      "loss": 0.0969,
+      "step": 2724
+    },
+    {
+      "epoch": 73.65,
+      "learning_rate": 0.00015296875,
+      "loss": 0.1427,
+      "step": 2725
+    },
+    {
+      "epoch": 73.68,
+      "learning_rate": 0.00015281249999999998,
+      "loss": 0.111,
+      "step": 2726
+    },
+    {
+      "epoch": 73.7,
+      "learning_rate": 0.00015265625,
+      "loss": 0.154,
+      "step": 2727
+    },
+    {
+      "epoch": 73.73,
+      "learning_rate": 0.0001525,
+      "loss": 0.1046,
+      "step": 2728
+    },
+    {
+      "epoch": 73.76,
+      "learning_rate": 0.00015234375,
+      "loss": 0.1227,
+      "step": 2729
+    },
+    {
+      "epoch": 73.78,
+      "learning_rate": 0.0001521875,
+      "loss": 0.161,
+      "step": 2730
+    },
+    {
+      "epoch": 73.81,
+      "learning_rate": 0.00015203125,
+      "loss": 0.121,
+      "step": 2731
+    },
+    {
+      "epoch": 73.84,
+      "learning_rate": 0.00015187500000000002,
+      "loss": 0.0868,
+      "step": 2732
+    },
+    {
+      "epoch": 73.86,
+      "learning_rate": 0.00015171875000000003,
+      "loss": 0.1297,
+      "step": 2733
+    },
+    {
+      "epoch": 73.89,
+      "learning_rate": 0.0001515625,
+      "loss": 0.1183,
+      "step": 2734
+    },
+    {
+      "epoch": 73.92,
+      "learning_rate": 0.00015140624999999998,
+      "loss": 0.1041,
+      "step": 2735
+    },
+    {
+      "epoch": 73.95,
+      "learning_rate": 0.00015125,
+      "loss": 0.1266,
+      "step": 2736
+    },
+    {
+      "epoch": 73.97,
+      "learning_rate": 0.00015109375,
+      "loss": 0.1053,
+      "step": 2737
+    },
+    {
+      "epoch": 74.0,
+      "learning_rate": 0.0001509375,
+      "loss": 0.0427,
+      "step": 2738
+    },
+    {
+      "epoch": 74.03,
+      "learning_rate": 0.00015078125,
+      "loss": 0.1066,
+      "step": 2739
+    },
+    {
+      "epoch": 74.05,
+      "learning_rate": 0.00015062500000000002,
+      "loss": 0.1107,
+      "step": 2740
+    },
+    {
+      "epoch": 74.08,
+      "learning_rate": 0.00015046875000000002,
+      "loss": 0.1108,
+      "step": 2741
+    },
+    {
+      "epoch": 74.11,
+      "learning_rate": 0.0001503125,
+      "loss": 0.1266,
+      "step": 2742
+    },
+    {
+      "epoch": 74.14,
+      "learning_rate": 0.00015015625,
+      "loss": 0.1344,
+      "step": 2743
+    },
+    {
+      "epoch": 74.16,
+      "learning_rate": 0.00015,
+      "loss": 0.0948,
+      "step": 2744
+    },
+    {
+      "epoch": 74.19,
+      "learning_rate": 0.00014984375,
+      "loss": 0.1369,
+      "step": 2745
+    },
+    {
+      "epoch": 74.22,
+      "learning_rate": 0.0001496875,
+      "loss": 0.1163,
+      "step": 2746
+    },
+    {
+      "epoch": 74.24,
+      "learning_rate": 0.00014953125,
+      "loss": 0.1094,
+      "step": 2747
+    },
+    {
+      "epoch": 74.27,
+      "learning_rate": 0.000149375,
+      "loss": 0.1025,
+      "step": 2748
+    },
+    {
+      "epoch": 74.3,
+      "learning_rate": 0.00014921875000000002,
+      "loss": 0.092,
+      "step": 2749
+    },
+    {
+      "epoch": 74.32,
+      "learning_rate": 0.0001490625,
+      "loss": 0.0941,
+      "step": 2750
+    },
+    {
+      "epoch": 74.35,
+      "learning_rate": 0.00014890625,
+      "loss": 0.1225,
+      "step": 2751
+    },
+    {
+      "epoch": 74.38,
+      "learning_rate": 0.00014874999999999998,
+      "loss": 0.1259,
+      "step": 2752
+    },
+    {
+      "epoch": 74.41,
+      "learning_rate": 0.00014859375,
+      "loss": 0.1285,
+      "step": 2753
+    },
+    {
+      "epoch": 74.43,
+      "learning_rate": 0.0001484375,
+      "loss": 0.1178,
+      "step": 2754
+    },
+    {
+      "epoch": 74.46,
+      "learning_rate": 0.00014828125,
+      "loss": 0.1156,
+      "step": 2755
+    },
+    {
+      "epoch": 74.49,
+      "learning_rate": 0.000148125,
+      "loss": 0.1488,
+      "step": 2756
+    },
+    {
+      "epoch": 74.51,
+      "learning_rate": 0.00014796875000000002,
+      "loss": 0.1093,
+      "step": 2757
+    },
+    {
+      "epoch": 74.54,
+      "learning_rate": 0.00014781250000000002,
+      "loss": 0.1204,
+      "step": 2758
+    },
+    {
+      "epoch": 74.57,
+      "learning_rate": 0.00014765625,
+      "loss": 0.1458,
+      "step": 2759
+    },
+    {
+      "epoch": 74.59,
+      "learning_rate": 0.0001475,
+      "loss": 0.1379,
+      "step": 2760
+    },
+    {
+      "epoch": 74.62,
+      "learning_rate": 0.00014734375,
+      "loss": 0.146,
+      "step": 2761
+    },
+    {
+      "epoch": 74.65,
+      "learning_rate": 0.0001471875,
+      "loss": 0.1132,
+      "step": 2762
+    },
+    {
+      "epoch": 74.68,
+      "learning_rate": 0.00014703125,
+      "loss": 0.1286,
+      "step": 2763
+    },
+    {
+      "epoch": 74.7,
+      "learning_rate": 0.000146875,
+      "loss": 0.1428,
+      "step": 2764
+    },
+    {
+      "epoch": 74.73,
+      "learning_rate": 0.00014671875,
+      "loss": 0.097,
+      "step": 2765
+    },
+    {
+      "epoch": 74.76,
+      "learning_rate": 0.00014656250000000002,
+      "loss": 0.1155,
+      "step": 2766
+    },
+    {
+      "epoch": 74.78,
+      "learning_rate": 0.00014640625,
+      "loss": 0.1041,
+      "step": 2767
+    },
+    {
+      "epoch": 74.81,
+      "learning_rate": 0.00014625,
+      "loss": 0.1392,
+      "step": 2768
+    },
+    {
+      "epoch": 74.84,
+      "learning_rate": 0.00014609375,
+      "loss": 0.121,
+      "step": 2769
+    },
+    {
+      "epoch": 74.86,
+      "learning_rate": 0.0001459375,
+      "loss": 0.1284,
+      "step": 2770
+    },
+    {
+      "epoch": 74.89,
+      "learning_rate": 0.00014578125,
+      "loss": 0.1291,
+      "step": 2771
+    },
+    {
+      "epoch": 74.92,
+      "learning_rate": 0.000145625,
+      "loss": 0.1422,
+      "step": 2772
+    },
+    {
+      "epoch": 74.95,
+      "learning_rate": 0.00014546875,
+      "loss": 0.1263,
+      "step": 2773
+    },
+    {
+      "epoch": 74.97,
+      "learning_rate": 0.00014531250000000002,
+      "loss": 0.1273,
+      "step": 2774
+    },
+    {
+      "epoch": 75.0,
+      "learning_rate": 0.00014515625,
+      "loss": 0.0808,
+      "step": 2775
+    },
+    {
+      "epoch": 75.03,
+      "learning_rate": 0.000145,
+      "loss": 0.1262,
+      "step": 2776
+    },
+    {
+      "epoch": 75.05,
+      "learning_rate": 0.00014484375,
+      "loss": 0.1359,
+      "step": 2777
+    },
+    {
+      "epoch": 75.08,
+      "learning_rate": 0.0001446875,
+      "loss": 0.0965,
+      "step": 2778
+    },
+    {
+      "epoch": 75.11,
+      "learning_rate": 0.00014453125,
+      "loss": 0.099,
+      "step": 2779
+    },
+    {
+      "epoch": 75.14,
+      "learning_rate": 0.000144375,
+      "loss": 0.1207,
+      "step": 2780
+    },
+    {
+      "epoch": 75.16,
+      "learning_rate": 0.00014421875,
+      "loss": 0.1174,
+      "step": 2781
+    },
+    {
+      "epoch": 75.19,
+      "learning_rate": 0.0001440625,
+      "loss": 0.1436,
+      "step": 2782
+    },
+    {
+      "epoch": 75.22,
+      "learning_rate": 0.00014390625000000002,
+      "loss": 0.1455,
+      "step": 2783
+    },
+    {
+      "epoch": 75.24,
+      "learning_rate": 0.00014375,
+      "loss": 0.1514,
+      "step": 2784
+    },
+    {
+      "epoch": 75.27,
+      "learning_rate": 0.00014359375,
+      "loss": 0.1628,
+      "step": 2785
+    },
+    {
+      "epoch": 75.3,
+      "learning_rate": 0.0001434375,
+      "loss": 0.1054,
+      "step": 2786
+    },
+    {
+      "epoch": 75.32,
+      "learning_rate": 0.00014328125,
+      "loss": 0.1361,
+      "step": 2787
+    },
+    {
+      "epoch": 75.35,
+      "learning_rate": 0.000143125,
+      "loss": 0.1235,
+      "step": 2788
+    },
+    {
+      "epoch": 75.38,
+      "learning_rate": 0.00014296875,
+      "loss": 0.1261,
+      "step": 2789
+    },
+    {
+      "epoch": 75.41,
+      "learning_rate": 0.0001428125,
+      "loss": 0.147,
+      "step": 2790
+    },
+    {
+      "epoch": 75.43,
+      "learning_rate": 0.00014265625000000002,
+      "loss": 0.1334,
+      "step": 2791
+    },
+    {
+      "epoch": 75.46,
+      "learning_rate": 0.0001425,
+      "loss": 0.1491,
+      "step": 2792
+    },
+    {
+      "epoch": 75.49,
+      "learning_rate": 0.00014234375,
+      "loss": 0.1471,
+      "step": 2793
+    },
+    {
+      "epoch": 75.51,
+      "learning_rate": 0.0001421875,
+      "loss": 0.1189,
+      "step": 2794
+    },
+    {
+      "epoch": 75.54,
+      "learning_rate": 0.00014203125,
+      "loss": 0.1265,
+      "step": 2795
+    },
+    {
+      "epoch": 75.57,
+      "learning_rate": 0.000141875,
+      "loss": 0.1162,
+      "step": 2796
+    },
+    {
+      "epoch": 75.59,
+      "learning_rate": 0.00014171875,
+      "loss": 0.1331,
+      "step": 2797
+    },
+    {
+      "epoch": 75.62,
+      "learning_rate": 0.0001415625,
+      "loss": 0.1303,
+      "step": 2798
+    },
+    {
+      "epoch": 75.65,
+      "learning_rate": 0.00014140625,
+      "loss": 0.1309,
+      "step": 2799
+    },
+    {
+      "epoch": 75.68,
+      "learning_rate": 0.00014125,
+      "loss": 0.0901,
+      "step": 2800
+    },
+    {
+      "epoch": 75.7,
+      "learning_rate": 0.00014109375,
+      "loss": 0.1257,
+      "step": 2801
+    },
+    {
+      "epoch": 75.73,
+      "learning_rate": 0.0001409375,
+      "loss": 0.1189,
+      "step": 2802
+    },
+    {
+      "epoch": 75.76,
+      "learning_rate": 0.00014078125,
+      "loss": 0.112,
+      "step": 2803
+    },
+    {
+      "epoch": 75.78,
+      "learning_rate": 0.00014062500000000002,
+      "loss": 0.1328,
+      "step": 2804
+    },
+    {
+      "epoch": 75.81,
+      "learning_rate": 0.00014046875,
+      "loss": 0.1301,
+      "step": 2805
+    },
+    {
+      "epoch": 75.84,
+      "learning_rate": 0.0001403125,
+      "loss": 0.1185,
+      "step": 2806
+    },
+    {
+      "epoch": 75.86,
+      "learning_rate": 0.00014015625,
+      "loss": 0.114,
+      "step": 2807
+    },
+    {
+      "epoch": 75.89,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 0.1431,
+      "step": 2808
+    },
+    {
+      "epoch": 75.92,
+      "learning_rate": 0.00013984375,
+      "loss": 0.112,
+      "step": 2809
+    },
+    {
+      "epoch": 75.95,
+      "learning_rate": 0.0001396875,
+      "loss": 0.1419,
+      "step": 2810
+    },
+    {
+      "epoch": 75.97,
+      "learning_rate": 0.00013953125,
+      "loss": 0.1186,
+      "step": 2811
+    },
+    {
+      "epoch": 76.0,
+      "learning_rate": 0.000139375,
+      "loss": 0.0933,
+      "step": 2812
+    },
+    {
+      "epoch": 76.03,
+      "learning_rate": 0.00013921875,
+      "loss": 0.134,
+      "step": 2813
+    },
+    {
+      "epoch": 76.05,
+      "learning_rate": 0.0001390625,
+      "loss": 0.1075,
+      "step": 2814
+    },
+    {
+      "epoch": 76.08,
+      "learning_rate": 0.00013890625,
+      "loss": 0.1239,
+      "step": 2815
+    },
+    {
+      "epoch": 76.11,
+      "learning_rate": 0.00013875,
+      "loss": 0.1001,
+      "step": 2816
+    },
+    {
+      "epoch": 76.14,
+      "learning_rate": 0.00013859375,
+      "loss": 0.1005,
+      "step": 2817
+    },
+    {
+      "epoch": 76.16,
+      "learning_rate": 0.0001384375,
+      "loss": 0.1004,
+      "step": 2818
+    },
+    {
+      "epoch": 76.19,
+      "learning_rate": 0.00013828125,
+      "loss": 0.1385,
+      "step": 2819
+    },
+    {
+      "epoch": 76.22,
+      "learning_rate": 0.000138125,
+      "loss": 0.1118,
+      "step": 2820
+    },
+    {
+      "epoch": 76.24,
+      "learning_rate": 0.00013796875000000002,
+      "loss": 0.1024,
+      "step": 2821
+    },
+    {
+      "epoch": 76.27,
+      "learning_rate": 0.0001378125,
+      "loss": 0.1426,
+      "step": 2822
+    },
+    {
+      "epoch": 76.3,
+      "learning_rate": 0.00013765625,
+      "loss": 0.114,
+      "step": 2823
+    },
+    {
+      "epoch": 76.32,
+      "learning_rate": 0.0001375,
+      "loss": 0.0962,
+      "step": 2824
+    },
+    {
+      "epoch": 76.35,
+      "learning_rate": 0.00013734375,
+      "loss": 0.1031,
+      "step": 2825
+    },
+    {
+      "epoch": 76.38,
+      "learning_rate": 0.0001371875,
+      "loss": 0.1458,
+      "step": 2826
+    },
+    {
+      "epoch": 76.41,
+      "learning_rate": 0.00013703125,
+      "loss": 0.1398,
+      "step": 2827
+    },
+    {
+      "epoch": 76.43,
+      "learning_rate": 0.000136875,
+      "loss": 0.136,
+      "step": 2828
+    },
+    {
+      "epoch": 76.46,
+      "learning_rate": 0.00013671875,
+      "loss": 0.1346,
+      "step": 2829
+    },
+    {
+      "epoch": 76.49,
+      "learning_rate": 0.00013656250000000002,
+      "loss": 0.0936,
+      "step": 2830
+    },
+    {
+      "epoch": 76.51,
+      "learning_rate": 0.00013640625,
+      "loss": 0.1374,
+      "step": 2831
+    },
+    {
+      "epoch": 76.54,
+      "learning_rate": 0.00013625,
+      "loss": 0.1673,
+      "step": 2832
+    },
+    {
+      "epoch": 76.57,
+      "learning_rate": 0.00013609375,
+      "loss": 0.1099,
+      "step": 2833
+    },
+    {
+      "epoch": 76.59,
+      "learning_rate": 0.0001359375,
+      "loss": 0.1414,
+      "step": 2834
+    },
+    {
+      "epoch": 76.62,
+      "learning_rate": 0.00013578125,
+      "loss": 0.1049,
+      "step": 2835
+    },
+    {
+      "epoch": 76.65,
+      "learning_rate": 0.000135625,
+      "loss": 0.1209,
+      "step": 2836
+    },
+    {
+      "epoch": 76.68,
+      "learning_rate": 0.00013546875,
+      "loss": 0.0938,
+      "step": 2837
+    },
+    {
+      "epoch": 76.7,
+      "learning_rate": 0.00013531250000000002,
+      "loss": 0.1244,
+      "step": 2838
+    },
+    {
+      "epoch": 76.73,
+      "learning_rate": 0.00013515625,
+      "loss": 0.1171,
+      "step": 2839
+    },
+    {
+      "epoch": 76.76,
+      "learning_rate": 0.000135,
+      "loss": 0.1243,
+      "step": 2840
+    },
+    {
+      "epoch": 76.78,
+      "learning_rate": 0.00013484375,
+      "loss": 0.126,
+      "step": 2841
+    },
+    {
+      "epoch": 76.81,
+      "learning_rate": 0.0001346875,
+      "loss": 0.1092,
+      "step": 2842
+    },
+    {
+      "epoch": 76.84,
+      "learning_rate": 0.00013453125,
+      "loss": 0.0757,
+      "step": 2843
+    },
+    {
+      "epoch": 76.86,
+      "learning_rate": 0.000134375,
+      "loss": 0.1444,
+      "step": 2844
+    },
+    {
+      "epoch": 76.89,
+      "learning_rate": 0.00013421875,
+      "loss": 0.1363,
+      "step": 2845
+    },
+    {
+      "epoch": 76.92,
+      "learning_rate": 0.0001340625,
+      "loss": 0.1319,
+      "step": 2846
+    },
+    {
+      "epoch": 76.95,
+      "learning_rate": 0.00013390625000000002,
+      "loss": 0.1078,
+      "step": 2847
+    },
+    {
+      "epoch": 76.97,
+      "learning_rate": 0.00013375,
+      "loss": 0.0859,
+      "step": 2848
+    },
+    {
+      "epoch": 77.0,
+      "learning_rate": 0.00013359375,
+      "loss": 0.0186,
+      "step": 2849
+    },
+    {
+      "epoch": 77.03,
+      "learning_rate": 0.00013343749999999998,
+      "loss": 0.1221,
+      "step": 2850
+    },
+    {
+      "epoch": 77.05,
+      "learning_rate": 0.00013328125,
+      "loss": 0.1148,
+      "step": 2851
+    },
+    {
+      "epoch": 77.08,
+      "learning_rate": 0.000133125,
+      "loss": 0.091,
+      "step": 2852
+    },
+    {
+      "epoch": 77.11,
+      "learning_rate": 0.00013296875,
+      "loss": 0.1241,
+      "step": 2853
+    },
+    {
+      "epoch": 77.14,
+      "learning_rate": 0.0001328125,
+      "loss": 0.113,
+      "step": 2854
+    },
+    {
+      "epoch": 77.16,
+      "learning_rate": 0.00013265625000000002,
+      "loss": 0.0908,
+      "step": 2855
+    },
+    {
+      "epoch": 77.19,
+      "learning_rate": 0.00013250000000000002,
+      "loss": 0.1192,
+      "step": 2856
+    },
+    {
+      "epoch": 77.22,
+      "learning_rate": 0.00013234375,
+      "loss": 0.1066,
+      "step": 2857
+    },
+    {
+      "epoch": 77.24,
+      "learning_rate": 0.0001321875,
+      "loss": 0.1037,
+      "step": 2858
+    },
+    {
+      "epoch": 77.27,
+      "learning_rate": 0.00013203125,
+      "loss": 0.1358,
+      "step": 2859
+    },
+    {
+      "epoch": 77.3,
+      "learning_rate": 0.000131875,
+      "loss": 0.1311,
+      "step": 2860
+    },
+    {
+      "epoch": 77.32,
+      "learning_rate": 0.00013171875,
+      "loss": 0.118,
+      "step": 2861
+    },
+    {
+      "epoch": 77.35,
+      "learning_rate": 0.0001315625,
+      "loss": 0.1086,
+      "step": 2862
+    },
+    {
+      "epoch": 77.38,
+      "learning_rate": 0.00013140625,
+      "loss": 0.1164,
+      "step": 2863
+    },
+    {
+      "epoch": 77.41,
+      "learning_rate": 0.00013125000000000002,
+      "loss": 0.1102,
+      "step": 2864
+    },
+    {
+      "epoch": 77.43,
+      "learning_rate": 0.00013109375,
+      "loss": 0.1025,
+      "step": 2865
+    },
+    {
+      "epoch": 77.46,
+      "learning_rate": 0.0001309375,
+      "loss": 0.0989,
+      "step": 2866
+    },
+    {
+      "epoch": 77.49,
+      "learning_rate": 0.00013078124999999998,
+      "loss": 0.113,
+      "step": 2867
+    },
+    {
+      "epoch": 77.51,
+      "learning_rate": 0.000130625,
+      "loss": 0.1178,
+      "step": 2868
+    },
+    {
+      "epoch": 77.54,
+      "learning_rate": 0.00013046875,
+      "loss": 0.1113,
+      "step": 2869
+    },
+    {
+      "epoch": 77.57,
+      "learning_rate": 0.0001303125,
+      "loss": 0.1055,
+      "step": 2870
+    },
+    {
+      "epoch": 77.59,
+      "learning_rate": 0.00013015625,
+      "loss": 0.0974,
+      "step": 2871
+    },
+    {
+      "epoch": 77.62,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 0.1138,
+      "step": 2872
+    },
+    {
+      "epoch": 77.65,
+      "learning_rate": 0.00012984375000000002,
+      "loss": 0.118,
+      "step": 2873
+    },
+    {
+      "epoch": 77.68,
+      "learning_rate": 0.0001296875,
+      "loss": 0.0903,
+      "step": 2874
+    },
+    {
+      "epoch": 77.7,
+      "learning_rate": 0.00012953124999999998,
+      "loss": 0.1417,
+      "step": 2875
+    },
+    {
+      "epoch": 77.73,
+      "learning_rate": 0.000129375,
+      "loss": 0.1122,
+      "step": 2876
+    },
+    {
+      "epoch": 77.76,
+      "learning_rate": 0.00012921875,
+      "loss": 0.0941,
+      "step": 2877
+    },
+    {
+      "epoch": 77.78,
+      "learning_rate": 0.0001290625,
+      "loss": 0.1002,
+      "step": 2878
+    },
+    {
+      "epoch": 77.81,
+      "learning_rate": 0.00012890625,
+      "loss": 0.1104,
+      "step": 2879
+    },
+    {
+      "epoch": 77.84,
+      "learning_rate": 0.00012875,
+      "loss": 0.0783,
+      "step": 2880
+    },
+    {
+      "epoch": 77.86,
+      "learning_rate": 0.00012859375000000002,
+      "loss": 0.1397,
+      "step": 2881
+    },
+    {
+      "epoch": 77.89,
+      "learning_rate": 0.00012843750000000003,
+      "loss": 0.094,
+      "step": 2882
+    },
+    {
+      "epoch": 77.92,
+      "learning_rate": 0.00012828125,
+      "loss": 0.1163,
+      "step": 2883
+    },
+    {
+      "epoch": 77.95,
+      "learning_rate": 0.00012812499999999998,
+      "loss": 0.0857,
+      "step": 2884
+    },
+    {
+      "epoch": 77.97,
+      "learning_rate": 0.00012796875,
+      "loss": 0.094,
+      "step": 2885
+    },
+    {
+      "epoch": 78.0,
+      "learning_rate": 0.0001278125,
+      "loss": 0.0843,
+      "step": 2886
+    },
+    {
+      "epoch": 78.03,
+      "learning_rate": 0.00012765625,
+      "loss": 0.1268,
+      "step": 2887
+    },
+    {
+      "epoch": 78.05,
+      "learning_rate": 0.0001275,
+      "loss": 0.0991,
+      "step": 2888
+    },
+    {
+      "epoch": 78.08,
+      "learning_rate": 0.00012734375000000002,
+      "loss": 0.1194,
+      "step": 2889
+    },
+    {
+      "epoch": 78.11,
+      "learning_rate": 0.00012718750000000002,
+      "loss": 0.1012,
+      "step": 2890
+    },
+    {
+      "epoch": 78.14,
+      "learning_rate": 0.00012703125000000003,
+      "loss": 0.1299,
+      "step": 2891
+    },
+    {
+      "epoch": 78.16,
+      "learning_rate": 0.00012687499999999998,
+      "loss": 0.1623,
+      "step": 2892
+    },
+    {
+      "epoch": 78.19,
+      "learning_rate": 0.00012671875,
+      "loss": 0.1133,
+      "step": 2893
+    },
+    {
+      "epoch": 78.22,
+      "learning_rate": 0.0001265625,
+      "loss": 0.0867,
+      "step": 2894
+    },
+    {
+      "epoch": 78.24,
+      "learning_rate": 0.00012640625,
+      "loss": 0.1229,
+      "step": 2895
+    },
+    {
+      "epoch": 78.27,
+      "learning_rate": 0.00012625,
+      "loss": 0.1188,
+      "step": 2896
+    },
+    {
+      "epoch": 78.3,
+      "learning_rate": 0.00012609375,
+      "loss": 0.0967,
+      "step": 2897
+    },
+    {
+      "epoch": 78.32,
+      "learning_rate": 0.00012593750000000002,
+      "loss": 0.0818,
+      "step": 2898
+    },
+    {
+      "epoch": 78.35,
+      "learning_rate": 0.00012578125000000002,
+      "loss": 0.1354,
+      "step": 2899
+    },
+    {
+      "epoch": 78.38,
+      "learning_rate": 0.00012562499999999998,
+      "loss": 0.1354,
+      "step": 2900
+    },
+    {
+      "epoch": 78.41,
+      "learning_rate": 0.00012546874999999998,
+      "loss": 0.1081,
+      "step": 2901
+    },
+    {
+      "epoch": 78.43,
+      "learning_rate": 0.0001253125,
+      "loss": 0.1196,
+      "step": 2902
+    },
+    {
+      "epoch": 78.46,
+      "learning_rate": 0.00012515625,
+      "loss": 0.1003,
+      "step": 2903
+    },
+    {
+      "epoch": 78.49,
+      "learning_rate": 0.000125,
+      "loss": 0.0961,
+      "step": 2904
+    },
+    {
+      "epoch": 78.51,
+      "learning_rate": 0.00012484375,
+      "loss": 0.1109,
+      "step": 2905
+    },
+    {
+      "epoch": 78.54,
+      "learning_rate": 0.00012468750000000002,
+      "loss": 0.1434,
+      "step": 2906
+    },
+    {
+      "epoch": 78.57,
+      "learning_rate": 0.00012453125,
+      "loss": 0.1214,
+      "step": 2907
+    },
+    {
+      "epoch": 78.59,
+      "learning_rate": 0.000124375,
+      "loss": 0.0878,
+      "step": 2908
+    },
+    {
+      "epoch": 78.62,
+      "learning_rate": 0.00012421875,
+      "loss": 0.1201,
+      "step": 2909
+    },
+    {
+      "epoch": 78.65,
+      "learning_rate": 0.00012406250000000001,
+      "loss": 0.1219,
+      "step": 2910
+    },
+    {
+      "epoch": 78.68,
+      "learning_rate": 0.00012390625,
+      "loss": 0.0653,
+      "step": 2911
+    },
+    {
+      "epoch": 78.7,
+      "learning_rate": 0.00012375,
+      "loss": 0.1462,
+      "step": 2912
+    },
+    {
+      "epoch": 78.73,
+      "learning_rate": 0.00012359375,
+      "loss": 0.1092,
+      "step": 2913
+    },
+    {
+      "epoch": 78.76,
+      "learning_rate": 0.0001234375,
+      "loss": 0.114,
+      "step": 2914
+    },
+    {
+      "epoch": 78.78,
+      "learning_rate": 0.00012328125,
+      "loss": 0.0955,
+      "step": 2915
+    },
+    {
+      "epoch": 78.81,
+      "learning_rate": 0.000123125,
+      "loss": 0.1004,
+      "step": 2916
+    },
+    {
+      "epoch": 78.84,
+      "learning_rate": 0.00012296875,
+      "loss": 0.0824,
+      "step": 2917
+    },
+    {
+      "epoch": 78.86,
+      "learning_rate": 0.0001228125,
+      "loss": 0.1205,
+      "step": 2918
+    },
+    {
+      "epoch": 78.89,
+      "learning_rate": 0.00012265625,
+      "loss": 0.1181,
+      "step": 2919
+    },
+    {
+      "epoch": 78.92,
+      "learning_rate": 0.0001225,
+      "loss": 0.1149,
+      "step": 2920
+    },
+    {
+      "epoch": 78.95,
+      "learning_rate": 0.00012234375,
+      "loss": 0.0888,
+      "step": 2921
+    },
+    {
+      "epoch": 78.97,
+      "learning_rate": 0.0001221875,
+      "loss": 0.108,
+      "step": 2922
+    },
+    {
+      "epoch": 79.0,
+      "learning_rate": 0.00012203125,
+      "loss": 0.0629,
+      "step": 2923
+    },
+    {
+      "epoch": 79.03,
+      "learning_rate": 0.000121875,
+      "loss": 0.1232,
+      "step": 2924
+    },
+    {
+      "epoch": 79.05,
+      "learning_rate": 0.00012171875,
+      "loss": 0.1077,
+      "step": 2925
+    },
+    {
+      "epoch": 79.08,
+      "learning_rate": 0.00012156250000000001,
+      "loss": 0.1101,
+      "step": 2926
+    },
+    {
+      "epoch": 79.11,
+      "learning_rate": 0.00012140625,
+      "loss": 0.1096,
+      "step": 2927
+    },
+    {
+      "epoch": 79.14,
+      "learning_rate": 0.00012124999999999999,
+      "loss": 0.1338,
+      "step": 2928
+    },
+    {
+      "epoch": 79.16,
+      "learning_rate": 0.00012109375,
+      "loss": 0.0924,
+      "step": 2929
+    },
+    {
+      "epoch": 79.19,
+      "learning_rate": 0.0001209375,
+      "loss": 0.1158,
+      "step": 2930
+    },
+    {
+      "epoch": 79.22,
+      "learning_rate": 0.00012078125000000001,
+      "loss": 0.1252,
+      "step": 2931
+    },
+    {
+      "epoch": 79.24,
+      "learning_rate": 0.000120625,
+      "loss": 0.1148,
+      "step": 2932
+    },
+    {
+      "epoch": 79.27,
+      "learning_rate": 0.00012046875,
+      "loss": 0.0997,
+      "step": 2933
+    },
+    {
+      "epoch": 79.3,
+      "learning_rate": 0.0001203125,
+      "loss": 0.1026,
+      "step": 2934
+    },
+    {
+      "epoch": 79.32,
+      "learning_rate": 0.00012015625000000001,
+      "loss": 0.0955,
+      "step": 2935
+    },
+    {
+      "epoch": 79.35,
+      "learning_rate": 0.00012,
+      "loss": 0.1331,
+      "step": 2936
+    },
+    {
+      "epoch": 79.38,
+      "learning_rate": 0.00011984375,
+      "loss": 0.1289,
+      "step": 2937
+    },
+    {
+      "epoch": 79.41,
+      "learning_rate": 0.0001196875,
+      "loss": 0.0969,
+      "step": 2938
+    },
+    {
+      "epoch": 79.43,
+      "learning_rate": 0.00011953125000000001,
+      "loss": 0.1362,
+      "step": 2939
+    },
+    {
+      "epoch": 79.46,
+      "learning_rate": 0.000119375,
+      "loss": 0.0985,
+      "step": 2940
+    },
+    {
+      "epoch": 79.49,
+      "learning_rate": 0.00011921875,
+      "loss": 0.0914,
+      "step": 2941
+    },
+    {
+      "epoch": 79.51,
+      "learning_rate": 0.0001190625,
+      "loss": 0.0852,
+      "step": 2942
+    },
+    {
+      "epoch": 79.54,
+      "learning_rate": 0.00011890625000000001,
+      "loss": 0.0929,
+      "step": 2943
+    },
+    {
+      "epoch": 79.57,
+      "learning_rate": 0.00011875,
+      "loss": 0.0918,
+      "step": 2944
+    },
+    {
+      "epoch": 79.59,
+      "learning_rate": 0.00011859375,
+      "loss": 0.1069,
+      "step": 2945
+    },
+    {
+      "epoch": 79.62,
+      "learning_rate": 0.0001184375,
+      "loss": 0.0958,
+      "step": 2946
+    },
+    {
+      "epoch": 79.65,
+      "learning_rate": 0.00011828125,
+      "loss": 0.0802,
+      "step": 2947
+    },
+    {
+      "epoch": 79.68,
+      "learning_rate": 0.000118125,
+      "loss": 0.0955,
+      "step": 2948
+    },
+    {
+      "epoch": 79.7,
+      "learning_rate": 0.00011796875,
+      "loss": 0.1181,
+      "step": 2949
+    },
+    {
+      "epoch": 79.73,
+      "learning_rate": 0.0001178125,
+      "loss": 0.1348,
+      "step": 2950
+    },
+    {
+      "epoch": 79.76,
+      "learning_rate": 0.00011765625,
+      "loss": 0.0907,
+      "step": 2951
+    },
+    {
+      "epoch": 79.78,
+      "learning_rate": 0.0001175,
+      "loss": 0.1238,
+      "step": 2952
+    },
+    {
+      "epoch": 79.81,
+      "learning_rate": 0.00011734375,
+      "loss": 0.0971,
+      "step": 2953
+    },
+    {
+      "epoch": 79.84,
+      "learning_rate": 0.0001171875,
+      "loss": 0.1465,
+      "step": 2954
+    },
+    {
+      "epoch": 79.86,
+      "learning_rate": 0.00011703125,
+      "loss": 0.1012,
+      "step": 2955
+    },
+    {
+      "epoch": 79.89,
+      "learning_rate": 0.00011687500000000001,
+      "loss": 0.1047,
+      "step": 2956
+    },
+    {
+      "epoch": 79.92,
+      "learning_rate": 0.00011671875,
+      "loss": 0.1115,
+      "step": 2957
+    },
+    {
+      "epoch": 79.95,
+      "learning_rate": 0.00011656250000000001,
+      "loss": 0.0997,
+      "step": 2958
+    },
+    {
+      "epoch": 79.97,
+      "learning_rate": 0.00011640625,
+      "loss": 0.1237,
+      "step": 2959
+    },
+    {
+      "epoch": 80.0,
+      "learning_rate": 0.00011625,
+      "loss": 0.1536,
+      "step": 2960
+    },
+    {
+      "epoch": 80.03,
+      "learning_rate": 0.00011609375,
+      "loss": 0.097,
+      "step": 2961
+    },
+    {
+      "epoch": 80.05,
+      "learning_rate": 0.0001159375,
+      "loss": 0.1006,
+      "step": 2962
+    },
+    {
+      "epoch": 80.08,
+      "learning_rate": 0.00011578125,
+      "loss": 0.114,
+      "step": 2963
+    },
+    {
+      "epoch": 80.11,
+      "learning_rate": 0.000115625,
+      "loss": 0.1439,
+      "step": 2964
+    },
+    {
+      "epoch": 80.14,
+      "learning_rate": 0.00011546875,
+      "loss": 0.0756,
+      "step": 2965
+    },
+    {
+      "epoch": 80.16,
+      "learning_rate": 0.0001153125,
+      "loss": 0.1177,
+      "step": 2966
+    },
+    {
+      "epoch": 80.19,
+      "learning_rate": 0.00011515625000000001,
+      "loss": 0.0859,
+      "step": 2967
+    },
+    {
+      "epoch": 80.22,
+      "learning_rate": 0.000115,
+      "loss": 0.0754,
+      "step": 2968
+    },
+    {
+      "epoch": 80.24,
+      "learning_rate": 0.00011484375,
+      "loss": 0.1289,
+      "step": 2969
+    },
+    {
+      "epoch": 80.27,
+      "learning_rate": 0.0001146875,
+      "loss": 0.1046,
+      "step": 2970
+    },
+    {
+      "epoch": 80.3,
+      "learning_rate": 0.00011453125000000001,
+      "loss": 0.0913,
+      "step": 2971
+    },
+    {
+      "epoch": 80.32,
+      "learning_rate": 0.000114375,
+      "loss": 0.0946,
+      "step": 2972
+    },
+    {
+      "epoch": 80.35,
+      "learning_rate": 0.00011421875,
+      "loss": 0.0704,
+      "step": 2973
+    },
+    {
+      "epoch": 80.38,
+      "learning_rate": 0.0001140625,
+      "loss": 0.1007,
+      "step": 2974
+    },
+    {
+      "epoch": 80.41,
+      "learning_rate": 0.00011390625000000001,
+      "loss": 0.0888,
+      "step": 2975
+    },
+    {
+      "epoch": 80.43,
+      "learning_rate": 0.00011375,
+      "loss": 0.1056,
+      "step": 2976
+    },
+    {
+      "epoch": 80.46,
+      "learning_rate": 0.00011359374999999999,
+      "loss": 0.0776,
+      "step": 2977
+    },
+    {
+      "epoch": 80.49,
+      "learning_rate": 0.0001134375,
+      "loss": 0.0851,
+      "step": 2978
+    },
+    {
+      "epoch": 80.51,
+      "learning_rate": 0.00011328125,
+      "loss": 0.0824,
+      "step": 2979
+    },
+    {
+      "epoch": 80.54,
+      "learning_rate": 0.00011312500000000001,
+      "loss": 0.1083,
+      "step": 2980
+    },
+    {
+      "epoch": 80.57,
+      "learning_rate": 0.00011296875,
+      "loss": 0.1096,
+      "step": 2981
+    },
+    {
+      "epoch": 80.59,
+      "learning_rate": 0.0001128125,
+      "loss": 0.1087,
+      "step": 2982
+    },
+    {
+      "epoch": 80.62,
+      "learning_rate": 0.00011265625,
+      "loss": 0.1232,
+      "step": 2983
+    },
+    {
+      "epoch": 80.65,
+      "learning_rate": 0.00011250000000000001,
+      "loss": 0.0854,
+      "step": 2984
+    },
+    {
+      "epoch": 80.68,
+      "learning_rate": 0.00011234375,
+      "loss": 0.0853,
+      "step": 2985
+    },
+    {
+      "epoch": 80.7,
+      "learning_rate": 0.0001121875,
+      "loss": 0.1086,
+      "step": 2986
+    },
+    {
+      "epoch": 80.73,
+      "learning_rate": 0.00011203125,
+      "loss": 0.094,
+      "step": 2987
+    },
+    {
+      "epoch": 80.76,
+      "learning_rate": 0.00011187500000000001,
+      "loss": 0.0951,
+      "step": 2988
+    },
+    {
+      "epoch": 80.78,
+      "learning_rate": 0.00011171875,
+      "loss": 0.1358,
+      "step": 2989
+    },
+    {
+      "epoch": 80.81,
+      "learning_rate": 0.0001115625,
+      "loss": 0.0969,
+      "step": 2990
+    },
+    {
+      "epoch": 80.84,
+      "learning_rate": 0.00011140625,
+      "loss": 0.0999,
+      "step": 2991
+    },
+    {
+      "epoch": 80.86,
+      "learning_rate": 0.00011125000000000001,
+      "loss": 0.1041,
+      "step": 2992
+    },
+    {
+      "epoch": 80.89,
+      "learning_rate": 0.00011109375000000001,
+      "loss": 0.103,
+      "step": 2993
+    },
+    {
+      "epoch": 80.92,
+      "learning_rate": 0.00011093749999999999,
+      "loss": 0.1045,
+      "step": 2994
+    },
+    {
+      "epoch": 80.95,
+      "learning_rate": 0.00011078125,
+      "loss": 0.0959,
+      "step": 2995
+    },
+    {
+      "epoch": 80.97,
+      "learning_rate": 0.000110625,
+      "loss": 0.0714,
+      "step": 2996
+    },
+    {
+      "epoch": 81.0,
+      "learning_rate": 0.00011046875000000001,
+      "loss": 0.1084,
+      "step": 2997
+    },
+    {
+      "epoch": 81.03,
+      "learning_rate": 0.00011031249999999999,
+      "loss": 0.1624,
+      "step": 2998
+    },
+    {
+      "epoch": 81.05,
+      "learning_rate": 0.00011015625,
+      "loss": 0.0972,
+      "step": 2999
+    },
+    {
+      "epoch": 81.08,
+      "learning_rate": 0.00011,
+      "loss": 0.1213,
+      "step": 3000
+    },
+    {
+      "epoch": 81.08,
+      "eval_cer": 0.07577324997376272,
+      "eval_loss": 0.3777328431606293,
+      "eval_runtime": 11.8935,
+      "eval_samples_per_second": 138.479,
+      "eval_steps_per_second": 2.186,
+      "eval_wer": 0.31130630170564805,
+      "step": 3000
+    },
+    {
+      "epoch": 81.11,
+      "learning_rate": 0.00010984375000000001,
+      "loss": 0.1183,
+      "step": 3001
+    },
+    {
+      "epoch": 81.14,
+      "learning_rate": 0.00010968749999999999,
+      "loss": 0.1111,
+      "step": 3002
+    },
+    {
+      "epoch": 81.16,
+      "learning_rate": 0.00010953125,
+      "loss": 0.1097,
+      "step": 3003
+    },
+    {
+      "epoch": 81.19,
+      "learning_rate": 0.000109375,
+      "loss": 0.0864,
+      "step": 3004
+    },
+    {
+      "epoch": 81.22,
+      "learning_rate": 0.00010921875000000001,
+      "loss": 0.111,
+      "step": 3005
+    },
+    {
+      "epoch": 81.24,
+      "learning_rate": 0.00010906250000000002,
+      "loss": 0.1246,
+      "step": 3006
+    },
+    {
+      "epoch": 81.27,
+      "learning_rate": 0.00010890625,
+      "loss": 0.0868,
+      "step": 3007
+    },
+    {
+      "epoch": 81.3,
+      "learning_rate": 0.00010875,
+      "loss": 0.1044,
+      "step": 3008
+    },
+    {
+      "epoch": 81.32,
+      "learning_rate": 0.00010859375000000001,
+      "loss": 0.0574,
+      "step": 3009
+    },
+    {
+      "epoch": 81.35,
+      "learning_rate": 0.00010843750000000001,
+      "loss": 0.0898,
+      "step": 3010
+    },
+    {
+      "epoch": 81.38,
+      "learning_rate": 0.00010828124999999999,
+      "loss": 0.1136,
+      "step": 3011
+    },
+    {
+      "epoch": 81.41,
+      "learning_rate": 0.000108125,
+      "loss": 0.0891,
+      "step": 3012
+    },
+    {
+      "epoch": 81.43,
+      "learning_rate": 0.00010796875,
+      "loss": 0.1096,
+      "step": 3013
+    },
+    {
+      "epoch": 81.46,
+      "learning_rate": 0.00010781250000000001,
+      "loss": 0.1225,
+      "step": 3014
+    },
+    {
+      "epoch": 81.49,
+      "learning_rate": 0.00010765624999999999,
+      "loss": 0.0908,
+      "step": 3015
+    },
+    {
+      "epoch": 81.51,
+      "learning_rate": 0.0001075,
+      "loss": 0.0967,
+      "step": 3016
+    },
+    {
+      "epoch": 81.54,
+      "learning_rate": 0.00010734375,
+      "loss": 0.1184,
+      "step": 3017
+    },
+    {
+      "epoch": 81.57,
+      "learning_rate": 0.00010718750000000001,
+      "loss": 0.1255,
+      "step": 3018
+    },
+    {
+      "epoch": 81.59,
+      "learning_rate": 0.00010703125,
+      "loss": 0.1108,
+      "step": 3019
+    },
+    {
+      "epoch": 81.62,
+      "learning_rate": 0.000106875,
+      "loss": 0.0888,
+      "step": 3020
+    },
+    {
+      "epoch": 81.65,
+      "learning_rate": 0.00010671875,
+      "loss": 0.0852,
+      "step": 3021
+    },
+    {
+      "epoch": 81.68,
+      "learning_rate": 0.00010656250000000001,
+      "loss": 0.1047,
+      "step": 3022
+    },
+    {
+      "epoch": 81.7,
+      "learning_rate": 0.00010640625,
+      "loss": 0.1032,
+      "step": 3023
+    },
+    {
+      "epoch": 81.73,
+      "learning_rate": 0.00010625,
+      "loss": 0.0976,
+      "step": 3024
+    },
+    {
+      "epoch": 81.76,
+      "learning_rate": 0.00010609375,
+      "loss": 0.0876,
+      "step": 3025
+    },
+    {
+      "epoch": 81.78,
+      "learning_rate": 0.00010593750000000001,
+      "loss": 0.1116,
+      "step": 3026
+    },
+    {
+      "epoch": 81.81,
+      "learning_rate": 0.00010578125,
+      "loss": 0.1013,
+      "step": 3027
+    },
+    {
+      "epoch": 81.84,
+      "learning_rate": 0.00010562499999999999,
+      "loss": 0.0688,
+      "step": 3028
+    },
+    {
+      "epoch": 81.86,
+      "learning_rate": 0.00010546875,
+      "loss": 0.1018,
+      "step": 3029
+    },
+    {
+      "epoch": 81.89,
+      "learning_rate": 0.0001053125,
+      "loss": 0.1017,
+      "step": 3030
+    },
+    {
+      "epoch": 81.92,
+      "learning_rate": 0.00010515625000000001,
+      "loss": 0.0873,
+      "step": 3031
+    },
+    {
+      "epoch": 81.95,
+      "learning_rate": 0.000105,
+      "loss": 0.1049,
+      "step": 3032
+    },
+    {
+      "epoch": 81.97,
+      "learning_rate": 0.00010484375,
+      "loss": 0.1072,
+      "step": 3033
+    },
+    {
+      "epoch": 82.0,
+      "learning_rate": 0.0001046875,
+      "loss": 0.1215,
+      "step": 3034
+    },
+    {
+      "epoch": 82.03,
+      "learning_rate": 0.00010453125000000001,
+      "loss": 0.1169,
+      "step": 3035
+    },
+    {
+      "epoch": 82.05,
+      "learning_rate": 0.000104375,
+      "loss": 0.0999,
+      "step": 3036
+    },
+    {
+      "epoch": 82.08,
+      "learning_rate": 0.00010421875,
+      "loss": 0.0836,
+      "step": 3037
+    },
+    {
+      "epoch": 82.11,
+      "learning_rate": 0.0001040625,
+      "loss": 0.0889,
+      "step": 3038
+    },
+    {
+      "epoch": 82.14,
+      "learning_rate": 0.00010390625000000001,
+      "loss": 0.0949,
+      "step": 3039
+    },
+    {
+      "epoch": 82.16,
+      "learning_rate": 0.00010375,
+      "loss": 0.0699,
+      "step": 3040
+    },
+    {
+      "epoch": 82.19,
+      "learning_rate": 0.00010359375,
+      "loss": 0.1065,
+      "step": 3041
+    },
+    {
+      "epoch": 82.22,
+      "learning_rate": 0.0001034375,
+      "loss": 0.1046,
+      "step": 3042
+    },
+    {
+      "epoch": 82.24,
+      "learning_rate": 0.00010328125,
+      "loss": 0.1216,
+      "step": 3043
+    },
+    {
+      "epoch": 82.27,
+      "learning_rate": 0.000103125,
+      "loss": 0.0906,
+      "step": 3044
+    },
+    {
+      "epoch": 82.3,
+      "learning_rate": 0.00010296875,
+      "loss": 0.089,
+      "step": 3045
+    },
+    {
+      "epoch": 82.32,
+      "learning_rate": 0.0001028125,
+      "loss": 0.0651,
+      "step": 3046
+    },
+    {
+      "epoch": 82.35,
+      "learning_rate": 0.00010265625,
+      "loss": 0.1099,
+      "step": 3047
+    },
+    {
+      "epoch": 82.38,
+      "learning_rate": 0.0001025,
+      "loss": 0.093,
+      "step": 3048
+    },
+    {
+      "epoch": 82.41,
+      "learning_rate": 0.00010234375,
+      "loss": 0.1037,
+      "step": 3049
+    },
+    {
+      "epoch": 82.43,
+      "learning_rate": 0.0001021875,
+      "loss": 0.0683,
+      "step": 3050
+    },
+    {
+      "epoch": 82.46,
+      "learning_rate": 0.00010203125,
+      "loss": 0.119,
+      "step": 3051
+    },
+    {
+      "epoch": 82.49,
+      "learning_rate": 0.000101875,
+      "loss": 0.1083,
+      "step": 3052
+    },
+    {
+      "epoch": 82.51,
+      "learning_rate": 0.00010171875,
+      "loss": 0.0866,
+      "step": 3053
+    },
+    {
+      "epoch": 82.54,
+      "learning_rate": 0.00010156250000000001,
+      "loss": 0.1214,
+      "step": 3054
+    },
+    {
+      "epoch": 82.57,
+      "learning_rate": 0.00010140625,
+      "loss": 0.0898,
+      "step": 3055
+    },
+    {
+      "epoch": 82.59,
+      "learning_rate": 0.00010125000000000001,
+      "loss": 0.1099,
+      "step": 3056
+    },
+    {
+      "epoch": 82.62,
+      "learning_rate": 0.00010109375,
+      "loss": 0.0728,
+      "step": 3057
+    },
+    {
+      "epoch": 82.65,
+      "learning_rate": 0.00010093750000000001,
+      "loss": 0.0974,
+      "step": 3058
+    },
+    {
+      "epoch": 82.68,
+      "learning_rate": 0.00010078125,
+      "loss": 0.08,
+      "step": 3059
+    },
+    {
+      "epoch": 82.7,
+      "learning_rate": 0.000100625,
+      "loss": 0.1104,
+      "step": 3060
+    },
+    {
+      "epoch": 82.73,
+      "learning_rate": 0.00010046875,
+      "loss": 0.1031,
+      "step": 3061
+    },
+    {
+      "epoch": 82.76,
+      "learning_rate": 0.0001003125,
+      "loss": 0.1421,
+      "step": 3062
+    },
+    {
+      "epoch": 82.78,
+      "learning_rate": 0.00010015625,
+      "loss": 0.106,
+      "step": 3063
+    },
+    {
+      "epoch": 82.81,
+      "learning_rate": 0.0001,
+      "loss": 0.0993,
+      "step": 3064
+    },
+    {
+      "epoch": 82.84,
+      "learning_rate": 9.984375e-05,
+      "loss": 0.0991,
+      "step": 3065
+    },
+    {
+      "epoch": 82.86,
+      "learning_rate": 9.96875e-05,
+      "loss": 0.1071,
+      "step": 3066
+    },
+    {
+      "epoch": 82.89,
+      "learning_rate": 9.953125000000001e-05,
+      "loss": 0.1179,
+      "step": 3067
+    },
+    {
+      "epoch": 82.92,
+      "learning_rate": 9.9375e-05,
+      "loss": 0.1045,
+      "step": 3068
+    },
+    {
+      "epoch": 82.95,
+      "learning_rate": 9.921875e-05,
+      "loss": 0.1176,
+      "step": 3069
+    },
+    {
+      "epoch": 82.97,
+      "learning_rate": 9.90625e-05,
+      "loss": 0.1371,
+      "step": 3070
+    },
+    {
+      "epoch": 83.0,
+      "learning_rate": 9.890625000000001e-05,
+      "loss": 0.0432,
+      "step": 3071
+    },
+    {
+      "epoch": 83.03,
+      "learning_rate": 9.875e-05,
+      "loss": 0.1135,
+      "step": 3072
+    },
+    {
+      "epoch": 83.05,
+      "learning_rate": 9.859375e-05,
+      "loss": 0.096,
+      "step": 3073
+    },
+    {
+      "epoch": 83.08,
+      "learning_rate": 9.84375e-05,
+      "loss": 0.0881,
+      "step": 3074
+    },
+    {
+      "epoch": 83.11,
+      "learning_rate": 9.828125000000001e-05,
+      "loss": 0.0988,
+      "step": 3075
+    },
+    {
+      "epoch": 83.14,
+      "learning_rate": 9.8125e-05,
+      "loss": 0.0965,
+      "step": 3076
+    },
+    {
+      "epoch": 83.16,
+      "learning_rate": 9.796874999999999e-05,
+      "loss": 0.1011,
+      "step": 3077
+    },
+    {
+      "epoch": 83.19,
+      "learning_rate": 9.78125e-05,
+      "loss": 0.0882,
+      "step": 3078
+    },
+    {
+      "epoch": 83.22,
+      "learning_rate": 9.765625e-05,
+      "loss": 0.0994,
+      "step": 3079
+    },
+    {
+      "epoch": 83.24,
+      "learning_rate": 9.750000000000001e-05,
+      "loss": 0.0811,
+      "step": 3080
+    },
+    {
+      "epoch": 83.27,
+      "learning_rate": 9.734375e-05,
+      "loss": 0.1131,
+      "step": 3081
+    },
+    {
+      "epoch": 83.3,
+      "learning_rate": 9.71875e-05,
+      "loss": 0.1248,
+      "step": 3082
+    },
+    {
+      "epoch": 83.32,
+      "learning_rate": 9.703125e-05,
+      "loss": 0.1052,
+      "step": 3083
+    },
+    {
+      "epoch": 83.35,
+      "learning_rate": 9.687500000000001e-05,
+      "loss": 0.1065,
+      "step": 3084
+    },
+    {
+      "epoch": 83.38,
+      "learning_rate": 9.671875e-05,
+      "loss": 0.1143,
+      "step": 3085
+    },
+    {
+      "epoch": 83.41,
+      "learning_rate": 9.65625e-05,
+      "loss": 0.104,
+      "step": 3086
+    },
+    {
+      "epoch": 83.43,
+      "learning_rate": 9.640625e-05,
+      "loss": 0.0946,
+      "step": 3087
+    },
+    {
+      "epoch": 83.46,
+      "learning_rate": 9.625000000000001e-05,
+      "loss": 0.0951,
+      "step": 3088
+    },
+    {
+      "epoch": 83.49,
+      "learning_rate": 9.609375e-05,
+      "loss": 0.0996,
+      "step": 3089
+    },
+    {
+      "epoch": 83.51,
+      "learning_rate": 9.59375e-05,
+      "loss": 0.0783,
+      "step": 3090
+    },
+    {
+      "epoch": 83.54,
+      "learning_rate": 9.578125e-05,
+      "loss": 0.1107,
+      "step": 3091
+    },
+    {
+      "epoch": 83.57,
+      "learning_rate": 9.562500000000001e-05,
+      "loss": 0.0947,
+      "step": 3092
+    },
+    {
+      "epoch": 83.59,
+      "learning_rate": 9.546875000000001e-05,
+      "loss": 0.1131,
+      "step": 3093
+    },
+    {
+      "epoch": 83.62,
+      "learning_rate": 9.531249999999999e-05,
+      "loss": 0.1038,
+      "step": 3094
+    },
+    {
+      "epoch": 83.65,
+      "learning_rate": 9.515625e-05,
+      "loss": 0.0743,
+      "step": 3095
+    },
+    {
+      "epoch": 83.68,
+      "learning_rate": 9.5e-05,
+      "loss": 0.0593,
+      "step": 3096
+    },
+    {
+      "epoch": 83.7,
+      "learning_rate": 9.484375000000001e-05,
+      "loss": 0.0998,
+      "step": 3097
+    },
+    {
+      "epoch": 83.73,
+      "learning_rate": 9.468749999999999e-05,
+      "loss": 0.0962,
+      "step": 3098
+    },
+    {
+      "epoch": 83.76,
+      "learning_rate": 9.453125e-05,
+      "loss": 0.0927,
+      "step": 3099
+    },
+    {
+      "epoch": 83.78,
+      "learning_rate": 9.4375e-05,
+      "loss": 0.0885,
+      "step": 3100
+    },
+    {
+      "epoch": 83.81,
+      "learning_rate": 9.421875000000001e-05,
+      "loss": 0.0916,
+      "step": 3101
+    },
+    {
+      "epoch": 83.84,
+      "learning_rate": 9.406249999999999e-05,
+      "loss": 0.107,
+      "step": 3102
+    },
+    {
+      "epoch": 83.86,
+      "learning_rate": 9.390625e-05,
+      "loss": 0.0641,
+      "step": 3103
+    },
+    {
+      "epoch": 83.89,
+      "learning_rate": 9.375e-05,
+      "loss": 0.0945,
+      "step": 3104
+    },
+    {
+      "epoch": 83.92,
+      "learning_rate": 9.359375000000001e-05,
+      "loss": 0.1283,
+      "step": 3105
+    },
+    {
+      "epoch": 83.95,
+      "learning_rate": 9.343750000000001e-05,
+      "loss": 0.1187,
+      "step": 3106
+    },
+    {
+      "epoch": 83.97,
+      "learning_rate": 9.328125e-05,
+      "loss": 0.108,
+      "step": 3107
+    },
+    {
+      "epoch": 84.0,
+      "learning_rate": 9.3125e-05,
+      "loss": 0.0792,
+      "step": 3108
+    },
+    {
+      "epoch": 84.03,
+      "learning_rate": 9.296875e-05,
+      "loss": 0.1172,
+      "step": 3109
+    },
+    {
+      "epoch": 84.05,
+      "learning_rate": 9.281250000000001e-05,
+      "loss": 0.1003,
+      "step": 3110
+    },
+    {
+      "epoch": 84.08,
+      "learning_rate": 9.265624999999999e-05,
+      "loss": 0.1083,
+      "step": 3111
+    },
+    {
+      "epoch": 84.11,
+      "learning_rate": 9.25e-05,
+      "loss": 0.1061,
+      "step": 3112
+    },
+    {
+      "epoch": 84.14,
+      "learning_rate": 9.234375e-05,
+      "loss": 0.0732,
+      "step": 3113
+    },
+    {
+      "epoch": 84.16,
+      "learning_rate": 9.218750000000001e-05,
+      "loss": 0.0908,
+      "step": 3114
+    },
+    {
+      "epoch": 84.19,
+      "learning_rate": 9.203124999999999e-05,
+      "loss": 0.0904,
+      "step": 3115
+    },
+    {
+      "epoch": 84.22,
+      "learning_rate": 9.1875e-05,
+      "loss": 0.1104,
+      "step": 3116
+    },
+    {
+      "epoch": 84.24,
+      "learning_rate": 9.171875e-05,
+      "loss": 0.0858,
+      "step": 3117
+    },
+    {
+      "epoch": 84.27,
+      "learning_rate": 9.156250000000001e-05,
+      "loss": 0.099,
+      "step": 3118
+    },
+    {
+      "epoch": 84.3,
+      "learning_rate": 9.140625e-05,
+      "loss": 0.0916,
+      "step": 3119
+    },
+    {
+      "epoch": 84.32,
+      "learning_rate": 9.125e-05,
+      "loss": 0.0849,
+      "step": 3120
+    },
+    {
+      "epoch": 84.35,
+      "learning_rate": 9.109375e-05,
+      "loss": 0.0879,
+      "step": 3121
+    },
+    {
+      "epoch": 84.38,
+      "learning_rate": 9.093750000000001e-05,
+      "loss": 0.1014,
+      "step": 3122
+    },
+    {
+      "epoch": 84.41,
+      "learning_rate": 9.078125e-05,
+      "loss": 0.0823,
+      "step": 3123
+    },
+    {
+      "epoch": 84.43,
+      "learning_rate": 9.0625e-05,
+      "loss": 0.075,
+      "step": 3124
+    },
+    {
+      "epoch": 84.46,
+      "learning_rate": 9.046875e-05,
+      "loss": 0.0803,
+      "step": 3125
+    },
+    {
+      "epoch": 84.49,
+      "learning_rate": 9.03125e-05,
+      "loss": 0.0944,
+      "step": 3126
+    },
+    {
+      "epoch": 84.51,
+      "learning_rate": 9.015625e-05,
+      "loss": 0.072,
+      "step": 3127
+    },
+    {
+      "epoch": 84.54,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 0.0889,
+      "step": 3128
+    },
+    {
+      "epoch": 84.57,
+      "learning_rate": 8.984375e-05,
+      "loss": 0.0981,
+      "step": 3129
+    },
+    {
+      "epoch": 84.59,
+      "learning_rate": 8.96875e-05,
+      "loss": 0.0687,
+      "step": 3130
+    },
+    {
+      "epoch": 84.62,
+      "learning_rate": 8.953125000000001e-05,
+      "loss": 0.0995,
+      "step": 3131
+    },
+    {
+      "epoch": 84.65,
+      "learning_rate": 8.9375e-05,
+      "loss": 0.1023,
+      "step": 3132
+    },
+    {
+      "epoch": 84.68,
+      "learning_rate": 8.921875e-05,
+      "loss": 0.0733,
+      "step": 3133
+    },
+    {
+      "epoch": 84.7,
+      "learning_rate": 8.90625e-05,
+      "loss": 0.1385,
+      "step": 3134
+    },
+    {
+      "epoch": 84.73,
+      "learning_rate": 8.890625000000001e-05,
+      "loss": 0.1207,
+      "step": 3135
+    },
+    {
+      "epoch": 84.76,
+      "learning_rate": 8.875e-05,
+      "loss": 0.1029,
+      "step": 3136
+    },
+    {
+      "epoch": 84.78,
+      "learning_rate": 8.859375e-05,
+      "loss": 0.0867,
+      "step": 3137
+    },
+    {
+      "epoch": 84.81,
+      "learning_rate": 8.84375e-05,
+      "loss": 0.0872,
+      "step": 3138
+    },
+    {
+      "epoch": 84.84,
+      "learning_rate": 8.828125000000001e-05,
+      "loss": 0.1121,
+      "step": 3139
+    },
+    {
+      "epoch": 84.86,
+      "learning_rate": 8.8125e-05,
+      "loss": 0.1031,
+      "step": 3140
+    },
+    {
+      "epoch": 84.89,
+      "learning_rate": 8.796875e-05,
+      "loss": 0.1192,
+      "step": 3141
+    },
+    {
+      "epoch": 84.92,
+      "learning_rate": 8.78125e-05,
+      "loss": 0.1064,
+      "step": 3142
+    },
+    {
+      "epoch": 84.95,
+      "learning_rate": 8.765625e-05,
+      "loss": 0.0801,
+      "step": 3143
+    },
+    {
+      "epoch": 84.97,
+      "learning_rate": 8.75e-05,
+      "loss": 0.0866,
+      "step": 3144
+    },
+    {
+      "epoch": 85.0,
+      "learning_rate": 8.734375e-05,
+      "loss": 0.0862,
+      "step": 3145
+    },
+    {
+      "epoch": 85.03,
+      "learning_rate": 8.71875e-05,
+      "loss": 0.1232,
+      "step": 3146
+    },
+    {
+      "epoch": 85.05,
+      "learning_rate": 8.703125e-05,
+      "loss": 0.1038,
+      "step": 3147
+    },
+    {
+      "epoch": 85.08,
+      "learning_rate": 8.6875e-05,
+      "loss": 0.0944,
+      "step": 3148
+    },
+    {
+      "epoch": 85.11,
+      "learning_rate": 8.671875e-05,
+      "loss": 0.1035,
+      "step": 3149
+    },
+    {
+      "epoch": 85.14,
+      "learning_rate": 8.65625e-05,
+      "loss": 0.1249,
+      "step": 3150
+    },
+    {
+      "epoch": 85.16,
+      "learning_rate": 8.640625e-05,
+      "loss": 0.0748,
+      "step": 3151
+    },
+    {
+      "epoch": 85.19,
+      "learning_rate": 8.625e-05,
+      "loss": 0.1159,
+      "step": 3152
+    },
+    {
+      "epoch": 85.22,
+      "learning_rate": 8.609375e-05,
+      "loss": 0.102,
+      "step": 3153
+    },
+    {
+      "epoch": 85.24,
+      "learning_rate": 8.59375e-05,
+      "loss": 0.1058,
+      "step": 3154
+    },
+    {
+      "epoch": 85.27,
+      "learning_rate": 8.578125e-05,
+      "loss": 0.076,
+      "step": 3155
+    },
+    {
+      "epoch": 85.3,
+      "learning_rate": 8.562500000000001e-05,
+      "loss": 0.1013,
+      "step": 3156
+    },
+    {
+      "epoch": 85.32,
+      "learning_rate": 8.546875e-05,
+      "loss": 0.0823,
+      "step": 3157
+    },
+    {
+      "epoch": 85.35,
+      "learning_rate": 8.531250000000001e-05,
+      "loss": 0.1278,
+      "step": 3158
+    },
+    {
+      "epoch": 85.38,
+      "learning_rate": 8.515625e-05,
+      "loss": 0.0859,
+      "step": 3159
+    },
+    {
+      "epoch": 85.41,
+      "learning_rate": 8.5e-05,
+      "loss": 0.0907,
+      "step": 3160
+    },
+    {
+      "epoch": 85.43,
+      "learning_rate": 8.484375e-05,
+      "loss": 0.0959,
+      "step": 3161
+    },
+    {
+      "epoch": 85.46,
+      "learning_rate": 8.46875e-05,
+      "loss": 0.1092,
+      "step": 3162
+    },
+    {
+      "epoch": 85.49,
+      "learning_rate": 8.453125e-05,
+      "loss": 0.06,
+      "step": 3163
+    },
+    {
+      "epoch": 85.51,
+      "learning_rate": 8.4375e-05,
+      "loss": 0.0816,
+      "step": 3164
+    },
+    {
+      "epoch": 85.54,
+      "learning_rate": 8.421875e-05,
+      "loss": 0.1027,
+      "step": 3165
+    },
+    {
+      "epoch": 85.57,
+      "learning_rate": 8.40625e-05,
+      "loss": 0.0924,
+      "step": 3166
+    },
+    {
+      "epoch": 85.59,
+      "learning_rate": 8.390625000000001e-05,
+      "loss": 0.0883,
+      "step": 3167
+    },
+    {
+      "epoch": 85.62,
+      "learning_rate": 8.375e-05,
+      "loss": 0.1288,
+      "step": 3168
+    },
+    {
+      "epoch": 85.65,
+      "learning_rate": 8.359375e-05,
+      "loss": 0.1351,
+      "step": 3169
+    },
+    {
+      "epoch": 85.68,
+      "learning_rate": 8.34375e-05,
+      "loss": 0.0948,
+      "step": 3170
+    },
+    {
+      "epoch": 85.7,
+      "learning_rate": 8.328125000000001e-05,
+      "loss": 0.1093,
+      "step": 3171
+    },
+    {
+      "epoch": 85.73,
+      "learning_rate": 8.3125e-05,
+      "loss": 0.1073,
+      "step": 3172
+    },
+    {
+      "epoch": 85.76,
+      "learning_rate": 8.296875e-05,
+      "loss": 0.0983,
+      "step": 3173
+    },
+    {
+      "epoch": 85.78,
+      "learning_rate": 8.28125e-05,
+      "loss": 0.0857,
+      "step": 3174
+    },
+    {
+      "epoch": 85.81,
+      "learning_rate": 8.265625e-05,
+      "loss": 0.068,
+      "step": 3175
+    },
+    {
+      "epoch": 85.84,
+      "learning_rate": 8.25e-05,
+      "loss": 0.0738,
+      "step": 3176
+    },
+    {
+      "epoch": 85.86,
+      "learning_rate": 8.234374999999999e-05,
+      "loss": 0.0774,
+      "step": 3177
+    },
+    {
+      "epoch": 85.89,
+      "learning_rate": 8.21875e-05,
+      "loss": 0.0996,
+      "step": 3178
+    },
+    {
+      "epoch": 85.92,
+      "learning_rate": 8.203125e-05,
+      "loss": 0.0684,
+      "step": 3179
+    },
+    {
+      "epoch": 85.95,
+      "learning_rate": 8.187500000000001e-05,
+      "loss": 0.0998,
+      "step": 3180
+    },
+    {
+      "epoch": 85.97,
+      "learning_rate": 8.171875e-05,
+      "loss": 0.0731,
+      "step": 3181
+    },
+    {
+      "epoch": 86.0,
+      "learning_rate": 8.15625e-05,
+      "loss": 0.0338,
+      "step": 3182
+    },
+    {
+      "epoch": 86.03,
+      "learning_rate": 8.140625e-05,
+      "loss": 0.1147,
+      "step": 3183
+    },
+    {
+      "epoch": 86.05,
+      "learning_rate": 8.125000000000001e-05,
+      "loss": 0.0998,
+      "step": 3184
+    },
+    {
+      "epoch": 86.08,
+      "learning_rate": 8.109375e-05,
+      "loss": 0.0724,
+      "step": 3185
+    },
+    {
+      "epoch": 86.11,
+      "learning_rate": 8.09375e-05,
+      "loss": 0.1032,
+      "step": 3186
+    },
+    {
+      "epoch": 86.14,
+      "learning_rate": 8.078125e-05,
+      "loss": 0.0977,
+      "step": 3187
+    },
+    {
+      "epoch": 86.16,
+      "learning_rate": 8.062500000000001e-05,
+      "loss": 0.0821,
+      "step": 3188
+    },
+    {
+      "epoch": 86.19,
+      "learning_rate": 8.046875e-05,
+      "loss": 0.0735,
+      "step": 3189
+    },
+    {
+      "epoch": 86.22,
+      "learning_rate": 8.03125e-05,
+      "loss": 0.1098,
+      "step": 3190
+    },
+    {
+      "epoch": 86.24,
+      "learning_rate": 8.015625e-05,
+      "loss": 0.1162,
+      "step": 3191
+    },
+    {
+      "epoch": 86.27,
+      "learning_rate": 8e-05,
+      "loss": 0.1002,
+      "step": 3192
+    },
+    {
+      "epoch": 86.3,
+      "learning_rate": 7.984375000000001e-05,
+      "loss": 0.0931,
+      "step": 3193
+    },
+    {
+      "epoch": 86.32,
+      "learning_rate": 7.968749999999999e-05,
+      "loss": 0.0782,
+      "step": 3194
+    },
+    {
+      "epoch": 86.35,
+      "learning_rate": 7.953125e-05,
+      "loss": 0.096,
+      "step": 3195
+    },
+    {
+      "epoch": 86.38,
+      "learning_rate": 7.9375e-05,
+      "loss": 0.1052,
+      "step": 3196
+    },
+    {
+      "epoch": 86.41,
+      "learning_rate": 7.921875000000001e-05,
+      "loss": 0.1054,
+      "step": 3197
+    },
+    {
+      "epoch": 86.43,
+      "learning_rate": 7.906249999999999e-05,
+      "loss": 0.0873,
+      "step": 3198
+    },
+    {
+      "epoch": 86.46,
+      "learning_rate": 7.890625e-05,
+      "loss": 0.0927,
+      "step": 3199
+    },
+    {
+      "epoch": 86.49,
+      "learning_rate": 7.875e-05,
+      "loss": 0.0933,
+      "step": 3200
+    },
+    {
+      "epoch": 86.51,
+      "learning_rate": 7.859375000000001e-05,
+      "loss": 0.0865,
+      "step": 3201
+    },
+    {
+      "epoch": 86.54,
+      "learning_rate": 7.843749999999999e-05,
+      "loss": 0.1242,
+      "step": 3202
+    },
+    {
+      "epoch": 86.57,
+      "learning_rate": 7.828125e-05,
+      "loss": 0.1059,
+      "step": 3203
+    },
+    {
+      "epoch": 86.59,
+      "learning_rate": 7.8125e-05,
+      "loss": 0.1133,
+      "step": 3204
+    },
+    {
+      "epoch": 86.62,
+      "learning_rate": 7.796875000000001e-05,
+      "loss": 0.0742,
+      "step": 3205
+    },
+    {
+      "epoch": 86.65,
+      "learning_rate": 7.781250000000001e-05,
+      "loss": 0.0877,
+      "step": 3206
+    },
+    {
+      "epoch": 86.68,
+      "learning_rate": 7.765625e-05,
+      "loss": 0.0511,
+      "step": 3207
+    },
+    {
+      "epoch": 86.7,
+      "learning_rate": 7.75e-05,
+      "loss": 0.1048,
+      "step": 3208
+    },
+    {
+      "epoch": 86.73,
+      "learning_rate": 7.734375e-05,
+      "loss": 0.0885,
+      "step": 3209
+    },
+    {
+      "epoch": 86.76,
+      "learning_rate": 7.718750000000001e-05,
+      "loss": 0.0934,
+      "step": 3210
+    },
+    {
+      "epoch": 86.78,
+      "learning_rate": 7.703124999999999e-05,
+      "loss": 0.0836,
+      "step": 3211
+    },
+    {
+      "epoch": 86.81,
+      "learning_rate": 7.6875e-05,
+      "loss": 0.0959,
+      "step": 3212
+    },
+    {
+      "epoch": 86.84,
+      "learning_rate": 7.671875e-05,
+      "loss": 0.0734,
+      "step": 3213
+    },
+    {
+      "epoch": 86.86,
+      "learning_rate": 7.656250000000001e-05,
+      "loss": 0.0831,
+      "step": 3214
+    },
+    {
+      "epoch": 86.89,
+      "learning_rate": 7.640624999999999e-05,
+      "loss": 0.0844,
+      "step": 3215
+    },
+    {
+      "epoch": 86.92,
+      "learning_rate": 7.625e-05,
+      "loss": 0.0899,
+      "step": 3216
+    },
+    {
+      "epoch": 86.95,
+      "learning_rate": 7.609375e-05,
+      "loss": 0.0718,
+      "step": 3217
+    },
+    {
+      "epoch": 86.97,
+      "learning_rate": 7.593750000000001e-05,
+      "loss": 0.0617,
+      "step": 3218
+    },
+    {
+      "epoch": 87.0,
+      "learning_rate": 7.578125e-05,
+      "loss": 0.1059,
+      "step": 3219
+    },
+    {
+      "epoch": 87.03,
+      "learning_rate": 7.5625e-05,
+      "loss": 0.1241,
+      "step": 3220
+    },
+    {
+      "epoch": 87.05,
+      "learning_rate": 7.546875e-05,
+      "loss": 0.0726,
+      "step": 3221
+    },
+    {
+      "epoch": 87.08,
+      "learning_rate": 7.531250000000001e-05,
+      "loss": 0.0744,
+      "step": 3222
+    },
+    {
+      "epoch": 87.11,
+      "learning_rate": 7.515625e-05,
+      "loss": 0.1049,
+      "step": 3223
+    },
+    {
+      "epoch": 87.14,
+      "learning_rate": 7.5e-05,
+      "loss": 0.0871,
+      "step": 3224
+    },
+    {
+      "epoch": 87.16,
+      "learning_rate": 7.484375e-05,
+      "loss": 0.0598,
+      "step": 3225
+    },
+    {
+      "epoch": 87.19,
+      "learning_rate": 7.46875e-05,
+      "loss": 0.1015,
+      "step": 3226
+    },
+    {
+      "epoch": 87.22,
+      "learning_rate": 7.453125e-05,
+      "loss": 0.1028,
+      "step": 3227
+    },
+    {
+      "epoch": 87.24,
+      "learning_rate": 7.437499999999999e-05,
+      "loss": 0.0996,
+      "step": 3228
+    },
+    {
+      "epoch": 87.27,
+      "learning_rate": 7.421875e-05,
+      "loss": 0.1077,
+      "step": 3229
+    },
+    {
+      "epoch": 87.3,
+      "learning_rate": 7.40625e-05,
+      "loss": 0.0754,
+      "step": 3230
+    },
+    {
+      "epoch": 87.32,
+      "learning_rate": 7.390625000000001e-05,
+      "loss": 0.0745,
+      "step": 3231
+    },
+    {
+      "epoch": 87.35,
+      "learning_rate": 7.375e-05,
+      "loss": 0.0792,
+      "step": 3232
+    },
+    {
+      "epoch": 87.38,
+      "learning_rate": 7.359375e-05,
+      "loss": 0.1177,
+      "step": 3233
+    },
+    {
+      "epoch": 87.41,
+      "learning_rate": 7.34375e-05,
+      "loss": 0.0991,
+      "step": 3234
+    },
+    {
+      "epoch": 87.43,
+      "learning_rate": 7.328125000000001e-05,
+      "loss": 0.1004,
+      "step": 3235
+    },
+    {
+      "epoch": 87.46,
+      "learning_rate": 7.3125e-05,
+      "loss": 0.1172,
+      "step": 3236
+    },
+    {
+      "epoch": 87.49,
+      "learning_rate": 7.296875e-05,
+      "loss": 0.0558,
+      "step": 3237
+    },
+    {
+      "epoch": 87.51,
+      "learning_rate": 7.28125e-05,
+      "loss": 0.0673,
+      "step": 3238
+    },
+    {
+      "epoch": 87.54,
+      "learning_rate": 7.265625000000001e-05,
+      "loss": 0.1192,
+      "step": 3239
+    },
+    {
+      "epoch": 87.57,
+      "learning_rate": 7.25e-05,
+      "loss": 0.0941,
+      "step": 3240
+    },
+    {
+      "epoch": 87.59,
+      "learning_rate": 7.234375e-05,
+      "loss": 0.1098,
+      "step": 3241
+    },
+    {
+      "epoch": 87.62,
+      "learning_rate": 7.21875e-05,
+      "loss": 0.0929,
+      "step": 3242
+    },
+    {
+      "epoch": 87.65,
+      "learning_rate": 7.203125e-05,
+      "loss": 0.0813,
+      "step": 3243
+    },
+    {
+      "epoch": 87.68,
+      "learning_rate": 7.1875e-05,
+      "loss": 0.0946,
+      "step": 3244
+    },
+    {
+      "epoch": 87.7,
+      "learning_rate": 7.171875e-05,
+      "loss": 0.0783,
+      "step": 3245
+    },
+    {
+      "epoch": 87.73,
+      "learning_rate": 7.15625e-05,
+      "loss": 0.0688,
+      "step": 3246
+    },
+    {
+      "epoch": 87.76,
+      "learning_rate": 7.140625e-05,
+      "loss": 0.1138,
+      "step": 3247
+    },
+    {
+      "epoch": 87.78,
+      "learning_rate": 7.125e-05,
+      "loss": 0.0797,
+      "step": 3248
+    },
+    {
+      "epoch": 87.81,
+      "learning_rate": 7.109375e-05,
+      "loss": 0.1044,
+      "step": 3249
+    },
+    {
+      "epoch": 87.84,
+      "learning_rate": 7.09375e-05,
+      "loss": 0.0403,
+      "step": 3250
+    },
+    {
+      "epoch": 87.86,
+      "learning_rate": 7.078125e-05,
+      "loss": 0.0972,
+      "step": 3251
+    },
+    {
+      "epoch": 87.89,
+      "learning_rate": 7.0625e-05,
+      "loss": 0.1075,
+      "step": 3252
+    },
+    {
+      "epoch": 87.92,
+      "learning_rate": 7.046875e-05,
+      "loss": 0.0939,
+      "step": 3253
+    },
+    {
+      "epoch": 87.95,
+      "learning_rate": 7.031250000000001e-05,
+      "loss": 0.0816,
+      "step": 3254
+    },
+    {
+      "epoch": 87.97,
+      "learning_rate": 7.015625e-05,
+      "loss": 0.1189,
+      "step": 3255
+    },
+    {
+      "epoch": 88.0,
+      "learning_rate": 7.000000000000001e-05,
+      "loss": 0.0922,
+      "step": 3256
+    },
+    {
+      "epoch": 88.03,
+      "learning_rate": 6.984375e-05,
+      "loss": 0.1062,
+      "step": 3257
+    },
+    {
+      "epoch": 88.05,
+      "learning_rate": 6.96875e-05,
+      "loss": 0.0759,
+      "step": 3258
+    },
+    {
+      "epoch": 88.08,
+      "learning_rate": 6.953125e-05,
+      "loss": 0.0974,
+      "step": 3259
+    },
+    {
+      "epoch": 88.11,
+      "learning_rate": 6.9375e-05,
+      "loss": 0.0759,
+      "step": 3260
+    },
+    {
+      "epoch": 88.14,
+      "learning_rate": 6.921875e-05,
+      "loss": 0.0752,
+      "step": 3261
+    },
+    {
+      "epoch": 88.16,
+      "learning_rate": 6.90625e-05,
+      "loss": 0.0906,
+      "step": 3262
+    },
+    {
+      "epoch": 88.19,
+      "learning_rate": 6.890625e-05,
+      "loss": 0.1055,
+      "step": 3263
+    },
+    {
+      "epoch": 88.22,
+      "learning_rate": 6.875e-05,
+      "loss": 0.0921,
+      "step": 3264
+    },
+    {
+      "epoch": 88.24,
+      "learning_rate": 6.859375e-05,
+      "loss": 0.0924,
+      "step": 3265
+    },
+    {
+      "epoch": 88.27,
+      "learning_rate": 6.84375e-05,
+      "loss": 0.0686,
+      "step": 3266
+    },
+    {
+      "epoch": 88.3,
+      "learning_rate": 6.828125000000001e-05,
+      "loss": 0.0855,
+      "step": 3267
+    },
+    {
+      "epoch": 88.32,
+      "learning_rate": 6.8125e-05,
+      "loss": 0.0736,
+      "step": 3268
+    },
+    {
+      "epoch": 88.35,
+      "learning_rate": 6.796875e-05,
+      "loss": 0.1042,
+      "step": 3269
+    },
+    {
+      "epoch": 88.38,
+      "learning_rate": 6.78125e-05,
+      "loss": 0.0674,
+      "step": 3270
+    },
+    {
+      "epoch": 88.41,
+      "learning_rate": 6.765625000000001e-05,
+      "loss": 0.1082,
+      "step": 3271
+    },
+    {
+      "epoch": 88.43,
+      "learning_rate": 6.75e-05,
+      "loss": 0.0725,
+      "step": 3272
+    },
+    {
+      "epoch": 88.46,
+      "learning_rate": 6.734375e-05,
+      "loss": 0.0851,
+      "step": 3273
+    },
+    {
+      "epoch": 88.49,
+      "learning_rate": 6.71875e-05,
+      "loss": 0.0597,
+      "step": 3274
+    },
+    {
+      "epoch": 88.51,
+      "learning_rate": 6.703125e-05,
+      "loss": 0.0659,
+      "step": 3275
+    },
+    {
+      "epoch": 88.54,
+      "learning_rate": 6.6875e-05,
+      "loss": 0.0799,
+      "step": 3276
+    },
+    {
+      "epoch": 88.57,
+      "learning_rate": 6.671874999999999e-05,
+      "loss": 0.1055,
+      "step": 3277
+    },
+    {
+      "epoch": 88.59,
+      "learning_rate": 6.65625e-05,
+      "loss": 0.0832,
+      "step": 3278
+    },
+    {
+      "epoch": 88.62,
+      "learning_rate": 6.640625e-05,
+      "loss": 0.0842,
+      "step": 3279
+    },
+    {
+      "epoch": 88.65,
+      "learning_rate": 6.625000000000001e-05,
+      "loss": 0.086,
+      "step": 3280
+    },
+    {
+      "epoch": 88.68,
+      "learning_rate": 6.609375e-05,
+      "loss": 0.0453,
+      "step": 3281
+    },
+    {
+      "epoch": 88.7,
+      "learning_rate": 6.59375e-05,
+      "loss": 0.1102,
+      "step": 3282
+    },
+    {
+      "epoch": 88.73,
+      "learning_rate": 6.578125e-05,
+      "loss": 0.0611,
+      "step": 3283
+    },
+    {
+      "epoch": 88.76,
+      "learning_rate": 6.562500000000001e-05,
+      "loss": 0.0737,
+      "step": 3284
+    },
+    {
+      "epoch": 88.78,
+      "learning_rate": 6.546875e-05,
+      "loss": 0.1026,
+      "step": 3285
+    },
+    {
+      "epoch": 88.81,
+      "learning_rate": 6.53125e-05,
+      "loss": 0.1114,
+      "step": 3286
+    },
+    {
+      "epoch": 88.84,
+      "learning_rate": 6.515625e-05,
+      "loss": 0.1002,
+      "step": 3287
+    },
+    {
+      "epoch": 88.86,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 0.0722,
+      "step": 3288
+    },
+    {
+      "epoch": 88.89,
+      "learning_rate": 6.484375e-05,
+      "loss": 0.1183,
+      "step": 3289
+    },
+    {
+      "epoch": 88.92,
+      "learning_rate": 6.46875e-05,
+      "loss": 0.0816,
+      "step": 3290
+    },
+    {
+      "epoch": 88.95,
+      "learning_rate": 6.453125e-05,
+      "loss": 0.1141,
+      "step": 3291
+    },
+    {
+      "epoch": 88.97,
+      "learning_rate": 6.4375e-05,
+      "loss": 0.0517,
+      "step": 3292
+    },
+    {
+      "epoch": 89.0,
+      "learning_rate": 6.421875000000001e-05,
+      "loss": 0.1365,
+      "step": 3293
+    },
+    {
+      "epoch": 89.03,
+      "learning_rate": 6.406249999999999e-05,
+      "loss": 0.1222,
+      "step": 3294
+    },
+    {
+      "epoch": 89.05,
+      "learning_rate": 6.390625e-05,
+      "loss": 0.0904,
+      "step": 3295
+    },
+    {
+      "epoch": 89.08,
+      "learning_rate": 6.375e-05,
+      "loss": 0.1267,
+      "step": 3296
+    },
+    {
+      "epoch": 89.11,
+      "learning_rate": 6.359375000000001e-05,
+      "loss": 0.1223,
+      "step": 3297
+    },
+    {
+      "epoch": 89.14,
+      "learning_rate": 6.343749999999999e-05,
+      "loss": 0.0922,
+      "step": 3298
+    },
+    {
+      "epoch": 89.16,
+      "learning_rate": 6.328125e-05,
+      "loss": 0.0918,
+      "step": 3299
+    },
+    {
+      "epoch": 89.19,
+      "learning_rate": 6.3125e-05,
+      "loss": 0.0906,
+      "step": 3300
+    },
+    {
+      "epoch": 89.22,
+      "learning_rate": 6.296875000000001e-05,
+      "loss": 0.1146,
+      "step": 3301
+    },
+    {
+      "epoch": 89.24,
+      "learning_rate": 6.281249999999999e-05,
+      "loss": 0.0831,
+      "step": 3302
+    },
+    {
+      "epoch": 89.27,
+      "learning_rate": 6.265625e-05,
+      "loss": 0.075,
+      "step": 3303
+    },
+    {
+      "epoch": 89.3,
+      "learning_rate": 6.25e-05,
+      "loss": 0.1198,
+      "step": 3304
+    },
+    {
+      "epoch": 89.32,
+      "learning_rate": 6.234375000000001e-05,
+      "loss": 0.0745,
+      "step": 3305
+    },
+    {
+      "epoch": 89.35,
+      "learning_rate": 6.21875e-05,
+      "loss": 0.0682,
+      "step": 3306
+    },
+    {
+      "epoch": 89.38,
+      "learning_rate": 6.203125000000001e-05,
+      "loss": 0.1247,
+      "step": 3307
+    },
+    {
+      "epoch": 89.41,
+      "learning_rate": 6.1875e-05,
+      "loss": 0.0832,
+      "step": 3308
+    },
+    {
+      "epoch": 89.43,
+      "learning_rate": 6.171875e-05,
+      "loss": 0.0842,
+      "step": 3309
+    },
+    {
+      "epoch": 89.46,
+      "learning_rate": 6.15625e-05,
+      "loss": 0.1074,
+      "step": 3310
+    },
+    {
+      "epoch": 89.49,
+      "learning_rate": 6.140625e-05,
+      "loss": 0.0771,
+      "step": 3311
+    },
+    {
+      "epoch": 89.51,
+      "learning_rate": 6.125e-05,
+      "loss": 0.0881,
+      "step": 3312
+    },
+    {
+      "epoch": 89.54,
+      "learning_rate": 6.109375e-05,
+      "loss": 0.0888,
+      "step": 3313
+    },
+    {
+      "epoch": 89.57,
+      "learning_rate": 6.09375e-05,
+      "loss": 0.1003,
+      "step": 3314
+    },
+    {
+      "epoch": 89.59,
+      "learning_rate": 6.0781250000000004e-05,
+      "loss": 0.1144,
+      "step": 3315
+    },
+    {
+      "epoch": 89.62,
+      "learning_rate": 6.0624999999999996e-05,
+      "loss": 0.0746,
+      "step": 3316
+    },
+    {
+      "epoch": 89.65,
+      "learning_rate": 6.046875e-05,
+      "loss": 0.1254,
+      "step": 3317
+    },
+    {
+      "epoch": 89.68,
+      "learning_rate": 6.03125e-05,
+      "loss": 0.0727,
+      "step": 3318
+    },
+    {
+      "epoch": 89.7,
+      "learning_rate": 6.015625e-05,
+      "loss": 0.0925,
+      "step": 3319
+    },
+    {
+      "epoch": 89.73,
+      "learning_rate": 6e-05,
+      "loss": 0.1144,
+      "step": 3320
+    },
+    {
+      "epoch": 89.76,
+      "learning_rate": 5.984375e-05,
+      "loss": 0.0572,
+      "step": 3321
+    },
+    {
+      "epoch": 89.78,
+      "learning_rate": 5.96875e-05,
+      "loss": 0.0869,
+      "step": 3322
+    },
+    {
+      "epoch": 89.81,
+      "learning_rate": 5.953125e-05,
+      "loss": 0.0869,
+      "step": 3323
+    },
+    {
+      "epoch": 89.84,
+      "learning_rate": 5.9375e-05,
+      "loss": 0.0814,
+      "step": 3324
+    },
+    {
+      "epoch": 89.86,
+      "learning_rate": 5.921875e-05,
+      "loss": 0.0625,
+      "step": 3325
+    },
+    {
+      "epoch": 89.89,
+      "learning_rate": 5.90625e-05,
+      "loss": 0.1082,
+      "step": 3326
+    },
+    {
+      "epoch": 89.92,
+      "learning_rate": 5.890625e-05,
+      "loss": 0.0851,
+      "step": 3327
+    },
+    {
+      "epoch": 89.95,
+      "learning_rate": 5.875e-05,
+      "loss": 0.0723,
+      "step": 3328
+    },
+    {
+      "epoch": 89.97,
+      "learning_rate": 5.859375e-05,
+      "loss": 0.1064,
+      "step": 3329
+    },
+    {
+      "epoch": 90.0,
+      "learning_rate": 5.8437500000000004e-05,
+      "loss": 0.0359,
+      "step": 3330
+    },
+    {
+      "epoch": 90.03,
+      "learning_rate": 5.8281250000000004e-05,
+      "loss": 0.0797,
+      "step": 3331
+    },
+    {
+      "epoch": 90.05,
+      "learning_rate": 5.8125e-05,
+      "loss": 0.1041,
+      "step": 3332
+    },
+    {
+      "epoch": 90.08,
+      "learning_rate": 5.796875e-05,
+      "loss": 0.0938,
+      "step": 3333
+    },
+    {
+      "epoch": 90.11,
+      "learning_rate": 5.78125e-05,
+      "loss": 0.1111,
+      "step": 3334
+    },
+    {
+      "epoch": 90.14,
+      "learning_rate": 5.765625e-05,
+      "loss": 0.0636,
+      "step": 3335
+    },
+    {
+      "epoch": 90.16,
+      "learning_rate": 5.75e-05,
+      "loss": 0.0996,
+      "step": 3336
+    },
+    {
+      "epoch": 90.19,
+      "learning_rate": 5.734375e-05,
+      "loss": 0.0857,
+      "step": 3337
+    },
+    {
+      "epoch": 90.22,
+      "learning_rate": 5.71875e-05,
+      "loss": 0.092,
+      "step": 3338
+    },
+    {
+      "epoch": 90.24,
+      "learning_rate": 5.703125e-05,
+      "loss": 0.1223,
+      "step": 3339
+    },
+    {
+      "epoch": 90.27,
+      "learning_rate": 5.6875e-05,
+      "loss": 0.0946,
+      "step": 3340
+    },
+    {
+      "epoch": 90.3,
+      "learning_rate": 5.671875e-05,
+      "loss": 0.093,
+      "step": 3341
+    },
+    {
+      "epoch": 90.32,
+      "learning_rate": 5.6562500000000006e-05,
+      "loss": 0.0725,
+      "step": 3342
+    },
+    {
+      "epoch": 90.35,
+      "learning_rate": 5.640625e-05,
+      "loss": 0.0785,
+      "step": 3343
+    },
+    {
+      "epoch": 90.38,
+      "learning_rate": 5.6250000000000005e-05,
+      "loss": 0.1073,
+      "step": 3344
+    },
+    {
+      "epoch": 90.41,
+      "learning_rate": 5.609375e-05,
+      "loss": 0.0846,
+      "step": 3345
+    },
+    {
+      "epoch": 90.43,
+      "learning_rate": 5.5937500000000004e-05,
+      "loss": 0.0856,
+      "step": 3346
+    },
+    {
+      "epoch": 90.46,
+      "learning_rate": 5.578125e-05,
+      "loss": 0.0776,
+      "step": 3347
+    },
+    {
+      "epoch": 90.49,
+      "learning_rate": 5.5625000000000004e-05,
+      "loss": 0.0957,
+      "step": 3348
+    },
+    {
+      "epoch": 90.51,
+      "learning_rate": 5.5468749999999996e-05,
+      "loss": 0.0729,
+      "step": 3349
+    },
+    {
+      "epoch": 90.54,
+      "learning_rate": 5.53125e-05,
+      "loss": 0.0699,
+      "step": 3350
+    },
+    {
+      "epoch": 90.57,
+      "learning_rate": 5.5156249999999996e-05,
+      "loss": 0.1098,
+      "step": 3351
+    },
+    {
+      "epoch": 90.59,
+      "learning_rate": 5.5e-05,
+      "loss": 0.0521,
+      "step": 3352
+    },
+    {
+      "epoch": 90.62,
+      "learning_rate": 5.4843749999999995e-05,
+      "loss": 0.0996,
+      "step": 3353
+    },
+    {
+      "epoch": 90.65,
+      "learning_rate": 5.46875e-05,
+      "loss": 0.0884,
+      "step": 3354
+    },
+    {
+      "epoch": 90.68,
+      "learning_rate": 5.453125000000001e-05,
+      "loss": 0.0695,
+      "step": 3355
+    },
+    {
+      "epoch": 90.7,
+      "learning_rate": 5.4375e-05,
+      "loss": 0.0888,
+      "step": 3356
+    },
+    {
+      "epoch": 90.73,
+      "learning_rate": 5.421875000000001e-05,
+      "loss": 0.0909,
+      "step": 3357
+    },
+    {
+      "epoch": 90.76,
+      "learning_rate": 5.40625e-05,
+      "loss": 0.0704,
+      "step": 3358
+    },
+    {
+      "epoch": 90.78,
+      "learning_rate": 5.3906250000000006e-05,
+      "loss": 0.0821,
+      "step": 3359
+    },
+    {
+      "epoch": 90.81,
+      "learning_rate": 5.375e-05,
+      "loss": 0.088,
+      "step": 3360
+    },
+    {
+      "epoch": 90.84,
+      "learning_rate": 5.3593750000000005e-05,
+      "loss": 0.088,
+      "step": 3361
+    },
+    {
+      "epoch": 90.86,
+      "learning_rate": 5.34375e-05,
+      "loss": 0.0831,
+      "step": 3362
+    },
+    {
+      "epoch": 90.89,
+      "learning_rate": 5.3281250000000004e-05,
+      "loss": 0.0838,
+      "step": 3363
+    },
+    {
+      "epoch": 90.92,
+      "learning_rate": 5.3125e-05,
+      "loss": 0.0766,
+      "step": 3364
+    },
+    {
+      "epoch": 90.95,
+      "learning_rate": 5.2968750000000003e-05,
+      "loss": 0.0903,
+      "step": 3365
+    },
+    {
+      "epoch": 90.97,
+      "learning_rate": 5.2812499999999996e-05,
+      "loss": 0.0708,
+      "step": 3366
+    },
+    {
+      "epoch": 91.0,
+      "learning_rate": 5.265625e-05,
+      "loss": 0.0129,
+      "step": 3367
+    },
+    {
+      "epoch": 91.03,
+      "learning_rate": 5.25e-05,
+      "loss": 0.0761,
+      "step": 3368
+    },
+    {
+      "epoch": 91.05,
+      "learning_rate": 5.234375e-05,
+      "loss": 0.0952,
+      "step": 3369
+    },
+    {
+      "epoch": 91.08,
+      "learning_rate": 5.21875e-05,
+      "loss": 0.0751,
+      "step": 3370
+    },
+    {
+      "epoch": 91.11,
+      "learning_rate": 5.203125e-05,
+      "loss": 0.0756,
+      "step": 3371
+    },
+    {
+      "epoch": 91.14,
+      "learning_rate": 5.1875e-05,
+      "loss": 0.0989,
+      "step": 3372
+    },
+    {
+      "epoch": 91.16,
+      "learning_rate": 5.171875e-05,
+      "loss": 0.083,
+      "step": 3373
+    },
+    {
+      "epoch": 91.19,
+      "learning_rate": 5.15625e-05,
+      "loss": 0.0746,
+      "step": 3374
+    },
+    {
+      "epoch": 91.22,
+      "learning_rate": 5.140625e-05,
+      "loss": 0.0925,
+      "step": 3375
+    },
+    {
+      "epoch": 91.24,
+      "learning_rate": 5.125e-05,
+      "loss": 0.0642,
+      "step": 3376
+    },
+    {
+      "epoch": 91.27,
+      "learning_rate": 5.109375e-05,
+      "loss": 0.0879,
+      "step": 3377
+    },
+    {
+      "epoch": 91.3,
+      "learning_rate": 5.09375e-05,
+      "loss": 0.0801,
+      "step": 3378
+    },
+    {
+      "epoch": 91.32,
+      "learning_rate": 5.0781250000000004e-05,
+      "loss": 0.0747,
+      "step": 3379
+    },
+    {
+      "epoch": 91.35,
+      "learning_rate": 5.0625000000000004e-05,
+      "loss": 0.0829,
+      "step": 3380
+    },
+    {
+      "epoch": 91.38,
+      "learning_rate": 5.0468750000000004e-05,
+      "loss": 0.0842,
+      "step": 3381
+    },
+    {
+      "epoch": 91.41,
+      "learning_rate": 5.03125e-05,
+      "loss": 0.0833,
+      "step": 3382
+    },
+    {
+      "epoch": 91.43,
+      "learning_rate": 5.015625e-05,
+      "loss": 0.0748,
+      "step": 3383
+    },
+    {
+      "epoch": 91.46,
+      "learning_rate": 5e-05,
+      "loss": 0.0738,
+      "step": 3384
+    },
+    {
+      "epoch": 91.49,
+      "learning_rate": 4.984375e-05,
+      "loss": 0.0616,
+      "step": 3385
+    },
+    {
+      "epoch": 91.51,
+      "learning_rate": 4.96875e-05,
+      "loss": 0.0695,
+      "step": 3386
+    },
+    {
+      "epoch": 91.54,
+      "learning_rate": 4.953125e-05,
+      "loss": 0.0951,
+      "step": 3387
+    },
+    {
+      "epoch": 91.57,
+      "learning_rate": 4.9375e-05,
+      "loss": 0.109,
+      "step": 3388
+    },
+    {
+      "epoch": 91.59,
+      "learning_rate": 4.921875e-05,
+      "loss": 0.0752,
+      "step": 3389
+    },
+    {
+      "epoch": 91.62,
+      "learning_rate": 4.90625e-05,
+      "loss": 0.0966,
+      "step": 3390
+    },
+    {
+      "epoch": 91.65,
+      "learning_rate": 4.890625e-05,
+      "loss": 0.092,
+      "step": 3391
+    },
+    {
+      "epoch": 91.68,
+      "learning_rate": 4.8750000000000006e-05,
+      "loss": 0.0543,
+      "step": 3392
+    },
+    {
+      "epoch": 91.7,
+      "learning_rate": 4.859375e-05,
+      "loss": 0.0957,
+      "step": 3393
+    },
+    {
+      "epoch": 91.73,
+      "learning_rate": 4.8437500000000005e-05,
+      "loss": 0.072,
+      "step": 3394
+    },
+    {
+      "epoch": 91.76,
+      "learning_rate": 4.828125e-05,
+      "loss": 0.1071,
+      "step": 3395
+    },
+    {
+      "epoch": 91.78,
+      "learning_rate": 4.8125000000000004e-05,
+      "loss": 0.0704,
+      "step": 3396
+    },
+    {
+      "epoch": 91.81,
+      "learning_rate": 4.796875e-05,
+      "loss": 0.08,
+      "step": 3397
+    },
+    {
+      "epoch": 91.84,
+      "learning_rate": 4.7812500000000003e-05,
+      "loss": 0.0567,
+      "step": 3398
+    },
+    {
+      "epoch": 91.86,
+      "learning_rate": 4.7656249999999996e-05,
+      "loss": 0.0747,
+      "step": 3399
+    },
+    {
+      "epoch": 91.89,
+      "learning_rate": 4.75e-05,
+      "loss": 0.0584,
+      "step": 3400
+    },
+    {
+      "epoch": 91.92,
+      "learning_rate": 4.7343749999999995e-05,
+      "loss": 0.0849,
+      "step": 3401
+    },
+    {
+      "epoch": 91.95,
+      "learning_rate": 4.71875e-05,
+      "loss": 0.0948,
+      "step": 3402
+    },
+    {
+      "epoch": 91.97,
+      "learning_rate": 4.7031249999999995e-05,
+      "loss": 0.0563,
+      "step": 3403
+    },
+    {
+      "epoch": 92.0,
+      "learning_rate": 4.6875e-05,
+      "loss": 0.0926,
+      "step": 3404
+    },
+    {
+      "epoch": 92.03,
+      "learning_rate": 4.671875000000001e-05,
+      "loss": 0.094,
+      "step": 3405
+    },
+    {
+      "epoch": 92.05,
+      "learning_rate": 4.65625e-05,
+      "loss": 0.0925,
+      "step": 3406
+    },
+    {
+      "epoch": 92.08,
+      "learning_rate": 4.6406250000000007e-05,
+      "loss": 0.0943,
+      "step": 3407
+    },
+    {
+      "epoch": 92.11,
+      "learning_rate": 4.625e-05,
+      "loss": 0.0527,
+      "step": 3408
+    },
+    {
+      "epoch": 92.14,
+      "learning_rate": 4.6093750000000006e-05,
+      "loss": 0.0796,
+      "step": 3409
+    },
+    {
+      "epoch": 92.16,
+      "learning_rate": 4.59375e-05,
+      "loss": 0.0798,
+      "step": 3410
+    },
+    {
+      "epoch": 92.19,
+      "learning_rate": 4.5781250000000005e-05,
+      "loss": 0.0811,
+      "step": 3411
+    },
+    {
+      "epoch": 92.22,
+      "learning_rate": 4.5625e-05,
+      "loss": 0.0897,
+      "step": 3412
+    },
+    {
+      "epoch": 92.24,
+      "learning_rate": 4.5468750000000004e-05,
+      "loss": 0.1032,
+      "step": 3413
+    },
+    {
+      "epoch": 92.27,
+      "learning_rate": 4.53125e-05,
+      "loss": 0.1023,
+      "step": 3414
+    },
+    {
+      "epoch": 92.3,
+      "learning_rate": 4.515625e-05,
+      "loss": 0.0797,
+      "step": 3415
+    },
+    {
+      "epoch": 92.32,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 0.0917,
+      "step": 3416
+    },
+    {
+      "epoch": 92.35,
+      "learning_rate": 4.484375e-05,
+      "loss": 0.0895,
+      "step": 3417
+    },
+    {
+      "epoch": 92.38,
+      "learning_rate": 4.46875e-05,
+      "loss": 0.0968,
+      "step": 3418
+    },
+    {
+      "epoch": 92.41,
+      "learning_rate": 4.453125e-05,
+      "loss": 0.0877,
+      "step": 3419
+    },
+    {
+      "epoch": 92.43,
+      "learning_rate": 4.4375e-05,
+      "loss": 0.0783,
+      "step": 3420
+    },
+    {
+      "epoch": 92.46,
+      "learning_rate": 4.421875e-05,
+      "loss": 0.0901,
+      "step": 3421
+    },
+    {
+      "epoch": 92.49,
+      "learning_rate": 4.40625e-05,
+      "loss": 0.0669,
+      "step": 3422
+    },
+    {
+      "epoch": 92.51,
+      "learning_rate": 4.390625e-05,
+      "loss": 0.0979,
+      "step": 3423
+    },
+    {
+      "epoch": 92.54,
+      "learning_rate": 4.375e-05,
+      "loss": 0.1081,
+      "step": 3424
+    },
+    {
+      "epoch": 92.57,
+      "learning_rate": 4.359375e-05,
+      "loss": 0.0779,
+      "step": 3425
+    },
+    {
+      "epoch": 92.59,
+      "learning_rate": 4.34375e-05,
+      "loss": 0.0904,
+      "step": 3426
+    },
+    {
+      "epoch": 92.62,
+      "learning_rate": 4.328125e-05,
+      "loss": 0.1005,
+      "step": 3427
+    },
+    {
+      "epoch": 92.65,
+      "learning_rate": 4.3125e-05,
+      "loss": 0.1041,
+      "step": 3428
+    },
+    {
+      "epoch": 92.68,
+      "learning_rate": 4.296875e-05,
+      "loss": 0.0615,
+      "step": 3429
+    },
+    {
+      "epoch": 92.7,
+      "learning_rate": 4.2812500000000004e-05,
+      "loss": 0.1069,
+      "step": 3430
+    },
+    {
+      "epoch": 92.73,
+      "learning_rate": 4.2656250000000003e-05,
+      "loss": 0.0719,
+      "step": 3431
+    },
+    {
+      "epoch": 92.76,
+      "learning_rate": 4.25e-05,
+      "loss": 0.0567,
+      "step": 3432
+    },
+    {
+      "epoch": 92.78,
+      "learning_rate": 4.234375e-05,
+      "loss": 0.1014,
+      "step": 3433
+    },
+    {
+      "epoch": 92.81,
+      "learning_rate": 4.21875e-05,
+      "loss": 0.0766,
+      "step": 3434
+    },
+    {
+      "epoch": 92.84,
+      "learning_rate": 4.203125e-05,
+      "loss": 0.0482,
+      "step": 3435
+    },
+    {
+      "epoch": 92.86,
+      "learning_rate": 4.1875e-05,
+      "loss": 0.085,
+      "step": 3436
+    },
+    {
+      "epoch": 92.89,
+      "learning_rate": 4.171875e-05,
+      "loss": 0.0935,
+      "step": 3437
+    },
+    {
+      "epoch": 92.92,
+      "learning_rate": 4.15625e-05,
+      "loss": 0.0812,
+      "step": 3438
+    },
+    {
+      "epoch": 92.95,
+      "learning_rate": 4.140625e-05,
+      "loss": 0.0826,
+      "step": 3439
+    },
+    {
+      "epoch": 92.97,
+      "learning_rate": 4.125e-05,
+      "loss": 0.1092,
+      "step": 3440
+    },
+    {
+      "epoch": 93.0,
+      "learning_rate": 4.109375e-05,
+      "loss": 0.0329,
+      "step": 3441
+    },
+    {
+      "epoch": 93.03,
+      "learning_rate": 4.0937500000000006e-05,
+      "loss": 0.1188,
+      "step": 3442
+    },
+    {
+      "epoch": 93.05,
+      "learning_rate": 4.078125e-05,
+      "loss": 0.0791,
+      "step": 3443
+    },
+    {
+      "epoch": 93.08,
+      "learning_rate": 4.0625000000000005e-05,
+      "loss": 0.0946,
+      "step": 3444
+    },
+    {
+      "epoch": 93.11,
+      "learning_rate": 4.046875e-05,
+      "loss": 0.1209,
+      "step": 3445
+    },
+    {
+      "epoch": 93.14,
+      "learning_rate": 4.0312500000000004e-05,
+      "loss": 0.0724,
+      "step": 3446
+    },
+    {
+      "epoch": 93.16,
+      "learning_rate": 4.015625e-05,
+      "loss": 0.0526,
+      "step": 3447
+    },
+    {
+      "epoch": 93.19,
+      "learning_rate": 4e-05,
+      "loss": 0.1153,
+      "step": 3448
+    },
+    {
+      "epoch": 93.22,
+      "learning_rate": 3.9843749999999996e-05,
+      "loss": 0.077,
+      "step": 3449
+    },
+    {
+      "epoch": 93.24,
+      "learning_rate": 3.96875e-05,
+      "loss": 0.059,
+      "step": 3450
+    },
+    {
+      "epoch": 93.27,
+      "learning_rate": 3.9531249999999995e-05,
+      "loss": 0.076,
+      "step": 3451
+    },
+    {
+      "epoch": 93.3,
+      "learning_rate": 3.9375e-05,
+      "loss": 0.0822,
+      "step": 3452
+    },
+    {
+      "epoch": 93.32,
+      "learning_rate": 3.9218749999999994e-05,
+      "loss": 0.0663,
+      "step": 3453
+    },
+    {
+      "epoch": 93.35,
+      "learning_rate": 3.90625e-05,
+      "loss": 0.0761,
+      "step": 3454
+    },
+    {
+      "epoch": 93.38,
+      "learning_rate": 3.890625000000001e-05,
+      "loss": 0.0872,
+      "step": 3455
+    },
+    {
+      "epoch": 93.41,
+      "learning_rate": 3.875e-05,
+      "loss": 0.0861,
+      "step": 3456
+    },
+    {
+      "epoch": 93.43,
+      "learning_rate": 3.8593750000000006e-05,
+      "loss": 0.0811,
+      "step": 3457
+    },
+    {
+      "epoch": 93.46,
+      "learning_rate": 3.84375e-05,
+      "loss": 0.0921,
+      "step": 3458
+    },
+    {
+      "epoch": 93.49,
+      "learning_rate": 3.8281250000000006e-05,
+      "loss": 0.0822,
+      "step": 3459
+    },
+    {
+      "epoch": 93.51,
+      "learning_rate": 3.8125e-05,
+      "loss": 0.0692,
+      "step": 3460
+    },
+    {
+      "epoch": 93.54,
+      "learning_rate": 3.7968750000000005e-05,
+      "loss": 0.0936,
+      "step": 3461
+    },
+    {
+      "epoch": 93.57,
+      "learning_rate": 3.78125e-05,
+      "loss": 0.0896,
+      "step": 3462
+    },
+    {
+      "epoch": 93.59,
+      "learning_rate": 3.7656250000000004e-05,
+      "loss": 0.0859,
+      "step": 3463
+    },
+    {
+      "epoch": 93.62,
+      "learning_rate": 3.75e-05,
+      "loss": 0.0883,
+      "step": 3464
+    },
+    {
+      "epoch": 93.65,
+      "learning_rate": 3.734375e-05,
+      "loss": 0.1111,
+      "step": 3465
+    },
+    {
+      "epoch": 93.68,
+      "learning_rate": 3.7187499999999996e-05,
+      "loss": 0.0506,
+      "step": 3466
+    },
+    {
+      "epoch": 93.7,
+      "learning_rate": 3.703125e-05,
+      "loss": 0.0875,
+      "step": 3467
+    },
+    {
+      "epoch": 93.73,
+      "learning_rate": 3.6875e-05,
+      "loss": 0.0749,
+      "step": 3468
+    },
+    {
+      "epoch": 93.76,
+      "learning_rate": 3.671875e-05,
+      "loss": 0.0991,
+      "step": 3469
+    },
+    {
+      "epoch": 93.78,
+      "learning_rate": 3.65625e-05,
+      "loss": 0.0736,
+      "step": 3470
+    },
+    {
+      "epoch": 93.81,
+      "learning_rate": 3.640625e-05,
+      "loss": 0.0949,
+      "step": 3471
+    },
+    {
+      "epoch": 93.84,
+      "learning_rate": 3.625e-05,
+      "loss": 0.0559,
+      "step": 3472
+    },
+    {
+      "epoch": 93.86,
+      "learning_rate": 3.609375e-05,
+      "loss": 0.0765,
+      "step": 3473
+    },
+    {
+      "epoch": 93.89,
+      "learning_rate": 3.59375e-05,
+      "loss": 0.1036,
+      "step": 3474
+    },
+    {
+      "epoch": 93.92,
+      "learning_rate": 3.578125e-05,
+      "loss": 0.0818,
+      "step": 3475
+    },
+    {
+      "epoch": 93.95,
+      "learning_rate": 3.5625e-05,
+      "loss": 0.0688,
+      "step": 3476
+    },
+    {
+      "epoch": 93.97,
+      "learning_rate": 3.546875e-05,
+      "loss": 0.0785,
+      "step": 3477
+    },
+    {
+      "epoch": 94.0,
+      "learning_rate": 3.53125e-05,
+      "loss": 0.0636,
+      "step": 3478
+    },
+    {
+      "epoch": 94.03,
+      "learning_rate": 3.5156250000000004e-05,
+      "loss": 0.1076,
+      "step": 3479
+    },
+    {
+      "epoch": 94.05,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 0.0822,
+      "step": 3480
+    },
+    {
+      "epoch": 94.08,
+      "learning_rate": 3.484375e-05,
+      "loss": 0.0857,
+      "step": 3481
+    },
+    {
+      "epoch": 94.11,
+      "learning_rate": 3.46875e-05,
+      "loss": 0.0912,
+      "step": 3482
+    },
+    {
+      "epoch": 94.14,
+      "learning_rate": 3.453125e-05,
+      "loss": 0.0924,
+      "step": 3483
+    },
+    {
+      "epoch": 94.16,
+      "learning_rate": 3.4375e-05,
+      "loss": 0.0495,
+      "step": 3484
+    },
+    {
+      "epoch": 94.19,
+      "learning_rate": 3.421875e-05,
+      "loss": 0.0694,
+      "step": 3485
+    },
+    {
+      "epoch": 94.22,
+      "learning_rate": 3.40625e-05,
+      "loss": 0.0981,
+      "step": 3486
+    },
+    {
+      "epoch": 94.24,
+      "learning_rate": 3.390625e-05,
+      "loss": 0.0993,
+      "step": 3487
+    },
+    {
+      "epoch": 94.27,
+      "learning_rate": 3.375e-05,
+      "loss": 0.0857,
+      "step": 3488
+    },
+    {
+      "epoch": 94.3,
+      "learning_rate": 3.359375e-05,
+      "loss": 0.084,
+      "step": 3489
+    },
+    {
+      "epoch": 94.32,
+      "learning_rate": 3.34375e-05,
+      "loss": 0.075,
+      "step": 3490
+    },
+    {
+      "epoch": 94.35,
+      "learning_rate": 3.328125e-05,
+      "loss": 0.0436,
+      "step": 3491
+    },
+    {
+      "epoch": 94.38,
+      "learning_rate": 3.3125000000000006e-05,
+      "loss": 0.0957,
+      "step": 3492
+    },
+    {
+      "epoch": 94.41,
+      "learning_rate": 3.296875e-05,
+      "loss": 0.0732,
+      "step": 3493
+    },
+    {
+      "epoch": 94.43,
+      "learning_rate": 3.2812500000000005e-05,
+      "loss": 0.0922,
+      "step": 3494
+    },
+    {
+      "epoch": 94.46,
+      "learning_rate": 3.265625e-05,
+      "loss": 0.0878,
+      "step": 3495
+    },
+    {
+      "epoch": 94.49,
+      "learning_rate": 3.2500000000000004e-05,
+      "loss": 0.0889,
+      "step": 3496
+    },
+    {
+      "epoch": 94.51,
+      "learning_rate": 3.234375e-05,
+      "loss": 0.0702,
+      "step": 3497
+    },
+    {
+      "epoch": 94.54,
+      "learning_rate": 3.21875e-05,
+      "loss": 0.095,
+      "step": 3498
+    },
+    {
+      "epoch": 94.57,
+      "learning_rate": 3.2031249999999996e-05,
+      "loss": 0.0893,
+      "step": 3499
+    },
+    {
+      "epoch": 94.59,
+      "learning_rate": 3.1875e-05,
+      "loss": 0.0639,
+      "step": 3500
+    },
+    {
+      "epoch": 94.59,
+      "eval_cer": 0.07157528598629664,
+      "eval_loss": 0.3788400888442993,
+      "eval_runtime": 11.9238,
+      "eval_samples_per_second": 138.127,
+      "eval_steps_per_second": 2.181,
+      "eval_wer": 0.2952711673986314,
+      "step": 3500
+    },
+    {
+      "epoch": 94.62,
+      "learning_rate": 3.1718749999999995e-05,
+      "loss": 0.0858,
+      "step": 3501
+    },
+    {
+      "epoch": 94.65,
+      "learning_rate": 3.15625e-05,
+      "loss": 0.0753,
+      "step": 3502
+    },
+    {
+      "epoch": 94.68,
+      "learning_rate": 3.1406249999999994e-05,
+      "loss": 0.0679,
+      "step": 3503
+    },
+    {
+      "epoch": 94.7,
+      "learning_rate": 3.125e-05,
+      "loss": 0.0984,
+      "step": 3504
+    },
+    {
+      "epoch": 94.73,
+      "learning_rate": 3.109375e-05,
+      "loss": 0.0836,
+      "step": 3505
+    },
+    {
+      "epoch": 94.76,
+      "learning_rate": 3.09375e-05,
+      "loss": 0.0819,
+      "step": 3506
+    },
+    {
+      "epoch": 94.78,
+      "learning_rate": 3.078125e-05,
+      "loss": 0.0948,
+      "step": 3507
+    },
+    {
+      "epoch": 94.81,
+      "learning_rate": 3.0625e-05,
+      "loss": 0.0804,
+      "step": 3508
+    },
+    {
+      "epoch": 94.84,
+      "learning_rate": 3.046875e-05,
+      "loss": 0.095,
+      "step": 3509
+    },
+    {
+      "epoch": 94.86,
+      "learning_rate": 3.0312499999999998e-05,
+      "loss": 0.0817,
+      "step": 3510
+    },
+    {
+      "epoch": 94.89,
+      "learning_rate": 3.015625e-05,
+      "loss": 0.1177,
+      "step": 3511
+    },
+    {
+      "epoch": 94.92,
+      "learning_rate": 3e-05,
+      "loss": 0.0911,
+      "step": 3512
+    },
+    {
+      "epoch": 94.95,
+      "learning_rate": 2.984375e-05,
+      "loss": 0.0767,
+      "step": 3513
+    },
+    {
+      "epoch": 94.97,
+      "learning_rate": 2.96875e-05,
+      "loss": 0.0992,
+      "step": 3514
+    },
+    {
+      "epoch": 95.0,
+      "learning_rate": 2.953125e-05,
+      "loss": 0.0284,
+      "step": 3515
+    },
+    {
+      "epoch": 95.03,
+      "learning_rate": 2.9375e-05,
+      "loss": 0.0921,
+      "step": 3516
+    },
+    {
+      "epoch": 95.05,
+      "learning_rate": 2.9218750000000002e-05,
+      "loss": 0.0721,
+      "step": 3517
+    },
+    {
+      "epoch": 95.08,
+      "learning_rate": 2.90625e-05,
+      "loss": 0.074,
+      "step": 3518
+    },
+    {
+      "epoch": 95.11,
+      "learning_rate": 2.890625e-05,
+      "loss": 0.0775,
+      "step": 3519
+    },
+    {
+      "epoch": 95.14,
+      "learning_rate": 2.875e-05,
+      "loss": 0.0746,
+      "step": 3520
+    },
+    {
+      "epoch": 95.16,
+      "learning_rate": 2.859375e-05,
+      "loss": 0.0747,
+      "step": 3521
+    },
+    {
+      "epoch": 95.19,
+      "learning_rate": 2.84375e-05,
+      "loss": 0.1195,
+      "step": 3522
+    },
+    {
+      "epoch": 95.22,
+      "learning_rate": 2.8281250000000003e-05,
+      "loss": 0.0877,
+      "step": 3523
+    },
+    {
+      "epoch": 95.24,
+      "learning_rate": 2.8125000000000003e-05,
+      "loss": 0.0809,
+      "step": 3524
+    },
+    {
+      "epoch": 95.27,
+      "learning_rate": 2.7968750000000002e-05,
+      "loss": 0.0864,
+      "step": 3525
+    },
+    {
+      "epoch": 95.3,
+      "learning_rate": 2.7812500000000002e-05,
+      "loss": 0.0666,
+      "step": 3526
+    },
+    {
+      "epoch": 95.32,
+      "learning_rate": 2.765625e-05,
+      "loss": 0.0659,
+      "step": 3527
+    },
+    {
+      "epoch": 95.35,
+      "learning_rate": 2.75e-05,
+      "loss": 0.0804,
+      "step": 3528
+    },
+    {
+      "epoch": 95.38,
+      "learning_rate": 2.734375e-05,
+      "loss": 0.0788,
+      "step": 3529
+    },
+    {
+      "epoch": 95.41,
+      "learning_rate": 2.71875e-05,
+      "loss": 0.0835,
+      "step": 3530
+    },
+    {
+      "epoch": 95.43,
+      "learning_rate": 2.703125e-05,
+      "loss": 0.0838,
+      "step": 3531
+    },
+    {
+      "epoch": 95.46,
+      "learning_rate": 2.6875e-05,
+      "loss": 0.0608,
+      "step": 3532
+    },
+    {
+      "epoch": 95.49,
+      "learning_rate": 2.671875e-05,
+      "loss": 0.0714,
+      "step": 3533
+    },
+    {
+      "epoch": 95.51,
+      "learning_rate": 2.65625e-05,
+      "loss": 0.0564,
+      "step": 3534
+    },
+    {
+      "epoch": 95.54,
+      "learning_rate": 2.6406249999999998e-05,
+      "loss": 0.0827,
+      "step": 3535
+    },
+    {
+      "epoch": 95.57,
+      "learning_rate": 2.625e-05,
+      "loss": 0.079,
+      "step": 3536
+    },
+    {
+      "epoch": 95.59,
+      "learning_rate": 2.609375e-05,
+      "loss": 0.0876,
+      "step": 3537
+    },
+    {
+      "epoch": 95.62,
+      "learning_rate": 2.59375e-05,
+      "loss": 0.1027,
+      "step": 3538
+    },
+    {
+      "epoch": 95.65,
+      "learning_rate": 2.578125e-05,
+      "loss": 0.0633,
+      "step": 3539
+    },
+    {
+      "epoch": 95.68,
+      "learning_rate": 2.5625e-05,
+      "loss": 0.044,
+      "step": 3540
+    },
+    {
+      "epoch": 95.7,
+      "learning_rate": 2.546875e-05,
+      "loss": 0.0958,
+      "step": 3541
+    },
+    {
+      "epoch": 95.73,
+      "learning_rate": 2.5312500000000002e-05,
+      "loss": 0.0852,
+      "step": 3542
+    },
+    {
+      "epoch": 95.76,
+      "learning_rate": 2.515625e-05,
+      "loss": 0.1179,
+      "step": 3543
+    },
+    {
+      "epoch": 95.78,
+      "learning_rate": 2.5e-05,
+      "loss": 0.0811,
+      "step": 3544
+    },
+    {
+      "epoch": 95.81,
+      "learning_rate": 2.484375e-05,
+      "loss": 0.074,
+      "step": 3545
+    },
+    {
+      "epoch": 95.84,
+      "learning_rate": 2.46875e-05,
+      "loss": 0.1032,
+      "step": 3546
+    },
+    {
+      "epoch": 95.86,
+      "learning_rate": 2.453125e-05,
+      "loss": 0.0994,
+      "step": 3547
+    },
+    {
+      "epoch": 95.89,
+      "learning_rate": 2.4375000000000003e-05,
+      "loss": 0.085,
+      "step": 3548
+    },
+    {
+      "epoch": 95.92,
+      "learning_rate": 2.4218750000000003e-05,
+      "loss": 0.0787,
+      "step": 3549
+    },
+    {
+      "epoch": 95.95,
+      "learning_rate": 2.4062500000000002e-05,
+      "loss": 0.0913,
+      "step": 3550
+    },
+    {
+      "epoch": 95.97,
+      "learning_rate": 2.3906250000000002e-05,
+      "loss": 0.0606,
+      "step": 3551
+    },
+    {
+      "epoch": 96.0,
+      "learning_rate": 2.375e-05,
+      "loss": 0.0565,
+      "step": 3552
+    },
+    {
+      "epoch": 96.03,
+      "learning_rate": 2.359375e-05,
+      "loss": 0.1186,
+      "step": 3553
+    },
+    {
+      "epoch": 96.05,
+      "learning_rate": 2.34375e-05,
+      "loss": 0.0755,
+      "step": 3554
+    },
+    {
+      "epoch": 96.08,
+      "learning_rate": 2.328125e-05,
+      "loss": 0.0971,
+      "step": 3555
+    },
+    {
+      "epoch": 96.11,
+      "learning_rate": 2.3125e-05,
+      "loss": 0.1001,
+      "step": 3556
+    },
+    {
+      "epoch": 96.14,
+      "learning_rate": 2.296875e-05,
+      "loss": 0.0632,
+      "step": 3557
+    },
+    {
+      "epoch": 96.16,
+      "learning_rate": 2.28125e-05,
+      "loss": 0.0746,
+      "step": 3558
+    },
+    {
+      "epoch": 96.19,
+      "learning_rate": 2.265625e-05,
+      "loss": 0.0841,
+      "step": 3559
+    },
+    {
+      "epoch": 96.22,
+      "learning_rate": 2.2499999999999998e-05,
+      "loss": 0.0751,
+      "step": 3560
+    },
+    {
+      "epoch": 96.24,
+      "learning_rate": 2.234375e-05,
+      "loss": 0.0714,
+      "step": 3561
+    },
+    {
+      "epoch": 96.27,
+      "learning_rate": 2.21875e-05,
+      "loss": 0.0803,
+      "step": 3562
+    },
+    {
+      "epoch": 96.3,
+      "learning_rate": 2.203125e-05,
+      "loss": 0.077,
+      "step": 3563
+    },
+    {
+      "epoch": 96.32,
+      "learning_rate": 2.1875e-05,
+      "loss": 0.0616,
+      "step": 3564
+    },
+    {
+      "epoch": 96.35,
+      "learning_rate": 2.171875e-05,
+      "loss": 0.0757,
+      "step": 3565
+    },
+    {
+      "epoch": 96.38,
+      "learning_rate": 2.15625e-05,
+      "loss": 0.0642,
+      "step": 3566
+    },
+    {
+      "epoch": 96.41,
+      "learning_rate": 2.1406250000000002e-05,
+      "loss": 0.0802,
+      "step": 3567
+    },
+    {
+      "epoch": 96.43,
+      "learning_rate": 2.125e-05,
+      "loss": 0.0571,
+      "step": 3568
+    },
+    {
+      "epoch": 96.46,
+      "learning_rate": 2.109375e-05,
+      "loss": 0.0663,
+      "step": 3569
+    },
+    {
+      "epoch": 96.49,
+      "learning_rate": 2.09375e-05,
+      "loss": 0.0719,
+      "step": 3570
+    },
+    {
+      "epoch": 96.51,
+      "learning_rate": 2.078125e-05,
+      "loss": 0.0689,
+      "step": 3571
+    },
+    {
+      "epoch": 96.54,
+      "learning_rate": 2.0625e-05,
+      "loss": 0.0925,
+      "step": 3572
+    },
+    {
+      "epoch": 96.57,
+      "learning_rate": 2.0468750000000003e-05,
+      "loss": 0.0541,
+      "step": 3573
+    },
+    {
+      "epoch": 96.59,
+      "learning_rate": 2.0312500000000002e-05,
+      "loss": 0.0836,
+      "step": 3574
+    },
+    {
+      "epoch": 96.62,
+      "learning_rate": 2.0156250000000002e-05,
+      "loss": 0.0547,
+      "step": 3575
+    },
+    {
+      "epoch": 96.65,
+      "learning_rate": 2e-05,
+      "loss": 0.0485,
+      "step": 3576
+    },
+    {
+      "epoch": 96.68,
+      "learning_rate": 1.984375e-05,
+      "loss": 0.0601,
+      "step": 3577
+    },
+    {
+      "epoch": 96.7,
+      "learning_rate": 1.96875e-05,
+      "loss": 0.0849,
+      "step": 3578
+    },
+    {
+      "epoch": 96.73,
+      "learning_rate": 1.953125e-05,
+      "loss": 0.0894,
+      "step": 3579
+    },
+    {
+      "epoch": 96.76,
+      "learning_rate": 1.9375e-05,
+      "loss": 0.0876,
+      "step": 3580
+    },
+    {
+      "epoch": 96.78,
+      "learning_rate": 1.921875e-05,
+      "loss": 0.0662,
+      "step": 3581
+    },
+    {
+      "epoch": 96.81,
+      "learning_rate": 1.90625e-05,
+      "loss": 0.0946,
+      "step": 3582
+    },
+    {
+      "epoch": 96.84,
+      "learning_rate": 1.890625e-05,
+      "loss": 0.0831,
+      "step": 3583
+    },
+    {
+      "epoch": 96.86,
+      "learning_rate": 1.875e-05,
+      "loss": 0.0673,
+      "step": 3584
+    },
+    {
+      "epoch": 96.89,
+      "learning_rate": 1.8593749999999998e-05,
+      "loss": 0.0891,
+      "step": 3585
+    },
+    {
+      "epoch": 96.92,
+      "learning_rate": 1.84375e-05,
+      "loss": 0.0792,
+      "step": 3586
+    },
+    {
+      "epoch": 96.95,
+      "learning_rate": 1.828125e-05,
+      "loss": 0.0732,
+      "step": 3587
+    },
+    {
+      "epoch": 96.97,
+      "learning_rate": 1.8125e-05,
+      "loss": 0.0689,
+      "step": 3588
+    },
+    {
+      "epoch": 97.0,
+      "learning_rate": 1.796875e-05,
+      "loss": 0.0391,
+      "step": 3589
+    },
+    {
+      "epoch": 97.03,
+      "learning_rate": 1.78125e-05,
+      "loss": 0.0871,
+      "step": 3590
+    },
+    {
+      "epoch": 97.05,
+      "learning_rate": 1.765625e-05,
+      "loss": 0.0644,
+      "step": 3591
+    },
+    {
+      "epoch": 97.08,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 0.0677,
+      "step": 3592
+    },
+    {
+      "epoch": 97.11,
+      "learning_rate": 1.734375e-05,
+      "loss": 0.0675,
+      "step": 3593
+    },
+    {
+      "epoch": 97.14,
+      "learning_rate": 1.71875e-05,
+      "loss": 0.088,
+      "step": 3594
+    },
+    {
+      "epoch": 97.16,
+      "learning_rate": 1.703125e-05,
+      "loss": 0.0739,
+      "step": 3595
+    },
+    {
+      "epoch": 97.19,
+      "learning_rate": 1.6875e-05,
+      "loss": 0.097,
+      "step": 3596
+    },
+    {
+      "epoch": 97.22,
+      "learning_rate": 1.671875e-05,
+      "loss": 0.1019,
+      "step": 3597
+    },
+    {
+      "epoch": 97.24,
+      "learning_rate": 1.6562500000000003e-05,
+      "loss": 0.0811,
+      "step": 3598
+    },
+    {
+      "epoch": 97.27,
+      "learning_rate": 1.6406250000000002e-05,
+      "loss": 0.0836,
+      "step": 3599
+    },
+    {
+      "epoch": 97.3,
+      "learning_rate": 1.6250000000000002e-05,
+      "loss": 0.1031,
+      "step": 3600
+    },
+    {
+      "epoch": 97.32,
+      "learning_rate": 1.609375e-05,
+      "loss": 0.0753,
+      "step": 3601
+    },
+    {
+      "epoch": 97.35,
+      "learning_rate": 1.59375e-05,
+      "loss": 0.0855,
+      "step": 3602
+    },
+    {
+      "epoch": 97.38,
+      "learning_rate": 1.578125e-05,
+      "loss": 0.0784,
+      "step": 3603
+    },
+    {
+      "epoch": 97.41,
+      "learning_rate": 1.5625e-05,
+      "loss": 0.0617,
+      "step": 3604
+    },
+    {
+      "epoch": 97.43,
+      "learning_rate": 1.546875e-05,
+      "loss": 0.0938,
+      "step": 3605
+    },
+    {
+      "epoch": 97.46,
+      "learning_rate": 1.53125e-05,
+      "loss": 0.0771,
+      "step": 3606
+    },
+    {
+      "epoch": 97.49,
+      "learning_rate": 1.5156249999999999e-05,
+      "loss": 0.0712,
+      "step": 3607
+    },
+    {
+      "epoch": 97.51,
+      "learning_rate": 1.5e-05,
+      "loss": 0.0777,
+      "step": 3608
+    },
+    {
+      "epoch": 97.54,
+      "learning_rate": 1.484375e-05,
+      "loss": 0.0949,
+      "step": 3609
+    },
+    {
+      "epoch": 97.57,
+      "learning_rate": 1.46875e-05,
+      "loss": 0.0921,
+      "step": 3610
+    },
+    {
+      "epoch": 97.59,
+      "learning_rate": 1.453125e-05,
+      "loss": 0.0648,
+      "step": 3611
+    },
+    {
+      "epoch": 97.62,
+      "learning_rate": 1.4375e-05,
+      "loss": 0.0741,
+      "step": 3612
+    },
+    {
+      "epoch": 97.65,
+      "learning_rate": 1.421875e-05,
+      "loss": 0.076,
+      "step": 3613
+    },
+    {
+      "epoch": 97.68,
+      "learning_rate": 1.4062500000000001e-05,
+      "loss": 0.0558,
+      "step": 3614
+    },
+    {
+      "epoch": 97.7,
+      "learning_rate": 1.3906250000000001e-05,
+      "loss": 0.1015,
+      "step": 3615
+    },
+    {
+      "epoch": 97.73,
+      "learning_rate": 1.375e-05,
+      "loss": 0.0703,
+      "step": 3616
+    },
+    {
+      "epoch": 97.76,
+      "learning_rate": 1.359375e-05,
+      "loss": 0.0659,
+      "step": 3617
+    },
+    {
+      "epoch": 97.78,
+      "learning_rate": 1.34375e-05,
+      "loss": 0.0675,
+      "step": 3618
+    },
+    {
+      "epoch": 97.81,
+      "learning_rate": 1.328125e-05,
+      "loss": 0.0916,
+      "step": 3619
+    },
+    {
+      "epoch": 97.84,
+      "learning_rate": 1.3125e-05,
+      "loss": 0.0818,
+      "step": 3620
+    },
+    {
+      "epoch": 97.86,
+      "learning_rate": 1.296875e-05,
+      "loss": 0.085,
+      "step": 3621
+    },
+    {
+      "epoch": 97.89,
+      "learning_rate": 1.28125e-05,
+      "loss": 0.0638,
+      "step": 3622
+    },
+    {
+      "epoch": 97.92,
+      "learning_rate": 1.2656250000000001e-05,
+      "loss": 0.0492,
+      "step": 3623
+    },
+    {
+      "epoch": 97.95,
+      "learning_rate": 1.25e-05,
+      "loss": 0.0629,
+      "step": 3624
+    },
+    {
+      "epoch": 97.97,
+      "learning_rate": 1.234375e-05,
+      "loss": 0.0677,
+      "step": 3625
+    },
+    {
+      "epoch": 98.0,
+      "learning_rate": 1.2187500000000001e-05,
+      "loss": 0.0354,
+      "step": 3626
+    },
+    {
+      "epoch": 98.03,
+      "learning_rate": 1.2031250000000001e-05,
+      "loss": 0.077,
+      "step": 3627
+    },
+    {
+      "epoch": 98.05,
+      "learning_rate": 1.1875e-05,
+      "loss": 0.0829,
+      "step": 3628
+    },
+    {
+      "epoch": 98.08,
+      "learning_rate": 1.171875e-05,
+      "loss": 0.1014,
+      "step": 3629
+    },
+    {
+      "epoch": 98.11,
+      "learning_rate": 1.15625e-05,
+      "loss": 0.0679,
+      "step": 3630
+    },
+    {
+      "epoch": 98.14,
+      "learning_rate": 1.140625e-05,
+      "loss": 0.0507,
+      "step": 3631
+    },
+    {
+      "epoch": 98.16,
+      "learning_rate": 1.1249999999999999e-05,
+      "loss": 0.0486,
+      "step": 3632
+    },
+    {
+      "epoch": 98.19,
+      "learning_rate": 1.109375e-05,
+      "loss": 0.066,
+      "step": 3633
+    },
+    {
+      "epoch": 98.22,
+      "learning_rate": 1.09375e-05,
+      "loss": 0.0987,
+      "step": 3634
+    },
+    {
+      "epoch": 98.24,
+      "learning_rate": 1.078125e-05,
+      "loss": 0.0813,
+      "step": 3635
+    },
+    {
+      "epoch": 98.27,
+      "learning_rate": 1.0625e-05,
+      "loss": 0.0573,
+      "step": 3636
+    },
+    {
+      "epoch": 98.3,
+      "learning_rate": 1.046875e-05,
+      "loss": 0.0636,
+      "step": 3637
+    },
+    {
+      "epoch": 98.32,
+      "learning_rate": 1.03125e-05,
+      "loss": 0.065,
+      "step": 3638
+    },
+    {
+      "epoch": 98.35,
+      "learning_rate": 1.0156250000000001e-05,
+      "loss": 0.0715,
+      "step": 3639
+    },
+    {
+      "epoch": 98.38,
+      "learning_rate": 1e-05,
+      "loss": 0.0919,
+      "step": 3640
+    },
+    {
+      "epoch": 98.41,
+      "learning_rate": 9.84375e-06,
+      "loss": 0.0688,
+      "step": 3641
+    },
+    {
+      "epoch": 98.43,
+      "learning_rate": 9.6875e-06,
+      "loss": 0.0741,
+      "step": 3642
+    },
+    {
+      "epoch": 98.46,
+      "learning_rate": 9.53125e-06,
+      "loss": 0.0686,
+      "step": 3643
+    },
+    {
+      "epoch": 98.49,
+      "learning_rate": 9.375e-06,
+      "loss": 0.0864,
+      "step": 3644
+    },
+    {
+      "epoch": 98.51,
+      "learning_rate": 9.21875e-06,
+      "loss": 0.0708,
+      "step": 3645
+    },
+    {
+      "epoch": 98.54,
+      "learning_rate": 9.0625e-06,
+      "loss": 0.0834,
+      "step": 3646
+    },
+    {
+      "epoch": 98.57,
+      "learning_rate": 8.90625e-06,
+      "loss": 0.0706,
+      "step": 3647
+    },
+    {
+      "epoch": 98.59,
+      "learning_rate": 8.750000000000001e-06,
+      "loss": 0.0905,
+      "step": 3648
+    },
+    {
+      "epoch": 98.62,
+      "learning_rate": 8.59375e-06,
+      "loss": 0.0823,
+      "step": 3649
+    },
+    {
+      "epoch": 98.65,
+      "learning_rate": 8.4375e-06,
+      "loss": 0.0559,
+      "step": 3650
+    },
+    {
+      "epoch": 98.68,
+      "learning_rate": 8.281250000000001e-06,
+      "loss": 0.04,
+      "step": 3651
+    },
+    {
+      "epoch": 98.7,
+      "learning_rate": 8.125000000000001e-06,
+      "loss": 0.0901,
+      "step": 3652
+    },
+    {
+      "epoch": 98.73,
+      "learning_rate": 7.96875e-06,
+      "loss": 0.0672,
+      "step": 3653
+    },
+    {
+      "epoch": 98.76,
+      "learning_rate": 7.8125e-06,
+      "loss": 0.0949,
+      "step": 3654
+    },
+    {
+      "epoch": 98.78,
+      "learning_rate": 7.65625e-06,
+      "loss": 0.0797,
+      "step": 3655
+    },
+    {
+      "epoch": 98.81,
+      "learning_rate": 7.5e-06,
+      "loss": 0.0715,
+      "step": 3656
+    },
+    {
+      "epoch": 98.84,
+      "learning_rate": 7.34375e-06,
+      "loss": 0.0608,
+      "step": 3657
+    },
+    {
+      "epoch": 98.86,
+      "learning_rate": 7.1875e-06,
+      "loss": 0.0703,
+      "step": 3658
+    },
+    {
+      "epoch": 98.89,
+      "learning_rate": 7.031250000000001e-06,
+      "loss": 0.0801,
+      "step": 3659
+    },
+    {
+      "epoch": 98.92,
+      "learning_rate": 6.875e-06,
+      "loss": 0.0995,
+      "step": 3660
+    },
+    {
+      "epoch": 98.95,
+      "learning_rate": 6.71875e-06,
+      "loss": 0.0731,
+      "step": 3661
+    },
+    {
+      "epoch": 98.97,
+      "learning_rate": 6.5625e-06,
+      "loss": 0.0771,
+      "step": 3662
+    },
+    {
+      "epoch": 99.0,
+      "learning_rate": 6.40625e-06,
+      "loss": 0.0216,
+      "step": 3663
+    },
+    {
+      "epoch": 99.03,
+      "learning_rate": 6.25e-06,
+      "loss": 0.0791,
+      "step": 3664
+    },
+    {
+      "epoch": 99.05,
+      "learning_rate": 6.093750000000001e-06,
+      "loss": 0.077,
+      "step": 3665
+    },
+    {
+      "epoch": 99.08,
+      "learning_rate": 5.9375e-06,
+      "loss": 0.0861,
+      "step": 3666
+    },
+    {
+      "epoch": 99.11,
+      "learning_rate": 5.78125e-06,
+      "loss": 0.0706,
+      "step": 3667
+    },
+    {
+      "epoch": 99.14,
+      "learning_rate": 5.6249999999999995e-06,
+      "loss": 0.0619,
+      "step": 3668
+    },
+    {
+      "epoch": 99.16,
+      "learning_rate": 5.46875e-06,
+      "loss": 0.0975,
+      "step": 3669
+    },
+    {
+      "epoch": 99.19,
+      "learning_rate": 5.3125e-06,
+      "loss": 0.081,
+      "step": 3670
+    },
+    {
+      "epoch": 99.22,
+      "learning_rate": 5.15625e-06,
+      "loss": 0.0854,
+      "step": 3671
+    },
+    {
+      "epoch": 99.24,
+      "learning_rate": 5e-06,
+      "loss": 0.0531,
+      "step": 3672
+    },
+    {
+      "epoch": 99.27,
+      "learning_rate": 4.84375e-06,
+      "loss": 0.0655,
+      "step": 3673
+    },
+    {
+      "epoch": 99.3,
+      "learning_rate": 4.6875e-06,
+      "loss": 0.0839,
+      "step": 3674
+    },
+    {
+      "epoch": 99.32,
+      "learning_rate": 4.53125e-06,
+      "loss": 0.0598,
+      "step": 3675
+    },
+    {
+      "epoch": 99.35,
+      "learning_rate": 4.3750000000000005e-06,
+      "loss": 0.0803,
+      "step": 3676
+    },
+    {
+      "epoch": 99.38,
+      "learning_rate": 4.21875e-06,
+      "loss": 0.0869,
+      "step": 3677
+    },
+    {
+      "epoch": 99.41,
+      "learning_rate": 4.0625000000000005e-06,
+      "loss": 0.0768,
+      "step": 3678
+    },
+    {
+      "epoch": 99.43,
+      "learning_rate": 3.90625e-06,
+      "loss": 0.078,
+      "step": 3679
+    },
+    {
+      "epoch": 99.46,
+      "learning_rate": 3.75e-06,
+      "loss": 0.087,
+      "step": 3680
+    },
+    {
+      "epoch": 99.49,
+      "learning_rate": 3.59375e-06,
+      "loss": 0.0875,
+      "step": 3681
+    },
+    {
+      "epoch": 99.51,
+      "learning_rate": 3.4375e-06,
+      "loss": 0.0504,
+      "step": 3682
+    },
+    {
+      "epoch": 99.54,
+      "learning_rate": 3.28125e-06,
+      "loss": 0.1065,
+      "step": 3683
+    },
+    {
+      "epoch": 99.57,
+      "learning_rate": 3.125e-06,
+      "loss": 0.0714,
+      "step": 3684
+    },
+    {
+      "epoch": 99.59,
+      "learning_rate": 2.96875e-06,
+      "loss": 0.0902,
+      "step": 3685
+    },
+    {
+      "epoch": 99.62,
+      "learning_rate": 2.8124999999999998e-06,
+      "loss": 0.0866,
+      "step": 3686
+    },
+    {
+      "epoch": 99.65,
+      "learning_rate": 2.65625e-06,
+      "loss": 0.07,
+      "step": 3687
+    },
+    {
+      "epoch": 99.68,
+      "learning_rate": 2.5e-06,
+      "loss": 0.0613,
+      "step": 3688
+    },
+    {
+      "epoch": 99.7,
+      "learning_rate": 2.34375e-06,
+      "loss": 0.0891,
+      "step": 3689
+    },
+    {
+      "epoch": 99.73,
+      "learning_rate": 2.1875000000000002e-06,
+      "loss": 0.0959,
+      "step": 3690
+    },
+    {
+      "epoch": 99.76,
+      "learning_rate": 2.0312500000000002e-06,
+      "loss": 0.0693,
+      "step": 3691
+    },
+    {
+      "epoch": 99.78,
+      "learning_rate": 1.875e-06,
+      "loss": 0.0711,
+      "step": 3692
+    },
+    {
+      "epoch": 99.81,
+      "learning_rate": 1.71875e-06,
+      "loss": 0.049,
+      "step": 3693
+    },
+    {
+      "epoch": 99.84,
+      "learning_rate": 1.5625e-06,
+      "loss": 0.0613,
+      "step": 3694
+    },
+    {
+      "epoch": 99.86,
+      "learning_rate": 1.4062499999999999e-06,
+      "loss": 0.0687,
+      "step": 3695
+    },
+    {
+      "epoch": 99.89,
+      "learning_rate": 1.25e-06,
+      "loss": 0.0926,
+      "step": 3696
+    },
+    {
+      "epoch": 99.92,
+      "learning_rate": 1.0937500000000001e-06,
+      "loss": 0.0721,
+      "step": 3697
+    },
+    {
+      "epoch": 99.95,
+      "learning_rate": 9.375e-07,
+      "loss": 0.0814,
+      "step": 3698
+    },
+    {
+      "epoch": 99.97,
+      "learning_rate": 7.8125e-07,
+      "loss": 0.0623,
+      "step": 3699
+    },
+    {
+      "epoch": 100.0,
+      "learning_rate": 6.25e-07,
+      "loss": 0.0302,
+      "step": 3700
+    },
+    {
+      "epoch": 100.0,
+      "step": 3700,
+      "total_flos": 4.877080420793346e+19,
+      "train_loss": 0.5632794735734226,
+      "train_runtime": 3321.7041,
+      "train_samples_per_second": 104.705,
+      "train_steps_per_second": 1.114
+    }
+  ],
+  "max_steps": 3700,
+  "num_train_epochs": 100,
+  "total_flos": 4.877080420793346e+19,
+  "trial_name": null,
+  "trial_params": null
+}